On Twitter Chris Mims expresses the entirely reasonable proposition that China’s low fertility is a prescription for long term socioeconomic disaster (they’re already beyond “peak worker”), with a link to an article in Quartz, China’s ratio of boys to girls is still dangerously high—and it’s the Chinese government’s fault. First, I’ve noted before that in East Asian societies where male preference was the norm this can shift very quickly. It happened to South Korea over the past 10 years, and it happened in Japan a generation ago. To my knowledge this was more of a matter of “bottom-up” cultural changes than government policy.
Second, the one child policy matters less than we think it does.
There’s another Census Projection out. Yes, I understand that the character of the children born today is going to have obvious impacts on the nature of the population 50 years from now, but we really need to heed the stupidity of past projections. Here’s a piece from 1930, A Nation of Elders in the Making:
To explain convincingly why we believe that we shall certainly not have more than 185,000,000 people here in 2000 A.D. and why we further believe that our population may cease to grow before that time, it is only necessary to make a rapid survey of our national trend of births and deaths….
I was at ASHG this week, so I’ve followed reactions to the election passively. But one thing I’ve seen is repeated commentary on the fact that Asian Americans have swung toward the Democrats over the past generation. The thing that pisses me off is that there is a very obvious low-hanging fruit sort of explanation out there, and I’m frankly sick and tired of reading people ramble on without any awareness of this reality. We spent the past few months talking about the power of polls, and quant data vs. qual (bullshit) analysis, with some of my readers going into full on let’s-see-if-Razib-is-moron-enough-to-swallow-this-crap mode.
In short, it’s religion. Barry Kosmin has documented that between 1990 and 2010 Asian Americans have become far less Christian, on average. Meanwhile, the Republican party has become far more Christian in terms of its identity. Do you really require more than two sentences to infer from this what the outcome will be in terms of how Asian Americans will vote?
Over at Econlog Bryan Caplan bets that India’s fertility will be sup-replacement within 20 years. My first inclination was to think that this was a totally easy call for Caplan to make. After all, much of southern India, and the northwest, is already sup-replacement. And then I realized that heterogeneity is a major issue. This is a big problem I see with political and social analysis. Large nations are social aggregations that are not always comparable to smaller nations (e.g., “Sweden has such incredible social metrics compared to the United States”; the appropriate analogy is the European Union as a whole).
There was a question below in regards to the high fertility of some extreme (“ultra”) religious groups, in particular Haredi Jews. The commenter correctly points out that these Jews utilize the Western welfare system to support large families. This is not limited to just Haredi Jews. The reason Somalis and Arabs have fertility ~3.5 in Helsinki, as opposed to ~1.5 as is the norm, is in part to due to the combination of pro-natalist subcultural norms, and a generous benefits state. Of course we mustn’t overemphasize economics. Israel’s decline in Arab Muslim fertility but rise in Jewish fertility in the 2000s has been hypothesized to be due to different responses to reductions in child subsidies by Muslims and the Haredi Jews. In short, the former reacted much more strongly to economic disincentives in relation to the latter.
A bigger question is whether exponential growth driven by ideology can continue indefinitely. I doubt it. Demographics is inevitable, but subject to a lot of qualifications. Haredi political power in Israel grants some benefits, but at the end of the day basic economics will serve as a check on the growth of the population of this sector. Similarly, barring massive productivity gains we’ll see some structural changes to the provision of government services across the aging developed world.
Below are some fertility numbers from the GSS. You see the median number of children for non-Hispanic whites born before 1960 from the year 2000 and later. I’ve compared the demographics of fundamentalists, non-fundamentalists, and those who are skeptical of the revealed nature of the Bible.
As someone with mild concerns about dysgenic (albeit, with a normative lens that high intelligence and good looks are positive heritable traits) trends, I’m quite heartened that Marissa Mayer is pregnant. Of course she’s batting well below the average of some of her sisters, but you take what you can get in the game of social statistics. Quality over quantitative thanks to assortative mating.
