CEA alum Ed Lazear points out the problems with the government's "jobs created" measure derived from the survey of funds recipients, available on www.recovery.gov. First, it's self-reported. But we also measure unemployment using self-reports, from individual workers in the Current Population Survey.
A second point is that the self-reported measure is a gross rather than net measure. If a job was "created" by poaching an employee from another firm or agency, the net effect is zero even when its gross effect, as would be reported by the poaching agency, is positive.
Lazear also cites the lagging nature of employment, which is certainly true. But he doesn't present a solution to the difficult problem of trying to assess the marginal effects of the stimulus package, a task that requires a counterfactual. Wouldn't it be nice if we'd had two economies rather than one, and to compare one with the stimulus to one without? Pre-stimulus forecasts are not a valid baseline because they are of poor quality. (Nobody knows how to forecast the length of a recession, let alone an impending one.)