American Enterprise Institute

November 14, 2008

[Edited transcript from audio tapes]

8:30 a.m.	Registration and Breakfast
9:00	Introduction:	Lawrence M. Mead, AEI and New York University
9:15	Panel I:	Field Researchers
	Panelists:	Tom Gais, Rockefeller Institute of Government
		Pamela Holcomb, Columbia Pike Revitalization Organization
		Irene Lurie, Rockefeller Institute of Government
		Karin Martinson, Urban Institute
	Moderator:	Ron Haskins, Brookings Institution
10:15	Break
10:30	Panel II:	Evaluators
	Panelists:	Dan Bloom, MDRC
		Demetra Nightingale, Urban Institute and Johns Hopkins University
		LaDonna Pavetti, Mathematica
	Moderator:	Douglas J. Besharov, AEI and University of Maryland
11:30	Panel III:	State Studies
	Panelists:	Stephen Camp-Landis, New York University
		Jason DeParle, New York Times
		Lawrence M. Mead, AEI and New York University
	Moderator:	Sheila Zedlewski, Urban Institute
12:30p.m.	Adjournment

Proceedings:

[Note: There were participants who spoke far from the microphone, hence, some words and phrases are indiscernible.]

Lawrence M. Mead:  What I’m going to do is summarize the questions, the mystery, really, that we’re addressing today and then I’ll introduce our panel leaders and they will introduce the individual speakers.

Now, welfare reform could have meant many things but in the ‘90s, what it primarily meant was efforts to move the term “welfare” into a work program.  What this meant specifically was the enactment of stronger work requirements for welfare mothers who were judged able to work; also time limits, limiting families to five years on the rolls; and greater benefits designed to support work, especially the EITC and child care.  These are both increased sharply.  The new demands and benefits came mainly from these three major pieces of legislation, the Family Support Act of 1988 which began the buildup of work requirements, the Personal Responsibility Act of 1996 which sharply increased those requirements, also strengthened child support, and then increases in the Earned Income Tax Credit in 1990 and 1993.

So welfare reform came to mean a combination of help and hassle, a designed above all to move welfare mothers into jobs.  Second only to the work goal was reducing dependency.  The hope was that if more of them went to work, we would see a reduction to the welfare rolls which recently had increased in the early 1990’s.

Early research had suggested that getting most welfare mothers to work or leave welfare would be difficult.  Statistical studies suggested that most of the mothers face substantial barriers to employment.  Evaluations of experimental work programs in the 1980’s and early 1990’s had suggested that these programs would likely have incremental effects on the caseload.  They might produce somewhat more employment and earnings by welfare mothers but this was insufficient to move most of them off welfare.

Here are some examples of some of the leading experiments: the initial San Diego Job Search and Work Experience Demonstration showed important, useful effects on employment but actually no impact on the caseload in that experiment.  The later Saturation Work Initiative Model in San Diego showed quarterly effects on the caseload of as much as 13 percent but no impact over the full course of the experiment, rather surprising.  In the GAIN Study in California, there were impacts of four or five percent in two of the three years and ten or 11 percent in Riverside, which was the strongest county of the six.  In the National Evaluation of Welfare-to-Work strategies involving 11 programs, the impact on AFDC in the last quarter of the second year was five to 13 percent, 22 percent in Portland, which was the strongest county in that study.

Anyway, the impacts are incremental.  So work programs that we are doing -- and that’s why we went ahead down that road towards this as a basis for welfare reform but they weren’t really expected to transform welfare.  But in fact, the caseload fall was much greater than anyone anticipated, even the proponents of reform.  The rolls peaked in 1994 and had fallen pretty much ever since.  The overall decline through 2007 is 66 percent for cases, 70 percent for recipients.  This is quite unusual.  In social policy, usually, its experimental programs outperform the program when taken to scale.  If you have an experiment, you usually have the especially able staff and you have often more funding and you do a very good job and then you take the program to scale to serve the whole caseload and you find smaller effects typically.  That was especially true in Head Start, for example, and many other cases.  So the experiment outperforms the program taken to scale.  But in this case, the structural form outperformed the experiments.  That’s the central issue.  Why is that?

Here’s the diagram we’ve often seen showing the trend in the caseload and what’s notable here is that the decline which -- this is what we’re really talking about here -- this period from 1994 down, but what it seems to have done is reverse, asking how did this initial increase occur in the 1960’s and early 1970’s and then we somehow managed to reverse that in the recent period.  So we’re back to levels of dependency characteristic of the late 1950’s.

Now, one question one might raise, the experiments were short.  They ran for a couple of years whereas the caseload fall was something that occurred over more years so maybe the rate of change isn’t so different.  However, if you look at the rate of caseload change, year to year -- these are calculations of change in the caseload each year over the previous year starting in 1994 to 1995 and you see changes that vary considerably but in many years, they’re quite large, much larger than we saw in the experiments, and interestingly, there is some large decreases even in the last few years.  So in percentage terms, the caseload fall in the real world appears to be greater than we saw in the experiments.

A number of experts have noticed this discrepancy.  For example, there is a very thorough summary of the effects of welfare reform by Jeff Grogger and Lynn Karoly, it’s called Welfare Reform:  Effects of a Decade of Change, and in this they note that the caseload fall was much greater than occurred under the waiver programs under AFDC prior to 1996.  So something about the structural reform is different from what we saw in the experiments, but no one has addressed this, no one has explained this and that’s what we’re trying to do today.

Now, notice, we’re not asking simply why did the caseload fall?  We have a number of econometric studies that address that.  They all, in various ways, say that the fall was due to a combination of the new welfare work requirements, excellent economic conditions in the late 1990’s and the support benefits and especially the EITC.  Studies differ about the relative importance that they give to these factors but they all mention those three and there is substantial consensus and that’s what’s involved. 

Now, we’re not getting into that question here, we’re not re-opening that issue.  Instead our issue is -- two issues:  Why was the caseload fall so much greater than previous research suggested?  How did the factors cited, in the statistical studies and perhaps other forces, interact at the local level to produce what amounted to a revolution?  What was the synergy that came together in the late 1990’s to produce this unexpected decline?  And secondly, why were there large diversion effects?  By diversion, we mean the tendency of many welfare mothers either to leave welfare before they were told to do so or to avoid welfare entirely by going directly to work.  This tendency, everyone and all observers agree, was a major factor in the caseload fall.  What was it that generated diversion?

Now, the issue is important because what it shows is that occasionally social policy -- which we often think is a frustrating endeavor -- sometimes social policy can have systemic effects.  It can produce much larger change than we think and that the change is more than marginal and that certainly was the case here.  How do we apply these lessons to other issues in social policy that we might contemplate today?  Especially the problems of poor men, are getting a lot of attention right now.  I’m working on that subject this year here at AEI.  My question is, how can we expand work programs for poor men so that they can be reached in the way the welfare mothers were?  In short, how can we have welfare reform for men?  How can we achieve comparable systemic effects for this group as we did for welfare mothers?  So to put it colloquially, we’re trying to figure out what happened in welfare reform and they were trying to bottle it and sell it.  That’s what we’re trying to do, we’re trying to bottle it and sell it.

 
Panel I – Field Researchers

Lawrence M. Mead:  Okay.  Today, we’re hearing from field researchers, by that mean people whose research on welfare reform involved direct contact either with the recipients or with local officials running welfare.  People who had what I call hands-on contact in what was going on rather than simply statistical analyses.  There are the people best able to speak to the issues that I pose.  And I’ve drawn especially on people involved in two projects which comprise much of our knowledge about the implementation of reform.  First, the Assessing the New Federalism Project of the Urban Institute and the State Capacity Study at the Rockefeller Institute of Government and I’m just delighted that virtually all of those that I asked to participate from those projects and others have agreed to be here today.  We really have a tremendous lineup.

Now, let me introduce our panel chairs.  First panel, Ron Haskins is a senior fellow at Brookings and also an adviser to the Annie E. Casey Foundation.  Ron was the chief staff person who drafted the Personal Responsibility Act.  In short, Ron was present at the creation of our current system.

Doug Besharov is a Jacobs Scholar in Social Welfare Policy at AEI.  He is also a professor at the School of Public Policy at the University of Maryland and he is currently president of the Association for Public Policy Analysis and Management.  Many of us heard him give his presidential address in Los Angeles just last week.  Doug has been a leading force in National Anti-Poverty Policy for over 20 years.

Sheila Zedlewski heads the Income and Benefits Policy Center at the Urban Institute where she was closely involved in UI’s project on Assessing the New Federalism.  Sheila, with all that input, we expect you to solve our question today and come up with the answer.  I’ll ask Ron, Doug, and Sheila to introduce their speakers as each panel begins.  So our first panel is Ron Haskins.

