I don’t have time to post anything new on the blog right now, but in case you’re interested, I got some interesting pushback from Earthly Knight and Matt in the comments of my post about Nima Sanandaji’s book the other day and I think the discussion that ensued is worth reading. Many issues came up in that discussion which I plan to discuss more on the blog at some point, but I have too much work for the moment, so it will have to wait.
17 thoughts
Comments are closed.
I have a follow-up as well. I did actually manage to find Census data on the poverty rates of Swedish-Americans (as well as Finnish-, Norwegian-, and Danish-Americans) after 1990. Specifically, the Census Bureau lists demographic information for ancestry groups for the years 2000 (actually 1999) and 2005-2015. The specific figure of 6.7% poverty for Swedish-Americans doesn’t hold for any year besides 1990 as far as I can tell, so that seems like the year that Sanandaji is referring to when he cites that statistic.
Since Sanandaji is comparing Swedish-American poverty with Notten and de Neubourg’s (2011) estimate of Swedish poverty for the year 2000, the figure we’re interested in is Swedish-American poverty in 2000. According to the Census Bureau, Swedish-American poverty for 1999 (one year earlier) was actually lower than the 6.7% figure Sanandaji uses. In 1999, 5.5% of self-identified Swedish-Americans were poor (see the table I created using the Census Bureau’s website. The poverty figure is on p. 7).
However, this doesn’t help Sanandaji’s case. To explain why, I need to go into a brief digression on how the Census Bureau calculates poverty.
The US government’s official poverty definition was devised by economist Mollie Orshansky in 1963. She calculated a “subsistence food budget” for a family of four, then multiplied that by 3. Any family of four falling below that threshold counts as poor (in other words, if the family has to spend more than a third of its income in order to maintain the “subsistence food budget”, that family is counted as poor, whether or not it actually does spend more than a third of its income on food; note that this is an “absolute poverty” measure, not a “relative poverty” measure). “Equivalency scales” are then used to calculate the poverty threshold for smaller and larger families. The poverty line has been adjusted for inflation every year since 1963. The income that goes into poverty calculation includes things like pre-tax labor income, interest, dividends, and cash transfers (e.g., Social Security, unemployment insurance, TANF). It does not take account of taxes, nor does it include in-kind transfers (e.g., food stamps, Section 8 housing vouchers) or tax-based transfers (EITC). It doesn’t include capital gains or losses.
I will call this kind of poverty “Orshansky-poverty”, more specifically, I will call it “gross income Orshansky-poverty”, since it does not include the impact of taxes and tax credits. You can also define a “net (or disposable) income Orshansky-poverty”, which would take account of the impact of the tax code. The latter measure is more interesting, since it doesn’t do me much good if my pre-tax income makes me gross income Orshansky-non-poor, but then the government taxes me until I’m net income Orshansky-poor. However, for some reason, the US Census Bureau continues to use the gross income Orshansky definition, and that is presumably the definition that the Census Bureau is using when it lists poverty rates by ancestry. So, when the Census Bureau says that the poverty rate for Swedish-Americans was 5.5% in 1999, it means that 5.5% of Swedish-Americans were gross-income Orshansky-poor in the year 1999.
Right off the bat, we can see that the Notten and de Neubourg (2011) data that Sanandaji relies on is incomparable to the Census Bureau’s data for Scandinavian-Americans, since different definitions of poverty are used. Notten and de Neubourg (2011) describe how their measure of poverty differs from the US government’s:
So, when Sanandaji observes that 6.7% (actually 5.5%, but whatever) of Swedish-Americans are gross-income Orshansky-poor, and 9.3% of residents of Sweden (including immigrants and their descendants) are some-other-kind-of-poor, this is true but irrelevant. He is equivocating on the word “poor”.
As it turns out, Notten and de Neubourg have another paper where they calculate Orshansky-poverty (both gross income and net/disposable income) for the same countries during the same years as Notten and de Neubourg (2011). In this paper, Notten and de Neubourg “selected household budget surveys for both [Europe and the US] that are reasonably comparable in terms of collection and variables…” and “used purchasing power parity (PPP) rates to convert the Orshansky thresholds to the price levels in each of the European countries” (Notten and de Neubourg 2007: 7). Notten and de Neubourg also calculated Orshansky-poverty rates using both gross income and net income, although they mainly focus on the latter.
