January 25, 2017

Changes in the American Black-White IQ Gap: 1916-2016

In this article, I am going to examine how the Black/White IQ gap has changed over time in the United States. After documenting basic trends, I’ll look at what caused the B/W IQ gap to shrink in the late 20th century. As will be seen, simplistic explanations based on changes in wealth and education are inadequate. We don’t know for sure why this convergence of test scores occurred, but desegregation is the most plausible hypothesis I have been able to come up with. We’ll also see that this narrowing of the IQ gap may have reflected a narrowing in test-taking ability rather than genuine intelligence. Finally, I will consider what implications this data has on the causes of racial intelligence disparities.

The First Fifty Years

IQ tests began to be widely used in the 1910s. Until the 1960s, IQ tests were not given to large and representative samples of the population and so estimating the Black/White IQ gap for this time period is less than entirely straightforward.

A vital resource on this topic is Audrey Shuey’s 1966 book “The Testing of Negro Intelligence“. In it, Shuey compiled and analyzed 382 studies which measured the Black/White IQ gap between the 1910s and 1965.

School Data

Within Shuey’s data, the closest we can get to representative samples of the national population comes from studies of school children. However, unlike today, a large fraction of people, especially Blacks, did not go to school during this time. Because poor and uneducated Black families were especially unlikely to send their children to school, the effect of this is to make studies under-estimate the true B/W IQ gap of the time.

Because the Black/White gap in highschool attendance was larger than the Black/White gap in elementary school attendance, this problem becomes more severe the older school children are.

Because Black/White gaps in educational attainment fell with time, this problem lessened as the years went on. This bias will therefore produce an increase in the B/W gap over time even if there actually was none among the nation as a whole.

That being said, here is Shuey’s data taken from 259 studies of preschool, elementary school, and high school, children tested between 1921 and 1965 (1).

Students Tested Before 1965.png

Based on this data, we can conclude that the B/W IQ gap was at least .84 SD in the first period, and at least 1.04 SD in the second period.

The B/W IQ gap in this data set increased over time. However, to some degree, this just reflects the fact that a more representative sample of Blacks attended school as time went on. This is probably why the largest increase was seen among high-school students.

War Data

Another important source of early IQ data comes from the military. Each time America drafted people for a major war they administered IQ tests to the draftees. The standardized B/W IQ gaps found in these samples can be seen below (2):

Army Gaps.png

Some authors have argued that poor southern Blacks were excluded from mental testing in the first world war and, because of this, the world war one data underestimates the true B/W gap at the time (3).

Like the data on school children, this data suggests that the B/Q IQ gap grew in the early 20th century. However, as with the data on school children, this might just be due to the representativeness of army draftees increasing between the first to the second world war.

Birth Cohort Data

Another source of information on the B/W IQ gap in the early 20th century comes from contemporary birth cohort studies. In these analyses, we look at the B/W IQ gap based on the year in which someone was born rather than the year in which someone had their IQ tested. This allows researchers to estimate the B/W IQ gap for decades prior to when their tests are administered.

The birth cohort data we are going to look at right now comes from the General Social Survey’s “wordsum” vocabulary test. As the name implies, this is just a vocabulary test, but some vocabulary tests correlate as highly as .75 with full scale IQ, and so can be used as a proxy for general intelligence (4).

That being said, the Wordsum test does show a smaller B/W gap than full scale IQ tests do. In the 1960s-2000s Wordsum B/W gaps were about 67% as large as the gaps shown on full IQ tests. We can use this figure to estimate the total IQ gap among birth cohorts (5).

To the 50s by Birth Year.png

Unfortunately, the correlation between life expectancy and IQ implies that these estimates probably still underestimate the actual IQ gap of the time. Even in the 1970s, when the GSS began collecting data, Black people born in the 1900s-1920s would have had above-average life expectancies and so above average IQs.

This will cause an underestimation of the true B/W IQ gap and this bias will be lesser the more recent the birth cohort is. Thus, this bias would, on its own, cause the B/W IQ gap to widen with time in the early 20th century.

As can be seen, this data does show an increase in the gap between the 1900s and the 1920s. It also shows a decline in the gap between the 1920s and the 1930s.

