ABSTRACT
This article investigates the source of predictability of emerging market (EM) local currency bond risk premia by using a dynamic factor approach based on a large panel of economic and financial time series. We find strong predictable variation in EM local currency excess bond returns that is associated with macroeconomic activity. We provide evidence that the main predictor variables are the factors based on real economic activity that are highly correlated with measures of industrial and manufacturing production; however, factors based on global financial factors also contain information about the future local currency bond returns. The predictive power of the extracted factors is both statistically significant and economically important. Our research has important implications for policymakers and pension fund managers.
Notes
1. The local currency government bonds are defined as bonds issued by the domestic government and denominated in local currency.
2. Ten EM Countries: Brazil, Mexico, Turkey, Russia, South Africa, Poland, Hungary, Indonesia, Malaysia and Thailand.
3. Hard currencies refer to globally tradable, reliable, and stable currencies such as the U.S. dollar, the yen, or the euro.
4. We also investigate the specifications with lagged values of the estimated factors and find that additional lags contain little information for future local currency returns.
5. This single factor is empirically grounded in the results of Cochrane (Citation2005) that take into account the presence of a common factor driving realized excess returns on U.S. government bonds.
6. One could also conduct an out-of-sample forecasting investigation in which the extracted factors are re-estimated recursively as a robustness check of our forecasting results. However our limited data set (e.g., thirty-five GDP data points) does not allow to run out-of-sample regression.
7. Konstantinou (Citation2005) provides some evidence in favor of the EH for the short end of the Polish interbank term structure.
8. Basci and Ekinci (Citation2005) find that inflation and default risks play important role to explain the observed bond premium in Turkey.
9. To save the space, we only provide detailed descriptions of the data sets and sources for Brazil (). Date sets and sources for Mexico, South Africa, and Turkey are available from the authors or in a working version of the article.