Introducing Advanced Macroeconomics:

Growth and business cycles

Chapter 8
PRODUCTIVE EXTERNALITIES AND ENDOGENOUS GROWTH

The McGraw-Hill Companies, 2005

Introduction to endogenous growth

In typical Western countries, income per capita has increased by factors between 6 and 8 over the last 100 years. How has this been possible? Technological growth: in all of our Solow models, long run economic growth is rooted in technological growth. But technological growth is unexplained in these models. An endogenous growth model explains/endogenizes the long run technological growth rate and hence the long run growth rate of output per worker. That is, the model shows how these growth rates depend on model parameters. Thereby, the models imply statements on how economic policy affects long run growth.

The McGraw-Hill Companies, 2005

Two types of endogenous growth models

R&D based: Explicitly describe the production of technological progress, i.e. contains a production function with output, At +1 At , depending on certain inputs. Externality based: No explicit production function for technological progress, but an assumption that (labour augmenting) technology in every firm, At , depends positively on aggregate capital (or output) because of productive externalities. This implies increasing returns in the aggregate production function allowing growth of GDP per worker in the long run without exogenous technological growth. This chapter: Endogenous growth based on productive externalities. The new feature: At = Kt (could be At = Yt , alternatively)
The McGraw-Hill Companies, 2005

A model of endogenous growth based on productive externalities

There is one representative firm, which we can see in two roles:
1. As the sole producer of aggregate output. 2. As the individual small firm that takes all aggregates as given.

When the firm decides its capital demand, Ktd , it takes the aggregate capital stock, Kt , as given because the firm is too small to influence the economys aggregates. But at the end of the day, that is, in equilibrium, one must have Ktd = Kt because there is only this one firm. At the level of the individual firm, the production function is: d d 1 Yt = ( K t ) ( At Lt ) , 0 < < 1, where At is taken as given, and there is CRS to ( Ktd ,Ld ) , t in accordance with the replication argument.
The McGraw-Hill Companies, 2005

 Because of productive externalities the individual firms At depends on aggregate capital, Kt :

At = Kt , > 0.

At the aggregate level the production function (in equilibrium) is:

Yt = K
t

( K Lt )
t

= Kt

+ (1 ) 1 t

L .

The sum of the exponents is 1 + (1 ) implying that there is IRS when > 0 . We have CRS at the individual firm level and IRS at the aggregate level.

The McGraw-Hill Companies, 2005

The micro economy is as in the Solow model: Competitive capital and labour markets implies real rental rates equal to marginal products for the inputs. But which marginal products are relevant? Since the optimality conditions come from the individual firms factor demands, it should be the marginal products at the firm level, where At is taken as given: 1 d d Kt Kt wt = (1 ) At . rt = , d d At Lt At Lt Then inserting Ktd = Kt , Ld = Lt and At = Kt etc. gives: t
Kt rt = K t Lt
1

K t wt = (1 ) At . Kt Lt

It follows that rt Kt = Yt and wt Lt = (1 ) Yt . Our theory of the functional income distribution is The McGraw-Hill Companies, 2005 as usual.

 How can we motivate the assumption of productive externalities? Empirics: estimates of the equivalent of our models 1 + (1 ) are often larger than one, e.g. around 1.5 (a fairly large estimate, however). This gives a of around . Theory (reasoning): learning by doing. Workers get more skilled (they are learning) as they use new capital (by doing). Workers become more productive, not only because there is more capital (the direct/internal effect), but because they learn new skills, which they keep if they are deprived of the new capital, e.g. if they get another job (the indirect/external effect). The last effect does not (in the longer run) accrue particularly to the individual firm, but to all firms.
The McGraw-Hill Companies, 2005

The complete model

consists of the equations for the factor prices plus:
Yt = ( K t

) (A L )
t t

, 0 < <1

At = Kt ,

St = sYt ,
K t + 1 = S t + (1 ) K t , Lt +1 = (1 + n ) Lt .

Parameters: , ,s, ,n . State variables: Kt and Lt . No equation At +1 = (1 + g ) At .