This brings me to a follow up of my post from yesterday, People wanted more children in 2000s, but had fewer. A reader was curious about limiting the data set to females. Therefore, I did. The same general pattern seems to apply (the limitations/constraints were the same). The only thing I’ll note is that there were only ~40 women in the data set with graduate degrees in the 1970s who were also asked these particular questions, so take this with a grain of salt.
Early this year I received an email from Dr. Peter Ralph, inquiring if I might discuss some interesting statistical genetic results from analyses of the POPRES data set which might have historical relevance. I’ve been excitingly waiting for the preprint to be made public so it could trigger some wider discussion. I believe that the methods outlined in the paper perhaps show us a path into the near future, where we might gain a much sharper perspective upon the recent past. So it’s finally out, and you can read it in full. Ralph and Dr. Graham Coop have posted put it up at arXiv, The geography of recent genetic ancestry across Europe. The paper uses ~500,000 SNPs from the POPRES data set individuals, and looks at patterns of identity by descent as a function of geography. By identity by descent, we’re talking about segments of the genome which are derived from a common ancestor. Because of recombination the length of the segments can give us a sense of the date of the last common ancestor; long segments indicate more recent ancestry because fewer recombination events have chopped up sequence.
Here’s the big takeaway of the paper: …There is substantial regional variation in the number of shared genetic ancestors: especially high numbers of common ancestors between many eastern populations likely date to the Slavic and/or Hunnic expansions, while much lower levels of common ancestry in the Italian and Iberian peninsulas may indicate weaker demographic effects of Germanic expansions into these areas and/or more stably structured populations. Recent shared ancestry in modern Europeans is ubiquitous, and clearly shows the impact of both small-scale migration and large historical events….
It is sometimes fashionable to assert that higher socioeconomic status whites are the sort who will impose integration on lower socioeconomic status whites, all the while sequestering themselves away. I assumed this was a rough reflection of reality. But after looking at the General Social Survey I am not sure that this chestnut of cynical wisdom has a basis in fact. Below are the proportions of non-Hispanic whites who have had a black friend or acquaintance over for dinner recently by educational attainment:
35% – Less than high school
36% – High school
47% – Junior College
45% – Bachelor
59% – Graduate
I thought this might have been a fluke, so I played around with the GSS’s multiple regression feature, using a logistic model. To my surprise socioeconomic status was positively associated with having a black person over for dinner, and age negatively associated. These two variables in fact tended to exhibit equal magnitude values in opposition, and always remained statistically significant. Just to clear, I created a variable Non-South vs. South below (being Southern increases likelihood of having had a black person over for dinner). All the individuals surveyed are non-Hispanic whites for the year 2000 and later. You can add and remove variables, but SEI and age tend to be rather stable, and statistically significant, throughout.
A few years ago I put up a post, WORDSUM & IQ & the correlation, as a “reference” post. Basically if anyone objected to using WORDSUM, a variable in the General Social Survey, then I would point to that post and observe that the correlation between WORDSUM and general intelligence is 0.71. That makes sense, since WORDSUM is a vocabulary test, and verbal fluency is well correlated with intelligence.
But I realized over the years I’ve posted many posts using the GSS and WORDSUM, but never explicitly laid out the distribution of WORDSUM scores, which range from 0 (0 out of 10) to 10 (10 out of 10). I’ve used categories like “stupid, interval 0-4,” but often only mentioned the percentiles in the comments after prompting from a reader. This post is to fix that problem forever, and will serve as a reference for the future.
First, please keep in mind that I limited the sample to the year 2000 and later. The N is ~7,000, but far lower for some of variables crossed. Therefore, I invite you to replicate my results. After the charts I will list all the variables, so if you care you should be able to replicate displaying all the sample sizes in ~10 minutes. I am also going to attach a csv file with the raw table data. As for the charts, they are simple.
- The x-axis is a WORDSUM category, ranging from 0 to 10
- The y-axis is the percent of a given demographic class who received that score. I’ve labelled some of them where the chart doesn’t get too busy
All of the charts have a line which represents the total population in the sample (“All”).
Update: There was a major coding error. I’ve rerun the analysis. No qualitative change.
