THE EFFECTS OF PUBLIC CAPITAL ON REGIONAL CONVERGENCE IN TURKEY Ozlem Onder, Metin Karadag, Ertugrul Deliktas Department of Economics, Ege University
Aim of the Study to investigate the impact of public capital on regional convergence in Turkey at NUTS 1 level by using dynamic panel data model for the time period
Table 1: Basic Data for Turkish Regions at NUTS I Level NUTS I RegionsPopulation in Year 2000 (000) Share of the total population in 2000 (%) GDP per capita in 2000 (YTL, in) (1987 prices) Real GDP as a Proportion of Turkey’s Average in 2000 Share of total Gross Domestic Product in 2000 Share of total Pub. Invest. Expend. on average ( ) Istanbul Western Marmara Aegean Eastern Marmara Western Anatolia Mediterranean Central Anatolia Western Black Sea Eastern Black Sea North Eastern Anatolia Middle Eastern Anatolia South Eastern Anatolia Turkey
Methodology Conditional Convergence Model (Mankiw et al. 1992) Dynamic Panel Data Model (Islam, 1995) = per capita real income, = real public capital stock,
Methodology Spatial Panel Data (Elhorst, 2003) Fixed Effect Spatial Lag Model (FSLM) Fixed Effect Spatial Error Model (FSEM)
Data Annual data on per capita real income and public capital stock at NUTS-1 level regions of Turkey between
Table 3: Estimation Results Dependent Variable: y it Pooled LSDVGMM-SYS Constant0.114 (0.98) (-1.35) y it *** (93.49) 0.880*** (35.39) 0.851*** (17.30) pubcap it (0.35) 0.056* (1.72) 0.154*** (2.30) Convergence rate λ R2R N252 Instruments diagnostics Hansen11.87 AR(1) c -2.87*** AR(2)-0.71
Table 4: Estimation Results with Spatial Effect Dependent variable: y it FSLM FSEM y it *** (18.481) 0.831*** (25.204) Pubcap it (1.345) (0.176) W.y it 0.09* (1.846) 0.613*** (0.000) Convergence rate λ R2R N252
Questions for further study Is it possible to apply spatial panel data models to dynamic panel data models Convergence rates What are specification tests for panel data