Distinguishing between potential sources of growth convergence for the Portuguese economy within the EU. A panel data - time series study of the aggregate production function

What are the potential sources of growth and how is the convergence process of the Portuguese economy within the EU characterised? We answer this question by determining the most suitable specification of the aggregate production function, CES or Cobb-Douglas, for the EU countries as in Duffy&Papageorgiou (2000). If the aggregate production technology is best described by a CES production function then the potential sources of growth are wider than the ones associated with a Cobb-Douglas technology. For instance, with an elasticity of substitution between inputs greater than one (ó>1) it is possible to have endogenous growth (see Jones&Manuelli (1990), Rebelo (1991)) while for ó<1 multiple equilibriums arise (see Azariadis (1993, 1996, 2001). To test for the most suitable production function specification we consider a sample of seventeen European countries between 1960 and 1987. The tests are conducted within a panel data and time series framework based on data retrieved from the STARS database of the World Bank. Three different kinds of samples were considered: a) all the seventeen countries; b) three of the cohesion countries, Portugal, Greece, Ireland, and Iceland; and c) each country separately, and two types of production functions ? one with raw labour and one with human capital adjusted labour. By considering groups of countries and not only each country separately it is possible to distinguish between each country?s behaviour and that of the average economy and also to characterise ó according to the income level of the different countries in our sample. Previous to the estimation of the non-linear production function by maximum likelihood and GMM techniques we tested the series for stationarity both in a time series and a panel data framework. We also used linear estimation techniques, generalised least squares with individual fixed effects and cointegration techniques. We conclude that it is not possible to reject the CES specification for the countries in our sample. Since ó>1, it is possible to have endogenous growth although the characterisation of our series does not allow us to ignore the spurious regression problem.