338 HANSON,MCLANAHAN,AND THOMSON prior to divorce are estimated from regression models in which resources at the first survey are treated as the dependent variable.Similarly,resource differences between divorcing and nondivorcing families after divorce are based on regres- sion models in which resources at the second survey are the dependent variable. Both sets of models include control variables discussed earlier,all measured at the first survey.s We present the results for pre-and postdivorce differences together in graphical form.Full model parameters and statistics are available on request. In describing our results,we will use the word "loss"or "gain"to describe changes in economic,parental,and community resources associated with the divorce process.Unless otherwise noted,we are speaking of changes in resources relative to that of families with continuously married parents(the"difference"in the "difference").Thus a loss or gain need not reflect an absolute change in resources.For example,if the absolute level of family resources is stable for the divorced between 1987-1988 and 1992-1994,but rises for the continuously married,then the divorced experienced a relative loss of resources during the two time periods. Adjusted Differences in Resources The graphs in Figure I show the relationship between divorce and economic resources.Graph 1.1 shows the difference in household income between families who did and did not divorce.The bar on the left-hand side of the graph shows the income difference prior to divorce,while the bar on the right-hand side shows the income difference after divorce.Couples who divorce between waves I and 2 had about the same income in 1987 as families that remain married.Recall that prior to controlling for differences in background characteristics,the former had about $7,000 less income than the latter(Table 1,panel 2,columns 1 and 3).Differences in parental age and education are primarily responsible for the gross association between 1987-1988 household income and subsequent divorce(not shown). The bar on the right-hand side of graph 1.1 shows that income differences increase after divorce.Mothers and children experiencing divorce or separation by 1992-1994 have about $20,000 less income than couples who remain married. Both the postdivorce income difference and the change in the difference between 1987-1988 and 1992-1994 are statistically significant. Graph 1.2 simulates how household income differences change as parents move closer to and further from divorce.According to the graph,income differences are relatively small and stable prior to divorce,they increase sharply immediately after divorce,and they gradually decline in the postdivorce period. 8 To examine whether changes in resources associated with divorce are statistically significant,we also estimated models that included controls for the level of resources at Wave 1 ("static-score" model). 9 Because income is measured for the calendar year prior to the survey year,we do not observe the income of respondents at the time of separation even for many parents who are very close in time to separating at Wave 1.This could inhibit our ability to see whether economic resources decline as parents move closer in time to divorce.A more fine-grained analysis revealed that those who wereprior to divorce are estimated from regression models in which resources at the first survey are treated as the dependent variable. Similarly, resource differences between divorcing and nondivorcing families after divorce are based on regres￾sion models in which resources at the second survey are the dependent variable. Both sets of models include control variables discussed earlier, all measured at the first survey.8 We present the results for pre- and postdivorce differences together in graphical form. Full model parameters and statistics are available on request. In describing our results, we will use the word ‘‘loss’’ or ‘‘gain’’ to describe changes in economic, parental, and community resources associated with the divorce process. Unless otherwise noted, we are speaking of changes in resources relative to that of families with continuously married parents (the ‘‘difference’’ in the ‘‘difference’’). Thus a loss or gain need not reflect an absolute change in resources. For example, if the absolute level of family resources is stable for the divorced between 1987–1988 and 1992–1994, but rises for the continuously married, then the divorced experienced a relative loss of resources during the two time periods. Adjusted Differences in Resources The graphs in Figure 1 show the relationship between divorce and economic resources. Graph 1.1 shows the difference in household income between families who did and did not divorce. The bar on the left-hand side of the graph shows the income difference prior to divorce, while the bar on the right-hand side shows the income difference after divorce. Couples who divorce between waves 1 and 2 had about the same income in 1987 as families that remain married. Recall that prior to controlling for differences in background characteristics, the former had about $7,000 less income than the latter (Table 1, panel 2, columns 1 and 3). Differences in parental age and education are primarily responsible for the gross association between 1987–1988 household income and subsequent divorce (not shown). The bar on the right-hand side of graph 1.1 shows that income differences increase after divorce. Mothers and children experiencing divorce or separation by 1992–1994 have about $20,000 less income than couples who remain married. Both the postdivorce income difference and the change in the difference between 1987–1988 and 1992–1994 are statistically significant. Graph 1.2 simulates how household income differences change as parents move closer to and further from divorce. According to the graph, income differences are relatively small and stable prior to divorce, they increase sharply immediately after divorce, and they gradually decline in the postdivorce period.9 8 To examine whether changes in resources associated with divorce are statistically significant, we also estimated models that included controls for the level of resources at Wave 1 (‘‘static-score’’ model). 9 Because income is measured for the calendar year prior to the survey year, we do not observe the income of respondents at the time of separation even for many parents who are very close in time to separating at Wave 1. This could inhibit our ability to see whether economic resources decline as parents move closer in time to divorce. A more fine-grained analysis revealed that those who were 338 HANSON, MCLANAHAN, AND THOMSON SSR625 @xyserv1/disk4/CLS_jrnlkz/GRP_ssrj/JOB_ssrj27-3/DIV_231a04 debb