Cost-Shift Theory Overview

While the cost-shift hypothesis has issued from multiple quarters, a representative version of the argument comes from Thomas Burke (Reference Burke2001, 1264–65):

Note that there are several aspects to Burke’s rehearsal of the theory. To begin, while Burke focuses on environmental regulation as a motivating example, the cost-shift theory is a general one. It has been applied to a wide range of substantive regulatory interventions, including antidiscrimination laws of all stripes, health care regulation, and constitutional torts under section 1983 of the US Code.Footnote ⁴

Moreover, the various articulations of the theory span at least three distinct modes of cost shifting, as depicted in Figure 1.Footnote ⁵ First, legislators who wish to achieve certain social ends can realize budgetary savings by forgoing government action financed by explicit taxes and, instead, impose a regulatory mandate—that is, a duty to do or not do something—that operates as an implicit, or “regulatory,” tax (Kelman Reference Kelman1999). The “accommodation mandates” at the heart of the ADA, as already noted, provide an obvious example. Another example comes in the environmental context. Legislators who are concerned about pollution and prefer explicit taxation as a regulatory approach could impose a general tax (on all citizens) or a targeted tax (on polluters only) and then use the proceeds to compensate those affected by pollution or subsidize their relocation away from emitters. Alternatively, legislators who wish to avoid explicit taxation and minimize on-budget costs could impose a regulatory mandate requiring heavy industry to deploy particular technologies to abate pollution. The result would be an off-budget, implicit tax on production.

Figure 1.

Three levels of cost shifting via regulatory choice. Note that cost shift is reflected in moves from left to right.

Once legislators have chosen a regulatory mandate over a tax-financed redistributive or other program, legislators can move regulatory costs off budget in two further ways: a shift in the institution charged with adjudicating the enforcement efforts or a shift in the identity of the enforcer (Burke Reference Burke2002).Footnote ⁶ With the first method, budget-conscious legislators can move enforcement efforts from “in-house” agency adjudications to agency-filed suits in the courts, or they can shift adjudication entirely off budget by relocating it from public bodies (agencies, courts) to private arbitration or other forms of privately funded “alternative dispute resolution.” With the second, legislators can shift responsibility for enforcement from public actors (agency administrators, prosecutor’s offices) to private actors (civil litigants). Combining both cost-shifting approaches yields budget-protecting moves that go well beyond the simple move from explicit to implicit tax financing. As a concrete example, cost shifting might entail a move from a scheme in which a public agency initiates and prosecutes enforcement actions against regulatory targets inside the agency before an administrative law judge—a power currently enjoyed by agencies like the Securities and Exchange Commission and the Federal Trade Commission—to one in which private litigants alone initiate and litigate civil actions against regulatory targets, whether in court or before private arbitrators.

Finally, and of critical importance for the present study, the cost-shift theory contains a key and empirically testable implication: cost shifting should be more attractive to legislators who face a “tight budgetary constraint” (Burke Reference Burke2001, 1265; see also Farhang Reference Farhang2008, 827). Legislators with a choice of regulatory instruments with which to achieve policy ends should, all else equal, prefer statutory mandates enforced via private litigation over on-budget regulatory approaches during lean economic times, when public budgets are stretched thin.

Critiques of the Cost-Shift Theory

Anatomizing the theory in this way suggests that the cost-shift theory is not without its problems. First, the theory lacks a limiting principle since virtually any regulatory mandate designed to mitigate social harms will realize on-budget savings in the modern social welfare state. To provide just one example, regulations designed to abate pollution may reduce costs incurred via publicly financed health care for poor children suffering from asthma.

Second, one might question the potency of cost shifting as a legislative strategy. For instance, the move from in-house administrative enforcement and adjudication to private lawsuits in court is not, as Sean Farhang (Reference Farhang2008, 828) puts it, a “free lunch.” Courts are expensive to run, especially where thick judicial procedures lead to more protracted proceedings than in a more procedurally streamlined administrative context. More fundamentally, the move from programs financed via explicit taxes to regulatory mandates that operate as implicit taxes may not yield on-budget cost savings at all except in the short term. This is because regulatory costs, whether borne by regulated entities alone or by regulatory beneficiaries via higher consumer prices or lower employee wages, can depress public tax receipts by reducing taxable production, consumption, or both (Kelman Reference Kelman1999; Donohue Reference Donohue2001). Where the reduction in receipts equals or exceeds the amount of explicit taxes that would have been necessary to fund alternative policies or programs, legislators who choose regulatory mandates over explicit taxation will realize only temporary budgetary savings. In other words, to the extent that we observe legislators acting with a cost-shifting motivation, they may be aiming to reduce visible short-term costs rather than reduce long-term budgetary pressure.Footnote ⁷ Such short-term motivations are consistent with the generally accepted view among political scientists that politicians are single-minded seekers of reelection (D. Mayhew Reference Mayhew1974). Even if the cost-shift strategy may not yield durable budgetary benefits, it may explain legislative decision making.

