Thursday, April 3, 2008

Immigration and Social Security financing

In an editorial entitled “How Immigrants Saved Social Security,” New York Times discusses how changes to modeling of immigration improved the long-term financing projection for Social Security in this year’s Trustees Report. the

Immigration is good for the financial health of Social Security because more workers mean more tax revenue. Illegal immigration, it turns out, is even better than legal immigration. In the fine print of the 2008 annual report on Social Security, released last week, the program’s trustees noted that growing numbers of “other than legal” workers are expected to bolster the program over the coming decades.

One reason is that many undocumented workers pay taxes during their work lives but don’t collect benefits later. Another is that undocumented workers are entering the United States at ever younger ages and are expected to have more children while they’re here than if they arrived at later ages. The result is a substantial increase in the number of working-age people paying taxes, but a relatively smaller increase in the number of retirees who receive benefits — a double boon to Social Security’s bottom line.

For what it’s worth, here’s the mental framework I use to think about immigration and Social Security financing. Roughly, I divide the Social Security program into two parallel systems – one for native born workers and their descendants, and a second for immigrants and their descendants. (Since immigration often marry native-born and have children, this isn’t flawless, but it will work for these purposes.) For immigration to help Social Security financing as a whole, the immigrants’ system has to be more than solvent over the long-term; it has to throw off additional money that can then be used to subsidize the native-born system and thus improve solvency overall.

What are the factors that determine whether this will happen? Among them:

  • Earnings: Since immigrants tend to have lower earnings and shorter work lives (which makes them appear even poorer to the progressive Social Security benefit formula), those who collect retirement benefits will tend to get a good deal. To the degree they receive an internal rate of return (IRR) exceeding the trust fund’s interest rate, these immigrants will receive more in benefits than they pay in taxes, and thus be a drain on the program.
  • Likelihood of collecting retirement benefits: As noted in the Trustees Report, many immigrants do not end up collecting retirement benefits. They may be illegal immigrants and thus ineligible to collect; likewise, they might return to their home country and not collect benefits, or fail to work the 10 years necessary to qualify for retirement benefits. These workers are a net gain to the program, since they pay taxes but don’t collect benefits.
  • Earnings levels of immigrants who don’t collect benefits: Social Security’s financing benefits from high income immigrants who don’t collect benefits, since these individuals pay more taxes. There is good reason to believe, however, that illegal immigrants and immigrants who return to their home countries will have below-average earnings levels. In part this is because illegal immigrants’ wages tend to below, and in part because those who succeed economically in the U.S. are more likely to remain here. The earnings levels of immigrants of any kind are not, to my knowledge, modeled by SSA’s actuaries; rather, they are treated as having the same average earnings as native-born. This is a weakness in the current model that SSA should aim to address.
  • Fertility of immigrants, legal and illegal: In general, fertility rates for immigrants are higher than those for native-born; this will tend to help long-run system financing. At the same time, fertility for immigrants varies more by education level than it does for native-born. Thus, lower educated (and presumably lower earning) immigrants may have disproportionate numbers of children. To the degree that these children have lower than average earnings, and thus higher than average benefits relative to earnings, the net gain to system financing may be reduced.
In sum, modeling the effects of immigration on long-run Social Security financing is a very complex issue and work in this area should not be considered finished. My instinct is that to do examine these issues in detail will involve a shift from the semi-aggregated “cell based” models currently used by the SSA actuaries to microsimulation models, such as CBO’s Long-Term model (CBOLT), the GEMINI modelPolisim model from the Policy Simulation Group, or the under joint development by SSA’s Office of Policy and Office of the Chief Actuary.

1 comment:

Bruce Webb said...

The NYT mischaracterized the issue, probably because reporters rarely if even get beyond the 15 page Overview (if indeed they get past the press release) to actually grapple with the data.

The actual analysis suggests that the overwhelming effect is from bullet point two.

"The most significant of these changes is a major revision in the methods used for projecting the other-immigrant (other than legal permanent resident) population. In previous reports, the other-immigrant population was projected using assumed annual numbers of net other immigrants with a static age-sex distribution. For this year’s report, the annual numbers of net other immigrants are projected by explicitly modeling other immigrants and other emigrants separately. Under this approach, a large number of other immigrants is assumed to enter the Social Security area at relatively young working ages, with the total annual number of other immigrants entering the area assumed to be about 1.5 million. Most of these immigrants are assumed to either: (1) leave the Social Security area (i.e., to depart from the area without having attained the legal status or work credits needed to become eligible for retired-worker benefits); or (2) attain legal permanent resident (LPR) status after several years of being in the other-immigrant population. Thus, this year’s report results in a much larger other-immigrant population projected at working ages and a smaller number remaining in the Social Security area into old age. This change, along with the additional births due to the larger other-immigrant population at younger ages, results in a substantial increase in the number of working-age individuals contributing payroll taxes, but a relatively smaller increase in the number of retirement-age individuals receiving bene fits in the latter half of the long-range period. This revision results in an increase (improvement) in the long-range actuarial balance of about 0.30 percent of taxable payroll."

Which is to say that the difference is not really do to absolute numbers of illegal immigrants but instead how they contribute to actual covered worker ratios in practice. This is hardly an argument for simply opening the borders, you could get the same effect by instituting an effective guest worker program with the explicit understanding that this would be limited to 39 quarters maximum and so converting FICA for these workers from an insurance premium to a straightforward employment tax.

Instead the whole thing gets dumbed down into the soundbite 'illegal workers benefit Social Security' where the real finding was 'actual impact of existing illegal workers on Social Security solvency refined and corrected under new more rigorous model'. Not sexy enough for the op/ed pages I guess.