Documente Academic
Documente Profesional
Documente Cultură
Agencies
This paper examines the role of credit rating agencies in the subprime crisis that
triggered the 2007–2008 financial turmoil. We focus on two aspects of ratings
that contributed to the boom and bust of the market for structured debt: rating
inflation and coarse information disclosure. The paper discusses how regulation
can be designed to mitigate these problems in the future. Our preferred policy is
to require rating agencies to be paid by investors rather than by issuers and to
grant open and free access to data about the loans or securities underlying structured
debt products. A more modest (but less effective) approach would be to
retain the ‘issuer pays’ model but require issuers to pay an upfront fee irrespective
of the rating, ban ‘credit shopping’, and prescribe a more complete format
for the information that rating agencies must disseminate.
— Marco Pagano and Paolo Volpin
played a key role in the crisis: insofar as many naively based their investment in
these securities mainly or solely on inflated credit ratings, these led to a massive
mispricing of risk, whose correction later detonated the crisis.
Issuers benefit from rating inflation if at least some investors are naive, that is,
do not realize that ratings are excessively optimistic. In this situation, rating inflation
leads to under-pricing of risk. A similar situation occurs even if all investors
are rational, but regulation forces them to buy highly rated securities (for instance,
only AAA-rated ones) and these are in scarce supply. A third possibility is that these
investors are rational portfolio managers who have an excessive incentive to take
risk, even when it is mispriced.
Ratings’ inflation and low informativeness may also reinforce each other. To the
extent that investors are rational, they will see through CRAs’ incentives to inflate
ratings and therefore will consider them as relatively uninformative. Conversely, the
coarseness of ratings may reinforce the tendency to inflate them, as it expands the
room for collusion between issuer and rating agency, and therefore the conflict of
interest with investors. For instance, if ratings are set on a discrete scale, friendly
rating agencies can suggest to issuers how to structure their securities or tranches so
as to just attain a given rating. So in each rating class a disproportionate number
of issues or tranches will have a risk corresponding to the low end of that class.
This enhances ratings inflation compared to a situation where ratings are set on a
finer grid.
Such sweeping changes will meet not only the likely opposition of CRAs, but also
that of regulators, due to their considerable transitional costs. Therefore, policymakers
may also want to consider a second-best policy, which tries to address the
above-discussed problems without overhauling the current setup. Specifically, they
may retain the ‘issuer pays’ model but constrain the way in which agencies contract
with issuers and are paid by them: issuers should pay an upfront fee irrespective of
the rating issued (the so-called ‘Cuomo plan’, named after New York Attorney
General Andrew Cuomo), and credit shopping should be banned. Similarly, regulators
could enhance transparency not by forcing issuers to grant open and free
access to all relevant data, but simply by prescribing a more complete and detailed
format for the information that CRAs must disseminate.
These more limited reforms may still be consistent with the current regulatory
delegation of vast powers to a select group of rating agencies. But their effectiveness
in addressing the failures of CRAs exposed by the current crisis is likely to be limited.
First, even if issuers must pay an upfront fee and cannot engage in explicit rating
shopping, implicit collusion may still be sustainable: issuers may systematically
patronize the agency that offers them the best ratings, which they can identify by
comparing the models that agencies use to rate securities.
Second, prescribing which pieces of information and statistics CRAs should disseminate
would shift the burden of identifying such information on the regulator,
which can be complex in the presence of very diverse financial products. It may
also expose such detailed regulation to the danger of becoming rapidly outdated in
the presence of innovations in the design of structured debt securities, some of
which may even be induced by regulation itself.
ucts were AAA-rated, in contrast to less than 1% of the corporate issues. Rating
agencies benefited a lot from the growth of structured products. By 2006, 44% of
Moody’s reported revenue came from rating structured finance products, with
respect to 32% of revenues from the traditional business of rating of corporate
bonds (Coval et al., 2008). In this way the issuers of structured products and the rating
agencies became very much dependent on each other, until the collapse in late
2007.
The extent of the crisis in the market for asset-backed securities can be best
appreciated by looking at the dynamics of the ABX price indexes reported in Figure
1. ABX indexes provide an indicative measure of the value of MBSs, as they
are based on the price of credit default swaps offering protection against the default
of baskets of subprime MBSs of different ratings. In other words, a decline in the
ABX index indicates an increase in the cost of insuring a basket of mortgages of a
certain rating against default. It is clear from the graph that the crisis was first felt
in March 2007 by the BBB-rated MBS. A few months later, in July 2007, all
tranches (even the AAA-rated securities) experienced a substantial drop in value,
as UBS shut down its internal hedge fund, Dillon Read, after suffering about
$125 million of subprime-related losses. As the crisis worsened, the indexes kept
declining across all ratings. The loss to the holders of structured debt securities was
compounded by the massive and severe downgrades of their ratings by CRAs in
2007 and 2008 (Benmelech and Dlugosz, 2009b).
The extent to which the performance of CRAs came to depend on the securitization
process in recent years can be exemplified by comparing the profits of Moody’s
– one of the three largest agencies – with those of Reuters – a leading financial
publisher and data provider – over the past 6 years. The choice of Reuters as a
benchmark is justified by the fact that CRAs define themselves as ‘financial publish-
1900
2100
2300
EBIT ($mil)
500
700
900
1100
1300
1500
1700
Moodys
Reuters
5%
6%
7%
8%
9%
60%
70%
80%
90%
100%
Return on Assets
3%
4%
40%
50%
Moodys (left axis)
Reuters (right axis)
Moody's Market Capitalization Relative to Reuters
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
Figure 2. Moody’s vs. Reuters
Notes: The top chart shows earnings before income and taxes (EBIT), and the second the return on assets
(ROA) of the two companies. The bottom chart displays the stock market capitalization of Moody’s relative
to that of Reuters.
Source: Worldscope.
RATINGS AGENCIES 409
Downloaded from http://economicpolicy.oxfordjournals.org/ by guest on April 30, 2016
ers’, and as such are treated by US law, even though their activity arguably goes
beyond the simple publication of financial ‘opinions’.3 The top chart in Figure 2
shows that Moody’s earnings before interest and taxes (EBIT) grew threefold from
$541 million at the beginning of 2003 to $1,439 million in the third quarter of
2007, and then reverted to $683 million by the third quarter of 2009, in perfect
synchrony with the crisis. In contrast, Reuters’ profits took only a modest dip at the
start of the crisis, and kept growing thereafter. The time pattern of the two companies’
return on assets (ROA) in the second chart paints a similar picture, although
the level of Moody’s profitability greatly exceeds that of Reuters throughout the
period (even at its lowest value in the third quarter of 2009, it stands at 40%
against Reuters’ 6%). Perhaps most tellingly, the ratio between the two companies’
market capitalization, shown in the bottom chart of Figure 2, peaked at approximately
80% between late 2005 and early 2006, and dropped continuously thereafter,
down to the current level around 20%.
To understand how securitization works, what information is made available to
investors and how CRAs contribute to it, it is best to illustrate it with reference to
a real subprime MBS. The special-purpose vehicle (SPV) shown in Table 2 is called
GSAMP-Trust 2006-NC2 and owns 3,949 subprime loans for an aggregate principal
of $881 million. The originator of the underlying loans is New Capital Financial,
at the time the second largest subprime lender in the US: it originated $51.6
billion in mortgage loans in 2006, and filed for bankruptcy in April 2007. The
arranger of the deal is Goldman Sachs, who bought the portfolio from the originator
and sold it to the SPV named GSAMP-Trust 2006-NC2. The SPV funded the
purchase of this loan through the issue of asset-backed securities (listed in Table 2).
These securities entitle their holders to the cash flow generated by the loan portfolio,
according to the seniority structure of their tranches (a ‘waterfall’ scheme):
holders of junior tranches can be paid only after senior tranches have been paid in
full. Therefore, in case of default the ‘junior’ tranche is the first to absorb losses
from the underlying collateral loans, and when it becomes worthless the ‘mezzanine
tranche’ starts absorbing further losses, with the senior tranche (typically AAArated)
being the most protected against default risk.
If sold to the public, these securities – or their tranches, if any – must be rated
by a CRA and must be described in a public prospectus. In our example, there
were 17 tranches: 5 AAA-rated senior tranches, 9 mezzanine tranches with ratings
ranging from AA+ to BBB), 2 B tranches with ratings BB+ to BB, and an equity
tranche X with no rating. It is worth noticing that the first 5 tranches representing
almost 80% of the total were AAA-rated. Tranche X (the riskiest one), being
unrated, was not sold to the public. The prospectus of this MBS is a document of
Fussing and Fuming over Fannie and Freddie: How Much Smoke, How Much Fire? W. Scott Frame and
Lawrence J. White he Federal National Mortgage Association and the Federal Home Loan Mortgage
Corporation-commonly known as Fannie Mae and Freddie Mac, respectively'-have led the way in dramatic
changes that have taken place in the structure of the U.S. residential mortgage markets since the 1970s.
Fannie Mae and Freddie Mac are quasi-private/quasi-public: for example, they have federal charters that
confer unique regulatory provisions; but their shares are publicly traded on the New York Stock Exchange. The
biggest advantage of Fannie Mae's and Freddie Mac's anomalous legal status arises because financial
markets treat their obligations as if those obligations are backed by the federal government-even though the
federal government explicitly does not do so. With the benefit of this special status, Fannie Mae and Freddie
Mac have grown into enormous financial institutions, with combined total assets of over $1.8 trillion in 2003.
One critic, Richard Carnell (2004), a former Assistant Secretary of the Treasury, has suggested that the two
companies' growth is at least partially a consequence of a "double game" that they play: " [They] tell Congress
and the news media, 'Don't worry, the government is not on the hook'-and then turn around and tell Wall
Street, 'Don't worry, the government really is on the hook."'
The preferential legal status of Fannie Mae and Freddie Mac serves as one of a number of mechanisms by
which the federal government encourages the consumption (and, arguably, the overconsumption) of housing in
the U.S. economy--this one with an on-budget cost of zero. But economists are congenitally economy--this one
with an on-budget cost of zero. But economists are congenitally suspicious of programs that seem to offer
something for nothing. After all, a federal suspicious of programs that seem to offer something for nothing.
After all, a federal guarantee of the deposits in savings and loans cost nothing for many decades- until the
early 1990s, when it cost taxpayers about $150 billion (U.S. Federal Deposit Insurance Corporation, 1997).
Furthermore, Federal Reserve Chairman Alan Greenspan (2004), among others, has suggested that the
anomalous situation of Fannie Mae and Freddie Mac may even pose systemic risks to the financial sector. This
article will offer a generalist's guide to the functions that Fannie Mae and Freddie Mac perform in the residential
mortgage financial markets and the con- troversies that swirl around them. Along the way we will highlight
some important-and perhaps underappreciated- changes that are occurring in the structure of U.S. residential
mortgage markets. Some Background What Do Fannie Mae and Freddie Mac Do? Fannie Mae and Freddie
Mac participate in the secondary mortgage market: Mortgage originators come to them with pools (bundles) of
mortgages and either swap these assets for securities or sell them outright to one of the two companies. Under
Fannie Mae's and Freddie Mac's "swap programs," an originator exchanges a mortgage pool for a
mortgage-backed security that is issued and guaranteed by one of the two companies and that represents an
interest in the same pool. Fannie Mae and Freddie Mac promise the security holders that the latter will receive
timely payment of interest and principal on the underlying mortgages, less an annual "guarantee fee" of about
20 basis points (0.20 percent) on the remaining principal. In essence, Fannie Mae and Freddie Mac are
providing insurance to holders of mortgage-backed securities against default risk on the underlying mortgages
and are thus bearing that risk themselves. This securitization activity illustrates one of their two core
businesses: mortgage credit guarantees. The other core business of Fannie Mae and Freddie Mac is their
investment portfolios. These portfolios consist largely of mortgage-backed securities that they have purchased
in the open market, as well as mortgages that they purchase from originators under their "cash programs."
Fannie Mae and Freddie Mac fund these assets largely by issuing debt, as the two companies are highly
leveraged with total equity that is less than 4 percent of total assets. The major differences between Fannie
Mae and Freddie Mac are in their historical roots. The National Housing Act of 1934 created the Federal
Housing Authority (FHA) and also provided for chartering national mortgage associations as entities within the
federal government. The only association ever formed was theW Scott Frame and LawrenceJ. White 161
National Mortgage Association of Washington in 1938, which eventually became the Federal National
Mortgage Association-now Fannie Mae. By issuing debt and purchasing and holding FHA-insured residential
mortgages from "mortgage banks," Fannie Mae was able to expand the available pool of finance to support
housing and also to provide a degree of unification to mortgage markets. During this time, mortgage markets
were localized for technological reasons, as well as for this time, mortgage markets were localized for
technological reasons, as well as for reasons rooted in laws that prohibited interstate banking and restricted
intrastate bank branches in many states during most of the twentieth century. In 1968, Fannie Mae was
converted into a private corporation, with publicly traded shares listed on the New York Stock Exchange,
although it retained a unique federal charter that is discussed further below. Apparently one major reason for
the privatization was that until 1968 Fannie Mae's debt was part of the federal debt; but when Fannie Mae
became a publicly traded company, that debt (which stayed with the company) was removed from the national
debt total. Fannie Mae was replaced within the federal government by the Government National Mortgage
Association (which became known as "Ginnie Mae"), an agency within the Department of Housing and Urban
Development (HUD) that guarantees mortgage-backed secu- rities that have as their underlying assets
residential mortgages that are insured primarily by the FHA or by the Department of Veterans Affairs (formerly
the Veterans Administration, or VA). Freddie Mac, by contrast, was created by Congress in 1970 to support
mortgage markets by securitizing mortgages originated by savings and loan associations (S&Ls). During the
1970s and 1980s, Freddie Mac was technically a private com- pany, with its equity shares held solely by the
twelve Federal Home Loan Banks (FHLBs) and by S&Ls that were members of the FHLBs. Freddie Mac's
board of directors consisted of the three board members of the Federal Home Loan Bank Board, which
regulated the S&L industry during that time. Freddie Mac was converted in 1989 into a publicly traded
company, also traded on the New York Stock Exchange, with the same special features as apply to Fannie
Mae. A major motivation for the conversion of Freddie Mac to a publicly traded company was the belief that a
wider potential shareholding public would raise the price of the shares held by the then ailing S&L industry and
thus improve the balance sheets of the latter. In its early history, Freddie Mac tended to securitize mortgages,
originated largely by S&Ls, whereas Fannie Mae tended to hold the mortgages that it bought largely from
mortgage banks. By the 1990s, however, the two companies' structures and strategies looked quite similar.
Current Size As of year-end 2003, Fannie Mae had $1,010 billion in assets and Freddie Mac had $803 billion
in assets, making them the second- and third-largest U.S. compa- nies, respectively, on this basis. In addition,
both Fannie Mae and Freddie Mac had significant quantities of mortgage-backed securities outstanding-that is,
net of
mortgage-backed securities carried on their balance sheets. As of year-end 2003, Fannie Mae had $1,300
billion in outstanding mortgage-backed securities, while Freddie Mac had another $769 billion outstanding.
Both Fannie Mae and Freddie Mac have grown rapidly over the past three decades and at a faster clip than the
residential mortgage market as a whole. As Table 1 shows, in 1980, the residential mortgage market consisted
of $1.1 trillion in obligations, of which Fannie Mae and Freddie Mac held or securitized only $78 billion, or
about 7 percent. By contrast, in 2003, these companies held or securitized over $3.6 trillion of the $7.7 trillion
in residential mortgage debt, or about 47 percent. The U.S. Congressional Budget Office (2001) estimated the
involvement of Fannie Mae and Freddie Mac in various slices of the mortgage markets as of 2000. For
example, for fixed-rate single-family mortgages that were eligible to be purchased or securitized by the two
companies (about half of the total residential mortgage market), the two companies accounted for 71 percent
of that slice (through either their securitizations or their portfolio holdings). Special Features of Fannie Mae and
Freddie Mac Fannie Mae and Freddie Mac differ from other U.S. corporations in that they were created by
Congress and maintain exclusive federal charters. These charters, in turn, confer a number of rights and
responsibilities on these companies (U.S. Congressional Budget Office, 1996, 2001; U.S. Department of the
Treasury, 1996; U.S. General Accounting Office, 1996).Fussing and Fuming over Fannie and Freddie: How
Much Smoke, How Much Fire? 163 Some of the advantages that Fannie Mae and Freddie Mac enjoy are as
follows. First, they are exempt from state and local income taxes. Second, the Secretary of the Treasury has
the authority to purchase up to $2.25 billion of Fannie Mae's and Freddie Mac's securities. Third, they issue
"government securities," as classified under the Securities Exchange Act of 1934, which in practice means that
their securities are eligible for use as collateral for public deposits, for purchase by the Federal Reserve in
open-market operations, and for unlimited investment by federally insured depository institutions. A further
implication is that they are exempt from the registration and reporting requirements and fees of the Securities
and Exchange Commission, although Fannie Mae voluntarily registered its stock and Exchange Commission,
although Fannie Mae voluntarily registered its stock with the SEC in March 2003, and they are exempt from the
provisions of many state with the SEC in March 2003, and they are exempt from the provisions of many state
investor protection laws. Fourth, they use the Federal Reserve as their fiscal agent, which means that their
securities are issued and transferred using the same system as U.S. Treasury borrowings. Finally, there is no
defined resolution process in place to deal systematically with insolvency at either Fannie Mae or Freddie Mac;
instead, congressional action is required (Eisenbeis, Frame and Wall, 2004). Fannie Mae's and Freddie Mac's
federal charters also present some disadvan- tages for these companies. First, their activities are restricted to
residential mort- gage finance. Second, they are restricted to the secondary market, which means that they
cannot originate mortgages directly. Third, there is a maximum size of mortgage that they can finance. These
mortgages are usually described as "con- forming" mortgages; larger mortgages are usually described as
'jumbos."2 The size is linked to an annual index of housing prices; for 2005, the limit for a single-family home is
$359,650. Fourth, the mortgages that they finance must have at least a 20 percent down payment or else have
mortgage insurance that is provided either by private firms or the federal government. Fifth, they are subject to
federal safety-and-soundness regulation, including minimum leverage and risk-based cap- ital requirements
and supervisory examinations, by the Office of Federal Housing Enterprise Oversight (OFHEO), an
independent agency within the Department of Housing and Urban Development (HUD). Sixth, they are subject
to "mission oversight" by HUD, which approves new housing finance programs and sets percent-of-business
housing finance goals. Currently, 50 percent of Fannie Mae's and Freddie Mac's business must benefit low-
and moderate-income families, 31 percent must benefit underserved areas, and 20 percent must serve
"special affordable" needs. 2 Other nonconforming mortgages (besides jumbos) are those that do not meet the
credit-quality standards of the two companies. Also, this limit applies only to a single-unit residence; higher
limits apply to two-unit, three-unit and four-unit residences and to multifamily housing. Also, limits for Hawaii,
Alaska and the Virgin Islands are, by law, 50 percent higher.164 Journal of Economic Perspectives A Halo of
Government Support By law, securities of Fannie Mae and Freddie Mac are required to include language that
they are not guaranteed by, or otherwise an obligation of, the federal government. However, Fannie Mae's and
Freddie Mac's special federal charters and the attendant package of special benefits directly lower their
operating costs and have created a "halo" of implied federal government support for the two companies. In
addition, past government actions have contributed to the percep- tion of implied government support. During
the late 1970s and early 1980s, Fannie Mae was insolvent on a market value basis and benefited from
supervisory forbear- ance (U.S. General Accounting Office, 1990; Kane and Foster, 1986). A summa- rizing
phrase for this halo is that the two companies are "government sponsored enterprises" (GSEs).3 enterprises"
(GSEs).3 Evidence suggests that financial markets believe that the federal government Evidence suggests that
financial markets believe that the federal government would come to the rescue of Fannie Mae and Freddie
Mac (and hence their creditors) in the event of financial difficulties. As a result of this perceived implicit
guarantee, Fannie Mae and Freddie Mac can typically borrow at interest rates that are more favorable than
those of a AAA-rated corporation (though not quite as favorably as the rates on government debt), even though
their stand-alone ratings would be about AA- or less. For Fannie Mae and Freddie Mac, empirical studies
suggest that this translates into roughly a 35-40 basis point debt funding advantage, although there is
significant variation in the estimates depending on the credit rating and maturity of the comparison bonds
(Ambrose and Warga, 1996, 2002; Nothaft, Pearce and Stevanovic, 2002). Other studies have found that the
compa- nies enjoy about a 30 basis point advantage in issuing mortgage-backed securities (U.S.
Congressional Budget Office, 1996, 2001; U.S. Department of the Treasury, 1996). The presence of Fannie
Mae and Freddie Mac in the secondary mortgage market influences the primary mortgage market. Most
notably, Fannie Mae's and Freddie Mac's activities result in "conforming" mortgages' carrying lower interest
rates thanjumbo mortgages. Several econometric studies have estimated this effect, and most found the
interest rate differential to be about 20-25 basis points, although the estimates vary depending on the empirical
specification, data sample and time period studied. For an introduction to this literature, see U.S. Congres-
sional Budget Office (2001), McKenzie (2002), Passmore (2003) and Ambrose, LaCour-Little and Sanders
(2004) and the references in these papers. 3 The Federal Home Loan Bank System, which serves as a
wholesale bank for many federally insured depository institutions (banks, S&Ls and credit unions), enjoys a
similar package of favorable features and is similarly described as a GSE. There are also GSEs that serve
agricultural credit markets (Farm Credit System and the Federal Agricultural Mortgage Corporation, or Farmer
Mac) and the student loan market (Student Loan Marketing Association, or Sallie Mae), although the latter is in
the process of privatization under the name SLM Corp. The financial pages of major newspapers often set
aside a separate box of "agency issues" to report the yields on GSE securities-a visible illustration of how they
are treated differently.W. Scott Frame and LawrenceJ. White 165 Fannie Mae and Freddie Mac may bring
other potential benefits to mortgage markets, although these claims are often controversial. For example, one
contro- versy is over whether Fannie Mae and Freddie Mac enhance the stability of the mortgage market by
acting as a large "market maker" in mortgage-backed securities and thereby reducing interest rate volatility in
that market (Gonzalez-Rivera, 2001; Naranjo and Toevs, 2002; Peek and Wilcox, 2003). We will discuss one
counter- argument to this point below--the possibility that the sheer size of these mortgage portfolios creates a
potential for systemic risk in the financial system. Another claim sometimes made is that Fannie Mae and
Freddie Mac can act as a focal point for market-wide standard setting with respect to technology and certain
"best practices." For example, Fannie Mae and Freddie Mac have each developed widely used automated
underwriting systems (known as Desktop Under- writer and Loan Prospector, respectively) that evaluate an
individual loan's credit risk, including whether it meets the companies' purchase requirements. These risk,
including whether it meets the companies' purchase requirements. These systems have greatly reduced the
time and cost of the mortgage origination process. systems have greatly reduced the time and cost of the
mortgage origination process. However, several large lenders that already had similar systems in place
resented the Fannie Mae and Freddie Mac systems, characterizing them as an unwarranted intrusion into the
primary mortgage market. Residential Mortgages: A Primer To appreciate the central role that Fannie Mae and
Freddie Mac play in U.S. residential mortgage finance, a brief tutorial on mortgage finance itself will be useful,
including the process of mortgage securitization. Some Interesting Aspects of Residential Mortgages At one
level, a residential mortgage is a simple debt instrument. The purchaser of a home borrows money to finance
the home purchase. The home serves as the collateral for the loan. The borrower makes monthly payments
that cover the interest on the outstanding principal and the amortization of that principal. The complexity of the
mortgage as a debt instrument arises because of the interactions among three properties of most U.S.
single-family residential mort- gages: lengthy maturities, fixed interest rates and "free" prepayment options.4
First, the term of most mortgages is for 15 or 30 years, with the average term of a new mortgage for a
single-family residence hovering at about 27-28 years over the past decade. Second, fixed-rate mortgages are
more common than adjustable-rate mort- gages, which have been one-quarter or less of the market in the last
6-7 years and exceeded one-third of the market in only a single year during the 1990s. Indeed, adjustable rate
mortgages were relatively unknown before the early 1980s, primarily 4 More detail on the characteristics of
residential mortgages can be found on the website of the Federal Housing Finance Board:
(http://www.fhfb.gov).166 Journal of Economic Perspectives because federal regulation prevented most
depository institutions from originating them. Finally, these long-term, fixed-rate mortgages are generally
prepayable with- out a penalty, in the sense that the borrower can accelerate the repayment of principal or
repay the entire amount at any time, at no additional cost. Though the borrower does not pay an explicit
penalty at the time of repayment, the cost of the option to the lender is incorporated into the contract interest
rate and the fees that the lender charges at the time of origination. A mortgage lender faces two kinds of risks.
First, credit risk bears on whether the lender will be repaid the principal amount that has been lent and the
contracted interest. Second, market risk refers to whether changes in market conditions- primarily interest rate
changes--will affect the value of the mortgage. The credit risk on most single-family residential mortgages is
quite low.5 After all, lenders only originate such loans after screening for adequate household income and a
good credit history. Further, the home itself serves as the collateral for the mortgage in the event of default.
Most lenders require a 20 percent down payment or some form of mortgage insurance. The credit risk losses
on mortgages payment or some form of mortgage insurance. The credit risk losses on mortgages held by
Fannie Mae and Freddie Mac averaged 5.4 basis points annually over the held by Fannie Mae and Freddie
Mac averaged 5.4 basis points annually over the 1987-2002 period, and the losses averaged only 1 basis point
annually for 1999- 2002 (Inside Mortgage Finance, 2003). A fixed-rate debt instrument also creates market risk
for the lender. If interest rates increase, the price of the instrument declines; and if interest rates decrease, the
price of the instrument rises. The longer is the maturity of the instrument, the greater are the associated price
swings. These risks are further complicated by changes in the rate of prepayment. A lower interest rate, which
would benefit the lender of a fixed-rate instrument, makes borrowers more likely to repay their existing
mortgages, either by refinancing the existing mortgage or deciding that the time is right to purchase a new
home. This quickening of the repayment rate deprives the lender of the potential capital gain on the mortgage
that would otherwise occur on a debt instrument that could not be repaid; equivalently, the greater pace of
repayment is occurring just when the lender doesn't want repay- ment, since the lender can then only relend
(or reinvest) the funds at the lower prevailing interest rates. Conversely, a higher interest rate leads to less
prepayment. In this case, the capital loss that the lender would have experienced on a fixed-rate debt
instrument as interest rates increase is compounded by the slackening of the prepayment rate; in essence,
prepayments are slackening just when the lender wishes that they would accelerate. In the specialized
literature, this phenomenon of 5 Two growth areas in residential mortgage lending that are exceptions to this
broad claim are "subprime" and high loan-to-value loans. Subprime loans are those made to borrowers with
material blemishes in their recent credit history. High loan-to-value loans, while generally made to individuals
with especially good credit, have principal amounts equal to or greater than the appraised value of the property
acting as collateral.Fussing and Fuming over Fannie and Freddie: How Much Smoke, How Much Fire? 167
additional adverse effects on the mortgage lender from decreases or increases in interest rates is described as
the "negative convexity" of the mortgage instrument. Mortgage Securitization Prior to 1970, mortgages were
largely a nontraded debt instrument.6 The initial lender that originated the loan usually held the loan until it
matured (or was prepayed), collecting interest and principal repayments in the interim. In 1970, Ginnie Mae
issued the first "pass-through" mortgage-backed securities, which cre- ated a claim on an underlying pool of
residential mortgages and meant that security-holders had the right to receive the interest and principal
repayments of the pool as a whole. The Ginnie Mae securities carried the federal government's promise of
timely payment of interest and principal, on top of the guarantees issued by the FHA and the VA on the
underlying mortgages. For the first time, the claim on a stream of mortgage payments could be readily traded.
A number of aspects of mortgage-backed securities are worth noting. First, securitization greatly widens the
potential market for the ultimate financing of residential mortgages. Anyone who buys the security directly or
indirectly (for example, through a mutual fund or pension fund) is, in essence, providing the mortgage
financing. mortgage financing. Second, investors in mortgage-backed securities are in a poor position to
assess Second, investors in mortgage-backed securities are in a poor position to assess the credit risk as to
what proportion of the individual mortgages in the underlying pool will be repaid on schedule, so they require
some assurance on these credit risks. Ginnie Mae, Fannie Mae and Freddie Mac offer direct guarantees with
respect to these repayment rates when they issue such securities. "Private-label" securitiza- tions offer other
kinds of reassurance: for example, private financial guarantees; "overcollateralization," in which a security
issuer pledges assets to back the securi- tization in an amount that exceeds the face value of the securities
being issued; or structuring the securities with senior and subordinated tranches, such that credit losses are
first absorbed by more subordinate securities up to certain values. Third, mortgage-backed securities holders
are not protected from the market risks associated with holding long-term, fixed-rate, prepayable mortgages.