Third, a thorough conceptual mapping of the cost-shift hypothesis suggests that whether and when the hypothesis holds will turn on context-specific aspects of the interest-group environment and the regulatory intervention at issue. In particular, budgetary constraints, which often reflect poor macroeconomic performance, can also create political constraints. Economic downturns stretch public and private budgets alike, leaving regulated entities especially resistant to new and costly statutory mandates. To that extent, we might expect that sagging macroeconomic fortunes, far from pushing legislators toward off-budget regulatory approaches, can have the precise opposite effect, making legislative adoption of such approaches less likely rather than more likely. Similarly, the validity of the cost-shift hypothesis is likely to turn on the extent to which regulatory costs are in fact shouldered by regulated entities or, instead, are passed through to regulatory beneficiaries, whether consumers (via higher prices) or employees (via lower wages) (Kelman Reference Kelman1999; Donohue Reference Donohue2001). At the theoretical limit, where pass-through of regulatory costs is relatively frictionless and complete, it is the regulated entities who might prove indifferent to an on- or off-budget regulatory approach, while the regulatory beneficiaries may have quite strong design preferences that are relatively immune to fiscal or economic health. (We recognize that this scenario is largely theoretical.)

In short, the cost-shift theory is hardly without theoretical and empirical ambiguities. It is thus highly problematic that the theory has never been subjected to rigorous empirical test (with one limited exceptionFootnote ⁸ ), despite being a mainstay in the scholarly literature on the rise of the American litigation state. The remainder of this article offers the first rigorous empirical test of the theory by focusing on an area where the cost-shift theory has been especially prominent: disability discrimination laws.

Postwar Trends in Social Provision for People with Disabilities and the Cost-Shift Hypothesis

Turning first to evidence of postwar policy trends, the cost-shift hypothesis finds support in the fact that the state-level drive to enact disability discrimination laws came amidst a seismic shift in the shape of American social provision for people with disabilities. The main development was a move away from publicly controlled rehabilitation programs centered around the provision of “curative” medical services and, for those deemed incurable, a system of state- and privately run residential institutions, which was part of a vast “rehabilitation empire,” in the words of a one historian (O’Brien Reference O’Brien2001, 98), or “an insular and extensive disability gulag,” according to a less forgiving account (Pelka Reference Pelka2012, 49). Gradually replacing it was a regime with three components: a more streamlined, voucher-based, privatized system of rehabilitation; community-based services and integrated public school education; and income-maintenance programs. The centerpiece of the latter was the creation in 1956 of the SSDI program, financed via payroll taxes and available to workers with disabilities who had paid into the system. It was strengthened in 1972 with the Supplemental Security Income (SSI) program, a means-tested income-maintenance program available to indigent people with disabilities and, among others, those who were disabled at birth or before they were old enough to hold a job.

Such a fundamental shift in policy emphasis did not come easily. Disability activists had to break the grip of a small army of “rehabilitation professionals” who populated, and benefited from, the incumbent regime and so actively opposed any legislation charting a new course (O’Brien Reference O’Brien2001, 63–65, 113–16; Pelka Reference Pelka2012, 16, 179). Paving the way for fundamental policy change was concern about the mounting costs of rehabilitation services. A key moment came during the Senate hearings in 1966 regarding federal disability programs, including testimony by Health, Education, and Welfare chief Wilbur Cohen that federal expenditures on behalf of people with disabilities had reached nine billion dollars per annum, spanning some eighty different programs and accounting for more than 12 percent of total federal outlays (Treanor Reference Treanor1993, 40).

Just as important were efforts by disability activists to effect a shift in prevailing conceptual approaches to disability, from the “medical” model that had underpinned the reigning curative-rehabilitative approach to a “civil rights” model that pushed away from concepts of dependency and capacity and instead saw autonomy and independence for people with disabilities as human rights (Scotch Reference Scotch2001; O’Brien Reference O’Brien2001; Bagenstos Reference Bagenstos2009). An important component of this view was that high rates of joblessness and poverty among people with disabilities were not the result of physical or mental impairments but, rather, societal attitudes and a built environment that did not take into account the needs of people with disabilities, thus violating those rights (O’Brien Reference O’Brien2001, 14; Scotch Reference Scotch2001; Bagenstos Reference Bagenstos2009). Bound up with this changing conception of disability was a new direction in disability politics. The disability community had long been riven by internal disputes among siloed, medically defined disability groups, much of it fueled by concern among organizational leaders that a cross-disability movement would not fully serve the needs of their particular members (Pelka Reference Pelka2012, 23, 468; Scotch Reference Scotch1989). The rights-based approach helped bridge differences among disability groups and removed the collective-action barriers that had long muted the voice and influence of people with disabilities within American politics (Treanor Reference Treanor1993, 33, 61; Scotch Reference Scotch1988, 161; Reference Scotch2001, 6; O’Brien Reference O’Brien2001, 11; Bagenstos Reference Bagenstos2009).

