Making Progress on “Welfare Reform”:
A Response to Scott Winship
September 30, 2016
Scott Winship of the Manhattan Institute contends that “welfare reform” is an “unambiguous”
success and reduced child poverty rates (measured in various ways), while I contend it is a
policy failure and increased the depth of poverty for many of our nation’s neediest families. This
paper is a response to a series of questions posed by Winship to me on Twitter, as well as various
related statements from his recent report, Poverty After Welfare Reform,
and articles about that
report and “welfare reform” generally, most notably in the National Review Online.
respect to the latter, I also draw on annotated comments using Genius in an exchange between
Shawn Fremstad of the Center for American Progress and Winship. (Both Fremstad and
Winship were recently commended for the civility of their discourse on this topic in an article by
Samuel Hammond of the Niskanen Center, “Celebrating Civil Discourse on Welfare Reform.”
Winship and I have very different views about “welfare reform’s” effects and policy
implications, stemming from (among other things) differences in: 1) the definition of “welfare
reform”; 2) standards for the evidence needed to draw causal inferences; 3) the appropriate
poverty measure; and 4) the importance of policy details in making policy recommendations.
After a discussion about what we mean by “welfare reform” and the problem of causal inference,
I highlight Winship’s questions to me and related statements from various papers; my response is
labeled “PC Response” (where PC is for “Peter the Citizen”).
I will also touch on some of Winship’s other recent statements about “welfare reform” and
briefly discuss the importance of understanding policy details in the policymaking process. It is
on this last issue, where most conservatives today have failed to provide constructive insights.
Winship’s paper about poverty after welfare reform is just that – a paper describing poverty
trends using one set of assumptions from before “welfare reform” to nearly 20 years after.
Winship’s paper is not a strong basis for causal inferences and his suggestion that “the lessons of
welfare reform should be extended to other safety-net programs” is irresponsible. (I will
elaborate on this last point in a forthcoming paper.)
WARNING: This “response” is long and “in the weeds”; its primary goal is to address
all of Winship’s concerns about my statements regarding “welfare reform” and to
highlight some of my concerns with his statements. This discussion draft along with
other responses to Winship will form the basis for a more complete paper on poverty
after “welfare reform” responding to conservative claims that the 1996 “welfare reform”
was, in Speaker Ryan’s words – an “unprecedented success.” In the end, Winship and I
may have to agree to disagree, but at least the reasons for our disagreements will be
clearer. Understanding these differences is important when considering the policy
implications of “welfare reform.”
What is “welfare reform”?
When I refer to “welfare reform,” I usually put the term in quotation marks, because I am
referring to the Temporary Assistance for Needy Families (TANF) block grant. (Indeed, my last
“response” to Winship was “Making Progress on TANF.”) I put the term “welfare reform” in
quotation marks when referring to TANF because TANF is not really welfare reform at all, but a
flexible funding stream with a myriad of dysfunctional federal requirements. When Winship
refers to “welfare reform,” he says he means the Personal Responsibility and Work Opportunity
and Reconciliation Act of 1996 (PRWORA), which included TANF. RRWORA also included
changes to the Supplemental Security Income (SSI) program, the Food Stamp (now SNAP) and
various child nutrition programs, child support enforcement, and child care; it also restricted
welfare and public benefits for immigrants.
TANF vs. PRWORA. Winship recently elaborated on the importance he places on the
distinction between TANF and PRWORA:
…I am arguing (and always have) that a specific piece of legislation – the Personal
Responsibility and Work Opportunity Reconciliation Act – benefited children. But
critics of PRWORA always shift the debate to a specific subset of TANF policies. It’s
like saying, “X was good,” and having critics say, “How can you possibly think part-of-
X-called-Y was good?” X can be good without Y being good. So critics need to rebut
my actual argument, which is about PRWORA, not certain parts of TANF that critics
don’t like (and that I’m not thrilled about either).
This distinction has not been clear in Winship’s papers, as his discussion of “welfare reform”
provisions is almost always TANF-specific. For example,
In “Conservative Reforms to the Safety Net Will Reduce Poverty,”
there is no mention
of PRWORA or any non-TANF PRWORA provision, but the article does refer to decline
in families receiving TANF cash assistance.