Ron Haskins:  Thank you, Larry.  All right.  So, the first panel, as Larry already said, has had direct hands-on experience, working not only with welfare clients themselves but with the caseworkers and with the administrators that are above the caseworkers so it’ll be interesting to see their take on this issue.  And I just want to say at the beginning, I really appreciate Larry having this session because this is an issue that I have pondered for a long time about why wasn’t the social science accurate, especially in view of the fact that almost always, when we do random assigned studies and get effects, then when you generalize and do it in ten sites or 20 sites or try to do it nationally, the effects are deminimus.  I offer the difference between Abecedarian and Perry Preschool and Head Start as an example.  And this happens repeatedly.  So here is a case of the opposite, the only one that I know of -- there may be others but -- so it’s a fascinating question.

So to address it, we have Irene Lurie from the Rockefeller Institute of Government at SUNY at Albany, and Tom Gais who is also at that same center.  And as you will probably find out, Tom is going to have very brief remarks.  And then we have Pam Holcomb and Karin Martinson from the Urban Institute.  Their biographical information is in your folders, and it’s my understanding that Karin is going to talk and Pam’s going to answer questions, is that correct?  That sounds like a good division of responsibility.

So our agenda will be they’ll all each have 12 minutes to talk.  I will be absolutely Germanic in keeping time even though Gais is not here -- we’ll punish him later.  And then I will ask some profound, penetrating questions and then I will turn it over to the audience to do the same thing and I request from the audience, we’d like to have questions and not big theories and so forth and so on.  Try to have pointed questions.

Lawrence M. Mead:  Or theories.

Ron Haskins:  Theories are okay.  Your event, okay, so talk as long as you want to.  You’ll regret that.  We’ll begin with Irene Lurie.

Irene Lurie:  Okay.  Great.  Thank you.  I did a study of TANF looking at the frontlines of the welfare agencies, actually looking at the encounters or the conversations between workers and people who came to apply for welfare and we did the observations in January to June in 2000 when caseloads were still dropping fairly quickly.  And by the way, I published the results of this study in a book that is published by the Rockefeller Institute of Government.  We looked at four states, Georgia, Michigan, New York, and Texas, three sites in each of these states except for New York where we only looked at two, and we observed almost a thousand encounters. 

Most of the encounters that we observed were people who were applying for TANF and then quite a few people who were being re-certified and then a few people who were leaving TANF, and then we also looked at what was happening when people are applying for childcare, when they were being sanctioned, when they just simply were at the application desk, the front desk, asking for an application.

So, to address Larry’s questions, I want to distinguish between four groups of potential TANF recipients, and the first two groups in green are people who actually became recipients and the groups in red are the people who for one reason or another were diverted, did not become recipients.  So group one is treated like the applicants under the standard, old AFDC program.  They applied for welfare, their application was completed and then they were subject to various mandates, work mandates including job search and other mandates but only after they became a recipient, okay?  So they were not diverted.

The second group applied for welfare and their application was pended, it was held up until they satisfied some sort of a Work-First requirement.  And I’ll talk a little bit about what that looked like soon.  Then if someone complies with that Work-First mandate, they become a recipient.

The third and fourth are the groups that Larry is particularly interested in and these are the ones that, for one reason or another were diverted from welfare, did not actually get on assistance.  So, the third group is like the first group; they applied, their application was pended and they, however, did not satisfy the various mandates.  They did not go to Work-First or they were already working and so they couldn’t go to Work-First or they just said, “Hey, if I’m going to have to work when I’m on welfare, I might as well just not go on welfare.  I’ll get a job on my own.”  Or maybe they decided that they would try to get more support from family members and friends.  So, this group was truly the “diverted.”

And then, we have the fourth group: I call them people who didn’t apply due to neighborhood effects.  These are people who got information from people they knew in the neighborhood or from the media that told them that there would be a Work mandate and other behavioral mandates and so they didn’t apply for welfare at all.  They never showed up at the welfare agency and, is this the group, Larry, that you’re referring to as the systemic impacts?  There was other information or something else happening in the broader system that caused people not to apply.

Now my study did not try to learn why people didn’t apply, we didn’t look at that group.  However, we can ask how much information did applicants seem to have when they came to the office.  And I would say that for the most part, they did not seem to have much information when they came to the welfare office, that they learned about the rules there and this was a first time introduction to the rules so the neighborhood effects, at least for that group, were not tremendous.

Another piece of evidence is that administrators in the welfare offices said that large numbers of people were still coming to apply for welfare, and they complained about caseloads of applicants.  So that people were still applying but it’s possible that people who came to apply were not actually counted in the statistics of people who came to apply because they never finished their application.  Their application got pended and so they didn’t show up in the count of applicants.  I mean, that’s a possibility and how applicants are counted is going to vary from one place to another, so data on applications to the extent that it exists may not be a true measure of everyone who came to the welfare agency.

Now, another question we could ask by looking at these four groups is the question of how TANF compares with the experiments that were done that Larry referred to and we need to ask when was the random assignment in these experiments conducted and, looking again at the first group, I think the random assignment was done after they became recipients.

Male Voice:  [indiscernible]

Irene Lurie:  Well, I don’t know.  Well, that’s a question.  How many studies was random assignment done on applicants versus recipients, and maybe people this afternoon or later today can answer that question because that’s really critical.  If the random assignment was done after people became recipients, you’re not going to see diversion, right?

Well, just briefly, I just want to talk a little bit about what the Work-First diversion program looked like.  And in Georgia, the goal was explicitly to divert people and people were asked to search for work from between 3-6 weeks before they could be eligible for TANF, they were pended for that period.  Texas, the goal was also explicitly diversion and there was an orientation at the workforce agency before someone could be eligible for TANF.  And then in Michigan, it was simply, the goal was not to divert people but to try to get people into jobs, and again, it was just an orientation.

But there were other factors that supported diversion.  In Georgia and Texas, mothers had to get their children immunized in order to get on welfare because they were going to be required to work and put their kids in childcare and unless they were immunized, the childcare agencies wouldn’t take them.  So they had to get their immunizations done and I heard a lot about that in the encounters.  There were also Personal Responsibility Agreements that people had to sign, child support enforcement which, of course, is not new.  Then importantly, there were funds for childcare that, as the caseloads fell, it freed up money that agencies could use to pay for childcare, Medicaid and welfare were de-linked so that you didn’t have to be on welfare in order to be eligible for Medicaid and that was explicitly told to people.  There were also statements by the welfare agencies and the workforce agencies that there a lot of jobs out there and so they were very encouraging to the clients about the prospects of employment.

And then finally, as I said earlier, some people were already working and they had other sources of support.  I also saw a lot of effort devoted to fraud detection and this had been increasing over the last years of AFDC and the early years of TANF, much greater access to the records of the employment agencies to learn whether people were actually working.  New York and Texas took electronic fingerprints of everybody which had a chilling effect on certain people.  Michigan did home visits and Texas had a tremendous amount of data on people that they had purchased from a company that provides credit card information so the welfare workers knew whether someone had just paid off part of their credit cards.  Well, if you just paid $200 on your credit card, where did you get that money?  So there was a lot of information.

And then finally, and this doesn’t address the diversion point but at re-certification, clients were told that they would have to work.  The time limit did not actually throw people off assistance during the period of my study.  Time limits hadn’t kicked in but workers invoked, they talked about the time limit and said, there is a time limit and you better get yourself ready to work for when your time limit is exhausted.  And then finally, there were sanctions that were imposed on people, much tougher sanctions than earlier.

So I really mostly want to draw our attention to this question about exactly when the random assignment occurred in the other studies and that might help answer some of Larry’s questions.  Thank you.

Ron Haskins:  Karin Martinson.

Karin Martinson:  Okay.  I’m Karin Martinson and I’m going to talk about some of the work we did at the Urban Institute that I did with Pamela Holcomb who has subsequently left the Urban Institute so I’m going to speak for both of us but I want her to feel free to pipe up, and we’re glad that she’s here because she directed a lot of these studies.

So, I kind of stepped back a little bit when I read the questions and I think these are some of the things Larry was raising about what caused the caseload to decline.  I know the conference is about diversion but I think a lot of these things tie in and they’re hard to talk about separately.  So just kind of my take on what the different factors and how they played out, well just a couple of things.  There are the economic factors that Larry mentioned which were particularly strong in the late 1990’s but I think what we’re seeing with the recessions in the early 2000’s, and the current recession, I kind of wonder if caseloads are ever going to come back up so I am raising questions about how much -- that I think it did play in but I don’t think you see the rise in caseloads when the economic conditions deteriorate.

I think strong work supports were also important.  I think Irene mentioned one that I was going to highlight which was de-coupling TANF from Medicaid because that was a big change.  I mean, Medicaid is very important to families and I think the fact that they could get that without TANF was a big change.  And then, of course, there are all the other things we’ve mentioned like EITC, childcare, and the earn disregards.

I’m mostly going to talk about the TANF program design elements because that’s what we studied in our program, but I think there are four of them: the work requirement, sanctions, diversion, and time limits, and the last thing I want to highlight is just that the law did have strong incentives for states to reduce caseloads and that’s kind of how their success was measured.  So in some ways I think it’s not exactly surprising that that’s what happened.  I think the caseload reduction credit that states can use to reduce their participation requirements was very important.  Because I worked at HHS when the law was passed, I think there was not a lot of attention to that particular feature when it passed but I think, the states understood how that could affect their participation, that was an issue.

