Disability insurance and the dynamics of the incentive-insurance tradeoff

Recent growth in the number of Disability Insurance claimants has led to calls for substantial scaling back of the program. We evaluate the incentive cost of the DI program against its insurance value to those in need. The main failure of the program is the number of severely work limited who do not receive insurance: the program is badly targeted.

Between 1985 and 2014, the proportion of Disability Insurance (DI) claimants in the United States more than doubled from about 1.8 percent to 4.8 percent of the working-age population. The cash benefits paid by the DI program were more than three times larger than those paid by Unemployment Insurance ($128 billion vs. $40 billion in 2012).

What are the welfare consequences of this large and growing program?  In our view, the answer hinges on three key questions.  First, are program claimants people who are genuinely unable to work?  Second, are those in need receiving insurance?  And third, how does the value of the insurance provided compare to the inefficiencies and incentive costs of the program? This article summarizes our recent research addressing these questions (Low and Pistaferri, 2015).

We evaluate how welfare is impacted by reforms to aspects of the DI program that determine the trade-off between insurance benefits and incentive costs. 

We evaluate how welfare is impacted by reforms to aspects of the DI program that determine the trade-off between insurance benefits and incentive costs.  Figure 1 captures the trade-off between “coverage”, which is the fraction of people with a severe work limitation who receive DI, and “false claimants”, which is the fraction of people on DI despite not having a work limitation.  Coverage (red line) is very low: only 60% of those with severe work limitations receive DI, and the fraction is much lower at younger ages. False claimants (solid blue line) are substantial: about 30% of successful claimants are false claimants, and the fraction is much higher at younger ages.

Figure 1: Insurance Coverage and False Applications


Any reform of DI should aim to improve coverage and reduce false claimants. But this is not straightforward because letting more people onto the programme is likely to mean more incentives to make false claims. The increase in false claims not only has a direct cost, but also reduces labour supply and leads to detachment from the labour force.

Our Approach

We believe it is important to model and estimate both the incentive costs and the coverage and insurance value of DI in the same framework.  Doing so insures that the analyses are consistent. We use a life-cycle framework so that decisions at one point in time have implications for the rest of the individual’s life. Such a structural model should also be able to replicate reduced form results obtained in the literature. Our model captures three key aspects of the problem.

First, we assume that individuals face several sources of risk over their lifetime: persistent productivity “shocks” unrelated to health (such as a decline in the price or marketability of skills), a disability or work limitation shock that reduces the ability to work, and labor market frictions (such as the probability of losing one’s job and/or inability to find a new one). Disability insurance is intended for those suffering severe and persistent shocks and so we allow for variation in the severity and persistence of these shocks over the lifetime.

Second, we assume that health affects individual welfare. People value consumption differently depending on their current health status: people may have greater consumption needs when they are disabled (such as for alternative transportation, domestic aid, etc.).

Finally, we model the decisions to apply for and award DI benefits. Those not truly disabled apply because they are stuck in permanently low-paying jobs; because their skills have been made obsolete by, say, new technologies; or because they are unable to find jobs. The government’s decision to award DI is made with imperfect information about the applicants’ genuine health status. Hence, we assume that the DI program screens applicants with errors. Moreover, we assume that awardees are re-assessed at regular intervals; these re-assessments, known as Continuing Disability Reviews, are performed periodically to check whether awardees are still disabled.

We have three key findings.  These come from estimating our model using self-reported disability status, employment,  wage and consumption data from the Panel Study of Income Dynamics (PSID) between 1986 and 2009.

  1. People with low skills are more likely to develop a disability and less likely to recover. This is because some low-skilled individuals work in occupations or industries where the risk of a disability is greater or work in ways that are less adaptable to health shocks. Low-skilled workers also tend to engage in activities that worsen overall health (smoking, bad eating habits, etc.) and invest less in “health capital” (such as preventive health activities or physical exercise).
  1. The effect on productivity of a disability shock is substantial: productivity declines between 6 and 18 percent depending on the severity of the health shock. Wages for those in poor health are even lower than this, but part of this difference is due to individual fixed effects: low productivity workers are also less healthy. On the consumption side, there are greater consumption needs when people become disabled.
  1. We find substantial rates of false rejection in the screening process, and smaller though important rates of false acceptance. Half of older workers and three-quarters of younger workers who are truly disabled are turned down on initial application, although over a 5-year horizon, many of these reapply successfully. False acceptances are particularly prevalent among those with at least some moderate disability. The cost of these false acceptances is high because individuals rarely leave the DI program.

The welfare implications of reforming the DI program

Given the rapid growth of the DI program, there have been many discussions regarding its reform. We use our model to analyze the impact of varying key policy features on welfare and on behavior.  These features are (a) the generosity of disability payments, (b) the stringency of the screening process, (c) the generosity of alternative social insurance programs, and (d) the re-assessment rate (Continuing Disability Reviews).

