还有些问题 补充的啊
帮忙帮忙~~呵呵
实在不知道该怎么翻哦
the estimated effects are very noisy里的noisy怎么翻呢?
No p-values for PSM are reported since the large number of caseworker fixed effects led to numerical problems in the bootstrap replications.
The third row therefore shows employment office fixed effects by including employment office dummies in the regression.
As an alternative to employment office fixed effects we can pursue an alternative approach to examine the potential degree of "selection-on-unobservables".
In the following rows we examine alternative specifications of the set of control variables X
In row 7 we add the ‘employability of the unemployed person’ as another control variable to the
main specification (of Table 1).
In addition, we include a bias reduction technique via regression on covariates within
matched pairs. More details on the estimator are given in the corresponding working paper (Behncke, Frolich, Lechner, 2008). Inference for the PSM estimator is based on the bootstrap by re-sampling caseworkers (together with all their clients) to account for possible dependencies among the unem-ployed counselled by the same caseworker. Following MacKinnon (2006), the t-statistic is boot-strapped and the bootstrap p-value is based on symmetric rejection regions of the t-statistic. There-fore, we report only p-values in the following tables, and suppress the standard errors to save space.