Is skin tone correlated with job prestige?

The General Social Survey rated the skin darkness of a sample of black Americans, ranging from "very dark brown" to "very light brown." Biologically oriented researchers might see the question as a rough measure of the percentage of European ancestry, while sociologists would see it as a measure of discrimination--lighter-skinned blacks getting better treatment. Respondents were also given a job prestige score that ranges from 16 to 80--16 is a shoeshine and 80 is a physician. Here are the job prestige means listed by skin tone:

You can see that average job prestige tends to rise with lighter skin. The mean for blacks with "very light brown" skin is roughly two-thirds of a standard deviation higher than the mean for "very dark skin" blacks. 

The pattern can be interpreted in at least two ways: 1) genetic--blacks with more European ancestry tend to rise in the status hierarchy much more than African blacks, or 2) sociological--whites discriminate more against darker blacks, and perhaps light-skinned blacks have white (privileged) relatives who gave them advantages. 

For several reasons, I'm inclined toward the genetic explanation. For one, my experience is that when a white person is interacting with a black person, his thought is, "I'm talking with a black person," not, "I'm talking with a light-skinned black person." For another, how do those Nigerian immigrants do so well in the US when their skin tends to be so dark? The sociologist would predict severe discrimination. The biologist would argue that African immigrants are a select group of Africans--above average in IQ and drive--and this overwhelms any bias they might experience. 

Do blacks and whites who grow up equally wealthy end up with the same mean IQ?

Some people argue that mean IQs for whites and blacks differ because blacks tend to come from poorer families. A black kid raised with resources equal to a white kid will have the same IQ. Is this the case? 

Using data from the General Social Survey, I calculated mean IQ for blacks, whites, and others for several ranges of father's socioeconomic index (SES): 1-2 standard deviations (sd) below average, 1 sd below average to average, average to 1 sd above average, 1-2 sds above average, and over 2 sds. The sample size is 12,016 (immigrants were excluded). Here's a graph that summarizes the results: 

Blacks and whites raised at the same SES level do not have the same mean IQs. The higher white average ranges from about 7 points among the poorest group to more than 10 points among the wealthiest group. By the way, other non-whites--mostly Hispanics--are closer to blacks than whites. 

Who is more likely to be murdered by a black offender: a white or an Asian?

 Wilfred Reilly (@wil_da_beast630) has a discussion going on at Twitter about which race currently faces the most "systemic racism." Some people are assuming that Asians are victimized by blacks more than whites are. Two sources of data are not very helpful here: hate crimes and victimization data from the National Crime Victimization Survey. Hate crimes are a tiny drop in the bucket of all violence, and you don't see NCVS tables with Asian victims by race of offender. (Let me know if I've missed them.) Victim data in general indicates that Asians face low levels of victimization, which makes sense because most crime is intra-racial, and Asians have very low rates of criminal offending. 

Perhaps homicide data is the best way to go here. I looked at expanded FBI homicide data and the Census to calculate rates for being murdered by a black person for two groups: 1) Whites/Hispanics, and 2) Asians/ Native Americans/Pacific Islanders. I'm afraid the FBI lumps people together that way. The rate for group 1 is 2.30 murders per 1 million population. For group 2, it's 1.44 murders per million. According to FBI data, blacks pose more of a threat to whites and Hispanics than Asians plus. 

Which low IQ groups have low arrest rates?

DGo (@Go321) on Twitter wondered which lower IQ groups still manage to be well-behaved as groups. Criminality is a decent measure of behavior, so I looked at General Social Survey data to see which groups with low average IQs also have lower than average arrest rates: 

The percent arrested for the whole country is 12.7%. The groups listed above with an arrest rate lower than that are: Filipinos, West Indians, and Non-Spanish West Indians. And let me add that these three groups have IQ means that are not very low: 96.7, 98.8, and 95.9, respectively. 

Most lower-IQ groups have high arrest rates, as expected.

UPDATE: After checking, I see that Arabs do not have a low mean IQ: it's 102.   

