a1 New York University
The stakes of political conflict involve contending values and issue definitions as well as policy. Welfare reform was the most important change in American domestic policy since civil rights. Its significance hinges crucially on how participants understood the issue, but existing research fails to resolve what their perceptions were. Most accounts suggest that welfare reform was an ideological contest concerning the proper scope of government, but there are other views. This study gauges the welfare agenda rigorously by coding speakers in congressional hearings on the basis of how they framed the issue and the position they took on it during the six chief episodes of welfare reform that occurred between 1962 and 1996. The reform efforts aroused four distinct divisions. Over time, positions moved rightward, but more important, the dominant issue changed: The ideological debate about government was overtaken by a more practical debate about how to manage welfare. This is the first study to track the substantive meaning of any issue in Congress over an extended period of time using hearing witnesses and a preset analytic scheme.
Lawrence M. Mead is a professor of politics and public policy at New York University. He has written several books on antipoverty policy and welfare reform, including Beyond Entitlement (Free Press, 1986), The New Politics of Poverty (Basic Books, 1992), and Government Matters: Welfare Reform in Wisconsin (Princeton University Press, 2004). He can be reached at LMM1@nyu.edu.
I am indebted most of all to my coders, Kendal Elliott, Ian Gold, and Frank Ortiz, who worked for seven years to produce the data reported here. I acknowledge funding from the Achelis and Bodman Foundations, Earhart Foundation, JM Foundation, Robert Wood Johnson Foundation, John M. Olin Foundation, and Randolph Foundation. I acknowledge helpful comments on earlier drafts of this paper from Scott Allard, Paul Burstein, Joe Dolan, Ron Haskins, Craig Ramsey, Michael Reinhard, Michael Wiseman, several discussants at conference presentations, and several anonymous journal reviewers.
List of Figures
Figure 1 Issues in Welfare Politics
Figure 2 Incidence of Issue Types, by Stage of Welfare Reform, 1962–96
Figure 3 Incidence of Issue Types as Lead Issue, by Stage of Welfare Reform, 1962–96
Figure 4 Liberalism of Witnesses Citing Issue Types, by Stage of Welfare Reform, 1962–96Note. Bars show percent of witnesses citing each welfare issue who were coded left on that issue.
Figure 5 Position Taken on Lead Issue by Witnesses, by Stage of Welfare Reform, 1962–96
Figure 6 Left and Right Witnesses Citing Main Issue Types, by Stage of Welfare Reform, 1962–96
List of Tables
Table 1 Intercoder Agreement
Table 2 Witnesses in Welfare Reform Hearings, by Group Affiliation and Stage
Table 3 Percent of Witnesses Coded as Liberal, by Group and Stage
Table 4 Witnesses Raising Ideological and Paternalist Issues, Overall and by Groups
Politics is about more than public policy. The contestants on any issue seek to change what government does but they also typically perceive larger stakes. They have competing visions of what the policy problem is and different political values. It is these wider concerns that drive them into the policy contest in the first place. Contestants' visions and values are also affected by outcomes. A government's decisions elevate one view of the problem while depressing another.
Major policy changes are morally complex and draw many participants for different reasons. One issue may dominate the debate, but there will likely be other issues provoking other divisions, which will be affected by the outcome. Consider the civil rights revolution of the 1950s and 1960s. Through a series of Supreme Court decisions and congressional enactments, black Americans obtained more equal rights. That changed the law but also expanded the role of the federal government, gave the courts a new prominence, and established a multiracial vision of the national community. The protagonists on all these issues now faced a new moral universe.
The most radical change in American domestic policy since civil rights was welfare reform. In the 1990s, the policy of cash assistance for families was recast to require many more welfare mothers to work as a condition of aid. That requirement, coupled with a strong economy and the provision of new benefits, drove most recipients off the rolls, mostly into jobs. What issues were at stake in this transformation? This article offers the first systematic analysis of the congressional politics behind welfare reform over a period of more than 30 years. I track what the participants thought the issues were and how these perceptions changed over time. The story is more complex than is usually perceived.
“Welfare” here originally meant “Aid to Families with Dependent Children” (AFDC), first enacted as part of the Social Security Act of 1935. AFDC supported needy, single-parent families. The program was run principally by the states, which controlled benefit levels, but the federal government helped finance state programs and thus had the authority to regulate them. AFDC became controversial in the 1960s as rising unwed pregnancy rates and falling work levels among poor adults inflated the rolls.
Starting in 1962, repeated efforts were made to reform welfare, some pressed by presidents and others by Congress. Some aimed to make the program more generous while others aimed to eliminate the incentives that AFDC seemed to create for poor adults not to marry or work. Most of the reform plans were enacted, but some were defeated. Over time, reformers concentrated chiefly on motivating and then requiring welfare mothers to work as a condition of aid. These efforts culminated in the passage of the radical Personal Responsibility and Work Opportunity Reconciliation Act (PRWORA) of 1996, which recast AFDC as “Temporary Assistance for Needy Families” (TANF), strengthened work and child support requirements, and limited welfare in other ways.
