Work and Welfare in the American States: Analyzing the Effects of the JOBS Program

This research seeks to determine whether the Job Opportunities and Basic Skills GOBS) program (established under the 1988 Family Support Act) was successful in reducing the number of welfare recipients among U.S. states for the period 1984 to 1996. Within the context of two theoretical perspectives-developmental and rational choice-we assess the impact of JOBS on AFDC participation rates using a pooled time-series design. At best, JOBS had a minimal effect. We estimate that states with higher proportions of their AFDC populations enrolled in JOBS programs had only slightly lower rates of participation in AFDC. Other forces were far more influential in reducing welfare participation. In particular, states with higher per capita income, lower female unemployment rates, lower poverty rates, and higher wages for low-paying jobs had the lowest welfare recipiency The AFDC participation rates of neighboring states had a significant effect, as well. The analysis showed that more generous AFDC benefits exerted strong upward pressure on a state's welfare rolls.

It required states to increase the proportion of able-bodied AFDC recipients that were engaged in work activities.
By the late 1980s, evidence was accumulating that welfare-to-work policies could be effective. The most extensive assessment of such programs came from the Manpower Demonstration Research Corporation (MDRC). During the 1980s, this group undertook a series of multi-state classical, random assignment field experiments to test the efficacy of work programs (Gueron and Pauly 1991). In brief, MDRC's evaluations concluded that the answer is clearly yes to the question of whether welfare-to-work programs produce positive results. While field experiments can generate valuable results, such research can be usefully supplemented by other studies that compare the effectiveness of work across multiple jurisdictions. The latter approach can take account of social, economic, and cultural forces that may influence the implementation of legislation such as the Family Support Act. Although a few studies have examined the operation of JOBS in one or several states (Mead 1997;Nathan 1993), almost no aggregate, comparative state studies assess the effectiveness of JOBS while controlling for other potentially important effects (however, see Mead 1995).
This research seeks to ascertain whether the JOBS program has affected welfare (AFDC) recipiency over a recent period of 13 years. We test for the influence of JOBS within a theoretical model that incorporates variables from two perspectives-developmental forces and rational choice. The dependent variable for all 50 states is an AFDC recipiency rate for the years 1984 through 1996. We test three JOBS-related measures: (1) percentage of welfare recipients active in JOBS; (2) the dollar per recipient amount spent by each state; and (3) a JOBS scope variable.

EXPLAINING INTERSTATE VARIATION IN WELFARE RECIPIENCY
Our framework for assessing the efficacy of the JOBS program draws on two somewhat different explanations, one that relies on a developmental interpretation and one based on individual self-interest. The work of Isaac andKelly (1981, 1982) provides the most thorough treatment of the "orthodox" development/modernization theory. An economic development interpretation views welfare expansion as a result of a sequential process. Basically, the demographic, social, and economic changes brought on by modernization create dislocations among certain groups. Those with fewer skills and less education get left behind as advanced polities become increasingly committed to an information-based global economy. Yet modernization clearly creates benefits for the larger society, primarily for our purposes in the form of economic surplus to support expanded social programs. Thus, the economic development thesis recognizes that advanced societies not only bring societal dislocations but also generate the resources to support policies to alleviate the ensuing distress. Gronbjerg (1977: 8-17) offers a refinement of the modernization approach. Relying on Shils (1975: 91-107), she argues that modernization leads to "mass societies" in which the values of the center spread to the periphery. In such polities, characterized by high levels of wealth, education, and voter participation, the well-being of the periphery becomes of greater concern to policymakers. These states spawn political cultures that support an expansion of economic and citizenship rights to include previously neglected and marginalized groups. One of the policy consequences, according to Gronbjerg (14), is more generous welfare payments, less punitive eligibility requirements, less stigma attached to public assistance, and therefore more people receiving such aid. Conversely, poor or less modern states may have a greater need for public relief, but they also are more inclined to take a restrictive and punitive approach to those who qualify. And these poorer states operate under greater fiscal constraints, leading to lower benefit levels aimed primarily at the "truly needy" and designed to discourage "abuse." In effect, in developed states, public assistance itself becomes a means of extending citizenship or social rights to a larger proportion of the population. During the period from 1960from -1970from , Gronbjerg (1977 found that welfare rolls rose predominantly in U.S. states with "high mass society status." Following Gronbjerg (1977), we end up with two somewhat different developmental interpretations for the changes in welfare rolls in the U.S. In backward states, absolute need is greater, increasing the pressure to