approach to thinking about social welfare resource allocations applies to them as well.

Transcription

1 Bad Draws, Bad Bets, Bad Apples, and Bad Policies Peter H. Schuck?? and Richard J. Zeckhauser???? 2/25/05 draft The public domain is awash in bad policies. They come in many different forms and are bad for many different reasons, 1 but they all share one characteristic: they accomplish less for society than would alternative policies using the same resources. Many programs seek to allocate resources to individuals who are members of a legally-defined target group that politicians and policymakers think deserve these resources. In this paper, we consider a subset of these beneficiaries: unfortunate, often low-income individuals whom we call bad draws. 2 It is natural to think that bad draws, as parties to a kind of social contract for insurance against random misfortunes, merit resource transfers. That is why the government helps to pay for the medical care for the sick, unemployment insurance for those who lose their jobs, and food stamps for low-income individuals. Our argument is that the bad draws category is too broad to guide policymakers in allocating social welfare resources wisely, and that bad policies often result when we allow two particular groups of bad draws to receive the standard resource transfers for their misfortunes. We call these two groups of bad draws bad apples and bad bets. Bad bets are individuals whom society can reasonably predict will benefit little from the resources that it might otherwise transfer to them relative to the benefits that others would receive from the same resources. Bad apples are individuals whose irresponsible, immoral, or illegal behavior causes deserving program participants (bad draws who are good apples) to benefit less from program resources than they otherwise would. This harms not only the good apples, but also third parties who care about their success, and the taxpayers who fund these transfers. We shall refine and analyze these two groups in some detail below, bad apples in part III and bad bets in part IV. Although the notion of a bad bet is quite straightforward a classic example is a chronically ill nonagenarian who receives costly, subsidized medical treatment that predictably yields little social benefit. By contrast, the concept of a bad apple is not at all straightforward. For this reason, we feel constrained to say more about it in the paper and in this introduction. In part III, we focus our discussion on two examples of bad apples: public school students who chronically disrupts class, thereby impairing the learning of others, and public housing tenants whose chronic misconduct damages the quality of life for other needy tenants. Doubtless, some readers will bridle at the very idea of labeling as bad apples people whom social programs seek to help and whose behavior may not even be criminal. Increasing the readers 1 Simeon E. Baldwin Professor of Law, Yale University. 2 Frank P. Ramsey Professor of Political Economy, Kennedy School of Government, Harvard University. 3 See part VI.. 4 Although universal entitlement programs are not our primary focus here, much of our approach to thinking about social welfare resource allocations applies to them as well. 1

2 unease, we predict, is the fact that for purposes of our analysis, we define our bad apples category based on misbehavior, 5 regardless of its causes, so long as those causes cannot be rectified through normal, short-term interventions. As explained in part III, we think that social welfare is enhanced when those properly classified as bad apples are separated from the good ones until, perhaps after rehabilitation, society concludes that they have reformed and will no longer impose harm. Americans often refer to them as bad apples, even while trying to rehabilitate and redeem them through alternative programs. We use the term in that same spirit. The analysis proceeds as follows. In part I, we present our definitions, assumptions, and analytical framework. In part II, we discuss the social policy stakes in thinking more rigorously and, dare we say it, more courageously about bad apples and bad bets. In part III, we consider the bad apples category closely, and in part IV we do the same for bad bets. Both of these discussions propose a number of policy reforms that could minimize these problems. Part V discusses why few politicians and policymakers acknowledge the bad apples and bad bets problems, and why even fewer pursue the reform imperatives and opportunities that these problems present. As we discuss in part I, this lack of candor and analytical rigor exhausts voters limited altruism, which predictably and systematically harms precisely the deserving lowincome people we should most want to help. Hence, those like us who want to maintain and expand social policies likely to aid poor people effectively and efficiently what we shall call well-targeted redistributionists are precisely the ones who should be most dismayed by bad apples and bad bets. They should want to recapture the resources that bad apples and bad bets waste, and redirect those resources to better apples and better bets. Alas, many well-targeted redistributionists are all too eager to deny the existence of bad apples and bad bets, or to sweep the allocation problems they present under the rug. Part V analyzes this set of self-defeating attitudes. Part VI moves beyond bad apples and bad bets to consider the larger universe of bad policies, distilling from the earlier discussion some of the policymaking patterns that proliferate them. Part VII presents a brief agenda for future research. I. Definitions, Assumptions, and Analytical Framework A. Definitions. Bad draws are people who were dealt a bad hand at birth or later, and who have suffered misfortune as a result. Modern welfare states, including the United States, characteristically seek to alleviate the suffering of people whose conditions are thought to deserve special social remediation. 6 Our purpose is to help policymakers perform this task more 5 We explain in part III why the charge that setting cause to one side in these chronic cases is blaming the victim misses the main point, which is to improve outcomes for the good apples while attending to the bad apples problems separately. 6 As we discuss in part VI, many of these programs score low on target efficiency; that is they benefit some good draws people who do not suffer from misfortune and thus do not need help. Examples include Medicare, free university tuition, and subsidized student loans. Poor targeting of benefits is an important and often overlooked feature of the welfare state. Some commentators deplore it while others favor it as a way to increase political support for various programs. For example, the universality of Social Security retirement benefits, along with its 2

