An analysis of the determinants of economic crises: Cross-sectional spatial regression and spatial panel models

In this study, the contagion of the 2008 global financial crisis among countries through geographical, financial and commercial interaction channels are analyzed with cross-sectional spatial analysis and spatial panel data analysis for thirty one developed and fourteen developing countries during the 2001-2016 period. The non-performing loans to total loans ratio (NPL) which is an important crisis indicator was defined as the dependent variable and sixteen macroeconomic indicators as explanatory variables by using the literature. While only geographical neighborhood relation is used in cross-sectional spatial models, financial and commercial neighborhood relations are also taken into account in spatial panel models as well as geographical spatiality. General-to-specific Hendry approach selected and specification tests applied to select the best models. Before the crisis (2007), during the crisis (2008), after the crisis (2009) periods were evaluated with the cross-sectional spatial analysis. According to our findings, there was a strong spatial relationship occurred only in 2007, in developed countries and the Spatial Autoregressive Model (SAR) was the best model. In the SAR model; one of the countries NPL is affected by the average NPL level of neighboring countries. Country's explanatory variables have a direct effect on itself and an indirect-spillover effect on neighboring countries. In spatial panel models; fixed effects models are preferred to random effects. In the geographical neighborhood between developed and developing countries SAR model, in the commercial neighborhood between developed and developing countries and in the financial neighborhood between developed countries Spatial Error Model (SEM), in the financial neighborhood between developing countries Spatial Durbin Model (SDM) is obtained as the best model. In the SEM model it is observed that the NPL ratio of countries is spatially affected by the error terms that are not included in the model and the explanatory variables only have a direct effect on their own country. In SDM model; there is a spatial interaction between the NPL ratios of countries and the explanatory variables, and the explanatory variables have a direct and spillover effect. When the best spatial panel models are analysed it is seen that the variables of current account balance, export to import ratio, share price index, real interest rate of deposits, trade balance and GDP growth are statistically and economically significant variables explaining the crisis in all models. In addition to identifying the determinants of the 2008 global crisis, this thesis makes a significant contribution to the literature by verifying the spread of the crisis through financial channels in spatial panel models for the first time.