expand the rolls. But, benefits may be niggardly. Among the more modern and wealthier states, there will be an expectation that people are entitled to assistance from the government. This presumption contributes to lenient or generous public aid policies among such jurisdictions. In effect, "both propositions suggest that welfare rolls will be high in both backward and mass society states-in the wealthy as well as the poor states" (Gronbjerg 1977: 15). Operationally, one might expect a curvilinear effect from state wealth. A developmental approach also stresses the importance of need or "demand" for welfare. Demand may arise from the consequences of family disorganization (especially among black families) and the subsequent growth of femaleheaded households, dependent children, and their resultant poverty status. Isaac and Kelly (1982) found urbanization to be the most useful surrogate for developmental effects. In the public mind, with some justification, welfare is seen as a large inner-city problem. Welfare dependency among ghetto minority mothers who have little exposure to the workplace may be the most intractable form of all. Some observers even consider the root cause of welfare dependency to be the values held by individuals who are isolated from mainstream America (Kaus 1992 chap. 7;Wilson 1987: 61). In the subsequent analysis, we discovered that both poverty and percentage black were useful predictors of AFDC recipiencyl Mead's (1997) research on the JOBS program among certain Wisconsin counties shows that those jurisdictions with a higher percentage of African Americans had more difficulty placing AFDC Presumably, more modem or progressive states should offer more generous benefits, leading to an expansion of the welfare rolls. But such states should also be more advanced economically, with more opportunities for marginal groups to gain employment. So, a person in such a state who might qualify for AFDC benefits has a twofold advantage. If she doesn't find a job, she may expect to receive more generous public assistance. Either way, a low-income mother may improve her position vis-a-vis someone in a less developed state. At some point, then, in an analysis of changes in welfare recipiency, one must pay attention to the process by which individuals make decisions about welfare and work.2 Economists have long engaged in such an enterprise, usually working with data on individual welfare recipients (Moffitt 1992). Using a microlevel theoretical perspective, these scholars employ a rational-choice (or labor market) framework to explain why poor women with children might choose welfare instead of employment. The relationship between welfare and work is thus seen in the context of individual self-interest, as a tradeoff between "income and leisure options" (Danziger et al. 1981: 979). The usual assumption, of course, is that if work doesn't pay enough, poor single mothers will opt for the dole. Considerable research shows that welfare programs do "create significant work disincentives" (Blau and Robins 1986;Danziger et al. 1981). Also, according to most studies, as AFDC benefits rise so do the welfare rolls (Tweedie 1994). An increase in earnings, on the other hand, reduces welfare participation (Moffitt 1992). Thus, if states reduce welfare benefits (increasing the cost of leisure), the act of remaining on AFDC becomes more expensive (Plant 1984). Recipients then might seek work instead. The analysis below will include each state's average monthly AFDC payment.
Economists and others also typically find that unemployment increases AFDC caseloads (Fording 1997;Moffitt 1983). If the economy lags, even wellmotivated recipients may encounter difficulty leaving AFDC. So, availability of jobs may be crucial, especially for low-income women. In short, according to rational choice theory, the key to understanding welfare recipiency is the relative recipients in jobs and had lower rates of case closure. In addition to percentage black and poverty, we tried a measure of urbanization-percentage of a state's population residing in cities of 250,000 or over, along with other measures of family disorganization. The large city measure was a good predictor of AFDC recipiency but declined in importance when we included it with the percentage black measure. Overall, two measures of need were most robust-percentage in poverty and percentage African American.
2 Some readers might question the use of microeconomic theory (rational choice) for research on political jurisdictions such as states. Economists Brennan and Buchanan (1980: 28-29) support the application of micro-level theories to collectivities by relying on an "as if' model of "government as an entity." In fact, Dunleavy (1991: 255) reports that neo-classical economists conventionally treat collective bodies as if they were individuals. Economists are not alone in this usage. For other studies applying individual-level explanations to the behavior of states or localities, see Peterson and Rom (1989), and especially Plotnick and Winters (1985).
attractiveness of work (labor market forces) compared to the income guarantee from AFDC (Ellwood 1994: 71).3 In addition to AFDC benefits, we should assess the effects of the labor market directly The whole purpose of JOBS, of course, was to move welfare parents from AFDC to self sufficiency, usually defined as work. So we need a wagerate variable that especially applies to women with few skills and little formal education. In the subsequent analysis, we use the average monthly wage for a person employed in the "food and drink" service area. This figure is one of the lowest wages among dozens of service wage rates listed in Department of Labor publications.