3 effectively, which will in turn increase the resources available to, and public support for, programs that assist bad draws, the vast majority of whom are both good apples and good bets. The connection between effective performance and public support for programs cannot be emphasized enough. Much of the opposition to these programs, we believe, arises from the public s belief that they devote too many benefits to the two groups of bad draws bad bets and bad apples that are our focus. Both of these groups drain public programs of vital fiscal and political resources. Needless to say, different voters define these two groups differently, and political controversy over some of those definitions is both inevitable and desirable. Bad bets are people who are unlikely to derive much benefit from a programmatic intervention on their behalf relative to both (a) the resources that the intervention would consume, and (b) the benefits that other individuals who are good, or at least better, bets would derive from some of those same resources (i.e., the opportunity costs of aiding the bad bets). Some bad bets are bad because the individual stands to gain little benefit from the resources. The classic example is a person in precarious health competing for a kidney transplant with another person who would benefit far more from the kidney. 7 Other bets are bad because the individuals who benefit do not need them. Tax expenditure policies and other programs targeted (often in disguised fashion) at the wealthy often include beneficiaries who are bad bets in the sense that including them will produce little or no policy bang for the buck. An important example is farm subsidies which, despite the populist rhetoric that surrounds them, primarily benefit giant agribusinesses, not family farms. Similarly, the tax deduction of mortgage interest for second homes, including very expensive ones, and even for homeowners in general does little to promote home ownership, the policy s chief purpose. Equally perverse, it disproportionately benefits the wealthy, who would probably buy their homes in any event. 8 In all of these examples (and all too many others), bad bets consume program resources that could be better used by other beneficiaries a widespread form of social waste. Bad apples are different: they may or may not be bad draws, and they may or may not be bad bets. In part III, we shall say much more about bad apples and about the costs that they impose. For now, the key point is that bad apples impose more harm than merely wasting scarce social resources. They reduce the already sadly-limited life prospects of the good apple-bad draws whom their behavior harms. They stigmatize these good apples in the programs, as well as the programs themselves. They jeopardize or exhaust voters limited altruism, demoralizing the larger society -- both literally and figuratively in the process. contributory feature, is an important source of the program s popularity. We touch on this issue here but do not pursue it at length. 7 The current priority system for kidneys in the United States, holding quality-of-match constant, allocates them first to the individuals who have been waiting the longest. This may well target kidneys to those who will receive the least benefit (in quality-adjusted life years (QALYs), discussed infra at ) from them. [infra cite to my personal example] 8 See Edward L. Glaeser and Jesse M. Shapiro, 17 Tax Policy and the Economy (2003). 3

4 Much of our analysis proceeds from the assumptions that voters altruism is limited and that they weigh the benefits that recipients receive as a function of the costs in higher taxes that they must bear. We think, and find casual evidence, that voters will agree to redistribute resources (through government programs or private charity) only when they think that those who receive them will secure more benefit from those resources than would the taxpayers who supply them. Needless to say, the political rhetoric that surrounds transfer programs often departs, sometimes substantially, from this benefit-cost framework. For example, redistributive rhetoric may justify transfers as flowing from the individual or group rights of recipients, as compensating for past injustices, or as simply helping the unfortunate. Whatever the rhetoric, politicians have their own reasons to support transfers. The basic rationale, however, is always the same: society is better off if some of the resources now held by As are instead spent on behalf of Bs. Only the state can effectuate significant redistribution to bad draws, and state action almost always follows the supposed preferences of voters. Voters recognize that their individual efforts and those of others like them will simply be too small to make a difference to large numbers of recipients. Even where private charity could make a difference, potential donors may not want those who are less charitable than they are to free-ride on their private beneficence. Compassionate voters can multiply their own charity many million-fold by getting the government to redistribute through a tax-and-transfer system. 9 Because such programs depend on popular support, voters must be convinced that the resources are not going to the wrong people. Voters, who are mostly middle-income, are altruistic toward bad draws -- but only to a point. 10 Their limited altruism will cease if they think that too many resources are going to bad apples -- or even to bad bets (particularly if they read part IV). Ordinary voters may not always be able to identify situations with too many bad bets or even bad apples as carefully as a professional policy analyst can, but they surely notice when costly government programs are not helping deserving recipients much. Our general argument, then, is that these three categories bad draws, bad bets, and bad apples are closely linked, and further, that the general inattention to the intricate empirical, moral, and practical relationships among them helps to explain the persistent political vulnerability of programs meant to help bad draws, most of them good apples who must depend on the programs for their well-being. It is true that programs aimed at low-income bad draws 9 Substantial private redistribution occurs; private philanthropy in the U.S. amounted to almost $250 billion in Doing Well and Doing Good, Economist, July 31, 2004, at 57. It may be undertaken to benefit specific individuals (e.g., one's destitute relatives), or to benefit donors themselves in some sense (e.g., salvation, tithing obligation, public esteem, supporting their favorite small-scale causes). Some private charity supports public goods, such as cancer research, that can benefit a large number of people, including the poor. In general, however, significant redistribution only occurs when voters, or their representatives, agree to bind themselves to contribute collectively. 10 It also appears that the race of the voter and of the presumed recipient is a powerful predictor of the trust in and hence the support for redistribution. Alberto Alesina & Edward L. Glaeser, Fighting Poverty in the U.S. and Europe: A World of Difference (2004). 4