The body of research on welfare reform is voluminous, but it has not pinned down what the issues were. Most scholars assume that the struggle was an ideological or partisan one, pitting liberals or Democrats who wanted a bigger government against conservatives or Republicans who wanted a smaller one. In this light, PRWORA is usually perceived as a conservative victory that downsized government. However, some scholars claim that other differences mattered more.
No one has yet closely examined the actual discourse. By measuring how the participants framed the issue over the years, I here show that the debate evoked several distinct divisions. All of these issues are still central to social politics today. The ideological dispute over the scale of government was indeed important, but this focus receded over time in favor of a more practical dispute over how to manage welfare programs. The discourse did become more conservative, but the change in the agenda—the dominant way the issue was framed—is more important.
The following sections summarize the key episodes of welfare reform, past research, the meaning of the agenda used here, my hypotheses and methodology, my findings, and conclusions.
The welfare reform controversy in Congress comprised six episodes in which major reforms to AFDC were proposed and debated.1
The final results of reform were radical. Starting in 1994, the rolls fell by around two-thirds through 2008, driven by rising work requirements as well as by a strong economy and new wage and child care subsidies. Work levels among low-income single mothers rose sharply, and poverty fell. Most mothers leaving welfare for jobs gained income, although many remained poor, and some who did not work lost ground (Blank and Haskins 2001; Duncan and Chase-Lansdale 2001b; Grogger and Karoly 2005). Reform efforts were criticized by advocates for the poor, but they were sufficiently successful to quiet controversy for the time being. TANF was reauthorized with even stronger work requirements in 2006.
The broader import of reform, however, is unclear. Several different issues have been at stake over the years, all of which are still central to social politics in America. Reform has acted as a kaleidoscope that put all these questions into play. The dominant view, as mentioned, is that welfare is an ideological or partisan issue that reflects the traditional division between the parties. Typically, liberals and Democrats want the government to “do more” for various groups; conversely, conservatives and Republicans typically want government to “do less.” In general, liberals have sought to raise benefits, extend coverage to new groups, and nationalize control of the system, as FAP and PBJI proposed. Conservatives have aimed to restrict benefits and eligibility while devolving more control to localities, goals advanced by PRWORA.
To liberals, welfare is part of the social safety net, a response to economic insecurity. The market economy fails to assure all workers jobs and wages that are sufficient to support families (Katz 1986). Welfare is a form of unemployment insurance, tiding families over periods when they cannot find work (Piven and Cloward 1993). To conservatives, however, the free market is not the problem, but the solution. It is welfare that produces poverty by rewarding people if they have children outside marriage and fail to work (Anderson 1978; Gilder 1981; Murray 1984). Liberals see such contentions as part of a broader assault on the welfare state (Marmor, Mashaw, and Harvey 1990; Noble 1997; Skocpol 2000). The debate on PRWORA triggered partisan fireworks along these lines, tending to confirm the idea that welfare is part of the larger battle over government.
However, one minority view holds that the deepest welfare division concerns obligation: Should welfare adults have to work in return for aid? That issue differs from the ideological question of how generous welfare should be. Liberals typically opposed even the mild work tests proposed by WIN or FSA, fearful that such demands would overstress families. Conservatives argued that it was fair to require employable recipients to work if they depended on the public. Conservatives attacked the more liberal plans—FAP and PBJI—for failing to clearly require adult claimants to work. It is important to note that liberals were more pro-government on the ideological issue, but conservatives were more pro-government about obligation. Conservatives sought to use government to enforce values, while liberals resisted this (Bryner 1998; Mead 1986; Mead 1992; Teles 1996).4
A third interpretation is that welfare politics is about opportunity. In and out of Congress, experts have debated whether it is even possible for recipients to support themselves. Are there “barriers” to employment, such as racial discrimination or a lack of jobs or child care, that prevent welfare adults from working? In Congress, liberals tended to assert this in rebuttal to conservative demands for work. However, conservatives typically believed that work was possible and hence obligatory. Most academic experts have adopted the first position (e.g., Danziger and Weinberg 1986), while advocates of cuts or work tests took the second (e.g., Mead 1992; Murray 1984). This issue is a battle not directly about welfare, but about the fairness of the opportunity structure outside government.
A fourth interpretation is that welfare politics is paternalist. Much welfare discourse is not about justice or fairness, but rather about the practical question of how government should manage welfare. How should it arrange welfare, child care, or child support in order to promote work, deter unwed pregnancy, and promote recovery from poverty? Liberals tend to want a more supportive regime for the poor, conservatives a more demanding one. The two sides debated how exactly recipients would be assisted or required to work, and what penalties nonwork would bring. Paternalist disputes often turn on questions of “what works” in promoting work or child support in light of research, evaluations, or experience. Several treatments of welfare politics stress the rise of this problem-solving style (Heclo 1974; Katz 1989; Leman 1980; Lenkowsky 1986; Moynihan 1965).5 This approach may have come to the forefront of debates simply because of the accumulation of information about welfare over time, thus breaking down ideological certainties and fostering more practical responses.
