Public Policy Sources #37:
Growth theory: the causes of economic growth

[Previous] [Contents] [Next]

Understanding the sources of economic growth is important. Our central concern, however, is with the causes of economic growth and it is difficult to say much about that without a theory of growth. Hence, in this section I will briefly discuss developments in the theory of economic growth over the past few decades. In the subsequent section, I will discuss this theoretical work in the Canadian context, and review some of the more recent empirical studies. I will attempt to keep the discussion as non-technical as possible. More technically minded readers are advised to read Barro and Sala-i-Martin 1995.

Neoclassical growth models

Although most analysts would date the birth of the modern theory of economic growth to the 1950s, the classical economists--Adam Smith, David Ricardo, and Thomas Malthus--were the first to discuss many of the basic ingredients of modern growth theory.2 In particular, their emphasis on competitive behaviour, equilibrium dynamics, and the impact of diminishing returns on the accumulation of labour and capital are integral elements of what is called the neoclassical approach to growth theory. During the 1950s, this approach to understanding growth was formalized by Solow (1956) and Swan (1956), and was later extended by Cass (1965) and Koopmans (1965). The basic assumptions underlying the neo­classical growth model are as follows.

• The productive capacity of the economy can be adequately characterized by a constant-returns-to-scale production function with diminishing returns to capital and labour.3
• Firms are price-takers in a competitive market place. In other words, no individual firm has any influence over market prices and individual firms are assumed to possess no market power.
• Technological change (i.e. productivity growth) is entirely exogenous (i.e. independent of the actions of the consumers and producers) and is available to all countries at no cost.

The implications of the neoclassical model of growth are straightforward. The first major implication is that sustained increases in per-capita income can be supported only by sustained increases in total factor productivity. In this model, output per worker can rise only if (a) the ratio of capital per worker increases or (b) total factor productivity increases. Since this model assumes diminishing returns to capital, there is a limit to how much capital accumulation can add to output per capita. Hence, the only way to increase output per worker in the long run is to have sustained productivity growth. This is a major weakness of the neoclassical growth model, since long-run growth is exogenous, i.e. determined by an element that is entirely outside of the model.

The second major implication of this model is the "conditional convergence thesis," which states that economies with lower initial levels of real output per worker relative to the long-run level should experience faster economic growth. This property follows from the assumption of diminishing returns to capital: the lower the ratio of capital per worker, the higher the return to investing in capital. Hence, the lower the ratio of capital per worker, the faster the rate of capital accumulation and the faster the growth rate of output per worker. This implies long-run convergence in output per capita. Convergence is said to be conditional here since the long-run level of capital per worker and output per worker depend on the saving rate, the growth rate of the population, and the existing technology--factors that are unlikely to be identical across countries. The convergence thesis is strengthened by the assumption that all countries can acquire technological progress at no cost.

Endogenous growth models

For several decades, the neoclassical growth model remained the benchmark model of economic growth. Starting in the 1980s, however, a number of newer, more sophisticated growth models have been developed. A key feature of these new models is that, unlike the neoclassical model, technological change is not assumed to be exogenous. Hence, the new growth models are sometimes dubbed "endogenous growth models" since a key task in these models is to explain where technologically driven productivity growth comes from. In particular, the accumulation of knowledge plays a key role in driving productivity growth in these models.

There are essentially two strands in the endogenous growth literature. The first strand takes its cue from articles by Romer (1986) and Lucas (1988).4 In this variety of endogenous growth theory, the assumption of constant returns to scale is dropped. In particular, knowledge is assumed to be an input of production with increasing returns to scale.5 In this class of models, it is possible for per-capita output to grow without bound. In addition, convergence of per-capita incomes need not occur in the long run. A survey of some of the developments in this area can be found in Romer 1994.

The second strand of endogenous growth models also takes its departure from an article by Romer (1990) but has been extended by other economists such as Grossman and Helpman (1991), Aghion and Howitt (1992), and others. In these models, an effort is made to model the micro­economic environment in which firms accumulate knowledge. In particular, the assumption of perfect competition is dropped. This is because the acquisition of knowledge through research and development activity is costly and can only be rewarded if firms have some ex post market power.6 Hence, in these models, firms are assumed to compete in a monopolistically competitive environment. As in the first class of endogenous growth models, per-capita output growth can occur without bound since there need be no tendency for the economy to run out of ideas. Furthermore, convergence across countries need not occur in the long run.

Much of the new research also includes models of the diffusion of technology (see Grossman and Helpman 1994 for a survey of this literature). As we noted earlier, in the neoclassical model it was simply assumed that technological change could be adopted at no cost by all countries. In many of the newer growth models, this assumption is dropped and an effort is made to analyze directly how technological progress is transferred across countries. One important implication of this research is that the location of research and development (R&D) activity may matter. If there are significant agglomeration effects associated with R&D activity, then the benefits of R&D are largely captured by the country in which R&D activity takes place.7

Another key feature of the endogenous growth models is that the long-run growth rate can depend on government actions. In the basic neoclassical growth model, government does not have an impact on the long-run growth rate. In an endogenous growth framework, however, government policy can affect the long run rate of growth, since government policy actions--taxation, provision of infrastructure, protection of intellectual property, regulations, maintenance of law and order, and so on--can affect the underlying rate of inventive activity. Government, therefore, has great potential for harm or good in these models.