We include one additional policy-related effect. Abundant research shows that policies of individual states are influenced by the behavior of their neighbors (Walker 1969; more recently Berry and Berry 1992). So, the analysis encompasses such an effect, or perhaps in a statistical sense, a spatial lag term. States learn from each other but also engage in competition, as policymakers try to enhance the economic attractiveness of their jurisdictions (Peterson 1981: 17-38). As applied to welfare, most observers assume that states generally wish to avoid generous benefits, along with high taxes, as a way of promoting a favorable business climate. If welfare payments are too high, the state might become overly attractive to the poor from neighboring states. Findings on the presence of welfare magnet states are inconclusive.4 Still, Case et al. (1989) have shown that states do take account of their neighbors' welfare policy They find that a state's per capita welfare expenditure is positively and significantly affected by the expenditure of its neighbors. Thus, a state with generous benefits might experience pressure to reduce payment levels to avoid attracting welfare recipients from nearby jurisdictions.
Finally, we come to the JOBS program. Can we have any hope that such government interventions can overcome the abundant external forces influencing the tradeoff between welfare and work? The judgment on JOBS is more favorable than for its immediate predecessor the WIN program. Previously, we mentioned the Manpower Demonstration Research Corporation's (MDRC) use of field exper-iments to evaluate the JOBS program. The MDRC findings are complex, but some state welfare-to-work programs do help AFDC recipients gain employment. Generally, however, such programs were not successful in helping their graduates find better paying jobs. In addition to the field experiments from MDRC, single state studies corroborate the potential of government programs to reduce welfare caseloads (Schiller and Brasher 1993). According to Mead (1995), states that more fully implemented JOBS had a smaller rise in welfare participation even controlling for other influences (also Mead 1997).
Regardless of favorable assessments forJOBS, Glazer (1995: 24) comments on the program's "remarkably limited numerical goals" for moving people off welfare and into jobs. Part of the problem was implementation. Nathan (1993) identifies several difficulties. In particular, he mentions rising state budgets and the national recession as constraints limiting the §tates' response to JOBS. Many jurisdictions had a hard time finding the money necessary to match federal funds available under FSA. Because of such constraints, in an average month from 1991 to 1993, JOBS served only about 11 percent of AFDC parents (U.S. General Accounting Office 1994, 2). Overall, JOBS had only limited success. According to Nathan (1993, 109), JOBS did not spur leaders to alter welfare or to engage in far-reaching welfare reform. Still, Schiller and Brasher (1990) claim that even the threat of a work requirement may have a significant deterrent effect on potential recipients.
In sum, despite research showing that JOBS may achieve desirable results in some instances, we subscribe to the null hypothesis. The evidence is too strong that the real explanation for the changes in AFDC rates comes from a vector of powerful external forces. So, the analysis to follow will evaluate the effects of JOBS by incorporating control variables from the rational-choice model plus other measures that appear to represent the influence of socioeconomic development among the American states.

DATA AND METHODS
The dependent variable for the analysis to follow is welfare participation or recipiency-the total number of AFDC recipients divided by the state's population (minus those age 65 and over) for each state for each year from 1984 to 1996.5 AFDC participation actually declined slightly during the early 1980s. Then, in about 1989, the curve reflects an abrupt and steep rise. Following this  84 85 86 87 88 89 90 91 92 93 94 95 96 YEAR Rate = AFDC recipients per 100 non-elderly persons. dramatic increase, the welfare rolls began to level off and show an actual decline by 1994 (Figure).
In the initial stage of the analysis, we tested for three possible effects of the JOBS program: (1) JOBS expenditures; (2) JOBS participation; and (3) a JOBS scope variable, an additive measure representing the degree to which states had implemented up to five optional services to JOBS participants-on the job training, community work experience, work supplementation programs, other work experience, and the presence of a "contract" between agency and participant. Only the JOBS participation measure (the number of "countable" JOBS participants divided those who were "mandated" to participate) was useful in the equation to follow.6 Nationwide, in 1994, about 44 percent of AFDC adults were required to participate in JOBS (U.S. Congress 1996). The JOBS variable is lagged one year so that its effects do not operate simultaneously with AFDC recipiency.