The literature does not resolve the question of which issue has dominated the discourse for two main reasons. First, accounts of reform battles in Congress concentrate on explaining outcomes rather than characterizing the debate. The questions are why the proposals made at each reform stage did or did not pass, and why reform has finally stressed work tests rather than more generous coverage and benefits. Accounts of PRWORA, for instance, trace the origins of the law to Bill Clinton's promise to “end welfare as we know it” during his 1992 campaign, as well as to the Republicans' 1994 platform of the “Contract with America,” which anticipated PRWORA (DeParle 2004; Haskins 2006; Weaver 2000). Accounts of earlier reform episodes also focus on the victory or loss of one side or another (Bowler 1974; Burke and Burke 1974; Lynn and Whitman 1981; Moynihan 1973; Steensland 2008). Little close analysis has been done of the arguments that participants actually made.
Second, no longitudinal research has been conducted on welfare politics. Welfare has been an issue for long enough that any view of its politics must encompass change over time. Yet most existing studies discuss only a single reform episode, principally the Nixon Family Assistance Plan (FAP) in 1969–72, or PRWORA in 1994–96. Three works discuss the controversy more broadly but give limited attention to the period prior to PRWORA (Bryner 1998; Teles 1996; Weaver 2000).
Again, identifying which framing of the welfare issue dominates an episode is crucial for determining the broader significance of reform. This framing affects which values were advanced or denigrated by the outcomes of reform. If welfare is an ideological battle, then the triumph of conservatism in FSA and PRWORA implies a diminution of the welfare state. But if reform was really about work tests or problem-solving, then welfare was changed more than downgraded, and public commitment to the poor remains.
This study is the first to gauge rigorously how the welfare issue has appeared in Congress and how the issue has changed over time. I trace the shape that the welfare agenda has taken on Capitol Hill from its first appearance in the 1960s through PRWORA in 1996. “Agenda” may be understood in various ways. In Kingdon's (1995) influential scheme, the term connotes the confluence of a recognized public problem with a feasible solution and the political will to enact it, producing a “policy window.” Such linkages have occurred in welfare reform, but this approach does not directly incorporate change over time.
Downs (1972) suggests that issues rise and fall in salience as the political class awakens to a problem and proposes solutions. Then, when change proves difficult or costly, people become disillusioned and turn to other concerns. Baumgartner and Jones (1993) speak of “policy monopolies” that are created and later torn down by criticism. These approaches do encompass change, but they say too little about how the politicians themselves understand the issue. The idea of “issue evolution” (Carmines and Stimson 1989) captures more of the intellectual substance of an issue as its meaning changes over time, but that approach assumes that there is a single dominant meaning at any point rather than multiple possibilities. Issue evolution also focuses mostly on electoral and party politics, whereas my focus here is on elites.
Other scholars argue that how disputes are “framed” or “constructed” affects the politics around them, an approach that does characterize issues. However, the framing literature chiefly concerns effects on mass opinion (Bobo and Smith 1994; Gilens 1999; Iyengar 1989, 1990; Ross 2000; Schneider and Ingram 1993), whereas here I focus on leaders in and around Congress.
More useful to this study is Schattschneider's (1960) idea of the agenda as simply the issue or division. The agenda connotes the dispute that is aroused by a policy issue. It is the way that players in the political arena define the choice that they face. Political change occurs not only when positions or coalitions on an issue change, but also when the issue itself changes. Each issue evokes polarization, but the meaning of “left” and “right” shifts along with the issue. One definition of conflict tends to drive out another—what Schattschneider calls “mobilization of bias.” This idea resembles Riker's “heresthetics,” in which elites compete to define issues and processes in terms favorable to themselves (1986), although Schattschneider suggests less amoral calculation. This concept works particularly well for the welfare issue, in which there are multiple possible issue definitions and the balance shifts over time.
We can use Schattschneider's approach to interrelate the four different readings of the welfare issue that have been emphasized in past research. Figure 1 arrays these four approaches in terms of two dimensions: political style and level. The two political styles I call “scale-of-government” versus “dependency.” The first focuses on impersonal questions of public commitment and values, while the second focuses on more personal issues of responsibility and competence. These issues play out on two political levels—basic principles versus concrete conditions.
The two scale-of-government issues share a concern for the role of government, but ideology is debated in terms of broad principles, whereas opportunity is discussed in more practical terms concerning whether it is possible for poor adults to work. Similarly, the two dependency disputes share a focus on personal responsibility and competence and concern the question of what the poor can reasonably be expected to do for themselves. Obligation is debated in terms of principles and paternalism in terms of the specifics of welfare reform programs, such as child care, work incentives, and work or child support requirements.