I also think there is the block grant structure where states keep their block grant regardless of whether caseloads decline, was an important issue.

So we did several studies looking at how programs were working in the field.  The main one, as Larry mentioned, was Assessing the New Federalism which took place between 1996 and 2000.  Here we did two rounds of site visits to 17 cities in 13 states.  We also did several other studies that primarily Pam worked on but we were in the field a lot during this early period looking at kind of how different features of the program were working.  I would emphasize that we were there in the initial years of Welfare reform and I think things could be different now and I haven’t been in the field as much lately.  But this was, like Irene’s work, it was during the period when the caseloads fell.  In our work, we did administrative interviews and onsite observations with program staff at all levels from management down to line staff and we looked at intake, Work-First programs, exemptions, sanctions, and a lot on organization and service delivery.

So I think, based on the program side, what we saw -- and I kind of put these in order of what I think the programmatic pieces that affected the caseload decline and I think first was the stringent work requirements and limited exemptions.  I think families were expected to do more and, before, I think there had been a large exempt population -- not large but at least substantial.  And now they are required to do something.  And I think as time moved forward and families realized these requirements that affected their ability to perhaps even apply in the first place.

I think another big issue was sanctions.  I think this was something that was not necessarily foreseen that half of the states in the early years moved to the Full-Family Sanction and that certainly affected the caseload decline.  We did look at diversion programs, kind of looked at the front door, and our take on it -- because just in the places we were at, in most places we didn’t really see a huge diversion effort, trying to get people not to come in.  There was a very strong message about encouraging work and letting them know the conditions of welfare had changed and really enforcing the work requirements.  So in our sites, that wasn’t a big part of the story but I think it is important.  And I also think it mattered in certain places.  We had only two sites in our study that kind of had the job search requirements before you could get cash assistance but I think it was important -- and that was Wisconsin and I think Mississippi.  So I think it matters in different places.

And finally, I agree with Irene on the time limits.  In the sites we were in, there were not a lot of families reaching their time limits during this period when the caseloads were really declining but there is certainly some message in there, but I think in terms of understanding decline, it wasn’t as important.

So, I know we’re supposed to focus on diversion but I guess I had to take this opportunity to say a couple of things about -- because I think as for calling it a success and something we want to kind of bottle and sell, I think there are some issues about the people that aren’t on TANF and so I just want to raise a couple of things that I wish we should think about in our discussion because falling caseloads can be a good thing but I think -- Sheila has done some work looking at disconnected families who are not working or receiving cash or have a working spouse and about 20 percent of welfare leavers were disconnected in 2002.  So I think people are moving to work but there is a question of people that are not moving to work.

And I also think you see low take-up rate of TANF among eligible families.  Under AFDC, 75 percent of those who were poor received cash and now that’s down to about 40 percent.  So I think it’s just not clear how this program is functioning now.

And just a couple of wrap-up thoughts.  I think the states had strong incentives to reduce the caseload because of the way the program was structured and they used a variety of strategies and I would particularly highlight the work requirements which states really moved quickly to implement and also sanctions, and I think that does have an effect on families and how they perceive the system.

One person that’s not here is Kristin Seefeldt from the University of Michigan and she has done some interesting qualitative work looking at families in Detroit who were initially on TANF, and now she goes back at regular intervals, to talk to them and really none of them are on TANF and many of them are not working and they think they view the system as something they don’t want to deal with.  They go on food stamps, but the TANF, I think it’s too little money for the effort.  I’m not saying that’s true of all families but I think I just heard on the radio coming in that UI applications for unemployment insurance have jumped dramatically this week and it’s not clear to me that you’d see that as some kind of increase in TANF, and I think that raises some questions for me about how it’s functioning as a safety net program and what the caseload decline really meant.  I’m under my time.

Ron Haskins:  Yes, you are.  Very good.  Tom Gais?  Tom, I want you to know that before you came in, I introduced you and the audience booed wildly.  You don’t have to go to the podium.  You can do it from there or podium, whatever you prefer.  Do you have slides?

Tom Gais:  I don’t have slides.  I am going to speak at the podium though.  Well, since I’ve been booed already maybe I should keep this very short.  Thank you very much, and I may actually say some of the same things that Irene said since I’ve learned so much from her but in that case, Irene, you could just yell at me.  But thank you very much.  We did, at the Rockefeller Institute, a lot of field research back in the 1990’s and we have continued to do so, thanks to some of the people in this room, research more recently as well.  We did field research in over 20 states in 1997 and 1998.  We revisited some of these sites back in the early 2000’s and more recently with support from the Department of Health and Human Services and in cooperation with the Lewin Group, we revisited five states in 2006 and 2007 and those reports are available as well. 

I just want to say a few things fairly briefly because I think that there are a lot of possible things to cover and I’m not quite sure all of which are relevant.  But I thought it was very interesting that Larry raised this question about what was unique or different about TANF that could be useful, perspectively, in thinking about future policy-making, especially in comparison to the AFDC waiver programs.  And so I just have a few points that I would like to make regarding that. 

AFDC waivers and the experiments involving those waivers were I thought very different from TANF in a number of very important ways.  They generally involved a limited range of policies, usually involving different treatment of AFDC clients, people already in the system whether it be time limits or work requirements, et cetera, although there was also quite a bit of an emphasis if you looked actually, the goals of the AFDC waiver experiments, there are a lot of family formation goals involved there about reducing out-of-wedlock births and so on and teen births, et cetera.  There were limited dollars, there was a requirement, at least a nominal requirement, for budget neutrality, whatever the actual reality is, it may be a different thing, but there were limited dollars involved there.  These programs were not flexible over time for legal as well as analytical reasons.  Once you set up these AFDC waivers, they were generally fairly fixed.  There were some exceptions which hardly created headaches for the evaluators, certainly, but these programs were in essence fairly well set and remained in effect for several years. 

And the other interesting thing about a lot of them as I go back to them is that they usually involve changing the behavior and culture of welfare agencies.  One of the things that we were involved -- and it makes a lot of sense because if we’re going to make a comparison between how people are treated with respect to policies, you may want to hold constant the kind of agencies or organizations that are administering the different policies and benefit packages.  Now, there were some exceptions to this but for the most part, it was involved -- they attempted to reform traditional welfare agencies when they were doing that.

Now TANF wasn’t constrained in any of these ways and although some of the states, in fact, a good number of states acted as if they were constrained in some of these ways, many of them didn’t and didn’t act that way at all and I think that really does raise some very interesting things about the prospects or the ideas that we ought to explore prospectively.

First, many of the places we went to involved large bundles of policy and benefit changes all of a sudden, in large packages.  There were changes in earnings disregards, there were changes in asset disregards, there were changes in work requirements, time limits, et cetera, just as the other speakers already mentioned.  There were changes in childcare, Medicaid, tax changes outside of AFDC and typical welfare policies altogether including major increases and establishments of state EITC’s, and there were also changes of course in child support enforcement.  These would have made a terrible situation for any kind of evaluation experiment but they were fairly comprehensive changes in policy.

The other obvious change I think in difference was that the dollars were plentiful.  They were less plentiful to begin with but even then, there was really no budget neutrality, there was a lot of money.  Some of the caseloads were declining since 1994 already and so there was a sense even in 1997 and 1998 when we were there that there was a fair amount of money, not anything like they saw the amount of money just a couple of years after that but there was lot of money to be able to throw at these programs.  And although it varied from state to state a great deal, it was also possible to make many changes in services, programs, procedures, contracts and so on, constantly -- or maybe not constantly but frequently.  These programs weren’t in any way constrained by having to stick with bad ideas.  And there were a lot of bad ideas and there were a lot of reversals that we saw. 

We saw a Michigan program, for example, which actually sounded wonderful.  It was a summer camp where they brought in about a thousand people with multiple barriers for two -- actually I think it was three weeks with their kids.  Kids went to day-camp, the parents went to multiple training sessions, parenting sessions, sessions for job preparedness, for managing their budgets, for nutrition, for cooking, et cetera, et cetera.  In the end, they found out that the kids really liked the program, they had a wonderful time but this really had no inroads to speak of for any of the people with the multiple barriers.  This was about in 2000, it cost $20 million to do this, a lot of money, and they ended it.  This was one of the more dramatic types of rollouts and rollbacks but there were a lot of things going on like this and a lot of adjustments over time.  We could go in and Larry could talk about the whole history of contracting in Milwaukee which was a work in progress and still is to some extent but it’s -- oh, okay.  But there were a lot of bad ideas out there and it’s very useful in retrospect that they could get rid of them and try out something different.