It's a clean sweep: IQ is more predictive of education, income, and job prestige than dad's social class

 Someone at Twitter, I forget now, wondered if IQ or one's social class was more important for adult success. Well, the General Social Survey can help with this. I threw in basic demographics as controls. 

Here are OLS results for income:

Looking at the betas, you can see that IQ is more strongly predictive of income than father's socioeconomic status (PASEI). Notice how race is not statistically significant when IQ is included in the model. 

And job prestige? 

IQ is much more predictive of job prestige than father's PASEI. 

The results for education should be even stronger for IQ:

 

The beta for IQ is much larger than for dad's social class. How far one goes in school depends much more on brains than dad's wallet (or his other influences). 

An updated analysis of the 2020 election: CPS data

In December, I posted data that make me skeptical about the integrity of the 2020 election. This is an update with new data.

In the December analysis, I reported that 158.4 million voted for US President but based on an estimate derived from the Current Population Survey (CPS), only 158.1 million were registered. Impossible. The CPS didn't have current data in December. Now they do, and they report that 168.3 million were registered in November 2020. This stunned me since the CPS estimated only 153.1 million in November 2018, so total registrations increased 15.2 million in only two years, 2018-20. For context, here's the total number registered by year since 1968:

Total number registered to vote (in millions)

1968       86.6

1972       98.5

1976       97.8

1980     105.1

1984     116.1

1988     118.6

1992     126.6

1996     127.7

2000     129.5

2004     142.1

2008     146.3

2012     153.2

2016     157.6

2018     153.1

2020     168.3  

The biggest four-year increase was 2000-2004 with 12.6 million additional registrations. Such a big increase makes the 10.7 million 2016-2020 plausible, but it's weird that the total went from 157.6 million in 2016 down to 153.1 million in 2018, then it increased 15.2 million in only two years, a bigger increase than any 4-year increase since 1968! Weird but possible. By the way, the margin of error for the CPS is less than a million.  

168.3 million registered people in November 2020 means 94.1% of registered voters voted for President. Election experts say that 90-95% of registered voter voting is indicative of election fraud. Countries like Australia that mandate voting can't get 94% of registered voters to vote. 

For context, here are the percentage of registered voters who voted going back to 1960:

Percent of registered voters who voted

1960   107.8

1964     95.1

1968     89.4

1972     79.8

1976     77.7

1980     76.5

1984     74.6

1988     74.5

1992     71.2

1996     65.9

2000     67.5

2004     70.0

2008     89.8

2012     84.3

2016     86.8

2020     94.1

The years 1960 and 1964 look screwy, but the only other year that looks suspicious is 2020. I'm still not convinced that there was not enough funny business to change the election outcome. 

By the way, according to CPS, 154.6 million people said they voted. According to the official count, 3.8 million additional votes were cast. 

What predicts virginity among young American men?

 According to the General Social Survey, 28% of American men under 30 report not having sex since they turned 18. I was baffled by that and wanted to know what predicts virginity among young men. 

Using General Social Survey data, I looked at a long list of possible predictors but found that most factors are unrelated. As you can see below, being younger predicts virginity (not surprisingly). More educated men are more likely to be virgins which cuts against the claim that high-status men are the ones who get sex. (Income and job prestige were not predictive.) 

Never going to bars and not watching an X-rated film in the past year predicted virginity. So porn doesn't look like a substitute for sex: sexually active men are more likely to watch porn. Perhaps celibate men tend to have lower levels of testosterone? Men who go to bars are more likely to have sex, so men who make an effort or who are more social are more likely to have sex. 

The GSS doesn't have many personality-type questions, so I wasn't able to look at traits like introversion or social awkwardness, but if you've got other ideas for unconventional variables, let me know. 

Which white groups have noticeably low IQs?

To answer this question, I calculated mean IQs for all self-described whites by the country they said their ancestors came from. Only people born in the US are included since immigrants are at a disadvantage on a vocabulary test (the measure of IQ used here). Here are the results:

Mean IQ

Africa  93.4
Mexico  90.0
Puerto Rico  93.2
Spain  97.2
American Indian  94.7
India  95.8
Other Spanish  95.9
American only  90.4

It looks like many of these people are racially mixed. They label themselves as white but often have some non-white ancestry. The mixed groups that do not have low mean IQs are self-described whites whose ancestors came from east Asian countries--Japan, China, or the Philippines. 