Size-of-government politics is about the good society: Should government intervene in the private sector to promote more equal opportunities and outcomes for ordinary Americans and, if so, how? In contrast, dependency politics is about good behavior: Should government seek to enforce central norms such as the work ethic and, if so, how? With size-of-government issues, divisions occur over the values that society should have, whereas with dependency issues, divisions occur over who is responsible for realizing values on which the public widely agrees, such as the work ethic.
Because of the mobilization of bias, the questions that are raised by each political style should tend to drive out those raised by the other. Scale-of-government debates tend to be abstract and impersonal arguments about justice or fairness among citizens, in which poor people's ability to “get ahead” is taken for granted. Conversely, dependency debates are more personal arguments about the morals and capacities of the poor, in which the meaning of justice in a more impersonal sense is taken for granted.
All four issues provoke division between liberals and conservatives, but the meaning of “left” and “right” shifts depending on the issue. In welfare issues, the liberal position on the ideological division favors doing “more” and nationalization, while the conservative position favors doing “less” and devolution. On the question of obligation, liberals defend entitlement (i.e., aid based on impersonal criteria and without behavioral conditions) while conservatives demand work tests, “social contract,” or “reciprocity.” On the question of opportunity, liberals claim that social conditions do not permit the poor to work while conservatives say they do. On the question of paternalism, liberals want the services and requirements of work or child support programs to be more supportive, conservatives more demanding. The terms of the debate shift as the issue shifts, just as Schattschneider says.
The question that motivates this study—What is welfare politics about?—has been raised by the existing literature. The hypotheses are drawn from my readings of congressional hearing transcripts in connection with my earlier research on welfare reform. Those inquiries made welfare politics seem more complex than past research has suggested. While most existing studies emphasize ideological division, the three other issues defined previously are also apparent. These other issues seemed to grow in prominence as reform proceeded, at least through the 1980s. The discourse by that point also seemed more conservative than portrayed in much past writing. In light of these impressions, I proposed the following hypotheses for the current, more rigorous analysis:
My research assistants and I tested these hypotheses by coding the speech of witnesses in congressional hearings on welfare reform during the six principal episodes of reform mentioned previously. I expected that these occasions would reveal the patterns of welfare politics most clearly because major policy changes were proposed, even if not enacted.
Earlier research on hearings suggested that each proposal prompted a wide-ranging debate that was constrained only slightly by the specifics of the reform plan. The proposal of the hour acted like a Rorschach test onto which the participants projected their general views about poverty and welfare in America. Many groups performed their own studies to express these wider concerns. The assumed stakes, again, included the framing of the issue and the values involved, not just the specific reform proposal on the table.
All witnesses appearing in the principal committee hearings during all six reform stages were coded.6 I limited the project by covering only those hearings that were held by committees with jurisdiction over AFDC/TANF—the Committee on Ways and Means in the House and the Finance Committee in the Senate—and omitting other committees. Only those hearings that concerned the main reform proposals were coded, with other hearings omitted.7 The witnesses appearing in either house numbered 49 for SSA, 57 for WIN, 160 for FAP, 119 for PBJI, 155 for FSA, and 284 for PRWORA, a total of 824 in all.
Only witnesses' oral statements to the committee were coded; their written statements were omitted, because they are less candid and pointed. I also omitted witnesses' colloquies with committee members to more clearly define what was coded. Statements by committee members were not coded. For each witness, coders recorded, as open-ended responses, what each speaker thought the main welfare issue was and the position that he or she took on it. Coders also recorded several measured variables. Those used in the current analysis include:
The results reveal which issue types bulked largest among witnesses at each stage of the controversy and how liberal or conservative opinion was on each issue. Results are shown both in general and for 14 different types of witnesses (e.g., administration spokespersons, interest groups, academic experts).
In content analysis, such as we are doing here, the categories used for coding should reflect prior, less-structured research. They should be exhaustive, mutually exclusive, and derived from a single principle. The classification of each case (here, each witness) should be made independently of the classification of other cases (Holsti 1969, 94–126). This analysis satisfies all these rules, except for the derivation of the coding scheme from a single principle. The four issue types differ in two dimensions—political style and level. Two dimensions simply seem necessary to capture the full range of welfare discourse.
