The GiveWell Blog

David Roodman’s draft writeup on immigration and current residents’ wages

As discussed previously, we are investigating the cause of labor mobility as a potential focus area within U.S. policy. Much of our investigation is focused on outlining potential giving opportunities; concurrently, we are interested in reviewing the academic literature on the merits (and possible drawbacks) of the policy changes that we would be working toward.

One key question around this cause is whether increasing immigration to the U.S. (something that we believe could be an excellent outcome in global anti-poverty terms) would result in lower wages for current U.S. residents. We commissioned David Roodman to provide a critical review of the literature on this question. David previously completed a project for us on the connection between infant mortality and fertility.

We haven’t yet fully vetted this writeup (something we are planning to do), but we believe it gives a thorough and convincing picture of the literature, and provides some reason to believe that immigration is unlikely to result in substantially lower wages (particularly over the long run) for current residents.

From the introduction:

As ever, the evidence base is not as sturdy as we would wish. Ironically, immigration policy is often arbitrary and even randomized (as in visa lotteries), which has allowed some high-quality measurement of impacts on immigrants (Gibson, McKenzie, and Stillman 2010; McKenzie, Gibson, and Stillman 2010; Clemens 2013b). Unfortunately, this randomness has not been as exploitable when assessing impacts on the receiving economy, because admissions have not been randomized across occupations, say, or cities. One family of studies attempts the next-best thing, exploiting natural experiments, and some are persuasive. The rest of the research is less experiment-like, for example, looking at correlations between wages and immigration flows across US cities over 20 years, and so must be taken with more grains of salt. Most of non-experimental studies reviewed here make a bid in the direction of natural experiments by instrumenting. But even when they aggressively check the instrumented results for robustness, it is always hard to be sure that the strategy is working.
Still, the available evidence paints a fairly consistent and plausible picture:

  • There is almost no evidence of anything close to one-to-one crowding out by new immigrant arrivals to the job market in industrial countries. Most studies find that 10 percentage point increase in the immigrant “stock” as a share of the labor force changes natives’ earnings by between –2% and +2% (Longhi, Nijkamp, and Poot 2005, Fig 1; Peri 2014, Pg 1). Although serious questions can be raised about the reliability of most studies, the scarcity of evidence for great pessimism stands as a fact. The economies of destination countries largely appear flexible enough to absorb new arrivals, especially given time.
  • The group that appears most vulnerable to competitive pressure from new low-skill migrants is recent low-skill migrants. This possibility is easy to miss when talking of the impacts of “immigrants” on “natives.” Yet it stands to reason: a newly arrived Mexican with less than a high school education competes most directly with an earlier-arrived Mexican with less than a high school education.
  • One factor dampening the economic side effects of immigration is that immigrants are consumers as well as producers. They increase domestic demand for goods and services, perhaps even more quickly than they increase domestic production (Hercowitz and Yashiv 2002), since they must consume as soon as they arrive. They expand the economic pie even as they compete for a slice. This is not to suggest that the market mechanism is perfect—adjustment to new arrivals is not instantaneous and may be incomplete—but the mechanism does operate.
  • A second dampener is that in industrial economies, the capital supply tends to expand along with the workforce. More workers leads to more offices and more factories. Were receiving economies not flexible in this way, they would not be rich. This mechanism too may not be complete or immediate, but it is substantial in the long run: since the industrial revolution, population has doubled many times in the US and other now-wealthy nations, and the capital stock has kept pace over the long term, so that today there is more capital per worker than 200 years ago.
  • A third dampener is that while workers who are similar compete, ones who are different complement. An expansion in the diligent manual labor available to the home renovation business can spur that industry to grow, which will increase its demand for other kinds of workers, from skilled general contractors who can manage complex projects for English-speaking clients to scientists who develop new materials for home building. Symmetrically, an influx of high-skill workers can increase demand for low-skill ones. More computer programmers means more tech businesses, which means more need for janitors and security guards. Again, the effect is certain, though its speed and size are not.
  • An important corollary of this last observation is that a migrant inflow that mirrors the receiving population in skills mix is likely to have the most benign effects. Especially once capital ramps up to match the labor expansion, a balanced inflow probably approximates a dilation of the receiving economy, with similar percentage increases in all classes of workers, concomitant growth in aggregate demand, and minimal perturbation in prices for goods, services, and labor. In particular, one way to cushion the impact of low-skill migration on low-skill workers already present is to increase skilled immigration in tandem.

In addition to summarizing the existing literature, Roodman also includes a technical appendix that replicates the analysis in two key papers (Ottaviano and Peri 2012, which finds minimal costs for U.S. natives from additional immigration, and Borjas, Grogger, and Hanson 2012, which criticizes Ottaviano and Peri 2012 and argues for larger costs). While we don’t feel equipped to assess all of the details, our impression is that Roodman offers novel technical arguments implying that Borjas, Grogger, and Hanson’s criticisms of Ottaviano and Peri undermine their own arguments. We hope that these arguments are subjected to appropriate scrutiny, given the major role that these papers have played in the literature.

We encourage our readers to check out the writeup and send in their thoughts. Like David’s previous writeup for us, we think it is an interesting read.

David Roodman’s draft writeup on immigration and current residents’ wages


  • Robert Wiblin on September 11, 2014 at 5:44 pm said:

    I just want to say that I find these reviews highly informative and appreciate GiveWell funding them.

  • Colin Rust on October 4, 2014 at 9:42 am said:

    You may want to fix the first bullet to match David’s revised wording.

    (The version above sounds like it is talking about the effect of a 10% increase in the number of immigrant workers. I don’t know about others, but I was initially surprised by how wide the error bars were in terms of the effect on native workers: it sounded like there was potentially a pretty large effect, positive or negative, on native workers. But in fact it is talking about the effect of a monster increase in the number of immigrant workers that represents 10% of total workers. So with that — correct — interpretation it is saying that the overall effect of immigrant workers on native workers might be of either sign, but it is likely pretty slight on average.)

  • Alexander on October 6, 2014 at 2:26 am said:

    Colin – fixed, thanks for the suggestion.

  • Colin Rust on October 6, 2014 at 8:47 am said:

    Thanks Alexander, but it doesn’t looked fixed. The old wording, which I still see above is:

    Most studies find that 10 percentage point increase in the immigrant “stock” changes natives’ earnings by between –2% and +2%…

    The new wording inserts “as a share of the labor force”:

    Most studies find that 10 percentage point increase in the immigrant “stock” as a share of the labor force changes natives’ earnings by between –2% and +2% …

    (On why this matters, see the parenthetical in my comment above.)

  • Alexander on October 6, 2014 at 10:42 am said:

    Colin, thanks for the catch; I had changed it from 10% to 10 percentage points but missed the labor force language. Actually fixed now.

  • Ben Rachbach on December 15, 2014 at 7:33 pm said:

    I recently completed an internal review of David’s writeup. I found his reasoning and conclusions to be basically sound, though I did not investigate the details of his analysis. (We complete internal reviews of most pages to check their use of evidence, analysis, and conclusions.)

Comments are closed.