As is often the case a 10 minute post using the General Social Survey is getting a lot of attention. Apparently circa 1997 web interfaces are so intimidating to people that extracting a little data goes a long way. Instead of talking and commenting I thought as an exercise I would go further, and also be precise about my methodology so that people could replicate it (hint: this is a chance for readers to follow up and figure something out on their own, instead of tossing out an opinion I don’t care about).
I’ve never watched Mad Men, but I really can’t help but hear all about the show. One thing that has struck me about the change from then, ~1960, to now, ~2010, is the alignment of quantitative demographic trends with impressionistic cultural ones. The 1970s were a disaster for the old urban order. Below are the top 10 cities by population in 1960 and 2010.
|1||New York||New York|
|10||St. Louis||San Jose|
The rise of the “Sun Belt”, housing bubble notwithstanding, is a real and awesome phenomenon. Below the fold I’ve taken some demographic trend data for the top 10 cities of 1960. The first two panels show raw population data. The second two panels show the decade-to-decade change in population in terms of multiples (i.e., 1.2 for 2010 means that the population in 2010 was 1.2 times that in 2000).
Prompted by Andrea Mitchell’s complaint that Iowa is not representative of America in racial terms the Audacious Epigone probed an American state’s typicality in terms of racial demographics, using the overall American population as a measure. One of the major issues with judging the typicality of a given state is that there is a great deal of residential segregation in even “diverse” regions. This comes up in our personal choices too. In 2008 ~10 percent of non-Hispanic whites married someone who was not a non-Hispanic white. Obviously more than ~10 percent of the population, particularly in the prime marrying demographic, are non-Hispanic whites, so you’re seeing a fair amount of homogamy. In some ways the homogamy is even more striking for minorities. ~31 percent of Asian Americans in this period married a non-Asian American. But, one has to keep in mind that using the American population as representative over 90 percent of the potential marriage partners are not Asian American!
The quest for a state that “looks like America” is understandable, but the reality of lived life is more complex. And not just in racial terms (e.g., the division in politics between the white suburbs of Maryland vs. Virginia on either side of D.C.). But keeping race in mind, one consistent finding in social science is that Americans actually tend to overestimate the number of minorities. Iowa is actually more typical than we think, despite the fact that it is not typical. In the year 2000 the General Social Survey asked respondents to estimate the number of various groups in the USA. The finding of a tendency to overestimate minorities, and underestimate non-Hispanic whites, was confirmed. But, I decided to break this down by demographic. The results are below in a table.
The first row are real counts from the 2000 Census. All the following rows are average estimates of a set of respondents in the year 2000.
Mike the Mad Biologist has a post up, A Modest Proposal: Alabama Whites Are Genetically Inferior to Massachusetts Whites (FOR REALZ!). The post is obviously tongue-in-cheek, but it’s actually an interesting question: what’s the difference between whites in various regions of the United States? I’ve looked at this before, but I thought I’d revisit it for new readers.
First, I use the General Social Survey. Second, I use the WORDSUM variable, a 10 question vocabulary test which has a correlation of 0.70 with general intelligence. My curiosity is about differences across white ethnic groups by region. To do this I use the ETHNIC variable, which asks respondents where their ancestors came from by nation. I omitted some nations because of small sample size, and amalgamated others.
Here are my amalgamations:
German = Austria, Germany, Switzerland
French = French Canada, France
Eastern Europe = Lithuania, Poland, Hungary, Yugoslavia, Russia, Czechaslovakia (many were asked before 1992), Romania
Scandinavian = Denmark, Norway, Sweden, Finland (yes, I know that Finland is not part of Scandinavia, Jaakkeli!)
British = England, Wales, Scotland
Northeast = New England, Middle Atlantic
Midwest = E North Central, W North Central
South = W S Central, E S Central, South Atlantic
West = Pacific, Mountain
The key method I used is to look for mean vocabulary test scores by ethnicity and religion. I also later broke down some of these ethnic groups by religion. Finally, all bar plots have 95 percent confidence intervals. This should give you a sense of the sample sizes for each combination.
First let’s break it down by race/ethnicity and compare it by region to get a reference:
There’s a variable in the GSS, GENESELF, which asks:
Today, tests are being developed that make it possible to detect serious genetic defects before a baby is born. But so far, it is impossible either to treat or to correct most of them. If (you/your partner) were pregnant, would you want (her) to have a test to find out if the baby has any serious genetic defects?
