Bond risk premia in macroeconomic models

In the first chapter we investigate the observed slope of inflation compensation in nominal government bond yields and demonstrate that it is very difficult to explain by the existence of large inflation risk premia. Using data from the UK and the US we show that a large class of micro-founded models fail to produce such premia in a way that is consistent with data. We propose a DSGE model with regime switching in the central bank's inflation target. Taking advantage of the well documented change in UK monetary policy to adopt inflation targeting, we estimate our model using nominal and inflation-linked Treasury bond data from the UK from 1985 to 2007. We find that this model can account for the term structure of inflation compensation in the nominal yield curve by generating regime dependent conditional expectations of future inflation.;In the second chapter we investigate the ability of DSGE models with heteroskedastic shocks to account for the time variation of excess returns on default-free bonds that has been observed in US postwar data. We find that a suitably calibrated simple model can produce sufficiently volatile bond risk premia for certain parameter values. Such calibrations however produce some implausible implications for the behaviour of real interest rates and macroeconomic variables. Moreover, the calibrated parameters cannot be recovered from an estimation of the model using US data.;The last chapter investigates the predictability of government bond returns by macroeconomic variables. It confirms previous findings that returns are correlated with past macroeconomic data in-sample. It also shows that this correlation declines and loses significance when returns on longer-maturity bonds are considered. Finally, it demonstrates that the factors exctracted from a large panel of macroeconomic variables, unlike interest rate forwards, have no predictive power for returns out of sample.