Given these considerations, Notten and de Neubourg (2007) should be a lot more interesting than Notten and de Neubourg (2011) when it comes to comparing Scandinavian poverty rates to Scandinavian-American poverty rates, so it’s remarkable that Sanandaji chose to use the latter paper.*
Unfortunately, Notten and de Neubourg (2007) didn’t have data on Swedish gross incomes, so they were only able to calculate the net income Orshansky-poverty rate for Sweden. By this measure, Sweden had a 5.7% poverty rate in the year 2000. This is slightly higher than the Swedish-American figure of 5.5% for that year, but we should remember (1) the Swedish figure includes immigrants, and (2) the Swedish-American figure is based on gross income while the Swedish figure isn’t, and (looking at Notten and de Neubourg 2007: 36, Figure 1) in no country was the gross-income Orshansky-poverty rate higher than the net-income Orshansky rate for the year 2000, and for almost all the countries measured it was lower (Only in Portugal was gross-income Orshansky poverty higher than net income Orshanky-poverty, although in 2000 the two measures converged).
It is quite probable, then, that the gross income Orshansky-poverty rate of ethnic Swedes living in Sweden was lower than that of Swedish-Americans in the year 2000, and possibly quite a bit lower.
Furthermore, Notten and de Neubourg do have gross income Orshansy-poverty data for Denmark and Finland (Norway was not included in the study), and the Census Bureau also has poverty data for Danish- and Finnish-Americans. In 1999, Danish-Americans had a poverty rate of 5.0%, and Finnish-Americans had a poverty rate of 6.3% (table). Eyeballing the graphs on Figure 1 on p. 36 of Notten and de Neubourg (2007), we can see that the gross income Orshansky-poverty rates for Finland and Denmark were well below 5%.
Remember also that Finland and Denmark have immigrants (though not nearly as many as Sweden), and that these people would be included in the poverty figures.
Unless I’m missing something, this appears to me to conclusively refute Sanandaji’s contention that Americans of Scandinavian ancestry have lower poverty rates than their cousins back home.
*Sanandaji actually does cite Notten and de Neubourg (2007) in his (2011), but he is very sloppy in this citation (as I’ve come to expect from him). He claims that “the poverty rate for Americans with Swedish ancestry is 6.7%….” (Sanandaji 2011: 24), citing “US Census 2006-2008” in a footnote. This is indeed about (though not quite) the average poverty rate for Swedish-Americans in the years 2006, 2007, and 2008 (6.5%, 6.5%, and 7.0% respectively. See this table. On the left, you can switch back and forth between the data for any years between 2005 and 2015). However, he goes on to compare this to a Swedish poverty figure of 6.7% (citing Notten and de Neubourg 2007), noting that this is exactly the same as that for Swedish-Americans (though he admits that the Swedish figure includes immigrants, without noting that this admission severely damages his case). However, Notten and de Neubourg (2007) don’t have any data for Swedish poverty in 2006-2008; the last year they include is 2000. In 2000, the Swedish net income Orshansky-poverty rate was 5.7%, not 6.7% (though in 1998 it was 6.7%) (Notten and de Neubourg 2007: 35, Table 2). Also, as I will mention in the text above, Notten and de Neubourg (2007) didn’t have data on Swedish gross incomes (2007: 34, Table 1), so they only measured Swedish net income Orshansky-poverty, whereas the Census Bureau’s figures for Swedish-American poverty is almost certainly based on gross income.
Notten and de Neubourg (2007)
Notten and de Neubourg 2011
Sanandaji (2011)
I’ll reply to this when I have some time, but did you see my latest reply on the other thread? I ask because I explained some of the things you say in your comment here, and I also gave a link to the data about poverty for Swedish-Americans from the Census Bureau, so I’m thinking that you may have missed it. On the other hand, what you say about Notten and de Neubourg, 2007 does address the point I made in my reply on the other thread, so perhaps you did see it but I just want to make sure. Anyway, like I said, I will reply to the points you make here later.
You’re right, I didn’t see your reply. However, I think the comment above addresses what you say there, since Notten and de Neubourg (2007) allows us to directly compare Danish- and Finnish-American gross income Orshansky poverty to Danish and Finnish gross income Orshansky poverty, which is a better method than any conceivable one based on indirect calculations. Once we make the direct comparisons, we can see that the Danish- and Finnish-American gross income Orshansky poverty rates were roughly at least twice as high as the poverty rates of their ancestral countries, in both cases.