The other thing to my mind that was intriguing when I go back and look at TANF, compared to some of the earlier AFDC waivers, is really the goal focus.  Now, not everybody will like this goal focus in the early years but it was really focused on work and employment.  The family formation issues were much less important.  They became more important later and maybe that was appropriate to be able to bring them in later after some of the work orientations were institutionalized but at first, at least in the initial years, the employment focus was very, very strong and many of the family formation issues that were evident in the waivers were really jettisoned or almost hoped for by-products of doing everything else.

Now, the other thing that was very important and I think this is especially critical, is that in many cases, states really didn’t try to re-engineer welfare agencies.  They simply added new agencies to the whole mix and created new complicated procedures and processes that brought these new actors, mostly workforce, investment to workforce development, bureaucracies or employment bureaucracies, but also a lot of private agencies in many cases, sometimes for-profits, most of the time, nonprofits, all with different, more work-oriented missions that they’ve had for many, many years.  They also brought them into new settings, different milieu that many of the people had actually seen.  This was not the case in lots of the settings but it was certainly the case in some of the more advanced states we’d bring in people into job centers.  In this case, it helped account for the rapidity of the change.  I mean, they didn’t have to retrain workers in many cases and in fact many of these AFDC welfare workers were doing all the same things, they were documenting, verifying, in fact, if anything, they were adding, as I think Irene probably would suggest, more upfront documentation paperwork requirements than had been the case before but they had simply added additional processes.  You had to show up for orientation, you had to show up to register at the new agency, often off-site, because many of the welfare workforce development agencies really didn’t want to be all that close to the welfare agencies.

Male Voice:  Could that have created diversion?

Tom Gais:  That could have created diversion.  I mean, I think that one of the interesting points about this is that there’re two ways of creating diversion when you have very complex upfront processes.  You can create diversion if these processes work.  You can create that is if people actually go to these agencies, find help, see that they’re part of a lot of the other job seekers, get assistance like they’d never had before and actually make progress towards finding a position.  It also helps diversion when these things don’t work.  So New Jersey had a lot of no-shows.  We also had Arizona, which never had a particularly good, close, connection even within its own agency between its work side and its eligibility side, failed miserably from a public administration point of view but they really helped diversion by having people simply not knowing exactly what they were supposed to do and they simply got sanctioned or failed to complete their applications.  That helps diversion as well.  So from the state point of view or from the caseload reduction point of view, it’s a pretty easy thing when you have this upfront complexity.

Now, I’m probably behind my time but -- and I’d be happy to talk about some of the examples later but I think all of these points are relevant prospectively and if you want to produce really big changes in behavior, it is useful, it probably is useful, hypothetically, to throw lot of policies and dollars at the problem, at least at first.  It’s probably very good to focus on an over-arching and well-accepted goal and not jumble it up with a lot of other things you would like to see result from your policy reforms.  And it’s probably good not to bother trying to re-engineer state and local agencies or even information systems initially -- and I could talk about the complexity of re-engineering them -- but to bring in other agencies and other systems that are already focused at that goal and simply cobble together some kind of new set of procedures that also, in a sense, present a new face of the social policy to the clientele.

And finally, it’s also very important to provide enough flexibility to jettison bad ideas and try out some new ones.  I would strongly suggest that to the extent that you actually do random assignment studies that require a number of years, you wait several years later when you get a settling down of what administrators seem to think work and not impose these staticity requirements -- I don’t think that’s a really good phrase -- in the early stages of program administration.  Thank you very much.

Ron Haskins:  Let me ask a small question first then I want to answer a bigger one.  This idea of new agencies, I think a lot of people especially on the Hill believe and maybe people in the audience that what welfare reform really did was it re-engineered welfare departments and that they made substantial changes and I know there were some that did that.  Is there quantitative information about how many programs or how many people applying for welfare or already on welfare had to go to a different agency, either state workforce agency or non-profit or a profit agency, do we know what percentage of -- if Tom’s right, this is a big part of the answer to Larry’s question.  So do we have quantitative information about what proportion of the caseload went somewhere else besides the old traditional welfare agency?

Irene Lurie:  Well, I know, in the states that I’ve looked at, if there was a work first program, everybody went.  You had to go before you could be on welfare so you might look -- to answer your question -- look at how many states had these very hardwired workforce programs.

Ron Haskins:  But the question now is, of those hardwired-- or whatever they did, how many of those were done with outside agencies because I think it’s a really crucial issues because the issues of changing a bureaucracy is a big deal for all of us on health, and every area that we’re interested in social policy and I think the common understanding is that welfare reform changed these agencies.  And Tom is saying, no, they’ve got new people that already believed in work and they just did what they always did.

Tom Gais:  That is a very strong statement of it with all kinds of qualifications but in general, that’s much more true than any other statement I could make.  Do we have quantitative data?  I don’t know.  Do you all have --?

Karin Martinson:  We don’t have quantitative information.  I think in our set of sites, we found about half started involving the workforce investment system and in general, we also saw what Tom saw, a lot of contracting with non-profits with experience providing employment.

Ron Haskins:  If that’s the case, it’d be more than half.  But if it’s half that do involve the workforce and then some additional ones got non-profit ---

Karin Martinson:  But they didn’t always have to go somewhere else.  Sometimes they would be all in the same place so there’s -- I mean, you’re talking about location.  I think there’re a lot of different models out there.  Just because you work with the workforce investment system, you could bring them all in the same place and some places did that.

Irene Lurie:  Or in fact they might just have -- a workforce development person might come in to the welfare office and do the work first orientation and leave.  Now, that’s a model where you do have a different bureaucracy coming in but it’s far different than in some of the other places where you had a much greater degree of integration or in fact that the work first development was really doing most of everything except for benefit calculation and the actual eligibility.  But I think that if you add in the use of contracting out to non-profits and to a certain degree, for profits, certainly then you have many, many more players involved in the whole welfare experience than you ever had before.

Ron Haskins:  If we review, I think everybody agreed on this, we emphasized the program really -- I think you need to turn these two microphones off, we’re getting a feedback or maybe it’s the way I talk -- there really were substantial changes.  Work requirements, sanctions.  I know a lot of people that are involved in welfare reform, especially people outside the Congress, doubted the states would really use sanctions, they’re going to be way too soft, they won’t really do it, but they did, right?  I mean, the states definitely used sanctions, not every state, a lot of but -- so that got a message across.  People actually lost their benefits.  Five-year time limit, I mean, a lot of changes that were not in any of these random assignment experiments.  Maybe one or two of them were, plus a flush statement straight out in the statute that says no one is entitled to a cash benefit, now that’s a huge difference.  So the states had a lot of flexibility to take benefits away, do a lot to make benefits contingent which they could approach and do somewhat in the random assignment experiments that preceded welfare reform but they couldn’t do anything like the whole package, and now we’re also identifying new things, like you bring in new actors that are really focused on making people work, that could make a big difference and so forth.  So when you look at all those factors, maybe there’s not as much mystery.

Irene Lurie:  To me, it is not a mystery after having looked at what happened.  It’s impossible to get a quantitative measure of how much each of these pieces contributed to the decline and I don’t think we’ll ever be able to do that but it’s --

Ron Haskins:  Larry’s going to do it anyway.  All right, here’s -- okay, go ahead.

Tom Gais:  In that list, you still didn’t mention again these upfront pre-qualification requirements which were not always very clearly stated in policy but they were a procedural requirement and I know, for example, just Georgia -- that’s an intriguing example, I know in some of the policy collections of various AFDC policies I’ve seen, there were not pre-qualification requirements but in fact, when we were just out in the field, there were -- in Macon County, for example, just to take one example, people coming in had to go through a four-week program where they would get training, they would have to make ten contacts, from the second through the fourth week, they would have to make 10 contacts with prospective job people in person per day and this was all long before they actually see a dollar.  They have a miniscule caseload.  And these things are not found in all these broad policies attached to work assistance but they’re there and those are very important and I think Irene’s study also clarifies their operations as well.

Ron Haskins:  So, let’s compare this to another common interpretation, compare making a lot of changes that really have direct effects on people like the ones you described in sanctions and a lot of other things that did exactly what they were designed to do, if people didn’t do what was required, they lost their benefits, that was the logic of the system.  And other people have said, well, you changed the whole culture of welfare.  That offices, they were check writing operations and now they’re job-finding agencies.  I mean, how many people have you heard say that and I still think that’s a predominant interpretation in the Congress.  Which one is right?  Have we changed the culture of welfare?

Irene Lurie:  I think the culture in the welfare agencies has not changed very much but that the work message is there and in order to implement the work message, typically other agencies are involved, agencies that specialize in helping people find work.  But the welfare agencies are still specializing in calculating and determining eligibility for assistance.

Karin Martinson:  I’m going to agree.  I mean, your eligibility worker is still pretty focused on collecting information and calculating benefit, making sure there are no errors but I think they bring a message of work and they bring you then to another set of services that are really focused on work.  But I think there are things like posters and you see the welfare office has certain messages but your welfare worker per se is really about eligibility.