Based on searches in JSTOR and Goehlert and Sayre (1982), no published research on welfare politics remotely like this exists. However, this project does fall within an emerging field that uses content analysis rather than voting records to study legislative politics on a range of issues. Baumgartner and Jones (1993) coded entire hearings in terms of whether they were “positive” or “negative” in tone. Burstein and Bricher (1997) and Burstein, Bricher, and Einwohner (1995) coded bill reports or legislation over time in terms of how problems were framed. Burstein and Hirsh (2007) coded witnesses in terms of information provided on policy proposals. Stein (2004, 26–84) coded congressional floor debates in terms of how program recipients were characterized. Previous research on witnesses' arguments coded them at one point in time or else used post-hoc analysis.9 This study is the first to track the substantive meaning of any issue in Congress over an extended period using witnesses and a preset analytic scheme.10
Coding was done by two research assistants. Following training and detailed instructions, these two individuals coded all witnesses in the specified hearings in chronological order. In each hearing, the coders rated each witness three times.11 For the first two cuts, they worked independently. After each cut, I calculated how closely they agreed on the different variables and advised them where agreement was high or low, without discussing individual witnesses. For the third cut, the coders worked together, debating codings and adopting the same ones in cases on which they agreed, although they were still free to disagree. Continuity was served by the fact that only three coders worked on the study over seven years, and one coder served throughout the entire study. Disagreement among the coders appeared to be random, rather than based on divergent interpretations of the categories. Accordingly, I have treated the third cut as most reliable, and I report those data here.12
Intercoder agreement was high. The top panel of table 1 shows the percentage of witnesses for whom the coders agreed on the different variables. Coders disagreed about many cases on the first cut, but by the third, agreement was usually over 90%. Agreement on whether paternalism was present ran somewhat lower than for the other issues, as this issue type was the least sharply defined. The bottom panel of table 1 shows Cohen's Kappa, a more conservative statistic that shows how far coding narrowed disagreement after allowing for agreement that would have occurred by chance. Nearly all the kappas were sizable and significant.13
The witnesses are important inasmuch as they had some influence on decisions. The pool of witnesses included, for instance, many members of Congress who later voted on reform. I also assumed that witnesses reflect the broader expert class that is concerned with welfare and poverty. I could not make that assumption if voices for or against reform were excluded from the hearings, but that does not appear to be the case. Table 2 shows that witnesses came from a wide range of different groups. While changes in the distribution of witnesses are more than random (a χ2 test of the table is significant), the shifts show no definite trend. Political views among the witnesses were also diverse, even though the general direction of reform is conservative. Table 3 shows the percent of witnesses who were coded as liberal on their leading issue, whichever that was, by group and stage of reform. There is a general drift to the right (see the following paragraphs), but liberal groups are strongly represented, and a majority of the witnesses remained liberal throughout.
Witnesses in Welfare Reform Hearings, by Group Affiliation and Stage
Note. χ2 (65 df) = 187.4, p = .000.
Percent of Witnesses Coded as Liberal, by Group and Stage
Note. Figures show percent of witnesses in the indicated cells who were coded “left” on their leading issue. Except for 11 unsure cases, all other cases were coded “right.” Dashes mean that no witnesses of this type appeared.
PRWORA poses the major question about representativeness, as it was both the most conservative reform stage and the one with a plurality of the witnesses. However, the range of groups represented in the PRWORA hearings was wider than participated in any previous stage except FAP. And despite the rightward shift, a majority of witnesses even during PRWORA were still left of center. This finding confirms prior research that “about half” of the PRWORA witnesses “generally were in favor of more rather than less-expansive programs for poor people” (Winston 2002, 94).
I would be especially concerned if the witnesses overall were slanted in service of a political agenda either for or against reform. However, there is almost no suggestion of such bias in the literature.14 A central reason for this absence is that in planning hearings, both majority and minority caucuses or their staffs name some of the witnesses. Thus, although each side chooses witnesses who support its own view of reform, a diversity of voices is assured. The keynote of welfare hearings is debate, not consensus. Even when elected or appointed officeholders are conservatives, the witnesses who appear before Congress are much the same as those who appear at outside conferences about welfare or poverty, many of whom are well left of center. Accordingly, it is plausible to view the witnesses as speaking not only for themselves, but also for the wider expert class.15
The findings largely confirm our hypotheses. Figure 2 shows the proportions of witnesses who mentioned each of the four issue types during each stage of the reform controversy. As expected, all four issues appeared at every stage. Equally important is the fact that only a handful of witnesses—36 out of 824—mentioned any issue other than the four previously defined. Only two witnesses failed to mention any of the four.16 Thus, these four main issues encompassed the lion's share of welfare discourse.
