In our discussion above of the logic of the model, we introduced four variables that tie macroeconomic performance to microeconomic behavior and that were v aried experimentally in the simulation runs. These variables were the ease of innovation, the emphasis on imita tion, the cost of capital, and the labor-saving bias of search. What ef fect do different settings of these variables have on the macroeco nomic time paths in the model?
We adopted a linear regression approach to this question. We con sidered t hree different macro economic variables: the Solow technol ogy index in year forty, the capital-labor ratio in year forty, and the four- firm concentration index . Our four experimental variables we designated XIN , XIM , XR, and XWT. We assigned the value one to these variables when (respectively) major innovation was relatively easy, search emphasis was on imitation, the required dividend rate was high, and the search was somewhat biased in a labor- saving direction.6
The effects on the p eriod-forty value of the Solow technology index are characterized by the following regression equation:
Figures in parentheses are significance levels. The conjecture that easier major innovation at a microeconomic level should lead to a faster rate of growth of total factor productivity at a macroeconomic level is strongly confirmed. This lends additional confidence that the model provides plausible and understandable connections between the microeconomic phenomena and macroeconomic phenomena of economic growth. Note that this is not a trivial result, since the rate of growth of total factor productivity and the level of the Solow tech nology index late in an economy’s evolution here are simply macro statistics, and do not correspond directly to features of the model.
Some interesting results also come out of regression analysis of the determinants of the capital-labor ratio in year forty.
The hypothesized effects of factors three and four are strongly con firmed. A higher price of capital, considered as a return that must be paid out and that is not available for reinvestment, does lead to a substantially less capital-intensive mode of production after a period of time. Considered as a growth rate effect, the rise in R from 0. 02 to 0.06 produces a decrease of 0.3 percentage points per period in the rate of change of the capital-labor ratio. The effect of the labor-saving search bias introduced by factor four is of comparable magnitude but, of course, in the opposite direction.
The magnitude and significance level of the coefficient of XIN comes as something of a surprise. Why should the capital-labor ratio be higher in a system in which search is less local? On reflection, one possible answer to this question seems to be the following. The gen eral direction of the path traced in input coefficient space does not depend on the localness of search. However, the rate of movement along the path is slower if search is more local. Therefore, given that the path is tending toward higher capital-labor ratios (as a conse quence of the level chosen for R and the neutrality or labor-saving bias of search), the capital-labor ratio that results after a given number of periods is lower when search is more local.
Another possible answer is more Schumpeterian . A high rate of technical progress may produce a high level of (disequilibrium) prof its, which in turn are invested. The resulting increase in the demand for labor results in a higher wage and deflects the results of profita bility comparisons in the capital-intensive direction. These possible answers are not, of course, mutually exclusive.
The regression result regarding concentration is :
Here, C4 is the four-firm concentration ratio. The imitation effect is clearly the most pronounced. We have suggested an explanation for this effect in terms of the “closer race.” There are actually two dis tinct mechanisms in the simulation model by which a closer tech nical race tends to keep concentration down, and both are quite plau sible as hypotheses about economic reality. First, as among fi rms in business, similarity in technique implies similarity in cost condi tions, hence in profit rates, and hence in growth rates. Thus, a closer race implies a smaller dispersion of firm growth rates and lower con centration. But, second, potential entrants also stay closer to the tech nical leaders when imitation is easy and perceived opportunities for profitable entry thus occur more frequently. Since entry tends to occur in a particular (and relatively low) scale range, the amount of capacity added by entry is higher when entry is higher. Consider ations of overall industry “equilibrium” imply that the infusion of capacity through entry is partially offset by lower investment by the firms previously in business. Since the latter are typically larger than the entrants, concentration is reduced.
The above analysis of the influences on the concentration of firms is illustrative of a fundamental difference between the neoclassical and evolutionary approaches to growth theory. Neoclassical growth theory is ai med at macro phenomena, and its micro details are in strumental to its macro purposes. Evolutionary theory treats the micro processes as fundamental and treats the macro aggregates as aggregates. Hence, it encompasses a wider range of phenomena; its treatment of the micro details is intended to be subj ect to test. Thus, for example, we can treat our simulation model not only as an abstract account of the phenomena of aggregate economic growth, but also as an abstract account of the size distribution of firm s. This we will do in a later chapter.
Source: Nelson Richard R., Winter Sidney G. (1985), An Evolutionary Theory of Economic Change, Belknap Press: An Imprint of Harvard University Press.