Understanding the reasons why the prices of sovereign bonds move the way they do has proven difficult. Studies of sovereign debt prices often focus on the risk of default. However, we have struggled to predict sovereign bond prices accurately using standard rational expectations models containing a single probability model for the future of an economy. Specifically, there appears to be an excess reward for buying bonds that risk the possibility of default.
Demian Pouzo of the University of California, Berkeley and Ignacio Presno of the Federal Reserve Bank of Boston show that reasonably high levels of risk aversion cannot generate high enough bond spreads to explain historical data. Having done so, they look to use ambiguity aversion to explain both empirical regularities and bond spread dynamics in the context of historical data.
Presno outlined this work in a poster session at the Ambiguity and Robustness in Macroeconomics and Finance conference. The authors designed a model in which indebted countries are sure about the distribution of their future incomes, but lenders are uncertain and recognize that they could have the wrong distribution in mind. In this way, lenders are ambiguous about the distribution of indebted country income. Pouzo and Presno offer several arguments for why this is the case: for instance, default-prone countries such as Argentina often deliberately interfere with or obfuscate data releases, official statistics are often restated with large revisions, and measurement error is always present.
In the paper, a country faces the decision of whether or not to default. If it defaults, then the nation is temporarily prevented from borrowing and also has direct impacts on its revenue. Unfortunately for lenders, if a country defaults, they have no recourse. Ordinarily, lenders would be able to know the likelihood of a country entering a state of the world in which it would prefer to default, and would price loans accordingly. In Pouzo and Presno, lenders aren’t sure about that probability distribution.
The authors calibrate their model to Argentinian data from 1983 until the end of 2001, when Argentina defaulted. The authors show that while previous attempts are unable to explain the high premium for facing default risk, their model using ambiguity aversion is able to match the premium. They also are more successful in explaining covariances between country production and default premia.
Interestingly, agents who have ambiguous aversion can often be equivalently rewritten as agents who simply have distorted, more pessimistic beliefs about the world. Pouzo and Presno generate a series of default predictions and compare them to what the risk of default agents are acting were true. On average, the quarterly probability of default for Argentina was measured at 0.74 percent, while their pessimistic agents act as if the quarterly probability of default was three times higher, around 2.25 percent.