Okay, sorry it took me so long to reply, but I finally have some time to do it.
Your argument is mistaken for the same reason that, as you pointed out, Sanandaji’s argument is mistaken, although his conclusion is true. You claim that, if you compare the poverty rate of Swedish-Americans from the Census Bureau with the poverty rate in Sweden as calculated by Notten and de Neubourg (2007), you find that the latter is significantly lower than the former.
You say that it’s a sound comparison since both the Census Bureau and Notten and de Neubourg (2007) use Orshansky’s definition of poverty. However, you are wrong about that, because although both the Census Bureau and Notten and de Neubourg (2007) used Orshansky’s definition of poverty, they didn’t use the same data for the US. Indeed, Notten and de Neubourg (2007) didn’t use the data from the the Census Bureau, but the data from the Panel Study of Income Dynamics, supplemented by the data from the Cross-National Equivalent File to allow the construction of pre-tax and post-tax income. (See p. 11 of their paper where they explain that.) In other words, you make the same mistake as Sanandaji, since your argument relies on a comparison between poverty rates that were not calculated in the same way.
You may think that it doesn’t matter that much, since at least both the Census Bureau and Notten and de Neubourg (2007) use the same definition of poverty, but it actually makes a big enough difference to change the conclusion of your argument. Indeed, according to Notten and de Neubourg (2007), the poverty rate in the US was only 8.7%, whereas it was 12.4% according to the Census Bureau. In other words, because they didn’t use the same data as the Census Bureau, Notten and de Neubourg (2007) calculated a poverty rate that is more than 24% lower than the poverty rate calculated by the Census Bureau, even though they used the same definition.
Now, as you can check using the advanced search function on the Census Bureau’s website, the poverty rate for Swedish-Americans calculated by the Census Bureau in 2000 was 5.5%. (If someone is reading this but can’t figure out how to find those data or doesn’t want to waste their time trying to figure it out, just email me and I’ll send you a pdf with the relevant data. I can’t give you a link because of the way in which the advanced research function on the Census Bureau’s website works.) Let’s assume that, relative to what it would have been if the data used for the calculation had been the same as in Notten and de Neubourg (2007), it was overestimated in the same proportion as the poverty rate for the US population as a whole. It follows that, had the Census Bureau used the same data as Notten and de Neubourg (2007) to calculate the poverty rate for Swedish-Americans, it would have been approximately 3.8% instead of 5.5%.
By comparison, according to Notten and de Neubourg (2007), the poverty rate in Sweden was 5.7% in 2000. As I noted in my previous reply to you, according to the Migration Policy Institute, 15% of the population in Sweden had a foreign background in 2000. Thus, in order for the poverty rate for the Swedish population as a whole to be 5.7% if we assume that people of Swedish ancestry in Sweden had the same poverty rate as Swedish-Americans, the poverty rate for people in Sweden who had a foreign background in 2000 would have to be approximately 16.4%. In other words, for that to be the case, the poverty rate for people in Sweden who had a foreign background in 2000 would have to be more than 4.3 times more likely to be poor than people of Swedish ancestry. For the same reasons I pointed out in my previous reply, although it’s even more true here, this is overwhelmingly implausible.
Therefore, I conclude that the poverty rate for Swedish-Americans in 2000 was almost certainly lower than the poverty rate for people of Swedish ancestry in Sweden, which is totally consistent with the calculation I did in my previous reply even though I didn’t use the same data. Of course, this calculation rests on the assumption that, relative to what it would have been if the data used for the calculation had been the same as in Notten and de Neubourg (2007), it was overestimated in the same proportion as the poverty rate for the US population as a whole, which we have no reason to believe is exactly true. But presumably it’s approximately true and, given how extreme the assumption we’d have to make for my conclusion to be false when assuming it’s exactly true, it’s overwhelmingly unlikely that my conclusion is false.
If we do the same calculation to compare the poverty rate of Danish-Americans with that of people in Denmark, we find that it’s approximately 3.5% for the former compared to 3.4% for the latter. Presumably, there were significantly less people with a foreign background in Denmark than in Sweden back then, but still some, so the poverty rate for people of Danish ancestry in Denmark was probably lower than for Danish-Americans. If we do the same thing to compare the poverty rate of Finnish-Americans with that of people in Finland, we find that it’s approximately 4.4% for the former compared to 4.9% for the latter. Since there were almost no people with a foreign background in Finland, at least with a non-European background, at the time, the poverty rate for Finnish-Americans was presumably lower than for people of Finnish ancestry in Finland. Notten and de Neubourg (2007) didn’t calculate poverty rates for Norway, so I can’t make the comparison, but I suspect it would have been favorable to Norway because of oil.
