Important Errors in a Highly Publicized Paper on the Economic Impacts of Japanese American Imprisonment (UPDATED 8/15)

            It is rare for an academic paper to make national news, but “The Causal Effect of Place: Evidence from Japanese-American Internment” managed the feat this past week, with a prominent article on Newsweek.com and in the science press.  The paper, by Daniel Shoag of the Harvard Kennedy School and Nicholas Carollo of the Economics Department at UCLA, builds on the fact that the roughly 110,000 Japanese resident aliens and Japanese American citizens forced from their West Coast homes in 1942 ended up in concentration camps in different parts of the country.    

            The claim that drew headlines was that “an internee’s randomly assigned location causally affected their economic outcomes and the economic outcomes of their descendants” (28).  That is, certain attributes of the place where a Japanese American was imprisoned influenced his or her economic fortunes later in life, and those of the next generations.  This is important, the authors say, because it can inform the choices of “policymakers” who are “struggling to resettle and integrate large refugee or immigrant populations today.”

            The paper performs a range of statistical analyses to support its conclusions, but there is a mistaken assumption at the heart of the paper that undermines its analysis. 

            Camps were sited in remote parts of Arizona (two camps), Arkansas (two camps), California (two camps), Colorado, Idaho, Utah, and Wyoming.  Shoag and Carollo use the year 2010 median income in the county where each camp was located as “a measure of the quality of the location to which internees were assigned.” (11)  In other words, the authors deem a camp sited in a more prosperous county to be a “richer camp” (14)  and a camp sited in a poorer county to be a “poorer camp” (25-26).  They then track the economic lives of the prisoners from each camp across the post-war years and conclude that those who were imprisoned in “richer” camps – and their descendants – did better economically than those imprisoned in the “poorer” camps. 

            It’s important to note that they are claiming not just correlation but causation:  regression analysis allows them to assert that it was the experience of the relative “wealth” of the camp – as approximated by its county’s 2010 median income – that caused post-war effects on economic fortunes. 

            Here is the problem: the median income of the county where a camp was sited had nothing to do with the conditions inside the barbed wire of the camps.  Nothing.  The camps were economic isolation zones.  They were in no sense artifacts of, or even meaningfully connected to, the economic conditions of the counties that surrounded them.  If they were similar to anything in terms of an economy, they were similar to each other regardless of location.

            The camps were built by the United States government to uniform specifications.  They were administered by officials who reported to a central agency in Washington, DC.  The regulations governing inmates’ lives were essentially the same at every camp.  Prisoners who worked in camp were paid on a single national pay scale regardless of their location -- $12 per month for unskilled labor, $16 per month for semi-skilled labor, and $19 per month for skilled labor.  The inmates ran camp commissaries and other services that used similar business models, charged similar prices, and reaped similar profits.  Outside of that they mail-order shopped from the same Montgomery-Ward and Sears catalogs.  They engaged in similar agricultural operations (dependent, of course, on climate).  They lived in barrack rooms similar to those at all of the other camps and ate meals in mess halls whose kitchens’ food allotments were uniformly regulated.  The inmates at all of the camps were forbidden from leaving camp to interact with the surrounding communities except with written permission.  The standards for granting such permission were similar from camp to camp, and permission was not frequently awarded – in part because surrounding communities typically did not want the prisoners circulating. The prisoners’ economic interactions with the communities that surrounded them were minimal.  Some white-owned businesses in the surrounding communities made considerable money off of the presence of the inmates (think of dairies and mortuaries, to take just two examples), but this was a one-way economic street. 

            I could go further, but the point should be clear:  there was no meaningful sense in which the median income of the surrounding county reflected anything about life inside the camps or the financial condition of the prisoners.

            To the extent we can speak of “rich camps” and “poor camps” at all, we can refer only to the varying wealth of Japanese Americans as they entered the camps and to their success (or lack of it) in holding onto their West Coast property while they were locked up in the interior.  But Shoag and Carollo do not look at anything at all about the inmates’ lives before camp, or even at anything about the actual conditions in camp, to determine the camp’s relative “wealth.”  They look only to a data point about the mostly white people who lived free lives in adjacent towns and villages, apart in every meaningful sense from the barbed wire enclosures in their midst. 

            We can press the point a bit further:  in many places, however unjustifiably, the surrounding white communities harbored economic resentments toward their Japanese American “neighbors.”  In many of these remote surrounding areas, residents were using outhouses, sending their children to old and shoddy school buildings, and traveling very long distances for medical care, often in clinics with outdated equipment or no equipment at all.  From their vantage point, the Japanese Americans in the camps were living lives with luxuries they could not imagine: indoor toilets, fine new high school buildings, and new and up-to-date hospitals.  This is to say that if Shoag and Carollo were to tell a 1942 resident of Powell, Wyoming (near the Heart Mountain Relocation Center) or McGehee, Arkansas (near the Rohwer Relocation Center) that  those towns’ economies were proxies for the camps’, the researchers would get quite an earful.

