Labor Mobility and Wages of the Rich Country Poor, Part Two: Instruments to Targets

June 29, 2017

In the last month three working papers were posted in the prestigious National Bureau of Economic Research (NBER) series on one very narrow question. The question is whether the massive and sudden migration flow of Cubans (around 90,000 arrived in six months April to October 1980) into Miami, USA from the Mariel Boatlift had an impact on the wages and/or employment of natives. The Mariel Boatlift is extensively studied because it is a nearly perfect natural experiment. In the classic study on the topic David Card (1990) compared Miami to other US cities and found little or no impact on the wages or employment of non-migrant workers, even those with a high school degree or less.

In 2015 George Borjas reanalyzed the Mariel experience and narrowed the question to whether the large scale arrivals of Cubans affected the wages or employment not of the “low skill” workers but of the “super low skill”—those without a high school degree. Peri and Yasenov (2017) estimate that there were 48,714 Mariel arrivals without a high school (HS) degree and in Miami in 1980 only 169,440 people without a HS degree in the labor force. For the “super low skill” segment the labor force was massive (almost 30 percent) and sudden. Borjas (2015) claims that for the Borjas sub-group (BSG)—non-Hispanic male natives, aged 25-59, with less than a HS degree—there was a large negative impact on wages. In a previous blog I show that two of the recent NBER papers argue the BSG finding is an artefact. Clemens and Hunt (2017) argue a sampling shift towards blacks explains the apparent wage impact. Peri and Yasenov (2017) show that the BSG is special and that nearly all other demographic sub-groups except the super low skilled (e.g. including women, including Hispanics, including younger and older workers) show no impact on wages.

Here I argue that, even if there were a robust and credible negative impact on the wages of the BSG from low skill migrant arrivals (which there isn’t), this would not justify limiting immigration as there are better instruments to achieve the same objectives, with much less cost.

First, economists’ have a standard response to distributional concerns about Pareto-improving policies: “instruments to targets.” Most market oriented economists think about how to maximize the size of the pie by making all factors as productive as possible and about how to ensure the pie is best distributed. But they are trained to think of these questions separately. The reason is that it is nearly impossible that the same policy instrument both maximizes productive efficiency (reached a Pareto Optimal outcome) and achieves distributional goals efficiently.

Trade economists, who have debated protectionists for centuries, understand “instruments to targets.” Protectionists point to the specific job losses that would result from lowering protection. Economists calculate the total losses to the economy from the higher prices to buyers (both consumers and producers of the product as an input). Trade economists calculate the “cost per job saved.” Sugar quotas imposed by the United States raise the prices of domestic sugar and in 2010 the estimated cost of “saving” 2,260 jobs in sugar production was $1.9 billion or $826,000 per job. In 2009 the US imposed tariffs on imports of Chinese tires and “saved” 1200 jobs at a cost of $1.1 billion annually for a cost of $900,000 per job saved. A 1986 review of protectionism in 31 industries in the USA estimated an average cost per job saved of $516,208.

Of course, international trade in goods is only one source of job reallocations in a market economy. The arrival of a Walmart in a locality almost always means that mom and pop single store retail outlets lose business and hence jobs. One could, in principle protect those jobs by keeping Walmart out of a region—but the cost per job saved has to be offset by both the jobs gained and the value lost to consumers from lower prices and more variety.

Technological progress also creates and destroys jobs. The occupational codes for the 1900 US census themselves illustrate the changes in a dynamic economy. Not only were farmers and agricultural laborers 34 percent of the labor force (compared to less than one percent today), but there were 212,104 Blacksmiths and 37,249 Livery and Stable Keepers. These particular jobs could have perhaps been protected—and perhaps even their wages maintained somehow—but at what economic loss per job “saved”?(Of perhaps some relevance is that the 1900 Census also listed 65,310 whose occupation was Hucksters and Peddlers).

Demonstrating an economic policy does not benefit literally everyone is not an argument against that policy, it is the acknowledgement that all policy changes, even those that are massively overall welfare-improving and hence potential Pareto-optimal, have winners and losers. The costs to those affected by economic changes are an important and legitimate concern to economists, policy makers, and to politicians. But there are almost always vastly more cost-effective policies or programs to help these workers than restrictions on markets—trade, competition, innovations—that impose costs on all consumers (and other industries) in the USA.

Second, the “instruments to targets” approach might seem unrealistic, or even callous to the potential losses of the already disadvantaged, if instruments to help the disadvantaged did not exist or those instruments were themselves ineffective or if the instruments could not feasibly be scaled to address the losses to the super low skilled BSG from labor mobility. But in the context of the USA absolutely none of those things is true. The newly created data source provides a treasure trove of data about what the government—federal and state—does in the USA. In 2014 the government in the USA spent $5,385 billion. Of that they report $862 billion was spent on “Standard of Living and Aid to the Disadvantaged.” The US state and federal governments spend $786 billion on “Education.” One particularly effective program for poorer workers is the Earned Income Tax Credit, which provides a refundable tax credit that, in essence, increases the wages of workers with low wages. This program provided $60.1 billion to working families in 2014.