The bottom line is that, although Sanandaji’s methodology was flawed, his claim that people of Scandinavian ancestry have a lower poverty rate in the US than in Scandinavia is true except in the case of Denmark where it’s presumably a little bit higher. Even if people of Scandinavian ancestry in the US had been positively selected, which for the reasons I gave in my previous reply is hardly obvious, it would presumably still be the case that they have a poverty rate that is roughly the same as that of people in the country of their ancestors.
(1) N&N (2007) used different data than the Census Bureau.
(2) N&N (2007)’s figure for the US in 2000 was 8.7, whereas the Census Bureau’s figure for the same year was 12.4%, which is higher.
Both of these things are true. But is (2) true because of (1), as you claim? That would follow if (1) was the only difference between the way the Census Bureau calculates poverty and the way N&N (2007) calculated poverty for the US. But it isn’t.
For one thing, the N&N’s 8.7% figure is based on disposable income, whereas the Census Bureau’s is based on pre-tax income. But also, N&N (2007) take account of a major in-kind transfer in their calculations for the US: food stamps (N&N 2007: 16).
Since Scandinavian countries (wisely) don’t have food stamps (which is an awful program in my opinion, though not one that should be eliminated before being replaced with an equivalent cash subsidy, preferably to everyone rather than just low-income people), this doesn’t affect the calculation of Scandinavian poverty. Moreover, since the Census Bureau does not acknowledge food-stamps as income for their purposes, food stamps are not going to be relevant for the poverty rates of Scandinavian-Americans as calculated by the Census Bureau. So it remains true that, unless the different data sets used are wildly distorting the estimates (and what you’ve said so far doesn’t give me any reason to think that), we obtain a more direct comparison by looking at the Census Bureau’s figures for Nordic-Americans and N&N (2007)’s figures for Nordic countries.
It also seems to me that if we’re going to take account food stamps for the US, we should also take account of the generous in-kind benefits that Scandinavian countries offer. Universal healthcare free of charge (no, Medicaid doesn’t begin to compare), child day care centers and childcare subsidies and homecare subsidies, and generous active labor market assistance programs are all things that Nordic countries have that the US doesn’t. If I had to guess, I’d say the total package of in-kind benefits Scandinavian countries offer is much more generous than what the US offers. So if we’re going to include food stamps, let’s include all in-kind benefits, and then see how poverty rates compare.
There are also various limitations of the data in N&N (2007) that make the kind of back-of-the-envelope calculations you’re performing more difficult than you realize. When calculating the net-income poverty rate for the US, N&N assume that the take-up rate for EITC is 100%, which, as they acknowledge, is too high, since obviously not everyone who is eligible for EITC benefits actually receives them. (N&N 2007: 18). So N&N’s estimate of US poverty is probably an underestimate, as they acknowledge.
Furthermore, as the also acknowledge, N&N’s estimate of Scandinavian poverty is probably an overestimate, at least for the purposes of international comparisons. Unlike the other countries, poverty rates for Denmark, Sweden, and Finland were based on register data rather than survey data. N&N point to a Finnish study that found that survey data reports higher incomes for the lower part of the distribution than does register data, and that survey data reports lower incomes for the higher part of the distribution than does register data. This suggests that the poverty rates of Finland, Denmark and Sweden may have been lower had N&N used survey data for these countries, like they did for all the other countries.
Finally, one reason I’m skeptical of the sort of back-of-the-envelope calculation you’re using is that it yields different conclusions depending on which set of data you apply it to. In a previous comment I mentioned Gornick and Jäntti (2011). Take a look at Table 1 on p. 26. This is data for 2004. Now, the Census Bureau only has data for Nordic-Americans for 2005, but this shouldn’t be too much of a problem. The US government’s poverty rate for Swedish-Americans in 2005 was 6.8%, and the overall poverty rate was 12.6%: a 0.54 ratio.
