The percentage of the program and control group working in each quarter was calculated as the average of the percentages in each of the three months making up the quarter. [...] The third column of the Table 3 shows the impact of the program, measured as the difference between the program and control group percentages. [...] Figure 1 shows the empirical hazards for the program and control groups in each of the 51 months for which hazard rates can be calculated; Figure 2 shows the difference between the hazard rate functions of the program and control groups.9 The greater speed with which the program group found full-time employment can be seen by noting that the height of the empirical hazard function for the program [...] Any comparison of means or proportions that is limited to program and control group members who worked full time will not be “comparing comparables.” The analysis of the impact of SSP on those who qualified for the supplement — the “effect of the treatment on the treated” — must therefore account non-experimentally for the differences between the two groups. [...] When we assume an exponential pattern of job loss after the censoring point, the mean durations of the first spell of full-time employment in the program group is 21.9 months while the mean for the control group is 20.3 months.