Did The Gap Really Widen?

Unfortunately, all three kinds of data we have on the early twentieth century have biases which will cause the appearance of the gap widening in the early 20th century. As we will see later, income and crime disparities between Blacks and Whites changed in the early 20th century in the direction that we would expect if the IQ gap widened while racial gaps in educational attainment did not. If I had to place a bet, I would say that the Black/Whtie IQ gap probably widened in the early 20th century. But, if given the choice, I wouldn’t bet at all. The available data is simply too problematic to justify any strong conclusions.

If I had to place a bet, I would say that the Black/White IQ gap probably widened in the early 20th century. But, if given the choice, I wouldn’t bet at all. The available data is simply too problematic to justify any strong conclusions.

The Convergence

The website “Human Varieties” has published an excellent meta-anlaysis, an of update to Shuey’s work, showing what the B/W IQ gap has been over the 20th century across 700+ samples (6).

A.png

In conjunction with the data already reviewed, this data set includes tests given to college applicants, standardized tests used in schools, IQ standardization data, tests given to criminals, and more.

As can be seen, this approach suggests that the B/W IQ gap rose up into the 1970’s, declined in the 1990’s, and has stagnated every since. It also shows that the Black/White IQ gap is almost as large today as it was prior to the 1960s.

IQ Standardization Samples

Of course, the primary objection to doing this sort of analysis is that not all data is equal in quality. The best data comes from IQ standardization samples which are likely to have very good measures of IQ and nationally representative samples of Americans.

A standardization sample is a nationally representative sample that an IQ test company administers their test to in order to determine how to weight each test item so as to achieve the desired mean and standard deviation and to measure their tests psychometric validity.

A new standardization sample is used each time a new version of a test is issued. There is variation in questions between versions of each test, but this is less than the variation between different kinds of tests and so variation in test items are less likely to cause changes in the Black/White IQ gap.

The samples used in standardization samples may be representative of the whole nation, but they are not always representative of every subset of the population. Specifically, the poorest and least educated people in the united states are sometimes underrepresented in these samples. This creates a problem when using them to estimate the absolute value of the B/W IQ gap for any time period but in no way inhibits our ability to measure changes in the B/W IQ gap over time. So long as the bias is stable, it’s not a problem.

In the early 20th century, IQ tests were rare and standardization samples were not reliably broken down by race. By the 1960s this situation had changed. I was able to find data on 8 IQ tests which had at least two standardization samples between the 1960s and now.

Within Test Standardized Gap.png

The following three show continual significant (more than 0.1) declines in the B/W IQ gap: the Wisc (1970s-2000s), the Standford-Binet (1960s-2000s), and the AFQT (1960s-90s) (8).

The following three tests show no significant change: the Kaufman (80s-2000s), the NAAL (90s-2000s), and the Wonderlic (1970s-2000s) (9). The GSS’s wordsum, analyzed by test administration year, also shows no significant change (1970s-2010s).

(The results for the Wordsum test without excluding Hispanics are included in parentheses for the last two columns.)

Finally, the following two IQ tests show a significant decline in the B/W IQ gap followed by a significant rise: the WAIS (1970s-2000s) and the Woodcock Johnson (1970s-90s) (10).

Clearly, analyzing tests this way gives us very mixed results. An alternative way of looking at this data is to go decade by decade.

One test reported data in the 60s and 70s and it showed a decline of .05 SD in the gap. Three tests reported data in the 70s and 80s and they reported an average .15 SD decline in the gap. Three tests reported data in the 80s and 90s and they reported an average .01 SD decline in the gap. Five tests reported data in the 90s and 2000s and they reported an average .03 SD decline in the gap. One test reported data in the 2000s and the 2010s and it showed an increase of .03 SD in the gap. These changes average out to a narrowing of the gap by .04 SD per decade.

Five tests reported data in the 70s and the 90s. On average, they showed a decline in the gap of .09 SD. This amounts to .045 SD per decade. Four tests reported data in the 70s and the 2000s. On average, they showed a decline of  .12 SD. This amounts to .04 SD per decade. Clearly, both trends are almost entirely explained by the .15 SD decline between the 70s and 80s.