The holder of a simple "pass-through" mortgage-backed security (described above) experiences the same
effects of interest rate changes-including the effects of prepayments-as does the owner of an otherwise similar
unsecuritized pool of mortgages. Fourth, the securitization process creates opportunities for "slicing and dicing"
the cash flows in ways that allow the risks to be better allocated among capital markets participants according
to their preferences (Fabozzi, 2001). The "senior/ subordinated" structure noted above is one such method. As
another example, the cash flows from a mortgage pool can be divided into a "principal-only" security and 6 As
was mentioned above, an exception to this pattern involved originations by mortgage banks, which
immediately sold the mortgages to Fannie Mae (which then held the mortgages in its portfolio).168 Journal of
Economic Perspectives an "interest-only" security, where the former has characteristics that heighten
interest-rate risks, while the latter (at least for modest changes in interest rates) ameliorates them.7 More
elaborate multiple-layer securities, in which some inves- tors have more buffering from early prepayment and
others have less, can also be created from Fannie Mae or Freddie Mac pass-through securities and are
described as "collateralized mortgage obligations" (CMOs) or "real estate mortgage invest- ment conduits"
(REMICs). Trends in Residential Mortgage Finance The structure of residential mortgage markets is
substantially different today than it was in 1970. At that time, the dominant pattern was as follows: The financial
institution-typically a savings and loan or a savings bank-that originated a fixed-rate mortgage loan also held it
in its portfolio, and the same financial institution collected the monthly payments and dealt with delinquencies.
The funding for the loan was provided by the institution's deposits, which were insured by the federal
government, through the Federal Savings and Loan Insurance Corporation (for S&Ls) or the Federal Deposit
Insurance Corporation (for com- mercial banks). As Table 2 shows, in 1970 savings and loan institutions held
over 56 percent of outstanding single-family mortgages, and commercial banks and savings and loans together
held over 70 percent. By contrast, the typical pattern today is for mortgage originators to share some or all of
the risks associated with fixed-rate residential mortgage loans with the secondary market. Depository
institutions typically do this by securitizing their conforming mortgages with Fannie Mae or Freddie Mac and
their FHA- and VA-insured loans with Ginnie Mae. The resulting mortgage-backed securities carry an
assurance as to the timely payment of principal and interest, which is backed by a full-faith-and-credit
guarantee of the federal government (Ginnie Mae) or an implied federal guarantee (Fannie Mae and Freddie
Mac). The originator will elect either to hold or to sell the mortgage-backed security, although it may continue to
"service" the underlying loans by collecting monthly payments and dealing with delinquencies. Nondepository
mortgage originators, such as mortgage banks, tend to sell their mortgages outright-often to Fannie Mae or
Freddie Mac. Only adjustable rate mortgages or those fixed-rate mortgages that exceed the
conforming loan limits ('jumbos") or that do not meet certain underwriting criteria are likely to be held in the
originator's portfolio (if the originator is a depository institution) or securitized in a "private label" offering. As
Table 2 also shows, as of year-end 2000, the share of "whole loan" single-family mortgages held by banks and
savings and loans had plummeted to below 30 percent, despite an increase in commercial banks' share of
single-family mortgage mortgages between 1970-2000. However, if one were also to include the banks' and
savings and loans' holdings of mortgage-backed securities (and thus the depositories' exposure to
mortgage-related market risk), their share would rise to slightly over 40 percent. This vertically dis-integrated
structure allows for greater specialization among institutions with respect to mortgage originations, collecting
payments, dealing with delinquent loans, funding, liability issuance and guarantees. Why the Trend to
Securitization and Dis-Integration of Mortgage Markets? Technological advancements-especially improved and
lower-cost data pro- cessing and telecommunications-have undergirded both the expansion of the
securitization process and division of the market into many interlocking providers. Loan originators are able to
gather information about prospective borrowers, analyze it, make judgments about who to lend to and transmit
that information and170 Journal of Economic Perspectives those judgments to others-notably, securitizers and
investors-in ways and over distances that weren't possible in the 1970s. Regulation, however, has also
contributed to the expanded role of Fannie Mae and Freddie Mac in residential mortgage markets, in at least
two ways. First, since 1988, the regulatory risk-based capital (net worth) requirements that apply to banks and
savings and loans have included a lower requirement of 1.6 percent for holding mortgage-backed securities
issued by Fannie Mae and Freddie Mac, compared with the 4 percent requirement for holding whole
(unsecuritized) residential mortgage loans. At first blush, since risk-based capital requirements for whole-loan
mortgages are at the same level as the minimum "leverage" requirements for an adequately capitalized
depository institution (a 4 percent ratio of a depository's overall capital to assets), there would seem to be little
market impact from these risk-based capital charges. However, in cases where a depository institution holds a
diversified loan portfolio that includes higher-risk loans requiring capital levels above the minimum leverage
requirements, its portfolio may be bound by risk-based capital require- ments. So at the margin, the lower
capital requirements for mortgage-backed securities would strongly encourage the institution to substitute
mortgage-backed securities for "whole" mortgage loans. Frame and White (2004b) discuss this "regulatory
capital arbitrage" in greater detail. Second, Fannie Mae and Freddie Mac are required to hold at least 2.5
percent capital against mortgages (or their own mortgage-backed securities) that they retain in their portfolios.
At first glance, this would seem to put such portfolio retention at a capital cost disadvantage as compared with
depositories' holdings of mortgage-backed securities (at 1.6 percent). But in comparison with depositories
mortgage-backed securities (at 1.6 percent). But in comparison with depositories that are bound by the 4
percent minimum leverage requirement (such as savings that are bound by the 4 percent minimum leverage
requirement (such as savings and loans that tend to specialize in mortgage lending), the purchases by Fannie
Mae and Freddie Mac for their own portfolios would have a capital cost advantage. Strategic management
decisions have also been an influence. When Freddie Mac became a publicly traded company in 1989 and
was freed from the constraints that had previously been imposed by the Federal Home Loan Bank Board, its
management soon realized that its favorable borrowing rates provided an excellent opportunity to expand its
income by earning a spread on the differ- ence between mortgages that it held in portfolio and its favorable
borrowing rates. Though Fannie Mae was chastened and restrained by its near-insolvency in the early 1980s,
by the 1990s, its management had also adopted this expan- sionist mentality. Ironically, Fannie Mae and
Freddie Mac did not expand as dramatically in the 1970s and 1980s, when their position as national operators
in mortgage markets should have given them a comparative advantage vis-a-vis their depository rivals who
were restrained by state and federal limitations on interstate branching. Appar- ently, it took technological and
regulatory changes, supplemented by some strate- gic management decisions, to bring about their
expansion.Fussing and Fuming over Fannie and Freddie: How Much Smoke, How Much Fire? 171 The Issues
The federal charters for Fannie Mae and Freddie Mac, and particularly the implied federal guarantee of their
financial obligations, raise the central policy issue. On one side, the implied guarantee on Fannie Mae's and
Freddie Mac's financial obligations allows the federal government to reduce mortgage interest rates for most
residential mortgages by about 20-25 basis points-without an annual appropriation. On the other side, this
policy creates a contingent liability for taxpayers in the event that either enterprise becomes insolvent and the
govern- ment elects to provide financial assistance, as well as causing additional distortions in the housing
market. These tradeoffs involving Fannie Mae and Freddie Mac can then be parsed into five related issues: a)
how the companies' activities mesh with other public policies that encourage the construction and consumption
of housing; b) the appropriate safety-and-soundness regulatory structure; c) the systemic risks that may flow
from the size of the companies; d) the effect of the companies' activities on allocative efficiency; and e) the
consequences of the two companies for efficient market structures. Encouragements for Housing and
Macroeconomic Efficiency U.S. public policy, at all levels of government, encourages the construction and
consumption of housing. The largest incentives pertain to income tax advantages: the exclusion of the implicit
income from housing by owner-occupiers, while allowing the deduction of mortgage interest and local real
estate taxes. Additional tax encouragements include the exemption of much owner-occupied housing from
capital gains taxation and accelerated depreciation on rental housing. Direct provision of rental ("public")
housing is another significant program. In terms of housing finance, some government programs operate
directly, like the mortgage insurance provided by FHA and VA (which allows lower down the mortgage
insurance provided by FHA and VA (which allows lower down payments) and the securitization of associated
mortgages by Ginnie Mae. Other payments) and the securitization of associated mortgages by Ginnie Mae.
Other programs support housing indirectly: federal deposit insurance for depository institutions whose
portfolios contain some residential mortgages, savings and loan charters with mandates to invest in residential
mortgages, and the Federal Home Loan Bank System-originally created by the federal government in 1932 as
a "wholesale" bank that would make low-cost loans to S&Ls. Federal sponsorship of Fannie Mae and Freddie
Mac are another indirect form of support for housing finance and hence housing consumption. The motives
underlying these public policy actions toward housing are diverse. One motive is to assist the middle class with
a major household expenditure. Another motive is to support revenues and employment in residential
construction, sales, and complementary industries. Yet another motive is that homeownership can be viewed
as a way of encouraging households to save (at least so long home values do not decline or other offsetting
reductions in saving or increases in borrowing don't offset the value of home equity). But for most economists,
the172 Journal of Economic Perspectives strongest arguments for government support of housing involve
either a form of in-kind redistributions of income toward lower-income households or the claim that
homeownership has positive externalities. For example, an owner is likely to care more about a residence and
the surrounding neighborhood than is an absentee landlord, which can result in positive externalities including
the external appearance of property, greater watchfulness leading to greater public safety and support for local
public goods. A modest but growing empirical literature provides some documentation for these positive
externalities for neighborhoods and even positive effects on owner-occupier families themselves (Green and
White, 1997; DiPasquale and Glaeser, 2002; Aaronson, 2000, and the references therein). If redistribution to
those with lower income or the positive externalities from a higher rate of homeownership are the goals, then
the logical policy would encourage low- and moderate-income households, who may be on the margin
between renting and owning, to become first-time home buyers. Such programs might aim to reduce down
payments, since the size of down payment can be a binding constraint for low-income households (Linneman
and Wachter, 1989; Quercia, McCarthy and Wachter, 2003), or to reduce monthly payments. However, most
housing programs are broad-based efforts that encourage more housing construction and consumption
throughout the income and social spectrum. For example, the income tax benefits from homeownership
operate as exemptions and deductions, which means that they tend to favor disproportionately higher-income
households in higher marginal tax brackets (Rosen, 1979; Gervais, 2002). Federal sponsorship of Fannie Mae
and Freddie Mac is of this broad-based nature. In 2002, the conforming loan limit for Fannie Mae and Freddie
Mac was $300,700. In that same year, according to the Federal Housing Finance Board, the median price of a
new home that was sold was $187,600, and an 80 percent mortgage on that sale price would have been
$150,080. Thus, the conforming loan limits allow Fannie Mae and Freddie Mac to purchase residential
mortgage loans limits allow Fannie Mae and Freddie Mac to purchase residential mortgage loans that are far
beyond the range that would encompass the low- or moderate-income, that are far beyond the range that
would encompass the low- or moderate-income, first-time buying household. Fannie Mae and Freddie Mac are
required to meet percent-of-business housing goals established by HUD involving annual purchases of loans
involving a) households with less than median incomes; b) underserved areas, such as low-income and
high-minority census tracts; and c) very low income households and low-income households living in
low-income areas. Nevertheless, the bulk of the mortgage purchases by the two companies have not involved
the groups that ought to be the target of homeownership-encouraging activities (U.S. Office of Management
and Budget, 2004). While some research has found that Fannie Mae and Freddie Mac have recently increased
the supply of mortgage credit available to low- and moderate-income households (Ambrose and Thibodeau,
2004), it does not appear that the companies' activities have appreciably affected the rate of homeownership in
the United States (Feldman, 2002; Painter and Redfearn, 2002; Freeman, Galster and Malega, 2003). Such
broad-based encouragements for housing imply that most beneficiariesW. Scott Frame and LawrenceJ. White
173 would have bought anyway, and the marginal effects are largely to cause them to buy larger and
better-appointed homes, on larger lots, and/or to buy second homes. In turn, this broad-based encouragement
means that the United States has invested in an inefficiently large housing stock, although the empirical
literature that attempts to measure the magnitudes of these broad social consequences is surprisingly small. In
one recent study, Gervais (2002) finds that the taxation of the implicit rents on owner-occupied housing
(accompanied by a compensating ad- justment in tax rates) could cause general consumption levels to
increase by almost 5 percent. Taylor (1998) finds that overinvestment in housing persisted during 1975-1995
and estimates the overinvestment in housing at over $220 billion per year (or $300 per month for each
owner-occupied home) for the late 1990s- consistent with prior research by Mills (1987a, b). These results can
be summarized bluntly. The United States has too much housing (and not enough of other goods and
services), and federal sponsorship of Fannie Mae and Freddie Mac exacerbates this problem. Moreover,
Fannie Mae and Freddie Mac do not do an especially good job of focusing on the low- and moderate-income
first-time buyer, where the social argument for support of home- ownership is strongest. Safety and Soundness
The implicit guarantee of Fannie Mae's and Freddie Mac's obligations suggests that attention be paid to the
companies' financial health. How could either enterprise become insolvent? One way is if the credit losses
mushroomed on the mortgages they guaranteed or held outright. This would occur if homeowners could not
repay their mortgages and the prices of housing fell below outstanding loan values. Another-and perhaps more
likely-way is if they failed to hedge their market risks adequately, and the value of their mortgage portfolios fell
below the values of their outstanding debt obligations. This happened to Fannie Mae, as well as to thousands
of savings and loans, in the early 1980s. What is the magnitude of this contingent liability to the federal
government? One way to answer this question is to consider how much the government poten- One way to
answer this question is to consider how much the government poten- tially would have to pay to a third-party
guarantor who would have the same tially would have to pay to a third-party guarantor who would have the
same probabilistic-contingent obligation to make whole the holders of Fannie Mae or Freddie Mac debt and
mortgage-backed securities in the event that one of them failed. Such estimates are approximated by the gross
benefits accruing to Fannie Mae and Freddie Mac from the implied guarantee in terms of both of the compa-
nies' debt and their mortgage-backed securities.8 A "back-of-the-envelope" calcula- tion of this can be
constructed using year-end 2003 data on their outstanding debt 8 Recall that the gross benefits from the
implied guarantee are simply the differential between the interest rates that the creditors of
government-sponsored enterprises actually demand (given their belief that the federal government would be
likely to bail them out) and what they would require if the two companies were wholly divorced from the federal
government.174 Journal of Economic Perspectives and mortgage-backed securities, coupled with estimates of
the interest rate advan- tages that the two companies enjoy (40 basis points on debt and 30 basis points on
mortgage-backed securities) as a result of implied government support, values for growth of portfolio and
mortgage-backed securities (4 percent),9 a discount rate (5 percent, based on long-term Treasury bond yields)
and a time horizon (25 years). These numbers suggest an estimated contingent liability borne by the federal
government of approximately $288 billion. The implied federal guarantee of Fannie Mae's and Freddie Mac's
financial obligations create a "moral hazard" problem. Because of the implied guarantee, creditors do not
monitor the firms' activities as closely as they otherwise would. As a consequence of this reduced monitoring,
the managements of Fannie Mae and Freddie Mac can engage in activities that involve greater risk (with
greater liability consequences for the government), since the companies' owners will benefit from the "upside"
outcomes while being buffered (because of the limited liability of corporate owners) from the full consequences
of large "downside" outcomes. Safety-and-soundness regulatory oversight is one way to deal with insufficient
market monitoring and discipline. However, in the case of Fannie Mae and Freddie Mac, the case for such
regulation must be more nuanced. The federal guarantee itself is implicit and based on expectations, rather
than a legally binding obligation of the government. The presence of safety-and-soundness oversight likely
rein- forces such expectations, and it is thus unclear whether safety-and-soundness regulation for Fannie Mae
and Freddie Mac actually reduces taxpayer exposure (Frame and White, 2004a; Greenspan, 2004). The Office
of Federal Housing Enterprise Oversight (OFHEO), within the Department of Housing and Urban Development
(HUD), is the safety and sound- ness regulator for Fannie Mae and Freddie Mac. OFHEO is authorized to set
risk-based capital standards, conduct examinations and take enforcement actions if unsafe or unsound
financial or management practices are identified.'0 However, OFHEO has been criticized for a perceived lack
of effectiveness, including the lengthy delays that the regulator experienced in issuing and finalizing its
risk-based capital regulation for Fannie Mae and Freddie Mac (U.S. General Accounting Office, 1997). More
recently, Treasury Secretary John Snow (2003) remarked that Office, 1997). More recently, Treasury Secretary
John Snow (2003) remarked that there is a "general recognition that the supervisory system for the housing
GSEs there is a "general recognition that the supervisory system for the housing GSEs 9 Though the overall
mortgage market has grown at an average annual rate of 8.5 percent over the past decade, this rate could not
be sustained indefinitely, nor could the even faster growth of Fannie Mae and Freddie Mac. Instead, we have
assumed a growth rate that would be roughly equal to the growth of nominal U.S. GDP. At the other extreme, if
the levels of the two portfolios and the mortgage-backed securities were to be frozen at year-end 2003 levels,
the contingent liability would be "merely" $183 billion. 10 Prior to the creation of OFHEO in 1992, the
Department of Housing and Urban Development (HUD) maintained exclusive regulatory oversight
responsibilities over Fannie Mae and (for 1989-1992) Freddie Mac. Prior to the passage of the Financial
Institutions Reform, Recovery and Enforcement Act of 1989, Freddie Mac was the responsibility of the Federal
Home Loan Bank Board.Fussing and Fuming over Fannie and Freddie: How Much Smoke, How Much Fire?
175 neither has the tools, nor the stature, to effectively deal with the current size, complexity, and importance
of these enterprises." Most important, OFHEO cur- rently lacks the power to place an insolvent
government-sponsored enterprise into receivership. A number of legislative proposals were introduced in
Congress during 2003 to enhance the safety and soundness regulation of Fannie Mae and Freddie Mac
(Frame and White, 2004a). These proposals would move the safety-and-soundness regulator for Fannie Mae
and Freddie Mac out of Housing and Urban Develop- ment (where the culture is more focused on housing) and
into either the Treasury (where the culture is more focused on safety and soundness) or to create a
freestanding agency outside the executive branch. This relocated regulatory agency might also be given some
additional powers. For example, it might have the ability to levy fees on Fannie Mae and Freddie Mac to fund
itself, thus removing the agency from the vagaries of annual budgetary appropriations. It would have a stronger
ability to set and revise the minimum capital requirements that the two companies must meet, as well as to
regulate new programs or new activities. Finally, the new agency might have the power to appoint a receiver
that could liquidate or otherwise dispose of either Fannie Mae's or Freddie Mac's assets in the event that one
of them became insolvent. At the time that this article was completed in fall 2004, no definitive legislative action
had been taken. Systemic Risk In recent testimony before Congress, Federal Reserve Chairman Greenspan
(2004) suggested that Fannie Mae and Freddie Mac may pose "systemic risks" to the U.S. economy. That is, if
one of the companies became financially distressed, enough harm to the overall financial system could be
caused such that a nontrivial reduction in general economic activity would result." Such concerns generally
begin with the observations that these companies are large, highly leveraged and focused on a particular asset
class that they dominate. As noted above, as of year-end 2003, Fannie Mae and Freddie Mac were the second
and third largest U.S. companies measured by assets. In terms of financial leverage, Fannie Mae and Freddie
Mac operated with ratios of total capital to total assets of 3.4 percent and 3.9 percent, respectively, as of that
same date. (By contrast, FDIC-insured deposi- tory institutions maintained a ratio of total capital to total assets
of 9.2 percent on December 31, 2003.) Finally, about 86 percent of the combined balance sheet of December
31, 2003.) Finally, about 86 percent of the combined balance sheet of Fannie Mae and Freddie Mac
constitutes mortgage-related assets. Almost 50 per- Fannie Mae and Freddie Mac constitutes
mortgage-related assets. Almost 50 per- cent of all the credit risk and 20 percent of all market risk associated
with U.S. residential mortgage assets are held by the two companies. The U.S. Office of 1 The discussion
below focuses on systemic risks emanating from either (or both) Fannie Mae or Freddie Mac. Fahey (2003)
and U.S. Office of Federal Housing Enterprise Oversight (2003) suggest that because of their implicit
government guarantees, the companies might also act as a source of strength to financial markets in the face
of external shocks.176 Journal of Economic Perspectives Federal Housing Enterprise Oversight (2003)
provides a detailed discussion of systemic risk as it may pertain to Fannie Mae and Freddie Mac. Concerns
about the systemic consequences of Fannie Mae and Freddie Mac have focused on the concentration of
market risk within the two companies. Fannie Mae and Freddie Mac do hedge these risks in their portfolios in
various ways, thereby distributing the market risk into the broader capital market. One way they do this is by
issuing "callable debt," so that if interest rates fall and a surge of prepaying home mortgage occurs, Fannie and
Freddie can refinance their existing debt as well. The companies also use derivative financial instruments, like
interest- rate swaps, to reduce their exposure to interest rate risk. Fannie Mae and Freddie Mac rely heavily on
"dynamic hedging," whereby they rebalance their portfolios in response to changing interest rates that
influence expected prepayment behavior (U.S. Office of Federal Housing Enterprise Oversight, 2003; Jaffee,
2003). A specific area of systemic concern has been the effect of portfolio rebalancing by Fannie Mae and
Freddie Mac on fixed-income markets. For example, a decline in the general level of interest rates often leads
to increased prepayment risk, which in turn reduces the duration of mortgage-related assets. Holders of these
assets, if they would like to maintain the original duration of their portfolios, would then have to purchase other
longer-term assets to add duration. Increased demand for these longer-term assets would increase the price of
these assets, or equivalently reduce the interest rates on them further. Perli and Sack (2003) present statistical
evidence that mortgage-related hedging significantly influences the behavior of the 10-year swap rate,
although their analysis is not focused on Fannie Mae and Freddie Mac specifically. The U.S. Office of Federal
Housing Enterprise Oversight (2003) notes several conditions that make an economy vulnerable to financial
sector shocks, including weak market discipline of institutions and poor public disclosure. "Market disci- pline"
refers to the ability of investors and creditors to track the changing financial condition and risk of firms and
securities, to price securities accordingly and, through pricing, to influence the actions of management (Bliss
and Flannery, 2001). Consistent with the presence of some market discipline of Fannie Mae and Freddie Mac,
Seiler (2003) finds that the share prices and debt yields of these companies respond to new information about
their respective financial risks. Fannie Mae and Freddie Mac have taken steps to strengthen their public
disclosure through a set of six voluntary initiatives announced in 2000: a) to issue subordinated debt; b) to
meet certain liquidity standards; c) to enhance credit-risk subordinated debt; b) to meet certain liquidity
standards; c) to enhance credit-risk disclosures; d) to enhance interest rate disclosures; e) to obtain annual
credit disclosures; d) to enhance interest rate disclosures; e) to obtain annual credit ratings; and f) to
self-implement and report their regulatory risk-based capital levels. (This sixth initiative was rendered obsolete
in 2002, when OFHEO's risk- based capital standard became effective.) Also, in 2002 both Fannie Mae and
Freddie Mac agreed to register their common stock with the Securities and Ex- change Commission, although
at the time of this writing only Fannie Mae has followed through. These steps all seem headed in the right
direction, althoughW. Scott Frame and Lawrence J. White 177 some improvements also seem possible (Frame
and Wall, 2002; Jaffee, 2003). A more direct approach to reducing systemic risk associated with Fannie Mae
and Freddie Mac would be to limit their debt issuance, thereby capping the size of their respective balance
sheets (Greenspan, 2004). Two other arguments about systemic risk related to Fannie Mae and Freddie Mac
have been discussed. One is that the companies have seemingly large expo- sures to a small number of
derivatives counterparties. Fannie Mae and Freddie Mac are significant end-users of interest rate derivatives,
and together the notional amount of these instruments outstanding as of year-end 2001 was $1.6 trillion for the
two companies. Moreover, as of the same date, five counterparties accounted for about 59 percent of their
over-the-counter derivatives. However, the notional amounts of derivatives often bear little direct relationship to
the actual credit exposure; as a simple example, an option to borrow $100 million at a certain fixed rate of
interest six months from today has a notional value of $100 million- although the price of the actual option is
much lower. Moreover, counterparties in a derivatives trade are required to post collateral if their net exposure
exceeds certain limits, with lower-rated counterparties posting proportionately more collat- eral. Indeed, U.S.
Office of Housing Enterprise Oversight (2003) reports that as of year-end 2001, the net uncollateralized
exposures were only $110 million for Fannie Mae and $69 million for Freddie Mac. In the event of counterparty
default, however, there is "rollover risk" to the extent that Fannie Mae and Freddie Mac find it difficult or
particularly expensive to replace their hedging positions (Jaffee, 2003). An introduction to derivatives and their
uses and misuses in this journal can be found in Stulz (2004). The other argument is that, under current
regulations, federally insured depository institutions can make unlimited investments in the obligations of
Fannie Mae and Freddie Mac. However, as illustrated in Frame and Wall (2002), although a large number of
banks hold such obligations in amounts that exceed their net worth (just over 50 percent as of year-end 2000),
most of these institutions are very small (over 95 percent as of year-end 2000). Further, in the event of
financial distress at either Fannie Mae or Freddie Mac, the losses incurred by mortgage- backed securities
investors would likely be minimal, given that the securities are collateralized by mortgages with very low
historic loss rates. Kulp (2004) assesses the exposure of FDIC-insured institutions to privatization of Fannie
Mae and Freddie Mac and finds minimal impact, although that analysis does not consider the impact in the
case of financial distress at either or both of the companies. Allocative Efficiency The benefits embedded in the
federal charters of Fannie Mae and Freddie Mac The benefits embedded in the federal charters of Fannie Mae
and Freddie Mac act as a barrier to entry in the secondary conforming mortgage market (Goodman act as a
barrier to entry in the secondary conforming mortgage market (Goodman and Passmore, 1992; Hermalin and
Jaffee, 1996). In that market, Fannie Mae and Freddie Mac can be characterized as duopolists. But instead of
raising prices, Fannie Mae and Freddie Mac cause mortgage interest rates to be below those that178 Journal
of Economic Perspectives the private market would otherwise provide. For this reason, Carlton, Gross and
Stillman (2002) conclude that the two companies do not raise antitrust concerns. However, an examination of
market power in the context of a government subsidy should not offer comparisons with an unsubsidized
market, but instead should ask whether the subsidy is completely passed through by competing firms to
customers. Some theoretical research has examined various equilibrium outcomes arising from interactions
between a perfectly competitive primary mortgage market and a less than perfectly competitive secondary
mortgage market (for example, Gan and Riddiough, 2004; Passmore, Sparks and Ingpen, 2002; Heuson,
Passmore and Sparks, 2001; Passmore and Sparks, 2000). Such studies examine issues related to mortgage
guarantee pricing as well as the distribution of mortgage credit risk (by risk classification) between mortgage
originators and Fannie Mae and Freddie Mac. The empirical evidence suggests that Fannie Mae and Freddie
Mac do retain some portion of their federal benefits and hence are not acting in a perfectly competitive manner.