A secondary shift in American disability policy as legislative campaigns got underway in the 1960s and 1970s took the form of rapid federalization. A good example is the Vocational Rehabilitation Act Amendments of 1965, which expanded federal funding to the states for vocational rehabilitation programs to three hundred million dollars and reduced the required state “match” from fifty–fifty to seventy-five–twenty-five, which then dropped further, to eighty–twenty in 1968 (Scotch Reference Scotch2001, 23).Footnote ¹⁷ Two other examples, detailed previously, are the creation of the SSDI program in 1956 and the SSI program in late 1971 and early 1972, which substantially increased federal funding of income-maintenance programs. Interestingly, both of these seismic shifts in American disability policy—the move away from rehabilitation and state-run residential facilities and creeping federalization—tend to cut against the cost-shift theory as an explanation for the enactment of state disability discrimination laws. Indeed, both shifts suggest a state-level fiscal burden that was decreasing, not increasing, as state legislative campaigns gathered momentum. However, any such conclusion ignores a third seismic development in social provision for people with disabilities that posed a significant threat to state budgets and thus offers far stronger support for the cost-shift hypothesis: the creation of the Medicaid program in 1966.

Two distinct phases of the Medicaid program’s evolution offer support for the cost-shift hypothesis. The first phase came during the first half of the 1970s, when states enacted an initial wave of disability discrimination laws. The years 1966–71 were startup years for Medicaid, with some states creating programs immediately and others needing several years to get underway (Klemm Reference Klemm2000). Only three states—Alaska, Nevada, and Illinois—enacted disability discrimination laws during this span. Soon after, however, the Medicaid program entered a period of rapid expansion, spurred by the amendments to the Social Security Act that created the SSI program in 1972.Footnote ¹⁸ As noted previously, those amendments continued the trend of federalizing the costs of income-maintenance programs for people with disabilities and so likely reduced certain state expenditures.Footnote ¹⁹ But they also had a key collateral effect: federal law required states to cover SSI recipients under Medicaid, and the SSI program’s launch was accompanied by extensive federal outreach efforts designed to identify and enroll eligible recipients. The result was “significant increases” in Medicaid enrollment among people with disabilities (Klemm Reference Klemm2000, 107). Driven in significant part by the inflow of enrollees with disabilities, total Medicaid enrollment increased by almost 5 percent per year over the period 1972–76. More importantly, total Medicaid expenditures grew by an average of 18 percent per year over the same period, as large numbers of people with disabilities—“among the most expensive groups in Medicaid”—flowed into the program (Klemm Reference Klemm2000, 107).

With Medicaid rolls swelling among people with disabilities, states were highly vocal about the resulting budgetary pressures. A signal example came in the litigation effort leading to the US Supreme Court’s decision in Alexander v. Choate.Footnote ²⁰ In this case, the US Supreme Court rejected a challenge by disability activists who sought to use section 504 of the Rehabilitation Act of 1973 to challenge Tennessee’s effort to cut back on Medicaid expenditures by reducing the number of inpatient hospital days provided to each eligible recipient.Footnote ²¹ The policy was discriminatory, the plaintiffs unsuccessfully argued, because Medicaid enrollees with disabilities had “greater medical needs,” to use the court’s phrasing, and so would be disparately affected by the state’s cost-saving measures.Footnote ²² What is most revealing about the case is the state’s detailed articulation of the budgetary pressures it was facing from Medicaid enrollees with disabilities in particular. More than one-quarter of Medicaid recipients, the state reported to the court, were disabled (90,668 out of 341,960), and they were heavy users.Footnote ²³ Enrollees with disabilities who used hospital services were four times as likely to require more than fourteen days of care than users without disabilities (27.4 percent versus 7.8 percent).Footnote ²⁴ Similarly, enrollees with disabilities were nearly one-and-a-half times more likely (18.6 percent versus 13.6 percent) to use inpatient hospital care each year.Footnote ²⁵ Though Tennessee conceded that federal-state matching meant it paid only 30 percent of total Medicaid reimbursements, the “astronomical growth” in Medicaid expenditures had been a “financial calamity” for the state, which could not be addressed via deficit spending because of the state constitution’s balanced-budget requirement.Footnote ²⁶

