Documente Academic
Documente Profesional
Documente Cultură
Flows
Apoorva Javadekar1
September 2, 2015
Broad Question
1. Question:
What causes investors to invest or withdraw money from
mutual funds?
)
2. Litarature:
Narrow focus on Winner Chasing phenomenon
)
Investors choices
Risk Choices by fund managers
)
)
Literature Snapshot
1. Seminal Paper: Chevallier & Ellison (JPE,
1997)
Flows(t+1)
Returns(t)
= Convex Fund Flows in Recent Performance!
2. Why Interesting? Non-Linear Flows (could)
mean
Bad and extremely bad returns carry same
information !
)
Non-Bayesian Learning
)
Behavioral Biases
)
= de c isio n
s x
= s ig n
=p r io
Data
1. Source: CRSP Survivor-Bias free mutual fund dataset
2. Time Period: 1980-2012.
3. Include:
)
)
4. Exclude
)
)
)
)
Performance Measures
1. Reputation: Aggregate performance of 3 or 5 years prior to
current period
2. How to Measure Performance?
Factor Adjusted: CAPM or 3-factor (Fama,French
(2010), Kosowski (2006))
Peer Ranking (Within each investment style):
(Chevallier,Ellison (1997), Spiegel (2012))
)
3. Which Measure?
Not easy for naive investor to exploit factors like value,
premium or momentum = factor-mimicking is valued
(Berk, Binsbergen (2013))
Flows more sensitive to raw returns (Clifford (2011))
)
Peer ranking within each style control for bulk of risk
differentials across funds
CAPM wins the horse race amongst factor models
(Barber
et.al 2014)
)
4.I use both the measures: CAPM and Peer Ranking but not
Main Variables
A i,t+1 (A it (1 + r i,t+1 ))
A it
Empirical Methodology
1. Interact Reputation With Recent Performance: To
understand how investors mix signals with priors
K
= 0 +
FF i ,t +1
k=1
i ,t 1
(rankit )
Zk
+
i ,t 1
k=1
2. Variables:
)
)
k Z k
rankit [0,
1]
3. Structure:
)
Capture learning technology
No independent effects of reputation(t-1) on
flows(t+1):
)
Reputation
Dep Var:FFit+1
Peer
CAPM
Peer
CAPM
Time Effects
Standard Errors
N
Adj R-sq
Yes
Fund Clustered
13512
0.137
Yes
Fund Clustered
13512
0.135
Yes
Fund Clustered
11468
0.158
Yes
Fund Clustered
11468
0.148
Constant
-0.088***
0.109***
(0.021)
(0.021)
0.216***
0.202***
(0.010)
(0.010)
-0.894***
0.808***
(0.183)
(0.178)
-0.031***
(0.005)
-0.027***
(0.005)
-0.098***
0.126***
(0.022)
(0.022)
0.207***
0.193***
(0.011)
(0.011)
-0.830***
0.761***
(0.193)
(0.188)
-0.010
(0.005)
-0.006
(0.005)
-0.002
-0.002
-0.011***
-0.008***
Rank(t+1)
Risk(t)
Log Size(t)
FFit +1
Peer
CAPM
Peer
CAPM
Unconditional Estimates
Rank(t)
0.043
(0.041)
Rank-Sq(t)
0.296***
(0.043)
0.117**
(0.041)
0.223***
(0.043)
0.308***
-0.031
(0.059)
(0.054)
0.285***
-0.031
(0.061)
(0.060)
Rank-Sq(t)
Rank-Sq(t)
0.116
0.210***
-0.0693
(0.062)
0.124
0.250***
-0.0741
(0.074)
Top Reputation(t-1)
.2
0
-.2
Flow Growth(%)
.4
.
5
.5
Rank (t)
.
5
Unconditional Estimates
.2
.1
0
-.1
.3
.2
.4
Rank(t)
95 % Confidence Interval
Growth % (t+1)
.6
.8
1
Flow
Top Reputation(t-1)
.2
0
-.2
Flow Growth(%)
.4
.
5
.5
Rank (t)
.
5
Top Reputation
-.2
.2
.4
Low Reputation
.
5
.5
.