One piece of casual evidence is the extraordinary profitability of these two firms. For the years 1998-2003, for
example, Fannie Mae earned an average return on equity of 32.6 percent while Freddie Mac earned an
average of 28.3 percent. By contrast, the industry return on equity for all FDIC-insured commercial banks for
the same six years was 13.6 percent.'2 A second piece of evidence is from studies of using Fannie Mae and
Freddie Mac as conduits for a mortgage interest rate subsidy. Feldman (1998) reviews the various
approaches, while U.S. Congressional Budget Office (2001, 2004) and Passmore (2003) offer more recent
analyses. For example, the U.S. Congressional Budget Office (2004) estimated that in 2003 the two
companies received gross benefits of $19.6 billion accruing from their federal charters, of which they passed
through $13.4 billion to homebuyers through lower mortgage rates and retained $6.2 billion for their
shareholders. Using a simulation exercise, Passmore (2003) estimates the median after-tax present value of
Fannie Mae's and Freddie Mac's net federal benefits at $72 billion, accounting for 60 percent of the companies'
combined market capi- talization. Outside analyses sponsored by Fannie Mae disputes various assumptions
and research methods used in these studies (for example, Toevs, 2001; Greene, 2004; Blinder, Flannery and
Kamihachi, 2004). Of course, if one believes that broad-brush public policies encourage too much housing
investment and that the activities of Fannie Mae and Freddie Mac exac- erbate this problem, then the exercise
of market power by these firms will improve allocative efficiency in the economy. 12 For the 15 years
1989-2003, Fannie Mae averaged 28.9 percent, while Freddie Mac averaged 24.9 percent. An alternative
profitability ratio, return on assets (ROA), is substantially lower for Fannie Mae and Freddie Mac than that for
commercial banks. The primary reason for this is that Fannie and Freddie are far more leveraged. As an
indicator of likely rents from less-than-completely-vigorous Freddie are far more leveraged. As an indicator of
likely rents from less-than-completely-vigorous competition, return on equity is the superior
measure.competition, return on equity is the superior measure.Fussing and Fuming over Fannie and Freddie:
How Much Smoke, How Much Fire? 179 Institutions for Greater Market Efficiency As an historical matter,
Fannie Mae and Freddie Mac have surely enhanced the liquidity of mortgage loans, improved the geographic
diversification of mortgage credit risk, and nationally integrated mortgage markets. Further, the presence of
Fannie Mae and Freddie Mac and their implied guarantees may well have been important for the innovation
and development of mortgage securitization in the 1970s and 1980s. Nevertheless, most of these benefits can
now be achieved as a result of geographic deregulation of banking, which has allowed large, nationwide
mortgage originators to emerge. Mortgage securitization is now a well-established technology of finance that
does not require the special status of Fannie Mae and Freddie Mac. So do Fannie Mae and Freddie Mac still
provide a necessary institu- tional underpinning for a more efficient mortgage market in the current economic
environment? Because of their special status, Fannie Mae and Freddie Mac can issue blanket credit-loss
guarantees on an entire pool of loans as well as avoid the costs of having their securities rated by rating
agencies or registered with the Securities and Exchange Commission. In contrast, private-label
mortgage-backed securities often have a structure that involves creating different levels of seniority of debt,
obtaining securities ratings from rating agencies like Moody's or Standard & Poor's and registering with the
Securities and Exchange Commission-all of which involve transactions costs. Further, investors in the senior
mortgage-backed securities follow the credit performance of the underlying loans, which results in monitoring
costs. In this way, the presence of Fannie Mae and Freddie Mac (and their special halo) might eliminate
significant transactions costs (Woodward, 2004). However, these transaction costs are actually shifted rather
than eliminated: investors believe that they are shielded from credit risk not only by Fannie Mae and Freddie
Mac, but also ultimately by taxpayers. Another argument is that the new-era securitization process is an
inherently more efficient way of providing mortgage finance and that the expansion of Fannie Mae and Freddie
Mac (with their implicit federal guarantee) at the expense of depositories' holdings of residential mortgages
(supported by the depositories' explicit federal deposit insurance) is evidence of this superior efficiency (Van
Order, 2000a, b; 2001). Three potential counterpoints come to mind here. First, securitization is a widely used
technology of finance in the United States: it permeates consumer credit markets (mortgages, credit cards,
auto loans) and is used by both depository and nondepository financial institutions. Second, as dis- cussed
earlier, an important reason why depository institutions securitize mortgages with Fannie Mae and Freddie Mac
rather than holding them in their portfolios is the differential treatment of these assets for purposes of required
regulatory risk-based capital. Finally, it is marginal debt funding costs that matter, rather than average costs.
While these marginal and average costs are probably roughly equal for Fannie Mae and Freddie Mac (that is,
they face a perfectly elastic supply of for Fannie Mae and Freddie Mac (that is, they face a perfectly elastic
supply of loanable funds), banks likely face a rising marginal cost curve as they tap intoloanable funds), banks
likely face a rising marginal cost curve as they tap into180 Journal of Economic Perspectives non-core
deposits or borrow money from wholesale lenders (like the Federal Home Loan Banks) or the capital markets.
Hence, the extent to which the dramatic changes in the structure of U.S. residential mortgage markets are
efficiency-driven rather than regulation-driven remains unclear.'3 Finally, there is no assurance that the current
organizational structures for Fannie Mae and Freddie Mac are the most efficient. Since the Congress has
issued only two charters of this particular kind, no competitive processes exist to reward more efficient firms
and winnow less efficient firms. The market for corporate control also cannot operate effectively, since the two
companies' large sizes and special status likely make them immune to a takeover by a firm or an investor
group. What Is to Be Done? First-Best There seems no strong efficiency reason for preserving the existing
structure of Fannie Mae and Freddie Mac. They mostly just add to an already excessive encouragement for
housing in the United States, using an implied guarantee that (to the extent that it would be honored) creates a
contingent liability for the U.S. government. Thus, a complete privatization of the two companies would be the
first-best outcome. In this vision, the two companies would no longer enjoy any special privileges, but also
would no longer be restricted to their current narrow slice of the financial world. The consequence of such a
step for residential mort-gage markets would be modest: Mortgage rates would probably rise by about 20-25
basis points (ceteris paribus). Because it appears that the United States already builds and consumes too
much housing, this would be a move in the right direction. Instead of backing a broad subsidy to housing
through Fannie Mae and Freddie Mac, the federal government ought to focus on assisting first-time home
buyers with low and moderate incomes (Calomiris, 2001; White, 2003). Second-Best The complete
privatization of Fannie Mae and Freddie Mac seems an unlikely outcome in the current political environment.
Accordingly, second-best measures should be considered. One useful step would be for government officials
to state clearly, whenever the subject comes up, that the federal government does not guarantee the debt of
13 As a related matter, whenever either of the two firms has expanded slightly in "horizontal" (sub-prime
lending) or "vertical" (providing underwriting software to mortgage originators) directions-or even publicly
contemplated such moves-critics have complained that the two companies' ability to expand arises solely from
the low-cost funding that they enjoy from the implicit guarantee and not because of any inherent efficiency
advantage.
W. Scott Frame and LawrenceJ. White 181 Fannie Mae or Freddie Mac and will not bail them (or their
creditors) out. No presidential administration has explicitly made such a statement. More typical are carefully
crafted comments that reiterate that the federal government is not re-quired to bail out Fannie Mae or Freddie
Mac, but fall short of flatly stating that the government will not do so (for example, Mankiw, 2004). One step
toward separat-ing the federal government from Fannie Mae and Freddie Mac has recently occurred. Five of
the 18 board members of each company were historically ap-pointed by the president; but in 2004, the Bush
administration announced that it would cease appointing such members. Second, the enterprises should be
forced to focus more on the lower end of the housing market. The loan limit for a conforming mortgage might
be frozen at its current level of $359,650 for some years. The Department of Housing and Urban Development
might strengthen the affordable housing goals that it sets for Fannie Mae and Freddie Mac. Indeed, in the fall
2004 HUD revised these goals in a way that should result in a marked increase in targeted lending. Third, the
safety-and-soundness regime should be strengthened so that it is more comparable to that of the federal
banking agencies. This step should include giving OFHEO or its successor a) responsibility for the approval of
new programs and other activities; b) the discretion to set both minimum and risk-based capital requirements;
and c) receivership authority. While the presence of a safety-and-soundness regulatory regime likely reinforces
the market perception of an implied guarantee, stronger oversight should serve to reduce taxpayer exposure.
The policy issues raised by Fannie Mae and Freddie Mac are complex and increasingly important as these
institutions grow and become a more pervasive influence on the financial sector. Although the first-best path of
privatization may well be politically unrealistic, we believe that some constructive second-best measures
deserve serious consideration.
scale economies in REITs on a more complete level than ever attempted before.
Recognizing that a serious limitation of the earlier studies of REIT economies of
scale is the inability to separate size-related advantages versus a period of rapidly
expanding, strong markets, we obtain 1,508 REIT year observations using a
sample of 187 equity REITs trading in U.S. markets between January 1990
and December 2001. Thus, our analysis covers an 11-year period that includes
a full market cycle of significant market expansion followed by contraction.
Using this relatively large data set, we test for economies of scale in REITs
by examining growth prospects, revenue and expense measures, profitability
ratios, systematic risk and cost of capital measures.
In addition to the size and time span of our data set, we also incorporate newsreported
information on merger activity into the analysis of economies of scale.
This is important as the costs of integrating a merger occur in the first year
or so, while efficiencies are realized largely subsequently. In the following
section, we examine the background of consolidation and economies-of-scale
arguments in real estate while defining our expectations. Next we discuss the
data and hypotheses. This is followed by a discussion of our empirical results
and conclusion.
Background
A Brief History of REIT Consolidation
Following banking deregulation in the 1980s, real estate investment, which
was already heavily debt financed, surged through the use of debt provided by
banks and savings and loans. However, the savings and loan crisis in 1989–
1990 resulted in a curtailment in debt financing for real estate, forcing many
industry leaders to turn to public capital markets. Some observers argued that the
real estate industry would have to follow the example of other capital-intensive
public firm-dominated industries and enter a period of significant consolidation,
with publicly traded companies leading the consolidation effort (e.g., Linneman
1997, 2002).1
Since 1994 industry watchers have routinely predicted a wave of consolidation
in the industry.2 The period of 1997–1999 was a particularly noticeable
See Ambrose, Highfield and Linneman (2000) for an overview of REIT consolidation.
1
See Campbell, Ghosh and Sirmans (2001) for a discussion of the REIT merger and
2
acquisition activity during the 1994 to 1998 period.
Real Estate and Economies of Scale 325
consolidation period in almost every property sector.3 Then, while the dotcom
bubble was bursting in early 2000 and investors finally abandoned overvalued
companies with little-to-no tangible assets, real estate stood out as a relatively
strong and stable investment. In fact, REITs averaged a 26% return for the year,
and 2000 was the first year for REITs to outperform both the Nasdaq and DJIA.
Thus, with the increase in funds and popularity, REITs again consolidated.
While the rate over the past 3 years has not been as substantial as the 1997–
1999 period, consolidation has continued.4 In fact, 2001 was the third-busiest
year for REIT and REOC merger and acquisitions activity, falling behind 1997
and 1998, respectively. Interestingly, while mergers and property acquisitions
continue on a monthly basis, research on these topics has been mixed.
In an early study, Allen and Sirmans (1987) examined acquirer returns from
REIT mergers during the period of 1977 to 1983, and they found significantly
positive REIT abnormal returns in reaction to merger announcements.
McIntosh, Officer and Born (1989) followed up with an examination of returns
for REIT merger targets during the period of 1969 to 1986. Again, they
found significantly positive returns. Overall, evidence from early research is
consistent with the notion that market participants believe that these mergers
would allow larger REITs to achieve economies of scale through better asset
utilization. However, as noted by Ling and Ryngaert (1997), REITs changed
significantly in the late 1980s and early 1990s. Thus, to see if the regime has
changed, Campbell, Ghosh and Sirmans (2001) examined REIT mergers during
the period of 1994 to 1998. In general, they found that acquirer returns
are slightly negative while target returns are significantly positive. While the
3 For example, in September 1997, Equity Office Properties (EOP) announced a
$4 billion merger with Beacon Properties that added 18.8 million square feet of property
to EOP. The combined holding of 33.4 million square feet of office space made EOP
the largest office owner in the United States, with 245 properties in 21 states and the
District of Columbia. In another transaction, in April 1998 Security Capital Pacific Trust
of Denver announced its acquisition of Security Capital Atlantic Inc. for $1.6 billion,
creating the third-largest apartment REIT in the country, Archstone Communities, with
90,166 apartments and a total expected investment of $5.6 billion in apartment communities.
Another example of REIT consolidation was the September 1998, $5.8 billion
acquisition of Corporate Property Investors by Simon Property Group. At the completion
of the merger, Simon held 241 properties in 35 states, and a staggering $1.3 billion
EBITDA. In addition, in October 1999, Equity Residential Properties Trust acquired
Lexford Residential Trust in a merger valued at about $732.8 million.
4 In February 2000, EOP announced a $4.6 billion merger agreement with Cornerstone
Properties, and in late February 2001, EOP announced a $7.2 billion acquisition of
Spieker Properties, Inc. At the time the EOP and Spieker merger was the largest REIT
merger and acquisition transaction in history until the December 2001 $8 billion takeover
of Security Capital Group byGECapital became the largest REIT merger and acquisition
transaction in history.
326 Ambrose, Highfield and Linneman
positive return for target firms is consistent with the scale economies argument,
the negative return to acquirer firms is not driven by diseconomies of
scale. Instead, because REITs often obtain significant geographical diversification
through mergers, these diversifying mergers can limit opportunities for
economies of scale. Thus, the authors explain that while diversification of all
kinds was beneficial to REITs in the 1970s and 1980s, geographical diversification
is not beneficial to the modern fully integrated REIT because it may
limit economies-of-scale opportunities.
Interestingly, McIntosh, Liang and Thompkins (1991) find evidence of a “smallfirm
effect,” that is, small REITs earn higher average rates of return than large
firms after accounting for risk. Using REIT returns from 1974 to 1988, they
find that although small REITs earn higher returns than large REITs, they are
no more risky than large REITs. While the “small-firm effect” contradicts the
scale economies argument that efficiency gains should be reflected in higher
returns, other studies that examined the impact of announcements concerning
property acquisitions on REIT share prices found insignificant wealth effects
for either sale or purchase transactions. For example, McIntosh, Ott and Liang
(1995) conclude that REITs do not experience significant wealth effects from
the announcement of either acquisition or divestment of properties. Unlike
McIntosh, Liang and Thompkins (1991), who find evidence against economies
of scale, the results of McIntosh, Ott and Liang (1995) find evidence neither
dismissing nor supporting scale economies. As a result, the early evidence
appeared to suggest that significant consolidation of property assets had not
clearly created significant gains in shareholder wealth, but research had also
not eliminated the hope for such gains.
A Brief History of REIT Economies of Scale Research
Unfortunately, as was the case for banking and other industries in the early
phases of their consolidation, it is extremely difficult to identify the presence
of economies of scale in the real estate industry. This problem arises for two
main reasons. First, given the relatively similar size of most REITs and their
recent integration, the statistical technology available to accurately measure
economies of scale is not sufficiently precise to fully capture cross-sectional
variations. For example, Mester (1996) notes that an implicit assumption in
the studies of economies of scale is that all firms in the study have access
to the same cost frontier and should be using the same technology. Second,
the effort required by growing firms to capture scale economies is difficult
and time consuming, with the pain of integration generally occurring prior to
the realization of the benefits. Thus, exploring cross-sectional variations in the
presence of significant merger and acquisition activity understates the benefits of
scale. Moreover, the market is not static as the consolidation activities described
Real Estate and Economies of Scale 327
above have been almost continuous in the real estate industry. As leading firms
merge and achieve competitive advantages, their competitors respond, making
it difficult to capture the effect of scale economies cross-sectionally.
Research from the 1980s and early 1990s focused on the set of very small REITs
that existed from the early 1970s and 1980s, which bear little resemblance to
today’s fully integrated REITs. Not surprisingly, limited statistical evidence
of scale economies exists based on data from the 1970s and 1980s, a period
when debtwas plentiful and the largest players were small by today’s standards.
For example, in the late 1970s and early 1980s the average equity REIT had a
market capitalization of only $28 million, but in 1990 the average equity REIT
had an inflation-adjusted market capitalization of $95 million. Furthermore,
regulations during this earlier period (such as shareholder concentration rules)
severely restricted the ability of REITs to raise sufficient capital to expand
and capture any meaningful economies of scale. Thus, it is not surprising that
researchers using REIT data from the same period find positive stock price
reaction to announcements of asset sales and attribute this to the belief that
scale economies do not exist for REITs, as these companies had no scale.
Research from the late 1990s and early 2000s utilized data from the 1980s and
early 1990s. The research in this period suggests that scale economies exist, at
least for larger REITs.5 These more recent studies attempt to isolate the effect of
economies of scale in REIT expenses, revenue growth and capital costs and build
on the hypothesis that larger REITs have higher earnings growth potential.6 For
example, evidence indicates that the nondiscretionary component of general
and administrative (G&A) expense increases at a decreasing rate as REIT size
increases, and other evidence indicates that scale economies exist in REIT
management costs.7 However, the studies that examine various REIT expense
items (G&A, interest expense, management fees, operating expenses) find that
economies of scale are most present in smaller expense items, suggesting that
while economies of scale exist, the gains from realizing these economies may
be insufficient to lead to massive consolidation in the REIT industry.8
Ambrose et al. (2000) compared REIT income growth and profitability relative
to changes for the market to explore economies of scale using data from the
5 See Bers and Springer (1997), Capozza and Seguin (1998) and Ambrose and Linneman
(2001).
6 See Zell (1997).
7 See Bers and Springer (1997, 1998) and Capozza and Seguin (1998).
Yang (2001) presents evidence for nonlinear economies of scale suggesting that diseconomies
8
of scale may exist for large REITs.
328 A mbrose, Highfield and Linneman
1990s. The results indicate that small REIT net operating income (NOI) growth
rates exceed average growth rates in the markets in which they held properties,
and thus small REITs appear to be generating revenue and operating economies.
However, this does not seem to be the case for the largest REITs. Their findings
indicate that NOI gains, relative to the market, were large prior to 1996, but
are no longer so, with REITs at the end of the 1990s outperforming the market
primarily via revenue enhancement, not cost reduction. Thus, the results from
the Ambrose et al. (2000) study call into question the ability of large REITs to
generate sufficient economies of scale based on income growth. However, this
conclusion must be interpreted with caution because the study was based on a
small sample of residential REITs.
Additional research has tested for economies of scale in REIT capital costs,
improving on previous studies by examining REITs that invest in multiple
property segments (residential, industrial/office and retail) and focusing on the
primary driver of REIT expansion, namely capital. Because REITs are very
capital intensive and the primary source of REIT expansion lies in their ability
to access capital, significant consolidation may be motivated by scale economies
in capital costs. Based on capital costs for equity REITs from 1997 and 1998,
the evidence indicates that REITs realized significant scale economies with
respect to capital costs. Although all REITs appear to generate scale economies
in capital costs, the results from this study show that the scale economies in
capital for large REITs are almost twice as large as the scale economies for
small REITs. As argued by Linneman (1997), the natural implication is that
large REITs are in a position to utilize their economies of scale in capital
costs—their main input cost—to further consolidate the real estate industry.
In addition to the problems inherent to identifying economies of scale in rapidly
changing industries, controversy exists regarding the technology employed in
estimating economies of scale. Estimation of economies of scale for various
industries and firms has a long history within the general finance and economics
literature. In particular, research on economies of scale in the banking
literature is extensive given the natural questions regarding size and efficiencies
in a highly regulated industry. As a result, the banking literature contains
numerous studies that demonstrate the relative merits of various methods of
measuring scale efficiencies. For example, Berger and Humphrey (1997) catalog
130 studies of financial institution economies of scale using a variety of data
and estimation techniques. In an attempt to bring order and clarity to the variation
in results reported in these studies, Berger and Mester (1997) conducted an
exhaustive comparison of the multiple efficiency concepts and differing econometric
measurement methods using a large data set containing over 5,000 U.S.
commercial banks. Their study notes that the most common estimation methods
involve both nonparametric techniques (data envelopment analysis [DEA]
Real Estate and Economies of Scale 329
and free disposable hull analysis) and parametric methods (stochastic frontier
approach, thick frontier approach and the distribution free approach). Berger
and Mester (1997) note that the nonparametric techniques have a number of
significant drawbacks, which lead to findings of lower efficiency means than
reported by the parametric methods. They also note that the parametric stochastic
frontier approach requires explicit and arbitrary distributional assumptions
regarding the components of the error term. Furthermore, Berger and Mester
(1997) discuss the choice of the functional form for the cost and profit functions
when estimating efficiency via one of the parametric methods. Although they
note that the translog form is popular, they indicate that other more flexible functional
forms may provide a better fit to the data. However, the results of their
comparison indicate that “choices made concerning efficiency measurement
usually make very little difference” (p. 897).
Integral to the discussion of estimation technology is the controversy regarding
how to measure economies of scale in REITs. Most empirical studies
of scale economies have focused on industries with readily identifiable and
quantifiable outputs. For example, economists have studied scale economies
in such diverse industries as brewing, where the output is quantified as barrels
or gallons of beer (Tremblay and Tremblay 1988); steel, where output is gross
tonnage of steel produced (McCraw and Reinhardt 1989); electrical utilities,
where output is measured in kilowatts-hours (Christensen and Greene 1976);
banking, where output is measured in terms of asset dollars (Noulas, Ray and
Miller 1990, Mullineaux 1978); and railroads (Caves, Christensen and Swanson
1981), shipping (Jansson and Shneerson 1978) and airlines (Borenstein
1991, Kim and Singal 1993), where output is measured in terms of freighttons
or passenger-miles hauled. The overriding feature across these industries
is that the firms produce essentially a homogenous product with readily identifiable
input factors necessary for production. REITs, however, invest in real
estate assets and derive profits through leasing space in these assets. Unfortunately,
real estate is not homogenous, and this creates difficulties in quantifying
firm output. For example, several studies have proxied REIT output
as total assets because it has the advantage of value-weighting the real estate
investment.9 Because total assets are recorded at book value, this creates difficulties
in comparing firms that may have acquired their assets at different
points in time. Other studies have suggested proxying output as total square
feet available for lease since this measure closely proxies the physical output
measured in other industries. However, this has the drawback of not recognizing
the inherent heterogeneity in real assets (i.e., location, product type and
quality).
See Anderson et al. (2002).
9
330 Ambrose, Highfield and Linneman
As a result of these problems, we focus on a simple metric of scale efficiency.
Our analysis, in the spirit of Altinkihc and Hansen (2000), estimates the effect
of firm size across multiple dimensions of revenue and cost parameters. In
order to capture the possibility that firm cost and profitability factors follow the
traditional U-shaped curve with respect to firm size, we include in our regression
analysis the natural log of total capitalization. However, in an effort to link
our simple efficiency measure with the prevailing literature, we also report an
analysis of REIT scale efficiency through stochastic frontier estimation of the
translog cost function.
Data
We examine a sample of 187 REITs trading on the Nasdaq, New York Stock
Exchange or American Stock Exchange from January 1990 through December
2001. The sample is restricted to REITs with financial data available from
the SNL REIT Datasource and monthly returns available from the Center for
Research in Securities Prices (CRSP) database. In addition to the SNL REIT
Datasource and CRSP data, we also used SNL Interactive and the Dow Jones
News Service to examine all REIT-related press releases from January 1988
through December 2001. From these press releases we compiled a listing of
historical mergers and the size of those mergers.
The summary statistics for our primary variables of interest are showninTable 1.
Because several variables commonly used in REIT research do not have a universal
and undisputed definition, it is important to clarify variable measurement.
First, we define total capitalization as the market value of common equity outstanding
plus the value of preferred shares plus the book value of debt.We then
examine measures of REIT growth prospects including implied capitalization
rates (ICR) and payout ratios. The ICR is rental net operating income (NOI) as
a percentage of average real estate value providing a rough proxy for pricing
REIT assets.10 That is, the ICR is effectively a going-in rate of return for investors:
The higher the ICR, the more the assets are “discounted” when sold
on a per share basis, and vice versa. This implies that lower ICRs indicate a
market premium while higher ICRs indicate a market discount on the value of
REIT assets. The payout ratio is the cash dividend per share of common stock
as a percentage of funds from operations (FFO) per share. Because capital is
costly, low payout ratios encourage growth.
To examine REIT revenue and expenses we include net operating income (NOI)
as a percentage of sales, rental revenue as a percentage of sales and general
and administrative (G&A) expenses as a percentage of sales. If economies of
scale are evident in REITs, then we would expect to find that both revenue
measures increase while the overhead costs fall as a proportion of revenue as
size increases.
We evaluate REIT profitability using return on equity (ROE) and funds from
operations yield (FFO Yield). ROE is defined as net income as a percentage
of average equity. Assuming that consolidation and growth leads to revenue
enhancement, we should find an expansion in net income relative to
increases in equity providing an increase in ROE. We define FFO Yield as
funds from operations as a percentage of the REIT’s market price per share
of common equity.11 In the presence of scale economies, we should find that
ROE and FFO Yield provide similar results; that is, profitability increases with
size.
Due to possible variations in risk across different firms, property types or organizational
structures, we estimate each REIT’s equity beta, a measure of
systematic variation in returns relative to the market, using the Capital Asset
Pricing Model (CAPM) framework. If economies of scale are present, we
should expect to find lower systematic risk as firm size increases.
Using CRSP we obtain the monthly returns for each REIT in the sample and
the CRSP value-weighted market index. We use the 3-month T-Bill secondary
market rate as a proxy for the risk-free rate.12 After converting the T-Bill yearly
rate to a monthly rate, we estimate REIT equity betas using the market model:
Ri ,t = ai + bi Rm
,t + ei ,t, (1)
where Ri ,t and Rm ,t represent the monthly returns for REIT i and the market
portfolio in excess of the risk-free rate for the 24 prior months, αi is the regression
intercept, βi is the estimated equity beta for REIT i and εi ,t is the standard
error term.13
11 As defined by SNL Datasource, funds from operations (FFO is typically derived by
taking GAAP net income, excluding gains and losses from sales of properties and debt
restructuring, and adding back real estate depreciation expenses.
12 T-Bill rates were obtained from the FRED_ database at the Federal Reserve Bank of
St. Louis (http://research.stlouisfed.org/fred/data/irates/tb3ms).
13 For example, December 2001 betas are estimated by regressing the REIT returns less
the risk-free rate against the CRSP value-weighted market index less the risk-free rate
for the period from January 2000 to December 2001.
Real Estate and Economies of Scale 333
We also examine the relationship between a REIT’s cost of capital and its size.
We calculate each REIT’s weighted average cost of capital (WACC) as follows:
WACC = kd
_
TD
TC
_
+ kp
_
PE
TC
_
+ ke
_
CE
TC
_
, (2)
where TC = TD + PE + CE and kd , kp and ke are the cost of debt (TD),
preferred equity (PE) and common equity (CE), respectively. The cost of debt
and preferred equity are estimated as the ratio of total interest expense to book
value of debt and the ratio of preferred dividends to book value of preferred
stock, respectively. The cost of equity is estimated via the CAPM using the
beta calculated in (1). It is intuitive that having a capital cost advantage in a
capital-intensive industry such as real estate should lead to a true competitive
advantage. In effect, a REIT with lower capital costs can hire better talent,
charge lower rents, and buy more expensive properties while still financially
outperforming competitors with a high cost of capital.
Next we calculate the return on capital (ROC) and economic value added spread
(EVA_ Spread) for each REIT. ROC is calculated by dividing NOI by capital
employed, and EVA_ Spread is defined as net operating profit after taxes minus
the capital charge, where the capital charge is the REIT’s WACC multiplied
by capital employed.14 In the presence of economies of scale we should
find a positive relation between total capitalization and both ROC and EVA_
Spread.
We also control for growth, leverage, industry, time and organizational structure
effects in our regression analysis. With regard to growth rates, asset growth is
the growth in the book value of total assets over the previous year, and FFO
Growth is the growth in FFO over the previous year. To control for leverage
effects we include the book value of total debt, expressed as a percentage of the
total capitalization of the REIT in dollar terms (Total Debt/Total Cap), and the
ratio of the book value of short-term debt to the book value of long-term debt
(ST Deb/LT Debt).
We control for industry effects through binary variables based on the
self-reported property focus of the REIT: Recreation, Restaurant, Healthcare,
Retail, Industrial, Residential, Self Storage, Hotel, Office and Other/
Diversified. Although all of the firms are publicly traded, most have a specific
property focus (such as hotels, apartments or self-storage facilities, etc.),
See Stewart (1991), Walbert (1994) and O’Byrne (1996) for a detailed discussion of
14
EVA_ calculations.
334 A
mbrose, Highfield and Linneman
with each sector effectively a separate industry with its own operating and competitive
dynamics. For example, Table 2 shows that recreation REITs typically
have the smallest total capitalization, approximately $358 million, while office
REITs have the largest market capitalization, approximately $1.6 billion.With
regard to growth prospects, residential REITs post the lowest ICR at 9.7%,
while hotel REITS have the highest average ICR at 12.8%. FFO Yields range
from 6.6% for restaurants, to a high of 10.2 for self-storage REITs. Looking
at revenue and expense measures, diversified REITs have the lowest NOI to
revenues ratio, 57.9%, and healthcare REITs have the highest NOI to revenues
ratio. The rental revenue to total revenue varies from 54.9 for hotels to 95.7
for self-storage REITs. While maximizing rental revenue, self-storage REITs
also minimize general and administrative (G&A) expenses as a percent of total
revenue at 3.7%, reflecting the fact that the costs of managing cubic space
is minimal. Lastly, regarding profitability, restaurants have the lowest ROE
(−0.7%) and highest payout ratio (134.7%), while residential and retail REITS
have the highest ROE (11.5%). Self-storage REITs provide the lowest payout
ratios among the groups considered.
Finally, we control time and organizational structure effects through a set of
binary variables reflecting self-managed and self-advised firms interacted with
the observation year. As shown in Table 3, after adjusting for inflation, during
the sample period the total capitalization for an average REIT increased at
a compound annual rate of approximately 22% (from $194 million in 1990
to $1,683 million in 2001). Table 4 displays the variation across operating
structures.
Methods and Results
Regression Analysis
The information in Tables 2, 3 and 4 reveals notable differences across property
types, time and organizational structures. So, are changes in growth prospects,
revenues and expense measures, profitability measures, systematic risk and
capital costs due to economies of scale or is it simply property focus, time or
organizational structure that drives these accounting and financial performance
measures? To answer this question we estimate multivariate regressions using
a random effects model with unbalanced panels.15
Growth Prospects
Table 5 provides the multivariate regression results for the impact of size
on growth prospects controlling for asset growth, FFO growth, industry effects,
time and organizational structure. Overall, the results show evidence of
economies of scale. For the ICR regression, we find a negative and significant coefficient
for firm size, and a significantly positive coefficient for the quadratic
effect, indicating that as REIT total capitalization increases there is a corresponding
nonlinear decrease in ICRs.16 The results indicate that the presence
of value-generating benefits obtainable by increasing size because the market
places a premium on larger REITs in the form of higher asset prices. In addition
to size, other factors contribute to the determination of implied capitalization
rates. We find that higher asset growth rates lower implied capitalization rates,
but higher debt usage has the opposite effect.