            The paper is marred by many other errors.  It attributes responsibility for the President’s executive order to the Navy (6) when actually the Army pushed for it.  It calls the first-generation Japanese immigrants the “Nisei” and their second-generation children the “Issei” (10), when in fact it’s the other way around.  It assumes Japanese Americans exercised free choice over where they went upon leaving camp, when in fact the War Relocation Authority went to great lengths both to channel them to specific cities and regions, and insisted on reviewing each person’s relocation plans before approving them.  And it repeatedly speaks of Japanese Americans being “interred” (which means buried) rather than “interned.”  If the authors of the paper were familiar with the literature on this historical period, they would know that even the word “interned” is incorrect, because “internment” is a lawful procedure that governments can deploy against foreign nationals in certain situations, while the imprisonment authorized by Roosevelt’s executive order did not invoke that legal procedure.

            But those errors are, in context, quibbles.  The important point is that their analysis proceeds from a historically mistaken starting point.  There really weren’t “rich camps” and “poor camps,” and if there had been, it wouldn’t have been because of anything related to the median income of the counties surrounding them.

            For this reason, Shoag and Carollo’s paper cannot inform current discussions about where countries should locate large influxes of refugees and immigrants.  Refugees do in fact enter the communities in which they live; they become (or at least try to become) part of its economic and social life.  One can plausibly imagine their economic fortunes being impacted by the fortunes of the community they’ve entered. 

            Japanese Americans did not enter any community when they entered camp apart from the isolated Japanese American “city” where they were confined.

            At a bare minimum, beyond correcting the many factual errors, Professors Shoag and Carollo ought to produce a convincing justification for an assumption that defies the realities of the camps and the communities where they were situated.  If they cannot do so, they should seriously rethink the claims they advance in their paper.

UPDATE, 7/15/17:  The more time I spend with this paper, the more serious errors I see.  The paper identifies the Gila River Relocation Center in southern Arizona as a comparatively "wealthy" camp because the 2010 median income in the county in which it sits is relatively high (compared to the counties in which the other nine camps were located).  Indeed, the paper names Gila River as the richest of the camps. (p. 11 and elsewhere).  The errors here are two.  

First, Gila River sat in Pima County, Arizona -- the same geographically vast county that contains the metropolitan area of Tucson, a major commercial and academic center.  (It's where the University of Arizona is.) Gila River was the only one of the ten Japanese American concentration camps situated in the same county as a sizable city.  It shouldn't surprise us, therefore, that Gila River appears to be the "richest" of the camps by Shoag and Carollo's measure. But how could we possibly expect any of the wealth of the distant city of Tucson to have "rubbed off" on the camp?  Tucson was inside the West Coast military exclusion zone from which all people of Japanese ancestry were excluded (unless they were in a concentration camp located within it).  It was the same exclusion zone that ran all the way up the Pacific Coast, through California, Oregon, and Washington.  So a prisoner at Gila River would not have been allowed to be at liberty in the rest of Pima County at all (except highly limited purposes specifically authorized by the camp's administrators). To link Gila River in any way with the overall wealth of Tucson is absurd.

Second, and this is the more interesting point, Gila River only sat "in" Pima County, Arizona, in a technical sense.  In actuality the government sited it on reservation land of the Akimel O'otham and Maricopa Indians of the Gila River Indian Community -- sovereign territory of those Native American tribes (though obviously under federal administration).  (The tribes tried to reject the siting of the camp on their lands but, true to historical form, the federal government put it there anyway.)  The more relevant administrative unit for Shoag and Carollo to examine should be the Indian reservation within whose jurisdictional lines it sat, not Pima County, dominated by distant Tucson. I don't have the numbers, but strongly suspect that if keyed to the Indian reservation and not Tucson, Gila River would zoom from the top of the "wealth" chart to near bottom.

FURTHER UPDATE: One might think that the talk of "richer" and "poorer" camps was just imprecise talk, and that the authors' claim is that the camp's surrounding county's median income is a useful variable because, if prisoners were likelier to remain in the state where they were incarcerated than to go elsewhere, it's the local economy that they would've entered.  Here's one basic trouble with this.  It's a data problem.  The place the authors look for destinations is records of where they were living in the early 1990s.  If one wants to know where prisoners went when they left camp, though, one doesn't need to wait until 1992 data to know (and extrapolate backwards in time).  One can look directly at the WRA records, all available online for each camp, that show, for each prisoner, where he or she went upon leaving camp.  It's the last column on the right in this image.  (The fourth column from the left shows where each prisoner came from.)

It doesn't appear that the paper used this available data source.  But if you want to know which city and state's economies the prisoners entered when they left the camps, that's the data source that actually documents it, down to each individual person.

Comments

[Anon]

Isn't doing this kind of "research" akin to relying on Mengele's "medical" abuses as research or the CIA's use of torture??

[Eric Muller]

Just a quick note to say that it's my policy not to respond to anonymous comments. I say this only to make clear that by not replying to what was posted above I am not endorsing it.

[J Burton]

Not that it changes your excellent points, but to be accurate the Gila River camp was in Pinal County.