Skip to content

“Why do the results of immigrant students depend so much on their country of origin and so little on their country of destination?”

Aleks points us to this article from 2011 by Julio Carabaña.

Carabaña’s article has three parts. First is a methodological point that much can be learned from a cross-national study that has data at the level of individual students, as compared to the usual “various origins-one destination” design. Second is the empirical claim, based on data from the Program for International Student Assessment (PISA), that the test scores of immigrant students depend so much on their country of origin and so little on their country of destination. Third is their discussion of possible explanations. The explanation they favor is national differences in cognitive or learning ability.

The first point makes complete sense to me.

The second point I’m not sure of, as I’d think that much would depend on how these data are collected and analyzed. The statement, “the results of immigrant students depend so much on their country of origin and so little on their country of destination,” could be a reasonable data summary or “stylized fact,” or maybe it’s not. I’m not saying it’s wrong; I just haven’t worked through the data and analysis so I’m not making a statement about it, one way or another.

The third point I’m not so sure about, in part because it’s not clear what’s so special about national origin. The destination countries are not uniform in their test scores—that’s a key point of the paper—so it’s not clear that why makes sense to treat the origin countries in that way. Different immigrant students come from different groups in the origin countries, right? Also, if it’s correct as stated in the paper that these differences last for generations, then, the concept of an origin country doesn’t seem quite right.

Again, I’m not saying the claims in the paper are wrong, exactly; I just don’t think they’re framed as clearly as it might seem at first. Once you start talking about differences of ethnic groups within a country, the whole concept of country of origin starts to come apart.


  1. Richard F says:

    “Different immigrant students come from different groups in the origin countries, right?” I’m wondering if that statement is, indeed, correct. I speculate that the students that have the opportunity to study abroad are a subset and, possibly, a more homogeneous subset of the society. I speculate that for some countries, the students that go abroad come from the more privileged part of their society or maybe they all come from the capital city. I’m still not sure why that would lead to donor country variation in success, though. It may even be that countries that send a more heterogeneous group of students have the better outcomes.

  2. Dale Lehman says:

    I have never looked at any of this research, so I am asking this of those more familiar with it. As I understand it, countries differ in their immigration policies, particularly when it comes to refugees. So, when we compare educational attainment across countries, immigrants may well vary in characteristics by destination country, even if the have the same origin nation. Richard F’s speculation may indeed be correct, but the issue seems far from certain to me. It is an empirical question and I wonder whether there is any research that sheds light on the relative homogeneity/heterogeneity of both immigrant groups by nation and destination county policies towards immigrants. My initial thought would be that both vary considerably.

  3. “Different immigrant students come from different groups in the origin countries, right?”

    This seems relevant:

    “We find that the disadvantage in educational outcomes between the second generation and their peers from majority populations is smaller for ethnic groups that are more positively selected in terms of educational attainment.”
    Selectivity of Migration and the Educational Disadvantages of Second-Generation Immigrants in Ten Host Societies
    European Journal of Population volume 35, pages347–378(2019)

  4. somebody says:

    I’m struggling to follow the claims in the paper because both the writing and data presentation are of a pretty low quality. But it does seem like they address your question about treating origin countries as uniform.

    > It is well known that emigrants never constitute a random sample from their country populations. Certain regions, social classes, ethnic groups and even personality traits often are over- or under-represented among them. Part of the deviations from the country of origin means observed in Tables 4 and 5 might simply derive from homologous deviations in immigrants’ social background. Economic emigrants from developed countries tend to be of higher socioeconomic background than their counterparts from developing countries. The children of those workers and peasants that emigrated from Italy, Portugal, Serbia and Spain into Central-Europe score below their origin countries’ means, but perhaps at one level with the children of the same social classes who have not moved abroad. We could then expect that controlling for socioeconomic and cultural characteristics of the parents would reduce still further the differences observed in Tables 4 and 5. Dronkers & Heus (2008) have actually calculated that controlling for cultural possessions at home, home educational resources, parental education and parental occupation, somewhat reduces the gap in question. On the other side, selective immigration policies could explain that emigrants to Australia outscore sometimes their native co-nationals, even if they remained in China (Table 6).

    Again, I can’t be certain I’m correctly following the argument, but it seems like the claim is that the differences between scores of immigrant children and children in the country of origin are explained by heterogeneity within country of origin?