UPDATE: The mean IQ for self-described whites whose ancestors are from east Asia is 99.8. 

 

Is Douthat right that Rush ruined the GOP?

Ross Douthat wrote in the NYT recently that the declining appeal of the GOP is the fault of Rush Limbaugh. He claims that Republicans were a 55% proposition during the Reagan Era because the message of conservatism was preached by well-bred men like William F. Buckley. Rush came along and turned conservatism into something angry and strident and that only appeals to 45% of voters. 

If Ross is correct, we should see greater support for Reagan among racial minorities--the groups we have a hard time winning over. After all, Buckley was a leading advocate for conservatism during that time and Rush was unknown.

Using the General Social Survey, we can compare minority deficits over time. I calculate this as the percent of a racial minority voting for the GOP presidential candidate minus the percent of all voters who cast a vote for the Republican candidate.

Minority deficit

1980
Blacks  -39.2
Mexicans  -18.2

1984
Blacks  -48.3
Mexicans  -19.5

Damn, impressive: under Buckley and Reagan sophistication, Blacks and Mex-Ams supported the GOP by only 18 to 48 points less than the general population! The power of positivity!

And since then? Bush Sr. wasn't strident. Perhaps he did better.

1988
Blacks  -41.9
Mexicans  -26.2

1992
Blacks  -34.6
Mexicans  -18.1

1996
Blacks  -26.2
Mexicans  -12.1

The numbers are slightly less bad with grumpy Dole. 

2000
Blacks  -42.5
Mexicans  -6.5

It looks like angry Rush didn't chase away Mex-Ams in 2000.

2004
Blacks  -37.6
Mexicans  -11.7

2008
Blacks  -34.7
Mexicans  -15.3

2012
Blacks  -33.3
Mexicans  -17.8

2016
Blacks  -38.2
Mexicans  -19.7

The Republicans did no better among minorities before strident Rush became influential in the early 1990s. 

Douthat ignores the obvious: the GOP is struggling because there were fewer than 20 million Hispanics during the Reagan Era. In 2021, there are more than 60 million--and this perennially poor group likes generous social programs and is susceptible to anti-white propaganda like all non-white groups. And probably will be for a very long time. 

Which religion has the most forgiving people?

The General Social Survey asked respondents if it is true that: "I sometimes try to get even rather than forgive and forget." Below you see the answers by religious affiliation. 

All groups have higher rates of vengefulness than Christians, including the "nones." This makes sense since Christianity stresses forgiveness.

Among Christians, those that attend church all the time are the most forgiving. 

And nowadays many of the people with no religion like to accuse devout Christians of being the haters.  

What is the mean IQ of whites who say their ethnicity is "American only"?

Based on the General Social Survey, if you ask Americans where their families originally came from, they will typically give you an answer. But about 1% of whites and 13% of blacks will tell you they are "American only." 54% of whites who give this answer live in the South, and Americans of Scots-Irish ancestry are known to sometimes give this answer. By contrast, "American only" blacks are concentrated in the mid-Atlantic and East North Central regions. 

And their mean IQs? 

 

  

Wow, it's around 90 for whites. That is low. 

I wouldn't generalize these findings to the Scots-Irish. Many of the whites, and perhaps some from other races, who say "American only" are probably from lower IQ families who simply have not kept track of where their people are from. Or lower IQ people might be more likely to prioritize their American identity. 

Are French Americans an underprivileged minority?

 In "The Son Also Rises," Gregory Clark presents historical data that suggests that French Americans are an underprivileged minority. They have been underrepresented among doctors and lawyers, for example. 

Do General Social Survey data support this view? 

Here are mean years of schooling for the French versus other white Americans:

No difference in education. Here's median income in 1986 dollars:

A lower median income for Americans of French descent. And mean job prestige:

The French have a slightly higher mean job prestige. And IQ?

Basically the same mean IQ.

Clark predicted to the French would eventually move to average levels. It looks like they've made it.  

