Sei sulla pagina 1di 11

NOTES ON THE NEOCLASSICAL APPROACHES TO TECHNOLOGICAL CHANGE

Alessandro Sterlacchini PRODUCTION FUNCTION


Implicit form

Q = A f(X) X= x1, xn

Time suffixes are omitted for simplicity: however, we are interested in how Q increases over time when inputs change.

SIMPLIFIED APPROACH Q = A f(K, L)


This implies that Q should be a measure of value added (rather than total output). Value added = at firm and industry level GDP = at country level (aggregate production function)

COBB-DOUGLAS PRODUCTION FUNCTION Q =AKL Cobb, C. and P. Douglas (1928) A Theory of Production, AER.
Charles Cobb was a mathematician. Paul Douglas (1892-1976) was an economist (University of Chicago) but also a politician. He served as a Democratic U.S. Senator from Illinois from 1949 to 1967.

LONG RUN ANALYSIS: THERE ARE NOT FIXED FACTORS OF PRODUCTION 1. CHOICE OF THE OPTIMAL COMBINATION (TECHNIQUE) OF K AND L: THAT MAZIMIZING THE FIRM PROFITS (maximum output for given costs or minimum costs for a given output) 2. ECONOMIES OF (OR RETURNS TO) SCALE 3. TECHNOLOGICAL CHANGE

COBB-DOUGLAS PRODUCTION FUNCTION Q =AKL


0< , <1 are the elasticities of output with respect to K and L. The marginal productivity of capital is: dQ/dK= AK-1L = (AKL)/K = *(Q/K) dQ/dK=*(Q/K) implies that =(dQ/dK)*(K/Q) which is the standard formula for an elasticity. The same applies to (prove it).

THE MARGINAL PRODUCTIVITIES (OR RETURNS) OF CAPITAL AND LABOUR ARE POSITIVE BUT DECREASING.
dQ/dK is clearly positive (see above) To find whether it is increasing or decreasing in K, we must look at the second derivative

d2Q/dK2= (-1)AK-2L = (-1)*Q/K2 .


Since <1, (-1)<0. This means that also d2Q/dK2<0, i.e. there are decreasing returns in K. The same applies to L (prove it).

K Graphically, assuming for instance that L is given, the relationship between Q and K will be positive but less than proportional (i.e. decreasing).

Note: the presence of decreasing returns in K and L implies the convexity of ISOQUANTS which is essential for guaranteeing the choice of the optimal technique.
The assumption of decreasing returns has important implications when we move to the aggregate production function and analyse the growth of a countrys GDP over time (i.e. when moving to the growth theory).

RETURNS TO SCALE
Returns to scale are different from the returns of single inputs. In fact, they are associated with an equal positive variation of both inputs, which can be interpreted as an increase of the production scale. So, the question is: how Q changes when there is an identical change in K and L? According to the C-D production function the answer is given by the sum of the elasticities and . If the sum is equal to 1 there will be constant returns to scale. If lower than 1 decreasing, and if greater than 1 increasing returns to scale. To prove it, just assume that both K and L double and the new level of output is Q*

Q* =A(2K)(2L). This can be written as Q* =A(2)+ KL and then Q* =(2)+ Q.


Accordingly, if +=1 (i.e. there are constant returns to scale) the output doubles precisely as the inputs. If the sum is greater than 1, the output increase will be greater than that of the inputs. The presence of increasing returns to scale explains a big deal of why firms tend to become larger and larger. At the level of aggregate economies, however, the very existence of scale economies (a term used to depict the same phenomenon) is less straightforward. Also for this reason, constant returns to scale are usually assumed.

LABOUR PRODUCTIVITY AND CAPITAL/LABOUR RATIO


Q =AKL
To find the labour productivity, divide both sides by L

Q/L = (AKL)/L = AKL-1


Then, multiply the right-hand side by (L/L), that is by 1, one gets

Q/L = A(K/L) L-1L = A(K/L) L+-1


If there are constant returns to scale (+=1; i.e. +-1=0 ) the equation becomes

Q/L = A(K/L)
and shows that the average productivity of labour only depends upon the A factor and the capital/labour ratio (elevated to ).

If the C-B production function is taken as the base of a growth model (as it occurs with the neoclassical traditional model) an important implication is that labour productivity increases less than proportionally with the increase in the capital/labour ratio. This is due to the decreasing returns from the physical capital K ( <1).

Q/L

K/L
So, by using this narrow framework, the prediction is that, as far as K will become larger, the increases in labour productivity will be lower and, at the limit, will approach to zero. We will be back to this important issue in discussing (old and new) growth theories.