This is relevant today especially. First, the technology is getting better and better. Second, couples are waiting longer to start families. Unfortunately this question was only asked in 1990, 1996, and 2004. But on the positive side the sample sizes were large.
I decided to combine 1990 and 1996 into one class. Also, I combined those who were very liberal with liberals, and did the same for conservatives. For political party ideology I lumped strong to weak identifiers. For intelligence I used WORDSUM. 0-4 were “dull,” 5-7 “average,” and 8-10 “smart.” For some variables there weren’t results for the 1990s.
The biggest surprise for me is that there wasn’t much difference between the 1990s and 2004. The second biggest surprise was that the differences between demographics were somewhat smaller than I’d expected, and often nonexistent. Below is a barplot and table with the results.
Fascinating story about the re-identification of people of Eurasian ancestry as white to get into elite universities. Some Asians’ college strategy: Don’t check ‘Asian’:
Lanya Olmstead was born in Florida to a mother who immigrated from Taiwan and an American father of Norwegian ancestry. Ethnically, she considers herself half Taiwanese and half Norwegian. But when applying to Harvard, Olmstead checked only one box for her race: white.
Asian students have higher average SAT scores than any other group, including whites. A study by Princeton sociologist Thomas Espenshade examined applicants to top colleges from 1997, when the maximum SAT score was 1600 (today it’s 2400). Espenshade found that Asian-Americans needed a 1550 SAT to have an equal chance of getting into an elite college as white students with a 1410 or black students with an 1100.
In the article Steve Hsu observes that the Ivy League universities have a suspiciously similar proportion of Asians, about 2/3 of the fraction of a “race blind” admissions college like Cal Tech. Here’s Alex Tabarrok with the numbers: “At Yale the class of 2013 is 15.5 percent Asian-American, at Dartmouth 16.1 percent, at Harvard 19.1 percent, and at Princeton 17.6 percent.” I assume that the “Asian Quota” will start to change as the current generation of Asian American students become established as alumni donors.
I’m not a big fan of the “Asian Quota” personally. But, I do think one can make a case for it based on the fact that children from families with an Asian background have a strong bias toward optimizing measured outcomes. But, this entails making a profile, or “stereotype,” of a population. I’m not someone who actually objects to this on principle, but I find the hypocrisy on this issue rather annoying, because the same administrators who would decry stereotypes feel they have to employ them implicitly for practical (so the alumni don’t see their university overwhelmed by “yellow hordes,” and so reduce giving) and idealistic reasons (to maintain some ethnic balance).
COMMENTS NOTE: Any comment which misrepresents the material in this post will result in banning without warning. So you should probably stick to direct quotes in lieu of reformulations of what you perceive to be my intent in your own words. For example, if you start a sentence with “so what you’re trying to say….”, you’re probably going to get banned. I said what I tried or wanted to say in the post. Period.
The New York Times has a article out about environmentalists who are now looking at population control again, after shying away from it. This is probably prompted by the hullabaloo over “7 billion.” This comes in the wake of a long piece, The Last Taboo, in the Lefty periodical Mother Jones.
The rationale for why environmentalists have moved away from population control is alluded to only elliptically in The New York Times piece. They make a big deal about abortion, but I don’t think this is the most terrifying issue in principle. Environmentalists tend to be on the pro-choice side of the Culture Wars anyway. To cut to the chase it is the fear of being called racist (and to be fair, racial nationalists from Madison Grant to John Tanton have synthesized ethnic concerns with genuine conservationist impulses). Only environmentalists with rock-solid credentials or a lean toward anti-humanistic Deep Ecology philosophies remained vocal about their opposition to mass immigration over the past few decades. David Brower, founder of the Sierra Club, was one such individual. And it’s no surprise that the founder of the radical Earth First movement is also an immigration restrictionist.
I wanted to clarify a few issues with the Census’ American Community Survey. These data come from the interval of 2006-2008, and they allowed me to query the proportional of various Latino/Hispanic groups who identified as white. I knew in the aggregate that the majority of America’s Latinos identified as white, but I was curious about two things:
1) The variation in white identification by group (by national origin)
2) The variation in white identification of Mexican Americans by selected states
Results below. There are stories in these data….
