FINANCIAL MARKETS’ ASSESSMENT: Term structure-based EMU 6


While the objective of proxying for global trends in credit markets is laudable, the diagnostics of model performance in- and out-of-sample overstate the model’s reliability. The major difficulty is that the independent and dependent variables, while stationary, are highly persistent. The high reported R2’s (ranging from 61% to 83% on 1300 daily observations over 1988-92) are therefore to be expected, while good out-of-sample forecasting performance over 1993 is also to be expected for highly persistent variables. Dickey-Fuller tests for stationarity of in-sample residuals are also run — the standard Engle-Granger test for cointegrated variables — but since the variables are stationary the relevance of this test is not apparent.

It does appear that the global variables used in assessing Morgan’s non-EMU scenario contain some information for post-1992 swap spreads. In particular, several countries experienced temporary declines in swap spreads relative to Germany over 1992-94, and again after 1995 (see Figure 3) that are in part correlated with Morgan s estimated non-EMU swap spreads.14 Overall, the regression approach appears somewhat preferable to just using average spreads over some interval as an estimate of the non-EMU scenario.15 Nevertheless, it must be recognized that the non-EMU estimate potentially contains considerable structural instabilities that could affect EMU probability inferences. People need payday loans like speedy cash because they all need money at some point. If you arrived at this point, you probably need a loan with low APR and fair terms that you can be sure about. You are welcome to visit website to get a loan like this, and we can guarantee you will not be sorry about this decision.

A further issue emphasized in Favero et al (1997), is that Morgan’s use of the post-1998 section of 10-year swap spreads ignores the possibility of delayed EMU entry. For instance, a 10-year Italian-German swap spread as of January 1, 1995 incorporates instantaneous forward spread predictions of future interest differentials over 1999 through 2005. Morgan’s EMU probability estimates should therefore be loosely interpreted as the probability of EMU entry on or some time after 1999.

Central bank reaction functions

Favero, Giavazzi, Iacone, and Tabellini (1997) use macroeconomic variables instead of financial variables in their forecast of the non-EMU scenario. Quarterly 3-month Euro-lira rates are regressed over 1987:I through 1996:II upon a lagged value, the inflation and growth gaps relative to Germany, a growth gap shift variable for the impact of German reunification, current and lagged log $/DM exchange rates, current and lagged 3-month Euro-DM interest rates, and dummy variable that eliminates the impact of the 1992:IV interest rate outlier.

The regression results are interpreted as representing the Bank of Italy’s reaction function to macroeconomic fundamentals over 1987-96. This regression, combined with current forecasts of the regressors,18 is used to forecast future interest rates, and this forecast is used as the non-EMU scenario. The probabilities of Italy joining Germany in a currency union by 1999 or by 2001 are then inferred using (7) above and instantaneous forward rates computed from Euro-lira rates and lira swap rates via Svensson’s extended Nelson-Siegel approach.

Favero et al emphasize that the decline in Italian forward rate spreads after 1996 was primarily attributable to improved fundamentals (e.g., falling Italian inflation) shifting the non-EMU forecast. The residual explanation of shifting EMU probabilities played less of a role. Their estimates also showed a relatively low probability over December 1995 through March 1997 of Italy joining the EMU in 1999, but a higher probability of it joining by 2002. The difference apparently reflects the sharply inverted term structure of forward rate spreads evident in Figure 3.