Differences in growth performance across western industrialized countries are surprisingly large and persistent. Recent growth data suggests, interestingly, that these countries belong to different growth clubs, with the membership of each club being determined by a threshold requirement on a candidate country's growth rate. Differences in the growth rates of countries belonging to different clubs can be striking. For example, the data indicates that, for the ten year period 1995–2004, large European countries like Germany and Italy lagged well behind smaller countries like Ireland and the Nordic countries as well as the United States. More specifically, while growth rates in Germany and Italy have averaged around 2% during the last 20 years, the Nordic countries and, particularly, Ireland, with her growth of approximately 7%, have outperformed Germany and Italy by a surprisingly wide margin.
As a first pass, it seems natural to hypothesize that the large observed differences in country-specific growth rates principally reflect important differences in the basic institutions underlying the functioning of these economies. An argument along these lines rests nicely on modern growth theory and can find a fair amount of support from empirical research on the factors explaining growth.
Modern growth theory emphasizes policy and other relevant institutions as important or even critical preconditions for a country to be able to enjoy growth gains. On the other hand, the statistical as well as economic significance of institutional factors for a country's growth has been vindicated in numerous empirical studies. Even so, important as institutional factors are, reference to the lack of an appropriate institutional framework supporting growth in the large European economies may not be as plausible as it seems at first sight. The reason is that these countries have already undergone a series of structural reforms and the basic institutions that research has found important for growth are already in place in all European countries.
Naturally, a number of other country-specific factors besides institutions may have contributed to the observed differences in growth, such as size of country, demographics and availability of natural resources. However, as long as we are considering growth in the policy context, the fundamental problem with these factors is that it is very difficult to design effective growth policies that decision-makers could implement in order to stimulate growth. Therefore, if we wish to incorporate a nontrivial role for policy, it is natural, and in line with the empirical growth literature, to focus on those features that can be influenced by policy.
A closely related issue that modern growth theory identifies as critical for growth relates to the role of knowledge or, in the more specific context of empirical measurement, to the role of education and research and development. A striking feature of the growth data is that, whereas it took roughly 60-70 years for countries to double their income levels in the 19th century, the best growth performers in the 20th century have shown that this can happen in approximately 15 years! Growth-supportive institutions are not enough to generate record growth rates. Knowledge and its rapid diffusion are both essential.
But modern growth theory also suggests that knowledge alone is not enough: it is the interaction between knowledge and growth-supportive institutions that underlies growth momentum. Hence, this approach puts much less emphasis on exogenous sources of technological progress, capital deepening and population growth as the main sources of economic growth, all factors which lie at the heart of growth thinking à la Solow.
The intriguing question now is what do fresh data say about the relative importance of each of these factors for growth? In a forthcoming Bank of Finland discussion paper, 'Why do growth rates differ? Evidence from cross-country data on private sector production', Juha Kilponen and Matti Virén approach the data in an attempt to provide an answer to this important question. As Kilponen and Virén rightly emphasize, one of the novel features of their analysis is the data. Specifically, they run regressions using annual data on business sector production, wages and R&D investment from 14 OECD countries covering the period 1960–2004. This is in contrast to most previous empirical work, which uses measures of aggregate output including public sector output. This implies, in particular, that Kilponen and Virén avoid all the complications emerging from the difficult measurement issues related to government output and capital stock as well as the productivity of government production. As pointed out by many economists engaged in empirical research on growth, these measurement problems alone may lead to highly misleading results concerning income shares and underlying production relationships. Furthermore, they tend to be more severe in the case of emerging economies.
The estimation results obtained by Kilponen and Virén suggest a set of interesting conclusions concerning the relative importance of the various sources of growth alluded to above. The authors start by estimating aggregate Cobb-Douglas production functions under the restriction that income shares are equal across countries but the rate of technical change can be country-specific. Using an earlier sample (1960–1994) for estimation, Kilponen and Virén then forecast GDP in-sample using actual data on capital and labour input for the period 1995–2004.
The surprising result is that the observed growth differences show up in the forecast errors from forecasting GDP for the period 1995–2004. Naturally, these forecast errors reflect not only different patterns of capital accumulation and employment growth, but also any other factors not taken into account in the estimation of the basic production function. However, when the authors decompose the forecast errors they find that approximately half of their variations across countries can be explained by R&D investment intensity, the ratio of R&D expenditure to business sector GDP. On the basis of this result alone, Kilponen and Virén tentatively conclude that the evidence puts some weight on the notion that R&D intensity has a role in separating good from bad growth performers.
The authors then proceed to explain the estimated variation in total factor productivity across countries. After extracting estimates of total factor productivity using a constant returns to scale Cobb-Douglas production function, Kilponen and Virén set up a statistical model for total factor productivity using a time trend, R&D intensity, openness of the economy and patents as well as cross terms of time, R&D intensity and patents as explanatory variables. To check for robustness, they also estimate an equation for labour productivity derived from profit maximization under a constant returns to scale production function using wages, a time trend and R&D intensity as explanatory variables.
One of the benefits of this latter specification comes from the fact that the authors need not assume that the elasticity of substitution between capital and labour is one. As far as the role of R&D intensity is concerned, the results do not seem to be particularly robust but are certainly interesting. If R&D intensity alone is included in the estimation equation, it enters in a statistically significant way. However, once additional controls are added, like openness of the economy, patents and a time trend possibly also allowing for cross terms, R&D looses its statistical significance.
In contrast, while R&D intensity becomes insignificant, openness of the economy and patents become highly significant factors. Kilponen and Virén interpret this as suggesting that R&D investments pay off once they produce useful innovations, ie patents. Interestingly, productivity enhancing patents seem to be more important in open economies, an observation supported by the estimated significance of the cross term of openness and patents. The authors obtain qualitatively similar results from estimating the labour productivity equation derived from profit maximization. In particular, R&D intensity as well as openness and patents retain their statistical significance even after allowing for time varying elasticity of substitution between capital and labour.
Finally, the authors check for time variation, in particular potential increases in the elasticity of substitution between capital and labour. There is a proposition in the reference literature on growth showing that when countries start at the same level of income, the one with higher elasticity of substitution will end up growing faster. This alone makes testing for increases in the elasticity of substitution particularly interesting.
In this respect, the results are somewhat disappointing. First of all, the estimated elasticity of substitution is fairly low, around 0.5. It seems to be slowly increasing, but not as fast as we could expect from previous results concerning the size of the elasticity. A further complication in interpreting these results comes from the fact that the elasticity of substitution is estimated very imprecisely, an outcome that has plagued much of the research on estimating production functions. Nevertheless, one of the profound lessons is that, although the results obtained by Kilponen and Virén are encouraging, there are still loose ends and much to be learned from investing research effort in this area. This is not the right time to stop digging deeper into the mysteries of economic growth. |