"They Don't Need Stata....They Have Word"

(See below for title explanation.)

Doug Besharov is Joseph J. and Violet Jacobs Scholar at the American Enterprise Institute. He is also nominally a colleague of mine at the University of Maryland (Doug has an appointment in the Public Policy School at UMd; see below for a related full disclosure moment) It's probably fair to say that he is the go-to guy for the MSM on welfare policy questions. This topic happens to be one on which I've done some academic research over the last few years.

Today Doug has an op-ed column in the NYT, titled End Welfare Lite as We Know It. While I don't usually blog about my research, I think it's probably worth making some comments here, because Doug does his usually brilliant job of

• citing basic statistics and characterizing largely uncontested facts, and then
  
• drawing the same old "cut-benefits, ratchet-up-the-pressure-on-the-poor" conclusion, regardless of the relationship - positive, negative, or zero - between the facts and the conclusion.

In today's article, Doug acknowledges that "it took more than welfare reform to end welfare as we knew it", with the remarkable boom of the late-1990s being the obvious candidate. This acknowledgment is welcome. On the other hand, his claim that the decrease in caseloads came amid "little sign of serious additional hardship" likely is the result of a focus on summary statistics (which cannot tell us what applied microeconomists call the "counterfactual", that is, what would have happened in the absence of either reform or the economy) as well as studies that focus on average treatment effects.

As one of my papers (coauthored with Marianne Bitler and Hilary Hoynes, who have nothing to do with this blog in general or this post in particular) shows, however, average effects can miss a lot. Basic economic theory suggests that reforms like those that preceded and were cemented by the 1996 Personal Responsibility and Work Opportunity Reconciliation Act (PRWORA) should increase earnings for some, reduce them for others, and have generally mixed or ambiguous effects on income. Our paper shows that average effects do a woeful job of characterizing the effects of reform. We do find some evidence of "additional hardship", and (for technical reasons) it is possible that the apparent additional hardship we do find is the tip of the iceberg. [Update: our paper involves only data from Connecticut's evaluation of its pre-PRWORA Jobs First program. Like other evaluations, this one involved random assignment of subjects to the new and old programs, which is an attractive feature for evaluating reform's effects (though such studies do have drawbacks). We chose this program because its features are both similar to those mandated by PRWORA and among the most substantial in terms of their difference from the pre-reform features. There is every reason (inlcuding preliminary analysis) to believe that results from other states' experiments would be similar. Since Jobs First is very similar to PRWORA, moreover, our results likely are generalizable to other states.]

None of this is to deny that income and earnings have gone up for some (perhaps even a lot) of women affected by welfare reform; in fact, our results show that these indicators must have improved for some women, and if the tip-of-the-iceberg effect mentioned above is real, then the gains for the winners would necessarily be even greater than is apparent in our paper. Doug writes that

the best estimates are that only about 40 percent to 50 percent of mothers who left welfare have steady, full-time jobs. Another 15 percent or so work part time. According to surveys in various states, these mothers are earning about $8 an hour. That’s about $16,000 a year for full-time employment. [JG note: I assume that these figures are averages, in which case they likely mask considerable heterogeneity of the same sort that motivated my paper, discussed above.]

Doug adds compassionately that

It is their story that the supporters of welfare reform celebrate, but $16,000 is not a lot of money, especially for a mother with two children.

That point is certainly true, but the fact is that no one in the welfare research world seriously thinks that these folks are going to do much better than $16,000 a year on average. The fact is that most women who have lengthy spells of welfare participation do not have the kind of skills that will translate into earnings much greater than this level (there's a reason they wind up on welfare in the first place). Barring a massive improvement in their human capital or a radical shift in the U.S. economy back toward career ladders based largely on experience, earnings just aren't going to be very high for these women.

Moreover, those folks - the ones who can get it together to find and hold a full-time job, if they are forced to - have never been the primary reason to worry about the stick-first, carrot-second style of reforms so dear to Doug and his comrades in arms. Rather, for me the biggest concern has always been the folks at the very bottom: those who will have very little chance of finding and keeping any job, even one that pays less than $8 (and almost surely has no benefits to speak of). Some critics of PRWORA suggested it would simply shift women from traditional cash assistance to other programs, with the disability and supplemental security insurance programs commonly mentioned.

Back when PRWORA was being debated in Congress and within the White House (a whole other but not uninteresting political story), PRWORA's supporters pooh-poohed these concerns. They pointed out that Medicaid would still be available, that states could exempt 20% of their welfare caseloads from work requirements and the 60-month lifetime limit on federal aid, and that states could supplement federal payments with their own dollars if they chose to. Thus, the claim was that the worst off folks would still have a safety net - we were just going to get serious with all those other folks who could make $16K per year working full-time.

In fact, there are some folks struggling to get off welfare even now. Doug writes today that

about a quarter of those who leave welfare return to the program, with many cycling in and out as they face temporary ups and downs.