5 constitute a relatively small proportion of our non-defense governmental expenditures. The largest transfer programs, Social Security and Medicare, overwhelmingly aid the non-poor. But voters would probably support substantially more transfers to the poor if they believed that their tax dollars were going to more deserving people and accomplished more social good. We hope to improve these bad policies by training a spotlight on them, illuminating the significant welfare losses, injustices, and demoralization caused by policies that devote scarce public resources to bad apples and bad bets rather than to those bad draws who are both good apples and good bets. B. Normative Assumptions and Analytical Framework Public policies in a liberal democracy like ours should promote the welfare of society. The principal ingredient of that social welfare is the aggregated well-being of individuals. Although this well-being is ordinarily measured by individuals self-assessment, society tends to be more paternalistic toward groups whose rational choices are compromised by immaturity, ignorance, or lack of free will for example, children, the uninformed, and addicts. 11 All social policies trade off the welfares of some groups against the welfares of others. Unfortunately, policymakers do not have a neat measurement device that tells them how much benefit an individual would receive from each proposed policy. The problem is not only one of measurement: philosophers and policy analysts vigorously debate which measures would be appropriate even if they were readily available. Although some would say that democratic processes solve the problem by legitimating whatever policies are generated by elections, legislative processes, and deliberation, this solution simply moves the difficulty back a step. Even a very compassionate citizenry, one that wants to help bad draws also wants its resources to go where they will do the most good. Voters want to know what proposed programs can accomplish, how much welfare they will provide to which beneficiaries, and at what cost. They can only know this if policymakers raise and resolve the kinds of issues that we analyze here. Allocating scarce resources to bad bets rather than to good bets is wasteful even when the bad bets are good apples for example, spending hundreds of thousands of dollars on an exemplary citizen in the closing days of his life whereas allocating resources to bad apples is not only wasteful but also unjust. Occasionally, sound policy may dictate benefitting even bad apples if this would create substantial welfare for the resources expended for example, giving medical care to tubercular criminals or free needles to chronic addicts will avoid infecting others. Efficiency maximizing social benefits per unit of cost is an important normative criterion even when the policy s primary goal is redistribution, and even when some beneficiaries of the redistribution are undeserving. 12 Felt compassion and calculated efficiency are complementary, not contradictory, virtues. Both payer As and recipient Bs share an interest in enabling the Bs to secure the greatest possible welfare gain for whatever resources the As make available for the benefit of the Bs, and this is so 11 Addicts possess some degree of free will when they make the choices that lead to their initial and perhaps their subsequent addictions; this fact may affect whether society considers them to be bets bets, bad apples, or just unfortunate bad draws. 12 Indeed, efficiency may be a special policy virtue in redistributive programs precisely because the moralizing politics and rhetoric, so characteristic of programs that emphasize group need or desert, goes to some lengths to obscure or marginalize efficiency considerations. 5

6 whether the As contribution is a voluntary donation or a legally-required tax. When As support a food stamp program for Bs rather than an increase in welfare benefits, then, we should assume that the As believe, at least implicitly and all things considered, that this is the best way to help the Bs. Where Bs consist of distinct sub-groups, As face a challenging two-part task. Given a fixed sum for redistribution, should more go to B1s or to B2s? And if the program tilts towards B2s (say, because they are sicker than B1s and can benefit more from the medical resources than B1s can), how can the program be designed to assure this outcome? When some B1s are bad apples and some are bad bets, our analysis points to allocating more resources to B2s and fewer to B1s, or perhaps all to B2s and none to B1s. Both compassion and efficiency values dictate that spending more on bad apples and bad bets is ordinarily unwise, irresponsible, and (in the case of bad apples) unjust. Some of our assumptions, while wholly consistent with the social welfare criterion explained above, may be sufficiently unfamiliar or even counter-intuitive that they bear some brief mention here. First, although we have characterized bad draws as individuals whose misfortune is not a result of their bad choices, society might still want to give them less than they need as defined by some objective measure. Individuals making decisions behind the Rawlsian veil of ignorance 13 (at least those who embrace a welfare economics framework like ours) should want to allocate resources so that the social welfare gained from an additional dollar spent would be the same across all social programs. If policymakers allocate in this way, needier claimants in a particular program will receive more resources than better off ones in that program and perhaps more than better off ones in any program -- but they may not receive enough to reach a standard that many would consider reasonable if attaining that standard would be too costly in relation to the benefits produced or to the benefits possible though alternative allocations. 14 Contrary to the Rawlsian scenario, of course, voters do not make their social policy choices behind a veil of ignorance. Rather, they already know whether they are more or less likely to benefit from particular policy proposals that is, whether they have turned out to be good draws or bad draws. Suburbanites know that they and their neighbors children are unlikely to suffer from failing inner-city schools, and college graduates understand that they will probably not have to rely on Medicaid. 15 To be sure, voters are public-spirited under certain conditions, and a healthy polity will nurture those conditions. Nevertheless, we assume that they will usually base their votes on the proposal s perceived benefits and costs to them, which includes whatever altruistic benefits they may enjoy. In the suburbs where most Americans live, the benefits of traditional welfare programs for the poor are overwhelmingly altruistic and only slightly out of fear that they may come to need these programs personally. 13 John Rawls, A Theory of Justice (1971). 14 Here, too costly is properly judged by voters and politicians, not the recipients or their advocates or (absent statutory discretion) program administrators. 15 See Richard J. Zeckhauser, Risk Spreading and Distribution, in Redistribution Through Public Choice, Harold M. Hockhman and George E. Peterson (eds.), New York: Columbia Univ. Press, 1974,