The McGraw-Hill Companies, 2005

 Take another look at the aggregate production function:

Yt = K
t

( K Lt )
t

= Kt

+ (1 ) 1 t

L .

If = 0 : the basic Solow model. Assume therefore that > 0 : increasing returns in the aggregate production function. If < 1, then + (1 ) < 1 : diminishing returns to the reproducible factor, capital, alone. Leads to semiendogenous growth. If = 1, then + (1 ) = 1 : constant returns to capital alone. Leads to truly endogenous growth. What about > 1 ? Gives an extreme model (Exercise 5).

The McGraw-Hill Companies, 2005

Semi-endogenous growth ( < 1 )

Define
% % kt kt / At = Kt / ( At Lt ) and yt = yt / At = Yt / ( At Lt ) . 1 % % From Yt = ( Kt ) ( At Lt ) , we get: yt = kt .

From At = Kt : Then:

At +1 Kt +1 = . At Kt

1 % kt +1 K t +1 / K t K t +1 / K t 1 Kt +1 % = ( A / A ) L / L = ( K / K ) L / L = 1 + n K kt t +1 t t +1 t t t +1 t t +1 t

Inserting Kt +1 = St + (1 ) K t gives:
1 % kt +1 1 Yt 1 % 1 + (1 ) 1 . % = 1 + n s K + (1 ) = 1 + n skt kt t The McGraw-Hill Companies, 2005

Rearranging gives the transition equation:

1 % % 1 % kt +1 = kt skt + (1 ) 1+ n

1 % = s kt 1 + n

( )

+ 1

+ (1 )

( )

% kt

1 1

The transition equation has properties: 1. It passes through (0,0). 2. It is everywhere increasing. % 3. There is a unique positive intersection, k * , with the 45o % % % line. Inserting kt +1 = kt = k gives: % 1 + (1 ) 1 (1 + n )1 / (1 ) (1 ) = sk 1 % 1 + n = sk

4. The transition eq. hits the 45 degree line from above (differentiate etc. and use n + > 0).
The McGraw-Hill Companies, 2005

s % k = 1 / (1+ ) (1 + n ) (1 )

1 / (1 )

% k * > 0 ( we assume that n + > 0)

% % k% t converges to k * implying that y t converges to:

% % y = k
*

( )
*

s = 1 / (1+ ) (1 + n ) (1 )

1 / (1 )

.
The McGraw-Hill Companies, 2005

This defines steady state.

Growth in steady state

% % When kt kt / At and yt = yt / At have converged to the % % constant steady state values k * and y * , respectively, kt and yt must grow at the same rate as At . The growth rate of At is endogenous! We can easily find its value in steady state: % % In steady state kt +1 / kt = 1 , from which:
1 % kt +1 1 Kt +1 K 1 / (1 ) = = 1 t +1 = (1 + n ) % Kt kt 1 + n K t

At +1 Kt +1 / (1 ) = = (1 + n ) At Kt

At +1 At / (1 ) = (1 + n ) 1 g se . At
The McGraw-Hill Companies, 2005

 Thus, our model implies convergence to a steady state with a common constant growth rate of kt , yt and At :
g se = (1 + n )
/ (1 )

1.

In fact, there is balanced growth in steady state. We have a steady state with endogenous growth: positive growth in yt without exogenous technical progress, and the growth rate depends on model parameters. But we only have g se > 0 if n > 0 : labour force growth is required for economic growth. Our steady state is therefore one of semi-endogenous growth: there is only economic growth if the labour force grows! Intuition: to utilize increasing returns, scale increases are required
The McGraw-Hill Companies, 2005

 The most obvious implication for structural policy is: in order to promote long run economic growth, promote population growth! Reasons for being cautious with such a policy:
the well-known thinning-out of capital effect. the empirics.

The McGraw-Hill Companies, 2005

Empirics for semi-endogenous growth

Plotting g i against ni , 1951-2000, across 55 countries:

The McGraw-Hill Companies, 2005

In contradiction to semi-endogenous growth, there is a clear negative correlation between g i and ni across countries, but: Direction of causality? Does n depend on income? Is our model a country by country model or is it for the World? The figure is in accordance with the models transitory growth if convergence is slow: perhaps we should not reject the model, but only the idea that the countries are in steady state?

