In particular, not only the estimated coefficients of several important control variables (e.g., School enrolment, Bank deposits, %∆bank deposits, Share 21+, Share 65+, Females, Blacks, and Immigrants) are substantially different in the mode regression, but more importantly the estimate of the coefficient of ln(rc), one of the main regressors of interest, is now negative and statistically significant. Although at first sight it may be surprising to find that a regressor has coefficients with opposite signs in two conditional measures of central tendency, this is indeed entirely possible as a result of heterocliticity.