Budgetary pressures in the first phase of Medicaid expansion were plainly felt well beyond Tennessee. Figure 2 plots the explosive growth of Medicaid spending for people with disabilities from 1975 (the first year for which disability-specific data is available) to 1995, five years after the ADA was enacted. During that roughly twenty-year span, total Medicaid costs for people with disabilities more than tripled, from $14.8 billion to $49.4 billion.Footnote ²⁷ Medicaid rolls also grew quickly during this time: total Medicaid enrollees with disabilities grew from 2.4 million to nearly 5.4 million, and the proportion of Medicaid recipients defined as “disabled” rose from 11 percent to more than 16 percent.Footnote ²⁸

Figure 2.

Total Medicaid payments on behalf of people with disabilities, 1975–95 (reported in billions of 2009 dollars).

The resulting budgetary pressures only worsened as the Medicaid program entered a second distinct phase during the 1980s. After ten years of rapid program growth, the 1980s saw a new turn in American welfare state politics, as rising concern about runaway entitlement spending and Ronald Reagan’s election fueled a range of efforts to retrench the social welfare state (Pierson Reference Pierson1994). Medicaid retrenchment came most prominently in the Omnibus Budget Reconciliation Act of 1981, in which Congress instituted a three-year reduction in federal financial participation by substantially cutting the federal matching rates that states enjoyed (Klemm Reference Klemm2000, 108).Footnote ²⁹ The result was that states were shouldering a growing share of rising Medicaid costs just as states enacted a new wave of disability discrimination laws, many of them of the stronger “accommodation mandate” variety, during the period from 1981 to 1985. As with the first phase, this second phase of Medicaid’s evolution offers suggestive support for the cost-shift hypothesis.

State-Level Legislative History and the Cost-Shift Hypothesis

In contrast to evidence from national policy trends, legislative history materials drawn from particular state-level campaigns to enact disability discrimination laws are sparser and more mixed in their support for the cost-shift hypothesis. In reviewing this evidence, we are conscious that legislators might have been motivated by the cost-shifting consequences of disability enactments without saying so; indeed, it is possible that some might have tried to hide that motivation. For that reason, we draw not only on records of public legislative debates but also on other records, which are unlikely to be salient to the public, on which cost-shifting legislators might have relied, such as committee reports and letters from interest groups.

We have sought to collect as comprehensive a legislative record as possible. Our research drew on state archives, state libraries, and the Library of Congress. We were able to find far more comprehensive legislative materials in some states than in others. We have dedicated our qualitative discussion below to the states in which we were able to find more comprehensive materials. The evidence is scarce because few states, even to this day, systematically record or maintain formal legislative history. Perhaps more importantly, what little formal legislative history exists suggests that the legislative debate over state disability discrimination laws was not especially heated, at least not publicly so, leaving little occasion for protracted and potentially revealing back-and-forth discussion of the sort that accompanies more contested legislative efforts. As reflected in Table 3’s accounting of the few states for which voting records are available, legislative votes at final enactment of disability discrimination laws were rarely close, even for laws that were imposing an accommodation mandate. The sole exceptions include New Mexico’s enactment of an accommodation mandate in 1983 and also an indication in legislative testimony in Vermont that the bill that ultimately made it into law in 1973 had only narrowly passed the House.Footnote ³⁰

Table 3.

Final legislative votes in enacting states

Note: Asterisk (*) denotes “accommodation” law.

Nor did the state-level drives exhibit a strong partisan cast. For instance, during Pennsylvania’s 1974 campaign, legislators described a bipartisan process, with a House committee chairman noting that the bill was “an example of effort on both sides of the aisle to pass a much needed piece of legislation.”Footnote ³¹ Indeed, as we show in Tables 5 through 7, partisan control of the legislature was not a significant predictor of enactment. This is not to say that the laws did not draw opposition. Business groups opposed them in many states, suggesting a standard left-right pattern of support and opposition that characterized legislative campaigns to enact other kinds of job discrimination laws (Chen Reference Chen2009). In Oregon, a representative of Associated Oregon Industries appeared at a hearing of the House Committee on State and Federal Affairs to warn that the bill would lead to affirmative action for the handicapped and might place employees with disabilities at risk of injury.Footnote ³² However, the company was alone in opposition among those registered to testify, leading the committee chair to cut short the hearing and turn immediately to a unanimous vote of all six committee members present.Footnote ³³ In other states, some business groups actively supported the proposed law. As Wisconsin considered a bill transforming its disability discrimination law from a simple antidiscrimination provision to an accommodation mandate in 1982, the Wisconsin Merchants Federation and the Wisconsin Association of Manufacturers and Commerce were in favor.Footnote ³⁴ No group—business or otherwise—appeared or registered against the bill that would ultimately win passage. Finally, an atmosphere of consensus prevailed in Rhode Island, where the Governor’s Commission on the Handicapped heard testimony from state representatives who unanimously supported the bill. Partisanship seemed relatively unimportant. Although four of the five representatives who presented were Democrats, Representative Gilbert Indeglia, a Republican, spoke in support of broadening the definition of handicap to include mental disabilities, in line with the federal definition.Footnote ³⁵

