The sample for the flow analysis is defined primarily by the starting date of the LBP assignment. In particular, the sample consists of each case with a first assignment to the LBP starting between November 1994 and April 1995 inclusive. The following discussion covers several issues in the sample design.
1. Sample Window
The flow analysis centers on cases with LBP assignments starting during a certain time period, or sample window. Our choice of the length and timing of this sample window was guided by three considerations. First, we wanted a window long enough to yield a sample that would be large enough to support subgroup analyses, such as a comparison of characteristics across cases that took different paths through the LBP. Second, we wanted a window short enough to make the data collection process efficient and economical given our objective to extract: (1) monthly case benefit histories for the 12-month LBP term and several months after this term from the IABC Case Master File, and (2) demographic characteristics measured near the starting date of the LBP from the IABC Individual Master File. The longer the window, the more versions of these files must be accessed to obtain these data. Third, we faced calendar constraints on the first and last months of the window. In particular, we wanted the first month to be late enough in LBP operation that the program and its data systems were fully implemented for the earliest assignments in our sample. At the other end, we wanted the last month to be early enough in time that we could follow each case for the full 12-month LBP period and three subsequent months.
Taking these factors into account, we selected a six-month sample window beginning in November 1994 and ending in April 1995. This window allowed us to obtain data from just the June 1995 and March 1996 versions of the Case Master File.(2) It also allowed us to use only two versions of the Individual Master File (December 1994 and March 1995) to obtain data within one month of the LBP starting date.
2. Unit of Analysis
The primary unit of analysis for the flow analysis is the case, or welfare benefit unit. Although assignments to the LBP are made for an individual, the LBP assignment affects the entire case to which that individual belongs, as cash assistance to the case is reduced and then terminated.(3) In light of this, we have structured the flow analysis around the case, reporting on the analysis in terms of cases assigned to the LBP, or LBP cases.
To obtain a unique observation per case in the sample for the flow analysis, it was necessary to address the possibility of multiple assignments to the LBP. Multiple assignments can happen for two reasons: (1) repeat assignments, in which a case is reassigned to the LBP after the initial assignment has ended, and (2) concurrent assignments, in which two individuals in a given case, such as a husband and wife, are assigned to the LBP at the same time. We eliminated multiple records per case as follows. First, we restricted our sample to initial assignments to the LBP; that is, we included only cases having an LBP assignment in the sample window and having no LBP assignment prior to that one.(4) This eliminated multiple observations arising from repeat assignments. Second, to eliminate multiple observations arising from concurrent assignments, we selected the single record associated with the case-name person.(5)
3. Resulting Sample
Table III.1 presents the sample for the flow analysis. The total number of LBP cases in the sample is 4,224. As mentioned above, the sample includes one observation per case for each case with a first assignment to the LBP starting in the six-month period from November 1994 through April 1995.(6) During this period, an average of 704 assignments started each month, with a low of 551 in December 1994 and a high of 841 in February 1995.
The 4,224 first assignments to the LBP made during the six-month sample window represent approximately 22 percent of LBP assignments made in the twenty-seven month period from April 1994 (the first month of LBP assignments) through May 1996. The average number of cases starting the LBP per month from April 1994 through May 1996 was approximately 644. This is 60 cases fewer than the average in the six-month sample window of November 1994 through April 1995. This lower average reflects a reduction over time in the number of cases assigned to the LBP that parallels an overall reduction over time in the FIP caseload.