7 The vast majority of real world voters, who know that lotteries have already been run and that they have been fortunate, are likely to favor redistribution less than would voters behind a Rawlsian veil, who recognize that they might end up in the disadvantaged class. We would not expect to find this difference, however, when the resources are being allocated between good and bad apples. Consider voters behind a veil of ignorance to whom one piece of information is revealed: they are told whether or not they will be parents of school children. We believe that these two groups would make similar decisions about how to allocate resources between good and bad apple students, assuming that the bad ones were placed in a separate school for disrupters. To be sure, the former would worry that their child might be a bad apple disrupter, but they would also worry that their child would be a good apple whose education was disrupted. Both would consider the relative gains and losses going to these two groups of schoolchildren if separated. Similarly, in an actual vote over allocations to the separate schools where it is not yet known which students will be disruptive, parents and non-parents (whose only personal stake in the issue is civic) should favor roughly equivalent allocations. 16 For the same reason, we also believe that these two groups would make the same decision about whether to place disruptive students in a separate school in the first place. The need for prioritizing is perhaps most apparent in social programs to provide medical care. Broadly speaking, such programs seek to enhance recipients well-being by increasing the quantity and quality of life by adding to individuals quality-adjusted life years (QALYs). 17 If we spend program funds efficiently (and other things being equal), then each additional dollar will yield the same number of QALYs regardless of which individual or medical condition receives them. But this QALY-maximizing standard would oblige us to deny additional funds (say, for heart-valve replacements) to individuals who are already beyond a certain age or certain levels of physical or emotional deterioration. The same standard would cause us to deny individuals with mental health problems as many visits to the psychologist as they might want (or, some might say, need). 16 Although little in our analysis turns on it, we imagine that policymakers are -- and should be encouraged to be -- more likely than voters to engage self-consciously in calculation, to recognize and manipulate incentives, and to look beyond narrow self-interest. If they do (and more to our point), they will be more likely to favor excluding bad bets and bad apples (or at least segregating the latter) from benefit programs. It is true that policymakers, unlike voters, are playing with other people s money and are influenced by political and bureaucratic factors [RJZ: cite to Downs? Niskanen?]. These inefficient and self-serving factors, however, are balanced to some extent by policymakers technocratic role definitions and (in some cases) training. 17 The concept, first presented in Richard J. Zeckhauser and Donald Shephard, Where Now for Saving Lives?, Law and Contemporary Problems 40 (4), Autumn 1976, 5-45, is widely used in cost-benefit and cost-effectiveness analyses of medical procedures. Although QALY analysis is usually deployed to address health-related issues, the analysis is perfectly general and can easily be used to measure life quality as affected by education, cultural events, and so forth. For a recent defense of QALYs in policy analysis, see Matthew D. Adler, Working Paper 05-01, AEI-Brookings Joint Center, Jan