The lack of publicly voiced opposition meant that floor debate was brief and rarely strayed beyond the invocation of the dignitary interests of people with disabilities. In Pennsylvania, one representative spent a day on the House floor in a wheelchair “to dramatize the plight of the handicapped in our society,”Footnote ³⁶ noting afterward that the simple difficulties of moving in a wheelchair were more trying than expected.Footnote ³⁷ Other Pennsylvania legislators drew attention to veterans returning home from Vietnam and noted the special obligations owed to those with service-related disabilities. “It is a poor welcome home [for veterans],” Representative Ivan Itkin, the sponsor of the 1974 Pennsylvania legislation said, “when our buildings prohibit them from taking a chemistry class.”Footnote ³⁸

The few states with more extensive legislative records offer at best mixed evidence that budgetary dynamics and cost shifting played a key role in the debate. In some states, budgetary and social welfare dynamics do not appear to have registered at all. During North Carolina’s passage of an accommodation mandate to replace its simple antidiscrimination provision in 1985, the otherwise voluminous legislative record contains no discussion of cost shifting at all. According to the minutes summarizing the North Carolina’s House Judiciary Committee’s deliberations on the ultimately successful bill, the main subject of discussion was the statement of the Governor’s Advocacy Council targeting three “trouble areas” in the prior simple antidiscrimination legislation.Footnote ³⁹ These areas were: “(1) no true definition of handicapped people; (2) no provisions for reasonable accommodations for handicapped people; (3) no viable means of enforcing the [law’s provisions].”Footnote ⁴⁰ Also discussed was the inclusion of “reproduction” disabilities in the definition of handicap as well as the affirmative defenses of employers. Nowhere were budgetary or redistributive issues noted.

In other states, the few statements by legislative supporters and disability advocates that explicitly raised budgetary costs were mostly brief and unelaborated, suggesting that reducing pressure on social welfare programs was not central to their messaging. In California, an “Enrolled Bill Report,” a document prepared for the governor by the legislative counsel upon an enacted bill’s presentment for signature, reported that a disability discrimination law would “make it possible to obtain more job opportunities for disabled citizens who are now on welfare rolls.” Somewhat more detailed evidence on cost questions comes from New York’s 1974 campaign and the packets of materials that accompanied transmittal of the bill to Governor Malcolm Wilson for his signature. For instance, the executive director of the New York State Nurses Association noted in a telegram to the governor when the bill was still under legislative consideration that the proposed law would not only provide for equal rights and social integration of people with disabilities but would also “decrease need for public welfare.”Footnote ⁴¹ Similarly, when a New York assembly member sent a letter to the governor apparently intended to defuse possible objections to the bill, he argued, first, that the bill contained sufficient protections for employers and, second, that “[m]illions of public dollars could be saved by giving the handicapped an opportunity to work.”Footnote ⁴² Neither letter, however, offered specifics.

A few more detailed statements about budgetary issues can be found in the memoranda and letters sent to the New York governor by executive departments and interest groups urging that he sign or veto the enacted bill. Only five of the twenty-one agency memos in the file made any mention of the budgetary implications of the bill. Of these, only two mentioned possible savings from social services. The others emphasized increased costs to the New York Human Rights Division, the state agency that would adjudicate disputes where a complainant asserting disability discrimination elected to proceed administratively rather than by filing a lawsuit in court. Similarly, of the eighteen letters in the file from interest groups, only two mentioned budget consequences. A lawyer representing a coalition of handicapped organizations argued that, if the bill became law, “[m]illions of dollars in public welfare and private support payments to the jobless, disabled or handicapped will be saved by giving them new hope and opportunities for employment, which they desperately seek.”Footnote ⁴³ Another organization supporting the bill argued that it would save money by reducing enrollment in state mental institutions.Footnote ⁴⁴ No other individuals mentioned any possible fiscal implications. Together, these communications suggest that cost-based arguments were plainly in play in the legislative debate but may not have occupied a central place therein (Appendix 1 catalogs these memos and letters and summarizes their comments on the bill’s budget effects).