Now multiply 0.54 by 9.4, the figure given for the US by Gornick and Jäntti, and we get a poverty rate (by G&J’s methodology) of about 5.15%. This is somewhat higher than G&J’s figure for Swedish poverty: 4.4%. Larger differences are obtained for Denmark and Norway. According to your method, Danish-American poverty based on G&J’s methodology should be about 4.5%, but G&J’s figure for Denmark was 2.9%. Norwegian-American poverty (by your method) should have been 5.15%, but Norwegian poverty was only 3.8% according to G&J. Only in Finland was the poverty rate roughly the same as that of the corresponding ethnic group: 5.50 as opposed to 5.45 (a difference like this is meaningless, given the roughness of the calculations; and, for all I know, G&J were rounding up).
When we compare child poverty (under-18 poverty) rates, we find that the advantage is even starker for Scandinavian countries. According to the Census Bureau (see Table 3), the child poverty rate for the US in 2005 was 17.6% (it wasn’t much different in 2004: 17.8%). The under-18 poverty rates for Nordic-American ethnic groups according to the Census Bureau were:
Danish-American – 6.4%
Finnish-American – 6.3%
Norwegian-American – 7.5%
Swedish-American – 7.3%
The child poverty rate for the US, according to G&J, was 11.8% for 2004. Applying your back-of-the-envelope calculation to these figures, we obtain the following Lemonine-Gornick-Jäntti child poverty rates for Scandinavian-American groups:
Danish-American – 4.3%
Finnish-American – 4.2%
Norwegian-American – 5.0%
Swedish-American – 4.9%
Now compare these to G&J’s figures for child poverty in the Nordic countries (G&J 2011: 27, Table 2):
Denmark – 1.9%
Finland – 3.4%
Norway – 2.6%
Sweden – 3.6%
These are all substantially lower than the child-poverty rates of the corresponding American ethnic groups. Danish-American child poverty, in particular, is 126% higher than Danish child poverty. Even Finnish-American child poverty is about 24% higher than Finnish child poverty.
Now let’s compare poverty rates for very young children. (G&J’s data is for children under 6, and so is the Census Bureau’s overall data for the US, but the Census Bureau’s data for ethnic groups is for children under 5. I’m assuming that this isn’t much of a problem.)
According to the Census Bureau, the poverty rate for families with children under 6 in the US in 2005 was 20.0% (same for 2004) (see Table 20). Also according to the Census Bureau, the poverty rate for families with children under 5 in the various Nordic-American ethnic groups was as follows:
Danish-American – 7.5%
Finnish-Americans – 8.7%
Norwegian-Americans – 10.1%
Swedish-Americans – 8.1%
According to G&J, the poverty rate for families with children under 6 in the US was 13.5%. By your method calculation, we obtain a Lemoine-Gornick-Jäntti under 5/6 child poverty rates of:
Danish-American – 5.1%
Finnish-American – 5.9%
Norwegian-American – 6.8%
Swedish-American – 5.5%
Now compare G&J’s figures for poverty rates for families with children under 6 in Scandinavia:
Denmark – 2.5%
Finland – 4.9%
Norway – 3.3%
Sweden – 5.1%
So, the poverty rates for children under 18 and under 6 are considerably lower in Scandinavian countries than they are for Americans of Scandinavian ancestry. Remember also that single motherhood (which reactionaries claim is the sole or overwhelmingly important cause of child poverty) is much higher among Scandinavians than it is among Americans of Scandinavian ancestry.
From these calculations, we can see that as far as poverty goes, the situation of people in Scandinavian countries is superior to that of Americans of Scandinavian background. Even in the case of Finland, whose poverty rate is roughly the same as that of Finnish-Americans by this methodology, their poverty rates for children under 18 and children under 5/6 are substantially lower. Especially since child poverty should be of special concern to humane people, we have to conclude that as far as poverty goes, Finland does better than Finnish-Americans do.
At least we would if we take your method of back-of-the-envelope calculation seriously. But since it yields different conclusions when apply it to Notten and de Neubourg and to Gornick and Jäntti, I’m not sure we should.
Without thinking very hard, I can think of a half dozen tweaks to these poverty measures that would change the figures slightly, so I don’t think this discussion is very meaningful. The basic point is that poverty rates for individuals of Scandinavian descent are similar, and very low, in different places and under quite different political-economic regimes, so it is unlikely that the regimes of different countries explain their different poverty rates. Rather, poverty rates are explained by the social and cultural characteristics of the population. I’m all in favor of people doing the hard work necessary to get the numbers as accurate as possible, but I don’t see anything in Matt’s discussion that changes the conclusion I have stated.
I don’t see anything in Matt’s discussion that changes the conclusion I have stated.