We include control variables from the two theoretical frameworks discussed above-developmental and rational choice. Under developmental forces we include five controls (source for all variables is U.S. Bureau of Census unless otherwise noted): Per capita income.7 Per capita income squared-the square of per capita income. We include this term, following Gronbjerg (1977), to test specifically for a curvilinear effect from our primary measure of development.8 Poverty-the percentage of families below the poverty level. Percentage black-the percentage of a state's population that is African-American.
Births to unmarried women-the percentage of all live births to unmarried women.
Female unemployment (rate)this variable gauges the pressure arising from fluctuations in the economy that may prompt women to consider AFDC as an alternative to employment. The rational choice approach typically tries to model the tradeoff between employment and welfare benefits. Economists routinely include the unemployment rate in their analyses of welfare participation. But we place that measure in the developmental category So, we end up with only the relationship between AFDC benefits and wages to represent rational choice: AFDC benefit-the average monthly AFDC family payment in hundred dollars.
Wages-the average monthly wage for the "food and drink services" category in hundred dollars (U.S. Bureau of Labor Statistics).
We also tried two policy variables that could affect AFDC recipiency, but without success. They included, first, a measure of child care (expenditures per AFDC recipient for child care services). The Family Support Act provided child care assistance to AFDC recipients engaged in some kind of work or JOBS activ- 7 We also included gross state product and the square of that term in the following equation. It performed less well than the per capita income measure. 8 Following Aiken and West (1991: 32-36), we center both the base and quadratic term to avoid or minimize multicollinearity We also tried adding squared terms for poverty and percentage black; neither was useful in the subsequent equation.
ity. Second, we included a "waivers" variable to capture differences across states in securing welfare waivers from the national government (coded 0 until a state obtained any JOBS waiver and 1 for each separate waiver in subsequent years (range 0-3)). Neither of these two measures were useful and contributed almost nothing to explained variance. We dropped them from the analysis. The final equation also includes a spatially lagged AFDC participation rate (recipients divided by the population under age 65). This variable, which takes account of conditions in neighboring states, is the average recipiency rate for adjoining states.9 (For Alaska and Hawaii the "adjoining" states are Washington and California, respectively.) As noted above, some scholars report that states set their AFDC benefits partly in response to those of their neighbors (Tweedie 1994).10 The analysis employs a time-series cross-section (TSCS) design (often called pooled time series or pooled cross section) for the years 1984-1996. This design is ideally suited for the assessment of changes over time in AFDC participation. It is capable of analyzing multiple units (states) for multiple years. JOBS data are available beginning only in 1990, yet a larger number of years can provide both a larger number of observations and a more stable time series. In some respects, one might consider the effects of JOBS as a policy intervention, since it is coded zero for all years before 1990. With 50 states and 13 years, the N for the equation is 650.
Despite its advantages, a TSCS design is subject to certain statistical problems, namely autocorrelation and heteroscedasticity (Sayrs 1989;Stimson 1985). Our design is cross-sectionally dominant; the number of states exceeds the number of years. According to Stimson (1985: 926), "Cross-sectional dominance simultaneously minimizes the threat of autocorrelated error and maximizes the possibility of bias (or inefficiency) from the specification of unit effects." Thus in our analysis, our most serious threat is biased estimators rather than autocorrelation. Still, we tested for autoregressive effects and found no evidence of contamination from correlated error terms. Therefore, our regression model is estimated with ordinary least squares. Following Beck and Katz (1995, 645), we determine statistical significance by calculating panel corrected standard errors. Table 1 displays the time-series regression model estimating AFDC recipiency for all 50 states for the years 1984 through 1996. All the proxies for devel-opmental forces are positive and reasonably strong. First, we might examine a prime measure of state modernization-per capita income. Both that measure and its square are useful regressors. The positive sign for income (b = 1.36) reveals that states with more resources tend to have more people on welfare. But this is not the whole story. The negative squared income term (b = -.88) indicates a curvilinear effect. In general, this finding partly confirms Gronbjerg's (1977) mass society hypothesis. State wealth apparently engenders a political culture that spawns more generous welfare programs, encouraging more participation and dispensing higher benefits.'2 As states reach a certain level of affluence, however, the recipiency rate levels off and begins to decline.

