Evaluating TFP differences by a mapping structure in a convergence framework

It is generally established that large differences in income levels across countries and regions are mainly due to differences in total factor productivity (TFP henceforth) levels (Klenow and Rodriguez Clare, 1997; Prescott, 1998; Hall and Jones, 1999). Typically, most of these studies show that about half of cross-country per capita income differences are left unexplained after taking into account for differences in physical or human capital. In addition, in recent TFP literature divergence in the form of twin-peaks phenomena with reference to income distribution emerge, and this evidence is usually attributed to cross country divergence in TFP rather than to factor accumulation rates. Feyrer (2003) finds that long run distributions of output per capita and TFP are bimodal while distributions of capital-output ratio and human capital are unimodal. The persistence of wide differences in total factor productivity levels among regions is empirically related to heterogeneity in technological characteristics as well as in institutions. Geographical proximity effects are also identified. From a theoretical perspective, the efficiency of an economy may be influenced by productivity improvements driven by firms’ innovation decisions. While the mechanism driving R&D decisions is analyzed in endogenous growth models, incorporating institutions in growth theories is currently dealt with in the literature, but much work must still be done . By ‘institutions’ we mean various aspects of law enforcement, the functioning of markets, inequality and social conflicts, democracy, political stability, government corruption, the health system, financial institutions, etc. Finally, regional studies have identified geographical spillovers and neighborhood effects that may explain uneven regional development. The theoretical explanation of such phenomena is based on models with convergence clubs, where history and geographical characteristics fundamentally condition the growth path. In all cases, the theoretical implications of technological, institutional, and geographic factors point in favor of multiple equilibria First, the paper aims at providing a suitable measure of the unknown TFP levels by introducing a mapping structure within the conditional convergence framework. We investigate differences in TFP levels by a spatial representation of the regional fixed effects in terms of some characteristics suggested by economic theory. The main (spatial) factors which drive differences in TFP levels across regions are identified without imposing an a priori spatial structure in the growth model specification. Second, besides traditional dynamic panel estimators, a generalized maximum entropy estimation procedure is developed with the aim of improving the efficiency of the estimates by dealing with ill-posed and ill-conditioned inference problems. Finally, we apply the proposed methodology to estimate a conditional convergence equation with reference to data collected from CRENOS for Italian regions (sample period 1960-1996), with the purpose to evaluate the convergence process, as well as the role of technological spillovers through human capital accumulation and agglomeration spillovers through the geographic distribution of economic activities (districts), both in the determination of steady-state GDP levels and in the accumulation process of TFP levels. Our procedure is used to detect the presence of multiple equilibria and club convergence, and results are compared to the empirical evaluation of TFP differences proposed by other authors.