Wrong. The whole of my discussion, especially my most recent post above, changes the conclusion, in the sense that it shows that’s it’s false.
Huh? Just for example, you conclude that Danish-Americans had an Orshansky-poverty rate of 5%, whereas Danes had an Orshansky-poverty rate “well below 5%.” Both of those rates are “similar, and very low.”
If, say, Danish Orshansky poverty is 2.5%, and Danish-American Orshansky-poverty is 5%, then presumably, if Denmark adopted an American-style tax-and-transfer regime, Danish Orshansky poverty would increase by 200%. That would be bad, and I hope the Danes don’t do it.
Or, actually, it would increase by 100%. What I meant to say is it would double (I made the same mistake in my other recent comment below). Still bad.
I didn’t respond to your point about GDP per capita in your reply on the other thread, so I’ll do that now.
First, note that there is an apparent disagreement between the Sanandaji brothers on this point. In Unexceptionalism (62), Nima calculates that Swedish-American per capita GDP 39 percent higher than the per capita GDP of Sweden, the latter figure presumably including immigrants. According to Tino, Swedish-American per capita GDP is 50% higher than Swedish per capita GDP excluding immigrants. It would be odd if the per capita GDP gap between Swedish-Americans and ethnic Swedes in Sweden was wider than the gap between Swedish-Americans and everyone in Sweden.
Leaving that aside, we can look at what Nima says about Denmark and Danish-Americans, and Finland and Finnish-Americans (as Nima admits, Norwegian GDP per capita, including immigrants, appears to be slightly higher than Norwegian-American GDP per capita, presumably because of oil). Since Denmark and Finland have relatively few immigrants, this should be an easier comparison.
I’m also too lazy to check Tino’s math, but Nima’s math is pretty simple (Unexceptionalism, 62-63). He checks the Census Bureau’s data for median household incomes for Americans of various Nordic ancestries, and compares them to the median household income of the US overall. If the median household income of Nordic group N is x% higher than the US median household income, he calculates the per capita GDP of N-Americans as US per capita GDP*[1+(x/100)]. Thus, he is assuming that, for example, since Norwegian-American median household income is 17% higher than the US median, Norwegian-American’s “contribution to [US] GDP is also 17% higher…” (Unexceptionalism, 62). This strikes me as a non-obvious assumption, but leave that aside.
Using this methodology, he finds that Danish-Americans’ GDP per capita is 37 percent higher than Denmark’s, and Finnish-Americans’ GDP per capita is 47% higher than Finland’s.
Now, as Nima admits, this is inconclusive, because we don’t know about possible differences in household composition (Unexceptionalism, 63). Not taking household composition into account can lead one to think that a difference in per capita GDP represents a real difference in something we’re actually interested in, when it in fact doesn’t. As I pointed out in a comment on the other thread, Charles Murray falsely assumed that because GNP per capita (as well as total GNP) increased more quickly in the 1970s than in the 1960s, the 1970s US economy must have been healthier than the 1960s economy, at least in that respect. He forgot to account for the fact that 1970s per capita GNP rose more quickly simply because the proportion of workers to dependents increased in the 1970s.
But let’s assume that (with the exception of oil-rich Norway), the per capita GDP of the various American Nordic ethnic groups is higher than that of their respective ancestral countries. Let’s assume it’s 50% higher in each case. And let’s assume that this difference in GDP per capita reflects a real difference rather than just differences in household composition.
Who cares? In particular, why should social democrats (one of the principal dialectical targets of your previous post) care? As a social democrat myself, I already knew that, with the exception of oil-rich Norway, US GDP per capita was higher than the GDP per capita of the Scandinavian countries. I also could have guessed that Americans of Scandinavian ancestry have better-than-average socioeconomic outcomes, and probably have per capita GDPs that are higher than US per capita GDP. It follows from this that Scandinavian-Americans have higher per capita GDPs than those of their ancestral countries. So I didn’t learn anything too surprising from this.
An informed social democrat who prefers the Scandinavian model to the American model presumably already knows that the Nordic social democracies are inferior to the US in terms of per capita GDP, and she could probably infer that Americans of Scandinavian background have higher per capita GDP than their respective countries. Therefore, this social democrat must have preferred the Scandinavian model despite its inferiority in terms of GDP per capita, so I’m not sure why this news (if it is news) should change her mind, especially if what I say about poverty is correct.