    A more worrying thing about this paper is that the data don’t clearly support the conclusion. The matrix on which the paper is premised is located at the end in the appendix, and it’s tough to see any trend. Admittedly, this is in part because of the fact that there’s no visualization to facilitate meaningful comparisons, but just eyeballing it seems like it’s dominated by EU scores that vary within a pretty tight range, and if scores of destination countries are pretty similar to scores of origin countries. The only big trend that sticks out to me is Turkish emigrants scoring substantially lower than in Turkey. There are also a number of some pretty standout cases of emigrants quite dramatically outperforming or underperforming relative to students in the country of origin. They seem to acknowledge this?

    > It is, however, possible to consider the question in other way. Most countries of immigration in Tables 4 and 5 clearly belong to one of two groups, Australia and New Zealand on the one side and Central-Europe on the other side, all with high PISA scores. There is no surprise in the fact that migrating among these countries leaves PISA scores unaffected. What asks for explanation is the unchanged performance of immigrants from countries with lower scores. No
    such countries send emigrants to Australia and New Zealand, except Greece in Table 4, whose emigrants’ score rises in Australia. Therefore, in Tables 4 and 5 the case of immigrants in Central-Europe coming from South Europe seems the only one needing explanation.

    So the speculation is restricted to specifically immigrants in Central Europe from Southern Europe. So it seems like the title of the paper should be “why do the results of immigrant students depend so much on their country of origin and so little on their country of destination except for the cases I’m ignoring, but specifically on the issue of Central European immigrants from Southern Europe”. Their explanatory hypothesis (big surprise) is cognitive ability. But even then, I’m struggling to see what they’re seeing. Again, since it’s just a big data matrix, all we can do is look up examples individually.

    In the 2006 data

    1. For Croatia, there’s one datapoint on emigrants to Austria. The emigrants test quite a bit lower than both country of origin and destination, and in fact the origin and destination are more similar to each other than the origin and the emigrants. This does not “pass the test” according to the paper, presumably some kind of test of significance based on difference in means.
    2. For France, there’s two, to Belgium and Luxemburg. Again, French emigrants to Belgium are lower than both France and Belgium, and are less similar to French kids than French kids are to Belgian kids. French emigrants to Luxemburg score higher than both origin and destination, and the same is again true about the relative closeness. Belgium “passes the test”, Luxemburg does not.
    3. For Italy, there’s Germany and Luxemburg and it’s basically the same story again. Emigrants score lower than both country of origin and country of destination, and natives of the two countries are more similar to each other. Both “pass the test”.
    4. Portugal to Luxemburg, same story as Croatia. It “passes the test.”
    5. Serbian and Romanian emigrants to Austria are seemingly the only examples of emigrants testing very similarly to their country of origin in spite of a substantive difference between the origin country and destination country. Neither “pass the test”.

    As far as I can tell, the entire argument of the paper is premised on the two examples of Serbia and Romania. You could argue that emigrants from Southern European countries to Central European countries are directionally more similar to their origin countries than destination countries in that both the origin countries and emigrants test lower than the destination, but it certainly does not support that scores of immigrant students “depend so much on their country of origin”, nor is there a “fact that the learning outcomes of immigrant students are very similar to the learning outcomes of their con-nationals in their countries of origin”.

    The paper also makes some pretty wacky sociological claims

    “We must, therefore, turn our attention to the origin countries in search for something that makes immigrant children from low-score countries resilient to the benefits of the better schools, the wealthier economies and the richer social environments in the countries of destination, at least in Central Europe. It has to be something that immigrants carry from their origin to their destination countries, and that they preserve in such a way that it continues to
    determine the scholastic achievement of their children, even of those born, grown up and educated in the host countries. What can it be?

    Nothing material is so lasting and influential. Wealth, income or living conditions are exactly what emigrants leave behind when they arrive to new countries. In any event, it would be strange if these factors were as efficient, or even more, in the destination as in the origin country.”

    Huh? Are they claiming people who leave their home country because they’re poor will spontaneously become rich upon arrival in a rich country, or at least that their former poverty is no longer important?

    Honestly, this whole paper looks like a poorly written excuse to write this line

    “Attempts to explain group differences in academic results by social scientists rarely question the assumption of equal cognitive abilities”

    You say you haven’t looked into the analysis; well there honestly isn’t any. As far as I can tell there’s really the premise supposedly but not actually supported by the two tables of means, only included to present the illusion of a data driven argument.

    • NickMatzke says:

      “ Honestly, this whole paper looks like a poorly written excuse to write this line

      “Attempts to explain group differences in academic results by social scientists rarely question the assumption of equal cognitive abilities””

      Yep agreed.

  5. Zhou Fang says:

    I thought it’s well established that cross-national comparisons of PISA scores is methodologically dubious?

  6. Joshua says:

    > The explanation they favor is national differences in cognitive or learning ability.

    Can’t read the paper – but my prior is that there’s no way that they can disaggregate the influence culture vs. cognitive/learning ability.

