Conventional theories have assumed that long-run economic growth is driven by exogenously determined technological progress and that policy can have only short-run effects on growth. More recent theories of `endogenous growth' incorporate possible longer-run effects of policy and shocks, which may switch the economy from one equilibrium path to another. Such theories also account for the possible influence on long-run growth of factors such as human and physical capital accumulation, income distribution, migration, production structure and financial intermediation. Both theoretical and empirical papers on these issues were presented at a conference on `Growth and Development: New Theory and Evidence' in Tokyo on 9/10 January. This was CEPR's fourth joint conference with the National Bureau of Economic Research and the Tokyo Center for Economic Research, organized by Takatoshi Ito, Professor of Economics at Hitotsubashi University, and Hiroshi Yoshikawa, Professor of Economics at the University of Tokyo. Financial support was provided to CEPR by the Ford Foundation, to NBER by the Mitsubishi Trust, and to TCER by the Mitsubishi International Fund.

Presenting his joint paper with Robert Barro, `Regional Growth and Migration: A Japan-U.S. Comparison', Xavier Sala-i-Martin (Yale University, NBER and CEPR) considered the proposition that poor countries or regions grow faster than rich ones. From an analysis of per capita incomes for two regional data sets 47 Japanese prefectures during 1930-87 and 48 US states during 1880-1988 he found clear evidence of convergence in both countries. Poor prefectures and states grew faster; and both intra-regional and inter-regional convergence of incomes occurred at rates of about 2% per year a `universal constant' that proved remarkably robust to the choice of data set and model specification.

Sala-i-Martin also reported that the cross-sectional standard deviation of income across prefectures and states has been in decline over the long run, although it has increased slightly in both countries since the early 1980s. He also found striking similarities in the determinants of the rates of regional in- migration: in both countries, there was a slow but highly significant adjustment of population to initial income differentials. Sala-i-Martin noted that migration itself may offer an explanation of this convergence, if poor regions' per capita incomes grow faster because fewer people stay in them. The estimated response of migrations to income differentials is too small, however, to account for the speed of the estimated convergence.

In the ensuing discussion, Rudiger Dornbusch (MIT, NBER and CEPR) emphasized that the steady-state path to which countries converge may itself depend on many variables, such as incentives to accumulate human capital, and this limits the predictive content of the results. Ryoko Okazaki (Bank of Japan) noted that these results were difficult to reconcile with those found by Romer, and he suggested that the Japanese results could be explained in part by fiscal transfers across prefectures.

In a paper entitled `The Market Size, Entrepreneurship, and the Big Push', Kiminori Matsuyama (Northwestern University and NBER) extended the MurphyShleiferVishny model, in which the government may play a major role in `concerted development strategies' by shifting the economy from a `bad equilibrium' dominated by unproductive, traditional sectors to a `good equilibrium' with a well-developed manufacturing sector. This shift cannot be spontaneous because entrepreneurs in the modern sector take no account of a `pecuniary externality': the additional income they generate will create extra demand for other entrepreneurs' output.

Matsuyama argued that the `big push' hypothesis only provides a rationale for comprehensive central planning if a coordinated expansion across different industries is difficult to achieve through spontaneous responses of creative entrepreneurs. To analyse this transition explicitly, he used a variant of the MurphyShleiferVishny model in which this coordination difficulty is represented by inertia in the entrepreneurial decision process. The level of `critical minimum effort' (above which growth can occur) is a function of the degree of inertia and market size: when this is strictly positive, the economy has a unique equilibrium path and is caught in the poverty trap; but when it is zero the economy can move from the bad to the good equilibrium and continue on a path of sustained development.

Rudiger Dornbusch argued that models of this type are not robust to economic openness: the economy may be caught in a poverty trap only if it cannot use exports to get out of it. The active role of government in the Japanese and Korean industrializations also casts doubt on the feasibility of spontaneous coordination of the type Matsuyama's model suggests. Masahiro Okuno-Fujiwara (University of Tokyo) also stressed the difficulty of accounting for such `ease of coordination', while various participants expressed doubts about the assumption of perfect foresight in growth models.

Ross Levine (World Bank) presented a paper, `Financial Intermediation and Growth: Theory and Evidence', in which he developed a model in which financial intermediaries raise the growth rate for three reasons. First, they eliminate productivity risk by allowing investors to hold diversified portfolios. Second, they remove liquidity risk by pooling society's savings. This reduces the shares of portfolios that individual agents have to hold as liquid assets, which improves the composition of investment. Third, they undertake screening activities that eliminate bad investments more efficiently.

His empirical testing of the link between financial development and growth using cross-country data for 87 countries during 1960- 89 indicated that there was little correlation between the overall size of the financial system and the growth rate, although measures of the relative size of the banking sector vis-à-vis the central bank and of private sector credit vis-à-vis that of the government and public sector enterprises are significantly correlated with growth. Countries in which private banks were important and the private sector's share of total credit initially high tended to grow faster.

In the discussion, Shinji Takagi (Osaka University) and Kazuo Ueda (University of Tokyo) welcomed Levine's paper as a significant improvement on earlier studies of financial development, but they expressed doubts about the reliability of his measures. They suggested supplementing these findings on the role of financial intermediation with evidence from case-studies. Rudiger Dornbusch pointed out that these results need not imply that the level of financial repression should be set to zero: this is a legitimate tax base that should be used optimally in conjunction with other instruments to equalize marginal distortions across instruments.