Empirical tests of the models

In recent years, significant empirical work has been conducted to test a number of the predictions of both the neoclassical and endogenous models of growth. Tests of the neoclassical model have focused on the conditional convergence thesis and the results have generally been mixed. While most studies reject the hypothesis of convergence across all countries, many find support for convergence across more homogenous subsets of countries or regions. For instance, Baumol (1986) finds support for the convergence thesis among OECD countries, while Barro and Sala-i-Martin (1991, 1992a, 1992b) find support for the convergence thesis across American states, regions of several European countries, and prefectures of Japan.

Much empirical work has also been done to test a weaker version of the convergence thesis. This weaker convergence thesis posits that convergence should occur among countries, holding constant such factors as initial levels of human capital, measures of government policy, political stability, and so on. The broadest cross-country sample supports this weaker form of the convergence thesis (see Barro 1991; Mankiw, Romer and Weil 1992). These studies find that the rate of convergence is about two percent per year. This implies that it takes about 35 years for an economy to eliminate half of the gap between its initial level and its long-run level of per-capita income.

Most other tests of modern growth theory take the form of regression analysis using data from a broad sample of countries. In these studies, variations in the growth rates of the per-capita GDP of these countries are analyzed for statistically significant association with a number of variables including the initial level of per-capita income, the initial quantity of human capital approximated by the average years of schooling, the ratio of government consumption expenditures to total output, measures of political stability, measures of market distortion by government, measures of openness to international trade, and the ratio of gross investment to output. Across many studies we find that growth seems to be positively related to human-capital attainment, negatively related to government consumption, positively related to the investment-to-output ratio, positively related to measures of openness to international trade, negatively related to measures of market distortion (i.e. economies with more extensive market distortions grow more slowly) and positively related to political stability (i.e. more politically stable countries experience faster growth). Measures of economic freedom also appear to be positively related to economic growth (see Gwartney, Lawson, and Block 1996). Using Mankiw, Romer and Weil's (1992) model, Easton and Walker (1997) estimate that increases in measured economic freedom raise the steady-state level of per-capita income.8 Other studies find that investment in research and development is also positively related to growth (see Coe and Helpman 1993), lending support to some of the basic ideas behind the endogenous growth literature.

There are several weaknesses to this approach to testing growth theory. One major flaw, which has been identified by Pack (1994) among others, is that these studies do not test the growth model under consideration directly. The problem with an regression analysis of growth that uses data from a sample of several countries is that such analysis can only identify correlations among the relevant variables. It does not tell us the direction of change nor does it test directly the growth process under investigation. Hence, definitive conclusions in support of one model or another cannot be drawn from these studies.

Another major weakness to this approach is that many of the variables that appear to be significant in regression analysis of growth using data from a sample of several countries do not survive sensitivity testing. Levine and Renelt (1992) conduct a sensitivity analysis of regression analysis of growth in a sample of several countries and find that very few variables appear to be consistently related to growth across different regression-equation specifications. In particular, they find that the only variable that appears to be robust across regressions is the investment-to-output ratio; none of the broad array of fiscal-policy variables and very few of the variables from trade policy and political stability that we have considered are robustly correlated with growth in per-capita income. Thus, the meaningfulness of much of the empirical work on growth theory remains in question. As is common in economics, the theoretical work in the economics of growth is far ahead of the empirical work.

Conclusions about
economic growth

A few broad conclusions can be drawn from this brief overview of the economic growth literature.

While endogenous and neoclassical growth models offer different explanations for the growth process, in both models, growth in total factor productivity (i.e. technological change) is an essential component of economic growth. In the neoclassical model, technological progress is essential for long-run growth in per-capita output. In endogenous growth models, productivity growth results from spill-overs from human capital accumulation or inventive activity and this is what generates long-run growth in per-capita income. Hence, productivity growth--"working smarter" as opposed to "working harder"--is an essential component of overall economic growth.

Straightforward capital accumulation and population growth is not sufficient for sustained growth in per-capita income. In the neoclassical model, the law of diminishing returns limits the extent to which raw factor accumulation can raise per-capita income. The emphasis, therefore, should be on accumulation of inputs of superior quality. The accumulation of capital and labour will increase the long-run rate of economic growth if this capital embodies more sophisticated technology and if workers are more skilled.

The development of new technologies and their diffusion across firms and nations are critical components of the growth process (Mokyr 1990; Lipsey 1996). Productivity-enhancing technological change does not fall from the sky. It is the result of purposive inventive activity on the part of economic actors. Hence, the incentives for individuals and firms to innovate and create new products, machines, tools, and production techniques is critical. In addition, it is costly for other economic agents to adopt new technologies. Clearly, a country that is able to adopt new technologies faster is able to grow faster. Moreover, if agglomeration effects are present, the location of R&D activity will matter, with countries that are host to R&D activity reaping the largest gains. Hence, the location and diffusion of technological progress across firms and nations are almost important as technological change itself.

Institutional factors--government regulations, taxes, provision of basic infrastructure, political stability, etc.--clearly matter for long run economic performance (see North 1990). This is because the accumulation of factors of production and the development of new technologies do not occur in a vacuum. Rather, economic exchange and production occurs in the real world where the incentives matter. How public policy is set over the long run will therefore influence productivity growth and economic growth since public policy is a critical determinant of the institutional environment.

[Previous] [Contents] [Next]

info@fraserinstitute.ca You can contact us at the above email address for any comments or information requests. Please report any dead links or technical problems.

If you know someone who would be interested in this web page, please enter their email address below, and we will forward this URL to them: Email Address:

Last Modified: Thursday, August 5, 1999.