Are Christians repressed and thus unhappy?

A common claim by liberals and anti-Christians is that Christians are unhappy because the religion is repressive. 

The General Social Survey Survey asks people how happy they are overall these days with answers ranging from not too happy (1) to pretty happy (2) and very happy (3). Here are the means for various Christians and those with no religion:

All five Christian groups have higher happiness averages than the "nones." 

What about devout Christians? Maybe they are the miserable ones.

Mean happiness rises with more frequent "repression" (church attendance). 

The research literature in general reports that religiosity is associated with a wide range of positive characteristics, but that doesn't stop smug, ignorant anti-Christians from going on about how religion is harmful. 

Has the mean IQ of American Jews fallen because of intermarriage?

Someone claimed on Twitter that the mean IQ of American Jews has been falling due to intermarriage. What do General Social Survey data have to say? Here are mean IQs by decade since the 1970s:

No evidence of a decrease here. 

GSS data: Do Jews have prestigious jobs because they are smart, or are there other factors?

 Another IQ issue arose on Twitter today: Does IQ explain the disproportionate number of Jews in high positions in the new Biden administration? Or is there some other factor? The General Social Survey can't answer this specific question, but we can look at job prestige in general. The table below shows that at the same IQ level, Jews tend to have noticeably higher average prestige scores than non-Jews. 

So, according to GSS data, Jews tend to have characteristics beyond IQ that boost their job prestige. What would those be? I'd be guessing, but in the case of Biden, it might not be a coincidence that most Jews are Democrats and are a key presence in the Democrat Party. My impression is that Jews are very achievement-oriented and are drawn to centers of power. In addition to that, I suspect that a Jew is also likely to enjoy the benefit of being born into a social network of high-achieving people. Connections matter.  

Does education increase IQ? The General Social Survey says no

The question was raised on Twitter the other day about whether more education causes increases in IQ. The General Social Survey can help a little with this. Below we have a chart of the mean years of schooling completed since 1977. The mean starts out at 11.71 and ends in 2018 at 13.89 years of education. Over the period, that is an increase of 19%. 

By contrast, here is the trend in IQ. In 1978, it was 97.45, and in 2018 it was 97.80--an increase of 0.3%. Basically, no change. US trends are consistent with the view that keeping people in school longer does not make them smarter. 

Should we trust the voter rolls? Hell no, eight states have more registered voters than people eligible to vote

In the previous post, I presented evidence that more people voted in 2020 than the total number of registered voters. One criticism is that we should not rely on surveys but government data to estimate the total number of registered voters. 

Using this data on state counts of registered voters and this website that calculates the total number of people in a state who are eligible to register, I constructed the table shown below that displays the number of registered voters as a percent of those eligible to vote. 

You can see that eight states have more registered voters than people eligible to vote, clear evidence that the voter rolls are wildly inaccurate. Another ten states have 95-99% of all eligible people registered which seems highly implausible. Does it make sense that 19 out of 20 eligible people are currently registered? Many states have a serious problem.

As we move down the table, the numbers get more and more plausible. Eight out of ten of all eligible people currently registered? Okay, maybe, but only four states are below that cutoff. 

It seems clear that the rolls in many of these states are larded up with outdated registrations, and the Current Population Survey (CPS) is a more trustworthy way to get an estimate of current registrations. And according to my analysis with CPS data, more people voted in 2020 than the total number of registered voters. 

Analyzing the Current Population Survey, more than 100% of registered voters voted this year

Using the Current Population Survey (CPS), there is evidence of widespread 2020 election fraud. The CPS is a monthly survey of around 60,000 U.S. households that is conducted by the Census Bureau. The survey is used by the Bureau of Labor Statistics to estimate the unemployment rate, and the Census uses it to estimate population characteristics in between decennial censuses. In other words, it's a good survey.

The CPS asks people if they are currently registered to vote. Let's use its estimates to calculate the percent of registered voters who voted in presidential elections since 1960, the first year CPS asked a question about voter registration. I'll use the numbers provided here.

