The Washington PostDemocracy Dies in Darkness

Heritage says immigration reform will cost $5.3 trillion. Here’s why that’s wrong.

Jim DeMint, the current president of the Heritage Foundation. (Win McNamee/Getty Images)

Robert Rector and Jason Richwine of the Heritage Foundation have made a splash by releasing a paper claiming that the immigration reform bill being weighed in the U.S. Senate will cost the government $5.3 trillion. Or, more precisely, that undocumented immigrants under current law will cost the government $1 trillion, and legalizing those immigrants will increase that to $6.3 trillion. Subtract one from the other and you get the $5.3 trillion total cost estimate.

The study represents the most notable attack on the reform effort to date from a conservative group, and in conjunction with the Economic Policy Institute's attempts to throw cold water over the high-skilled immigration sections of the bill, suggests the effort is facing flack from both sides.

So does the Heritage estimate hold up? Not really. They make a lot of curious methodological choices that cumulatively throw the study into question. It's likely that immigrants would pay a lot more in taxes, and need a lot less in benefits, than Heritage assumes, and that other benefits would outweigh what costs remain.

What is a benefit?

First, let's review the basic methodology of the Rector/Richwine study. When they're talking about "cost," they really mean static fiscal cost. They add up the amount of money they expect undocumented immigrants to contribute in taxes, and subtract that from the amount of money they expect those immigrants to receive in government services. If the result is positive, then the immigrants are a net fiscal drain. If it's negative, they're net contributors.

Surprise: It's positive. Undocumented immigrants tend to earn less in their careers than the median American (though far more than they would have earned had they not migrated). That means they pass less in taxes, because we have a tax system that's moderately progressive. We also have a social services system that's designed to provide services to the less economically fortunate, so undocumented immigrants, when eligible, get more access to those services than the median American. So it's not exactly shocking that Rector and Richwine found that the average immigrant gets more than what he or she puts in.

But let's pause a second and look at what Rector and Richwine count as "benefits," and what they exclude. They break benefits into five broad categories: direct benefits (e.g. Social Security, Medicare), means-tested benefits (e.g. Earned Income Tax Credit, Medicaid), public education, population-based expenditures (parks, police and fire departments, etc.), and interest payments on past benefits. They exclude the latter from their calculations, as it's not actually a benefit currently being enjoyed by residents.

From this, they subtract "pure public goods," or government programs that are (a) nonrival (any number of people can use them without diminishing others' enjoyment) and (b) can be extended to new users at zero cost. Their main example is defense spending, the logic being that the benefit of having a military to defend a country does not diminish when more people move there, but they also include medical research and veterans' benefits.

The latter choice is utterly baffling. Veteran's health care and benefits is rival (people can't enjoy it simultaneously), and it cannot be extended to new users at zero cost. By Rector and Richwine's own standards, it is absolutely not a pure public good.

Correspondingly, the choice to exclude classic public goods like parks, police and fire departments, roads, and so forth from that category is also questionable. The logic presented by Rector and Richwine is that those services expand with population. New York has more police officers than Hanover, N.H., say. But the cost of many such goods don't expand linearly with population. Ten thousand people moving to New York may make Grand Central Park more expensive to maintain, but one person moving there doesn't increase the cost much at all. Depending on where immigrants settle, the inclusion of these goods could artificially inflate the cost of immigration by Rector and Richwine's methodology.

Also notable is what benefits are excluded from Rector and Richwine's calculus. They don't count tax expenditures, which economists of all stripes usually count as spending, as constituting a benefit. Or, rather, they do, but only when they benefit poor people (as in the EITC). The mortgage interest tax deduction, the charitable deduction, the employer health-care tax exclusion, the preferential treatment of capital and dividend income — these are all massive benefits. The Joint Committee on Taxation estimates the benefit of the employer health-care exclusion at $760.4 billion over five years, and preferential capital treatment at $616.2 billion over five years.

The vast majority of these tax benefits accrue disproportionately to wealthy Americans. That makes the overall tax and transfer system less progressive than it may initially appear, according to Rector and Richwine's methodology. They're excluding the single biggest category of benefits for the rich.

They also exclude regulatory restrictions that cost the public billions but benefit a certain segment of the population. For example, intellectual property laws generate billions in revenue for content-creating industries at the expense of consumers. The distributive impact of those regulatory frameworks is almost certainly regressive. But it doesn't count as a benefit to rich Americans according to Rector and Richwine.

The rest of the study is only as good as the estimates of benefits-received and taxes-paid that Rector and Richwine produce initially. And the benefit estimates are woefully incomplete.

What about taxes?

The tax section is slightly less problematic. Rector and Richwine simply count all tax revenues at all levels of government, divided by the payers' education levels, and create an average payment from undocumented immigrants based on that. But they also add some adjustments. Some — such as assuming that undocumented immigrants use goods like parks less, and that their children use fewer medical services — are designed to give their opponents the benefit of the doubt. But others badly skew the results in their favor.