The ideological issue dominated the early stages of reform but then declined as time passed. In SSA, every witness mentioned it, but by PRWORA, only 72% did. Meanwhile, the paternalist issue rose more sharply, from only 18% of witnesses mentioning it in the SSA hearings, to over 80% doing so in the FSA and PRWORA hearings. This shift is even clearer in figure 3, which shows the share of witnesses who chose each issue type as their leading issue. Ideological discourse entirely dominated at the outset but then fell steeply, until by the time of FSA in 1986–88, only 35% of witnesses led with this issue. Meanwhile, paternalism rose from 0% to 59% in the FSA hearings. That increase might partly reflect FSA's heavy focus on strengthening work and child support programs, concerns likely to elicit paternalist commentary. In PRWORA in 1994–96, the ideological issue rebounded somewhat, as one would expect, because of the sharp debate that occurred over changes in the federal role. Yet paternalism, while declining slightly, retained its dominance, with 57% of witnesses leading with this issue type. PRWORA's debate over work, family, and child support details outweighed the classical partisan battle over whether to devolve and limit welfare, although both occurred.17
The obligation and opportunity issues proved to be less prominent than anticipated, although each peaked at the point that one would expect. In figure 2, obligation surged during the WIN and FAP hearings in the late 1960s, when work tests for recipients were initially debated, while opportunity peaked during PBJI in the late 1970s, when the economy was troubled and thus the ability of welfare adults to work was thrown most into doubt. By the FSA hearings in the late 1980s, few disputed that adult recipients should and could work, and the discourse shifted toward the details of such requirements—to paternalism. One close participant observed that “after 1992 the debate was about the amount of work required, the specific conditions of work, and the consequences for individuals and states of not working rather than on whether recipients should be required to work” (Haskins 2001, 17).
Opinion shifted rightward on all the issues as expected, but the shifts were limited.18 Table 3 shows that views drifted rightward over time, while most witnesses remained left of center. Figure 4 shows the percentages of witnesses citing each issue type at each stage of the hearings who were coded left on that issue, with liberalism again dominating. Well over half the witnesses—and often three-quarters—were rated left of center on most of the issues during most stages. Liberal views on paternalism did decline after PBJI, but only on obligation did the liberal position collapse, vanishing by the time of PRWORA. Especially on the size-of-government issues (ideology and opportunity), liberalism remained robust, even in PRWORA, when 63% of ideologues remained liberals. Figure 5 shows the percentage of witnesses who were left or right on their leading issue, whichever that was. Here, the fall of liberal positions and the rise of conservatism is more marked, but liberals still slightly outnumbered conservatives by the time of PRWORA.
It is fair to say that our first four hypotheses were supported, with the proviso that obligation and opportunity issues played a smaller role in reform than expected. Primarily, the agenda moved away from ideology toward paternalism. This shift is more marked than the rightward shift in positions on the issues. The discourse shifted from ideological contention toward problem-solving, even though most witnesses remained left of center. The nature of the shift helps to explain why, as I note later, reform finally assumed a form that was conservative but not antigovernment.
While discourse did shift toward the dependency style and the concrete as hypothesized, this occurred mainly because paternalism displaced ideology as the dominant discourse. Mobilization of bias did not operate strongly, in the sense of one issue type driving out another. Witnesses who mentioned ideological issues were only slightly less likely to mention paternalist issues than witnesses who did not. Conversely, paternalist mentions only slightly depressed ideological mentions (details not shown).
Our fifth hypothesis was that the shifts would be general across different types of participants. Table 3 has shown that opinion drifted to the right during the welfare controversy for most witness groups, although there were exceptions. Were the shifts in the agenda toward a different mix of issues also general? Some might regard paternalist discourse as another form of conservatism. If so, then we should expect to see the shift toward paternalism occurring mainly among conservative witnesses, not liberals.
Figure 6 shows the proportion of witnesses who were left and right on their leading issue, whichever that was, who raised ideological and paternalist issues at each stage of the hearings. For both left and right adherents, the focus on ideology fell somewhat, while the incidence of paternalism rose more sharply, as is true overall. At most, the decline of ideological mentions and the rise in paternalism were somewhat steeper among witnesses on the right than on the left. Witnesses left of center usually maintained their ideological stances, yet they increasingly addressed paternalistic issues as well. Again, mobilization of bias appears to have been slight.
Table 4 shows the proportion of each witness group that raised ideological or paternalist issues at each stage of the hearings. For most groups, ideological mentions fell somewhat over time, while paternalist mentions rose more steeply, as is true overall. The paternalist trend tended to peak with FSA, then fall slightly with PRWORA. No group clearly led the changes. The groups that were most extreme in ideological terms tended to maintain those stances, yet they too gave rising attention to paternalism. Thus, the changes in discourse were broad-based.
Witnesses Raising Ideological and Paternalist Issues, Overall and by Groups
Note. Civil rights/ideological, and advisory/other categories are combined because of low N. Dashes signify that this type of group did not testify in these hearings. Base for percentages is all witnesses of that type in that stage. For numbers of witnesses by group in each stage, see table 2.
Our final hypothesis states that shifts in the welfare agenda would parallel changes in national policy. As mentioned, past research on welfare reform chiefly seeks to explain legislative outcomes, while our findings apply to the terms of the debate.
Some might expect little connection between the two. Welfare reform, after all, was not crafted in committee hearings but by executive and congressional leaders. Their politics might be quite different from those of the witnesses, and often were. Indeed, some analysts assert that welfare policymakers were out of touch with broader expert or mass opinion, either to the left or right. Some observers say that the FAP and PBJI plans of the 1960s and 1970s were designed by liberal economists who attempted to foist guaranteed income on America, although there was little support for this (Davies 1996; Mead 1986; Moynihan 1973). More recently, others have charged that the reform plans of the 1980s and 1990s were crafted by conservative moralists who stigmatized unwed pregnancy, even though most Americans were more tolerant (Gans 1995; Mink 1998).