5
Rank ( t)
95 % CI
Growth %
Flow
Implications
1. Shape:
)
)
2. Level:
)
3. Slope:
)
Robustness Checks
Dep Var:FFit+1
Rank(t)
Peer
CAPM
Reputation
Peer
CAPM
Unconditional Estimates
0.0345
0.0871*
(0.0435)
(0.0430)
Rank-Sq(t)
0.276***
0.232***
(0.0453)
(0.0448)
Low Reputation
Rank(t)
-0.0978
-0.140*
(0.0592)
(0.0630)
Section II:
Risk Shifting
2. Previous Papers:
)
Basak(2007):
)
R tm +
s xs
i
lo a d ing
s x substantial (rit )
2. Fact: Factors (e.g market) explain
= p r ic e
i,2t
for
s2nd half
2t
sx
median for 2nd half
Some Statistics
Table:Summary Statistics For Risk Change
Reputation Category
Variables
Low
Med
Top
1.04
1.03
0.19
1.02
1.00
0.15
1.02
1.00
0.20
0.12
0.084
0.14
0.09
0.066
0.09
0.12
0.091
0.11
Annual Beta
Mean
Median
Dispersion
Risk
Mean
Median
Dispersion
Regression Results
Table:Risk Shifting
Unconditional
Peer
CAPM
Peer
CAPM
Time Effects
Style Effects
Standard Errors
Yes
Yes
Fund Clustered
Yes
Yes
Fund Clustered
Yes
Yes
Fund Clustered
Yes
Yes
Fund Clustered
Constant
0.298*
**
(0.015)
0.542*
**
(0.008)
-0.007
(0.004)
0.291*
**
(0.015)
0.543*
**
(0.008)
-0.007
(0.004)
0.304*
**
(0.015)
0.534*
**
(0.008)
-0.006
(0.004)
0.305*
**
(0.016)
0.532*
**
(0.008)
-0.006
(0.004)
-0.000
(0.001)
-0.000
(0.001)
-0.000
(0.001)
-0.002
(0.001)
Log Size(t)
Perf. Rank(H1)
-0.243***
(0.029)
-0.228***
(0.029)
Result Continued
Peer
CAPM
Peer
CAPM
Low Reputation(t)
Perf. Rank(H1)
-0.355***
(0.042)
-0.378***
(0.043)
Perf. Rank(H1)2
0.377***
0.410***
(0.048)
Medium Reputation(t)
(0.049)
Perf. Rank(H1)
-0.250***
(0.032)
-0.227***
(0.034)
Perf. Rank(H1)2
0.219***
(0.033)
0.208***
(0.034)
Top Reputation(t)
Perf. Rank(H1)
-0.0573
-0.0472
(0.042)
(0.043)
Discussion of Results
Section III
Model Of Fund
Flows
Model Overview
1. Question: What explains the heterogeniety in observed
Fund-Flow schedules
2. Possible Answer:
)
3. Basic Intuition:
)
)
Model Outline
1. Basic Set-Up:
)
it+1 N 0, ( ) 2
= + (r
)
2
it +1
it
it
i ,t +1
(t )2 + ()2
2. Beliefs:
)
)
Loss-Averse Investors
1. Assumptions:
)
)
u (t ) = t 1 (t < 0) + t 1 (t 0) If sell
0
If no sell
)
Evolution of
t
t +1 =
rj ,t +1
t+1 + ri,t+1
If no sell
If shift to fund j I i
If exit from industry
E [u ( +
) | ]
1t
exit
t
t
, ) = E [u ( +
) | ]
1t
r
= P (t + r1t+1 0) E t [t + r1t+1|t + r1t+1 0]
t
, ) = E [u ( +
) | ]
1t
r
= P (t + r1t+1 0) E t [t + r1t+1|t + r1t+1 0]
t
, ) = E [u ( +
) | ]
1t
= rP (t + r1t+1 0) E t [t + r1t+1|t + r1t+1 0]
1t
exit
t
, ) = E [u ( +
) | ]
1t
r
P (t + r1t+1 0) E t [t + r1t+1|t + r1t+1 0]
=
+ P (t + r1t+1 < 0) Et [t + r1t+1|t + r1t+1 < 0]
= Q ( + )
t
V sell ( , )
t 2t
(r
1t
u (t ) + E [u
t
) | ]
2t+1
2t
exit
t
) | ]
1t
V sell ( , )
t 2t
(r
E [u ( +
1t
u (t ) + E [u
t
) | ]
2t+1
2t
exit
t
) | ]
1t
V sell ( , )
t 2t
(r
E [u ( +
1t
u (t ) + E [u
t
) | ]
2t+1
2t
= u (t ) + P (r2t+1 0) E t [r2t+1|r2t+1 0]
exit
t
, ) = E [u ( +
) | ]
1t
r
P (t + r1t+1 0) E t [t + r1t+1|t + r1t+1 0]
=
+ P (t + r1t+1 < 0) Et [t + r1t+1|t + r1t+1 < 0]
= Q ( + )
t
V sell ( , )
t 2t
(r
1t
u (t ) + E [u
t
) | ]
2t+1
2t
= u (t ) + P (r2t+1 0) E t [r2t+1|r2t+1 0]
+P (r2t+1 < 0) E t [r2t+1|r2t+1 < 0]
=
=
=
Vsell ( ,
t 2t
E [u ( +
) | ]
1t
exit
t
1t
) = u (t ) + E [u
) | ]
t
2t+1 2t
(r
= u (t ) + P (r2t+1 0) E t [r2t+1|r2t+1 0]
+P (r2t+1 < 0) E t [r2t+1|r2t+1 < 0]
= u( )t + Q ( )
2t
Properties of Q()
1.Expression for Q(), R
.
Q () = + ( 1)
. .
(
2Q ()
=
1)
. .
<
(t )
1t
( , ) <
some
1t 2t
0
1t 2t
2. Understanding Why?
Q()
<=
s x
ut()
s
=Marginal
Lossx
skill
Note If shifted
Gain =
Q()
(2t 1t
) =
Loss
t
Optimal Policy
1. Result 3: Loss-Holding Region Increases in
1t
) Why? Relative gain from shifting (
) decreases as
2t
increases
1t
1t
)
)
2t
1t
2t
1t
( , ) >
some
1t 2t
0
If Q ( ) > , liquidate any gain.
No liquidation is optimal if
1t
>
2t
2. Solution:
.
2 t
i = it2
it
3. Discussion:
)
Dynamics Of Investor-Base
Figure: Dynamics Of Investor-Base
Alternative Theories
1. Lynch & Musto (JF,2003):
)
)
)
Conclusions
nd
Thank You !