In addition, when they’re off welfare, some of these families survive only because they still receive government assistance through food stamps (an average of more than $2,500), the Women, Infants and Children program (about $1,800 for infants and new mothers), Supplemental Security Income (an average of over $6,500), or housing aid (an average of $6,000). Their children also qualify for Medicaid. In reality, these families are still on welfare because they are still receiving benefits and not working -- call it "welfare lite."

So, yes, welfare reform reduced welfare dependency, but not as much as suggested by the political rhetoric, and a great deal of dependency is now diffused and hidden within larger social welfare programs.

Note that this set of results is precisely how PRWORA's proponents had suggested things should work: those who can't make it in the labor market would be exempted from new requirements and limits of the welfare system, allowing them to stay on (or perhaps cycle on and off) welfare as necessary, while their needs would be met using other government programs.

So what is Doug's response to this situation? You might think he would declare victory and say, "See, liberals? PRWORA's enactment hasn't left any women or children behind!". But Doug is no optimist about PRWORA when it comes to this particular topic. No. Apparently the appropriate response to full-time workers making an average of $16k per year and a bunch of folks who are at least partly insulated from the trilogy of their own bad luck, lack of skills and the welfare reforms of the 1990s is not the celebration so commonly entertained by PRWORA's cheerleaders.

Instead, Doug concludes (cue screeching eagle and rousing, Colbert Report-like music) thusly:

The tougher work and participation requirements added in this year’s reauthorization of the law could help states address the deeper needs of welfare families. But many states are already planning to avoid these new strictures with various administrative gimmicks, like placing the most troubled and disorganized families in state-financed programs where federal rules do not apply. This would only further obscure the high levels of continuing dependency.

For now, welfare reform deserves only two cheers. Not bad for a historic change in policy, but not good enough for us to be even close to satisfied.

Evidently we have met the enemy, and it's name is any provision of PRWORA -- or fiscal federalism, for that matter: how dare those states act like laboratories of democracy! -- that provides even the barest insulation for the would-be welfare population.

Let's review Doug's argument:

1. For all the talk about welfare reform's successes, lots of people are still in bad shape -- cycling in and out of welfare and other programs, making little money (say, because they work full-time for just $8 an hour). Some of them are even being protected by those lousy states!
   
   
2. Therefore we should crack down, increasing work requirements, eliminating exemptions, and (I guess) further restricting state flexibility.

Only someone committed to the "first know answer, then plug in possibly related but if necessary unrelated assertions of fact to justify answer" style of argument could come up with this combination. Regarding women who make $8/hour working full-time, it is just silly to say that increasing work requirements will somehow make them better off. Regarding women who continue to cycle through TANF or use other programs, I've said enough above.

This is the sort of article that reminds me why I unsubscribed from Doug's op-ed listserv (to which I'd never asked to be subscribed, incidentally).



Title explanation: Stata is a statistical software package that empirical researchers use (among other things) to assess the relationships between behavior and policies. Word is a wordprocessing package that (among other people) pundits use to claim things that suit their preferred policy outcomes. The title of this post is a comment made to me today by an economist friend who is no liberal and no stranger to Doug's writing after he read Doug's column from today's NYT.

Full disclosure moment--I provide this discussion to be clear that Doug is not my favorite guy for those of my two readers who may think I have an axe - as opposed to Doug's current shoddy argument - to grind: A few years ago (2002 if I remember correctly), one of my other papers---a revised version of which was recently published; see this link for an abstract---was the subject of a debate on the Tavis Smiley Show on NPR. One of my coauthors spent a few minutes discussing the earlier version of the paper with Tavis Smiley, after which there was a debate between Doug and some (non-economist, if I remember) person I'd never heard of and whose name I don't remember. The full story is a very long one, but the day before the show I spent a considerable amount of time on the phone with Doug and one of his AEI/UMd (both) associates answering questions they had about our paper. I did a nontrivial amount of additional statistical research to address concerns Doug expressed about our paper, the upshot of which was that they showed his concerns were almost certainly unfounded. None of this stopped Doug from going on the air and distorting my research, claiming we hadn't looked at things we had, and insinuating that the data we used were somehow inferior to his preferred source of analysis (which, as it happens, had been done on precisely the same dataset).

This stunt was pretty low in my book, not least because Doug was nominally a senior colleague of mine at Maryland (though the policy school and econ department are administratively separate). In his final email to me on this matter, Doug later claimed that the issue was my supposed "insistence that" my research "*must* be correct". The research itself has both strengths and weaknesses, as does all social science research; I'd have been delighted to acknowledge that point - indeed, I was willing to spend my own time investigating Doug's concerns precisely out of a commitment to healthy skepticism (if Doug were more committed to empirical research and less committed to the ensuing politics surrounding given empirical findings, he might have realized that fact). What made this interaction different was that Doug's particular criticisms were empirically unfounded, a fact upon which I certainly did insist, and his mode of pressing them was intellectually dishonest.