8 Such prioritizing a euphemism for rationing is now common, though most often concealed or inadvertent, and will be even more common in the future (for reasons explained in part II). Public and political resistance to this prioritizing is greatest when the losers under such systems the elderly and severely ill, for example are readily identifiable. Managed care organizations, for example, routinely place roadblocks in the way of expensive procedures that they think will yield relatively few benefits for the resources used. Oregon s Medicaid program allocates according to a list of subsidized interventions ranked according to benefit-cost calculations. 18 In the United Kingdom, individuals beyond a certain age [*specify] are ineligible for kidney transplants. Many other formal and informal rationing methods are employed, even as those doing the rationing may, for a number of reasons, deny the fact. Even where policymakers see the need for rationing and are prepared to accept its allocative and political consequences, implementation is difficult. EPA, for example, is now embroiled in a debate about whether its policies should be assessed according to QALYs gained, or should look instead to the criteria of lives saved or willingness-to-pay. 19 We take no position here on the desirability of particular ranking methods, but we do maintain that a compassionate society should generally prioritize its expenditures with a view to maximizing the social benefits generated. We would not confine this principle to health-related issues. Indeed, aggregate welfare would be increased if we could make accurate tradeoffs among the social benefits of cash transfers, health services, education, and other forms of welfare enhancement. Giving a poor person $1000 in cash may give him more QALYs than $1000 spent on his health care. If so, and if this choice had no other effects, our efficiency criterion would dictate giving him the cash. But however policymakers tally up benefits, they must still prioritize their expenditures. This precept has far-reaching implications for the substantial resources now going to bad bets and bad apples. Although we have referred categorically to good and bad bets and good and bad apples, it is extremely difficult to define these groups and determining which individuals fit into which group and it is crucial to get these definitions and classifications right. Society can distinguish bad bets from good ones only by analyzing social costs and benefits, empirical determinations that for familiar reasons may be elusive in practice. 20 It can only distinguish bad apples from good ones by making highly contested normative judgments. 18 The Oregon Health Plan: An Historical Overview, Oregon Department of Human Services, Office of Medical Assistance Programs (2004), 19 See, e.g., Katharine Q. Seelye and John Tierney, E.P.A. Drops Age-Based Cost Studies, N.Y. Times, May 8, 2003, A34. The QALYs approach tends to assign lower weight to the elderly, relative to other groups, than the other two methods. 20 This difficulty is exemplified by the problem of allocating organs, discussed in part _. 8

9 But even if we could define the bad and good apple categories uncontroversially, the task of assigning particular individuals to them (as any social program must do) would be difficult and controversial, raising complex issues of procedural fairness and even constitutional rights. Moreover, it is unclear which kinds of institutions and competencies professional, bureaucratic, judicial, religious, or other are best suited to make the factual and normative determinations on which such assignments ultimately depend. We discuss these challenges in the remainder of the paper. We have no neat solutions to offer and doubt that anyone else does either. But the task is no less important than it is daunting. We are convinced that conscientious policymakers must think much harder about these categories, gather the information and build the institutions that are necessary to fairly and accurately apply them to individuals, and then make resource allocation decisions accordingly. Accomplishing this will require some combination of technical diagnostics, hands-on program experience, and close attention to the normative debates among ordinary citizens concerning their moral judgments about which behaviors and bets are good, bad, and in-between, and about which tradeoffs they deem acceptable. 21 From this combination, we hope, better policies will emerge. One more group of social welfare program consumers should be mentioned. We call them resource drains. These are individuals whose use of program resources is vastly disproportionate to their number and sometimes to their moral claims. Examples include the frequently admitted and re-admitted hospital patients, often serial substance abusers, who account for a remarkably high share of total hospital costs; 22 patients who are on life support in a vegetative state for long periods of time; 23 recidivist criminals; 24 long-term welfare recipients; 25 and the chronically unemployed. 26 Some resource drains are bad apples (e.g., the criminals) or 21 The importance of this normative debate about resource allocations in public programs and the responsibility of officials to facilitate and structure it can scarcely be exaggerated. For some different approaches to this goal, see Foundations of Administrative Law (P. Schuck ed., 2d ed., 2004), at Christopher J. Zook and Francis D. Moore, High-Cost Users of Medical Care, 302 N.E.J.M (1980). The authors found that hospital costs were concentrated on a few patients; on average, the high-cost 13% consumed as many resources as the low-cost 87%. 23 [data] 24 [DiIulio data?] 25 E.g., Mary Jo Bane and David Ellwood, Welfare Realities: From Rhetoric to Reform (1994); U.S. Dept. of Health and Human Services, Indictors of Welfare Dependence: Annual Report to Congress 2004, espec. sec. III, indicators [data. Many of these people end up on SSDI, as Mashaw has shown] 9