In addition to lower implied capitalization rates, we find evidence that firm
size affects payout ratios in a similar, nonlinear fashion. The regression results
indicate that larger firms have higher payout ratios. As expected, we find that
higher debt ratios negatively affect payout ratios and payout ratios appear to
vary across property types. These results suggest diminishing returns to scale
with respect to retained earnings. While this issue as a whole is driven by
dividend discipline of the board rather than operating or capital market issues,
firms that lower payout ratios are effectively lowering their cost of capital by
using retained earnings in place of the more expensive capital markets. As a
result, smaller firms, with lower payout ratios, appear to be capitalizing on the
ability to utilize retained earnings to fund growth opportunities. However, as
firm size increases, retained earnings are no longer sufficient to fund significant
growth opportunities.
Revenue and Expense Measures
One of the main driving forces behind consolidation in the real estate industry
is the belief that larger REITs should obtain some measure of efficiency gains
in improving profit margins by increasing revenues and lowering expenses.
Estimates of these effects are presented in Table 6.
The ratio of NOI to total revenue is a popular proxy for firm profitability.
We find that asset growth, FFO growth, total debt ratios and property focus
are significant in determining NOI as a percentage of total revenue, but we
also find a positive and significant coefficient on firm size, and a statistically
significant, negative coefficient for the quadratic size effect. Thus, firm NOI
increases at a decreasing rate as the firm grows larger. Similarly, examining
rental revenue, we find that total debt ratios and property focus are significant
in explaining rental revenue as a percentage of total revenue. Interestingly, we
find no evidence to suggest a relationship between firm size and rental revenue
as a percentage of sales; however, we do find that mergers, in the current year
or previous year, have a negative effect on rental revenue as a percentage of
total revenue. Thus, while larger firms can increase profit margins, most likely
through lower expenses, rental revenue as a percentage of sales is unaffected by
size, but the effects of a recent merger may negatively impact rental revenue.
As mentioned previously, another popular argument for economies of scale
in real estate is that as a REIT grows larger, it should decrease its portion of
G&A expenses as a percentage of revenue. In other words, it can allocate its
G&Aexpenses over a larger asset base. Again, property focus shows significant
explanatory power in determining the ratio of G&A expenses to total revenues,
and the significantly negative coefficient on firm size suggests that larger REITs
are increasing shareholder value by lowering overhead expense ratios. Overall
the evidence here supports the notion that larger firms are better at increasing
revenues while decreasing costs, a true-scale economies competitive advantage
we should expect to be evident in profitability measures.
Profitability Measures
In Table 7 we report the models regarding profitability measures, beginning
with net income to total equity (ROE). In general, we find evidence that larger
firms achieve higher returns for shareholders. The positive and significant size
coefficient and the negative and significant quadratic coefficient suggest that
ROE increases at a decreasing rate. Again, this result is not surprising given
the findings for size effects on NOI and overhead costs. We also find strong
evidence that property focus affects ROE.
Next, we consider the effect of size on FFO Yield. Again, we find strong support
for the theory that size influences FFO Yield. The results indicate a nonlinear
effect, with FFO Yield increasing at a decreasing rate, as total capitalization
increases. As expected, in the presence of scale economies, we should find that
ROE and FFO Yield provide similar results; that is, profitability increases with
size.
the assumption that a REIT’s cost will deviate from the industry cost frontier
as a result of random noise and inefficiency. Following Mester (1996), we use
a translog cost function as the functional form, and we estimate the model
using maximum likelihood techniques. After estimation, we calculate the scale
economies measure (SCL) where SCL < 1 implies scale economies, SCL =
1 implies constant returns to scale, and SCL > 1 implies diseconomies of
scale.17 Table 10 reports the scale economies measure evaluated at the mean
output, input price levels and financial capital cost levels over the entire sample
and size quartiles based on total assets. For the entire sample we find a SCL
of 0.988, which implies that a 1.0% increase in the level of output, holding
risk and the quality of the output constant, would lead to a 0.988% increase
in total costs. As shown in Table 10, all size quartiles support the findings
presented above. This suggests that economies of scale continue to exist for
REITs.
Conclusions
REITs have experienced significant growth and received increased attention in
the financial literature during the past decade.With this increased attention and
growth, many experts have raised the possibility of economies of scale in real
estate, and theory suggests that, as REITs have grown and merged, they provide
an almost perfect laboratory for examining any such scale economies.
Building on past research in the REIT economies-of-scale debate, we test for
economies of scale in REITs by examining growth prospects, revenue and expense
measures, profitability ratios, systematic risk and cost of capital measures.
Utilizing REIT property data from the SNL REIT Datasource as well as REIT
and market return data from the CRSP database, we examine differences in
REITs across organizational structures, across time and across property focus.
We then estimate multivariate regression models to determine whether REITs
are gaining economies of scale with respect to firm size.
A serious limitation of the earlier studies of REIT economies of scale is the
inability to separate size-related advantages versus a period of rapidly expanding,
strong markets. Thus, our updated analysis now covers an 11-year period
that includes a full market cycle (significant market expansion followed by contraction),
the importance of which should not be understated. In addition to the
size of our data set, we also improve on other REIT economies-of-scale studies
by using only real dollar amounts, leveraging SNL data with news-reported
information on merger activity and specifically controlling for REIT property
focus.
As a whole, our results suggest that small REITs have available efficiency gains
in the area of growth, and we find strong evidence suggesting that large REITs
are succeeding at lowering costs, specifically G&A expenses, and increasing
profit margins. Thus, it is no surprise that there is a direct relationship between
firm profitability, as measured by return on equity, and firm size.We also confirm
these results with the FFO yield.
With respect to equity betas, we find an inverse relationship between betas
and firm size indicating that as REITs grow larger they find ways of lowering
systematic risk. These findings are supported by the inverse relationship between
REIT size and WACC. In fact, for all cost-of-capital measures we find
significant economies of scale, such as the positive and significant relationship
between firm size and ROC and EVA_ Spreads. In addition to the regression re348
Ambrose, Highfield and Linneman
sults discussed above, we also estimate a stochastic econometric frontier model
using a translog cost function as the functional form. The results of the scale
economies measure calculated from the efficient frontier suggests that REITs
can lower costs while increasing output for all size classes considered.
Overall, we find continued support for economies of scale in real estate through
growth and consolidation in REITs, but we should note that these results, while
particularly strong, could be underestimated, particularly in the areas of revenue
and expenses. In general, econometric analysis of economies of scale is often
difficult due to “local” data relative to the true scale, the lag of integration
and expansion efficiency as well as the nature of competition. While we have
adjusted for scale issues by using ratios when available, converting all dollar
amounts to real dollars, and controlling for lag effects by incorporating merger,
merger amount and development lags, competition may lead to understated
results in econometric models. Like other industries where economies of scale
abide, competition responds in a manner that erodes competitive advantages.
This means that costs can decline while returns remain constant simply because
all market participants are doing the same thing. Thus, the results presented in
this article are effectively downward biased, but they nonetheless support the
theory of economies of scale in the REIT market.
Furthermore, it should be noted that the old generation of REIT data, dating
back to the 1980s, is of limited relevance because many of those firms were in
their infancy and not operating anywhere near a productive level of efficiency.
During the early and mid-1990s the REIT industry experienced a period of rapid
growth, and by the late 1990s and early 2000s we have seen a market contraction
and slower period of growth. As such, the REIT market has changed and many
of these changes may not become apparent until 5 or 10 years into the future.
Finally, the market has proven that REITs have dramatically increased in size
over the past decade, and they will continue to do so. In 1990 a $300 million
REIT was a large company, but now a $300 million company is one of
the smallest. However, we cannot misinterpret our results as a projection of
doom and gloom for small REITs. As suggested by Stigler (1958), there is
no single, magical, optimal firm size. Just because a REIT is smaller or larger
than our estimates does not mean that it cannot compete. In fact, some small,
well-managed REITs are sustaining competitive advantages over large, poorly
managed REITs. A large, inefficient REIT is simply a wasted opportunity, and
just increasing size will never guarantee success—a point clearly illustrated by
the negative shareholder return for REITs engaging in geographical diversifying
mergers (Campbell, Ghosh and Sirmans 2001). As stated by Linneman
(2002), a large public company is not a “silver bullet” that will solve all problems,
and, moreover, it should not be the goal of all REITs to simply become
Real Estate and Economies of Scale 349
larger. The real point here is that economies of scale are visible in REITs, with
part of this effect due to increases in operating efficiency, and the remainder a
reward to risk. As a result, the market has generally forced REITs to expand
and consolidate.
Back to black
The Treasury squashes hopes that the agencies may ever be private again
SINCE 2008 Fannie Mae and Freddie Mac,
America’s two housingfinance
giants, have
been on life support, spared from
insolvency by an intravenous drip of
taxpayer cash. Lately, however, the
companies have shown signs of life: earlier
this month both reported their biggest
profits since being forced into
“conservatorship” four years ago (see
chart).
That has sent a frisson through investors
clutching preferred shares issued back
when the companies minted money by
using their quasigovernmental
status to borrow cheap and buy or guarantee most
residential mortgages in America. Between March and early August, many of Fannie’s old
preferred shares, which now trade over the counter, jumped from around $1.50 to more
than $3 (still a fraction of their $25 par value).
Several factors explain the turn in the companies’ fortunes.
As home prices have stabilised, unemployment has
gradually declined and troubled loans have been
restructured or written off, the two have set aside ever
smaller provisions for loan losses. In the second quarter
Fannie reversed some prior provisions, adding $3 billion to
the bottom line. More importantly, mortgages issued after
2008 now make up more than half of the companies’
portfolios. Thanks to stricter underwriting terms and slowing
homeprice
falls, those mortgages sport far lower loantovalue
ratios and delinquency rates than “legacy” mortgages
issued during the bubble years of 200508.
As a result, both companies are firmly back in the black.
Fannie recorded a $7.8 billion profit in the first half of this
year, compared with a loss of $16.9 billion in all of 2011. Her sibling earned $3.6 billion,
against a loss of $5.3 billion.
Under the terms of the original bailout,
the Treasury invested just enough each quarter in
senior preferred shares to cover the companies’ losses and to keep their net worth above
zero. In return the companies pay the Treasury a 10% dividend on those shares.
Perversely, to distribute those dividends both companies have routinely had to draw even
more cash from the government. But in the second quarter, neither had to. That has
begun to bring down the net cost of the bailout,
from a peak of $151 billion at the end of
2011 to $142 billion now. A decade from now, the administration reckons the tally will be
just $28 billion.
The good news had one unwelcome sideeffect,
however. With profits now exceeding
their dividends, the companies’ net worth began to grow, arousing hope that one day
dividends might resume on their old shares. On August 17th the Treasury drove a stake
through those hopes, announcing that rather than pay a 10% dividend, the companies
would henceforth simply send every penny of profit its way. For those who did not get the
message, the Treasury rammed it home: the move underlined that they “will not be
allowed to retain profits, rebuild capital, and return to the market in their prior form.”
Fannie’s old preferred shares promptly sank back to about $1.
While the companies’ status as public utilities now appears crystal clear, the federal
government’s longterm
role in housing finance is as muddy as ever. In the short run, it is
indispensable. Fannie, Freddie and the Federal Housing Administration, another
government agency, currently back some 90% of newly originated mortgages. In the long
run, Democrats and Republicans agree that Fannie and Freddie should be wound down,
but concur on little else. The Obama administration has proposed several options for a
smaller federal role in backstopping mortgages: the companies’ regulator is exploring
how to draw private insurers into the mortgage market via losssharing
arrangements with
Fannie and Freddie. Mitt Romney and congressional Republicans want to wind both
companies down, but have not specified any remaining role for government. As with
every policy of consequence in America, the fate of Fannie and Freddie must await the
election.
The housing sector is now (September 2007) at the root of three distinct but related problems: (1)
a sharp decline in house prices and the related fall in home building; (2) a subprime mortgage problem
that has triggered a substantial widening of all credit spreads and the freezing of much of the credit
markets; and (3) a decline in home equity loans and mortgage refinancing that could cause greater
declines in consumer spending. Each of these could by itself be powerful enough to cause an economic
downturn. The combination could cause a very serious recession unless there are other offsetting forces.
In this paper, I discuss each of these and then comment on the implications for monetary policy.
Housing, Credit Markets and the Business Cycle
Martin Feldstein*
The housing sector is now (September 2007) at the root of three distinct but related
problems: (1) a sharp decline in house prices and the related fall in home building ; (2) a
subprime mortgage problem that has triggered a substantial widening of all credit spreads and
the freezing of much of the credit markets; and (3) a decline in home equity loans and mortgage
refinancing that could cause greater declines in consumer spending. Each of these could by itself
be powerful enough to cause an economic downturn. The combination could cause a very
serious recession unless there are other offsetting forces. In this paper, I discuss each of these
and then comment on the implications for monetary policy .
Falling House Prices
Robert Shiller (2007) presents evidence of a sharp surge in house prices after the year
2000. More specifically, national indexes of real house prices., real rents and real construction
costs moved together for 25 years until the year 2000. After that, while construction costs and
rents continued to move in parallel, real house prices then rose rapidly so that by 2006 they were
70 percent higher than equivalent rents., driven in part by a widespread popular belief that
houses were an irresistible investment opportunity. How else could an average American family
buy an asset appreciating at 9 percent a year , with 80 percent of that investment financed by a
mortgage with a tax deductible interest rate of 6 percent, implying an annual rate of return on the
initial equity of more than 25 percent?
But at a certain point home owners recognized that house prices – really the price of land
– wouldn’t keep rising so rapidly and may decline. That fall has now begun, with a 3.4 percent
decline in housing prices over the past 12 months and an estimated 9 percent annual rate of
decline in the most recent month for which data are available (a Goldman-Sachs estimate for
June 2007.) The decline in house prices accelerates offers to sell and slows home buying,
causing a rise in the inventory of unsold homes and a decision by home builders to slow the rate
of construction. Home building has now collapsed, down 20 percent from a year ago, to the
lowest level in a decade.
Edward Leamer (2007) explained that such declines in housing construction were a
precursor to 8 of the past 10 recessions. Moreover, major falls in home building were followed
by a recession in every case except when the Korean and Vietnam wars provided an offsetting
stimulus to demand..
Why did home prices surge in the past 5 years? While a frenzy of irrational house price
expectations may have contributed, there were also fundamental reasons. Credit became both
cheap and relatively easy to obtain. When the Fed worried about deflation it cut the Fed funds
rate to one percent in 2003 and promised that it would rise only very slowly. That caused
medium term rates to fall, inducing a drop in mortgage rates and a widespread promotion of
mortgages with very low temporary “teaser” rates.
Mortgage money also became more abundant as a result of various institutional changes
as Ben Bernanke (2007), Ned Gramlich (2007) , and Richard Green and Susan Wachter (2007)
explained. Subprime mortgages were the result of legislative changes (especially the
Community Reinvestment Act) and of the widespread use of statistical risk assessment models
by lenders,. In addition, securitization induced a lowering of standards by lenders who did not
hold the mortgages they created. Mortgage brokers came to replace banks and thrifts as the
primary mortgage originators. All of this had been developing since the 1990s but these
developments contributed to the current mortgage problems when rates fell after 2000.
If house prices now decline enough to reestablish the traditional price-rent relation1 ,
there will be serious losses of household wealth and resulting declines in consumer spending.
Since housing wealth is now about $21 trillion, even a 20 percent nominal decline would cut
wealth by some $4 trillion and might cut consumer spending by $200 billion or about 1.5 percent
of GDP. The multiplier consequences of this could easily push the economy into recession.
A 20 percent national decline would mean smaller declines in some places and larger
declines in others. A homeowner with a loan to value ratio today of 80 percent could eventually
find himself with a loan that exceeds the value of his house by 20 percent or more. Since
mortgages are effectively non-recourse loans, borrowers can walk away with no burden on future
incomes. While experience shows that most homeowners continue to service their mortgages
even when the loan balances slightly exceed the value of their homes, it is not clear how they
would behave if the difference is substantially greater. The decision to default would be more
likely if house prices are expected to fall further. .
Once defaults became widespread, the process could snowball, putting more homes on
the market and driving prices down further. Some banks and other holders of mortgages could
see their highly leveraged portfolios greatly impaired. Problems of illiquidity of financial
institutions could become problems of insolvency.
Widening Credit Spreads
I turn now to the second way in which the housing sector is affecting our economy: the
impact of subprime mortgages on credit spreads and credit availability.
For several years now, informed observers have concluded that risk was underpriced in
the sense that the differences in interest rates between U.S. Treasury bonds and riskier assets
(i.e., the credit spreads) were very much smaller than they had been historically.
Some market participants rationalized these low credit spreads by saying that financial
markets had become less risky. Better monetary policies around the world have reduced
inflation and contributed to smaller real volatility. Securitization and the use of credit derivatives
were thought to disperse risk in ways that reduced overall risk levels. Most emerging market
governments now avoid overvalued exchange rates and protect themselves with large foreign
Housing, Credit Markets & the Business Cycle.092307 -4-
exchange reserves. There was also the hope based on experience that the Federal Reserve would
respond to any financial market problems by an easing of monetary policy.
Many of us were nevertheless skeptical that risk had really been reduced to the extent
implied by existing credit spreads. It looked instead like the very low interest rates on high grade
bonds were incenting investors to buy riskier assets in the pursuit of yield. Many portfolio
managers were enhancing the return on their portfolios by selling credit insurance – i.e., by using
credit derivatives to assume more risk – and by using credit to leverage their investment
portfolios on the false assumption that the basic portfolio had relatively small risk. Investors
took comfort from the apparent risk transfer in structured products. And less sophisticated
investors were buying such structured products without actually recognizing the extent of the
risk.
Most of the institutional investors who thought that risk was mispriced were nevertheless
reluctant to invest based on that view because of the cost of carrying that trade. Since virtually
all such institutional investors are agents and not principals, they could not afford to take a
position that involved a series of short term losses. They would appear to be better investment
managers by focusing on the short term gains that could be achieved by going with the herd to
enhance yield by assuming increased credit risk.
But these investors also shared a widespread feeling that the day would come when it
would be appropriate to switch sides, selling high risk bonds and reversing their credit derivative
positions to become sellers of risk. No one knew just what would signal the time to change.
It was the crisis in the subprime mortgage market that provided the shock that started the
wider shift in credit spreads and credit availability.
Subprime mortgages are mortgage loans to high risk borrowers with low or uncertain
incomes, high ratios of debt to income, and poor credit histories. These are generally floating
rate mortgages, frequently with high loan-to-value ratios and very low initial “teaser” rates. A
realistic assessment would imply that the borrowers would have trouble meeting the monthly
payments once the initial teaser rate period ended and the interest rate rose to a significant
premium over the rates charged to prime (i.e., low risk) borrowers.
Borrowers with subprime credit ratings nevertheless took these adjustable rate loans with
low teaser rates because they wanted to get in on the house price boom that was sweeping the
2This reflected in part the use of data on borrower experience in recent years in which
Is the 2007-2008 U.S. sub-prime mortgage financial crisis truly a new and different phenomena? Our
examination of the longer historical record finds stunning qualitative and quantitative parallels to 18
earlier post-war banking crises in industrialized countries. Specifically, the run-up in U.S. equity and
housing prices (which, for countries experiencing large capital inflows, stands out as the best leading
indicator in the financial crisis literature) closely tracks the average of the earlier crises. Another
important parallel is the inverted v-shape curve for output growth the U.S. experienced as its economy
slowed in the eve of the crisis. Among other indicators, the run-up in U.S. public debt and is actually
somewhat below the average of other episodes, and its pre-crisis inflation level is also lower. On the
other hand, the United States current account deficit trajectory is worse than average. A critical question
is whether the U.S. crisis will prove similar to the most severe industrialized-country crises, in which
case growth may fall significantly below trend for an extended period. Or will it prove like one of
the milder episodes, where the recovery is relatively fast? Much will depend on how large the shock
to the financial system proves to be and, to a lesser extent, on the efficacy of the subsequent policy
response.
I. Introduction
The first major financial crisis of the 21st century involves esoteric instruments,
unaware regulators, and skittish investors. It also follows a well-trodden path laid down
by centuries of financial folly. Is the “special” problem of sub-prime mortgages this time
really different?
Our examination of the longer historical record, which is part of a larger effort on
currency and debt crises, finds stunning qualitative and quantitative parallels across a
number of standard financial crisis indicators. To name a few, the run-up in U.S. equity
and housing prices that Graciela L. Kaminsky and Carmen M. Reinhart (1999) find to be
the best leading indicators of crisis in countries experiencing large capital inflows closely
tracks the average of the previous eighteen post World War II banking crises in industrial
countries. So, too, does the inverted v-shape of real growth in the years prior to the
crisis. Despite widespread concern about the effects on national debt of the early 2000s
tax cuts, the run-up in U.S. public debt is actually somewhat below the average of other
crisis episodes. In contrast, the pattern of United States current account deficits is
markedly worse.
At this juncture, the book is still open on the how the current dislocations in the
United States will play out. The precedent found in the aftermath of other episodes
suggests that the strains can be quite severe, depending especially on the initial degree of
trauma to the financial system (and to some extent, the policy response). The average
drop in (real per capita) output growth is over 2 percent, and it typically takes two years
3
to return to trend. For the five most catastrophic cases (which include episodes in
Finland, Japan, Norway, Spain and Sweden), the drop in annual output growth from peak
to trough is over 5 percent, and growth remained well below pre-crisis trend even after
three years. These more catastrophic cases, of course, mark the boundary that
policymakers particularly want to avoid.
II. Post War Bank-Centered Financial Crises: The Data
Our comparisons employ a small piece of a much larger and longer historical data
set we have constructed (see Reinhart and Kenneth S. Rogoff, 2008.) The extended data
set catalogues banking and financial crises around the entire world dating back to 1800
(in some cases earlier). In order to focus here on data most relevant to present U.S.
situation, we do not consider the plethora of emerging market crises, nor industrialized
country financial crises from the Great Depression or the 1800s. Nevertheless, even in
the smaller sample considered in this paper, the refrain that “this time is different”
syndrome has been repeated many times.
First come rationalizations. This time, many analysts argued, the huge run-up in
U.S. housing prices was not at all a bubble, but rather justified by financial innovation
(including to sub-prime mortgages), as well as by the steady inflow of capital from Asia
and petroleum exporters. The huge run-up in equity prices was similarly argued to be
sustainable thanks to a surge in U.S. productivity growth a fall in risk that accompanied
the “Great Moderation” in macroeconomic volatility. As for the extraordinary string of
outsized U.S. current account deficits, which at their peak accounted for more than twothirds
of all the world’s current account surpluses, many analysts argued that these, too,
4
could be justified by new elements of the global economy. Thanks to a combination of a
flexible economy and the innovation of the tech boom, the United States could be
expected to enjoy superior productivity growth for decades, while superior American
know-how meant higher returns on physical and financial investment than foreigners
could expect in the United States.
Next comes reality. Starting in the summer of 2007, the United States
experienced a striking contraction in wealth, increase in risk spreads, and deterioration in
credit market functioning. The 2007 United States sub-prime crisis, of course, has it
roots in falling U.S. housing prices, which have in turn led to higher default levels
particularly among less credit-worthy borrowers. The impact of these defaults on the
financial sector has been greatly magnified due to the complex bundling of obligations
that was thought to spread risk efficiently. Unfortunately, that innovation also made the
resulting instruments extremely nontransparent and illiquid in the face of falling house
prices.
As a benchmark for the 2007 U.S. sub-prime crisis, we draw on data from the
eighteen bank-centered financial crises from the post-War period, as identified by
Kaminsky and Reinhart (1999), Gerard Caprio et. al. (2005), and Andrew Mullineux
(1990).
These crisis episodes include:
The Big Five Crises: Spain (1977), Norway (1987), Finland (1991),
Sweden (1991) and Japan (1992), where the starting year is in parenthesis.
Other Banking and Financial Crises: Australia (1989), Canada (1983),
Denmark (1987), France (1994), Germany (1977), Greece (1991), Iceland
5
(1985), and Italy (1990), and New Zealand (1987), United Kingdom
(1973, 1991, 1995), and United States (1984).
The “Big Five” crises are all protracted large scale financial crises that are
associated with major declines in economic performance for an extended period. Japan
(1992), of course, is the start of the “lost decade,” although the others all left deep marks
as well.
The remaining rich country financial crises represent a broad range of lesser
events. The 1984 U.S. crisis, for example, is the savings and loan crisis. In terms of
fiscal costs (3.2 percent of GDP), it is just a notch below the “Big Five” 1 . Some of the
other 13 crisis are relatively minor affairs, such as the 1995 Barings (investment) bank
crisis in the United Kingdom or the 1994 Credit Lyonnaise bailout in France. Excluding
these smaller crises would certainly not weaken our results, as the imbalances in the runsup
were minor compared to the larger blowouts.
III. Comparisons
We now proceed to a variety of simple comparisons between the 2007 U.S. crisis
and previous episodes. Drawing on the standard literature on financial crises, we look at
asset prices, real economic growth, and public debt. We begin in Figure 1 by comparing
the run-up in housing prices. Period T represents the year of the onset of the financial
crisis. By that convention, period T-4 is four years prior to the crisis, and the graph in
1 Thefiscal costs of cleaning up after banking crises can be enormous. The fiscal cleanup from Sweden’s
1991 crisis was 6 percent of GDP and Norway’s 1987 crisis was 8 percent. Spain’s post-1977 cleanup cost
over 16 percent of GDP. Estimates for Japan’s bill vary widely, with many in excess of 20 percent of GDP.
6
each case continues to T+3, except of course in the case of the U.S. 2007 crisis, which
remains in the hands of the fates.2 The chart confirms the case study literature, showing
the significant run-up in housing prices prior to a financial crisis. Notably, the run-up in
housing prices in the United States exceeds that of the “Big Five” .
Figure 1: Real Housing Prices and Banking Crises
95
100
105
110
115
120
125
130
135
t-4 t-3 t-2 t-1 T t+1 t+2 t+3
Index
Average for banking crises in
advanced economies
US, 2003=100
Index t-4=100
Average for the "Big 5"
Crises
2 Forthe United States, house prices are measured by the Case-Shiller index, described and provided in
Robert Shiller (2005). The remaining house price data were made available by the Bank for International
Settlements and are described in Gregory D. Sutton (2002).
7
Figure 2 looks at real rates of growth in equity market price indices. (For the United
States, the index is the S&P 500; Reinhart and Rogoff, 2008, provide the complete listing
for foreign markets.)
Once again, the United States looks like the archetypical crisis country, only more
so. The Big Five crisis countries tended to experience equity price falls earlier on than
the U.S. has, perhaps because the U.S. Federal Reserve pumped in an extraordinary
amount of stimulus in the early part of the most recent episode.
Figure 2: Real Equity Prices and Banking Crises
80
85
90
95
100
105
110
115
120
125
130
135
t-4 t-3 t-2 t-1 T t+1 t+2 t+3
Index
Average for banking crises in
advanced economies
US, 2003=100
Index t-4=100
Average for the "Big 5"
Crises
8
Figure 3: Current Account Balance/GDP on the Eve of Banking Crises
-7
-6
-5
-4
-3
-2
-1
0
t-4 t-3 t-2 t-1 t
percent of GDP
Average for banking crises in
advanced economies
U.S.
Figure 3 looks at the current account as a share of GDP. Again, the United States
is on a typical trajectory, with capital inflows accelerating up to the eve of the crisis.
Indeed, the U.S. deficits are more severe, reaching over six percent of GDP. As already
mentioned, there is a large and growing literature that attempts to rationalize why the
United States might be able run a large sustained current account deficit without great
risk of trauma. Whether the U.S. case is quite as different as this literature suggests
remains to be seen.
Real per capita GDP growth in the run-up to debt crises is illustrated in Figure 4.
The United States 2007 crisis follows the same inverted V shape that characterizes the
earlier episodes. Growth momentum falls going into the typical crisis, and remains low
9
for two years after. In the more severe “Big Five” cases, however, the growth shock is
considerably larger and more prolonged than for the average. Of course this implies that
the growth effects are quite a bit less in the mildest cases, although the U.S. case has so
many markers of larger problems that one cannot take too much comfort in this caveat.
Figure 4: Real GDP Growth per Capita and Banking Crises
(PPP basis)
-2
-1
0
1
2
3
4
5
t-4 t-3 t-2 t-1 t t+1 t+2
Percent
Average for banking crises in
advanced economies
US
Average for the "Big 5"
Crises
Figure 5 looks at public debt as a share of GDP. Rising public debt is a near
universal precursor of other post-war crises, not least the 1984 U.S. crisis. It is notable
that U.S. public debt rises much more slowly than it did in run-up to the Big Five crisis.
