GVAR Model Research Articles

Regional unemployment forecasts with spatial interdependencies

We forecast unemployment levels for the 176 German labour-market districts on a monthly basis. Because of their small sizes, strong spatial interdependencies exist between these regional units. To account for these, as well as for the heterogeneity in the regional development over time, we apply different versions of a univariate spatial GVAR model. When comparing the forecast precision with that of univariate time series methods, we find that the spatial model does indeed perform better, or at least as well. Hence, the spatial GVAR model provides an alternative or complementary approach to commonly used methods in regional forecasting which do not consider regional interdependencies.

Linking global economic dynamics to a South African-specific credit risk correlation model

In order to address practical questions in credit portfolio management it is necessary to link the cyclical or systematic components of firm credit risk with the firm's own idiosyncratic credit risk as well as the systematic credit risk component of every other exposure in the portfolio. This paper builds on the methodology proposed by Pesaran, Schuermann, and Weiner [Pesaran, M.H., Schuermann, T., and Weiner, S.M., (2004), Modeling regional interdependencies using a global error correcting macroeconometric model, Journal of Business and Economic Statistics, 22, 2, 129–169.] and supplemented by Pesaran, Schuermann, Treutler and Weiner [Pesaran, M.H., Schuermann, T., Treutler, B., and Weiner, S.M., (2006), Macroeconomic dynamics and credit risk: a global perspective, Journal of Money, Credit, and Banking, Volume 38, Number 5, August 2006, 1211–1261.] which has made a significant advance in credit risk modelling in that it avoids the use of proprietary balance sheet and distance-to-default data, focusing on credit ratings which are more freely available. In this paper a country-specific macroeconometric risk-driver engine which is compatible with and could feed into the GVAR model and framework of PSW (2004) is constructed, using vector error-correcting (VECM) techniques. This allows conditional loss estimation of a South African-specific credit portfolio but also opens the door for credit portfolio modelling on a global scale, as such a model can easily be linked to the GVAR model. The set of domestic factors is extended beyond those used in PSW (2004) in such a way that the risk-driver model is applicable for both retail and corporate credit risk. As such, the model can be applied to a total bank balance sheet, incorporating the correlation and diversification between both retail and corporate credit exposures. Assuming statistical over-identification restrictions, the results indicate that it is possible to construct a South African component for the GVAR model that can easily be integrated into the global component. From a practical application perspective the framework and model is particularly appealing since it can be used as a theoretically consistent correlation model within a South African-specific credit portfolio management tool.