In a paper entitled `Growth, Productivity, and the Access to the World Financial Markets', Daniel Cohen (University of Paris I, ENS, CEPREMAP and CEPR) criticized recent work by Barro and Sala- i-Martin on how free access to world financial markets affects developing countries' ability to catch up with industrialized countries. Such models assume that there exists at each moment a steady state to which the economy is `converging'. Since this steady state is itself moving all the time, however, their regression results provide little information about the speed of convergence towards or indeed the existence of such a steady state. This state is conventionally proxied by the level of secondary school enrolment, which is implicitly assumed to proxy the value the country places on education; but if secondary enrolment is an endogenous function of the country's stage of development, it will not correspond to the true steady state.

Cohen therefore developed an overlapping generations model of convergence which also incorporated human capital accumulation, which he then estimated on a sample of 66 countries for 1970-87. His results indicated that the impact of access to financial markets on large debtors was unambiguously lower than in the rest of the world, although this appears not to have resulted from endogenous differences. This may be either because their access to financial markets was insufficient even in the 1970s to speed up their physical capital accumulation or because the accumulation of physical capital is much less significant than that of human capital in driving growth in all countries.

The discussion focused on the effects of capital mobility on the convergence results. Participants noted that so long as there is some accumulation of immobile human capital, physical capital mobility need not affect these results. If the only cumulative factor is physical capital, however, income equalization should be expected when it is mobile but not otherwise. Xavier Sala-i- Martin disputed Cohen's assumption that wide differences in the marginal product of capital across countries indicate capital immobility. Differences in tax systems create wedges between the private and physical marginal products of capital, and only the former should be equalized when capital mobility is perfect.

In a joint paper with Thierry Verdier entitled `Historical Accidents and the Persistence of Distributional Conflicts', Gilles Saint-Paul (Département et Laboratoire d'Economie Théorique et Appliquée, Paris, and CEPR) argued that transitory episodes may have permanent effects on growth and income distribution if parents' investment in their children's education reacts differentially to taxation changes depending on their income. In a model in which taxation to finance public education is the outcome of the political process, he showed that a high tax rate can increase inequality by deterring the poor but not the rich from investing in private education.

Saint-Paul noted that this is more likely to occur when the initial income distribution is highly unequal, which will strengthen political support for redistributive schemes. High inequality may be persistent, while low inequality may lead to even lower inequality and income homogenization over time, since a low tax rate will then be more likely to induce the poor to invest in private education. Historical episodes that temporarily affect the dynamics of redistribution may therefore have long-run effects on growth and inequality. For example, it may be desirable to delay extending political rights to the poor to ensure convergence towards an egalitarian society: extending them while income inequality remains too high may result in too high a tax rate and a highly unequal equilibrium.

In the ensuing discussion, Tsuneo Ishikawa (University of Tokyo) suggested relating the paper to the literature on the economic and sociological determinants of education. He also noted that the empirical evidence on the effects of public education on growth and inequality is mixed.

In a paper entitled `On Productivity Change: Case Study of the Prewar Japanese Cotton Industry', written with Tetsuji Okazaki and Hajime Wago, Hiroshi Yoshikawa (University of Tokyo and TCER) considered the procyclical behaviour of observed total factor productivity. This empirical regularity is consistent both with the `real business cycle theory', in which shocks to technical progress drive economic fluctuations, and also with Keynesian theories in which labour hoarding causes observed productivity to fall during recessions because the labour input to the production function is over-estimated. It has been difficult to disentangle these interpretations because of problems in measuring the production function.

Yoshikawa described a data set on the pre-war Japanese cotton industry, however, for which the capital stock is well measured in terms of physical units, and which also contains an excellent measure of the rate of capacity utilization. His direct estimates of the production function on the capital stock alone indicated that productivity was procyclical, while similar estimates that also included capacity utilization as an explanatory variable indicate that it was acyclical. He concluded that this case study supported the Keynesian rather than the real business cycle theory.

In the discussion, Xavier Sala-i-Martin stressed that the real business cycle theory considered not only supply shocks but also real, non-monetary, demand shocks (such as preference shocks) as determinants of the cycle.

Jisoon Lee (Seoul National University) presented his paper, `Optimal Magnitude and Composition of Government Spending', which extended Barro's earlier work on the growth effects of government spending. He used a model in which the government has to allocate tax receipts optimally across various components (e.g. government consumption, investment and transfers). He found that slow- growing countries have high levels of government spending, of which transfers account for a large share, while fast-growing countries will have a small share of government spending, most of which will be allocated towards government investment.

In the discussion, various participants expressed doubts regarding the detrimental effects of transfers on growth. Xavier Sala-i-Martin argued that lump-sum taxation may be less efficient than economists traditionally believe, which may explain why it is not implemented in practice.

In a joint paper with Giorgio Brunello, Yasushi Ohkusa and Yoshihiko Nishiyama on `Corporate Hierarchy, Promotion and Firm Growth: Japanese Internal Labor Market in Transition', Kenn Ariga (Kyoto University and TCER) developed a theoretical model of the incentive system and hierarchical design of the Japanese firms, with internal promotion rather than monetary reward as the main incentive device. Applying his model to cross-sectional data on job ranks and wages, he found it consistent with the observed allocation of the span of control and age profile of wages prevailing in Japanese firms. The need for productive efficiency implies a hierarchical structure different from that generated by the need for incentives, which will therefore not be affected by factors such as the economy's long-run growth rate. Ariga argued that the recent productivity slow-down in Japan has led to conflict between the two structures and compelled firms to develop alternative systems of reward, new grading systems and alternative promotion ladders; and it has stimulated a greater spread of relatively small subsidiaries.

The papers presented at this conference will be published in a special issue of the Journal of the Japanese and International Economies in December 1992.