1. Corr" is the correlation between # of calls and (# of calls) X. . The first row is a regression with only the number of calls, without the second term. The number of calls is the log of the number plus one. It is a probit regression with robust standard errors. The first line in Table 14 shows a significant positive relationship between each program and the number of calls, as Davis and Taylor found. The remaining lines add regressors that are the number of calls with increasingly small powers, creating increasingly more collinearity. The probit regression results show clear evidence of SvOVB with respect to the house calls. Coefficients greatly expand with more collinearity, and the t-ratios do not decline. This finding is in accord with the fact that the police did not make house calls in some of the cases that were randomly assigned to the receive calls. It also suggests that these cases were different from REFERENCES Belsley;D;Regression Diagnostics: Identifying Influential Data and Sources of Collinearity,1980
2. Spurious regressions and near-multicollinearity, with an application to aid, policies and growth;J B Chatelain;Journal of Macroeconomics,2014
3. A proactive response to family violence: the results of a randomized experiment;R C Davis;Criminology,1997