A similar story emerges in Vermont, one of few states for which relatively intact legislative history is available due to unusually detailed records of hearings before the Senate General Affairs Committee in 1981. The hearings featured representatives from a wide range of government agencies and interest groups, including the Office on Aging, the Coalition of Vermont Elders, the Vermont Coalition of the Handicapped, the Vermont Association of Business Industries and Rehabilitation, as well as the State Chamber of Commerce and Retail Association. Yet, as in other states, the bulk of the debate either centered on the dignitary interests of people with disabilities or else tracked other concerns about antidiscrimination laws: whether state prosecutors and agency administrators might prove to be overzealous enforcers or whether a small employer could be liable for hiring his or her own child over an applicant with a disability.Footnote ⁴⁵

When Vermont bill supporters addressed the cost of social welfare programs serving people with disabilities, their statements only skirted the possible fiscal benefits of disability discrimination laws. Lee Viets, the president of the Vermont Coalition of the Handicapped, testified that “[s]ociety can no longer afford the questionable luxury of rehabilitating and providing education and training for handicapped children and adults only to relegate them to the sidelines cast aside as non-productive children unable to repay the effort expended on them.”Footnote ⁴⁶ But Viets did not directly articulate public cost savings as a benefit of the proposed law, and this portion of her testimony produced no follow-up from committee members. Similarly, no senators followed up on the issue when Betsy Hummell of the Vermont Association of Business and Rehabilitation noted in passing that the people with disabilities she knew who were now employed had previously been on public assistance.Footnote ⁴⁷

Perhaps most revealing in the Vermont legislative drive is that, when cost questions surfaced most explicitly, the focus was on how to mitigate employer costs, even where doing so would increase, not decrease, pressure on government programs and public budgets. In a letter to Vermont House Labor Committee, chair John Murphy, a lawyer for an employer association called Vermont Life, noted that, in contrast to companies that contract with the federal government and can pass on to it the costs of section 554’s requirement of reasonable accommodation, a Vermont company covered by the proposed accommodation mandate would not have “a choice” but to accept the “costs it will incur in meeting obligations to accommodate the handicapped” and, moreover, would not “be able to pass on its costs to its customers if competing manufacturers from other states do not have a similar handicapped law.” “The effect the ‘reasonable accommodation’ aspect of this bill,” the letter concluded, “is to shift to Vermont business many of the costs and expenses of meeting the needs of the handicapped in our society.”Footnote ⁴⁸ Testimony by a representative of the Vermont Retail Association sounded the same note, asking rhetorically:

A similar line of argument aired during Washington’s successful drive to enact a disability discrimination law in 1973. One senator asserted that the bill in question would reduce the burden on the state, but he quickly noted that his main concern was employers: “Shouldn’t any increased health care cost, he asked, be borne by the state rather than employers?”Footnote ⁵⁰ The senators quickly settled on a solution, adding the following amendment:

In both states, these dynamics suggest that, even if legislators had a cost-shifting strategy in mind, they faced formidable political constraints in acting on it.Footnote ⁵² This underscores a possibility raised previously: there may be basic political limits on cost shifting as a legislative strategy.

Canvassing available state legislative materials permits a final observation: while state legislative campaigns featured substantial debate about implementation choices, such as whether to grant government agencies the authority to enforce the new disability discrimination laws or, instead, to provide an aggrieved individual with a private right of action, rarely did actors expressly frame the choice in budgetary or cost-shifting terms. Thus, in Oregon, legislators unsuccessfully sought to move the determination of what constitutes a “reasonable accommodation” to the Bureau of Labor, rather than to the courts, on the grounds that such a technical and potentially evolving consideration “lends itself to administrative rule.”Footnote ⁵³ But the resulting debate made no mention of budgetary dynamics. Similarly, the legislative drive in Massachusetts featured dueling bills, with some providing for purely administrative enforcement by the Massachusetts Commission against DiscriminationFootnote ⁵⁴ —established in the 1940s when the state enacted one of the first laws prohibiting job discrimination on the basis of race, national origin, and religion (L. Mayhew Reference Mayhew1968)—and others providing for civil actions for damages. However, at no point did the public debate explicitly reference possible budgetary cost savings from judicial adjudication compared to the agency adjudication of complaints.Footnote ⁵⁵ Only in California did the debate about institutional choices even implicitly connect the choice of enforcer to budgetary dynamics: when California’s new disability discrimination law was presented to the governor for signature, the “Enrolled Bill Report” noted previously contained a “veto” recommendation from the state’s Department of Industrial Relations and the Fair Employment Practices Commission on the ground that the agencies did not want to be responsible for the law’s implementation without additional resources.Footnote ⁵⁶

In short, the qualitative evidence drawn from state-level legislative campaigns in support of the cost-shift theory is mixed. Cost shifting plainly figured, though not prominently, in at least some state-level debates. But cost shifting was forcefully articulated, not by disability rights advocates touting budgetary savings but, rather, by employers who stood to most directly shoulder the antidiscrimination mandate’s costs. To that extent, the debate suggests tight political limits on the ability of legislators to deploy a cost-shifting strategy.