In “Welfare Reform Reduced Poverty and No One Can Contest It,”
the focus is on
children in single-parent families – TANF’s primary focus. Many of the other PRWORA
provisions focus on other populations, e.g., SNAP’s work requirement is aimed at able-
bodied adults without dependents and PRWORA’s immigrant provisions do not
differentiate by family structure. In discussing the responsiveness of various safety net
programs to the Great Recession, Winship compares TANF to SNAP. There is no
discussion of the responsiveness of the other programs included in PRWORA.
In “Welfare Reform’s Success and the War on Immobility,”
there is one reference to the
“Personal Responsibility and Work Opportunity Reconciliation Act,” but the discussion
of welfare reform is limited to TANF. Indeed, Winship asserts that TANF’s time limit
and work requirements were a central part of the law’s success. There is no reference to
any non-TANF provision of PRWORA.
In “Was Welfare Reform a Success?,”
PRWORA is mentioned by name, but its
description is limited to the conversion of AFDC to TANF, along with some of TANF’s
main provisions – the block grant, work requirements, and time limits. There are also
several references to AFDC waivers prior to the 1996 law as “welfare reform” – again,
suggesting the focus is cash assistance.
Even in Poverty After Welfare Reform there is virtually no discussion of PRWORA’s
In short, there is nothing in anything Winship has written that would suggest “welfare reform” is
anything other than TANF. Winship acknowledges that he has not been clear about this:
I tried to make this clear in my previous essay, but I guess I didn’t do a great job. The
question is “what would’ve happened if PRWORA hadn’t passed? Would kids be better
off?” It’s not, “What would’ve happened if we’d designed specific parts of TANF
better?” I mean, everyone’s free to ask that question and attempt to answer it, but it’s
different from asking whether “welfare reform” was a success.
The problem with this refinement is that the effects of PRWORA as a whole are of less interest
than the effects of TANF, because the latter is what many believe exacerbated poverty and what
conservatives believe to be a model for reform. And, virtually everyone who discusses “welfare
reform” is referring to TANF. Winship cites Senator Moynihan’s concerns about the 1996 law
The late senator Daniel Patrick Moynihan famously surmised that the new five-year time
limit on eligibility for federal cash benefits “might put half a million children on the
streets of New York in 10 years’ time.”
Clearly, Senator Moynihan was worried about TANF. Similarly, when Speaker Ryan discusses
“welfare reform” today, he thinks of TANF:
In 1996, we created a work requirement for welfare. But that was just one program. We
have to fix all the others now.
Obviously, Speaker Ryan doesn’t believe we “fixed” the non-TANF programs in PRWORA, as
he said just “one program.” This belief is reflected in his various budget proposals, with
proposals to block grant SNAP and Medicaid.
Virtually all conservatives who use a pre-post analysis of poverty rate trends to assess “welfare
reform” refer to TANF, not PRWORA, and they use the same poverty statistics as Winship to
make their case (without all Winship’s adjustments).
Robert Doar of the American Enterprise Institute (AEI) wrote, “TANF has been a success
– Let’s make it better.”
He cited the decline in child poverty rates.
Rachel Sheffield and Robert Rector of The Heritage Foundation wrote, “[T]he 1996
welfare reform is a rare example of a policy that actually reduced welfare dependence
and poverty while cutting welfare costs.”
They went on to describe TANF and the
decline in child poverty rates in the immediate aftermath of the law’s passage.
Ron Haskins highlighted the decline in child poverty rates after the law passed,
describing the legislation as follows: “A primary goal of the legislation was to help,
encourage, and cajole mothers to work. The law did three things to try to increase work
rates: it ended the legal entitlement to welfare payments, thereby clearing the way for
cash benefits to be contingent on working or preparing to work; it placed a five-year time
limit on receipt of cash welfare for most mothers; and it required states to place half their
welfare caseload in programs designed to help recipients find work or prepare for
He described TANF, not PRWORA.
Those who read Winship’s papers are likely to come away with this same conclusion. And,
indeed, many of them do:
Ramesh Ponnuru, a senior editor for National Review and a Visiting Fellow at AEI,
recently claimed, “Welfare reform may be the last great bipartisan success story.”