Percent of registered voters who voted

1960   107.8

1964     95.1

1968     89.4

1972     79.8

1976     77.7

1980     76.5

1984     74.6

1988     74.5

1992     71.2

1996     65.9

2000     67.5

2004     70.0

2008     89.8

2012     84.3

2016     86.8

While most years seem perfectly reasonable, notice how the 1960 election--the election stolen from Nixon--yields an impossible number. 95.1% voting in 1964 seems fishy, too. Election experts say that over 90% is a red flag for fraud. 

CPS has not provided 2020 numbers yet. The most recent is 2018, and it estimated 153.1 million registered voters with a margin of error at about 750,000.  If we average increases in registered voters since 1980. that is 1,027,000 per year, or 2,054,000 for two years. Perhaps 2018-2020 saw a much better than average increase. Let's be generous and say 5 million new people got registered during that time. That puts us at 158.1 million registered in 2020. The problem is, 158.4 million voted for president in 2020 according to Ballotpedia. That gives us an estimate of 100.2% of registered voters voting for president in 2020. Impossible.

A response to this is that we can simply use state counts of registered voters, and the World Population Review totals those to be 214 million. The problem with using that approach is that voter rolls are notoriously larded up with outdated or improper registrations: dead people (who ever calls the officials to let them know Granny just died), people who have moved away, etc. There are many sources of error. For example, recently in California, non-citizens were being registered to vote when they got their driver's licenses. Government seems curiously lax on cleaning up voter rolls. 

The CPS is considered a gold standard of surveys, and you shouldn't have undercounts due to people legitimately being registered but being shy about admitting it. If anything, it might be like responses about voting where people want to look good and say they voted when they did not. Registering to vote is what good citizens do.  

The 1960-2016 CPS results shown above seem valid with the clear exception of 1960, the first year the CPS asked about voter registration, and the year when an election was stolen. I'm sure methodological refinements have been made over the past 60 years. 

For 2020 to have the voting rate we saw in 2016--86.8% of those registered voting--there should be 182.5 million people currently registered. The CPS will be off by a little but not almost 25 million voters. All this suggests funny business in 2020. 

UPDATE: I show here that voter rolls are unreliable for this analysis.

Why do people from large families earn less income?

 According to conventional economic theory, growing up in a large family predicts less income as an adult because parents were unable to invest as much in each child. By contrast, genetic theory would predict that that family size would not matter for how much income you earn as an adult; rather, income would be predicted by one's IQ. Let's test these two competing hypotheses using General Social Survey data. 

Respondents were asked how much income they earned in the past year, and they were also given a ten-word vocabulary quiz, which makes a decent proxy for IQ (N = 19,902).

This table shows the estimates for a regression model that includes personal income in constant dollars as the dependent variable and the number of siblings as the predictor. You can see that each additional sibling results in a predicted reduction in one's income of $1,139.  (I believe these are 1986 dollars.)

Looking at this table above, we can see that IQ is positively related to income, and the beta indicates that the relationship is of considerable magnitude. It is predicted that each additional IQ point will result in an additional $423 in income. 

The unstandardized coefficient for number of siblings has dropped from $1,139 down to only $586. In other words, much of the reason why a large family predicts a smaller income is due to the correlation between having many siblings and having a lower IQ. This finding supports, to some extent, the genetic hypothesis. On the other hand, we see that the sibling coefficient is still statistically significant, so even after controlling the influence of IQ, the number of siblings is still negatively correlated with income. The economic hypothesis appears to have something to it. 

Race trumps: Race, not social class, predicts 2016 voting

You often see the argument that the central political divide is social class, not race and ethnicity. If this were true, then we should see it in voting patterns. 

The General Social Survey asked participants who they voted for in 2016, and they also asked about annual income and race. The results below are estimates from a logistic regression model that predicts 2016 voting for President with income and race. The three racial categories are white, black, and other race. Whites are the reference category and so are omitted from the model (sample size = 1,360).

From the p-values (probability) you can see that, once you adjust for race, REALINC (inflation-adjusted income) does not significantly predict who you voted for in 2016. By contrast, blacks and people of other non-white races were less likely than whites to vote for Trump, regardless of one's social class. 

Race trumps.