However, if one were to incorporate the huge buildup in private U.S. debt into these
measures, the comparisons would be notably less favorable.
10
The correlations in these graphs are not necessarily causal, but in combination
nevertheless suggest that if the United States does not experience a significant and
Figure 5: Public Debt and Banking Crises
90
110
130
150
170
190
210
230
250
t-4 t-3 t-2 t-1 T t+1 t+2 t+3
Index
Average for banking crises in
advanced economies
US, 1997=100
Index t-10=100
Average for the "Big 5" Crises
protracted growth slowdown, it should either be considered very lucky or even more
“special” that most optimistic theories suggest. Indeed, given the severity of most crisis
indicators in the run-up to its 2007 financial crisis, the United States should consider
itself quite fortunate if its downturn ends up being a relatively short and mild one.
11
IV. Conclusions
Tolstoy famously begins his classic novel Anna Karenina with “Every happy
family is alike, but every unhappy family is unhappy in their own way.” While each
financial crisis no doubt is distinct, they also share striking similarities, in the run-up of
asset prices, in debt accumulation, in growth patterns, and in current account deficits.
The majority of historical crises are preceded by financial liberalization, as documented
in Kaminsky and Reinhart (1999). While in the case of the United States, there has been
no striking de jure liberalization, there certainly has been a de facto liberalization. New
unregulated, or lightly regulated, financial entities have come to play a much larger role
in the financial system, undoubtedly enhancing stability against some kinds of shocks,
but possibly increasing vulnerabilities against others. Technological progress has plowed
ahead, shaving the cost of transacting in financial markets and broadening the menu of
instruments.
Perhaps the United States will prove a different kind of happy family. Despite
many superficial similarities to a typical crisis country, it may yet suffer a growth lapse
comparable only to the mildest cases. Perhaps this time will be different as so many
argue. Nevertheless, the quantitative and qualitative parallels in run-ups to earlier postwar
industrialized-country financial crises are worthy of note. Of course, inflation is
lower and better anchored today worldwide, and this may prove an important mitigating
factor. The United States does not suffer the handicap of a fixed exchange rate system.
On the other hand, the apparent decline in U.S. productivity growth and in housing prices
does not provide a particularly favorable backdrop for withstanding a credit contraction.
12
Another parallel deserves mention. During the 1970s, the U.S. banking system
stood as an intermediary between oil-exporter surpluses and emerging market borrowers
in Latin America and elsewhere. While much praised at the time, 1970s petro-dollar
recycling ultimately led to the 1980s debt crisis, which in turn placed enormous strain on
money center banks.3 It is true that this time, a large volume of petro-dollars are again
flowing into the United States, but many emerging markets have been running current
account surpluses, lending rather than borrowing. Instead, a large chunk of money has
effectively been recycled to a developing economy that exists within the United States’
own borders. Over a trillion dollars was channeled into the sub-prime mortgage market,
which is comprised of the poorest and least credit worth borrowers within the United
States. The final claimant is different, but in many ways, the mechanism is the same.
Finally, we note that although this paper has concentrated on the United States,
many of the same parallels hold for other countries that began experiencing housing price
duress during the 2007, including Spain, the United Kingdom and Ireland. Indeed, as
Appendix Table 1 reminds us for the panic of 1907 and the crash of 1929, the global
dimension of the current crisis is neither new nor unique to this episode. As history has
shown, there can be similarities across unhappy families, too.
I. INTRODUCTION
After a number of warning signs, the U.S. “subprime mortgage crisis” became a headline
issue in February 2007. Notwithstanding the bankruptcy of numerous mortgage companies,
historically high delinquencies and foreclosures, and a significant tightening in subprime
lending standards, the impact thus far on core U.S. financial institutions has been limited.
And while some structured credit hedge funds have suffered large losses, mortgage
securitization appears to have helped disperse the impact throughout the financial system, in
contrast to the Savings & Loan crisis of the early 1990s. The credit cycle is thus largely
playing out in the securities and derivatives markets, rather than on bank balance sheets.
This paper reviews the history and structure of the subprime market. The results suggest that
new origination and funding technology appear to have made the financial system more
stable at the expense of undermining the effectiveness of consumer protection regulation.
Potential solutions to the management of this trade-off are then explored.
II. ORIGINS AND HI STORY OF THE SU BPRIME M RTGAGE MA
O RKET
Subprime mortgages are residential loans that do not conform to the criteria for “prime”
mortgages, and so have a lower expected probability of full repayment. This assessment is
usually made according to the borrower’s credit record and score, debt service-to-income
(DTI) ratio, and/or the mortgage loan-to-value (LTV) ratio. Borrowers with low credit
scores, DTIs above 55 percent, and/or LTVs over 85 percent are likely to be considered
subprime. So-called “Alt-A” loans fall into a gray area between prime and subprime
mortgages. These began as a more flexible alternative to prime loans, mainly for borrowers
who met all of the credit score, DTI, and LTV prime criteria, but did not provide full income
documentation.
Several legal milestones facilitated the development of the modern subprime mortgage
market. Interest rate caps imposed by states were preempted by federal legislation in 1980
while lenders were allowed to offer adjustable-rate mortgages from 1982. Also the Tax
Reform Act of 1986 left residential mortgages as the only consumer loans on which the
interest was tax deductible. This made home equity withdrawal (for instance, through “cashout”
refinancing of a mortgage) a preferred means of financing home improvements and
personal consumption, relative to other forms of consumer loans (Klyuev and Mills, 2006).
Automated underwriting and securitization were also key developments in reducing the cost
of subprime mortgage lending. Automated underwriting (using computer models rather than
loan officer judgment) has made loan origination more cost efficient, while advances in
statistical credit scoring have led to more accurate and consistent assessments of borrower
4
credit risk.2 Securitization also facilitated market growth by dispersing risk, providing
investors with a supply of highly-rated securities with enhanced yield, and opening up the
origination business to non-depository specialty finance companies (Box 1).
Consequently, subprime lending
developed as a specialist loan class in the
mid-90s and facilitated a substantial
expansion of home ownership (Figure 1).3
These developments allowed a relaxation
of credit rationing for borrowers—such
as the poor, or those in minority
communities—previously considered too
risky by traditional lenders.4
Loans for subprime borrowers were once
predominantly guaranteed by the Federal
Housing Association (FHA). However,
subprime loans have displaced FHAguaranteed
lending (Figure 2) due to the
FHA’s less aggressive product mix, its
lack of flexibility to changing market
conditions, and its low lending limits.5
Mortgage originators also complain that
the fees that they earn on FHAguaranteed
mortgages do not adequately
compensate for their higher processing
costs.
Despite advances in credit scoring techniques, the subprime market experienced its first
“crisis” in 1998-99. Subprime loan delinquencies transpired to be higher than anticipated by
the new models while the East Asian and LTCM crises reduced investor appetite for higherrisk
mortgage securities. As a result, the majority of the largest subprime lenders went
bankrupt.6
PID RE
III. THE RA CENT EX
PANSION OF SU
BPRIME LE NDING
Until 2003, the majority of mortgage originations were “prime conforming” loans. These
were then purchased by two government-sponsored housing enterprises (GSEs - Fannie Mae
and Freddie Mac). However, by 2006, over half of all originations did not meet the GSEs’
“conforming” criteria.
This paper links the current sub-prime mortgage crisis to a decline in lending standards
associated with the rapid expansion of this market. We show that lending standards declined
more in areas that experienced larger credit booms and house price increases. We also find
that the underlying market structure mattered, with entry of new, large lenders triggering
declines in lending standards by incumbent banks. Finally, lending standards declined more
in areas with higher mortgage securitization rates. The results are consistent with theoretical
predictions from recent financial accelerator models based on asymmetric information, and
shed light on the relationship between credit booms and financial instability.
Recent events in the market for mortgage-backed securities have placed the U.S. subprime
mortgage industry in the spotlight. Over the last decade, this market has expanded rapidly,
evolving from a small niche segment to a major portion of the U.S. mortgage market.
Anecdotal evidence suggests that this trend was accompanied by a decline in credit standards
and excessive risk taking by lenders, and possibly by outright fraud. Indeed, the rapid
expansion of subprime lending, fueled by financial innovation, loose monetary conditions,
and increased competition, is seen by many as a credit boom gone bad. Yet, few attempts
have been made to link empirically lending standards and delinquency rates in the subprime
mortgage market to its rapid expansion. To our knowledge, our paper is the first to do so.
How does the recent increase in delinquency rates relate to the boom? How did lending
standards change over the expansion? How did changes in local market structure and
financial innovation affect lender behavior during the boom? What was the role of monetary
policy? To answer these questions, we use data from over 50 million individual loan
applications combined with information on local and national economic variables.
Reminiscent of the pattern linking credit booms with banking crises, current mortgage
delinquencies appear related to past credit growth. In particular, delinquency rates rose more
sharply in areas that experienced larger increases in the number and volume of originated
loans (Figure 1). We find evidence that this relationship is linked to a decrease in lending
standards, as measured by a decline in loan denial rates and a significant increase in loan-toincome
ratios, not explained by an improvement in the underlying economic fundamentals.
Consistent with recent theories suggesting that banks behave more aggressively during
booms than in tranquil times, the size of the boom mattered. Denial rates declined more and
loan-to-income ratios rose more where the number of loan applications rose faster.
The subprime boom also shared other characteristics often associated with aggregate boombust
credit cycles, such as financial innovation (in the form of securitization), changes in
market structure, fast rising house prices, and ample aggregate liquidity. We find evidence
that all these factors were associated with the decline in lending standards. Denial rates
declined more in areas with a larger number of competitors and were further affected by the
entry into local markets of large financial institutions. The increasing recourse to loan sales
and asset securitization appears to have affected lender behavior, with lending standards
deteriorating more in areas where lenders sold a larger proportion of originated loans.
Lending standards also declined more in areas with more pronounced housing booms.
Finally, easy monetary conditions also played an amplifying role. These effects were more
pronounced in the subprime mortgage market than in the prime mortgage market, where loan
denial decisions were more closely related to economic fundamentals.
We obtain this evidence in an empirical model where, in addition to taking into account
changes in macroeconomic fundamentals, we control for changes in the distribution of
4
applicant borrowers and for the potential endogeneity of some of the explanatory variables.
Specifically, we develop a two-stage regression framework, explained in detail later on, that
exploits individual loan application data to control for changes in the quality of the pool of
loan applicants. We focus on loan applications rather than originations to reduce further the
concern for simultaneity biases. For further robustness, we run an instrumental variable
specification of our model, where we instrument the subprime applications variable with the
number of applications in the prime market.
The contribution of this paper is twofold. First, the paper sheds some light on the origins of
the current crisis by establishing a link between credit expansion and lending standards in the
subprime mortgage market, and by identifying increased loan sales and changes in the
structure of local credit markets as factors amplifying the decline in denial rates and the
increase in loan-to-income ratios.
Second, the boom-bust cycle of the subprime mortgage industry, beyond being of interest in
itself, provides an excellent “lab case” to gain insights into the black box of credit booms in
less developed credit markets. Subprime borrowers are generally riskier, more
heterogeneous, can post less collateral, and have shorter or worse credit histories (if any) than
their prime counterparts. These are all features often prevalent in developing countries. At
the same time, the wealth of information available and the geographical variation (Figure 2)
in this market allow us to control for several factors, such as changes in the pool of loan
applicants, that are difficult to take into account when studying episodes of aggregate credit
growth. Thus, the subprime mortgage market provides an almost ideal testing ground for
theories of intermediation based on asymmetric information (and adverse selection in
particular).
From a policy point of view, the evidence in this paper alerts against the dangers arising from
lax standards during credit booms, and it is relevant for the debate on cyclical management of
prudential regulation and on the potential effects of monetary policy on banks’ risk-taking
(Jimenez et al., 2007). To the extent that during booms standards decline more than justified
by economic fundamentals, our findings are consistent with the view that bankers have “an
unfortunate tendency” to lend too aggressively at the peak of a cycle and that most bad loans
results from this aggressive type of lending. That said, credit booms may still be beneficial.
While, in light of the recent financial turmoil, it is easy to argue that standards were
excessively lax, it is much harder to compute the benefits associated with greater access to
credit and, hence, the net welfare effect of the subprime expansion.
The rest of the paper is organized as follows. Section II reviews the related literature. Section
III provides a description of the data and introduces some stylized facts. Section IV describes
our empirical methodology. Section V presents the results. Section VI concludes.
5
II. RELATED LI TERATURE
Several studies examine the interaction between economic fluctuations and changes in bank
credit (Bernanke and Lown, 1991, Peek and Rosengren, 2000, Black and Strahan, 2002, and
Calomiris and Mason, 2003). However, little evidence has been collected on how lending
standards are related to credit booms. Asea and Blomberg (1998) find that loan
collateralization increases during contractions and decreases during expansions, while Lown
and Morgan (2003) show that lending standards are associated with innovations in credit.
Jimenez, Salas, and Saurina (2006) find that during booms riskier borrowers obtain credit
and collateral requirements decrease.
A few papers have examined the recent boom from a house-price perspective, while not
strictly focusing on the subprime market (Himmelberg et al., 2005, and Case and Shiller,
2003). The literature on subprime mortgages has instead largely focused on issues of credit
access and discrimination and on what determines access to subprime versus prime lenders.
Our loan level analysis builds on a model from Munnell et al. (1996) who show that race has
played an important, although diminishing, role in the decision to grant a mortgage. A few
papers examine how local risk factors affect the fraction of the market that uses subprime
lending (Pennington-Cross, 2002). Other studies focus on how borrowers choose a mortgage
and on their decision to prepay or default on a loan (Deng et al., 2000, Campbell and Cocco,
2003, and Cutts and Van Order, 2005).
A few recent papers focus on how securitization affects the supply of loans (Loutskina and
Strahan, 2007) and mortgage delinquencies. Demyanyk and Van Hemert (2007) find that
delinquency and foreclosure rates of subprime borrowers are to a large extent determined by
high loan-to-value ratios. Mian and Sufi (2007) link the increase in delinquency rates to a
disintermediation-driven increase in loan originations, while Keys et al. (2007) find that
loans that are easier to securitize default more frequently.
Most theoretical explanations for variations in credit standards rely on financial accelerators
based on the interaction of asymmetric information and business cycle factors (Bernanke and
Gertler,1989, Kiyotaki and Moore, 1997, and Matsuyama, 2007; see Ruckes, 2004, for a
review of this literature). Others focus on the potential for herding behavior by bank
managers (Rajan, 1994), on banks’ limited capacity in screening applications (Berger and
Udell, 2004), or on how strategic interaction among asymmetrically informed banks may
lead to changes in lending standards during booms (Gorton and He, 2003, and Dell’Ariccia
and Marquez, 2006).
III. DATA AND DE SCRIPTIVE ST ATISTICS
We combine data from several sources. Our main set of data consists of economic and
demographic information on applications for mortgage loans. We use additional information
on local and national economic environment and on home equity loan market conditions to
construct our final data set.
6
The individual loan application data come from the Home Mortgage Disclosure Act
(HMDA) Loan Application Registry. Relative to other sources, including LoanPerformance
and the Federal Reserve Bank’s Senior Loan Officer Opinion Survey, this dataset has the
important advantage of covering extensive time-series data on both the prime and subprime
mortgage markets. The availability of data on the prime mortgage market provides us with a
control group generally unavailable to studies focusing on aggregate credit. By comparing
prime and subprime mortgage lenders we are also able to identify differences between the
two lending markets. Given the different risk profiles of the prime and subprime markets, we
include variables that proxy for the risk characteristics of a loan application to enhance
comparability of the results across the two markets.
Enacted by Congress in 1975, HMDA requires most mortgage lenders located in
metropolitan areas to collect data about their housing-related lending activity and make the
data publicly available. Whether a lender is covered depends on its size, the extent of its
activity in a metropolitan statistical area (MSA), and the weight of residential mortgage
lending in its portfolio. Comparisons of the total amount of loan originations in the HMDA
and industry sources indicate that around 90 percent of the mortgage lending activity is
covered by the loan application registry (Table 1).
Our coverage of HMDA data starts from 2000 and ends in 2006. This roughly corresponds to
the picking up of both the housing boom and the rapid subprime mortgage market expansion
(Figure 3). HMDA data does not include a field that identifies whether an individual loan
application is a subprime loan application. In order to distinguish between the subprime and
prime loans, we use the subprime lenders list as compiled by the U.S. Department of Housing
and Urban Development (HUD) each year. HUD has annually identified a list of lenders who
specialize in either subprime or manufactured home lending since 1993. HUD uses a number
of HMDA indicators, such as origination rates, share of refinance loans, and proportion of
loans sold to government-sponsored housing enterprises, to identify potential subprime
lenders.
Since 2004, lenders are required to identify loans for manufactured housing and loans in
which the annual percentage rate (APR) on the loan exceeds the rate on the Treasury security
of comparable maturity by at least three (five, for second-lien loans) percentage points and
report this information to the HMDA. The rate spread can be used as an alternative indicator
(to the HUD list) to classify subprime loans. For the years with available data, the ranking of
subprime lenders using the rate spread variable alone coincides closely with the ranking in
the HUD list. The HUD list of subprime lenders is also preferable to the rate spread
information for a number of reasons. First, rate spreads are not available prior to 2004.
Second, subprime loans do not necessarily have APRs that are three (or five) percentage
points above a comparable Treasury rate but may reflect fees and yield spread premiums or
other borrower characteristics determined by the lender. Third, and most importantly, the rate
spread in HMDA is available only for originated loans, making it impossible to calculate
denial rates for prime and subprime applications separately.
7
We remove some observations with missing HMDA data from the sample and also focus on
the subset of loans that are either approved or denied. First, we drop applications with loan
amounts smaller than $1,000 because loan values are expressed in units of thousands of
dollars and rounded up to the nearest number. Second, applicant income is left-censored at a
value of $10,000. We therefore eliminate applicants with missing applicant income or
applicant income of exactly $10,000. Third, we drop loans for multi-family purpose from the
sample, as this is a distinct market from the overall mortgage market for single family homes.
Fourth, we drop federally insured loans and refinancing loans as their risk profile is likely to
differ from that of other loans. Finally, and importantly, we eliminate all application records
that did not end in one of the following three actions: (i) loan originated, (ii) application
approved but loan not originated, or (iii) application denied. Other actions represent dubious
statuses (e.g. application withdrawn by applicant) or loans purchased by other financial
institutions. Including purchased loans would amount to double-counting as these loans are
reported both by the originating institution and the purchasing institution.
We supplement the HMDA information with MSA-level data on economic and social
indicators published by federal agencies, including annual data on macroeconomic variables,
such as personal income, labor and capital remuneration, self-employment, and population
from the Bureau of Economic Analysis (BEA), data on unemployment and inflation
(consumer price index) from the Bureau of Labor Statistics (BLS), data on total population
from the Census Bureau, and data on house price appreciation in a given MSA (based on a
quarterly housing price index) from the Office of Federal Housing Enterprise Oversight
(OFHEO). We also obtain data on “seriously delinquent” subprime loans, defined as
subprime loans with 60 or more days delay in payment, from LoanPerformance, a private
data company. Data on these delinquency rates are available only for 2004 onwards.
Over the last decade, subprime mortgage lending has expanded rapidly both in terms of the
number of loans originated and the average loan amount. Subprime mortgage originations
almost tripled since 2000, reaching $600 billion in 2006. Against an also fast growing market
for prime mortgages, this boom brought the share of subprime lending from 9 percent in
2000 to 20 percent of all mortgage originations in 2006. Average loan amount also grew
reaching $132,784 in 2006 or 90 percent of the prime mortgage average amount. In absolute
terms, the subprime market reached a size of about $1.3 trillion in 2006.
A first look at our data suggests that rapid growth in subprime loan volume was associated
with a decrease in denial rates on subprime loan applications and an increase in the loan-toincome
ratio on the loans originated by subprime lenders (Figure 4). These casual
observations lend some support to the view that lending decisions are influenced by market
conditions and rapid credit growth episodes tend to beget trouble later on. In the next
sections, we explore these relations in a more formal setting.
Table 2 presents the name and definitions of the variables we use and the data sources. Table
3 presents the sample period summary statistics of these variables at the loan application and
8
MSA levels. The data cover a total of 387 MSAs for a period of 7 years (2000 to 2006),
amounting to a total of 2,709 observations. For the entrant and incumbent variables,
summary statistics are based on data for the period 2001 onwards only, as entry data is
missing for the first year of the sample period. The summary statistics show that about one in
five loan applications is denied, while about one-fourth of all loans are extended by subprime
lenders. As expected, the denial rate of subprime lenders is much higher (about 2.5 times)
than the denial rate of prime lenders.
IV. EMPIRICAL ME THODOLOGY
We rely on two main indicators of lending standards: the application denial rate and the loan
to income ratio. We focus primarily on regressions at the MSA level. We control for changes
in the economic environment in the MSA by including variables that have been shown to be
good predictors of loan denial decisions at the individual level (see Munnell et al., 1996),
such as average income, income growth, the unemployment rate, and the self-employment
rate. We include a measure of house price appreciation to take into account the role of
collateral. The number of competing lenders is a proxy for the competitive conditions in the
MSA. Finally, we include the number of loan applications as a measure of credit expansion.
We find this variable preferable to the number of loans originated or the growth in credit
volume as it is arguably less endogenous to the dependent variable (i.e., denial rates).
Endogeneity may remain a concern to the extent that potential borrowers might be deterred
from applying for a loan if denial rates are generally high in their area. For this reason, we
also estimate an instrumental variable specification of the model (details later in the paper).
In addition, we control for time-invariant MSA specific factors and for time-variant
nationwide-uniform factors by including MSA and time fixed effects.
We estimate the following linear regression model:
DRi t = αt + γ
i+ β1 AVGINCi t + β2 INCGROWi t + β3 UNEMPi t + β4 SELFEMPi t + β5 POPi t+
β6 COMPi t + β7 HPAPPi t-1 + β8 APPLi t + εi t,
(Eq. 1)
where DRi t is the average denial rate of mortgage loan applications for home purchase and
refinance purposes in MSA i in year t. It is computed as the number of loan applications
denied divided by the total number of all loan applications in a given MSA using loan-level
data at individual banks, and hence, takes on values between 0 and 1.2 All explanatory
variables are also measured at the MSA level. AVGINC denotes average income, INCGROW
is income growth, UNEMP is unemployment rate, SELFEMP is self-employment rate, POP
is the log of total population, COMP is the number of competing lending institutions, HPAPP
2 Weestimate regression equation 1 using ordinary least squares as well as using truncated regression methods.
The results remain the same.
9
is the annual change in house price appreciation, and APPL is the log of the number of loan
applications. The error term εi t has the standard properties. MSA and time fixed effects
control for time-constant regional idiosyncrasies and nationwide changes in economic
conditions. The first five variables are meant to control for the general economic and
demographic conditions in the MSA. We expect areas with higher per capita income and
income growth to have lower denial rates; areas with higher unemployment rates and larger
proportions of self-employed people (whose income may be considered less stable) to have
higher denial rates; and areas with larger populations, proxying for market size, to have lower
denial rates.
The number of competing lenders in the MSA is meant to capture the effects of competition
on lending standards. Since theory does not deliver unambiguous predictions, we do not have
a strong prior on the sign of this coefficient.3 The house price appreciation variable is
computed over the same period as the denial rates, although the results are not sensitive to
using one-period lagged changes in house price appreciation. We expect this variable to have
a negative coefficient. Price increases raise the net worth of borrowers, reducing their default
risk. At the same time, lenders may gamble on a continued housing boom to evergreen
potentially defaulting borrowers. Finally, our working assumption is that if banks did not
change their lending standards during the boom, the variable measuring credit expansion
should not be statistically significant after controlling for the other factors affecting the
banks’ decision. If instead banks lent more leniently in regions and times of fast credit
expansion, we should find a negative and significant coefficient for this variable. In that case,
we would have established a link between credit expansion and loan quality that, in turn,
would explain why we now observe higher delinquency rates in regions that experienced
greater booms.
Theoretical models focusing on adverse selection (such as Broecker, 1990, and Riordan,
1993) predict that an increase in the number of competing lenders in a market may have the
perverse effect of increasing lending interest rates and make banks more choosy (tighten
standards). However, it has also been argued that, when local borrowers have an
informational advantage, the threat of new entry may induce incumbents to cut standards and
trade loan quality for market shares. We test for these effects by focusing on the behavior of
incumbent lenders when large nationwide institutions entered local markets. To that purpose
we augment the model in equation (1) with a variable measuring the market share of new
entrants that belong to the top 20 lending institutions in the country. We compute the market
share in terms of number of loan applications, not originations, to limit concerns about
endogeneity. We expect the coefficient on the entrants variable to be negative since we
already control for the adverse selection effect considered by Broecker (1990), among others,
with the number of competing lenders in the region.
3 See Dell’Ariccia and Marquez (2006) for a discussion of this issue.
10
For robustness purposes, we construct an alternative denial rate-based measure of lending
standards. We borrow and augment the empirical model presented in Munnell et al. (1996) to
estimate bank’s loan approval decision with individual application data, though we do not
have all variables they consider. Specifically, we do not have data on borrower credit scores.
We augment their specification by including several new variables, including whether or not
the loan is being used for refinancing purposes and whether or not the household income of
the loan applicant is below the poverty line (as applicable in the year of loan application). We
expect the latter to be particularly important in the case of subprime loans because applicants
for subprime loans tend to have low income. We estimate the following logit specification at
the loan application level for the year 2000:
Djk = αk + γ1INCj + γ2LIRj + γ3POVj + γ4REFINj + γ5OCCj + γ6Fj + γ 7Bj + γβ8Wj + εjk,
(Eq. 2)
where j denotes loan application j, k denotes lender k, α
k denotes lender-specific fixed
effects, and Dj k is a dummy variable that takes a value of one if lender k denied loan
application j in year 2000, and zero otherwise. All explanatory variables are measured at the
loan application level. INC is applicant income, LIR is the loan-to-income ratio, POV is a
dummy variable denoting whether or not the applicant income is below the poverty line for a
family of four, REFIN is a dummy variable denoting whether or not the purpose of the loan is
to refinance an existing loan, OCC is a dummy variable denoting whether or not the property
financed by the loan is intended for owner occupancy, F is a dummy variable indicating
whether or not the applicant is female, B is a dummy variable indicating whether or not the
applicant is black, and W is a dummy variable indicating whether or not the applicant is
white (the default option being of Hispanic origin).
Next, we use the estimate coefficients of Regression (2) to forecast the denial rate for
mortgage applications in subsequent years, and aggregate the residuals of this regression at
the MSA level. Finally, we use this constructed measure of prediction errors as the dependent
variable for our main model. The advantage of this two-stage regression approach over using
simple, unadjusted denial rates is that it takes into account changes in the pool of applicant
borrowers that are difficult to control for at the MSA level.
As an alternative measure of lending standards, we consider the average loan-to-income ratio
in the MSA. Other things equal, an increase in this ratio would signal a looser attitude in
banks’ decisions to grant loans. We estimate the following regression model:
LIRi t = αt + γi + β1 AVGINCi t + β2 INCGROWi t + β3 UNEMPi t + β4 SELFEMPi t + β5 POPi t+
β6 COMPi t + β7 HPAPPi t-1 + β8 APPLi t +εi t,
(Eq. 3)
where the set of explanatory variables is the same as in regression model 1.
11
Our final set of regressions are aimed at measuring the effect of securitization on bank
lending decisions. The specifications presented above are augmented with a variable
measuring the percentage of loans in an MSA that are sold within a year from origination.
According to the view that securitization causes moral hazard in loan originators, we should
find that lending standards are looser (lower denial rates and higher loan-to-income ratios) in
MSAs with higher securitization rates.
V. EMPIRICAL FI NDINGS
We find robust evidence that lending standards eased in the subprime mortgage industry
during the fast expansion of the past few years. After controlling for economic fundamentals,
lenders appear to have denied fewer loan applications and to have approved larger loans.
Results for the denial rate regression, controlling for MSA fixed effects, are in Table 4.
Column (1) reports results for all lenders, while columns (2) and (3) report results separately
for either only prime lenders or subprime lenders (where subprime lenders are defined
according to the annual list compiled by the HUD). This sample breakdown between prime
and subprime lenders allows us to identify different characteristics of the two lending
markets, including differences in the evolution of lending standards. Most coefficients have
the expected sign. Starting from our main variables of interest, a faster rate of house price
appreciation was associated with lower denial rates. This, as discussed before, is consistent
with the notion that lenders were to some extent gambling on speculative borrowers, but may
also reflect the positive effect of higher borrower net worth on creditworthiness. Notably, this
effect was much more pronounced in the subprime relative to the prime mortgage market
where both these factors are likely to be more relevant. Denial rates in both markets are also
lower in MSAs where applicants tend to have higher income. In the subprime mortgage
market, denial rates were lower in more competitive markets as measured by the number of
competitors in the MSA. This coefficient was, instead, not statistically significant for the
prime market. In the subprime mortgage market, the denial rate was also negatively and
significantly associated with the number of loan applications in the MSA. Given that we are
including MSA fixed effects and thus effectively estimating regressions in first differences,
this result suggests that the lending boom (as captured by changes in the number of
applications) was associated with a reduction in lending standards (as captured by changes in
denial rates). In the prime market, however, denial rates are positively and significantly
associated with the number of applications, consistent with the notion that the lending
standards in the prime market were tightened as applications grew. This suggests different
credit boom dynamics in these two markets. In the subprime market, the decline in standards
associated with the rise in the number of applications is consistent with theories of
intermediation where asymmetric information among lenders plays an important role. In the
prime market, the publicly available credit history of borrowers makes these frictions less
likely to be relevant, and the tightening of standards in reaction to a growing number of
applications may reflect an expected deterioration in the quality of the pool of applicants.
