Convertibility Risk, Default Risk, and the Mexdollar Anomaly: A Reply to Gruben and Welch
回应Gruben和Welch对作者可兑换性风险假说的质疑，通过重新估计模型发现，预期比索贬值与美元化比率仍呈负相关，而不良贷款比率的影响不显著，支持了原假说。
In Rogers (1992a,b) I put forth the convertibility risk hypothesis to explain the seemingly anomalous negative relationship, depicted in Figure 1, between the expected rate of peso depreciation (forward discount) and the dollarization ratio-the ratio of Mexdollars (dollar-denominated demand deposits held in Mexican banks) to peso deposits. Estimation using either VARs (Rogers 1992b) or cointegration equations, single-equation OLS estimates, and error-correction models (Rogers 1992a) produces a statistically significant, negative relationship between the dollarization ratio and expected depreciation, which I took as evidence in favor of the convertibility risk hypothesis. Gruben and Welch (1996) examine the effect of deteriorating bank loan quality, and find (i) a negative relationship between nonperforming loans and the dollarization ratio and (ii) the conventional positive relationship between expected peso depreciation and dollarization. The first result suggests an additional factor influencing money demand in Mexico, the second is evidence against my convertibility risk hypothesis. I find some evidence for their first result, but the preponderance of evidence runs counter to the second result. Using data on the nonperforming loans ratio provided by Gruben and Welch, I reestimated (5.2M) from my paper starting in 1979 to conform with the GrubenWelch data set. The results, displayed in the first column of Table 1, indicate that there is a negative and significant relationship between changes in the dollarization ratio, l\MM, and lagged changes in expected peso depreciation, l\EM (lM is the three-month Mexican Treasury bill rate). Adding four lagged changes of the nonperforming loans ratio, /\(h, reveals that the coefficients on l\EM remain negative and significant, while the coefficients on /\(O are insignificant (column 2). I then reestimated models for the numerator and denominator of the dollarization ratio sepa-