Working Paper: CEPR ID: DP2591
Authors: Carlo A. Favero; Francesco Giavazzi
Abstract: This paper applies a full-information technique to test for the presence of contagion across the money markets of ERM members. We show that whenever it is possible to estimate a model for interdependence, a test for contagion based o a full information technique is more powerful. We test for the presence of contagion after having identified episodes of country-specific shocks, whose effect on other European markets is significantly different from those predictable from the estimated channels of interdependence. Using data on three-month interest rate spreads on German rates for seven countries over the period 1988?1992, we are unable to reject the null of contagion. Our evidence suggests that contagion within the ERM was a general phenomenon not limited to a subset of weaker countries, the exception in our sample being France. Our result are mute as to the question of what lies behind these episodes of contagion; they show, however, that it is not always that one only detects contagion when one applies poor statistical techniques.
Keywords: contagion; ERM; interdependence
JEL Codes: E30; F30; F40
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| contagion within the ERM (F44) | increase in market correlations during crises (G01) |
| country-specific shocks (F69) | effects on other markets (F69) |
| specific shocks (E32) | significant deviations from predicted interdependence outcomes (C62) |
| robust statistical techniques (C38) | detection of contagion (F44) |
| full-information technique (D83) | more powerful than other methods (C90) |