Three tests reported data in the 80s and 2000s. On average, they showed a decline in the gap of .11 SD. This amount to .04 SD per decade. However, this trend is entirely driven by the WISC test.

The other two tests which reported data in the 80s and 2000s show an average decline of .005 SD, virtually zero, in the gap. Moreover, the two tests are consistent on this point: the Wordsum shows a change of 0.00 SD while the Kaufman shows a change of -0.01 SD.

Given this, the most reasonable interpretation of the within test data is the same as what is implied by Human Varieties aggregate data: the B/W IQ gap declined between the 70s and 80s and has been stagnant since.

School Tests

Standardized tests given to school children are probably the second best source of evidence. Compared to IQ testing companies, they have far larger samples but worse measures of cognitive ability.

By far, the SATs are the best cognitive test administered by schools. Scores on these tests correlate at .86 with full-scale IQ, making them essentially an IQ test (11). SAT scores by race show a significant decline in the B/W gap in both verbal and math scores from 1975 to 1986. From 1987 to 2013 the B/W gap was largely stagnant (12).

Final SAT Gap Line Graph.png

NAEP tests focus more on acquired knowledge, and so are probably biased in favor of crystallized intelligence. However, unlike the SATs, they are administered to multiple age groups, which adds an interesting dimension to the data.

Here is how the NAEP math gap has changed by age and decade (13):

by-decade-math

analysis-1

And here is the same analysis, but for the NAEP’s reading tests:

by-decade-reading

analysis-2

A few things of note emerge from this data. First, for both tests, the largest convergence happened between the 70s and 80s just as we saw with IQ tests. Second, the reading test shows continued, smaller, gains in subsequent decades while the math test shows no significant convergence after the 1980s. Thirdly, convergence was larger for the reading test than it was for the math test. And fourthly, for both tests gains were stronger for 17 and 13 year olds than they were for 9 year olds.

Finally, we have ACT data. The ACT is a newer test that has not been administered nationally for long. The B/W ACT gap is virtually identical today to what it was 15 years ago (14).

ACT Gap by Year.png

In totality, the school data suggests that the B/W IQ gap probably converged between the 1970s and 1980’s and has probably stagnated since. This is the same conclusion supported by the aggregated and within test IQ data.

Though IQ test data and school data both paint the same general picture, the magnitude of the convergence is much greater in school tests. On average, each IQ test that had a standardization sample during and before the 1970s, and after the 1970s, showed a narrowing of the gap of 13%. By contrast, school test data showed a narrowing of 27%.

Average Narrowing.png

(NAEP data for 17-year-olds was used in this table.)

There is also NEAP data suggesting that the B/W IQ gap fell by about 15% between the 1950s and the 1960s (28). It was not included in the above data because it is not available on the NAEPs website but, rather, comes from an academic paper, and shows different absolute gaps in later years then the NAEP website does today.

By Birth Cohort

A final source of evidence on changes in the B/W IQ gap in the later 20th century comes from birth cohort data. Since we are dealing with birth cohorts, a convergence in the 1970s should be reflected as a convergence among people born between the 40s and 60s.

As can be seen, data from the wordsum test is consistent with data from school tests and IQ tests.

wordsum-gap-non-hispanics

(These numbers differ from those used previously because I excluded Hispanics to stop immigration trends which began in the 1970s from affecting the results. Non-Hispanic data could not be used for the 1900s and 1910s because the GSS did not begin asking people whether or not they were Hispanic until 2000 by which point there were very few respondents born in these early decades.)

Another birth Cohort data set comes from the 2008 WAIS standardization sample (15). The WAIS is the most popular full scale IQ test in existence. Thus, the test itself is a much better measure of general intelligence than is Wordsum. However, the WAIS sample is much smaller and because the results come from the late 2000’s, people born in the early 20th century will virtually all be senior citizens.

By Birth Cohort.JPG

As can be seen, this data set shows a decline in the B/W IQ gap which begins with people born in the 1917-1942 period and continues up to people born between 1963 and 1977, stagnates, and then picks up again with people born between 1988 and 1991. This last bit, the gap decline resuming among people born between 1988 and 1991, is the only aspect of this data which is inconsistent with the other data sets we’ve looked at.