Pamela Holcomb:  It’s the client who gets a broader message by virtue of having to do more than just fill out the application form.  But I would just say, as long -- in one sense, it’s not mystifying when you look at all these changes all together, I don’t think it’s mystifying.  But I was thinking about it and I think that in certain ways though, it is mystifying, still, to me because if you just took the New York’s and the Wisconsin’s and examples of the Georgia where it’s very clear.  There are certain places where I was sort of amazed anybody ever actually got on welfare because there were so many requirements and they were administered in such a one, two, three, if you passed one, you might not pass two and three and four.  So in that sense, I think it’s de-mystified.  But still, places that were pretty soft and didn’t necessarily have all of that, California where you didn’t have a Full Family sanction, where your time limits didn’t matter.  You have other states which were taking a much more soft approach or supportive approach and still, caseloads were going down, not a little but a lot, and that piece of it still remains kind of the mystery part to me.

Ron Haskins:  Do you think that implies a broader cultural interpretation?  Not just the office but the whole society and governors and public officials saying you ought to work, personal responsibility and that even when you implemented it softly, it still worked.

Irene Lurie:  Did you go to Hawaii?  Did you see what --?

Ron Haskins:  No!

Irene Lurie:  I think you do maybe have to go and see what’s going on there.

Pamela Holcomb:  Well, I’ll come back to the institute and do that.

Ron Haskins:  I’ll do that.  I volunteer.

Tom Gais:  Now, Hawaii is part of the country now.

Ron Haskins:  Okay.  So, time for the audience -- are you going to say something more?  Go ahead.

Tom Gais:  I was just going to say -- you still see some --I mean, the caseloads have dropped quite a bit everywhere and I was always curious about that as well although there are still some variations that make some sense.  One of the places we re-visited, just recently, was Missouri, my home state, and they still don’t really have immediate attachment orientation at all.  They’re thinking about it, they were thinking about it as of 2007, I think it was but the -- or they were thinking very seriously about it, I should say, but they hadn’t actually implemented it, and their caseloads since the 1990’s, 1996 I think it was, through 2005 dropped about 24 percent.  I mean it was nothing close to what was happening and it hasn’t declined anything like you’ve seen in a lot of the other states.

Karin Martinson:  I would just add that I think the Work piece was a piece that was implemented most consistently across the states.  I think we saw states really moved quickly on that, moved to meet the requirements, they brought in the partners.  I think other things you see in more variation like sanctions, the diversion policies, so I think it’s the work that’s really pervasive across the states, even the states that are softer, they’re still focused on work.

Ron Haskins:  Audience.  Wade Horn.  Take the microphone and introduce yourself and ask your question.

Wade Horn:  Hi, I’m Wade Horn.  I’m with Deloitte Consulting.  Much of your discussion is about what happened to people once they were on the caseload, yet diversion as I understand is mostly about people who might’ve been on the caseload but were diverted, either self-diverted because they chose not to go into caseload or were formerly diverted or diverted through a formal program such as an up front program.  What’s interesting to me is the 40 percent penetration rate of TANF which means that 60 percent of the people who are eligible for TANF, today, are not on the program, so almost by definition, there are people who diverting themselves off from the program.  What do we know about that 60 percent?  Are they informally diverting?  How much of them are informal diversion programs so hence not really on the rolls, they’re not being counted?  I mean, what do we know about the 60 percent who are eligible for TANF but are not on the rolls right now?

Karin Martinson:  We can ask Sheila but I think -- I don’t actually know the answer to that.  I think some of them are working but not earning enough.  We have this group of disconnected families that are not working and not on assistance and don’t have a working spouse.  Their means of support are unclear so I think there’s that group in there.  And I think we have groups of families just more relying on food stamps, community resources and just staying off TANF and trying to cobble together other benefits that they can get but I don’t have a breakdown but I know there’s a group of disconnected families.

Ron Haskins:  I think we know at least two things Doug may want to add to this but one thing that we know for sure, I think, is that the number of at least single moms in census bureau records over a period of a year is not longitudinal data, of course, cross sectional but a number of moms that do not have cash welfare and do not have earnings in a given year has doubled since roughly the mid-1990’s.  And then the second we know is that we shouldn’t assume that they’re in miserable shape because they don’t have that kind of income and about a half of them -- and this is Doug that has looked in this more carefully than I have -- about half of them live with someone who has an income so they’re not necessarily even in poverty even though they don’t have cash welfare income.  So as you probably know there are a lot of people in Washington who are concerned about this.  I’m sure there’s going to be legislation introduced to give states more money to deal with this particular group outside welfare or at least that’s what I know some people will propose.  Doug is that about right, about half of them living with someone who has an income?

Douglas J. Besharov:  I don’t remember the exact numbers.  You know, we’re at the mercy of the Census Bureau Data on a lot of this and those data are just not great.  They miss a lot of people at the bottom, they miss a lot of income at the bottom and we also know that the Census Bureau Data underreport TANF recipiency as they underreport Medicaid recipiency and so forth.  So, you have to take all those numbers with a grain of salt unless you’re on the Hill and then you spin your story the way you need to.

Ron Haskins:   Actually you have to do little bit more work than that, Doug, because you have to explain how the number would double unless it’s less reliable now than it was then, right?  I mean, it’s the longitudinal, that’s the change over time that you have to count for and the most straightforward explanation is that there really are people who don’t -- single moms who don’t have cash income and don’t have welfare.

Douglas J. Besharov:  Well, this is your panel but actually the trend of underreporting has increased and we know a lot of this because people are no longer in AFDC, they don’t know what they’re in.  They can be in the jobs opportunity program and not know that that’s cash assistance and the Census Bureau works very hard to get that.  The other thing just to mention here is whether you call it diversion or not, there are just a lot more people in disability programs and in other programs that we don’t count as cash assistance which provide the same kind of benefit, but I think Sheila wants to say something.

Sheila Zedlewski:  When we define disconnected, we always exclude those who have a disability benefit from any type of system, so they wouldn’t be included in that count.  I did a study where we looked at -- we did in-depth interviews of 100 mothers who were clearly eligible for TANF, they weren’t working, and we wanted to know why weren’t they on welfare, why didn’t they apply and most importantly, how were they getting by?  Getting by was interesting because of food stamps and housing assistance whether it’s public or private, can make a huge difference in your life and that combination plus a little bit of help from family members was how a lot of them were getting by. 

The other factor that we haven’t talked about here yet that came up as important is child support and while I think Larry already said that it’s certainly true that we’ve been collecting -- assigning child support for a long time in the welfare office, it was really strengthened right about the same time that welfare reform came about and there were a significant number of mothers who said, “I get a little bit of support from my child’s father,” “I don’t want to turn him over to the authorities,” “I’d rather get by with the child support,” and so on and so forth.  So other kinds of income and put a plug in here that when we count poverty, we don’t include food stamps, we don’t include housing assistance and we’re missing some very important sources of income.

Lawrence M. Mead:  This is actually relevant what I was going to say, the thing that I think we’re trying to bottle and sell today is not caseload fall but really structural change, that is we get a large change in the structure, that’s what we want to understand.  In this context, it largely meant driving welfare mother’s away from government, away from welfare.  But in another context, and in particular, child support, it might mean just the opposite, it might mean trying to get men who are now outside of government to come in from the cold, to join the child support system, see that as part of a way to reconcile with their families and help support them, so structural change might operate to increase an institution in government. 

In fact, the child support system, many people don’t realize this, has actually become the main child support collection operation for all of America, not simply poor peoples, not simply people on welfare.  The majority of child support today is collected through what was the 4D system.  So it has become, actually, an almost universal system for many people involved in child support and that is a structural change.  And to get the men to work more readily through some kind of welfare reform for men would probably mean expanding certain caseloads and certain institutions.  So structural change doesn’t necessarily mean smaller government, I think, in this case it did but we should indeed be concerned about the disconnected families but we want to keep an eye in the ball here which is how to understand that structural change.

Ron Haskins:  Time’s up.  Do we have time to take another question?

Lawrence M. Mead:  Let’s see.  LaDonna.

LaDonna Pavetti:  I just would like to ask a question on that.  There was a study in Arizona that looked at people who were diverted and they actually looked at participation in the child support enforcement system, and for those who are on [audio glitch] at the point of application, there were about five percent of TANF and of the diverted who were involved in the child support enforcement system.  When they did the follow-up which was I believe a year later, 75 percent of TANF were involved in the child support enforcement system and 25 percent of the diverted.  So diverting people is actually keeping people out of that system, at least from the custodial parent’s point, I mean, sort of getting them involved.

Ron Haskins:  Was that 75 percent [indiscernible].

LaDonna Pavetti:  No, no, no.  That was another part of it because I thought, this is big, but then if you look at payments, they actually end up looking quite similar.  So getting into that system doesn’t actually, necessarily translate into getting the payment.

Ron Haskins:  That’s because Wade used to run it.

LaDonna Pavetti:  When you get to the payment, it looks quite different and they look quite similar.

Ron Haskins:  Break, ten minutes?  Okay, 15.  Thank you.  And thanks to the panelists.