Scott Winship, a careful researcher at the conservative Manhattan Institute, has a
new report arguing that welfare reform deserves to be remembered fondly. It
imposed work requirements on many welfare recipients and gave states more
flexibility in how they spent welfare funds.
This is a description of TANF, not PRWORA. (Work requirements were added for able-
bodied adults without dependents in the SNAP program, but this is irrelevant for an
analysis of child poverty, the main focus of Winship’s analysis.)
In “Did Welfare Reform Work?,” Danny Vinik of Politico wrote:
In a major new report released Monday, Scott Winship says … that the law has
actually succeeded in reducing child poverty…Clinton’s reform, officially the
Personal Responsibility and Work Opportunity [Reconciliation] Act, created a
new block grant, known as Temporary Assistance to Needy Families, to replace
the old welfare program, called Aid for Families with Dependent Children, which
offered financial assistance to low-income single mothers and which critics
argued fostered dependency. The TANF program imposed work requirements
and time limits on welfare beneficiaries.
Again, this describes TANF; there is no reference to non-TANF provisions of PRWORA.
If Winship really believes that his causal analysis is about PRWORA and not TANF, then he
should correct everyone who misinterprets his findings. For example, Vinik wrote:
The paper offers further support for Republicans like House Speaker Paul Ryan (R-Wis.)
who want to further experiment with safety-net programs, including giving states more
flexibility to impose work requirements and time limits. Many Republicans would also
like to turn programs like food stamps into block grants.
According to Winship, this conclusion is inappropriate, because, in his words, “…critics need to
rebut my actual argument, which is about PRWORA, not certain parts of TANF that critics don’t
like (and that I’m not thrilled about either).” Vinik pulled out certain parts PRWORA,
suggesting that work requirements, time limits, and even block grants are the lessons to be
learned from Winship’s article. Using Winship’s line of reasoning, Vinik has no basis to suggest
that his article offers any support for specific provisions, whether it is the block grant or work
requirements or time limits. As described below, however, Winship himself violates his own
rule constantly, as he draws on the TANF experience to discuss “lessons” from reform, including
the importance of eliminating the entitlement and of work requirements (features central to
Conservatives promote TANF as a model for reform – not PRWORA as a whole. And, to the
extent that the focus is on child poverty and single mothers – TANF is the main driving factor. If
Winship wants to make an argument about PRWORA, he should explain the various non-TANF
provisions of the law and how they affect child poverty; indeed, he should explain why he is
limiting his analysis to child poverty.
NOTE: The debate over whether “welfare reform” is PRWORA or TANF should not be
overblown. A simple examination of poverty rate trends over time is a very weak basis
for causal inference because it ignores external factors that could influence the outcomes
of interest. If Winship’s approach were a statistical study, there would be a dependent
variable (e.g., the poverty rate) and there would be a variety of independent variables
(including an intervention variable). In this debate, we both look at largely the same data
(and I am happy to use Winship’s data) to try to assess the impact of the 1996 law. In
Winship’s case, the intervention variable (which is also an independent variable) is all of
PRWORA. In my case, I make claims about TANF and that is the intervention variable.
For me PRWORA’s non-TANF provisions are like other independent variables that can
explain changes in the dependent variable. Neither of us is doing a formal evaluation.
Our conclusions vary not so much on whether the intervention is PRWORA or TANF,
but a variety of other factors.
TANF as a Fixed and Flexible Funding Stream. Winship misunderstands what TANF really
is; he suggests that TANF is similar to the welfare reform that preceded it – AFDC with waivers:
But I have been squishy in sometimes arguing that the pre-PRWORA waivers also
constitute “welfare reform.” I suppose what I really mean to say is that the waivers –
which made AFDC less appealing, generally, but not always – were beneficial AND
PRWORA was beneficial. That is, I think the waivers beat what we’d have seen absent
any waivers, and I think PRWORA beats what we’d have seen if we just continued the
existing waivers, because it promoted work and independence (thereby reducing child
poverty) in states that would not have adopted the tougher parts of PRWORA.
As described below, there are important differences between AFDC with waivers (real welfare
reform) and TANF (a form of revenue sharing with a myriad of dysfunctional federal
requirements). One important difference is that the waiver approach included a requirement for
rigorous evaluation (generally a randomized control trial). We could actually determine if
“waivers beat what we’d have seen absent any waivers.”