Given the rapid growth of the DI program, there have been many discussions regarding its reform. We use our model to analyze the impact of varying key policy features on welfare and on behavior. 

The effect of policy changes on social welfare depends on two parameters: the change in insurance coverage for the truly disabled and the change in the incentive costs of the program, captured mainly by changes in the number of non-disabled workers who apply for DI.[1]  The number of DI applications by moderately disabled people who have alternative options is sensitive to policy changes (a measure of the “moral hazard” problem). In contrast,  the severely disabled are insensitive to policy changes: they apply out of need, not in response to changes in economic incentives.

In the first experiment, we consider the effects of decreasing generosity. As the payout decreases, there is a fall in “false” applications from those with only moderate disabilities but also a fall in applications from those with true disabilities; however, the number of “false” applicants declines much faster as generosity falls, than the decline in the number of truly disabled applicants because false applicants return to work if generosity is low. On the other hand, as the payout decreases, those who are severely disabled and on DI are less well insured.  Overall, social welfare falls as the payout decreases because of the worse insurance for the severely disabled.

In the second experiment, we change the strictness of the screening process for DI. This is a “reform” that was tried in 1980 and that led to sharp declines in the number of DI awards and significant removal of DI recipients. It was repealed in 1984, resulting in more liberal admission criteria that allowed applicants with mental and musculoskeletal disabilities. Increasing the strictness of the screening process reduces the number of false applicants granted insurance, but increases the likelihood that more severely disabled workers will also be rejected due to imperfect screening. The result is a net welfare loss, because of  the large false rejection rate, especially among younger severely-disabled workers.  Younger severely-disabled individuals are particularly ill-equipped to insure against disability risk: even if they do not get DI, they do not return to work and they have not had time to accumulate enough assets to self-insure. Thus reducing strictness to admit more of the truly needy outweighs the value of rejecting those who are undeserving.

The third policy we study is a change in the generosity of the food stamp program, rather than a change to the DI program itself. In our model, food stamps are a stand-in for all means-tested welfare and social insurance programs not contingent on health status. DI interacts in important ways with these welfare programs. We show that an increase in the generosity of means-tested programs reduces DI application rates by non-disabled workers and increases insurance coverage among disabled workers. This positive combination arises for two reasons.   First, marginal undeserving applicants treat the means-tested program as a substitute for DI to avoid the uncertainty and hassle of a DI application. These are individuals whose skills have been made obsolete by technological shocks and who at the same time have some moderate to low disability; given screening inefficiencies, they have a chance at entering the DI program. Second, truly disabled workers use the means-tested program as a complement to DI to finance their consumption while waiting for a decision and also in case of DI rejection.

Finally, we change the rate at which people are reassessed following entry into the program. An increase in the reassessment rate discourages false applications from those who are not severely disabled. Specifically, an increase in the average frequency with which people on DI get reassessed (from 2 percent per quarter to 8 percent per quarter), lowers the proportion of false applications from 54 to 30 percent. The threat of reassessment has a marked discouragement effect, and leads to a decline in the number of undeserving candidates who receive insurance. However, more frequent reassessment also results in reduced coverage for the severely disabled, as some also end up being removed from DI in error. Overall, increasing the reassessment rate increases welfare, albeit modestly.


We study the welfare implications of reforming Social Security Disability Insurance using a rich model of individual behavior. We document extensive disability risk as well as large screening errors, which vary by the severity of applicants’ disabilities and by age.

Our counterfactual reforms show that it is welfare improving to reduce the generosity of the program in terms of wage replacement rates, but at the same time to reduce the strictness of the screening process.

Our counterfactual reforms show that it is welfare improving to reduce the generosity of the program in terms of wage replacement rates, but at the same time to reduce the strictness of the screening process. Increasing generosity of means-tested programs moves individuals off DI, without the same long-term disincentives to work.  DI is currently serving a need for long-term unemployment insurance which is unavailable in the US. It is doing so particularly for low-wage individuals displaced by skill-biased technological changes.

We see two limitations to our analysis. First, we do not account for the fact that the probability of receiving a negative health shock is partly under an individual’s control through choices of occupation and lifestyle. These decisions will be affected by the disability insurance program. Second, we have ignored the health insurance component of the program: DI recipients receive Medicare, and this means we underestimate the insurance value provided by the program. One interesting, and less debated, aspect of Obamacare is that if people stay on DI primarily because of the value they attach to the health insurance component provided, then this “program lock” effect should disappear once the health care reform runs its full course.

[1] We measure welfare as “consumption willingness to pay”: how much consumption individuals are willing to pay (or need to be paid) to be in the new policy environment.

Further reading

Hamish Low and Luigi Pistaferri (2015), “Disability Insurance and the Dynamics of the Incentive-Insurance Tradeoff”, American Economic Review, 105(10): 2986-3029

Jeffrey Liebman (2015) “Understanding the increase in Disability Insurance Benefit Receipt in the United States” Journal of Economic Perspectives 29(2): 123-150