Yet another birth cohort data set comes from Charles Murray’s analysis of birth cohorts in a Woodcock Johnson IQ test’s standardization sample (16). This sample showed a decline of the gap from 1.33 SD among people born between 1920 and 1939 to 1.08 SD among people born between 1940 and 1955. The gap continued to decline until 1972 when it was .83 SD and then rose to .98 SD in 1987-1991. This data set is fully consistent with the other data sets we’ve seen so far except for the slight increase in the B/W IQ gap among those born between 1987 and 1991.

A final birth cohort data set comes from Charles Murray’s analysis of Peabody tests administered in the National Longitudinal Survey of Youth (17). Looking at people born between 1970 and 1995, Murray found a stagnant gap in reading scores and a widening of the B/W gap in math and vocabulary.

PeaBody Scores by Birth Date, Reading Gaps Stayed the Same, Math and Vocab Increased.JPG

Looking at the GSS, WAIS, WJ, and Peabody, data together it is clear that all the data sets are consistent with a closing of the gap among people born between 1920 and sometime in the early 1960s. Following this, the gap stagnated at least until the late 1980s birth cohorts. At this point the data sets diverge: GSS data shows the gap remaining stagnant, WAIS data shows it declining again, and Peabody and WJ data show it increasing.

On the whole, birth cohort data does not seriously challenge the conclusions supported by aggregate data, within IQ test data, or school test data on the convergence of the gap in the late 20th century.

However, birth cohort data does imply that the gap began narrowing earlier, perhaps in the 1940s or 1950s. This is in sharp contrast with Shuey’s data, consisting of tests actually administered during this time period, which shows the gap increasing rather than decreasing.

Convergence by Sub-Group

At this point, we hopefully all agree that a narrowing of the B/W IQ gap clearly happened in the late 20th century and a widening of the gap may have happened between the two world wars. Let’s look at how these changes differed by sex, age, ability level, and test item G loading, before trying to answer what caused these IQ changes.

Gains by Sex

First, let’s look at how changes in the B/W IQ gap have differed by sex. Averaging across ages and subjects, the NAEP data we looked at before shows that the female gap fell by an average of .33 SD compared to .34 SD for males. Thus, they made virtually identical total gains.

However, females closed the gap in reading score by .47 SD compared to .42 SD for men while men closed the math gap by .26 SD compared to .19 SD for women. Thus, there was a slight tendency for males to gain more on math tests and women to gain more on reading tests.

The Standardized NAEP Gap by Decade, Sex, and Age.png

Wordsum data can also be broken down by sex. The increase in the B/W IQ gap shown in Wordsum data for the very early 20th century is mostly caused by changes among women. In the early 20th century the female B/W Wordusm gap was larger than the male B/W Wordsum gap by .06 SD. This difference grew to .25 SD by the 1920s.

In terms of the late 20th-century convergence, comparing the 1950’s to the 1970’s, we see that the female gap fell by .27 SD while the male gap fell by .23 SD, suggesting little difference between the two trajectories in this time period.

Wordsum by Decade and Sex.png

Gains by Age

Next, let’s look at how changes in the gap have varied by age. Several lines of evidence suggest that the Black/White IQ gap changed more over time among adults than it did among elementary school aged children.

Shuey’s data comparing 1921-1944 and 1945-1965 found that the gap among elementary and middle school-aged children did not change and the gap among high school students increased by .53 SD.

NAEP data shows that the gap among 9-year-olds fell by .06 SD in math and .31 SD in reading. The gap among 13-year-olds fell by .25 SD in math and .38 SD in reading. The gap among 17-year-olds fell by .25 SD in math and .54 SD in reading. Thus, 9 year olds closed their gap the least, and 17 year olds closed their gap the most.

Finally, Sacket and Shen meta-analyzed data from 80+ school samples from the 1960s-2007 (19). They found that the standardized B/W gap in cognitive ability among elementary school children fell from 1.04 SD in the 1960s to .98 SD in the 1970s and .81 in the 1980s, after which the gap became stagnant. Thiu, the gap fell a total of .23 SD.