    IOW, I tend to doubt that they settled the nature/nurture question (which is a false dichotomy, BTW) in the process of writing this paper.

    • steven t johnson says:

      I advocate replacing “nature/nurture” with “development/biography.” I suppose you could technically see this as pushing the false dichotomoy onto “biography versus history,” except, does anyone really see biography and history as dichotomous?

      • Joshua says:

        I don’t really understand. What is the difference between biography and history?

        • Martha (Smith) says:

          I see “biography” as the history of one individual. “History” is a broader word that includes biography, but is more often used to refer to the history of groups, nations, and other large groups of people.

        • steven t johnson says:

          I don’t think there truly is the kind of opposition between biography and history in the way people imagine an opposition (dichotomy) between nature and nurture. Which is why I speculated nobody (not literally) does see such a dichotomy. The point is there is no such dichotomy between nature and nurture.

          (Also, “nurture” is misread as “only formal education,” I think. And “nature” is something like character is destiny or some such vaguely secularized version of religious thinking. This latter in particular I think is the real difficulty with nature/nurture dichotomy.)

  7. Mendel says:

    My guess is that emigration and immigration correlates with low economic status (poverty), and so does school performance.

  8. TO says:

    Critical thinking is an essential research skill. Many immigrants have to learn a new language. Learning and writing exams in a second language require more mental energy. Coupled with the daily communication challenges with peers, educators etc, immigrants generally are tasked mentally beyond what is normal, which may contribute to some of the disparities in achievements. I hope they control for effects like this.

  9. mpledger says:

    I would say language has an impact – not that many migrants go to a country with the same first language. It’s hard to do well on a test when it’s in a language that you have only been exposed to for a couple of years.

  10. James B says:

    A lot of the confusion melts away once you allow yourself to consider the obvious but taboo explanation: IQ is hereditary, differs on average across countries of origin, and explains a large share of variance in test scores.

  11. Happy to see Andrew help promote this funny but honest book chapter. I found it sometime back early this year, promoted it on Twitter, and it has now found its way to the Gelman blog. It continues Andrew’s new hobby of doing readings in race science by studying national gaps and their origins. The Streisand effect from the politically motivated retraction of Cory Clark et al paper only helps gather more attention.

    Another case of these studies could be Martin Voracek’s studies of suicidality, like this one:

    “Conclusion: Following the logic of the migrant study design of genetic epidemiology, the correspondence of IMM and COB suicide rates is consistent with the assumption of population differences in the prevalence of genetic risk factors for suicide.”


    • Andrew says:


      1. Regarding “funny but honest,” recall that honesty and transparency are not enough. Scientists should always be honest and transparent, but that doesn’t mean that they’re doing useful or competent science.

      2. This is not a hobby or a new hobby or whatever. Sometimes people point me to research that might be of interest and I post on it. I think this is more useful to the world to post and discuss in public than to reply by email to one person.

      3. If you think the retraction by Clark et al. was politically motivated, or if you think they retracted it in order to gather more attention, you should take it up with the authors, as they’re the ones who retracted it.

      • Re. turning emails into posts, I have in fact started doing this on my own blog following your example.

        You know who said it was political? It was professor Gelman! “It’s pretty clear that all this fuss is about the article’s political content, not its data problems.”.

        • Andrew says:


          Fair enough: I too think the fuss is about the article’s political content, and I think the whole thing is political. But is it a politically motivated retraction? That I don’t know. For that you’d have to ask the authors of the paper, as they’re the ones who did the retraction.

    • Joshua says:

      Emil –

      > The Streisand effect from the politically motivated retraction of Cory Clark et al paper only helps gather more attention.

      It’s much easier to cry “politicization” than to critique the critiques.

      It’s also far less useful, unless your goal is to justify self-victimization.

      Why not step up to the plate and make an actual argument rather than simply argue by assertion?

  12. Anonymous says:

    “Different immigrant students come from different groups in the origin countries, right?”

    coupla points:

    1) no, not necessarily; the US favors immigration apps of people who already have family in the US, so certain extended families, and thus ethnic groups, can develop pipelines to the US. I suspect the family preference is common in other developed countries too because it ensures the immigrant will have financial support.

    3) even though ethnic groups or whatever may have persistent characteristics, governments also have policies that may – in theory at least – apply across ethnic groups.

  13. Michael Nelson says:

    Immigrants tend to settle among others from their country, not always by choice. And immigrants from certain countries get a more welcoming reception than others. It’s easy to see that, for example, not everyone who immigrates to America lives in the same country. It may be a mistake to code them as if they do.

Leave a Reply