One of the things I really hate are unqualified linear projections. They’re so useless most of the time. A science fiction magazine will give you more insight about the future than the United Nations population projection for the year 2100. This is just as much of an issue when it comes to American Census demographic projections. As I’ve noted before population projections of the coming non-Hispanic white minority 2040 to 2050 are sensitive to the assumptions behind the basic parameters. The logic of the projection is crystal clear and airtight, but just because a certain set of assumptions holds today, does not mean that those assumptions will hold indefinitely (though the Census projections are much more plausible than the United Nations projections because two generations are so much more strongly impacted by by the inertia of current conditions that four generations). In the 18th and 19th century white Americans, and especially the Anglo-Saxon founding stock, were a highly fertile folk. They took over the American Southwest and the Northwest in large part due to their demographic assault. In New England the 30,000 of 1650 became the 700,000 in 1790 in large part due to fertility rates on the order of 7 per woman! Today no one would expect that Anglo-Saxon Americans would be so fertile, let alone the New Englanders who were prominent in the population control movements of the 20th century. In the 17th and 18th century the Jews of Eastern Europe were a highly prolific group, and the gentile majority in places like Poland viewed the waxing of the proportion of this minority with great suspicion. Today no one views the Ashkenazi Jews as demographic engines, though in places like Israel the fecund Haredi have now helped close the “birth gap” with the Arab population, as its fraction of the Jewish population keeps increasing. I can give you other “counter-intuitive” examples from the recent past, but a little history goes a long way in teaching suspicion (e.g., in the Balkans in the late 19th century rural Christian populations had much higher fertility than urban Muslim ones).
These sorts of reversals are not inexplicable. Fertility shifts occur, sometimes within a generation or two. This is why Thomas Malthus turned out to be wrong: he didn’t predict the demographic transition. But we shouldn’t be complacent and assume we’ve reached the “end of history” when it comes to fertility transitions. In the early 20th century there was great terror in the American elite due to the immigration of what would later be termed “ethnic whites,” in particular Jews and Southern Europeans. And yet the Jewish proportion of the American population peaked in the late 1940s at ~5%. What about the other groups? The General Social Survey has large sample sizes for some ethnic groups, so I decided to look there.
In my long post below, Celts to Anglo-Saxons, in light of updated assumptions, I had a “cartoon” demographic model in mind which I attempted to sketch out in words. But sometimes prose isn’t the best in terms of precision, and almost always lacks in economy.
In particular I wanted to emphasize how genes and memes may transmit differently, and, the importance of the steps of going between A to Z in determining the shape of things in the end state. To illustrate more clearly what I have in mind I thought it might be useful to put up a post with my cartoon model in charts and figures.
First, you start out with a large “source” population and a smaller “target” population. Genetically only the migration from the source to the target really has an effect, because the source is so huge that migration from the target is irrelevant. So we’ll be focusing on the impact upon the target of migration both genetically and culturally.
To simplify the model we’ll imagine a character, whether genetic or memetic, where the source and target are absolutely different at t = 0, or generation 1. Also, these are discrete generations, and the population is fixed, so you can presume that it’s at carrying capacity. Migration of the outsiders into the target population from the source means less of the original native population in absolute terms (to be realistic this is bidirectional, so people are leaving the target too, but that’s not our concern here).
There are two time series which illustrate the divergent dynamics on both the genetic and memetic dimensions. In one series you see gradual and continuous migration from the source to the target population over 13 generations. In another there are two generations of massive migration, before and after which there is no migration. For the genetic character, imagine disjoint allele frequencies at generation 1. So at generation 1 the target population is at 100% for allele A, while the source is at 100% for allele B. Therefore migration of from the source to the target results in a decrease in the proportion of allele A, which is what is being measured on the y-axis. For the memetic character, imagine that it’s language. So at generation 1 100% in the target zone speak language A, while everyone in the source zone speak language B. Again, the frequency on the y-axis is of the proportion who speak language A in the target zone.