The McGraw-Hill Companies, 2005

The McGraw-Hill Companies, 2005

 Country by country tendency: population growth decreases and economic growth increases from the first to the second subperiod. This speaks against semiendogenous growth. If convergence is very slow, lower population growth in combination with higher economic growth can still be in accordance with the transitory growth of the model. We want a model with observable predictions.

The McGraw-Hill Companies, 2005

Endogenous growth ( = 1 ): the AK model

In the model just considered ( < 1 ) convergence becomes very slow as 1 . Assuming that = 1, we approximate the case of a large just below one. We also assume n = 0 , and thus Lt = L otherwise our model becomes extreme. The complete model boils down to two equations, derived from the aggregate production function and the capital accumulation equation, respectively:
Yt = K
t

( At L )

=K

( K L)
t

= Kt

+ (1 ) 1

= Kt L1 ,

Kt +1 = sYt + (1 ) Kt .

The McGraw-Hill Companies, 2005

 The AK model:
Yt = AKt ,
1 A L ,

Kt +1 = sYt + (1 ) Kt .

Dividing both sides of both equations by L gives yt = Akt and kt +1 = syt + (1 ) kt which combine to:
kt +1 = sAkt + (1 ) kt = ( sA + 1 ) kt
kt +1 k t = ( sA ) kt kt +1 kt = sA ge . kt

We assume that sA > ensuring that ge > 0 .

The McGraw-Hill Companies, 2005

 Since yt = Akt and At = Kt = kt L , all of kt , yt and At grow at the rate ge : there is balanced growth.
kt+1 = k t

kt+1 - kt kt+1 - kt

45

kt

kt

kt

We do not have convergence to a steady state: kt grows at rate ge all the time.

The McGraw-Hill Companies, 2005

 We have one common, constant growth rate for kt , yt and At :

ge = sA

Truly endogenous growth: positive growth in yt without exogenous technological growth or population growth; the growth rate is given by model parameters, among these s . Structural policies: higher s and lower give everlasting higher growth in GDP and consumption per capita. Critique: = 1 is a knife-edge case (not a valid objection, = 1 is an approximation of less than, but close to one). Scale effect: remember that A = L1. A larger constant population results in higher growth and an increasing population gives accelerating growth. The McGraw-Hill Companies, 2005

Empirics for endogenous growth

The most important prediction: larger rate of investment, s i , increases the growth rate, ge . Plotting g i against s i , 19602000, across 90 countries gives:

The McGraw-Hill Companies, 2005

 This is in nice accordance with the endogenous growth model. But it could also be explained as the transitory growth of an exogenous growth model. The latter is not the case, however, if there is a positive correlation between s and the growth rate of the technology variable, At (this rate is exogenous in exogenous growth models, but equal to ge = sA in this model). Plotting the growth rate of At as determined by growth accounting against s i, 1960-2000, across 84 countries gives:

The McGraw-Hill Companies, 2005

This is probably the most important single argument in favour of endogenous growth models.
The McGraw-Hill Companies, 2005

Conclusions
1. Productive externalities, or learning-by-doing spillovers can motivate increasing returns to capital and labour at the aggregate level at the same time as constant returns at the firm level. 2. This can result in a well behaved growth model with long run growth in income per worker without this being generated by exogenously assumed technological progress. 3. Sufficiently weak spillovers lead to semi-endogenous growth, where the long run economic growth rate is mainly determined by, and depends positively on, the growth rate of the labour force.

The McGraw-Hill Companies, 2005

4. Strong spill-overs lead to truly endogenous growth, where the long run economic growth rate depends positively on the investment rate. This feature is empirically plausible. However, the scale effect is unattractive. 5. Externalities in production an interesting mechanism, but it is difficult to get firm evidence on the performance of this endogenous growth model.

The McGraw-Hill Companies, 2005