Indeed, the coefficient for the prime market loses significance when we control for changes
12
in the characteristics of the applicant pool (see below). The rest of the control variables have
the expected sign, but are generally not significant.
The finding that denial rates are negatively associated with the number of competitors only in
the subprime, and not in the prime mortgage market, also suggests that the decrease in
lending standards was associated with different forces in the prime and subprime mortgage
markets. In the subprime market, the evidence is consistent with a decline in standards linked
to lenders’ strategic interaction under asymmetric information and speculative behavior. In
contrast, for the prime market, it is more difficult to reject the hypothesis of a fundamentaldriven
decline in lending standards. This is consistent with our prior that, relative to
fundamentals, the deterioration in lending standards was more pronounced in the subprime
mortgage market where the class of borrowers tends to be riskier than in the prime market.
A comparison of year effects across the different specifications shows that denial rates
decreased until the end of 2003 and then increased from 2004 onwards, though only in the
prime mortgage market. In the subprime mortgage market, after controlling for other factors,
denial rates did not vary much over the period 2002 to 2006. Following several years of low
interest rates, the U.S. started tightening monetary policy in mid-2004 by increasing interest
rates. While denial rates in the prime mortgage market closely mimic the evolution of interest
rates in the U.S., with denial rates increasing sharply in 2005 compared to 2004, this is not
the case for the subprime market, where denial rates do not increase in 2005 compared to
2004 (although they do increase somewhat in 2006). This suggests that, while in the prime
market monetary policy changes reflected quickly in the denial rate likely through their effect
on loan affordability,4 this did not happen for subprime mortgages. Indeed, a regression
specification replacing the year fixed effects with the Federal Fund rate returned a positive
coefficient for the prime market, but not for the subprime (not reported).5
The economic effect of our main findings is substantial. From regression (3) in Table 4, it
follows that changes in the number of loan applications (a proxy for credit expansion) have a
particularly strong effect on denial rates in the subprime market. For example, a one standard
deviation increase in the log of the number of applications reduces MSA-level denial rates of
subprime lenders by 4 percentage points, which is substantial compared to a standard
deviation of subprime denial rates of 8 percentage points. The effect of applications on denial
4 This is also consistent with the idea of a negative relationship between bank risk-taking and the monetary
policy rate. This hypothesis is explored at length, though in a different context, in Jimenez et al. (2007).
5 One explanation for this result relies on the fact that prime mortgages are mostly fixed-rate and are by
definition underwritten for the fully-indexed cost while subprime mortgages are mostly adjustable-rate loans
with low teaser rates. It is possible that lending standards in the subprime market were already flawed in the
sense that the denial decision was based on initial debt-to-income ratios calculated using the teaser rate instead
of considering payment shocks that would occur with the reset of the loan rate. In that case, denial rates would
not respond to higher short-term interest rates, concealing the potential impact of monetary policy on lending
standards.
13
rates is significantly more negative in the subprime market than in the prime market. In fact,
the effect is positive and significant in the prime market. A one standard deviation increase in
the number of competitors reduces MSA-level subprime denial rates by 3 percentage points,
slightly smaller than the effect of applications though still substantial. For the prime market,
we obtain no significant relationship between denial rates and the number of competitors.
Finally, a comparison of coefficients across regressions (2) and (3) shows that a one standard
deviation increase in house price appreciation reduces MSA-level denial rates by 2
percentage points in the subprime market compared to only 1 percentage point in the prime
market (compared to a standard deviation of denial rates of about 7 percent in both markets).
A. Effects of Changes in the Pool of Applicant Borrowers
Changes in the pool of applicant borrowers not captured by aggregate controls could partly
explain our findings on the association between the number of applications and denial rates.
The results, however, are broadly the same when, following the two-step approach described
above, we control for changes in the underlying borrower population using data on individual
borrower characteristics.
To this end, we first identify in Table 5 which characteristics are likely to explain the
decision on a loan application. We follow earlier studies on mortgage lending to form a list
of variables that would account for the economic factors that might shape the financial
institution’s decision.6 These regressions are based on a sample of close to 5 million loan
applications in 2000, and include lender-specific fixed effects. The regression coefficients
presented in Table 5 are odds ratios. We find that loan applications are more likely denied if
borrowers have low income, though this effect is only significant in the prime mortgage
market. Applications with higher loan-to-income ratios, denoting riskier loans, are more
likely denied in the subprime mortgage market, as expected, though we find the opposite
effect in the prime mortgage market. Taken together, these results indicate that applicant
income affects lending decisions in a nonlinear fashion, and differently in prime and
subprime markets. This is in part because applicants with higher incomes, who primarily
apply for prime loans, also tend to apply for larger loans. Loan applications are also more
likely denied for male applicants in the subprime market and for female applicants in the
prime market, while applications of African-American descent are more likely denied in both
markets (as compared to white applicants or applicants of Hispanic descent). White
applicants also appear to be less likely denied a mortgage in the prime market. Finally, loan
applications for refinancing purposes are more likely denied, while owner occupation does
not significantly affect the loan denial decision.
Next, we estimate the regression model with the MSA-level aggregated prediction errors
from the model estimated in Table 5 as dependent variable. The results of these regressions
6 See Munnell et al. (1996) and references therein.
14
(all of which include MSA fixed effects) are reported in Table 6. These results, where we
abstract from certain borrower characteristics that determine a lender’s decision on a loan
application, are broadly consistent with the findings in Table 4. Again, we find that denial
rates in both prime and subprime markets tend to deteriorate more in areas with a stronger
acceleration in house price appreciation. Subprime denial rates also respond negatively to an
increase in competition, as measured by an increase in the log of the number of competitors,
and to an increase in the number of loan applications, capturing the expansion of the credit
market.
B. Identification and Robustness Issues
One should be careful in interpreting the estimated coefficients as causal relationships. As
proxy for credit market expansion, the loan application series has arguably a smaller
endogenous component than the loan originations series. That said, at least in theory, there
remains some potential for reverse causality to the extent that potential borrowers may be
deterred from applying for a loan if denial rates are generally high in their locale. While our
focus on total applications (rather than applications in the subprime market only) partly
assuages the potential for an endogeneity bias, for further robustness we estimate an
instrumental variable (IV) specification of our model. In this particular specification, we use
the log of applications in the subprime market as our main regressor, but we instrument it
with the log of the number of prime applications. These two series are highly correlated (the
correlation coefficient is over 0.8), while, at least in theory, there should not be a direct
negative link between the denial rate in the subprime market and the number of applications
in the prime market. If anything, this relationship should be positive, as higher denial rates in
the subprime market would make the prime market more attractive. Indeed, the correlation
between the denial rates in the subprime and prime markets in our sample is only about 0.1,
suggesting that denial rates in both markets are largely independent from one another. For
comparison purposes, we also include the OLS regression of the specification that includes
the number of applications in the subprime market. These OLS and IV results are presented
in columns (1) and (2) of Table 7. The IV estimates broadly confirm our earlier results,
suggesting that our findings are not the product of an endogeneity bias. The F-test of
excluded instruments supports the choice of our instrument. The evidence supports the notion
of a negative causal link between an increase in the number of applications and denial rates
in the subprime market. For robustness, we also estimate our model using the number of
originated loans and the total loan volume as alternative measures of market expansion,
obtaining similar results (Table 7, columns 3 and 4).
Similarly, house price changes may be affected by lending standards to the extent that a
decline in standards and an increase in the local supply of mortgages leads to an increase in
demand for housing. To address this concern, we consider a specification where we lag the
house price variable one period. The results, presented in column (5) of Table 7, confirm our
earlier findings that denial rates are negatively affected by (lagged) house price appreciation.
Admittedly, some concern about endogeneity between denial rates and house price
15
appreciation remains since, while lagged house price changes are not directly affected by
lending standards, it is conceivable that the expectation of a decline in standards, and hence,
of an increase in the supply of mortgage liquidity may trigger speculative pressures on the
housing market. Crowe (2008) finds that in MSAs with a larger portion of the population
belonging to Evangelical churches house prices tend to rise disproportionately when the
“Rapture Index” rises.7 This index maps current events into a subjective probability of an
imminent coming of a time of “extreme and terrible” events and as such is independent from
denial rates at the MSA level. We can then use the interaction term of the share of
Evangelicals in the MSA population and change in the Rapture index as an instrument for
house price appreciation. The results of this exercise are reported in column (6) of Table 7
and confirm our original estimates.
C. Sensitivity Analysis: Time and Size Effects
So far we have imposed the coefficient linking lending standards to market expansion to be
invariant across time and markets of different sizes. Yet, anecdotal evidence suggests that the
relaxation of standards associated with housing boom frenzy was more prevalent in large
metropolitan areas and that abundant liquidity due to loose monetary condition was at least in
part to be blamed. In this section, we attempt to formally test those assumptions by explicitly
allowing the “credit boom” coefficient to vary over time, with changes in monetary policy,
and across markets of different sizes.
First, we consider how this relationship has evolved over our sample period by
interacting the log number of applications with a simple linear trend (Table 8, column 1). The
negative and significant sign of the coefficient of this interacted term suggests that the link
between boom and standards become increasingly stronger over the sample period.
We test for the role of monetary policy by interacting the log number of applications with the
Federal Fund rate. The positive and significant coefficient of this interacted term suggests
that the effect of credit expansion on lending standards is indeed stronger when monetary
policy is loose (Table 8, column 2). Notably, this effect survives when we control for a time
trend (column 3).
Next, we confirm that the relationship between booms and standards was stronger in
relatively larger markets in a specification interacting our boom variable with the log of the
MSA population (Table 8, column 4). While the linear coefficient for the boom variable is
positive and significant, the overall relationship is negative for essentially all markets and
becomes significant for markets above the 25th percentile of the population distribution.
7 The Rapture Index is available at http://www.raptureready.com/rap2.html
16
The relationship between lending standards and credit expansion also appears to depend on
the size of the boom itself. Table 9 shows that the coefficient of log number of applications is
larger and more significant when our baseline specification is estimated on subsamples of
MSAs with the number of application above the median and the growth rate of applications
above the median.
D. Effects of Entry and Changes in Market Structure
We further refine our analysis by assessing the impact on denial rates of credit expansion by
new entrants (rather than incumbent institutions). In Table 10 we report the results of our
analysis of the effects of entry by new players on incumbent lending standards. Consistent
with asymmetric information theories of competition in credit markets implying that an
increase in the number of competing institutions increases adverse selection (Broeker, 1990,
and Riordan, 1993), we find that an increase in the number of entrants (i.e., competing
institutions) increases the denial rates of incumbent institutions in the overall mortgage
market (column 1). In this regression, we use the market share of entrants, computed as the
sum of each entrant’s share in total loan applications, rather than the simple number of
entrants, to control for the size of each entrant and capture overall market power of entrants.
The evolution of denial rates in the subprime mortgage market, however, supports the notion
of incumbents cutting their lending standards in reaction to the entry of new (and large)
competitors (column 3). As the industry expanded and more subprime lenders entered
specific metropolitan areas, denial rates by incumbent lenders went down. We take this as
direct evidence of a reduction in lending standards in this market. We find a similar, though
much less pronounced, effect in the prime market (column 2).
Denial rates of incumbent institutions are unlikely to affect the entry of new lenders to the
extent that they reflect underlying applicant fundamentals. Then, by focusing on the effect of
new entrants on the denial rates of incumbent lenders we are able to assess the independent
effect of market entry (and expansion) on incumbent lending standards. That said, high
denial rates could conceivably attract entry if they reflect collusion among incumbent lenders
rather than the underlying fundamentals in the MSA. However, a close-to-zero correlation
between the incumbent denial rate level (lagged) and our entry variable suggests that this is
unlikely to be the case. The evidence in this section suggests that, as for small business
lending (see Petersen and Rajan, 2002), information technology may have reduced but has
not eliminated the importance of geography in the mortgage market
E. Alternative Proxies for Lending Standards
We now turn to the loan-to-income (LTI) ratio regressions (Table 11). As mentioned earlier,
LTI ratios can be regarded as an alternative proxy for lending standards. We find that higher
average LTI ratios are associated with lower unemployment rates and are more common in
high income areas and where there is a larger percentage of the population that is self
employed. Turning to our variables of interest, the results indicate that LTI ratios grow with
the number of loan applications, particularly in the subprime market, confirming the notion
17
of a boom effect on lending standards. The effect of competition is also confirmed with
higher LTI ratios in MSAs with larger number of competing lenders. The house price
appreciation variable enters only significantly in the subprime market regression, suggesting
that LTI ratios in the prime market are not much affected by house price appreciation. In the
subprime market, LTI ratios are strongly positively associated with house price appreciation.
F. Effects of Loan Sales
The increased ability of financial institutions to securitize mortgages over the past decade
may have contributed to both the expansion of the mortgage market and the documented
decline in lending standards. In Table 12, columns 1 to 3, we explore how the increasing
recourse to securitization of mortgages has affected denial rates in the prime and subprime
mortgage industry. For each originated loan in the HMDA database, the variable “Purchaser
type” denotes whether the loan was securitized kept on the books of the originating
institution or sold through a private sale to another financial institution. We use this
information to compute the share of loans sold within a year of origination and use this as a
proxy for the ability to securitize loans in a given MSA. Given the share of sold loans
changes dramatically over the period,8 we allow this coefficient to be different for the 2000-
2003 and the 2004-2006 periods.
The results indicate that denial rates were lower in MSAs where a greater proportion of
originated loans were sold within one year from origination, consistent with findings by Mian
and Sufi (2007) and Keys et al. (2007). During the first part of the sample period, this effect
was more pronounced for the prime than the subprime market. However, during the second
part of the sample period, when securitization of subprime loans increased dramatically, the
effect turns more pronounced for the subprime mortgage market.
In the remainder of Table 12, we document the relationship between the number of
applications and overall credit market expansion, as measured by changes in the ratio of
credit to income at the MSA level. We scale credit by income to control for changes in the
level of income in the MSA. Notice that the contribution to credit market deepening of higher
denial rates is much stronger in the subprime market compared to the prime market,
indicating that the link between deteriorations in lending standards and credit expansion is
stronger in the subprime market. This is not surprising given that the subprime mortgage
market tends to concentrate on high-risk mortgages.
Securitization also appears to have favored the expansion of overall credit with a positive and
significant effect on credit-to-income ratios, particularly during the second part of the sample
period. This evidence partially supports the view that disintermediation through
securitization provides lenders with incentives to extend riskier loans.
8 See Ashcraft and Schuermann (2007).
18
VI. DISCUSSION AND CONCLUSIONS
This paper provides robust evidence that the recent rapid credit expansion in the subprime
mortgage market was associated with easing credit standards and that the current troubles in
this market are more severe in the areas where the expansion was faster. We link the change
in lending standards to four main factors. First, we find evidence that standards declined
more where the credit boom was larger. This lends support to the assertions that rapid credit
growth episodes tend to breed lax lending behavior. Second, lower standards were associated
with a fast rate of house price appreciation, consistent with the notion that lenders were to
some extent gambling on a continuing housing boom, relying on the fact that borrowers in
default could always liquidate the collateral and repay the loan. Third, change in market
structure mattered: lending standards declined more in regions where (large and aggressive)
previously absent institutions entered the market. Finally, we find that disintermediation
played a role, with standards declining more in regions where larger portions of the lenders’
loan portfolios where sold to third players.
Our results are robust to a number of alternative specifications, including controlling for
economic fundamentals using out-of-sample data, using alternative measures of lending
standards, and introducing variables that capture the effect of new entrants on the denial rates
of incumbent lenders. The latter approach allows us to assess the independent effect of
changes in local market structure on lending standards. The results are also robust to using
instrumental variables to identify the independent effect of the number of applications and
changes in house prices on loan denial rates. This mitigates concerns that our results are
confounded by endogeneity between loan denial rates and the volume of loan applications.
Finally, the effects we identify for the subprime market are either much weaker or absent in
the prime mortgage market, lending additional support that the deterioration in lending
standards was more pronounced in the subprime mortgage market. Our evidence suggests
that while in the prime market lending standards were largely determined by underlying
fundamentals, for subprime loans lending market conditions and strategic interactions played
an important role in lending decisions.
Our results also shed some light on the effects of monetary policy on banks’ lending
standards. The evolution of U.S. interest rates mimics the evolution of denial rates in the
prime market remarkably well, with denial rates increasing in 2005 following monetary
tightening in 2004. Although we do not find such a relationship for the subprime market,
where denial rates remain relatively low, we find evidence suggesting that the negative
impact of rapid credit growth on lending standards was more pronounced when interest rates
were lower, suggesting that lax monetary policy may exacerbate the effects of booms on
lending standards.
ABSTRACT
This paper analyzes options for reforming the U.S. housing finance system in view of the failure of
Fannie Mae and Freddie Mac as government sponsored enterprises (GSEs). The options considered
include GSE reform, a range of possible new governmental mortgage guarantee plans, and greater
reliance on private mortgage markets. The analysis also considers the likely consequences of adopting
alternative roles for government in the U.S. housing and mortgage markets. We start by reviewing
the history of the GSEs and their contributions to the operation of U.S. housing and mortgage markets,
including the actions that led to their failure in conjunction with the recent mortgage market crisis.
The reform options we consider include those proposed in a 2011 U.S. Treasury White Paper, plans
for new government mortgage guarantees from various researchers and organizations, and the evidence
from Western European countries for the efficacy of private mortgages markets.
I. Introduction
The two large Government Sponsored Housing Enterprises (GSEs),1 the Federal
National Mortgage Association (“Fannie Mae”) and the Federal Home Loan Mortgage
Corporation (“Freddie Mac”), evolved over three quarters of a century from a single
small government agency, to a large and powerful duopoly, and ultimately to insolvent
institutions protected from bankruptcy only by the full faith and credit of the U.S.
government. Between 2007 and Q2 2011, the two GSEs had realized losses of $247
billion, and they required draws of $169 billion under the Treasured Preferred Stock
Purchase Agreements to remain in operation. (See Federal Housing Finance Agency
2011). This paper traces the transformation of the GSEs from privately held institutions
with powerful direction and political influence to vassals reporting to an administrative
agency in the Department of Housing and Urban Development (the Federal Housing
Finance Agency, FHFA).
Within the next few years, the agencies will have to be restructured. Proposals for
reform include recapitalizing them in some form as Government Sponsored Enterprises
(GSEs), reconstituting them as agencies of the federal government with more narrowlyspecified
missions, or privatizing the organizations. There are also proposals to replace
the GSEs with a variety of new government mortgage guarantee/insurance programs. The
GSE reform and mortgage guarantee proposals are both nested within the larger question
of what are the likely consequences of alternative roles for government in the U.S.
housing and mortgage markets. This paper is intended to help in the deliberations about
1 Athird, much smaller, Government Sponsored Housing Enterprise is the Federal Home Loan Bank
System (FHLBS). The issues for reforming the FHLBS are similar to many of the issue raised in this paper
for Fannie Mae and Freddie Mac, although we have not analyzed separately the FHLBS or other nonhousing
government enterprises.
3
“what to do” about these costly failures. We briefly review the history of the housing
enterprises and their performance, including the recent housing crisis. We document the
contributions of Freddie and Fannie to the operation of U.S. housing markets, and we
analyze the role of the agencies in the recent housing crisis. We search for evidence on
the importance of Freddie and Fannie in achieving other important housing goals. We
compare U.S. policies with those adopted in other developed countries.
This is not the first time we have provided some analysis of the reform options in
housing finance, either individually (Jaffee, 2010b, 2011; Quigley 2006) or jointly (Jaffee
and Quigley, 2010). However, it is our first attempt to consider all the history and all of
the options.
In section II below we discuss the background and origin of the GSEs and of the
federal role in supplying housing credit. Section III provides a brief summary of
homeownership and government policy. Section IV describes the broader objectives and
goals of the GSE institutions and analyzes the most recent failures of the credit market
and the secondary housing market. Section V links the current housing crisis to the
insolvency of credit institutions. Section VI describes likely the consequences of a series
of plans concerning the restructuring of these institutions and alternative mechanisms for
government support of the U.S. mortgage market. It also provides a brief summary of the
GSEs under their government conservatorship since September 2008.
II. Background
With the public sale of its stock and its conversion into a government sponsored
enterprise in 1968, the Federal National Mortgage Association (FNMA) emerged from
obscurity as an agent in the market for home mortgage credit. The FNMA had been
4
established in 1938, based on provisions in the 1934 National Housing Act, after the
collapse of the housing market during the Great Depression. The 1934 Act had
established the Federal Housing Administration (FHA) to oversee a program of home
mortgage insurance against default. Insurance was funded by the proceeds of a fixedpremium
charged on unpaid loan balances. These revenues were deposited in Treasury
securities and managed as a mutual insurance fund. Significantly, default insurance was
offered on “economically sound” self-amortizing mortgages with terms as long as twenty
years and with loan-to-value ratios up to eighty percent.
Diffusion of the new FHA product across the country required national
standardization of underwriting procedures. Appraisals were required, and borrowers’
credit histories and financial capacities were reported and evaluated systematically. The
Mutual Mortgage Insurance Fund, established to manage the reserve of FHA premiums,
was required to be actuarially sound. This was generally understood to allow very small
redistributions from high income to low income FHA mortgagees. By its original design,
the FHA was clearly intended to serve the vast majority of homeowners.
In the 1934 Act, Congress had also sought to encourage private establishment of
National Mortgage Associations that would buy and sell the new and unfamiliar insured
mortgages of the Federal Housing Administration. By creating a secondary market for
these assets, the Associations sought to increase the willingness of primary lenders to
make these loans. No private associations were formed, however. When further
liberalization of the terms under which associations could be organized was still
unsuccessful, the Federal National Mortgage Association was chartered in 1938 by the
Federal Housing Administrator following the request of the President of the United
5
States. Federal action was precipitated particularly by concern over the acceptability of
new FHA ninety-percent twenty-five-year loans authorized that year.
At first, the Association operated on a small scale, but its willingness to buy FHA
mortgages encouraged lenders to make them. A 1948 authorization to purchase
mortgages guaranteed by the Veterans Administration led the Association to make
purchases, commitments, loans, and investments that soon approached the
congressionally authorized limit of $2.5 billion. Since the maximum interest rate on VA
mortgages was below the market rate, FNMA’s advance commitments to buy VAguaranteed
mortgages at par assured windfall gains to private borrowers or lenders. The
1954 Housing Act reorganized Fannie Mae as a mixed-ownership corporation with
eligible shareholders being the federal government and lenders that sold mortgages to
Fannie Mae. FNMA was then able to finance its operations through sale of its preferred
stock to the U.S. Treasury, through sale of its common stock to lenders whose mortgages
it bought, and by the sale of bonds to the public.
The Housing and Urban Development Act of 1968 transferred FNMA’s special
assistance and the management and liquidation of part of its portfolio to the newly
constituted Government National Mortgage Association. Its secondary market operations
remained with FNMA, now owned entirely by private stockholders. Commercial banks
were the primary beneficiaries of FNMA’s secondary market activities in FHA and VA
mortgages -- since the banks specialized in originating the government-guaranteed
mortgages. In contrast, the mortgages originated by Savings and Loan Associations
(S&Ls) and Mutual Savings Banks (“Thrift Institutions”) were primarily “conventional”
mortgages, meaning they received no government guarantee. The thrift institutions
6
lobbied for equal treatment, and were rewarded in 1970 with the establishment of the
Federal Home Loan Mortgage Corporation (“Freddie Mac”) under the regulatory control
of the Federal Home Loan Bank System, the S&L regulator. Freddie Mac stock first
became publicly available in 1989, although shares owned by Freddie Mac’s financial
partners had been traded on the New York Stock Exchange starting in 1984.
III. Homeownership and Government Policy
According to de Tocqueville (1835), Americans have long been obsessed with
owner-occupied housing. Richard Green (2011) sees this as a political issue, as societies
are less disposed to make revolution when personal and real property is augmented and
distributed among the population. Other recent work emphasizes the external benefits of
owner-occupied housing, and a large social science literature has developed exploring the
connection between higher levels of homeownership and the economic and social
outcomes of households. Appendix Table A1 reports some of the findings linking
homeownership to social outcomes. Two other papers (Dietz and Haurin, 2003; Haurin,
Dietz and Weinberg, 2002) provide an exhaustive comparison of the economic and social
consequences for those living in owner-occupied and rental housing.
Most of the research supports the conclusion that homeownership has some
positive effects upon the social outcomes for individuals and households. But the
research does not conclude that the effect is very large. But even if the effect were large,
nothing supports the conclusion that homeownership should be supported by the
institution of the GSEs or their policy choices. In particular, the primary impact of
instruments that focus on lowering the cost or expanding the availability of mortgages
7
will be larger mortgages, which makes those instruments ineffective and costly relative to
direct subsidies for homeownership.
This is important -- for as noted below many of the popular arguments in support
of subsidies for the GSEs are based upon the promotion of homeownership in the
economy.
IV. Policy Objectives for the GSEs
A. Primary Objectives
The GSE charters are quite explicit in stating the goals and responsibilities of the
enterprises, but they do not state homeownership goals directly. Instead, they seek to:
1) provide stability in the secondary market for residential mortgages;
2) respond appropriately to the private capital market;
3) provide ongoing assistance to the secondary market for residential mortgages
(including activities relating to mortgages on housing for low- and moderate-income
families involving a reasonable economic return that may be less than the return earned
on other activities) by increasing the liquidity of mortgage investments and improving the
distribution of investment capital available for residential mortgage financing;
4) promote access to mortgage credit throughout the Nation (including central cities,
rural areas, and underserved areas) by increasing the liquidity of mortgage investments
and improving the distribution of investment capital available for residential mortgage
financing; and
5) manage and liquidate federally owned mortgage portfolios in an orderly manner,
with a minimum of adverse effect upon the residential mortgage market and minimum
loss to the Federal Government.
8
This section reviews the key activities of the GSEs with respect to providing
stability, assistance, and liquidity to the secondary market for residential mortgages. The
specific objectives of the secondary market activities have varied over time, including
operations to reinforce or offset fiscal and monetary policy, to increase residential
construction, to make a market in federally underwritten mortgages, to reduce regional
yield differentials, and to act as a mortgage lender of last resort. (See Guttentag, 1963, for
an extensive discussion of these key activities.)
A.1 Quantitative Impact of the GSEs on the U.S. Home Mortgage Market
Table 1 reviews the quantitative role of the GSEs in the US mortgage market over
the recent past. The top panel reports the outstanding amounts of whole home mortgages
at the end of each decade from 1950 through 2010. Through 1960, all whole home
mortgages were directly held in portfolios, and even by 1970 the only exception was $3
billion of mortgage-backed securities (MBS) issued by the newly established
Government National Mortgage Association (GNMA). The largest portfolio investor has
always been the set of depository institutions, commercial banks and thrift institutions
(savings and loan associations, savings banks, and credit unions).2 The market investor
portfolios include capital market investors ranging from pension funds and mutual funds
to insurance companies. Starting in 1980, increasing amounts of whole home mortgages
have been held within MBS pools. The top panel of Table 1 separates the three main
categories of MBS pools: pools issued by the GSEs, by GNMA, and by private label
securitizers (PLS).
2 The
GSE category covers the Fannie Mae on-balance-sheet portfolio through 1970 and the sum of the
Fannie Mae and Freddie Mac portfolios thereafter.
9
The middle panel of Table 1 shows each of the investor categories for whole
home mortgage holdings as a percentage of the total amount outstanding. One major
trend is apparent; portfolio holdings declined steadily from 100 percent of the total in
1960 to 37 percent of the total by 2010. Among the portfolio investors, both depository
institution and market investor holdings declined steadily starting in 1970. The GSE
portfolio holdings of whole home mortgages, five percent of the total in 2010, remained a
small percentage of the total throughout the history, with fluctuations within the narrow
band of three percent to eight percent of the total.
The second major trend reported in the middle panel of Table 1 is the steady rise
in mortgage pool holdings as a percentage of the total, starting at one percent in 1970 and
reaching 63 percent of the total by 2010. GSE pools show the most rapid rise, reaching 41
percent of total outstanding home mortgages by 2010. The PLS pools also grew steadily,
reaching twelve percent of the total by 2010. The GNMA pool share of total outstanding
mortgages, ten percent at year-end 2010, fluctuated in a narrow range between ten
percent and fifteen percent of the total from 1980 to the present.