Note: if the returns to scale are not constant the level of labour productivity (Q/L) can be also affected positively (negatively) by the scale variable L+-1 if there are increasing (decreasing) returns to scale.

TOTAL FACTOR PRODUCTIVITY


Let start again with Qt =At KtLt (the only novelty is that we now insert time suffixes in order to emphasise that we study the function over time). Transform all the variables in natural logs

lnQt = lnAt + lnKt + lnLt


To find their annual growth rates one can take the log differences: for the output variable we get lnQt1 - lnQ t0. To simplify the notation and avoid time suffixes, such a difference can be labelled as lnQ

lnQ = lnA + lnK+ lnL


Remember: to find the average annual growth rates over a given period of time one can take the mean of the annual log differences (alternatively, the compound annual rate can be used). The shift factor A is a measure of Total Factor Productivity (TFP), a comprehensive index of productivity or efficiency concerned with all the inputs. Labour productivity is a partial rather than total index of efficiency since it refers to one input only. In our case, we have that TFP=Q/(KL) =A The growth rate of A is then a measure of the TFP rate of change and can be computed as a residual

lnA = lnTFP = lnQ lnK lnL


4

According to the traditional neoclassical approach (Solow, 1957), the growth rate of TFP is interpreted as the rate of change of output due to exogenous technological change (manna from heaven). Technological progress is then depicted as a (positive) shift of the production function and explains why labour productivity can continue to improve over time, even in presence of decreasing returns from capital.

Q/L

t2 t1

K/L

Above, you can find a graphical representation of the above phenomenon: from time 1 to time 2 the production function shifts upward: this implies that, for the same level of K/L, labour productivity reaches a level higher than that recorded at time 1.

GROWTH ACCOUNTING
Suppose of having data for Q (GDP), K and L of a given country for a given period of time. In order to find the growth rate of TFP by means of the equation

lnTFP= lnQ lnK lnL


you must know the values of the parameters and . If there are constant returns of scale [+=1 and =1- ] and the assumption of perfect competition holds, the two factors of production are paid according to their marginal productivity expressed in monetary terms. This means that

(dQ/dK)p=r or (dQ/dK)=r/p; similarly (dQ/dL)p=w or (dQ/dL)=w/p


where p is the output price (GDP price index), r is the average price of capital and w denotes the average wage, that is the price of labour: r/p and w/p are then the price of capital and wage expressed in real terms. As we saw before dQ/dK=(Q/K) so that (Q/K)= r/p and then =(rK)/(pQ). Thus, can be interpreted as the share of GDP that goes to capital and we can term it sK.

Similarly, it is easy to prove that =1- is equal to the share of income distributed to labour [that is (wL)/(pQ)]. Such a share, termed sL, is by definition, equal to 1- sK. In short, the share of labour income is precisely one minus the share of income that goes to capital. In general, it is easier to find data for sL and, then, obtain sK residually. As a consequence the growth rate of GDP (Q) can be decomposed in the following manner:

lnQ = lnTFP + (1-sL) lnK + sL lnL


The above equation is the classical (and simplest) example of a so-called growth accounting exercise. Accordingly, the TFP change can be easily computed as

lnTFP = lnQ (1-sL) lnK sL lnL


Alternatively, since there are constant returns to scale, we can start by expressing the production function in terms of labour productivity (see above).

Q/L = A(K/L)
Assuming for notational simplicity q=Q/L and k=K/L, we have: q = Ak. By taking logs and computing annual log differences one gets lnq = lnA + lnk where the growth rate of labour productivity can be decomposed into two elements: the rate of change of A (or TFP) and that of the capital/labour ratio. Again, the change in A (or TFP) can be obtained as a residual

lnA = lnTFP = lnq lnk


As before, can be substituted with 1-sL, i.e. the share of income distributed to capital.

The table below shows the average annual rates of growth of labour productivity (lnq) and total factor productivity (lnA) concerned with four developed countries and different time periods from 1963 to 1998.

1963-73 lnq lnA Japan UK USA Italy 8.6 3.3 2.2 6.5 6.1 2.0 1.5 4.7

1973-79 lnq lnA 3.2 1.3 0.6 2.4 1.8 0.2 -0.1 0.5

1980-90 lnq lnA 2.7 2.3 1.1 2.0 1.6 n.a. 0.8 1.2

1990-98 lnq lnA 1.8 1.5 1.4 2.3 0.7 1.2 1.1 1.1
6

Many stylized facts are well represented in the table. Over the first period, Japan and Italy were experiencing an intense process of catching-up. From 1973 to 1979 (i.e. after the international crisis due to the first oil shock and the end of the dollar standard) there was a productivity slowdown. The growth rates of the golden age were never achieved in the following years. Finally, while from 1963 to 1990 they were clearly losing ground, during the 1990s the United States have witnessed a recovery, especially in terms of TFP changes.