Multivariate Analysis of Cost Shifting and Fiscal Stress

Before our regression analysis, we offer a time-varying visualization of the relationship between enactments and fiscal stress, using one of our principal independent variables—the three-year moving average of change in per capita revenue. Figure 3 shows when each state passed antidiscrimination and accommodation enactments; the vertical axis displays the three-year average percentage change in revenue. The dashed line, meanwhile, plots the three-year moving average percentage revenue change across all states: states below the line passed legislation in times of decreased revenue (compared to other states). The cost-shift hypothesis predicts that most enacting states will be below the line. As in the simple bivariate confidence-interval plots above, the results are mixed, and we do not draw firm conclusions.

Figure 3.

Enactments and revenue change over time.

We turn, then, to regression analysis. Our ability to leverage both the timing and the strength of state-level disability discrimination laws is a virtue of our empirical study, but it also poses significant modeling challenges. An initial modeling puzzle concerns the proper dependent variable and how best to account for the fact that states enacted two different types of disability discrimination provisions of different strengths. We addressed this puzzle in two ways. First, we estimated the effect of fiscal stress separately for simple and accommodation enactments. In the set of models presented here, we coded state years as one when an antidiscrimination enactment occurred, zero beforehand, and missing afterwards; in a second set of models, we took the same approach for accommodation enactments but included a control variable that is one in states with previous antidiscrimination enactments and zero otherwise. For each dependent variable, we estimated linear probability models of the form described in Equation 1, provided below.

Second, as a way of retaining as much of the data as possible for our models, we combined the two types of provisions in a simple index. We coded years in which no change occurred with a zero; years in which a state passed a simple enactment or shifted from a simple to an accommodation enactment with a one; and years in which states enacted an accommodation law where no simple enactment was present with a two. In other words, we understood the passage of a simple enactment and the shift from a simple to an accommodation enactment as similar one-unit changes in the dependent variable. In both approaches, once a state has enacted an accommodation law, it drops out of the dataset. This reflects the data structure for a discrete event history model—a hazard model adapted to a context in which time is measured in discrete units (here, years).Footnote ⁶³ Discrete event history models often employ logistic regression, but we used linear ordinary least squares regression instead. We made this choice in order to avoid the problems associated with including multiple fixed effects in logit regressions.Footnote ⁶⁴ We estimated panel regressions of the form:

(Equation 1) $${Y_{it}} = {\beta _0} + {\beta _1}{X_{it}} + {\beta _2}{J_{it}} + {e_{it}}$$

where Y _it is a three-level outcome variable for state i in year t, J is the independent variable of interest (revenue, unemployment, welfare, and welfare/revenue, respectively), X is a vector of control variables (Democratic control, political competition, contagion, and veterans per capita), and ┌ represents fixed effects for states and years. We omit state years after accommodation enactments since those states are no longer at risk of enacting or strengthening a law. An advantage of this approach is that it estimates the effect of changes in our independent variables on changes in the probability of enactment; the model produces an average of individual difference-in-difference effects (Goodman-Bacon Reference Goodman-Bacon2019).

We readily acknowledge the central weakness of this approach: none of our independent variables is exogenously determined. For example, if welfare expenditure is associated with enactments (as our models suggest is the case), this could be the result of a third variable—perhaps, to name just one possibility, growing social concern for vulnerable populations. In the absence of a plausible quasi-experimental research design, we presented these admittedly muddy results as the best available approach. Given the absence of any previous empirical research on the cost-shift hypothesis or causes of disability enactments, we believe that these results have significant descriptive value. Table 4 presents simple descriptive statistics on our independent and dependent variables. Tables 5 and 6 present the results of models that examine the relationship between our four principal independent variables and the enactment of simple and accommodation disability antidiscrimination laws, respectively.

Table 4.

Descriptive statistics

Table 5.

Antidiscrimination enactments

Notes: * p < 0.05; ** p < 0.01; *** p < 0.001.

Standard errors are in parentheses.

Models include state and year fixed effects, and standard errors are clustered by state.

Models 1 and 4 have fewer observations given limited revenue data.

Table 6.

Accommodation enactments

Notes: * p < 0.05; ** p < 0.01; *** p < 0.001.

Standard errors in parentheses.

Models include state and year fixed effects, and standard errors are clustered by state.

Models 1 and 4 have fewer observations given limited revenue data.