Our results deprecate such views. Reform proposals were not crafted in isolation from broader expert opinion. The plans ran parallel to the opinion tapped by the hearings, whether or not proposals were enacted. In the early stages of controversy, both the proposals and the politics were strongly ideological and liberal. Reformers chiefly aimed to build up benefits and services for the needy, and whether to do this was the chief issue in the hearings. Later, both proposals and politics became more conservative, but above all, more paternalist. Starting with the PBJI hearings in the late 1970s and strengthening through FSA and PRWORA in the 1980s and 1990s, proposals gave less stress to spending and benefits and more to crafting institutions that could manage the poor in order to achieve work and child support. Committee hearings trended the same way. PRWORA was radical in both ideological and paternalist terms, as it proposed both to downsize and devolve welfare and to construct the toughest national work and child support regime yet imagined. Accordingly, we find both the ideological and paternalist divisions strongly reflected in those hearings.
These results do not reveal why discourse shifted toward paternalism. The reasons could include the general movement of partisan politics to the right (especially if paternalism is seen as a form of conservatism), the public's strong will to enforce work in welfare, and the rising influence of experts in all areas of national policy. The existing accounts of welfare reform politics may have more to say about causes than does the evidence here.
These results, however, bear directly on the values at stake in the controversy and, hence, its significance. Welfare reform is commonly seen as a defeat for the left. Since the 1960s, Congress has largely rejected liberal proposals to expand welfare or raise benefits while embracing conservative proposals to impose work and child support tests and devolve the program. However, this view assumes that welfare reform was an ideological struggle. Our results show that this interpretation oversimplifies the conflict. Reform was finally shaped by a paternalist discourse in which “left” and “right” had different meanings and in which the outcome was not principally antigovernment.
The practical effects of reform confirm this conclusion. PRWORA included antigovernment features, but its main impact was to build up welfare work programs and support services rather than cut back welfare defined broadly. Recipients of cash aid under AFDC/TANF did plummet in the later 1990s, but caseloads grew in other means-tested income programs—food stamps and the Earned Income Tax Credit. Federal and state spending on all three programs actually rose from $77 to $79 billion between 1994 and 2001, a fall of only 14% in constant dollars.19 Large new sums were also spent on child and health care to enable poor mothers to leave AFDC/TANF for work (U.S. Congress 2004, 9.52; Besharov and Samari 2001, 463–64). So, overall, the nation spent more to reform welfare than it had spent on the unreformed system.
The advance of paternalism reflected lessons drawn from the earlier stages of controversy. The political class accepted that the main task of reform was not to alter the scale of the welfare state. Sharp changes to either the left or the right were strongly resisted. Rather, factions agreed on the need to manage welfare better. Paternalism did not totally displace partisan controversy, as the intense disputes over PRWORA demonstrated. However, welfare battles have shifted toward a narrower terrain. Disputes are less about justice and more about how to improve programs in practical ways. Individuals or groups seeking to do more for the needy used to argue simply that the needy deserved it. Today, they are likelier to say, more practically, that greater generosity will achieve better development for children (Duncan and Chase-Lansdale 2001a).
Very likely, the overall weakening of ideological dispute could be found on many issues. Welfare reform might well be an instance of what Francis Fukuyama has called the end of history. Fukuyama claimed that at the end of the Cold War, there was no longer principled opposition on the world scene to democracy and capitalism (1992). Some would dispute that the ideological division has faded in international politics, but it clearly has in the internal politics of Western nations. Compared to 50 years ago, disputes over the welfare state are seen less as struggles between polarized social visions than as practical problems of cost and dependency. Western Europe has begun to reform welfare much as the United States has done, chiefly by expecting work from adult recipients. As in the United States, the motive is primarily to foster social integration; saving money is secondary. Parties of both the right and the left have contributed to this agenda, setting aside the deeper divisions of the past (Lødemel and Trickey 2001). The meaning is not that conservatives have won the old, ideological battle, but that a new agenda has developed, in which the issues are much more managerial.
That shift is frustrating to those who seek a broader debate about the role of government in welfare, either on the right or left. However, that change may be a sign of maturity. Welfare reform has taught Washington that the ideological debate is important, but that it does get in the way of actually improving the welfare system. Politics has reshaped welfare, but welfare has also educated the public mind.
1 There is, of course, a popular politics of welfare, discussed by such scholars as Gilens (1999) and Piven and Cloward (1993), but I focus here on the controversy that occurred in and around Congress.
2 Coverage had been extended to two-parent families in 1961, but only as a state-by-state option and with restrictive conditions. The new proposals mandated coverage and eased those conditions.