[Break]

Panel II – The Evaluatiors

Lawrence M. Mead:  Let’s return to our seats if you could.  We want to start our second panel.  Let’s return to our seats.  The break is over.  We are cracking the whip here.  Please sit down, please sit down.  We’re going to start our second panel.  Please sit down.  Our new chair, Doug Besharov, will introduce our speakers.

Douglas J. Besharov:  Hi, second good morning to you all.  A pleasure to have you here at AEI, pleasure to have Larry Mead here this year.  In preparing to introduce our sterling panel, I read their bios to underline the part that I wanted to emphasize and I’m going to introduce them one by one as they’re about to speak.  And the most striking thing about their bios -- and I’ll try to emphasize that as I introduce them -- is how much things have changed, the cultural bios have changed, Larry, if not welfare offices, and every bio talks about employment and work as opposed to -- and I don’t think I saw a mention of dependency and so forth.  It’s very interesting.  Sometimes, one reads too much into these things but there you have it.  All right. 

I hated it in grammar school when I was always first because it was size places and so I would like reverse order and we’re going to do reverse order here starting with the P’s as in LaDonna Pavetti.  Donna has been an important contributor to welfare analysis for many years now.  She is a senior fellow at Mathematica and from the beginning she has been studying the implementation of TANF.  I was quite impressed to see that her research has taken her to half the states.  My guess is that means 90 percent of the people because there are a couple of small states.  And I’m reading this, Donna is currently working on several projects designed to examine how states and local offices are changing their policies to meet the higher work participation rates of the Deficit Reduction Act of 2005.  My only advice to Donna is work fast because those participation requirements will disappear in a few weeks.

LaDonna Pavetti:  You’ll be surprised how fast we’re going to work because we’re almost done.  So there are two studies, actually, that I think are relevant.  One is I am working on a study but we also are doing a study right now of diversion, just really trying to understand what states are doing and how they’re changing their diversion policies and sort of in the more recent, I did a study of what states were on diversion policies at the very beginning of welfare reform and this is a current look.  But I’m not going to talk about that since this was really about evaluation.  What I decided I would like to do is to talk a little bit about what are the studies that have been done on diversion and what do we learn from them and what are the implications of really thinking about how do you both harness sort of the good things and think about what are the unintended consequences as well that have happened.

So first, again, people have talked about this but thinking about diversion is there is, you can sort imagine that somebody is in need, their first decision is do I go on or not, do I apply or not.  And then once they apply, there is, do I complete this process and do I go on?  And then if I go on -- regardless of what your path is, in the end you end up with people in basically three positions, oversimplifying, they’re either working, they’re on TANF, or they’re not working.  So if you think of that, I think the important thing is to think about that what do we know about diverted and is there any ways to compare them to other people, to people who’ve went on the TANF.  It’s not perfect but I think that we don’t have good experiments and it’s probably the best that we have.

First is -- so I looked at seven studies, one in Illinois, one in Texas, one in South Carolina, Arizona, California and Maryland, actually that’s six.  There’s also one in Milwaukee which was a little bit different.  So the first thing is what can we learn about the relative importance of people coming on versus finishing the application process, and I think, what you see is there really is some of both.  So just for an example, in Illinois, in 1997, there were 14,419 applications and by 2001, that had gone down to 3,037, so there was a huge drop in applications in Illinois.  Of those that were approved, in 1997, it was about 10,000 of the 14,000.  In 2001, it was about 1,500, to about a half.  So both the entry down and the rate went down.

If you look at Georgia which is another state that has a pretty significant upfront process, you see sort of a decline from the number of applications but you see a much greater decline in the percentage that are approved, so the approval rate went from 50 percent in 2001 to actually 22 percent in 2006.  So you have a lot of people who are coming on and not going on TANF so I think the diversion, if you think about is sort of these two stories that people who didn’t come on but also people coming on and not completing the process or not getting on the TANF.

So not surprisingly, the studies that are out there actually look at who’s diverted.  So they’re looking at the people who came on and then didn’t go on the TANF.  And in some cases, there are comparisons to people who came on.  So I would just like to tell you, I’m not going to give you numbers.  I just am going to give you a broad brush of what are the characteristics, what do we know from those studies.  One is that the characteristics that people gather look pretty much the same.  So you don’t have information that really sort of distinguishes from just their pure characteristics, much about who’s going on and who isn’t.

If you look at the employment rates, they are for the most part -- there are variations here.  Sometimes, they are less likely to be employed, sometimes they are more likely to be employed.  Even when they are more likely to be employed, the employment rates are quite low.  So in Texas, people who were diverted, the employment rate was 28 percent.  You also have this other category called redirected, and that was higher, it was 55 percent.  In Illinois, the percent that were employed is 36 percent. 

There’s only one study and that’s in Maryland where you see a higher employment rate than we see sort of in general leaver studies.  So in general, what we see is quite low employment rates when you look across the board.  But again, sometimes those employment rates are higher than for people who went onto TANF which suggest that there may have been what was intended which was getting people who could work to do that.  The other case studies that is in those studies is looking at some of the hardship measures, and when you look at measures of hardship in pretty much across the board, you see higher rates of hardship particularly around housing and food issues.  And then the other is -–

Lawrence M. Mead:  Among the diverted?

LaDonna Pavetti:  Pardon?

Lawrence M. Mead:  Among the diverted?

LaDonna Pavetti:  Among the diverted.  So you do see quite high housing needs and you see higher than you see in the others among food --

Lawrence M. Mead:  Higher than the recipients?

LaDonna Pavetti:  Yes, higher than people who went on to TANF.  The other is that they’re less likely to receive other benefits so you see lower take-up rates for food stamp and you see lower take-up rates for Medicaid although the gap for food stamps tends to be a little bit higher.  And just as an example, in the Arizona study, at the start there were 37 percent in the recipient group and 20 percent in the diverted group who were receiving food stamps.  For those who went on to TANF, that went up to 73 percent, it went up to 41 percent for the diverted, and then in the second-time point which is about 18 months out, there were 50 percent of the recipient and 32 percent of the diverted.  So you see two things going on.  One is they may have lower levels of employment but they’re missing sort of that link of getting into other benefits by doing that.  And the other thing is what we had just talked about, they are more likely to live with others so some of sort of not having the resources, benefits is there is sort of this countervailing piece of actually living with other people that may mitigate some of that.

So what we want to do is it seems to me that if that’s the story, then it is the story that we started talking about, that we need to be concerned about.  It does feel like there is a story of a group of people who were diverted and have positive outcomes but there is also this story that is a story about deep poverty and disconnectedness.  And I asked Sharon Parrot from the Center on Budget, if I could use her slides which she has put together to demonstrate this because this is a case where I think a picture is worth a thousand words. 

So this is just sort of deep poverty rates for children and this is using the National Academy of Science definition and you just can see this is sort of the racial breakdown but there are about 4.3 million children who have incomes below 75 percent and about 1.7 million below 50 percent.  So there is this substantial problem of people with deep poverty rates.  The thing that I think is important -- next slide -- is -- oh, I guess I can use the clicker -- is that not going on to TANF is part of the story, is that means-tested benefits are lifting a much smaller share of children out of deep poverty than in the 1990’s and so that again, if we had maintained the rates of recipients that we had, we would not see the same level of deep poverty as we do.

This is the receipt of what’s happening with AFDC, TANF and food stamps, and you can see that the share of eligible families has declined dramatically and it has continued to decline on the AFDC-TANF.  You can see that there was rebound in food stamps.  There were efforts to really try and increase participation and it really does look like that as well as maybe some things in the economy have really made a difference there.

This is -- if you look at states and how they differ, and you can’t really see the states but I will tell you, the states that we know -- at least I know from my work and we know from what’s written, they have very strong upfront work requirements are at the tail with the lowest percentage of people actually who are eligible receiving.  So you’ll see Georgia down there, you’ll see Illinois; those are states that have very stringent upfront requirements.  I think the interesting thing, Larry, would be to sort of try and find in this a state that has a strong diversion and has a higher rate of participation because that then suggests that they have been able to get people on and keep people on who are most in need.

Lawrence M. Mead:  The percent of children below 50 percent--

LaDonna Pavetti:  So, what percent of people below 50 percent are receiving TANF?  So if you look at the very poor -- so this I sort of a TANF recipiency rate around people who you would expect to be eligible.

Lawrence M. Mead:  I guess I don’t understand, it looks like California is on the right?

LaDonna Pavetti:  It’s on the right and you have higher than those 50 percent because of their benefit level because it’s so much higher.  So California is way on the right and it’s way up there, way up high.  So the percentage is -- whatever the first one is, basically, on the state, there’s ten percent of people who are on TANF, and if you’re in California you’re way up above 100 percent.  So again, this is sort of what’s the extent to which you’re actually addressing people, people who we would think need assistance are actually getting it.  And again, I think the interesting thing is here, you do see that cluster of those with very strong upfront requirements at the tail there, but again, I think the interesting thing would be to try and find one that has a higher rate to really look at that, as well as how have they been able to manage a dual sort of purpose.