For simple enactments, welfare expenditure (both alone and as a proportion of revenue) is the only significant predictor of enactments.Footnote ⁶⁵ More precisely, Model 3 in Table 5 suggests that a one thousand dollar increase in per capita welfare expenditure would increase the probability of an enactment by over three hundred percentage points. That is obviously not a realistic prediction, but in the context of a median per capita welfare expenditure of twenty-seven dollars, this implies that moving from the twenty-fifth to the seventy-fifth percentile in welfare expenditure increases the probability of an enactment by six percentage points. This association supports the cost-shift hypothesis, but it must be balanced against the lack of any association of enactments with sinking revenue or rising unemployment.Footnote ⁶⁶ One of the control variables also stands out: the number of veterans per capita is associated with enactments in all specifications.Footnote ⁶⁷ This may offer some support for a political demand theory of these enactments. Given the unanimity of many of the votes on enactments, this association could suggest that legislators from both parties perceived these enactments as politically beneficial.

Table 6 presents similar results for accommodation enactments, including a control variable for the presence of previous antidiscrimination enactments. There is only one significant relationship in this table, and it is in the opposite range of the predicted direction: higher state tax revenue predicts enactments. This result runs directly counter to the cost-shift hypothesis, but it is subject to the same caveats that we raised with respect to the more intuitive welfare expenditure result regarding the endogeneity of independent variables. Table 7, finally, presents the results of models that examine the relationship between our four principal independent variables and the enactment or strengthening of disability antidiscrimination enactments. In this specification, none of the independent variables is a significant predictor of enactments. One way to understand this lack of results is that the results in Tables 5 and 6 largely wash one another out when combined. The lack of any statistically significant associations also suggests that any large associations are unlikely. For example, a one standard deviation decrease in revenue would increase the probability of an enactment by only about 2 percent, even if one were to assume that the estimate here is two standard errors smaller than the actual effect.

Table 7.

Index of enactments

Notes: * p < 0.05; ** p < 0.01; *** p < 0.001.

Standard errors in parentheses.

Models include state and year fixed effects, and standard errors are clustered by state.

Models 1 and 4 have fewer observations given limited revenue data.

Cost Shifting and Public versus Private Enforcement

So far, our models have only evaluated the hypothesis that states sought to shift costs from social services to employers by enacting disability discrimination laws. But, as explained above, a second version of the hypothesis is equally plausible: that states facing budgetary stress also sought to shift costs off-budget by incentivizing more private enforcement of antidiscrimination mandates. We tested that hypothesis with novel data on state litigation incentive enactments (see Table 2). We constructed a simple index: in a year in which a state enacts any single litigation incentive (attorneys’ fees, compensatory damages, or punitive damages), we coded the state as a one in that year; if it enacts no incentives, we coded it as a zero. If the state enacts two incentives in the same year, we coded the state as a two in that year, and if it enacts three incentives in the same year, we coded it as a three in that year. We only considered years beginning with the first year in which a state enacted either a simple or accommodation enactment. We omitted state years after states had enacted all three incentives. We then estimated a regression of the same form as Equation 1, but now with this incentive index as the dependent variable. We also added a control variable for whether the state has a simple or accommodation enactment on the books.

Table 8 shows the results, again including a column for each independent variable as well as the political and demographic control variables discussed above.Footnote ⁶⁸ The results in Table 8 largely match our main results predicting the probability of simple and accommodation enactments. Some of our results are consistent with the cost-shift hypothesis, but they provide only qualified support for it. The only statistically significant association is between revenue per capita and litigation incentives: our model suggests that a one thousand dollars per capita decrease in revenue leads to a 2.6 percentage point higher chance of passing a litigation incentive. Since that chance is lower in any given year—about 6.5 percent across all eligible state years—that 2.6 percentage point change is about a 40 percent increase.Footnote ⁶⁹ None of the other associations between fiscal stress and the passage of litigation incentives is distinguishable from zero, however.

Table 8.

Litigation incentives

Notes: * p < 0.05; ** p < 0.01; *** p < 0.001.

Standard errors in parentheses.

Models include state and year fixed effects, and standard errors are clustered by state.

In sum, the models offer only highly qualified evidence for the cost-shift hypothesis. Although we do find a statistically significant association between welfare expenditures and enactments, this association does not hold for accommodation enactments, and, in fact, we find that decreasing revenue has the opposite effect in the context of accommodation enactments. Finally, in our models examining the association between fiscal stress and the passage of litigation incentives, we find that decreases in revenue are associated with the passage of litigation incentives—but that our other measures of fiscal stress are not. Together with our qualitative analysis of the discussions that occurred around the passage of the relevant bills, these results suggest that cost shifting did not play a large role in spurring the adoption of disability discrimination statutes.