Among highschool children, the gap changed from 1.08 SD in the 1960s to 1.23 SD in the 1970s, to 0.92 SD in the 1980s, and 0.82 SD in the 1990s. Thus, the gap fell  .25 SD if we take the 1960s as our start point or by .41 SD if use the 1970s as our start point. Either way, the gap converged more among highschools than it did among middle school and elementary school aged children.

In terms of very young children, Shuey found that the gap among pre-school children increased by .47. Picking up where this research left off, Jason Malloy analyzed 34 samples of cognitive tests administered to Blacks and/orWhite aged 3 or younger which were administered between the early 1960’s and the early 2000’s (18). In each decade, the average B/W IQ gap was found to be 1 SD. This suggests that the racial gap among children too young to go to school may not have converged in the late 20th century at all.

This 1 SD gap is also very close to the 1.07 SD gap that Shuey found among pre schoolers tested between 1945 and 1965. Thus, the B/W IQ gap among infants looks to have been fairly constant since the second world war.

In summary, multiple sources of evidence suggest that the Black/White IQ gap fluctuated more strongly among adults than among younger people. Moreover, the late 20th century lessening of the gap may not have happened at all among very young people.

Gains by Ability Level

Two studies, using NAEP data on the one hand and AFQT standardization data on the other, have looked at how gains in Black IQ were distributed across ability level (20). Both found that gains were low for high ability Blacks and especially large for low ability Blacks. In other words, the mean Black IQ value has risen in the late 20th century primarily because of increase in values at low, rather than high, percentiles.

Gains by Ability Level.JPG

Gains by g loading

Two studies have looked at the question of whether or not the late 20th-century convergence in IQ scores was also a convergence in G (21). Both papers found a negative correlation between a test item, or sub factor’s, g loading and its degree of convergence. However, Dickens and Flynn also extracted each races score on the G factor from WISC and WAIS standardization samples and found that, in spite of the previously mentioned negative correlation (r=-.28), gains on G were 93% as large as total IQ gains were. Thus, while gains were disproportionately not on G this effect was slight.

Hollow Gains?

Let’s now turn to asking whether or not changes in the B/W IQ gap coincided with changes in racial differences in income, education, and crime, all of which have well-established causal relationships to IQ.

If these changes reflect real changes in intelligence, we would expect that the rate of convergence between the races in terms of income and education should be higher than average between the 1970s and 1980s and lower than average between the 1910s and 1940s.

If changes in intelligence do not predict changes in IQ related life outcomes, then relative rises in black performance on IQ tests may be caused by an increase in test taking ability which was not accompanied by an increase in actual intelligence.

The census provides easily accessible data on the Black/White income gap going back to the 1940s (22). Using this data, we can see that, on average, the Black/White income ratio increased by .04 per decade.

Total Income.png

This ratio actually fell by .01 between the 70s and 80s and rose by only .02 between the 80s and 90s, contradicting what we would expect if the convergence in IQ scores reflected a real convergence in intelligence.

Unfortunately, the only data I’ve been able to find which breaks income down by race and which extends back to the 1910s is only for men and, as we’ve already seen, Wordsum data indicates that the possible early 20th century widening of the B/W IQ gap mostly happened among women (23).

However, using this data we can see that, on average, the Black/White male income ratio rose by.019 per decade. Between the 1910s and the 1940s, this ratio rose by .01, or .0025 per decade. This is consistent with what we would expect if the early 20th-century rise in the B/W IQ gap reflected a genuine change in relative intelligence.

Income Ratio.png
This data also shows an increase of .02 between the 70s and 80s. This rate of increase is perfectly average for this 90 year time period and so not what we would expect if there was a real convergence of intelligence.

Turning to education, we can see that, on average, between the 1940s and the 2010s the increased probability of finishing high school if a person was White rather than Black decreased by .27 per decade (24). The same figure for finishing college is .30 per decade.

Between the 70s and 80s, the high school figure fell by .27 and the college figure fell by .39. Between the 80s and the 90s, the high-school figure fell by .11 while the college figure fell by .06. Thus, no exceptional convergence in education is seen during this time, contrary to what we would expect if the gap reflected a genuine change in intelligence.

White - Black probability of finishing school.png

The best education data I could find which went all the way back to the 1910s was the census’s records of the average number of years of schooling completed by race and sex (23). This data is broken into five-year periods rather than decades.