The creation of TANF eliminated the
requirement for rigorous evaluation, giving states a blank check with virtually no accountability.
That is why we are having this debate.
Winship’s comparison of AFDC with waivers to PRWORA is really a comparison to TANF.
The non-TANF programs in PRWORA either had no waiver authority (and weren’t part of the
waiver-based reform process) or limited waiver authorities. AFDC with waivers represented real
“welfare reform.” These waivers permitted states to test changes to the AFDC program – some
state provisions made welfare “less appealing” (e.g., by adopting time limits, stricter sanctions,
and family caps), but others made it more generous (e.g., expanding earnings disregards, asset
limits, and eligibility for two-parent families).
TANF is not “welfare reform” in the same way AFDC with waivers was. While TANF sent a
symbolic message about the importance of work requirements and time limits, in practice,
neither of these elements have been implemented in the way Congress intended. In fact, TANF
is not “welfare reform” at all, but a fixed and flexible funding stream that has failed to provide an
adequate safety net or an effective welfare-to-work program.
Perhaps the best way to think about TANF is to divide “welfare reform” into various phases –
this highlights the fact that “welfare reform” is not the same over time. This is important for
those who say TANF’s lessons should be extended to other safety net programs. The so-called
“lessons” vary from period to period – and, state-to-state.
Pre-TANF (1992-1996): This is a period during which states continued to implement the
JOBS program of the Family Support Act of 1988 and many began experimenting with
reform through waivers. The waiver approach was subject to “cost neutrality,” which is
very different than a fixed block grant. During this period, the focus of “welfare reform”
was on experimentation with cash assistance and welfare-to-work programs.
TANF I (1997-2000): This period is one in which TANF’s provisions were being phased
in – the required work rates were low (often 0%) and no one had reached TANF’s 5-year
time limit. Most states focused on traditional welfare reform activities, as they generally
continued their waiver policies. This was also a period in which states received a large
windfall in federal resources, because the block grant was based on historic funding
levels when caseloads had peaked. The GAO estimated that if all states had received the
full TANF block grant in 1997 (some didn’t because they had until July 1, 1997, to
implement TANF) they would have $4.7 billion more than they would have spent in 1997
under AFDC: “On average, given the actual caseload in 1997, we estimated that states
would have had about 25 percent more budgetary resources under TANF than they would
have had under AFDC funding rules.”
“TANF I” was implemented in the midst of
strong economic growth, expanded aid to the working poor, and a large windfall of
money. It is therefore not surprising that TANF looks successful during this period.
TANF II (2001-2007): The windfall disappears, as inflation erodes the value of the block
grant and states begin changing the way they use TANF dollars, partly in response to a
congressional directive. During “TANF I” many states had built up large reserves
because they did not spend all their block grant funds, saving a portion for the future. On
March 16, 1999, former Rep. Nancy Johnson, then chair of the House Ways and Means
Subcommittee on Human Resources, wrote individually to all 50 governors warning that
more TANF funds needed to be spent or states risked having Congress take back some of
the unspent funds or would have future grants reduced.
From this point, states begin
shifting more of their TANF/MOE funds to activities other than traditional “welfare
reform,” such as child welfare, preK, and college scholarships. Notably, this does not
necessarily represent new spending, but often the supplantation of existing state
TANF III (2008-present): Inflation continues to erode the block grant and the rise in
poverty associated with the Great Recession makes TANF a weaker safety net. In many
states, TANF continues to become even more of a form of revenue sharing than a safety
net for needy families. Prior to TANF, most states spent about 70 to 80 percent of their
funds on basic assistance. In FY 2014, 10 states spent less than 10 percent of their funds
on basic assistance, 24 states spent less than 20 percent on basic assistance, and 40 states
spent less than 30 percent on basic assistance.
And, in most states, this spending hasn’t
been diverted to work-related activities, as this accounted for just 8 percent of spending.
The changing nature of TANF has important implications for assessing TANF’s effectiveness.
