USC Marshall

The objective of this research is to characterize the variation of real interest rates across 152 countries over available years of data since the floating FX rate period.

We assembled lending rates, inflation rates, population, GDP growth, budget deficits, current account deficits, and openness indicators. We will use a cross-sectional analysis by year, for 1975-2009. This cross-sectional approach allows us to get better statistics on the coefficients of interest and we can also study possible time patterns. For example, the opening of the capital markets has been a continuous process, and it may be that we can detect shifts in the importance of the various determinants of real rates as more economies become open.

The data-gathering process have been frustrating because certain variables turn out to be hard to find, even for developed countries. For this project, the definition of the variables has to be comparable across countries. For this reason we have limited ourselves to IFS, World Bank, the OECD, and Penn World Tables data.

We have also collected, geographic, geopolitical (wars, regimes, independence) as well as religion data for the countries in our sample. Further, we are continuing our efforts to collect good indicator data for "qualitative" institutional variables, such as governance, openness, and freedom.

At this stage we only have very preliminary results.

Macroeconomic fundamentals should affect exchange rates prominently but empirical confirmation of this proposition has been scarce. The current version of the paper employs a very general econometric specification to determine which scheduled announcements about macroeconomic fundamentals affect the DM/$ exchange rate over the 1980-1998 period in daily data.

Although both countries’ macroeconomic fundamentals significantly affect exchange rate dynamics, the U.S. announcements appear to be much more important. We analyze the market’s responses in part by examining their impact on money market returns by estimating simultaneous models for the U.S. and the German money market, together with the FX market. Several announcements exhibit significant and asymmetric responses. We find that FX rates and yields react to money supply announcements in the Volcker period but not later in the sample.

At the suggestion of a referee, we decided to expand the sample and include data for the Euro and the Yen. We plan to perform the same tests across the three currencies (which have different sample sizes and dates) and draw general conclusions about the behavior of FX rates, rather than study only the DM/$ rate. This extension is now possible because macro announcement data are now available for the Euro and the Yen through MMS.

We also find that higher than expected real economic growth simultaneously appreciates the currency and increases short term yields. These two effects taken together lead to the conclusion that the yield increases represent real yield increases.

This joint behavior is in contrast with the well-known uncovered interest parity (UIRP), which predicts that an increase in the national yields ought to be associated with a depreciation of the currency. Our findings add to the list of UIRP failures but in addition point to the likely reason for this failure, namely that yield movements are at least in part changes in real yields, and that the exchange rate reacts differently to real compared to nominal yield movements.

I derive a dynamic version of the Dornbusch “overshooting” model in which real yields and inflation vary stochastically, and the exchange rate (FX) delivers UIRP in expectations.

Tests using the model provide support for the UIRP proposition. Simulations show that the “disconnect” of FX rates from fundamentals as well as their very high volatility is a necessary consequence of UIRP when real yields are autocorrelated. I also show that FX rates display some predictability as required by the model. Finally, the model shows that the statistical patterns on which “carry trade” is based are consistent with equilibrium.

We show that even when fundamentals are i.i.d., a three-period OLG competitive model produces negatively autocorrelated expected returns, market risk premia, and prices; conventional models deliver zero autocorrelation for the same fundamentals. The negative autocorrelation we find arises out of market interactions among consumers that are ex-ante identical but differ by age. This negative autocorrelation is pervasive in our model and persists regardless of the autocorrelation structure of the “fundamentals”. Thus the model provides a simple alternative explanation for the observed autocorrelation of returns. A byproduct is low correlation between aggregate consumption and wealth, a puzzle for conventional models.

We examine the impact of budget deficits on both real yields and the current account and develop tests to fully assess the neutrality of government deficits. Earlier studies that only examine deficits’ effects on interest rate are valid only in closed economies, because even in large open economies, “crowding out” effects are likely to appear primarily in the current account. We should be able to detect even economically small effects of deficits, since our data include large variations in deficits. We find no evidence that current or expected budget deficits materially influence real yields. We do find evidence of a significant and positive transitory impact of deficits on the current account.

The findings of the paper are:

1. Budget deficits clearly do not cause real interest rates to rise. This result holds up under fairly extensive sensitivity analysis. In some partial specifications deficits appear significant in the estimates but then the results suggest that an increase in deficits would lower real interest rates. Since we are not aware of a theory that predicts such a relation we attribute it to missing conditioning variables in those partial regressions.

2. Budget deficits, even in raw $ form, are stationary series –I(0)– while the current account is nonstationary –I(1). Thus deficits cannot be contributing to the trend behavior of the current account.

3. At the same time, using an error-correction framework, we find that deficits are associated with increases in the current account; as much as 50% of an increase in deficits may be borrowed from abroad.