SOLOW RESIDUAL
Robert Solow (1957) was the first scholar employing a Cobb-Douglas aggregate production function to compute TFP as a residual. Moreover, he interpreted its growth over time as exogenous (and disembodied) technological progress. Taking the data for the US economy from 1909 to 1949 he first computed the annual log levels of TFP as residuals

lnTFP = lnA= lnQ - (1-sL) lnK sLlnL


and found (by means of a regression analysis) that, over the years considered, the US labour productivity doubled and about 87% of the increase was attributable to technical change (i.e. to the residual). Thus, changes in the capital/labour ratio did play a minor role. The problem was that the major responsible of labour productivity growth was something that could not be directly measured (namely, a residual) and Solows interpretation of it as exogenous technological change was indeed too restrictive. Moreover, the Solow interpretation of technological change was limited to process innovations, i.e. the improvements allowing a more efficient exploitation of given inputs. Product innovations (improved or entirely new outputs) were neglected. In an almost contemporaneous work, Moses Abramovitz (1956), analysing the US economy in the long-run, came to a similar conclusion about the size of the residual. However, he stressed that for its own nature the residual should be interpreted as a measure of our ignorance. It could be due to a variety of causes such as increases in knowledge and education. The latter are examples of efficiency-enhancing factors that cannot be considered exogenous since they depend upon intentional investments.

HOW TO REDUCE OR EXPLAIN THE RESIDUAL? REDUCING THE RESIDUAL


Capital and labour inputs are not well measured. The more recent vintages of physical capital are more productive since they embody new pieces of knowledge (embodiment hypothesis). So, the measure of capital should be adjusted for quality changes. The same applies to the measurement of 7

labour input: it should be adjusted for the different composition in terms of skills and levels of education. Thus, one hour worked in 1949 is likely to be much more productive than one in 1909. The same applies to a capital unit. If labour and capital are not adjusted for quality changes, their increased productivity is captured by the residual. This contributes to explain its bigness.

EXPLAINING THE RESIDUAL


Technological change is not exogenous because it also depends on the intentional investments undertaken by firms to introduce process and product innovations. Starting from the 1970s some applied economists used a production function that included, along with traditional inputs, an additional factor called technological or knowledge capital whose value depends on the cumulated investment in R&D undertaken by firms (Griliches, 1979). The starting point is the implicit production function

Q= TFP f(K, L)

[where TFP=A; see our starting equation Q= A f(K, L)].

Then, it is assumed that TFP= g(A, T) where A= et (the shift factor A is an exponential function of time) and T is a measure of KNOWLEDGE OR TECHNOLOGICAL CAPITAL and can be computed as

Tt= RDt + (1-)RDt-1+ (1-)2RDt-2 + (1-)3RDt-3..


Where RD is the flow of R&D investment undertaken in previous years, diminished each year by the rate of depreciation (the idea is that, at the end of year t, the R&D invested in year t-1 is less productive than that undertaken during year t). Moving to the explicit form, we get the so-called "augmented" Cobb-Douglas Production Function

Q= et T K L(1-)
Taking logs [the log of et is t, where t is time] one gets

lnQ = t + lnK +(1- ) lnL + lnT


Then, in terms of rates of change we have

lnQ = + lnK +(1- ) lnL + lnT


[the rate of change of t over time is constant and equal to ] In presence of constant returns the scale and perfect competition in input markets, we have

lnQ - (1-sL) lnK - sL lnL= + lnT


8

The left-hand side of the above equation is precisely the rate of change of TFP. So we get

lnTFP = + lnT
which states that the growth rate of TFP is explained by a rate of exogenous technological change () and the change in knowledge capital T multiplied by the parameter . R&D expenditures include the costs for specific capital (scientific instruments), materials or intermediate goods used for laboratory experiments and proofs, and, above all, skilled and highly educated workers (researchers, scientists and engineers). If data are not corrected for the above double counting and there is a positive contribution to output of technological capital, this means that the specific inputs devoted to R&D activities are more productive than the traditional factor of production. In this sense, should be considered an "excess" elasticity. To be stressed is that the introduction of knowledge capital into the (augmented) C-B function solves the problem due to the assumption of decreasing returns for traditional physical capital (see below).