10 bad bets (e.g., those on long-term life support), but many are good apples and a few may even be good bets (though their extremely high costs make this less likely). Despite the importance of resource drains, we shall not discuss them further unless they are also bad apples or bad bets both because they do not neatly fall into these categories, which are our chief interest, and because those who are good apples and good bets do not present promising opportunities for policy reforms. We sprinkle this paper with many real-world examples in order to illustrate how our analytical principles would play out in actual cases. We hope to make our most basic claim -- that policymakers should take the bad apples and bad bets problems seriously when formulating and implementing policy more compelling by showing that we are grappling with genuine policy problems, not tilting at windmills or attacking straw men. Some of our readers will criticize one or another of our examples. Some may insist that bad behavior does not mean that the actor is a bad apple. Others will say that it is wrong to exclude people of certain ages or medical condition from public benefits simply because they are bad bets, and still others will dispute our empirical claims for example, that removing disruptive students will improve the life chances of those who remain. Although we are eager to refine our understanding of the specific examples we cite, we are most interested in learning about other examples that would strongly illustrate our points. Our chief goal is to establish correct principles and then provide examples that can help test the plausibility of these principles. Readers who find our framework convincing should test it in their own policy domains. We expect, for example, that thinking about anti-poverty programs in terms of bad draws, bad bets, bad apples, and bad policies can extend the lessons of, say, the 1996 welfare reform, discussed in Part III, to a broader array of social problems and interventions, both actual and proposed. II. The Social Policy Stakes The stakes in clear thinking about the role of bad apples and bad bets in our social programs are quite high, for three reasons. First, as a fiscal matter, allocating program resources to bad apples and bad bets is very costly and wasteful, and becoming more so a particularly worrisome development at a time of growing budget stringency. Second, as a substantive matter, bad apples (though not bad bets) often prevent us from achieving our specific policy goals education for low-income children, for example. Third, as a political matter, the current misallocation of resources to bad apples and bad bets should be high on the agenda of welltargeted redistributionists, for the presence of bad apples and bad bets among program recipients provides much political ammunition for those who oppose these programs, and diminishes the social benefits produced per tax dollar, which is what concerns most citizens. We now take up each of these kinds of impacts. A. Fiscal Impact. Resources available for social programs are always limited relative to the demands made on them, of course, but rarely more than today. Although our economy is now much larger than it was when most of these programs were established, the claims against those resources have grown even more. The federal government is now running a deficit that is at a historical high in absolute terms. Although the deficit as a percentage of GDP is not as high as it was in the mid-1980s, it still represents a larger share of the economy than at any time since 199_. [RJZ find] The two largest expenditure categories in the federal budget Social Security 10

11 and defense are growing rapidly and will certainly increase in the future. Although the Social Security Trust Fund is now in surplus and will continue to be until roughly 2042, 27 both major parties seem to recognize the need to begin shoring it up soon with some combination of additional revenues from taxes and contributions, and from benefit reductions. Military expenditures, which declined significantly as a percentage of the budget during the 1990s, are expanding due to the demands of technological change, higher force levels, open-ended wars in Afghanistan and Iraq, growing global instability, and the threat of terrorism. In addition, both political parties are committed to substantially increased federal expenditures on health care for an aging population that, protected by Medicare from most of the cost, demands all the costly technologies that are flooding the medical marketplace. 28 The 2003 amendments to Medicare that create a new drug benefit for the elderly are just the beginning. Expanded medical benefits for children not now covered by Medicaid seem certain as the government moves, piecemeal, to reduce the size of the large and growing uninsured population. At the same time, neither party seems willing to support higher taxes; the 2004 Democratic platform proposed only to restore the pre-bush tax rates on those earning more than $200,000, which would do little to curb the deficit. Given these fiscal constraints, public expectations, and political commitments, there is simply no good alternative to the effort to stretch programmatic resources by targeting benefits more on those who can make the best use of them and who most deserve the public s solicitude and support. This means substantially excluding bad apples and bad bets. 29 We strongly suspect that the incidence of bad apples in most social programs grows absent reform. As time goes on, the always-strong incentive for bad apples to exploit public programs that are not meant for them increases as they gain more skill and sophistication in 27 See Henry Aaron and Peter Orzag, The Impact of an Aging Population, chap. 5 in Alice M. Rivlin and Isabel V. Sawhill, Restoring Fiscal Sanity: How to Balance the Budget (Brookings, 2004), p Many of the new technologies particularly aid the young, making them especially beneficial from a QALY perspective. E.g., Gardiner Harris, Panel Reviews New Vaccine That Could Be Controversial, N.Y. Times, Oct. 27, 2004, at A12 (new treatment for bacterial meningitis is the first time government has had to decide whether a vaccine is worth the cost). 29 Two ostensibly alternative ways to increase social benefits under these fiscal constraints deserve brief mention. The first, attacking fraud and abuse in public programs, is always popular, and every administration promises to achieve it. Like everyone else, we favor this strategy, but it is largely outside this paper s purview. (We discuss it very briefly in part III). Indeed, as part III explains, the kinds of bad apples who principally concern us here are especially worrisome precisely because they are not engaged in fraud and abuse but rather are considered legitimate program participants. A second alternative, shifting the costs of public policies to the private sector through regulatory mandates, merely changes the cost-bearer; it does not effect a more efficient and fairer distribution of the benefits, much less expand them. 11

