1. The calendar year dummies reflect growth not accounted for by the explanatory variables. The quarterly dummies do not play a role in the simulations because the number of quarters in the full period, and each subperiod, is a multiple of four.
2. Each of the two recessionary periods includes approximately one year following the end of the official recession, during which the economy continues to be sluggish.
3. The vertical distance between a point on the labor market line corresponding to a specific quarter and the point corresponding to 1989.4 is the estimated contribution of labor market strength to the caseload for that quarter relative to labor market strength for 1989.4. An analogous interpretation applies to the IRCA series. The IRCA series begins in 1988.1, the first quarter in which there were legalizations under IRCA.
4. For the Basic caseload, growth not accounted for is 1.8 percent, and for the UP caseload in the 19 states it is 6.1 percent. Because the UP caseload is a very small share of the total caseload, total caseload growth not accounted for is only slightly above the 1.8 percent figure -- very close to 2.0.
5. "National" estimates refer only to the 19 states with UP programs for the entire sample period.
6. Personal correspondence, October 1996.
7. Personal correspondence, October 22, 1996.
8. Personal correspondence, October 22, 1996.
9. Personal correspondence, November 4, 1996.
10. Wiseman (1996) also concludes that economic growth has had a greater impact on the national caseload than on the Wisconsin case load in the last two years.
11. Personal correspondence, October 1996.
12. This is the approach taken by Cromwell et al. (1986), although the participation measure was Medicare AFDC enrollees, not the AFDC caseload.
13. Steven Thompson has developed a separate time-series model for Prince George's County because the behavior of the caseload in that county is differs substantially from the behavior of the caseload in the rest of the state.