FINDINGS
The positive sign for poverty shows that need also stimulates higher levels of welfare participation. As the poverty rate rises by 1 percentage point, AFDC recipiency should also increase about .11 percentage points. A similar effect occurs for our prime manifestation of social disorganization-percentage of births to unmarried women. Here a 1 percentage point growth in unwed motherhood produces about a 10 increase in the participation rate. Welfare recipiency is only moderately sensitive to percentage black. A 1 percentage point difference among the states in this variable should produce about a .03 difference in the dependent variable. However, the unemployment rate among women is one of the stronger developmental effects. As female unemployment goes up by 1 percentage point, the proportion of the population on welfare also should climb about .30 percentage points. For example, the number of AFDC recipients for the average state for this 13-year period is 238,150 while the mean for population (minus age 65 and above) is 4,378,352, indicating a 5.44 percent participation rate. Assume female unemployment increases by 1 percentage point, driving AFDC participation up by .30 to 5.74. This results in about 13,266 additional AFDC recipients (.00303 * 4,378,352 = 13,266). In all, the level of economic and social development among states produces the expected effect-higher levels of modernization provide resources that enable states to respond more generously to those in need. At the same time, modernization produces certain dysfunctional consequences. The result: greater need for public assistance and more welfare recipients. Now we turn to those measures representing the tradeoff between work and welfare. The relationship between benefits, wages, and the AFDC rate is consistent with rational choice theory. The amount of the monthly AFDC check has a powerful stimulative effect on participation. The coefficient of .91 indicates that a $100 increase in the amount of the family benefit is associated with an estimated .91 percentage point rise in the AFDC recipient rate. The effect of low-- .60 N = 650 *p < .05, one-tailed test ** = p < .01, one-tailed test wage employment is not quite so strong. A $100 average improvement in earnings should produce a .52 decrease in AFDC participation. This contrasting effect of work and welfare suggests that being on the AFDC rolls carries benefits beyond just the monthly payment itself. Observers have long pointed to the ancillary advantages of welfare, beyond the mere money it provides. We are unable to test for other influences directly. But the larger effect for the AFDC dollar compared to a dollar of increased earnings is entirely consistent with the argument that costs increase, especially for child care and health care, when women move from welfare to employment.
The results from Table 1 also show that state welfare policy responds to other external factors, especially interstate competition (Soule and Zylan 1997). The coefficient for the spatial lag term is positive and statistically significant. A I percentage point average increase in AFDC recipients (per population less than age 65) in adjoining states increases AFDC participation in the state they all adjoin by .19 percentage point. This variable could be capturing regional effects, or perhaps states try to avoid becoming "welfare magnets" by adopting benefit levels similar to those of their neighbors. Now we come to the JOBS measure. Its coefficient of -.01 shows a modest effect, one that falls just short of the .05 level of statistical significance (p = .10; a t ratio of -1.65 is needed for the .05 level; one-tailed test). Still, this means that _  Table 1 is reasonably robust (R2 = .60) with an acceptably low standard error. Per capita income and its squared term are correlated at .99, and each has a very high variance inflation factor (Fox 1991). Yet, both variables are statistically significant, and the model is stable when we add and drop each of them successively from the equation.
We might include one more example to illustrate the estimated effects of the different independent variables for a single state. Minnesota provides a good case study Its population is close to the national average for states, and Minnesota fits the model well. Table 2 shows the effects of changes experienced from 1995 to 1996 to illustrate these dynamics. Applying the model to actual data indicates an increase of 4,600 AFDC recipients owing to higher per capita income. An increase of 0.6 percentage points in the poverty rate is associated with 2,786 additional recipients. Births to unmarried mothers increased by half a percentage point, swelling the rolls by 2,118. The one-tenth of a point increase in unemployment translates to 1,234 new recipients. A slight increase in the proportion of African Americans works out to 66 additional recipients.
The remaining variables are associated with decreases in AFDC participation. The average monthly AFDC check in Minnesota dropped by $44 for 1996. This change is associated with a decrease of 16,235 recipients. Meanwhile wages increased by 26 dollars a month, which would lead us to expect a decrease of 5,527 more people. An additional decline of 3,310 is expected due to neighbor effects. This leaves the JOBS measure. Minnesota increased its JOBS participation measure by 2 percentage points from 1995 to 1996. The mean for all states for all years (24 percent) is about the same as Minnesota's during these two years (19.17 percent and 21.16 percent, respectively). As Table 2 indicates, Minnesota's increase in JOBS participation is associated with a decrease of 652 AFDC recipients. This is not a large effect, of course, when compared with the other components of the model.