The bottom panel of Table 1 shows the direct GSE share of the home mortgage
market, computed as the sum of whole mortgages held in the GSE portfolios and their
outstanding MBS. While this GSE share rose steadily from 1950, the primary increase
started in 1990, with the share reaching 46 percent of all outstanding home mortgages in
2010. This direct share does not include MBS from other issuers that were held in the
GSE portfolios, a topic to which we turn below.3
3 Quantitatively,
including the GSE holdings of other MBS would raise the total GSE share to 47 percent
and 48 percent for 2000 and 2010 respectively. This ratio actually peaked in 2003, reaching fifty percent.
10
While Table 1 accounts for all outstanding home mortgages, it does not
distinguish among the investor groups holding the MBS instruments created by the
mortgage pools. This issue is addressed in Table 2, in which ownership of the MBS pools
has been allocated among the various investor classes. These values are then combined
with the portfolio holdings of whole mortgages to determine the ownership structure of
all home mortgages, whether held as whole mortgages or as investment in MBS pools.4 It
is apparent from Table 2 that, starting in 1980, market investors were expanding relative
to the depository institutions and the GSEs, and that by 2010 the market investors were
the largest investor class for the sum of whole mortgages and mortgage securities.
Figure 1 reports the percentage of outstanding whole mortgages held directly in
portfolios for each of the three investor classes. The depository institutions have always
been the predominant holder of whole mortgages. At year-end 2010, the depository
institutions held 76 percent of all whole mortgages that were directly held in portfolios,
with the market investors and the GSEs each holding a twelve percent share.
Figure 2 reports the percentage of outstanding MBS for the three holder classes.5
It is apparent that the market investors have always been dominant in holding MBS
positions. At year-end 2010, market investors were holding 67 percent of the outstanding
MBS, with depository institutions holding 21 percent and the GSEs twelve percent.
Figure 3 combines the results for Figures 1 and 2, reporting the share for each
holder class of their combined positions in whole mortgages and MBS. By 2010, the
market investors had the largest position, representing 47 percent of all home mortgages,
4 As far as we are aware, this integration of whole mortgage portfolio holdings and MBS pools by investor
has not been available previously.
5 The graphs start in 1970, since there were no outstanding MBS before that year.
11
with depository institutions in the second position, holding 41 percent of all home
mortgages. At the same time, the GSEs were holding twelve percent of all home
mortgages (as either whole mortgages or MBS) a share just below their average over the
last three decades.
Figure 3 indicates that the GSE combined holdings of whole mortgages and MBS
has always represented a relatively small share of total U.S. home mortgages outstanding.
In this sense, closing the GSEs now, in an orderly way, would have a minor impact on the
U.S. mortgage market. That is, the twelve percent GSE share could be readily replaced by
a combination of market investors and depository institutions (who are already holding
88 percent of U.S. home mortgages). There are, however, two other measures of potential
GSE benefits with regard to outstanding whole mortgages and MBS: (1) the contribution
of MBS issued by the GSEs, and (2) stabilization of the U.S. home mortgage market
through countercyclical activities by the GSEs. We now consider these in turn.
A.2 The Role of GSE-Issued MBS
Figure 4 shows the relative shares of outstanding home mortgage MBS by issuer
class. The GSE share has been dominant since 1990, representing 65 percent of all
outstanding MBS in 2010. The share of private label securitizers (PLS) has been steadily
rising, but still represented only 19 percent of outstanding MBS at year-end 2010. The
GNMA share has been steadily declining, reaching a 16 percent market share by year-end
2010.
The dominant historical position of GSE MBS in the current U.S. home mortgage
is sometimes used to justify a future role for the GSEs in the market. But, at its core, the
GSE dominance of the MBS market for home mortgages has been largely derived from
12
the assumption of market investors—reinforced by GSE marketing--that the GSE MBS
had an implicit government guarantee (and which turned out to be correct, after
imposition of the GSE Conservatorships in 2008). In this sense, the dominant GSE MBS
position is just an example of crowding out, whereby any asset with a low-cost
government guarantee against loss will likely replace private activity in the same market.
If the government guarantee were eliminated, there is every reason to expect that private
market activity would simply replace the activity of the government entity.
A brief review of the history of U.S. MBS development is valuable for
understanding the limited contribution of the GSEs to MBS innovations:6
1968: GNMA creates first modern MBS by securitizing FHA/VA mortgages;
1970s: GSEs expand MBS market based on their implicit government guarantee;7
1980s: Salomon Bros. securitizes multi-class, non-guaranteed, MBS instruments;8
1990s: Multi-class (structured finance) mechanism is first applied to wide range of
asset-backed securities, including auto, credit card, and commercial mortgage loans;
2000s: Subprime lending becomes the most important application of MBS/ABS
methods.
6 US mortgage securitization probably actually began soon after the founding of the Republic. Following
the war of 1812, the US federal government was desperate for revenue and extended loans to homesteaders
for property on the Western frontiers. Without the resources to make and hold these loans, the government
pooled and sold these loans to investors. By the 1920s, securitization was already a well accepted format
for selling loans to investors. These mortgage-backed securities failed during the real estate crisis of the
1930s, and it was decades before U.S. securitization was reactivated in 1968. See Quinn (2010) for a new
history of the U.S. housing policy and the origins of securitization.
7 The GSEs could point to their $2.25 billion line of credit at the US Treasury as backing for their
guarantee, a significant factor only in the early years when their scale of operations was relatively small. It
also helped the GSE case that the US government never firmly and officially rejected the notion of an
implicit guarantee.
8 The colorful development of private-label MBS under Lewis Ranieri at Solomon Brothers is wonderfully
chronicled in Liars Poker by Lewis (1990).
13
Credit for the modern innovation of single-class MBS belongs to the government
itself with the creation of the GNMA MBS. GNMA was, and remains, an agency within
the Department of Housing and Urban Development. Likewise, credit for the innovation
of the multi-class MBS belongs to the private sector with the development of structured
MBS by Salomon Bros. in the 1980s. In fact, the GSEs have always been followers, not
innovators, in the MBS market. The success of the GSEs in establishing the market for
their own MBS depended entirely on the perception of capital market investors that they
faced no credit risk as the result of the implicit federal guarantee. Absent this government
guarantee, the single-class GSE MBS would have simply lost out in the marketplace to
the multi-class, private-label, MBS.
GSE proponents often argue that the GSEs reduced securitization costs and
mortgage interest rates. Here, too, the reality is that the GSEs provide no benefit other
than the implicit guarantee. A case in point is the TBA (“to be announced”) forward
market for GSE and GNMA MBS. While this market arguably expands the liquidity of
the traded MBS, the benefit depends completely on the market’s perception that the
guarantees—explicit for GNMA and implicit for the GSE MBS—make credit risk
irrelevant in the pricing and trading of the securities. It is equally noteworthy that the
markets for asset-backed securitization, for the securitization of credit card, auto, and
commercial mortgage loans, and other loan classes as well, expanded rapidly starting in
the early 1990s without any contribution from the GSEs. Indeed, as with the original
GNMA MBS, the GSEs benefited from the innovation by others, creating their own
14
structured finance offerings once the market demand for such securities had been
expanded through private market innovation.9
Finally, the claim is sometimes made that the GSE MBS activity is critical for the
survival of the thirty-year, fixed-rate, residential mortgage. This claim is unwarranted. In
fact, two features of the GSE MBS instrument were clearly detriments to the expansion
of the long-term, fixed-rate, mortgage:
First, the GSE MBS transferred the entire interest rate risk imbedded in the fixedrate
mortgages to the market investors who purchased the instruments. The GSEs took no
action to mitigate this risk;
Second, the GSE MBS generally disallowed prepayment penalties on all the
mortgages they securitized. While borrowers may have felt they benefitted from this
“free” call option, it greatly magnified the interest rate risk imposed on investors in the
GSE MBS, and led to higher interest rates on the fixed-rate mortgages.
Finally, a number of Western European countries successfully use long-term,
fixed rate, mortgages, but have no entity comparable to the GSEs, Denmark is the most
conspicuous example. The use of covered bonds allows European banks to hold longterm
mortgages on their balance sheets, while passing a substantial part of the interestrate
risk to capital market investors. We further discuss the experience of Western
European countries in Section A5 below.
9 See
Downing, Jaffee, and Wallace (2009) for a discussion of how the GSEs profited by restructuring their
simple passthrough MBS into more complex multi-tranche securitizations.
15
A.3 The Limited GSE Contributions to Mortgage Market Stability
The GSEs also claim credit for taking actions to stabilize the U.S. mortgage
markets. The U.S. Government Accountability Office (2009), however, finds little
evidence of such benefits:
“… the extent to which the enterprises have been able to support a stable
and liquid secondary mortgage market during periods of economic stress,
which are key charter and statutory obligations, is not clear. In 1996, we
attempted to determine the extent to which the enterprises’ activities
would support mortgage finance during stressful economic periods by
analyzing Fannie Mae’s mortgage activities in some states, including oil
producing states such as Texas and Louisiana, beginning in the 1980s.
Specifically, we analyzed state-level data on Fannie Mae’s market shares
and housing price indexes for the years 1980–1994. We did not find
sufficient evidence that Fannie Mae provided an economic cushion to
mortgage markets in those states during the period analyzed.”
Reports by the Congressional Budget Office (1996, 2010) come to similar conclusions.
The academic literature also generally concludes that the GSE contribution to U.S.
mortgage market stability has been modest at best. This view is stated in early studies by
Jaffee and Rosen (1978, 1979) and more recent studies by Frame and White (2005) and
Lehnert, Passmore, and Sherlund (2008). In contrast, Naranjo and Toevs (2002), a study
funded by Fannie Mae, found evidence of effective stabilization by the GSEs, as did
other studies carried out internally by the GSEs. Unlike the previous studies, Peek and
Wilcox (2003) focused on the flow of mortgage funds, and not on mortgage interest rates,
and found the GSE contribution to be countercyclical. Of course, this research was all
conducted before the subprime housing bubble and its collapse. In this event, as we now
document, the GSE participation was decidedly destabilizing.
A.4 The GSE Role in the Subprime Mortgage Boom and Crash
The losses reported by the GSEs starting in 2008 leave no doubt that the GSEs
acquired a significant volume of risky mortgages during the subprime boom. However,
16
the extent, timing, and significance of these acquisitions is debated. For example, Jaffee
(2010) describes the GSE role as “expanding” the subprime boom, especially in 2007,
whereas Wallison (2011, p.2) concludes that GSE activity, based on their housing goals,
was a primary “source” of the crisis. In this section, we evaluate the role played by the
GSEs in the subprime mortgage boom and crash.
A quantitative evaluation of the GSE role in the subprime crisis faces a number of
significant data issues:
1) Definitions for subprime and Alt A mortgages differ across data sets, and certain highrisk
mortgages are not included under either label.
2) Defining high-risk mortgages (including subprime and Alt A instruments) is
necessarily complex because mortgage default risk arises from numerous factors
including borrower and property attributes (FICO scores, loan-to-value ratios, etc.),
special amortization options (interest only, negative amortization, etc.), and fixed-rate
versus adjustable-rate loans.
3) The GSEs could not acquire any mortgages with an initial loan amount above the
conforming loan limit (so-called jumbo mortgages).
Our analysis starts by reviewing a newly compiled mortgage origination dataset
from the GSE regulator, the Federal Housing Finance Agency (2010a).10 These data
compare the risk characteristics of all mortgages acquired by the GSEs (whether
securitized or held in retained portfolios) with the risk characteristics of all conforming,
conventional, mortgages that were included in private label securitizations (PLS),
tabulated by year of mortgage origination. Because the dataset has nearly complete
coverage and is restricted to conforming mortgages, it provides the best available direct --
10 We thank Robin Seiler of the Federal Housing Finance Agency for providing us with a roadmap for the
intricacies of these data.
17
“apples to apples” -- comparison of the GSE acquired mortgages relative to the
comparable market. Nevertheless, there are two limitations. First, while the FHFA data
include all the conforming mortgages that collateralized PLS MBS instruments, the GSE
holdings of PLS tranches are not so identified. We do not expect a significant bias in the
comparisons from this source, however, because the GSE PLS holdings were almost
entirely AAA tranches with little ex ante credit risk.11 Second, the FHFA data exclude
conforming mortgages that were not securitized (i.e., they were retained in lender
portfolios). To the extent that lenders did retain conforming mortgages with high-risk
attributes, the FHFA dataset will undercount the high-risk dimensions of the overall
conforming origination pools, and will therefore overstate the GSE share of all high-risk
originations. Here too, we do not expect a significant bias in our comparisons, because
most subprime and Alt-A mortgages were securitized, and the securitization rate was
even higher among those high-risk loans that were also conforming mortgages.12
Panel A of Table 3 shows the dollar amount of the conforming mortgages by
origination year and various risk attributes. Rows (1) to (3) report on loans with one of
the identified high-risk factors: high loan-to-value (LTV) ratios, low FICO scores, and
adjustable rate mortgages (ARMs) respectively. However, there is some double counting
11 See Thomas and Van Order (2011) for further discussion. PLS tranches as a share of total GSE
acquisitions reached its high point at 22.9 percent in 2005, but had fallen to 7.4 percent by 2007.
Furthermore, actual cash flow losses on GSE PLS positions have been modest to date, although the GSEs
have recognized significant mark to market valuation losses on these positions.
12 For example, 2007 data from Inside Mortgage Finance indicate that only $33 billion (or 7%) of the
subprime/Alt A mortgages originated that year were not securitized. Even if these were all conforming
mortgages, their share of total conforming originations that year would be less than 3 percent. Furthermore,
Inside Mortgage Finance indicates that over 31% of subprime MBS and 9 percent of Alt A MBS in 2007
were “GSE eligible”—i.e. conforming mortgages eligible for GSE purchase--further reducing the incentive
of portfolio lenders to hold these mortgages in unsecuritized form. It is also noteworthy that while there is
no consensus conclusion from the expanding literature on whether securitization created lax underwriting
standards—see for example the contrast between Bubb and Kaufman (2009) and Keys etal. (2010)—there
is no finding that portfolio lenders were systematically retaining high-risk mortgages.
18
since some loans have more than one of these attributes. The aggregate high-risk
originations shown in row (4) net out all double counting.13 Row (6) shows the
percentage of high-risk mortgages as a share of total conforming mortgages (in row 5).
This high-risk share of total conforming originations rose steadily through 2004 and then
declined steadily thereafter.
Panel B of Table 3 computes the share of the conforming mortgages acquired by
the GSEs—whether as backing for guaranteed MBS or to hold on their balance sheets--
for each risk attribute. For example, in 2001, the GSEs acquired about 92.2 percent of all
conforming mortgages with LTV ratios above 90 percent. For all 3 of the risk attributes,
the GSE share fell steadily through 2005 and then expanded rapidly through 2007. By
2007, the GSEs were acquiring 79.9 percent of the high-risk, conforming, mortgage
originations. In interpreting these numbers, however, it must be recognized that, as shown
in row (11), the GSEs represent a large share of the overall conforming mortgage market;
as their overall market share approaches 100 percent, their share of each risk attribute
would necessarily do the same.
Panel C corrects for the large GSE share of the conforming market by computing
a “relative intensity,” dividing the GSE market share for each risk attribute in Panel B by
the overall GSE market share in Row (11). A coefficient of one indicates the GSEs are
holding the “market portfolio,” whereas coefficients below one indicate they are avoiding
risky mortgages and coefficients above one indicate the GSEs are actively acquiring risky
mortgages. The pattern for each of the three risk attributes shows the relative intensity
13 Forexample, for the fixed-rate mortgage originations in 2007, 2.2 percent had LTV > 90 percent and
FICO score < 620. For adjustable rate mortgages in 2007, 19.2 percent had either LTV > 90 percent or
FICO score < 620. Overall, in 2007 4.7 percent of the originated mortgages had more than one of the highrisk
attributes.
19
rising steadily starting in 2005. In each case, the high point of the seven-year history was
reached in 2007. Since the relative intensities over the full time span are generally less
than one, it would appear the GSEs were not leading the market for high-risk lending as
the subprime boom took off.14 But the jumps in the relative intensities in 2007 for most of
the indicators suggest that the GSEs then rapidly expanded their participation in the
subprime boom. This is one key basis for our conclusion that the GSEs were a
destabilizing influence on the conforming mortgage market as the subprime boom headed
to its peak in 2007.
The analysis has so far focused on the GSE acquisition of high-risk mortgages as
a share of the overall conforming mortgage market. We now consider the GSE
acquisition of high risk mortgages as a share of their total acquisitions. Table 4 reports
the first three attributes high LTV ratios; low FICO scores; and ARMs; as reported in
Table 3. The time pattern is again distinctive, with the share of the GSEs new business
dedicated to mortgages with these high-risk attributes generally rising starting in 2004,
the only exception being the declining share of ARM acquisitions by Fannie Mae. The
companies also reported their acquisitions of interest-only, condo/coop, and investor
mortgages; and here too the pattern is generally rising from 2004. 2007 represents the
year of maximum share for each high-risk mortgage attribute with the exception of
Fannie Mae ARMs and Freddie Mac interest-only mortgages. These data thus present a
second independent basis for our conclusion that the GSEs were a decidedly destabilizing
influence on the conforming mortgage market as the subprime boom headed to its peak in
2007.
14 Thomas and Van Order (2011), although using different datasets, come to the same conclusion.
20
A.5 Mortgage Markets Without GSEs
The analysis above leaves little doubt that the GSEs destabilized the U.S.
mortgage market during the later stages of the subprime boom, but there is a further
question how the U.S. mortgage markets would function without the GSEs. To help
answer this, in this section we consider evidence from two sources: (1) how the U.S.
mortgage markets have performed without GSEs, and (2) the performance of the
mortgage markets in Western European countries.
The evidence that private mortgage markets have operated effectively in the U.S.
economy can be summarized with three comments on the historical role of private
markets within the U.S. mortgage market. First, private markets have always originated
100% of U.S. mortgages, and closing the GSEs would not affect this. Second, the GSEs
have never held a significant share of the outstanding U.S. home mortgages, this share
being, for example, 12 percent at year-end 2010. Third, the GSE MBS share of total
home mortgages first exceeded 30% only in 2007. This confirms that the private
markets—depository institutions and capital market investors--are capable of holding or
securitizing the large majority of U.S. mortgages. It is also noteworthy that the market for
jumbo mortgages—mortgages that exceed the conforming loan limit--has generally
functioned quite satisfactorily.
Turning to the European evidence, the European economies and housing markets
are sufficiently similar to the U.S. to provide a potentially interesting comparison, while
they have the key distinction that government intervention in these housing and mortgage
markets is far less than for the U.S.; in particular, none of these countries has entities with
21
any significant resemblance to the U.S. GSEs.15 This conclusion is stated very clearly by
Coles and Hardt (2000, p. 778):16
“There is no national or European government agency to help lenders fund
their loans. Mortgage loans have to be funded on the basis of the financial
strength of banks or the intrinsic quality of the securities. EU Law (Article
87 and 88 of the EC treaty) outlaws state aid in the form of guarantees as
there may be an element of competitive distortion.”
Table 5 compares the U.S. and Western European mortgage markets for a range
of quantitative attributes from 1998 to 2010 based on a comprehensive data base of
housing and mortgage data for fifteen European countries from the European Mortgage
Federation (2010). Column 1 compares the most recent owner occupancy rates for the
U.S. and European countries. The U.S. value is 66.9 percent, which is just below its peak
subprime boom value. It is frequently suggested that the high rate of homeownership is
the result of the large U.S. government support of the mortgage market, including the
GSEs. It is thus highly revealing that the U.S. rate is just at the median— eight of the
European countries have higher owner occupancy rates—and slightly below the average
value for the European countries. Furthermore, the lower owner occupancy rates in some
of the countries, Germany for example, appear to be the result of cultural preferences
rather than government inaction. A full analysis of the determinants of owner occupancy
rates across countries should also control for the age distribution of the population, since
younger households, and possibly the oldest households, may have lower ownership rates
in all countries. Chirui and Jappelli (2003) provide a start in this direction, showing that
lower downpayment rates are a significant factor encouraging owner occupancy after
15 SeeEuropean Central Bank (2009) for an extensive review of housing finance in the European Union
countries.
16 Hardt was the Secretary General of the European Mortgage Federation at the time.
22
controlling for the population age structure in a sample of fourteen OCED countries. The
U.S. has also generally benefitted from very low downpayment rates, but it still has an
average ownership rate, reinforcing the conclusion that the government interventions
have been largely ineffective in raising the U.S. home ownership rate relative to its peers.
Column 2 measures the volatility of housing construction activity from 1998 to
2010 based on the coefficient of variation of housing starts as a measure of relative
volatility. The U.S. relative volatility is third highest out of the 16 countries, implying
that the government interventions have failed to reduce U.S. housing cycles relative to
those in Western Europe. Column 3 measures the volatility of house price changes based
on the standard deviation of the annual house price appreciation from 1998 through 2010.
Here the U.S. stands fifth, meaning the country has faced a relatively high rate of house
price volatility. This negative result is all the more significant because the U.S. is far
larger than any of the individual European countries, and thus the benefits of regional
diversification should have lowered the observed U.S. volatility.
Column 4 compares the level of mortgage interest rates in Western Europe and
the U.S., using “representative variable mortgage rates” for Europe and the Freddie Mac
one-year ARM commitment rate for the U.S. The column shows that the U.S. has the
sixth highest average mortgage interest rate from 1998 to 2010, and exceeds the Western
European average by 27 basis points. Since overall interest rates also vary across
countries, as a further test, column 5 shows the average spread between the mortgage rate
and the Treasury bill rate for each country. The U.S. ranks third highest based on the
spread and exceeds the Western European average by 70 basis points. Of course,
numerous factors determine these mortgage rates and spreads, including the precise terms
23
of the variable rate mortgages, other contract features such as downpayment
requirements, and the generally greater credit risk of U.S. mortgages. Nevertheless, the
fact remains that despite the government subsidies and other interventions in the U.S.
residential mortgage markets, U.S. mortgage rates have remained among the highest
levels compared with the countries of Western Europe. Finally, Column 6 shows the
20109 ratio of home mortgages outstanding to each country’s annual GDP, a standard
measure of the depth of a country’s mortgage market. The U.S. ratio is 75.5 percent
which puts it sixth within this group of sixteen developed economies. A relatively high
U.S. result would be expected, given the large mortgage subsidies provided through the
GSEs and other channels. It is noteworthy, therefore, that five Western European
countries achieved even higher ratios without substantial government interventions in
their mortgage markets
The overall conclusion has to be that Western European mortgage and housing
markets have outperformed the U.S. markets over the full range of available measures.
Although data are not provided here, a similar conclusion would hold for the Australian
and Canadian mortgage markets; see Lea (2010). There are, of course, a wide range of
possible explanations for the superior performance of the European mortgage markets.
The key point for present purposes is simply that the superior performance of the
European mortgage markets is not explained by greater government intervention. In the
absence of GSEs, almost all Western European mortgage lending is carried out privately
by banks, primarily funded by bank deposits or covered bonds. Other indirect forms of
government support, such as the tax deductibility of mortgage interest and property taxes
are also notably absent in most European countries.
24
B. Other Justifications for GSE Subsidies
The activities of the GSEs are justified by the particular benefits accruing to
specific classes of borrowers, or more specifically, to all home purchasers and
homeowners from the activities supported by these institutions. As noted above, benefits
have been claimed for the stabilization of the mortgage supply and corresponding
reductions in the volatility of housing construction and home sales. But there are at least
three other classes of potential benefits arising from the GSE:
Increases in the extent of mortgage credit accruing to income and demographic
groups that policy-makers appear to have deemed particularly deserving -- credit which
augments that supplied by the private marketplace;
Increases in the lending support provided to builders, owners, or residents of
specific types of housing, e.g., multifamily rental housing, which would otherwise not be
provided in the market;
Subsidies accruing more broadly to housing market participants, for example, to
all home purchasers in the form of lower interest costs arising from the increased
liquidity afforded by the GSEs and the implicit guarantee of repayment provided by those
institutions;
This section reviews the evidence on the extent and distribution of these benefits.
1. Increased Credit to Targeted Groups and Geographical Areas
The original charter establishing Fannie Mae as a GSE in 1968 recognized a
“national goal of providing adequate housing for low and moderate income households,”
and it authorized the Secretary of the Department of Housing and Urban Development
(HUD) to require that a reasonable portion of Fannie Mae’s purchases of home
25
mortgages be related to this goal. Although regulations requiring the GSEs to allocate a
fixed percentage of mortgage purchases to lower-income households were advanced in
the 1970s, mandatory rules were not proposed in Congress until after the passage of the
Financial Institutions Reform, Recovery, and Enforcement Act (FIRREA) of 1989.
Ultimately, the Federal Housing Enterprises Financial Safety and Soundness Act of 1992
modified and made more explicit the “housing goals” to be promoted by the GSEs. The
Act directed the HUD Secretary to establish quantitative goals for mortgages to “lowand
moderate-income” households and for mortgages originated in “underserved areas.”
It also imposed a “special affordable housing goal” for mortgages for low-income
housing in low-income areas. The 1992 legislation stipulated two-year transition goals,
but after that period, the HUD Secretary was empowered to promulgate more detailed
regulations.
Under the HUD regulations, finalized in December 1995, the first goal (“low- and
moderate-income housing”) directs that a specified fraction of new loans purchased each
year by the GSEs be originated by households with incomes below the area median. The
second goal (“underserved areas”) requires that a specified fraction of mortgages be
originated in census tracts with median incomes less than 90 percent of the area median,
or else in census tracts with a minority population of at least 30 percent and with a tract
median income of less than 120 percent of area median income. The third goal (“special
affordable housing”) targets mortgages originated in tracts with family incomes less than
60 percent of the area median; or else mortgages in tracts with incomes less than 80
percent of area median and also located in specific low-income areas. Any single
mortgage can “count” towards more than one of these goals. (For example, any loan that
26
meets the “special affordable housing” goal also counts towards the “low- and moderateincome”
goal.)
The numerical goals originally set by HUD for 1996 were modest – requiring, for
example, that 40 percent of the GSEs’ mortgage purchases be loans made to households
with incomes below the area median. Over time, the goals for new business set by HUD
have been increased.17 The goal for mortgages to low- and moderate-income households
has been increased from 40 percent in 1996 to 56 percent by 2008. Until 2007, mortgage
originations by both Fannie Mae and Freddie Mac had reached their primary goals every
year. The HUD goal for “underserved areas” was increased from 21 percent in 1996 to 39
percent in 2008. Originations by the larger GSE, Fannie Mae, exceeded this goal in every
year; originations by Freddie Mac exceeded the goal in each year until 2008. The “special
affordable” housing goal was increased by HUD from 12 percent in 1996 to 27 percent in
2008. Both GSEs surpassed this goal in loan originations each year until 2008.
Figures 5, 6, and 7 report the HUD goals and GSE progress in achieving those
goals from their publication in 1995 to the federal takeover of the GSEs in 2008.
Figures 8, 9, and 10 provide another perspective on the magnitude of the goals set
by HUD for the GSEs. They report each of the three goals as well as an estimate of the
share of all newly-issued mortgages in each of the categories. For example, in 2000 the
HUD-specified “low- and moderate-income goal” was to reach 42 percent of new
purchases for the GSEs. However, in 2000 low- and moderate-income mortgages,
according to the same definition, constituted about 59 percent of all new mortgages. At
17 Note,
however, that at the time that the 1992 act was debated in Congress, only 36 percent of Fannie
Mae’s single-family deliveries were for housing whose value was below the area median. (See FHFA
Mortgage Market Note, The Housing Goals of Fannie Mae and Freddie Mac, February 1, 2010.)
27
that time, the “underserved areas” goal was 21 percent of GSE mortgages, while these
mortgages constituted more than a 30 percent market share of new mortgages. In virtually
all cases, the goals imposed were a good bit lower than the share of mortgage loans of
that type originated in the economy. There is no evidence that the goals were set so that
the GSEs would “lead the market” in servicing these groups of households.
2. Increased Credit to Targeted Housing Types: Multifamily
Numerical goals for purchases of multifamily mortgages are not mentioned in the
Financial Safety and Soundness Act of 1992, but there was considerable concern at the
time that the GSEs were not financing their “fair share” of multifamily housing,
especially small multifamily properties. For example, in 1991, small multifamily units
accounted for less than five percent of Freddie Mac’s multifamily unit purchases. At that
time, small multifamily units constituted 39 percent of all recently-financed multifamily
units. (See Herbert, 2001.) Thus, the first rules for implementing the 1992 Act put
forward by HUD also included explicit goals for multifamily housing.