This is sort of the issue with what Ron mentioned about, the disconnected single mothers.  You see this steady increase of people who really are neither connected to TANF, nor connected to the labor force.  I think the interesting thing, Becky Blank has done some work on this, is that these mothers appear to be connected to the labor force, at least at some point in time, it’s this movement in and out.  It was interesting when I was reading her paper, I felt like it was a repeat of the welfare dynamics but with the labor market, and that rather than people are moving out of the labor market onto welfare, what they were doing is moving out of work into no assistance.

So that’s sort of what the story that I think is there and what I would like to do -- I have one minute -- is to talk about, what I think the implications are and what we need to be thinking about as far what does this mean for policy, both for TANF and for, if we want to think about diversion in other contexts.  One is, I think one of the reasons why we have this story is that we knew from the start that people who came on to TANF were a very heterogeneous group and we’ve never figured out how to develop a system that really acknowledges that heterogeneity so that when people come on, if you’re going to acknowledge that, what you need to be able to do is to triage and be able to identify who can -- is the work requirement the right thing and something they can realistically accomplish and who is that really just not feasible for.  And I think we’ve never been able to develop, figure out how do we do that and do it well.  And so I think you know what has happened is that we have ended up with a system that has a strong work focus and strong set of work supports but we haven’t figured out, again, how do we allow for this diversity and different paths to work.  I don’t think the issue is whether people get to work.  I think the issue is what’s the right way for them to get there and whether there’s a difference.

I think the other is the labor market.  I think there was an assumption when we started welfare reform, that if we got people on that first rung on the ladder, they would move up and I think we must’ve assumed that when they lost their jobs, they’d come back.  And I think that we haven’t really figured out a safety net that works well for people’s situations that are not static, that really they do work sometimes but they don’t work all the time.  And so I think that part of that is really trying to again think about how do you come up with a system that really acknowledges that and that works for people when their situations change.

And finally, I think that we really need to think about, can we do this within TANF?  Are there ways in which we can make a broader set of work paths feasible?  Or, do we need to change other safety net programs to make them more accessible whether it’s housing, food stamps, whatever?  Do we need to create new programs, particularly for people who may be partially disabled?  I think one of the things you see is really high levels of mental health issues, physical health issues among people who are on TANF and some people who are not on TANF so is there really a need to sort of rethink a different kind of structure that again may be part of this triaging?  So I’ll stop there and can talk more on that when we have questions.

Douglas J. Besharov:  Donna, thank you very much.  Our next speaker is Demetra Nightingale, and when I said to Demetra, “Is it all right with you if we reverse alphabetical order,” she must think about twice as fast as I do, she said, “That’s fine.”   Everyone here seems to work faster than I do.  Anyway, Demetra is a research scientist in employment -- I’m going to read the list -- welfare, poverty, and social policy at Johns Hopkins at the Institute for Policy Studies there and she is a principal research associate and program director in the Labor and Social Policy Center at the Urban Institute -- was, okay, was the program director.  Her recent research focus is, again, on the nation’s economy, and again, the functioning of the labor market and her most recent publication is Reshaping the American Workforce in a Changing Economy, and boy, is the economy changing.  Demetra, thank you for being here.

Demetra Nightingale:  Thanks.  I’m really glad to be here, and in a way, listening to the previous speakers, I’m not sure I have that much more to add but I have been looking at these programs for a long time, I’ve done research in every state and most recently, in some of my insights that I’ll be conveying come from my recent work in South Carolina, New York City, Wisconsin, the Welfare-to-Work grants, programs which no one has mentioned yet but I will, and currently, I’m doing the Colorado Works Evaluation which is really very interesting and informative because some of you may know, the 64 counties in Colorado, Colorado is the most de-centralized welfare system, so there’s every range of diversity that you can imagine.  And I’m also involved in the designing the next round of HHS demos that will be looking at innovative strategies to increase employment and self-sufficiency.

First, before I begin, I want to just go back, one of the slides that Larry started with showed the decline that we all know in the caseload.  But it’s also important to remember that over the past few years, in at least half of the states, the caseload is actually going up.  So we shouldn’t forget about that.  I have just a few points that I want to make.  I can do this myself, right?  I can do it.

First, the public message has changed and I’m trying to focus my comments on things that other people haven’t said, if I don’t mention it, it’s probably because I agree with previous presenters.

I think it’s important to acknowledge the role that the president’s message had in welfare reform and as others have already mentioned, Tom I think mentioned this and others, that the caseload declined even in states where there was nothing going on and part of that was because the message was on the television, it was on the airwaves and the message clearly changed with leadership coming from the top.  It also combined with another message coming from the business community at a time when they needed workers at all levels, particularly at the low end in this regard, and it reinforced the message and the welfare to work partnership at the national level with the major companies agreeing on television with the president and Republicans and Democrats alike to hire welfare recipients, really helped to remove some of the stigma attached with hiring welfare recipients even if the welfare hires that they claimed weren’t always AFDC-TANF; that’s a topic for another session.

The message to welfare workers also changed which we’ve also already talked a little bit about.  I just wanted to add a couple of points here.  One is often, those who’ve been looking at these programs for way too many years, before some of you were born, we know, and this is still the case as Irene said, that income maintenance and the eligibility workers are doing basically the same thing they’ve always done and we know that how their work is structured is with a checklist of things that they have to do.  Well, there are more things on the checklist now, more boxes to check, which Tom also mentioned, but that’s very important.

I think the other thing that happened we haven’t mentioned is that there was an initial purge of the caseloads that occurred.  And if you look at Larry’s first chart, you can see that after 2000, the decline really leveled off a little bit, most of the purge happened even before 1996 but certainly before 2000.  So since then, the decline has been much slower and sort of purging the rolls always leads to some immediate smoking out of people who either were already working or don’t want to be hassled or some were dead in some states.  So, it did certainly change.

The message in the offices has changed, as Irene said.  That’s clear.  There is a new message, there are new signs the message is different.  Many of the workers tell us though that the number of families in need haven’t decreased, as Donna implied, and they may actually still be in the system somewhere so often they’re on food stamps and Medicaid as we mentioned, but also, some are on SSI which isn’t necessarily under the auspices of the welfare system but they’ll be getting food stamps through that.  Some on SSI are employable or in child welfare or in child-only cases.  So it’s important to understand that just because the TANF caseload went down, doesn’t mean they’re not there.

Also there’s more funding for non-cash and employment services, others have already mentioned this.  The TANF reserves and the surplus from the federalist negotiated financial agreements that led to welfare reform in 1996 allowed states more money that they could expand to different things and that allowed funding different employment services which also was a major change.  The Welfare-to-Work grants which was also part of that deal, the deal to get welfare reform passed in 1996 also expanded the capacity, the employment and training system capacity, not necessarily in the welfare agencies but in the workforce system and in non-profit organizations in the late 1990’s.  Ironically, at that time, they didn’t really need that money.  Boy, they’d love to have it now but now it’s gone.

Lump sum diversion payments to provide options to cash assistance and non-cash components that are increasingly now targeting on non-welfare working families.  Some of the diversion payments also were specifically targeted to people who had housing needs which is consistent with what Donna was saying.  So sometimes, that diversion was because of housing problems and the funds were to help in that regard.

The employment side of welfare changed, as I already mentioned.  A couple of people have mentioned, I think, where is Ron?  But anyway, certainly the institutional arrangements have changed some but the workforce system called different things at different times has always been a major part of the welfare system.  In fact, it declined a little bit in the early 1990’s, mid-1990’s, and then sort of reverted back.

Just two little anecdotes, recently in Colorado, I was in some small county somewhere and two women came up to me, one said, “Oh, I’m doing the same thing that I did when you were here 31 years ago talking to me,” and that was also very interesting.  And another one from a different county nearby said, “Actually, we’re doing things a little bit differently.  Now, we’re going back to what we were doing 30 years ago.”  So that the shift in the institutional arrangements I think has to be understood within the context of the political and organizational system in the field.

The gateway has changed, and others have already mentioned the decoupling of Medicaid and TANF and expanded SCHIP and also the simplified outreach that Pam talked about for the food stamp procedures.  All of these things have clarified what we have heard anecdotally and through focus groups for many, many years, that it was not necessarily the small amount of the AFDC grant that was important but in fact the package of services and the fact that it was AFDC that was the gateway to these services.  So some of the reduction in applications, where it exists, and some of the willingness to take diversion is because they can get some of these other services.

Finally, will it continue?  I have just a few points here I want to make.  One is that the reserve funds are dwindling and this may lead to cuts in employment services, we’re seeing that in some states that have expanded their employment activity and now cutting back.  It doesn’t mean they’re cutting back on their work requirement responsibilities and their checklists but in terms of the contracting for services, the trend may stop.

New participation, pressure with the DRA.  Changes may increase diversion because a lot of states are trying to reduce their denominator so that they meet their federal participation rate.  So that means the trend that we’ve seen in the past may stop, the trend in the decline of the caseload.