A 28 percent increase in less than HS education migrants (the Mariel Boatlift experience) would mean about 4.2 million new migrants. The gain in (PPP adjusted) wages for each of those new migrants would be about $15,000 (CMP). So the first and most obvious consequence would be a gain to the movers of about $63 billion dollars and their total wages in the USA would, conservatively, be about $82.3 billion dollars.

Suppose that Borjas (2015) is right and the 28 percent increase in the supply of low-skill labor in Miami did cause wages for the Borjas Sub-Group to fall (and there is no evidence it affected any workers but that group so we are going to stick to just that group). Using the interface for the Public Use MicroSample (PUMS) for the March 2016 wage of the Current Population Survey we can see that the BSG has 1.8 million workers of a total USA labor force of 150.8 million—1.2 percent of the labor force. The average annual wages of this group are $32,996. Using the EITC formula a married couple with this level of income and two children would be eligible for cash refund of $3,627. So, there already exist many instruments to help families with low wages and just one of those, the EITC, provides a 10 percent boost to their income.

Suppose that a 28 percent increase in the super-low-skill labor force from 4.2 million migrants caused a four percent wage fall for the BSG. While Borjas claims much higher numbers for this narrow group from Mariel Boatlift I think the combination of shift in sampling (Clemens and Hunt 2017), methodological issues (Peri and Yasenov 2017) and the incredibly small samples imply a four percent loss is a generous estimate that cannot be rejected statistically as being too small for the BSG.

A four percent wage fall for the BSG would mean a wage loss of $59.8*.04=2.4 billion dollars.

There are several ways of putting this number of the potential wage losses in perspective.

First, it is 0.28 of one percent of the $862 billion US governments spend on assistance to the disadvantaged.

Second, it is 3.6 percent of the $67 billion in EITC payments. So a 3.6 percent increase in EITC payouts could fully compensate the affected BSG.

Third, the loss to the BSG is only 2.9 percent of the total wages of the newly admitted migrants (on the conservative assumption these new migrants with less than HS only make 80 percent of the existing wages of less than HS educated of $24,512). This implies that, suppose we wanted to have a “compensation fund” for the wage losses to the BSG that was fully funded out of a tax on the newly admitted migrants (bracketing for one minute the fairness of that) that would leave the movers massively better off and compensate the estimated wage losses. In fact, in the current situation many undocumented workers pay their full Social Security tax rate of 6.2 percent but will never see any benefits. Estimates of the net payments to Social Security are around $12 billion a year—far more than the wage losses to the BSG.

Third, once one puts magnitudes on the losses to less-skilled natives and takes an “instruments to targets” perspective the use of this argument against greater levels of low-skill labor mobility starts to look facetious, if politically astute.

That is, suppose one were devoted to improving the well-being of US citizens that were economically disadvantaged and were seeking the top five or ten most cost-effective ways to make their lives better. An expansion of EITC—a program that encourages work and labor force attachment, provides cash benefits, and has almost no administration cost—looks really attractive. Housing vouchers of a “moving to opportunity” type have, with fungibility, at least the impact of a cash transfer plus, with recent evidence on long-term benefits, perhaps some positive impacts on inter-generational transmission of poverty. Spending resources that prevent people from ending up as high school drop-outs, particularly early childhood spending on disadvantaged children, is argued by James Heckman to be cost-effective. Citizens with less than high school complete are disproportionately black and latino and so perhaps efforts to reduce racial and ethnic discrimination would be cost effective. If I saw someone making the strong, evidence-based argument for improving the lives and livelihoods of the disadvantaged and on that list were restrictions on low skill migration assessed on a cost-benefit basis with other potential interventions, that is worthy of serious consideration.

On the other hand, suppose I were opposed to the entry of low-skill migrants into my country and wanted to make the argument that had the broadest possible political and social acceptability and made me look noble. I obviously would not use the argument that “we” just don’t like “them” as this has the look and feel of widely distasteful racial and ethnic arguments made (and widely accepted) in the past. I obviously would not use arguments that “they” raise diversity and I don’t like diversity. I obviously would not argue that I just don’t like having poor people around. I obviously would not use the argument that “they” will debase our culture (at least not for a broad audience). Arguing that migrants hurt disadvantaged natives is, on the other hand, however weak empirically and pointless from an “instruments to targets” analytic basis, pure political gold.


CGD blog posts reflect the views of the authors, drawing on prior research and experience in their areas of expertise. CGD is a nonpartisan, independent organization and does not take institutional positions.