CONCLUSION
In recent years, until about the mid-1990s, the nation's welfare rolls had risen abruptly Under growing public pressure, U.S. policymakers sought various ways of slowing what seemed an inexorable trend. Determined to keep welfare from becoming a way of life, lawmakers searched for ways to encourage poor women to eschew welfare in favor of gainful employment. Many observers hailed the 1988 Family Assistance Act, which established the JOBS program, as a valuable step in the effort to move recipients from welfare to work. Field studies revealed that JOBS did achieve its desired effect under certain conditions in certain states. But few attempts have assessed the efficacy of the program in a broader context over time.
The research reported here tested the effectiveness of JOBS among all states over a 13-year period. Its impact, operationalized as a participation measure, was modest, at best. As more and more AFDC recipients became involved in JOBS activities, the proportion of the state's nonelderly population on AFDC declined ever so slightly The time-series regression equation showed that a 1 percentage point increase in JOBS participation, on average, should have reduced the proportion of nonelderly persons on welfare by only about .01 percentage points.
From the outset, we had assumed that the external forces influencing welfare participation would overwhelm the effects of such a limited policy intervention as JOBS. Indeed, the analysis showed that interstate variation over time resulted primarily from two major forces-the degree of state modernization and individual self-interest on the part of low-income mothers. Several proxies for state development or modernization were useful in estimating changes in welfare recipiency over time. For example, if female unemployment went up, so did the welfare numbers. We discovered a curvilinear effect for state resources; both affluent states and those with more poverty had higher AFDC participation rates. States with higher proportions of African Americans also had more people on welfare. In addition, the analysis showed that welfare policy in one state has a positive spillover effect on neighboring states.
The results also are consistent with a rational choice interpretation. More generous AFDC payments had a strong positive impact on welfare recipiency For each additional 100 dollars of monthly benefits a state might pay, welfare rolls should go up almost 1 percentage point. Wages had the opposite effect. Higher pay for low-income employment tended to reduce AFDC rolls.
Although our results suggest that the JOBS work requirement was largely ineffectual, we should acknowledge the apparent success of the 1996 welfare reform legislation. TANF, above all, is predicated on the efficacy of work. And, the last few years have brought dramatic changes in public assistance. The nation's strong economy coupled with welfare reform has brought an extraordinary transformation in the welfare rolls. From January 1994 to September 1997, the welfare caseload dropped by a remarkable 30 percent (U.S. General Accounting Office 1998). Despite this remarkable success, we might offer a few brief observations on what may lie ahead.
First, with the recent dramatic decline in welfare participation, the easiest cases undoubtedly have been resolved. As the rolls downsize further, the task of moving clients from welfare to work clearly becomes much harder. Indeed, some observers estimate that up to one-third of welfare recipients are basically unemployable (Milbank 1997). Second, of those former recipients who do find jobs, many will require substantial and continuous support to remain employed. And, even with employment, most former welfare recipients will not escape poverty (Glazer 1995: 29). Jencks (1997), among others, also argues that many former welfare mothers with new jobs will be worse off financially than before because of child care, health, and transportation costs. Third, some local officials worry about a potential job shortage for former welfare clients, especially if the nation's economy slows significantly. Localities may have to commit ever more resources to job search assistance, training, even public employment programs.
Finally, we might reflect on the origins of these recent large-scale changes in public aid. These developments would seem to confirm Teles's (1996: 164-66) basic theme. Writing before the recent changes in law, he argued that a fundamental transformation was unlikely unless or until the nation's elites could reconcile their ideological differences over welfare. The public's view has been unequivocal; every competent adult should work, and government policy should further that end. The barrier to this objective, according to Teles, has been the disparity between public preferences and elite attitudes. Almost from the beginning, Teles contends that welfare, loaded down with moral baggage, has been the source of major ideological divisions among political elites. Especially in recent years, certain leaders have used welfare politics as a wedge issue to wage what he calls a "vicious cultural war" (p. vii). This failure of elites to agree over the necessary changes in AFDC had produced an impasse, ensuring that public policy did not reflect the country's views on public aid. With the election of a Republican majority in the House in 1994, the stage was set for a convergence between elite and public attitudes. The president was on record favoring the ending of welfare as we know it. And the House majority was committed both to imposing stringent work requirements and handing policy implementation over to the states. With the 1996 law, the long-standing dissensus between public and elite attitudes had finally been resolved.