3 This list omits the important changes to welfare that occurred in 1981, when Ronald Reagan persuaded Congress to cut AFDC eligibility and allow states to craft more demanding welfare work programs. These changes were enacted as part of omnibus budget legislation. Thus, they occasioned too little discussion in their own right to permit coding. I have also omitted the 2006 reauthorization of TANF, as it did not consider or enact any major changes to existing policy.
4 Work requirements are sometimes dubbed “paternalism.” However, that term applies properly not to the principle of work tests, but rather to detailed oversight of clients by case managers. Accordingly, we treat obligation and paternalism here as separate—albeit related—issues.
5 Some individuals or groups believe that welfare is a racial issue, viewing those who are more opposed to blacks as also more hostile to welfare (Gilens 1999). However, welfare debates at either the federal or state level have contained little overt racial content. Thus, this theory must assert that nonracial discourse is implicitly “coded” with racial content. This study concerns overt divisions, so I do not test this theory here.
6 To be coded, witnesses had to speak about the reform of AFDC. Some reform bills also contained changes in welfare for the aged, blind, and disabled, or in Social Security or health programs. Witnesses speaking only on these issues were omitted.
7 Welfare reform legislation was not importantly shaped by the other committees. A listing of the hearings covered is available from the author upon request.
8 Some recent methods have been formulated that allow researchers to sort political language into categories by computer (see Hopkins and King 2008). Using an early such method, Reinhard (2003) analyzed some welfare hearings in 1988 and 1993–94 and produced results similar to mine. My study predates the new computer-driven methods. However, even had they been available, these programs can classify language according to only one set of categories. Here, language was classified by several dimensions—which issues were present, left-right position on the issues, and which issues were most important. I also judged whether issues other than the four defined were present. Only hand coding, in which individual coders wrestle with witnesses' full language, could have dealt with all these dimensions..
9 Del Sesto (1980a, 1980b) coded congressional hearings for argumentative content using witnesses, but the analytic structure is post hoc and based on a single hearing in 1973–74. Bero et al. (2001) coded testimony and public comment about tobacco regulation in Maryland and Washington state for argumentative content, but without a preset analytic scheme and at only one point in time. MacLeod (2002) coded witnesses rather than whole hearings in terms of whether they supported or opposed the AT&T telephone monopoly. However, the arguments used are not categorized with the depth I offer here, and witnesses are not coded in terms of them.
10 I did not use data from the Policy Agendas Project, created by Bryan Jones, John Wilkerson, and Frank Baumgartner, because that study is based on coding entire hearings in terms of subject and other variables. My data are based on witnesses. I also did not use data from the Congressional Bills Project, because those data are about bills, not witnesses.
11 The coders were trained through practice coding of hearings separate from those coded for the project. The coding instructions are available from the author upon request.
12 Specifically, I report Kendal Elliott's third cut, as she worked all stages of the hearings. Her third cut is a joint product with her partners, Ian Gold (who worked SSA through PBJI) and Frank Ortiz (FSA and PRWORA). I could have used an additional coder to resolve differences, but an extra individual would have meant more complexity and cost, and would have put continuity more in doubt.
13 For most of these variables, expected agreement, based on marginal distributions, was over 50% and sometimes over 90%, so it was not easy to improve on that level. See Cohen (1960).
14 The only author to assert a slant of which I am aware is Steensland (2008, 172), who claims that Finance Committee hearings held on the Nixon FAP in 1972 were “stacked” against approval. However, the evidence for this argument is a single newspaper article that claims that administration spokespersons were not invited to these hearings, and that the initial witnesses were conservative. No systematic attempt to slant the hearings is mentioned (Hunter 1972). According to our coding, the 1972 Senate witnesses were only slightly more conservative than those witnesses who appeared earlier in the Senate hearings or, indeed, in the FAP hearings as a whole. The 1972 Senate witnesses also represented a wide variety of groups with diverse views of reform, as was the case in the earlier FAP hearings.
15 One sign of representativeness is that the shift in the hearings discourse away from ideological and toward paternalist discourse, shown later, is mirrored in academic writing about poverty and welfare. Earlier compendia of expert analyses, such as Danziger and Weinberg (1986), stress the structural causes of poverty, while later ones, such as Blank and Haskins (2001), give much more attention to behavioral causes.
16 These other concerns mostly had to do with the organization and management of programs. Fifteen of the 36 cases arose during PRWORA, with only 21 arising in the five stages prior to that.
17 Reinhard (2003) reaches similar conclusions. Had I framed the issues only for PRWORA, issues such as marriage that are here subsumed under paternalism might have been treated separately.
18 Coders were instructed to rate positions left or right of center using a single scale over all the stages, not with reference to a single stage. The continuity of the coders made this possible.
19 These calculations use caseload and spending figures from the U.S. Congress (2004, 7.31, 7.59, 13.41, 15.9, 15.24, 15.44, 15.86, 15.88), with spending figures adjusted for inflation using the consumer price index for all urban consumers.