12 doing so for example, with tax shelters, nursing home reimbursements, and minority small business contracts. Thus, people learn how to turn ordinary expenditures into investments in order to claim investment tax credits, siphon off assets from a person entering a nursing home, or use a minority front to qualify for a business subsidy. A second reason for the growing bad apples problem is moral hazard. This is the tendency of people, knowing that others (usually a public or private insurer) will pay the expected cost of risks they run, either to take on more risk or to choose socially undesirable conduct for which they will be partly or fully compensated (e.g., not starting a new job until unemployment insurance benefits run out). Ex ante moral hazard encourages risky pre-harm conduct; ex post moral hazard encourages post-harm conduct that increases both program costs and the claimant s compensation. 30 Some programs benefits for the elderly, widows, wounded war veterans, and the seriously disabled, for example entail little or no moral hazard. This is either because the beneficiaries have no way to affect the occurrence or the magnitude of the compensable event (old age, spousal death), or because the compensable event serious disability, for example is so harmful that no beneficiary would willingly bring it on himself. Historically, and particularly during the New Deal, social welfare policy was designed inadvertently at times to minimize moral hazard. That is, it targeted bad draws whose misfortunes occurred through no fault of their own. Since these misfortunes could befall anyone, the risks of such losses could be reduced or socialized by being spread among a much larger pool of people. Social insurance of this kind, and many other programs as well, is like an insurance contract in which one does not know ex ante (except in an anonymous, statistical sense) who will be sick and who healthy, who disabled and who able-bodied, who laid off and who employed, and so forth. Such programmatic features help to justify the transfer of resources to those who turn out to be the unfortunate bad draws. We suppose that a rational and risk-averse citizen behind the Rawlsian veil of ignorance would favor such a system. These features also help make the programs actuarially feasible. More recent policies, however, create much more moral hazard, encouraging far greater utilization of program resources. Examples include the mandated equal treatment of mental health and physical health coverage, the expansion of Social Security to cover more temporary 30 [cite to Trebilcock article] Flood disaster programs, for example, generate ex ante moral hazard; they encourage people to locate in flood plains. Similarly, deposit insurance encourages banks to make riskier investments. Neither of these programs charges premiums that are effectively risk-rated. A well-documented example of ex post moral hazard is unemployment insurance, which encourages recipients to malinger or delay re-employment, as officials cannot easily verify the genuineness of their job searches and choices. [cite to Feldstein]. Workers compensation programs are plagued by easy-to-claim and hard-to-disprove back pain complaints. [cite] Extreme forms of ex ante moral hazard are policies that reward self-destructive behavior, Examples include public coverage of costly treatments for emphysema (almost always caused by smoking), or of organ transplants for heavy drinkers. 12

13 and partial disabilities, and drug coverage under Medicare. 31 Nor is the legislation-induced moral hazard limited to social welfare programs. Quite the contrary; federal deposit, pension, mortgage, and other commercial guarantees subsidizing business dwarf those involving personal moral hazard. 32 This sequence is perfectly predictable. We should expect politicians hoping to appeal to voters by expanding government largesse to begin by proposing programs that address risks to which most or all people are exposed, that program beneficiaries are helpless to avoid, and that can therefore be operated on sound actuarial principles. Once such median voter programs 33 are in place, the ambitious politician will move on to propose others that exhibit fewer of these advantages. In this way, politicians work their way down the ladder of social program possibilities. As they descend this ladder, their programmatic choices inevitably entail evergreater compromises with the ethical and efficiency goals of minimizing moral hazard, producing diminishing social returns. Substantive Impact. Bad apples, unlike bad bets, inflict much more harm than merely squandering precious program resources. Perhaps more important, they sometimes undermine our substantive social policy goals. Drunk drivers do not simply eat up a disproportionate share 31 Professor Mashaw points out that virtually every income support program, from SS pensions to the EITC, is constructed with elaborate attention to moral hazard issues... He notes that vocational rehabilitation and other service programs have elaborate restrictions on overserving bad bets and that Medicare spending on catastrophic care has caps and nursing home care is restricted to 3 months. dated Dec. 27, See David A. Moss, When All Else Fails: Government as the Ultimate Risk Manager (2002); U.S. General Accounting Office, Financial Management: Status of Governmentwide Efforts to Address Improper Payment Problems (Oct. 2003). A prominent social policy analyst uses this proliferation of pro-business moral hazard to argue in favor of creating even more of it (in the form of what he calls universal insurance against sharp income declines) so long as it favors families rather than businesses. Jacob S. Hacker, False Positive, The New Republic, Aug. 16 & 23, 2004, p.17. But Hacker s view of moral hazard is too simple, ignoring its opportunistic character. His moral hazard-increasing proposal would enable some families to exploit it say, by reducing work effort at the expense of those who (1) do not try to game the system (i.e., families that do not reduce their work effort), and (2) pay for the system (i.e., taxpayers). Proposals like Hacker s, due to their other more attractive features, gain support and let moral hazard snowball. Another example of well-intended measures that create more moral hazard is the policy of giving homeless families the top priority in getting subsidized apartments. See Leslie Kaufman, Homeless Families Blocked From Seeking U.S. Housing Aid, N.Y. Times, Oct. 20, 2004, at B1 (policy making poor families believe that fastest way to get own apartment is to become homeless).[i think Ellickson has some harder evidence on the effect of New Haven s generous policies] 33 [explain and cite] 13