The most egregious case of that is what they do with FICA taxes for Social Security and Medicare. "Since 45 percent of unlawful immigrants are believed to work 'off the books,' the federal and state income tax and FICA tax payments that Census imputes for each household were reduced by 45 percent among unlawful immigrant households," they write. But why? Two thirds of undocumented immigrants were contributing to Social Security as of 2007. A 33 percent discount would be more justified for FICA taxes. Nowhere in the paper is the 45 percent assumption justified.


To compute the average "fiscal deficit" of the typical undocumented immigrant, the study weighs spending on each identified government benefit program by the estimated share of the program enjoyed by undocumented immigrants. For Social Security, the share is practically 0; for aid to Women, Infants, and Children (WIC), it's over 10 percent. This is fine so far as it goes but also runs into the problem around public goods identified previously. The study suggests that each immigrant household added $476 a year in costs to the highway system, and $313 a year to parks and recreation costs. That doesn't really pass the smell test, I don't think.

Once they have the average cost estimate in all its flawed glory, they subtract the amount a typical undocumented worker pays in taxes to get the total "fiscal deficit" for undocumented immigrants. That figure is about $14,387 a year, per household, according to Rector and Richwine.

What reform does

Initially, Rector and Richwine estimate that, by requiring the payment of back taxes and putting workers back "on the books", the immigration reform bill will reduce the fiscal deficit for now-undocumented (then, presumably, documented) workers to $11,455. But then, by making these immigrants eligible for means-tested benefits like Obamacare insurance premium subsidies and EITC, and universal benefits like Social Security, that will increase to $28,000, or $22,700 a year for retired workers. The per-household cost is increased by over 50 percent relative to pre-reform.

That sounds really bad! But, unsurprisingly, it's much more complicated than that. For one thing, Rector and Richwine assume no income growth for affected immigrants. But that's beyond implausible. Plenty of studies has found that legalization increases immigrant incomes. One found that they grow by 15.1 percent due to legalization. That raises tax receipts, reduces benefit payouts, and generally makes the fiscal picture less bleak than the one painted by Rector and Richwine.

They also assume no other economic impact of any kind. That's so implausible that even the CBO, which is famously conservative with regards to incorporating economic effects of policies, took direct economic effects of adding people to the workforce into account when evaluating the 2006-7 reform bill. They found that legalization, even paired with increased border security spending, would mildly cut the deficit over 10 years, by about $12 billion. If you take a broader, "dynamic scoring" approach, they estimated that the bill would increase GDP by between 0.3 percent and 0.4 percent from 2007 through 2011 and by 0.8 percent to 1.3 percent from 2012 through 2016. That's a big deal, one that could reach tens if not hundreds of billions more in new revenue and saved social spending.

This isn't a really controversial point. Even critics of immigration reform like Harvard's George Borjas admit that it increases the economy at large, with over $30 billion in gains going to non-immigrants. To quote, er, the Heritage Foundation, "While the presence of low-skill migrant workers can be construed as a challenge to low-skill native workers, the economic effects are the same as the effects of free trade-a net positive and a leading cause of economic growth."

The most bizarre portion of the Rector and Richwine study is how they construct their counterfactual for what would happen absent reform. They seem to assume that a huge number of immigrants would just…leave. On their own. No, for real, this is what the paper says. "If one assumes that under current law, most unlawful immigrants will return to their country of origin around age 55, the lifetime fiscal costs of unlawful immigrants under current law are comparatively low: only around $1 trillion," they write. Why would someone assume that?

The authors justify this statement through reference to the fact that there are very few immigrants over the age of 55. They theorize that "unlawful immigrants, being unable to access the U.S. welfare and retirement systems under current law, simply go back to their country of origin as they get older."

The alternative theory, they note, is that "unlawful immigrants, arriving as young adults over the past 15 to 20 years, have simply not yet reached age 50." That seems by far the likelier theory, especially given that many older undocumented immigrants were legalized by the 1986 amnesty law, and that the number of elderly immigrants has been growing.

By the author's own admission, by far the biggest cost source for undocumented immigrants is public education. That isn't affected by legalization.


Rector and Richwine are certainly correct that making currently ineligible immigrants eligible for means-tested benefits and retirement entitlements has a real budgetary cost. But in the long run, we know that immigration is a net economic boon, and in particular for immigrants, which reduces their fiscal cost and increases our ability to pay for what benefits they do receive. And the best study we have on the fiscal effects of immigration reform, from the CBO, finds the impact to be minimal or positive.

Pay attention to that study. Pay attention to whatever score the CBO puts out of the Gang of 8 bill. But the Heritage numbers simply are not credible.

Correction: a photo that I thought was the Grand Canyon but was not was briefly in this post. A picture from a real national park has replaced it. We regret the error.