People on the left tend to already have various criticisms of metrics like GDP per capita. Two prominent ones are that GDP per capita says nothing about the distribution of income, and it says very little, if anything, about poverty (absolute or relative). Since social democrats are much more concerned with income inequality and poverty than they are with GDP per capita itself, then, if what I say about poverty is true, the Sanandaji brothers’ estimates don’t give social democrats much reason to stop being social democrats.
In particular, Rawlsian social democrats are going to clearly prefer Scandinavia to the US, regardless of GDP per capita statistics. Imagine contemplating a choice between two societies, (A) and (B), in the original position behind a veil of ignorance. The two societies have the following features:
(A) Your ethnic group has a per capita GDP that is 50% higher than your ethnic group’s per capita GDP in society (B), but you are 200% more likely to live in poverty than you would be if you lived in (B).
(B) Your ethnic group’s per capita GDP is 50% smaller than it is in society (A), but you are only half as likely to live in poverty as you would be in (A).
The choice seems like a no-brainer to me, but in any case, a Rawlsian is certainly going to choose (B).
I don’t quite know what my theoretical distributive justice commitments are, but I can say I find Rawlsianism to be prima facie a lot more plausible than GDP-per-capita-consequentialism.
I should also note that Tino Sanandaji (in a paper I think I previously mistakenly attributed to Nima) explicitly claims that “the United States is more just than Western Europe” by Rawlsian criteria (T. Sanandaji 2012: 53). If what I’ve said about poverty is correct, then this is false at least with regard to Denmark and Finland, probably with regard to Sweden, and quite possibly with regard to Norway as well.
The citation is here.
This should read: “Your ethnic group’s per capita GDP is two-thirds of what it is in society (A)…”
For what it’s worth, Tino used the same method to impute GDP per capita from income data, so I don’t know why their estimate is inconsistent. But I’m not going to spend time trying to figure this out, since regardless of the reason for this discrepancy, the GDP per capita of people of Scandinavian ancestry is without a doubt higher than that of ethnic Scandinavians in Europe, as even you apparently concede. Similarly, the difference between the raw figure is so huge that I already know that, even if we made completely implausible assumptions about household composition and the GDP per capita of people with a foreign background in Scandinavia, this would remain true. Again, since you seem to concede that it’s true, I won’t bother doing the math.
Now, you say that since people of Scandinavian ancestry in the US are less likely to be poor than ethnic Scandinavians in Europe, the fact that the former have a higher GDP per capita than the latter should not convince a Rawlsian that the social-democratic model in Scandinavia is preferable to the American model. But as I have argued in my reply to you above, it’s not true that people of Scandinavian ancestry in the US are less likely to be poor than ethnic Scandinavians in Europe.
Moreover, even for a Rawlsian, the probability to be poor is presumably not the only thing that matters, when you consider how well-off the worst-off in society are. You also need to consider social mobility at the bottom or how fluid the group of the worst-off is and, as I plan to argue in another post at some point, it’s more fluid in the US than in Scandinavia, despite a widespread belief to the contrary which is the result of a conceptual confusion about social mobility. So I don’t think it’s true that even a Rawlsian should prefer the social-democratic model in Scandinavia to the American model.
With regard to poverty, your objection here is handled by my reply above. With regard to social mobility, you misunderstand the Rawlsian position. Rawlsians don’t care about social mobility as such, they care about fair equality of opportunity (FEO). A society could theoretically satisfy FEO perfectly, and yet have zero social mobility. This would occur, for instance, if socioeconomic status (SES) is substantially heritable (as it probably is), and a society has achieved FEO by eliminating all environmental sources of variation in SES. Then, there would be zero social mobility, because SES would be entirely a matter of genes, but FEO would be fully satisfied.
So, even if you are right that Scandinavia has lower social mobility than the US, this could be because Scandinavian societies have made substantially more progress in eliminating environmentally caused inequalities than the US has. If that were the case, opportunity would be more fairly equal in Scandinavia than it is in the US, even though social mobility is lower. Although, for a Rawlsian, neither the US nor Scandinavian countries fully satisfy FEO (no country does), what you’ve said does not rule out Scandinavia satisfying FEO to a greater degree than the US does.
Apparently there is some evidence of positive selection among Danish migrants. Tyler Cowen (no social democrat) refers to this paper, which states (p. 20) that artisans and craftsmen were overrepresented among late 19th century Danish emigrants, while unskilled laborers were underrepresented.