A common refrain among opponents of social-welfare programs is that theyâ€™re too costly and act as a drag on future growth because someday taxpayers will have to pick up the tab. Thatâ€™s a crude and short-sighted view: Itâ€™s long been understood that some social programs not only pay for themselves but yield positive returns; they also play an important role in reducing inequality, which is crucial for boosting the economy.
Yet some programs are better than others, and itâ€™s often difficult to know which we should invest in. My Harvard colleague Nathan Hendren and Harvard graduate student Ben Sprung-Keyser help answer that question in a new paper that looks at 133 different government programs, including social insurance, education and in-kind transfers such as food stamps.
It can be hard to compare the outcomes of different types of programs directly. Education subsidies and food stamps, for example, kind of look like apples and oranges. How would you know whether increasing education or nutrition is a better use of government funding? To make the comparison apples-to-apples, Hendren and Sprung-Keyser introduce a new measure called the marginal value of public funds. The idea is to add up beneficiariesâ€™ willingness to pay for a program, and divide that by the net cost to the government. This measures a programâ€™s bang for the buck for its beneficiaries: a value much larger than 1 means that the policy is creating a lot of value for each dollar of spending; a value of exactly 1 means the policy is on par with a dollar-for-dollar cash transfer to recipients; close to or below 0 means weâ€™re just burning taxpayer money.
You can compare bang for the buck across very different types of programs, making it easy to see which offers the better return. Hendren and Sprung-Keyserâ€™s measure shows, for example, that investing in childrenâ€™s education typically pays off much more than in-kind aid to families. For comparing policies that affect different parts of the population, the measure doesnâ€™t automatically tell us which programs we want, but it does help us get a sense of the trade-offs.
Hendren and Sprung-Keyser apply their measure to 133 programs for which we have solid estimates of the benefits to recipients and costs to the government.
Programs that help children tend to come off looking the best: if you improve a childâ€™s health and/or education, thereâ€™s a future payoff in terms of reduced need for social services later in life, as well as higher-paying employment and increased tax revenue. Expanding Medicaid to cover pregnant women, infants, and young children paid for itself in full, as do many programs that improve childrenâ€™s access to college. Programs targeting adults that have spillovers to children - such as a program that helped people move to lower-poverty neighborhoods - can pay for themselves, as well.
A number of adult healthcare expansions clock in next; although these policies donâ€™t pay for themselves, they can create much more value for recipients than they cost on the margin.
Food stamps and the earned income tax credit tend to be roughly comparable with each other and to direct aid; they both have marginal values close to 1 on average.
Lowest under the measure are programs that donâ€™t affect outcomes much, such as tuition-tax deductions that donâ€™t actually increase college enrollment. Perhaps surprisingly, job-training programs tend to register as having a low return, as well. Unemployment and disability insurance also appear to have a relatively low return under Hendren and Sprung-Keyserâ€™s measure, but this is hardly surprising - many people receiving these forms of social insurance wonâ€™t return to the labor force, making the costs of those programs less likely to be offset by future tax revenue.
Of course, a low marginal value of public funds doesnâ€™t automatically mean we shouldnâ€™t implement a policy. But the measure tells us about the trade-offs. For the same beneficiary population, a program with a higher payoff is better.
At a minimum, though, there is no excuse not to expand policies that pay for themselves - as Hendren and Sprung-Keyser point out, the marginal returns on those policies are literally infinite. Why wouldnâ€™t we, for example, want to improve childrenâ€™s health and education if the government will make money from doing so in the end?