R&D SPILLOVERS (POSITIVE EXTERNALITY)


Firms can benefit from their own R&D investment (private returns) but also from the overall investment in knowledge (spillover effect, a positive externality).

Qi= et Ki Li (1-) Ti Ta

where Ta = i Ti

Examples: the technological capital of an industry a is the sum of that of the firms i belonging to it. Alternatively, the technological capital of a country a is sum of that of its industries i. Thus, at the aggregate level, the elasticity of output with respect to technological capital will reflect both private and social returns to R&D and, as a consequence, it will be higher than at micro level.

Qa= et Ka La (1-) Ta+


The above contributions introduced in the 1970s were limited to the explanation of TFP growth. They were based on an endogenous treatment of technological change. However, formal theoretical models of economic growth based on the same insights (and overcoming the limits of Solow-Swan model) were developed later.

NEOCLASSICAL MODELS OF ECONOMIC GROWTH


SOLOW (SWAN) GROWTH MODEL
The notion of growth as increased stocks of capital goods (means of production) was codified as the Solow-Swan Growth Model. In this view, the role of technological change became crucial, even more important than the accumulation of capital.

This model, developed by Robert Solow and Trevor Swan in the 1950s, was the first attempt to model long-run growth analytically. It assumes that countries use their resources efficiently and that there are diminishing returns to capital and labour increases. From these two assumptions, the neoclassical model makes three important predictions. First, increasing capital relative to labour creates economic growth, since people working with more capital can be more productive. Second, poor countries with less capital per person will grow faster because each investment in capital will produce a higher return than that achievable by rich countries with greater capital. Third, because of diminishing returns to capital, economies will eventually reach a point at which no new increase in capital will create economic growth. This point is called "steady state". The model also stresses that countries can overcome this steady state and continue growing only thanks to technological change. However, the process by which countries may go on in their growth process (despite the diminishing returns to physical capital) is "exogenous" and represents the creation of new technology that allows to get the same output with fewer inputs. The empirical evidence does not support many of this model's predictions, in particular, that all countries grow at the same rate in the long-run or that poorer countries should grow faster until they reach their steady state.

ENDOGENOUS GROWTH MODELS


Growth theory advanced again with the theories of economist Paul Romer in the late 1980s and early 1990s. Other important new growth theorists include Robert Lucas (1988) who focussed on the role of human capital. Unsatisfied with Solow's explanation, economists worked to "endogenize" technology. They developed the endogenous growth theory by means of a formalized explanation of technological advancement. These models also incorporated a new concept of human capital, the skills and knowledge that make workers productive.

From a resource-based to a knowledge-based economy


Unlike physical capital, knowledge (technological) and human capital exhibit increasing rates of return. New ideas are not scarce resources (like labour and physical capital) because they can be accumulated without limit (Romer, 1990). New pieces of knowledge are not rival goods. The marginal cost of using them is negligible. The non-rival nature of ideas is the attribute that drives economic growth. However, contrary to pure public goods they are partially excludable thanks to patents, trademarks and copyright (IPR), and also secrecy and, as a consequnce, they can be produced by profit-seeking private firms. Therefore, economies never reach a steady state. Growth does not slow as capital accumulates, but the rate of growth depends on the types of capital a country invests in. This implies that, rather than a unique steady state equilibrium, there are multiple equilibria. From the traditional concept of convergence we move to the concept of conditional convergence (or 10

convergence clubs) depending on the levels of knowledge and human capital that are available in a particular country. The above arguments have opened the door to policy interventions in support of knowledge and human capital investment with a view to sustaining economic growth. Thus, they have a lot to do with the growth-enhancing strategy launched by the EU in the 2000 Lisbon Council.

References
Abramovitz, Moses (1956) Resource and Output Trends in the United States since 1870, American Economic Review. Aghion, P. and P. Howitt (2009) The Economics of Growth, London: The MIT Press. Chapters: 1 (sections 1.1-1.2) and 5 (sections 5.1-5.5). Griliches, Zvi (1979) Issues in Assessing the Contribution of Research and Development to Productivity Growth, Bell Journal of Economics. Lucas, Robert (1988) On the Mechanics of Economic Development, Journal of Monetary Economics. Romer, Paul (1990) Endogenous Technological Change, Journal of Political Economy. Solow, Robert (1956) A Contribution to the Theory of Economy Growth, Quarterly Journal of Economics. Solow, Robert (1957) Technical Change and the Aggregate Production Function, Review of Economics and Statistics. Swan, Trevor (1956) Economic Growth and Capital Accumulation, Economic Record.

11

Potrebbero piacerti anche