Are the government bond yield determinants of developed and developing countries substantially different in- and outside crisis periods and can the bond yields be predicted using rolling-window estimation?

In this thesis we examined whether the determinants of government bond yields of twenty-four developed and ten developing countries worldwide are substantially different in- and outside crisis periods. We observed the ten-year government bond yields (Federal Reserve Bank of St. Louis) for the period 1991 quarter one to 2014 quarter four. We distinguished the yield determinants (Federal Reserve Bank of St. Louis, CSP, IMF, NBER, World Bank) in financial market, real market and political or institutional factors. We estimated three different panel data models (pooled OLS, fixed effects and first-differences) to quantify the relationships between our observed economic variables. We showed that factors from the financial market (e.g. OECD recession and total reserves) and real market (e.g. inflation and export per capita) are important determinants to explain the variation of bond yield series of developed countries. For developing countries, we showed that real market- (e.g. GDP growth and import per capita) and financial market (e.g. government expenditures) factors are important yield determinants. We used rolling-window schemes to forecast the bond yield series and showed that rolling-window forecasting is more accurate than traditional (i.e. out-of-sample) forecasting during crisis periods.