These goals have been in the form of dollar-based targets. Goals in 1996-2000
were approximately 0.8 percent of the mortgage purchases of Fannie Mae and Freddie
Mac recorded in 1994; goals in 2001-2004 (2005-2007) were 1.0 percent of each GSE’s
estimated mortgage purchases in 1997-1999 (2000-2002). Beyond the achievement of
these numerical goals, multifamily mortgage purchases also qualified for “bonus points”
towards the achievement of the three goals specified in the 1992 law. It has been argued
that these “bonus points” (discontinued in 2004) were a major inducement leading to an
increase in participation by the GSEs in the multifamily housing market, particularly in
their financing of small multifamily properties. (See Manchester, 2007.)
28
Figure 11 reports the dollar goals for multifamily dwellings specified by HUD
regulations and the performance of each of the GSEs. As noted in the figure, until quite
recently purchases of multifamily dwellings exceeded the HUD goal by a substantial
amount.
V. Broad Benefits to Homeowners and Purchasers
a. The Effectiveness of the GSE Goals in Directing Mortgage Credit
Of course, the finding that the GSEs have achieved the annual goals specified in
regulations need not imply that Freddie and Fannie have been very effective in increasing
mortgage credit to targeted groups. For example, many suggest that the numerical goals
set for the GSEs have been far too low (e.g., Weicher, 2010), and that, as a result the
GSEs have simply followed the market with a lag of a few years. Indeed, the data in
Figures 5, 6 and 7, provide no evidence that Freddie Mac or Fannie Mae purchased more
than their “fair share” of mortgages in any of these areas of congressional concern. GSE
purchases of mortgages that satisfied any of these congressional goals – as a fraction of
all new purchases – were consistently smaller than their “market share” in all newlyissued
mortgages.
Similarly, Figure 11 indicates that the GSEs’ new purchases of “special
multifamily” mortgages greatly exceeded the dollar goals mandated by HUD in every
year.
Finally, Figure 12 demonstrates that the GSEs’ multifamily housing business was
only a small fraction of the mortgage purchases of the GSEs in any year. It never
amounted to even seven percent of either GSEs’ purchases.
29
Figure 13 reports the aggregate amount of commercial mortgage backed security
(CMBS) and multifamily originations between 2003 and 2009 as reported by the
Mortgage Bankers of America. Mortgage originations by Freddie Mac and Fannie Mae
were small – less than $9 billion in any year. Until 2008, GSE originations were less than
twenty percent of all such mortgage banker mortgage originations. Note, however that in
2008-2009, CMBS and commercial banks left the market entirely; originations by life
insurers declined as well. Since the conservatorship in 2008, virtually all multifamily
mortgages have been originated by the GSEs.
These simple comparisons suggest that any causal effect of the GSEs on lending
to specific income classes, neighborhoods, and property types is not likely to be large –at
least before 2008. Economic analysis of the potential impacts of the GSEs is also
complicated by other public programs in effect. For example, in 1977, the Community
Reinvestment Act (CRA) was passed to encourage banks to exert further efforts to meet
the credit needs of their local communities, including lower-income areas. In identifying
neighborhoods of special concern in administering the CRA, neighborhoods (census
tracts) with median incomes below 80 percent of the area median income are targeted. As
noted above, “underserved areas” of concern in GSE regulation are census tracts with
median incomes below 90 percent of the area median income. In addition, many
borrowers targeted under GSE criteria are also eligible for Federal Housing
Administration (FHA) loans or Veterans’ Administration (subsidized) loans.
The existence of parallel government programs under the CRA, FHA, and VA
raises the possibility that the GSE purchases of qualifying mortgages simply displaced
lenders who would have made the same mortgage under one of the other programs. To
30
the extent that this has been the case, the GSE purchases would have had no noticeable
impact on the mortgage market for the qualifying borrowers. Of course, it is a subtle
empirical problem to determine whether the GSE purchases were simply displacing loans
from the other programs. Nevertheless, a number of academic papers have sought to
identify and quantify the effects of the GSE goals on local and neighborhood housing
markets and on classes of borrowers.
Table 6 summarizes much of this research.
An early paper by Canner, Passmore and Surette (1996) examined loans eligible
for insurance under the FHA. The authors evaluated how the risk associated with these
loans is distributed among government mortgage institutions, private mortgage insurers,
the GSEs, and banks’ in-house portfolios. The results indicated that FHA bears the
largest risk share associated with lending to lower-income and minority populations, with
the GSEs lagging far behind. Bostic and Gabriel (2006) analyzed the effects of the GSE
mortgage purchase goals upon homeownership and housing conditions in California. A
careful comparison of neighborhoods just above the GSE cutoff for “low-moderateincome”
and “special affordable” designation with nearby neighborhoods just below the
cutoff found essentially no differences in the levels and differences in home-ownership
rates and housing conditions during the decade of the 1990s.
In a more sophisticated analysis using a similar comparison of neighborhoods
“just above” and “just below” the GSE cutoff, An, et al, (2007) focused on three
indicators of local housing markets: the home ownership rate, the vacancy rate, and the
median home value. The authors related (an instrument for) the intensity of GSE activity
in a census tract to these outcomes, using a variety of control variables. The results
31
indicated that increases in GSE purchase intensity were associated with significant but
very small declines in neighborhood vacancy rates and increases in median house values.
The authors conclude that the “results do not indicate much efficacy of the GSE
affordable housing loan-purchase targets in improving housing market conditions (2007,
p. 235).”
Two papers by Bhutta (2009b, 2010) adopted a regression discontinuity design to
test the effects of the “underserved areas” goal upon the supply of credit to those areas.
Rather than attempt to match similar neighborhoods for statistical analysis, Bhutta
exploited the facts that census tracts qualified for CRA scrutiny if their median incomes
were 80 percent of the local area, and they qualified for scrutiny under the HUD GSE
goals if their median incomes were 90 percent of the area median design. Bhutta merged
tract-level data on mortgages (from the Home Mortgage Disclosure Act) with
neighborhood (census) data. Bhutta’s results (2009a) do find a significant effect of the
“underserved area” goal on GSE purchasing activity – but the effect is very small (2-3
percent during the 1997-2002 period).
A more recent paper by Moulton (2010), also using a regression discontinuity
approach, finds no effect of the GSEs -- on individual loans rather than aggregate credit
allocations. Moulton uses micro data on mortgage loan applications to examine whether
the GSE’s affordable housing goals altered the probability that a loan application was
originated by a mortgage lending institution or that a loan was purchased by one of the
GSEs. The analysis led to the conclusion that the GSE affordable housing goal had no
effect at all on mortgage lending or on GSE purchases.
32
The consistent finding of little or no effect of the GSE goals on housing
outcomes, mortgage applications, or mortgage finance could suggest that there is little
effect of the GSE rules upon FHA lending as well. But several papers have reported that
an increased market share of GSE mortgages in a census tract is associated with a decline
in the FHA share of mortgages (An and Bostic, 2008; Gabriel and Rosenthal, 2010).
These results may explain why the increases in lending mandated by the HUD
regulations to achieve the congressional goals of the 1992 Act have had very little net
impact on housing and neighborhood outcomes. Small increases in GSE activity have
been offset by roughly comparable declines in FHA activity.
The extent to which an expansion of GSE activity simply crowds out private
mortgage purchases remains an open research question. For example, Gabriel and
Rosenthal (2010) argue that increased GSE activity in the mortgage market involved little
or no crowd-out until about 2005. After that, GSE activity crowded out private activity
until the crash in mortgage markets in 2007.
But even if there were a complete crowd-out of private mortgage activity arising
from GSE behavior , it is hard to attribute any of this to the goals set by the 1992 Act –
especially since the goals were substantially less than the share of these new mortgages in
the market.
To summarize: the academic and scientific literature has generally found little
effect from housing goals as they operated through the GSEs. The goals were low.
Despite appearances, they provided no incentive for the GSEs to “lead the market” in
providing credit to potentially riskier housing investments. They accomplished nothing in
increasing credit for riskier loans.
33
But there is a view that the housing goals were actively harmful in facilitating the
subprime housing crisis.
This position has been put most forcefully by Peter Wallison (2011) in his rebuttal
statement to the Financial Crisis Inquiry Commission. He argues that the requirement to
meet the housing goals “forced” the GSEs to make substandard loans, which is why they
ultimately acquired such large positions in subprime mortgages and subprime mortgage
securities. Indeed, Wallison claims that the HUD goals actually “caused” the subprime
crisis. There is no question that the GSEs ultimately acquired large portfolios of subprime
mortgages and securities -- see our discussion in Section IV.A.4 above -- but Wallison
provides no evidence at all that these subprime portfolios had anything to do with the
GSE goals.
However, an impressive journalistic account of recent history in the mortgage
market argues forcefully that the housing goals in the 1992 act led directly to the
subprime mortgage debacle of 2008 (Morgenson and Rossner, 2011). Our analysis of the
academic literature supports no such claim. It is certainly possible that the passionate
rhetoric from the GSEs provided a convenient “cover” for the trend towards lower
quality, even toxic, mortgages by 2004-2005. However, there is no evidence that this
rhetoric increased GSE lending to targeted groups during the 1990s. Ironically (or
perhaps diabolically), the rhetoric about “affordable housing” from the GSEs had little
effect upon their own mortgage purchases until the subprime crisis was well underway.
As noted above, the empirical evidence simply fails to support a claim that the
GSE housing goals were a primary source of the subprime crisis. First, there are simple
questions of timing. The GSE goals were enunciated in a law passed in 1992; it is
34
implausible that their effect was not felt until a quarter century had elapsed. Further, as
noted below, the GSE accumulation of subprime mortgages accelerated only in 2007, too
late to have “caused’ the subprime bubble (but certainly early enough to have accelerated
it).
Second, as noted above, it appears that the GSE mortgage purchases in support of
the housing goals were principally loans that would otherwise have been made by other
lenders.
Most importantly, the subprime crisis has a long list of proximate causes,
including U.S. monetary policy, a global savings glut, the error of assuming a national
housing pricing collapse was highly unlikely, etc. (see Jaffee, 2009 for further
discussion.)
b. Benefits to all housing market participants
There has been active research seeking to establish the value of the enhanced
liquidity and subsidy to homeowners. In principle, the subsidy provided by the implicit
guarantee can be calculated. Freddie Mac and Fannie Mae issue debt in the same market
as other participants in the banking and finance industry participate. The yield difference
(“spread”) between the debt of the GSEs and that of other firms can be applied to the
newly issued GSE debt to compute the funding advantage in any year arising from the
GSE status. Of course, it is not quite straightforward to apply this principle and to
produce credible estimates. The relevant benchmark estimate (i.e., the appropriate sector
and bond rating) is not without controversy, and a comparison with broad aggregate
indices combines bonds containing a variety of embedded options. Pearce and Miller
(2001), among others, reported comparisons of GSE and AA-rated financial firms,
35
suggesting that the agencies enjoyed a 37 basis point (bps) spread. More sophisticated
comparisons by Nothaft, et al, (2002) suggest that the relative spreads are about 27 bps
(vis-à-vis AA-minus firms). Table 7 summarizes available comparisons. A careful
analysis of yields at issue for GSE debt and the option-free debt issued by a selection of
finance industry corporations (Ambrose and Warga, 2002) concludes that the GSEs enjoy
a spread of 25-29 bps over AA bank bonds and 37-46 over AA financials. Quigley (2006)
provides a terse summary of available estimates.18
The substantial subsidies arising from the funding advantage of the GSEs means
that mortgage rates for all homeowners can be lower than they otherwise would be, that
is, the subsidy can improve the well-being of homeowners and home purchasers.
But of course, in the first instance the subsidy is provided directly to private
profit-making firms with fiduciary duties to their shareholders. It is thus not obvious that
all, or even most, of the funding advantage provided by the public subsidy is passed
through to homeowners. As documented by Hermalin and Jaffee (1996), the secondary
market for mortgage securities (at least for those securities composed of loans
comparable to the rules under which Fannie and Freddie operate) is hardly a textbook
model of atomistic competition. The two GSEs are large, and each has a large market
share of the conforming segment of the market. There are high barriers to entry, and the
MBS product is more-or-less homogeneous. Moreover, mortgage originators have an
inherent first-mover advantage in deciding which newly-issued mortgages to sell to
Fannie and Freddie. This may force the GSEs to pay a premium for the mortgages they
18 These
estimates are in the range of the spreads which have been assumed (41 bps) by the Congressional
Budget Office (CBO, 2001) in estimating the annual federal subsidy to the GSEs. They are similar to the
estimates of spreads (40 bps) used by Passmore, (2005) in a more recent exercise.
36
purchase in the market. These factors, duopoly and adverse selection, may mean that
much of the subsidy accrues to the shareholders of the GSEs or to the owners of other
financial institutions, not to homeowners or home purchasers.
The effects of the GSEs upon mortgage rates can be calculated by estimating the
spread between the interest rates on mortgages which conform to the loan limits and
underwriting guidelines of the GSEs and the rates on otherwise comparable mortgages.
As in the analysis of funding advantages, it is not quite straightforward to apply this
principle and to produce credible estimates. (For example, most research compares the
rates paid by borrowers with loans one dollar below the conforming limit with rates paid
by borrowers with loans one dollar above the limit. But the latter group of borrowers
differs from the former group, or else they surely would have made an additional cash
payment and taken a conforming loan.)19
Early analyses, e.g. by Hendershott and Shilling (1989) comparing interest rates
on jumbo and conforming mortgages, indicated that this spread was 24-39 bps. More
recent studies, e.g., by Passmore, et al (2002), by McKenzie (2002), and by the CBO
(2001), conclude that the spread is 18-23 bps. These more recent studies differ mostly in
their application of more complex screens to insure comparable data for conforming and
nonconforming loans. Table 8 summarizes these comparisons. More recent work by
Passmore, et al (2005) suggests that this spread may be as low as 16 bps.
19 Ofcourse, other reasons besides the greater liquidity provided by the GSEs could explain some of an
observed spread between jumbo and conforming mortgages. Jumbo mortgages are generally prepaid more
aggressively -- borrowers have more at stake, if nothing else. This means that investors will require higher
rates on jumbos merely to compensate for the increased prepayment risk. On the other hand, borrowers
with jumbo mortgages have better credit, and they make larger down payments, which should create lower
rates on jumbo mortgages. See, also, Ambrose, et al (2001), Heuson, et al (2001), or Woodward (2004b).
37
In summary, it appears that the GSEs’ funding advantage is about 30-40 bps, and
the effect of this is to reduce mortgage rates by 16-25 bps. Stated another way, on the
order of half of the subsidy rate to the GSEs is transmitted to homeowners in the form of
reduced mortgage interest rates. Presumably, the remainder is transmitted to the
managers of the GSEs, the shareholders of the enterprises or to the owners of other
financial institutions.20
VI. Where Do We Go From Here?
As noted in the introduction, most commentators agree that the current structure
of the housing finance system must be reformed in the very near term. A question of firstorder
importance is then the likely consequences of the role of government in support of
the U.S. housing and mortgage markets, whether as a modification or replacement of the
GSEs.
The research results reported in this paper make it clear, we think, that the public
benefits arising from the GSEs have been quite small. The establishment of Fannie Mae,
a half century ago, and the establishment of Freddie Mac, forty years ago, did stimulate a
more stable national market for housing finance and did substantially improve the
liquidity and access of the market. As reported above, however, the specific benefits
arising from the GSE structure have been minor. In any event, these benefits -- with some
contributions from the GSEs -- were achieved by the 1980s. There now exists a national
market for home mortgages. The GSEs have followed reform in the secondary market
and have benefited from private innovation.
20 Of
course, the net effects of the GSEs upon public welfare and the economy has greatly exceed the three
effects upon housing market participants discussed here. Indeed, the evidence suggests that the macro
economic effects of the structure and operation of the GSEs during the past half decade has been much
more important for the economy than the direct housing-market effects of the institutions.
38
There have been surprisingly few benefits to deserving households or
neighborhoods which can be attributed to the GSEs. There has been more political or
partisan attention to the cause of homeownership among lower-income households as a
result of powerful advocacy by the interests of GSEs, but there is little evidence that
lower-income homeownership was stimulated at all, at least not until the run up to the
housing bubble.
It is true that the GSE structure has reduced interest rates on home mortgages, by
about a quarter percent or so. But this benefit to homeowners has arisen from the federal
guarantee for GSE debt. And the public cost of the subsidy has been far more than the
benefits of lower interest rates to homeowners. About half of the overall subsidy has
accrued to GSE employees, shareholders, and other market intermediaries. These large
losses are directly attributable to the GSE structure which was created in 1968.
As noted below, we also conclude that the structure of the GSEs themselves has
made regulation of the housing market far less transparent and has extended some of the
consequences of the housing bubble of the past half decade.
A. The Appropriate Role for Government in the U.S. Residential Mortgage
Market
If the GSEs in current form are to be closed, the fundamental policy question is to
decide which government interventions, if any, should replace GSE functions and which
should be performed by the private sector? Once that is decided, there is also the delicate
issue of how to manage the transition from the current GSE conservatorship. Fortunately,
there are two quite flexible instruments available to close down the GSEs in a smooth,
safe and dependable manner: (i) steadily reduce the conforming loan limit until it reaches
39
zero; and (ii) steadily raise the fee charged by the GSEs for guaranteeing MBS. Although
we will return to questions of the dynamic transition below, the key question is to
determine the appropriate role of government in the U.S. mortgage market.
A large number of proposals have been offered for the reform of the U.S.
mortgage market, ranging from a mortgage market managed primarily by private sector
entities to recreation of the GSEs as public/private hybrids (albeit with new controls).
Summaries and analyses of the general approaches are available in U.S. General
Accountability Office (2009), Congressional Budget Office (2010), and Bernanke (2008).
The following is an annotated list of the three primary proposals scrutinized:
Reestablish GSEs with tighter controls and explicit guarantees. The entities would
continue their organization as public/private hybrids, but with tight government
controls, sometimes described as a “public utility” model. In most plans, the
government guarantees would apply to the underlying mortgages, not the newly
created entities. A cooperative structure such as that of the current Federal Home
Loan Banks is an alternative version. The number of entities to be chartered varies by
proposal.
Restructure GSE functions explicitly within a government agency. A simple version
would create a government agency that would explicitly insure mortgages up to some
conforming limit and then securitize pools of these mortgages, very much as the
current FHA and GNMA agencies operate. The support for underserved borrowers
and areas, including multi-family housing, currently covered under the GSE housing
goals, would then continue in a revised form as explicit government programs.
40
Privatization of the U.S. mortgage market. This proposal would create a fully
privatized mortgage market, with no special federal backing for the secondary
mortgage market, although this could include spinning out the GSEs as new private
entities.
More recently, in February 2011, the U.S. Treasury and Housing and Urban
Development agency, U.S. Treasury/HUD 2011), issued a white paper that offered an
alternative list of three policy options. The policy options were based on three principles
(White paper, p. 11):
1. Pave the way for a robust private mortgage market by reducing government support for
housing finance and closing down Fannie Mae and Freddie Mac on a responsible
timeline;
2. Address fundamental flaws in the mortgage market to protect borrowers, to help ensure
transparency for investors, and to increase the role of private capital;
3. Target the government's vital support for affordable housing in a “more effective and
transparent manner.”
In effect, these principles rule out the reestablishment of the GSEs as
private/public hybrids.
The White paper then offers three options for long-term mortgage market reform:
Option 1: A privatized system of housing finance with the government insurance role
limited to FHA, USDA and Department of Veterans’ Affairs’ assistance for narrowly
targeted groups of borrowers.
41
Option 2: A privatized system of housing finance with assistance from FHA, USDA and
the VA for narrowly targeted groups of borrowers and a guarantee mechanism to scale up
during times of crisis.
Option 3: A privatized system of housing finance with FHA, USDA and the VA
assistance for low- and moderate-income borrowers and catastrophic reinsurance behind
significant private capital.
Since the publication of the White paper, most discussions of specific proposals
among academics, public interest groups, and market participants have centered on
versions of the “Option 3.” The alternative views expressed in these discussions mainly
concern the extent and form in which the government’s mortgage guarantees would be
provided. Of course, if the government guarantee is sufficiently limited, “option 3” is no
different from “option 2.” While these discussions have focused on the form of the
government mortgage guarantee, most commentators agree that the abusive mortgage
market practices that evolved during the subprime boom must be ended through
regulation; see U.S. Treasury/HUD (2011, pp.15-18). In fact, Federal Reserve (2008)
actions to modify the Truth in Lending Act and a wide range of requirements in the
Dodd-Frank Act have already gone a long way to eliminating any possible replay of such
abusive practices in the U.S. mortgage market. Most commentators also appear to agree
that the GSE housing goals should be replaced with an explicit and transparent system of
targeted support for access and affordability. An obvious solution, and one endorsed by
the White Paper, is to strengthen and expand the FHA for this purpose. The White Paper
also proposes a public commitment to affordable rental housing.
42
B. Government Insurance of U.S. Mortgages
We now review the major issues and differences among the plans that are
proposed as the mechanism to replace the GSEs with a program of federal government
mortgage insurance. Specific versions are available from Acharya, Richardson Van
Nieuwerburgh, and White (2011), the Center for American Progress (2010), Ellen, Tye,
and Willis (2010), and Hancock and Passmore (2010). While the plans differ in details
and specificity, a composite can be summarized:
1) The plans anticipate government regulations will set the underwriting standards to be
met by all mortgages that underlie the qualifying MBS, roughly comparable to the
standards historically applied by the GSEs. The plans also generally anticipate a size limit
roughly equivalent to the conforming loan limit historically applied to the GSEs;
2) Investors in the qualifying MBS will be protected from all default risk by a
combination of private capital and government guarantee. The government guarantee
component is considered essential. The various plans differ primarily in the split between
private capital and government guarantee;
3) Risk-based insurance premia will be paid to the private capital and the government as
compensation for the risks they bear.
For simplicity, we refer to this structure as the “government insurance proposal.”
A key feature of the insurance proposal relative to any plan that would recreate the GSEs
is that the government would set the underwriting standards and be compensated for the
risk it bears.
The immediate question is whether the government can be effective and efficient
in carrying out such a mortgage insurance program. Evidence is available from a variety
43
of existing government insurance programs. Perhaps the most positive evidence is the
FHA program itself. As noted earlier, this program has existed since 1934, sets its
premiums on an actuarial basis, and has never required a government subsidy or bailout
for its self-supporting programs. Most interestingly, as documented in Jaffee and Quigley
(2010), the FHA effectively sat out the subprime boom, allowing its overall market share
to fall from a peak share of twenty-five percent in 1970 to under two percent by 2006.
Even more dramatically, its market share of loans to minority borrowers, which had been
close to fifty percent of this market as recently as 2000, fell to well below ten percent by
2006. In effect, the FHA took no action to deter its traditional clients from switching to
private market lenders and the GSEs as the source of their mortgage loans. While this
inaction could not protect the FHA from the rising loss rate that is now affecting most
segments of the U.S. mortgage market, it has certainly minimized the dollar amount of
the losses that the FHA could still potentially impose on U.S. taxpayers.
The FHA thus provides a model, or even a precise mechanism, for a broad
government guarantee program, possibly covering the same market share—at times fifty
percent of the overall market—that was traditionally served by the GSEs. Indeed,
operating within its traditional programs, the FHA market share of total mortgage
originations has already jumped dramatically from under two percent in 2006 to over
twenty percent in 2010. The issue is whether the FHA mechanism, which has worked
well serving a well-defined set of lower-income clients, can scale efficiently to serve
what could be as much as three quarters of the entire U.S. mortgage market (summing a
50 percent GSE share with a traditional 25 percent FHA share). The major concern is
whether the FHA -- or any comparable government insurance plan -- can resist the
44
political pressures to reduce its underwriting standards and to subsidize its risk-based
insurance premiums. The evidence here is not encouraging.
An interesting and comparable case is the National Flood Insurance Program
(NFIP). The NFIP was created in 1968, following a series of disastrous mid-western
floods that caused a large part of the private insurance industry to stop offering flood
coverage. The NFIP legislation required premiums to be set on an actuarial basis,
including risk-based premiums, to discourage the construction of new homes in flood
zones. This noble goal floundered, however, when the owners of existing properties in
dangerous flood plains successfully lobbied to obtain special “grandfathered” premium
reductions. This all become evident when there were insufficient reserves to pay the
losses created by Hurricane Katrina, thus requiring taxpayer bailout of the NFIP on the
order of $22 billion. Further discussion of the NFIP see Michel-Kerjan and Kunreuther
(2011) and of failed government insurance programs in general see Jaffee and Russell
(2006).
The Terrorism Risk Insurance Act (TRIA) provides an alternative approach to
government insurance and may provide a useful structure for a government mortgage
insurance program. TRIA was first passed by Congress in 2002, following the terrorism
attack of September 2001. The issue was that, as a result of their World Trade Center
losses, virtually all property insurers were refusing to renew policies on large commercial
buildings unless there was a substantial government reinsurance program to cap their
potential losses. TRIA accomplished this goal with a structure in which the government
provides the insurers protection against possible catastrophic losses while placing the
insurers in the first-loss position with a series of deductibles and coinsurance
45
requirements. Roughly speaking, TRIA 2002 required the industry itself to cover most of
the losses that would have resulted from another event comparable to the sabotage of
2011, but provided quite complete government protection against any losses above that
level. TRIA has now been renewed two times, and both times the deductible and
coinsurance requirements have been raised, so a taxpayer loss would now occur only with
truly extreme events.21
The specific proposals offered by Acharya, Richardson, Van Nieuwerburgh, and
White (2011) and Hancock and Passmore (2010) both reference “catastrophe insurance”
as the coverage to be provided under their plans. A particular concern, however, is that
MBS investors might not consider government catastrophe coverage to be a sufficient
inducement for them to take the first-loss positions on portfolios of U.S. mortgages. For
example, while the property insurers may have been most concerned with the last twenty
percent of the tail risk from terrorist attacks, investors in residential mortgage pools may
be primarily concerned with the first twenty percent of the risk distribution. In that case,
for a government mortgage insurance program to be effective, it may have to mimic the
NFIP more than TRIA. In other words, even if the starting point were the principle of a
backstop to catastrophe, the political process may create a plan that covers high-risk
mortgages at subsidized rates, i.e., GSEs with a different “cover.”
This appears to be the conundrum for creating a feasible program for government
insurance of U.S. mortgages. While a true catastrophe government insurance plan appears
feasible, investors and other market participants will, of course, have incentives to push
as much of the first-loss risk as possible under the government’s coverage. If the political
21 On the other hand, the government’s TRIA coverage is provided without charge.
46
process can stand firm on the issue, then it is quite possible that private incentives will
create an efficient market for U.S. mortgages. After all, it is hard to believe that only the
countries of Western Europe have the ability to create effective mortgage markets while
maintaining a low level of government intervention.
C. The Role of GSE Mortgage Market Activity under the Conservatorship
In concluding, it is relevant to comment on the role of GSE mortgage market
activity since the two firms were placed under a government Conservatorship in
September 2008. Relevant data on the home mortgage acquisitions of the GSEs and for
the total home mortgage market are shown in Table 9 for 2009 and 2010. The raw
numbers suggest a significant GSE and overall government role. For 2009 and 2010,
annual GSE mortgage acquisitions as a percentage of total home originations was 63
percent. FHA and VA activity averaged 24 percent of total home originations over the
same period, so government programs participated in 87 percent of all mortgage
originations for 2009 and 2010.
The high GSE market share under the Conservatorship, however, can be
misleading. First, 80 percent of all GSE mortgage acquisitions were refinanced loans, so
only 20 percent of the GSE activity represented loans for home purchase. The GSE
refinancing activity includes the refinancings that occurred under the Home Affordable
Refinance Program (HARP). In comparison, for the overall mortgage market, home
refinancings represented 68 percent of total mortgage originations, leaving 32% of the
originations for home purchase activity. The conclusion is that while the GSEs dominated
U.S. mortgage market activity in 2009 and 2010, most of this activity was simply the
refinancing of mortgage loans that had already been guaranteed by the GSEs. To be clear,
47
refinancing activities are certainly beneficial to the borrowers, and generally so for the
GSEs as well (since they reduce the likelihood of default on these loans for which the
GSEs are already at risk). On the other hand, refinancing is a zero-sum game, since the
investors who are holding the higher rate mortgages will have to reinvest their money at
the now lower market rates. Indeed, the Federal Reserve, U.S. Treasury, and GSEs are
major holders of these GSE mortgage securities, so the HARP program is far from costfree
for the government itself.22
The GSEs also participate in the Home Affordable Modification Program
(HAMP), along with servicers for non-GSE home mortgages. As of September 2011, the
GSE share of total HAMP modifications was 52 percent, only slightly above the GSE
share of all outstanding home mortgages. This suggests that the participation rate in
HAMP modifications was about the same for GSE and non-GSE mortgages. Perhaps
more importantly, the HAMP program is widely considered to be a disappointment: as of
September 2011, just over 800 thousand loans had been modified, compared to the earlier
hopes of 3 to 4 million loans.
The overall conclusion is that the primary mortgage market result of maintaining
the GSEs under the government Conservatorship through 2011 appears to have been their
role as a catalyst for the refinancing of their existing mortgages. In terms of funding for
home purchase loans, private market lenders have actually been more active than the
GSEs, even without the benefit of a government guarantee.