Government debt, learning and the term structure

Abstract

Does a change in the public׳s holdings of government debt affect the term structure of interest rates? Empirical analysis using a VAR model indicates that a rise in these holdings of the real debt-to-GDP ratio increases both the three-month and ten-year U.S. nominal yields in a statistically significant manner. The maturity composition of debt is also found to matter: innovations in holdings of long-term debt affect the term structure, while increases in short-term debt affect inflation expectations. These effects of changes in holdings of debt on the yield curve can be derived in a general equilibrium model in which the government issues exponentially-maturing riskless debt, financed by lump-sum taxes, and the optimizing agents are adaptive learners. On calibrating the average maturity of debt in the model to match that of U.S. Treasury debt since the 1980s, I find that positive innovations in government debt lead to increases in asset yields. This is because agents do not learn the principle of Ricardian equivalence exactly, and perceive increases in holdings of government bonds as a rise in their net wealth. Imposing rational expectations on the agents eliminates this channel, and changes in holdings of government debt have no effect on yields. The learning model also implies that as the real debt-to-GDP ratio increases, and the average maturity of debt becomes longer, the agents become less likely to learn that Ricardian equivalence holds.