How Much Disincentive Is Too Much?

Matt Zwolinksi wrote a brief in favor of the BIG. I replied. Matt responded to objections raised by Robert H. Frank and me.  This is my reply to Matt, with respect to the parts of his response directed to my arguments.

In his current response, Matt now accepts that it is “uncontroversial” that the Negative Income Tax (NIT) has been shown in randomized experiments to cause a reduction in work effort.  But he first raises some technical issues: (1) the decrease was small, (2) it was caused primarily by non-employed people remaining unemployed longer, (3) it would be less under an unconditional BIG, and (4) it was subject to bias because it was survey-based, and that under alternative measurements using administrative data, “much of the apparent unemployment effect disappeared altogether.” [italics in original].  He then proceeds to return to the argument in the alternative from his original essay that a reduction in work is not necessarily a bad thing.  I’ll address the technical issues first, and then move on to the values argument.

Matt’s technical issue (4) seems to undercut his acceptance that a reduction in work effort in these experiments was uncontroversial.  To my knowledge, two of the NIT experiments were subject to post hoc reanalysis in which administrative data was used after the fact to try to create different impact estimates that were then compared to survey-based estimates.  Contrary to my reading of Matt’s statement, in neither case would the administrative data have made the unemployment effect “disappear altogether.”  Even according to the paper that Matt cites as evidence, in one of these experiments “much” of the employment effect would disappear if measured using administrative data, and in the other the change in effect was “somewhat smaller.”  The NIT experiments were designed from the start to measure effects using surveys (and often in fact set up a parallel tracking system), and it turns out that there are general pros and cons to using survey versus administrative data in social experiments.  Further, as I reviewed in Uncontrolled, the later series of more than 30 welfare experiments in the 1990s were designed from the beginning to be read using administrative data, and these experiments consistently found that the absence of work requirements led to statistically significantly lower employment and earnings using this data.  There is no credible interpretation of the NIT experiments and later welfare reform experiments in which we do not observe a causal reduction in work effort in response to the elimination of various kinds of work requirements. 

Matt says, however, that this reduction in work effort was small and would be even smaller under an unconditional BIG.  One problem with this is that the NIT experiments measured the impact of this policy change versus the baseline of the American welfare system circa 1975.  The current welfare system has much more stringent work requirements, which were proved in the1990s welfare experiments to increase work effort. Therefore a hypothetical 2014 BIG would be expected to have a larger reduction in work than the same system would have had in 1975.  I understand (though I do not accept) Matt’s argument from theory that somehow a truly unconditional BIG would partially offset this, because it would have, ceteris paribus, less impact on work effort.  I think any rational observer should remain unconvinced on this theoretical point until it is demonstrated experimentally – unlike similar theories that were disproved in the original NIT experiments.

In the end, all of these debates are about quantity, not sign.  All reliable evidence seems clear that a BIG would reduce work effort. 

This brings us to the second part of Matt’s argument – less work would not necessarily be a bad thing.  As Matt correctly notes, this is not completely independent of the size of the work reduction.  If a BIG would cause a total reduction of one person-hour per year in all of America, it would technically reduce work, but would presumably not matter to us.  Matt argues, again correctly, that we must compare the costs of work reduction (and presumably other drawbacks) to the benefits created by a BIG.

But Matt’s argument breaks down both in the method of measurement of this trade-off, and in where he places the burden of proof.

With respect to method, I think the key passage form Matt’s response is the following:

Work is a central component of both personal and social well-being. But it is not the only component. And it may well be a component that we, as individuals and as a culture, tend to err on the side of over­-valuing. [Italics in original]

This is closely related to my original critique that Matt wants to assume away democratic governance.  Matt may be right, and the American electorate may be wrong, but it is clear that they disagree on this point.  To grossly oversimplify, the amount of work disincentive is prohibitive versus the benefits of leisure in the view of an electorate that has never supported a BIG in two hundred years, and has been moving away from this concept and toward stricter work requirements for decades.  Matt needs to change a lot of minds before this would become law in a democratic society. 

So what is the basis for Matt’s appeal to the electorate that they should change their views?  Matt writes that:

Like most other issues of social policy, then, the issue of work disincentives involves questions of trade-offs. One cannot criticize a BIG merely by showing that it would create some disincentives to work. One must show that it would create too much – that the moral, economic, and cultural costs of those disincentives would not be sufficiently compensated for by the moral, economic, and cultural benefits. [Italics in original]

Sometimes it is hard to determine the reference case in policy debates, but this is not one of them: Matt is self-consciously proposing a radical departure from current policy and established practice over centuries. He is then asserting that he can demand that those who oppose the change must provide evidence that it would have net negative effects. The burden of proof lies with Matt, and he has yet to make his case.  I will look forward to hearing it from a writer as persuasive and talented as he is.

Also from this issue

Lead Essay

  • Matt Zwolinski argues that a basic income guarantee (BIG) could very easily do better than our current welfare state by many different criteria. It would be far more efficient. It would be less subject to rent-seeking. It would be easily accessible by the poor, and its benefits would flow to them rather than to the middle class. Although there are many libertarian objections to a BIG, Zwolinski nonetheless argues that when faced with a choice between a BIG and the status quo, libertarians should be open to making the change.

Response Essays

  • Michael Huemer argues that while a basic income guarantee might be better than the status quo, this amounts to some rather faint praise. A basic income guarantee would necessarily violate some people’s rights, while a fully legitimate government must never violate anyone’s rights. The problem of political authority will likely remain a barrier to all similar proposals, even if we may happen to find this problem’s full implications troubling.

  • Jim Manzi doubts that a basic income guarantee would emerge from our political process while still bearing its purportedly beneficial features. Compromises would proliferate, as would paternalistic controls. The interests of the bureaucracy would assert themselves, and the temptation to make exceptions would prove overwhelming to the electorate. Moreover, when a basic income guarantee has been tried in practice, the result has consistently been a withdrawal of participants’ labor. Scaled to an entire society, the result of such a withdrawal may be dire.

  • Robert H. Frank agrees with Matt Zwolinski that a basic income guarantee would achieve the welfare state’s goals more effectively than our current patchwork of programs. But he argues that a basic income guarantee sufficient to end poverty would spawn massive taxpayer resentment. Incentives to work would be undermined both for recipients and for those whose tax dollars funded them. Frank recommends a combination program that would include a significantly smaller cash grant and a standing offer of public employment for any who desired it. Frank defends taxation against libertarian objections and offers several additional taxes that he believes should be implemented. He argues that these should help pay for the expensive programs here being considered.

Letters to the Editor

  • The Cato Institute’s Michael D. Tanner examines the Basic Income Guarantee and finds that its simplicity wouldn’t survive the political process. Difficulties abound, arising both from practical politics and from the realities of our current welfare expenditures. Tanner recommends consolidating our welfare system and simplifying it, but he does not endorse a Basic Income Guarantee.

  • Economist Ed Dolan shares some of his findings on the Basic Income Guarantee. He finds that work disincentives will indeed exist under a BIG, and yet these may be smaller than the work disincentives we already experience owing to the welfare state as it now exists. Libertarians should not be tempted by the so-called “gospel of work,” he says; libertarianism, rather, is about the gospel of freedom of choice.