14 of police, emergency rescue, and hospital resources. They also destroy the lives and well-being of their victims. Much of the analysis in part III addresses the substantive impact of bad bets. There we summarize some data on two particularly troubling and widespread examples chronically disruptive public school students who reduce the educational opportunities of their classmates, and chronically disruptive public housing tenants who impair the other tenants quality of life. C. Political Impact. We think that the experience with social welfare policies particularly the 1996 welfare reform, the Earned Income Tax Credit (EITC), and [* student loan program?] suggests an important but often-ignored political lesson that is highly relevant to the present discussion. Taxpayers are hostile to programs that benefit people, poor or rich, 34 whom they consider, rightly or wrongly, to be bad apples. Politicians not only reflect this popular animosity; they skillfully exploit it in their own interests. But this is only half the story and the more familiar half. Less well understood is the fact that taxpayers, and their representatives, are sometimes willing to support social programs that target benefits to poor people whom they consider good apples by which they seem to mean those who are poor because they are bad draws, who have not caused their own misfortunes by making immoral choices and squandering their opportunities, and who are prepared to work hard to improve their prospects if given the chance. In our terms, many taxpayers and politicians who support well-targeted redistribution criticize existing programs less because of the programs goals than because of the bad apples who taint them. If we are right about this, then programs that screen out or remove bad apples will tend to increase their political legitimacy, which in turn can improve their long-run public support and funding at least on a per capita good apple basis. The leading recent example of this political dynamic is the welfare reform law of Among other things, this law imposed a stiff work requirement designed to prevent recipients deemed capable of work bad apples, in Congress s view -- from remaining on the rolls indefinitely. The new program (Temporary Assistance for Needy Families (TANF)) maintained the pre-reform funding level in absolute terms and, because the number of those who remained on the rolls (good apples) declined, increased per capita good apple funding significantly this, despite the election of a much more conservative President and Congress. 35 Another example is the Earned Income Tax Credit (EITC), a federal transfer program that augments the income of low-wage working people. Since 1988, Congress and the IRS have taken a number of steps to reduce fraud and other noncompliance with the program s requirements. Ostensibly, these measures have reduced the number of bad apples receiving the EITC and 34 [re the rich: discuss tax/citizenship penalties for offshore tax evaders] 35 The TANF program places significant discretion with the states on how to spend its monies. In many states, a part of the expenditures do not go directly to recipients but are used for job training, child support, and other work-related programs. 14

15 stemmed the loss of political support suffered by the pre-reform program. 36 In this way, the reforms that eliminated bad apples helped to more than double the real dollar value of EITC benefits. 37 We cannot say, of course, what would have happened if the welfare program had not been reformed, or if EITC compliance rules had not been tightened, so we can only infer how the political legitimacy of these recipient groups would have fared absent these reforms. What we can say, however, is that the bad apples critique of these two programs is less prominent today even in a decidedly more conservative era politically than it was a decade ago. 38 In the welfare reform case, for example, Congress provided additional funds with the hope and expectation that the reform, if successful, would eventually lead to savings, as transfer payments were replaced by wage income. Sometimes, the political linkage between bad apples reform and more resources for the good apple poor can be inferred even when it is not altogether explicit. Christopher Jencks points out that the Clinton administration s welfare policymakers thought that they needed to expand the EITC before they attempted welfare reform because they had to convince both themselves and those to their left that single mothers with minimum wages could make it. Congress also raised the minimum wage within days of passing welfare reform, despite the hostility of the House leadership under Newt Gingrich to such proposals. Finally, the states that reduced the number of welfare beneficiaries spent a lot of money on childcare subsidies and other job supports, and some states raised their EITCs. 39 As Jencks puts it, Was that linkage? 36 For example, a 1997 study by the IRS of tax year 1994 found 26 percent of dollars were overclaimed, down from 35 percent in Although efforts to reduce error were instrumental in maintaining political support for the EITC, the most recent estimates of EITC non-compliance are that 30 percent of dollars are paid in error. [* cite] This reduction in bad apples is even more impressive given that the larger EITC benefits that these reforms facilitated politically increased recipients incentives to engage in fraud and other non-compliance. 37 See Jeffrey Liebman, The EITC Compliance Problem, Joint Center for Policy Research News, Summer Unfortunately, as Liebman observes, though the reforms have improved compliance 1998 overpayments were 21% as opposed to an estimated 28%, absent reforms non-compliance remains high. Congress has continued to pass reforms that it hopes will reduce non-compliance and thus enhance the program s legitimacy. (Jeffrey Liebman, personal communication, September 22, 2004). Apparently, political support for programs depends more on the perceptions of the number of bad apples than on the reality. 38 [cite to Liebman s work] 39 A recent report suggests that they are continuing to do so. Robert Pear, States Pocketbooks Are Fuller, But Health Costs Stall Recovery, N.Y. Times, Dec. 17, 2004, at A26 ( States have poured money into education, training and child care to help welfare recipients get and keep jobs. Forty-four states said they would maintain cash assistance benefits in 2005 at the levels in effect this year. Five states...said they planned to increase cash assistance benefits next year... ). 15