Some states are beginning to think more about outcomes rather than process and participation.  Many administrators and staff say that that’s what they always wanted to focus on but that they couldn’t because of participation rates.  Now we’re seeing that some of the programs and components that are funded with state funds, the state segregated funds are thinking more about outcomes meaning employment.  So we may see the trend continue.

The cash caseload is low.  As I said, we may have plateaued.  And in terms of the national decline, the continuing decline, it’s really coming as we all know from California and New York, so if California and New York plateau, then we’re going to see an end to the caseload decline.  And the cash caseloads may rise with declining economy, we’re seeing that in some local offices although not very much yet, in part because they’re using lump sum diversion grants for those formerly employed people who don’t qualify for unemployment insurance.

So, what would I bottle up if I could, from Larry’s point?  I think, one is having money for employment services because the value of the TANF grant has decayed due to inflation over the years.  We don’t have the welfare to work grants there anymore and the importance of the employment services shouldn’t be forgotten and we probably need some designated funds for that.

Second, the broadened availability of other supports, like healthcare, have been very important to allowing families who could, to seek other options.

Third, which is consistent with what Donna said and some of the others, we need to make sure that we allow the programs to be able to focus on working families and working poor whether or not they’re on welfare.  There’s also an increasing evidence, I think talking to local people about the disconnected population of eligibles, why aren’t they coming forth and is there an unnecessary barrier to enrollment?  And so in a lot of places where the caseloads are so low and the workers know that there are poor families, they’re beginning to do more outreach to bring them in; whether or not they get cash assistance is a different story.

And finally, I would highlight what I started with, that it’s very important to have the political support at the highest levels; that it’s both changed the stigma about hiring welfare recipients and others have done studies looking at the employer reactions and certainly that has changed the role of welfare in the overall mix of income in some working families.  And the political support on welfare per se was a very important factor in helping to change the culture of welfare to the extent that it did change.  Thanks.

Douglas J. Besharov:  Demetra, thank you very much.  Our final speaker on this panel is Dan Bloom who is the director of the Welfare and Barriers to Employment policy research area at MDRC, and he is currently directing the evaluation of the Center for Employment Opportunities program on ex-offenders.  Again, the message, it’s very interesting, about the research that’s now being done, about barriers to employment and so forth.  Since joining MDRC in 1988, Dan has written more than 30 research reports and a book summarizing the lessons learned from the studies of Welfare-To-Work programs.  By the way, I would encourage you to take a look at that book.  And I think, like the other people in this panel, he has seen a lot of programs.  Dan, thank you very much for being with us.

Dan Bloom:  Thanks, Doug.  As I was sitting here, thinking of the scary notion of talking without PowerPoint for the first time in years -- I designed a really interesting random assignment experiment about PowerPoint but we’ll go into that another time but anyway -- I thought I would start with just a note sort of maybe humility.  I haven’t visited all the states so Donna and Demetra are way ahead of me on that, but I’ve done lots of field visits and interviews with welfare staff and managers over the years and I think that kind of data is tremendously useful in understanding programs and how they work and why they don’t work and so on. 

I’m a little leery about relying too much on that kind of data to understand a big question like a caseload decline.  And the reason is I think that staff and managers in welfare agencies can only see a fairly narrow slice of their clients lives, number one.  And number two, I think they’re bound to overestimate the importance of things that they can see clearly like changes within their programs which is just naturally -- And then most important, they’re not going to know a lot about people who never come forward to apply for their programs which is one of the big things we’re talking about today.  So I think surveys, like the one Sheila was mentioning, ethnographic kind of work, talking to current, former, potential recipients is potentially very valuable to look at questions like this.  But anyway, now that I’ve discredited everything else I’m going to say, I’ll move on.

So the first question Larry asked is why did the welfare caseload go down so much more than random assignment studies would have predicted?  And since MDRC did many, maybe even most of the studies, I can paraphrase the question, why did all those MDRC studies get the wrong answer?  So that’s what I started to think about.  I want to start by saying it’s not actually a new phenomenon.  I dug back into some old reports, if you look at MDRC reports and leave them on you shelf for a long time, the color actually changes, I learned that. 

Anyway, so I looked back at the GAIN, remember the Riverside GAIN program that got a lot of attention back in the late 1980’s and early 1990’s and the random assignment study found that they decreased the caseload, I think, by about ten percent more or less.  The interesting thing which I had never actually realized before is the county’s welfare caseload was going up during the period of that study.  So the Riverside GAIN program evaluation was finding a reduction in the welfare caseload.  The county’s welfare caseload in 1988 when the study started was 20,000 cases, and in 1993, when the study ended, it was 27,000 cases so it’s actually a big increase.  And it was a period where the welfare caseload, nationally, was going up, it’s not surprising but I don’t think it means that the study is less valuable.  I think it suggests probably the caseload would’ve gone up even more if it hadn’t been for the GAIN welfare-to-work program. 

But I think it also points out that random assignment studies are designed to evaluate the impact of specific changes in policies or practices and usually for current recipients -- we can talk about that issue a little more later.  They’re not designed to explain changes in the caseload overall which can be affected by lots of other things, the economy, the changes in national tax policy, whatever.  Later in the 1990’s, we saw obviously somewhat different pattern, random assignment studies again, we did this waiver evaluations that people have talked about.  They usually showed, again, fairly modest declines in the welfare caseload where the overall picture was that the caseload was dropping like a stone. 

And so -- I guess on one level, the answer to Larry’s question is pretty clear.  The caseload went down more than we predicted or more than we measured because it was affected by factors that we didn’t try to measure in the studies; the strong labor market, expansion of the EITC, de-linking of Medicaid from TANF, all the things we’ve talked about today and those things apply to both the program and control groups in the study.  We controlled for them to try to isolate the impact of “welfare reform,” and of course, for the most part -- I’ll just talk for a second, I actually made a call during the break to try to get the answer to your question about applicants and whether they were part of these studies.  We often did have applicants in the studies. 

I think the big question is, where did we exactly do the random assignment in the application process, and I talked to Karin about this at the break.  I think we usually did it at the point where an application was likely to be approved.  It wasn’t like the first day that somebody walked into the office; it was a little further upstream. So we’re probably not getting a great picture of the effects on the application process itself, and we’re getting no picture of the big, big question which is, did more people apply for benefits in the first place or fewer?  Because of welfare reform, that’s just not part of the studies.  So, we’re missing lots of things. 

I think the other thing that was different in the late 1990’s is that the changes in rules and practices in the welfare system as we’ve talked about lots here, triggered broader changes in the society and perceptions about welfare and these perceptions almost certainly affected people’s behavior; either to leave welfare or not to come on to welfare.  And we could protect the control groups from the mechanical changes of welfare reform and the studies were fairly well done, not just MDRC’s but everybody’s were fairly well done and the control group really didn’t have time limits and they really didn’t get the expanded disregard, they really didn’t get the other things that we were testing but we couldn’t protect them from this general message, this increased stigma, this fear about welfare ending, whatever. 

We used to say that we didn’t put the control group in that soundproof room that they use on the game shows.  They were out there in the world and they were talking to their friends and reading the paper and watching TV.  So maybe what the studies did is they captured a very direct impact of the welfare changes but they didn’t capture an indirect impact of the changes and maybe in this case the indirect change was bigger than the direct change.  I will say it felt a little silly at times giving briefings saying that welfare reform modestly decreased the caseload where the world was standing there saying welfare reform caused the caseload to drop by half; people looked at me like I was living on another planet, I think, sometimes.  But anyway, all of that is a little unsatisfying in the sense that we know that the caseload decline was driven, at least in part or maybe a large part, by things we couldn’t or didn’t measure in random assignment studies but that leaves open these key questions: which elements of welfare reform mattered most?  And, how important was welfare reform overall relative to other things, EITC, economy, all the other stuff that was going on? 

On the first question, I think the key elements of welfare reform, the things that changed in the mid to late 1990’s, I think we’ve talked about sanctions, time limits, Work-First, all that.  And I’m just going to say one word about time limits because it’s just something I know more about than other things.  You know, on the surface, as people said, it’s very hard to argue that time limits had much of an effect at all.  Very few people reach time limits during the period of the big caseload decline.  Most of the biggest states don’t even have time limits, to be honest.  I mean, they have very loose versions of time limits, some don’t have them at all.  But we need to remember that we didn’t know all that back in the mid- to late 1990’s and a lot of the states that I knew, although they didn’t end up applying time limits very strictly, they pretended to throughout most of the 1990’s, they said they were going to and a lot of potential recipients may well have believed that and it didn’t become clear until much later that they were giving a lot of extensions and exemptions and everything, there was sort of a big bluff that went on in a lot of states and it could have had an effect. 

So all I’m saying by that is I can’t prove that time limits did have a big effect but they may have even though you can’t see it in people reaching the time limits.  And again, I’m not sure we did a great job.  We did a few random assignment studies where the program group had a time limit and the control group didn’t and I’m not sure that we did a great job of measuring this because no matter how clearly we told the control group that they didn’t have a time limit, I thin