…child poverty declined, and this decline occurred concurrently with dramatically falling
welfare rolls and increasing work among single mothers. That doesn’t prove causality by
any stretch, but my conclusions about the Personal Responsibility and Work Opportunity
Reconciliation Act (PRWORA) would be very different if all three of these trends hadn’t
The positive concurrence of trends Winship is referring to occurred only during the “TANF I”
period. Between 2000 and 2013, the employment rate for single mothers fell (from 72.8 to 65.3
percent) and the poverty rate for single mother families rose (from 31.8 percent to 38.0 percent)
– meanwhile, the TANF caseload fell another 25 percent.
It’s easy for TANF to be a success
when the economy is strong and states are given a windfall, as in “TANF I.” But Winship’s
“concurrent trends” argument falls apart after 2000. In short, even if one believes TANF was
successful at one time (most notably during the “TANF I” period), it cannot be successful now.
I demonstrated this with an example in TANF is Broken! by comparing the plight of Governor
Walker of Wisconsin and the challenges he faces (during “TANF III”) compared to those of
Governor Thompson (during “TANF I”). See Exhibit 1. Winship should explain to Governor
Walker and Speaker Ryan just how TANF (or PRWORA) is a success now, since the number of
poor families with children in Wisconsin has risen since 1996 and the state keeps failing to meet
its work requirements.
The foregoing discussion is about how TANF has changed generally over time. It is also
important to understand that its implementation has varied enormously across states. Some
states, most notably California, still focus on core welfare reform purposes; others, like Texas
and Georgia (to name but a few) have essentially eliminated cash welfare and instead use TANF
as a giant slush fund. Winship looks at national data, but to understand TANF, one needs to look
trends and implementation on a state-by-state basis.
A Tale of Two Governors: The Best of Times and the Worst of Times
(The Wisconsin Story)
Wisconsin, a state long lauded by conservatives for its welfare reforms, is a prime example of
TANF’s differential effects over time. To see the impact of a fixed funding stream, compare the
deal Governor Thompson got when TANF was created in 1997 to the deal Governor Walker got in
2012. Governor Thompson benefitted from a windfall of over $100 million in federal funds (in
2014 dollars), compared to what the state spent in FY 1996, because Congress based the block
grant on the FY 1994-95 period, when caseloads were much higher. But, caseloads had already
started coming down before TANF and would have continued to decline without it, due to the
strong economy, expanded aid to the working poor, and the waiver-based welfare reform. Indeed,
the state continued to receive a federal windfall through the mid-2000s, and from FY 1998 through
FY 2006, it faced a work participation rate requirement of 0 percent.
Governor Walker was not so lucky. He not only has a TANF block grant that is about 30 percent
lower than the 1997 value, but he had to deal with a 38 percent increase in poor families with
children. As a result, he had about 50 percent less federal funding per poor family (and that is
before adjusting for the fact that state politicians have diverted TANF funds to fund other
activities). And, unlike Governor Thompson, he has faced a 50 percent target for the work
requirement from FY 2012 through at least FY 2014, a rate the state did to meet. Wisconsin thus
faces the prospect of significant financial penalties.
Except for the initial group of governors who were around when TANF was enacted, TANF is a
bad deal and getting worse each year.
A Tale of Two Governors: The Best of Times and the Worst of Times
Gov. Thompson (1997)
Gov. Walker (2012)
TANF Block Grant (2014$)
Windfall/Deficit vs. 1996 (2014$)
# of poor families w/children
$ per poor family w/children
Work Rate Targets
Sources: CBPP for poverty data; GAO for state-specific 1996 spending and block grant amounts. For Wisconsin work rate targets, see:
Wisconsin Legislative Fiscal Bureau, Wisconsin Works (W-2) and Other Economic Support Programs, January 2015. Wisconsin’s
deficit in FY 2012 is relatively smaller than most states because it got one of the biggest windfalls when TANF was enacted. This
deficit will continue to grow in the future.
Welfare Reform vs. the Safety Net. TANF cash assistance is just a small part of the broader
means-tested safety net programs. In 2014, spending on cash assistance (federal and state)
accounted for about 1 percent of federal-state spending on means-tested programs (e.g., TANF,
SSI, SNAP, assisted housing, EITC, Child Care Tax Credit, and Medicaid/CHIP). Even
excluding Medicaid/CHIP, TANF cash assistance was just 2 percent of total spending on these
Winship and I agree that the focus of “welfare reform” does not include the broader safety net,
with one very important exception. Winship states:
One last point: while I think that the EITC expansion was also helpful (as discussed in the
piece), I don’t include it or other safety net expansions in “welfare reform” – which I
define as specifically relating to PRWORA – UNLESS there were safety net expansions
that occurred BECAUSE OF PRWORA.
Winship never explains how he determines whether a “safety net expansion” was “because of
PRWORA,” and his analysis of poverty rate trends essentially takes credit for all of the
expansions in these programs. This issue is discussed more below.
“Facts” vs. Causal Claims
Winship recognizes the challenges associated with making causal statements about “welfare
reform” and its impact on poverty:
The real question of interest is not whether welfare reform was the only cause of
declining child poverty, nor whether it was the best conceivable way to reduce child
poverty. The important question is whether welfare reform reduced child poverty or not
– are children better off today than they would be if we had not passed welfare reform?
That is a difficult question to answer because none of us knows what the counterfactual
would have been in the absence of reform. It would be silly for me to claim to have an
I don’t think PRWORA was the single best policy that could’ve been enacted. But it was
better than no PRWORA. That’s my only claim and I don’t make it or try to justify it in
the new paper, which is about establishing basic facts. I titled it as I did, without causal
inference, for a reason. My conclusion discusses the question of how difficult it is to
Despite these few cautious statements about causality, Winship frequently makes causal
statements and does “claim to have an airtight case,” as in the title of his article “Welfare Reform
Reduced Poverty and No One Can Contest It.”
Winship’s “basic facts” are limited to his
construction of poverty trends adjusting for a variety of factors, most notably non-cash benefits
and tax credits. This part of his work does make for a reasonable discussion of poverty after
“welfare reform.” The problem arises when Winship asserts that “welfare reform” caused
poverty rates to fall, as in articles like “Welfare Reform’s Success and the War on Immobility”
and “Happy birthday, welfare reform: The law signed by Bill Clinton in 1996 helped millions of
American families rise out of poverty.”
In Poverty After Welfare Reform, he suggests that
TANF’s “lessons” should be extended to other safety net programs:
The idea that rolling back welfare reform would help the poor is wholly unjustified by the
evidence. Obviously, much depends on the details of future proposals, but the facts do
not even imply that extending the lessons of welfare reform to other safety-net programs
would be harmful to the very poor.
Winship’s analysis of poverty rate is not a basis for making causal inferences about the effects of
“welfare reform.” There are many factors that affect poverty beyond “welfare reform” – the
economy; the expansion in health spending and aid to the working poor; changes to other
policies (e.g., periodic increases in the minimum wage); and changes in drug use, crime, teen
pregnancy, and other social behaviors. In the 1980s and 1990s federal and state governments
spent well over $100 million conducting random assignment experiments to isolate the effect of
welfare reform policies on outcomes like poverty. Apparently, Winship believes he can do this
simply by looking at a trend line – disentangling the impact of the economy, demographic
factors, and other variables. He goes well beyond providing “basic facts”; he makes unsupported
causal claims and worse, offers policy recommendations based on those claims.
Unlike the AFDC waiver experiments, which could be evaluated using random assignment to
assess their impacts on welfare dependency and self-sufficiency, TANF cannot be evaluated in
this manner (or using other conventional evaluation methods). TANF is not a “program”; it is
first and foremost just a funding stream. There is no counterfactual that could be constructed in
any rigorous way to determine its impacts. To assess TANF, it is not only important to examine
trends in key outcomes, but also to examine how the law is written, how the policies are
implemented, and apply a good dose of common sense. That is the approach I take in TANF is
Broken! It’s Time to Reform “Welfare Reform.”
Addressing Winship’s Questions and Comments
Scott Winship: So do I have this right? We agree that:
(1) welfare reform was a lousy job training/jobs program,
Note: In a separate response to Shawn Fremstad of the Center for American Progress Winship
It’s one thing to ask to define terms, though, and another to ask that I adopt someone
else’s terms though. My frustration with Germanis is that he refuses to engage me on
my terms. He seems to want me to say welfare reform failed because TANF is a lousy
work program. But the former doesn’t necessarily follow from the latter.