The average change per 5 years in the White advantage in the average number of years in school was -0.20 or men and -0.26 for females. Between 1911 and 1940, the change for men was 1.87, or -.31 per five-year period, and the change for women was  -1.77, or  -.30 per five year period. Thus, contrary to what we would expect if the Black/White intelligence gap was widening, the B/Q education gap seems to have converged at a faster than average rate during this period.

White Ad in Years of Schooling.png

Data on both income and educaiton suggests that there was no real convergence of intelligence in the late 20th century. Income data, but not education data, is consistent with a real widening of the intelligence gap in the early 1900s.

Crime data should also reflect changes in intelligence. Unfortunately, the data I found only goes back to the 1920s (24). However, between the 1920s and the 1940s the Black to non-Black incarceration ratio increased by .47 per decade compared to an average increase of .10 per decade across this entire data set. This is the kind of thing we would expect if Black intelligence was falling relative to White intelligence during this time.

Incarceration.png

By contrast, the ratio increased by .35 from the 70s to the 80s and by .37 from the 80s to the 90s. This is totally contradictory to what we would expect if a real convergence in intelligence was happening at this time.

This evidence increases my confidence that the IQ gap may have widened between the two world wars. Data on income and crime both match up with what this hypothesis would predict, though education data does not.

By contrast, education, crime, and income, data all suggest that there was no real convergence in intelligence in the late 20th century. If Blacks really did become more relatively intelligent then the 70s and 80s should stand out as a time of exceptional Black improvement in terms of income, education, and crime. It does not.

Given this, I think it is reasonable to conclude that in the 1970s Black Americans may have improved in their ability to take tests but not in terms of their actual intelligence.

Why Did The Gap Narrow?

A few authors have looked at the causes of the late 20th century convergence by examining whether, in their samples, Blacks and Whites of equal SES had the same IQ convergence as did their total Black and White samples. The results of such analyses have been contradictory (25).

However, it is obvious that no simple story about SES can explain this convergence. After all, racial differences in income and education lessened throughout the early and very late 20th century when there was no convergence of IQ. Since convergence only significantly happened in the 60s/70s, something particular to the 60s/70s must be the cause.

The best hypothesis I can come up with has to do with desegregation. Modern analyses by the Department of Education have shown that the proportion of a school which is Black negatively correlates with the test scores of both Whites and Blacks in that school (26). This is likely because, as research has shown, Black students are more likely than White students to be disruptive, commit violence, and otherwise damage the educational environment (27).

Because of this, having Africans in a school inhibits student’s ability to reap the rewards of formal schooling. Chief among said benefits may be a familiarity with test taking procedures, mental heuristics to use to succeed in formal testing, and a knowledge of the kinds of answers that teachers/researchers look for on tests.

Beginning in the 1950s, there was a huge legal push to end the segregated schooling which was prevalent in the south. It could be that this caused the IQ convergence that we see in the 70s.

A prediction which this hypothesis makes is that the converging test scores in the 1970s should be caused both by an increase in Black test scores, who benefited from desegregation, and a decline in White scores, who were hurt by it.

In order to test this we need to look at the absolute scores of Whites and Blacks on cognitive tests ranging at least from the 1950s to the 1970s. The only two sources I know of which meet these conditions are the GSS wordsum test data (by birth year) and SAT data.

The wordsum data shows exactly what this hypothesis predicts: the absolute scores of Whites rose from the 1900s to the 1950s and then fell. Thus, part of the convergence in the wordsum B/W IQ gap is caused by the White score falling.

The SAT data I have by race only goes back to the 1970s. It shows the total White score falling 19 points between 1975 and 1980 while the Black score rose by 8.

We can extend this analysis back further. Because White people were the vast majority of the population before the 70s, looking at changes in the national mean is a pretty good proxy for changes in the White mean.

Average SAT scores rose throughout the 1950s and 1960s and then fell in the 1970s. We know that this falling cannot be due to Black SAT scores falling at an equal or greater rate than White SAT scores since the SAT gap during this period fell. This data is consistent with What SAT scores falling and Black SAT scores rising.

This hypothesis is superior to vague claims about general socio-economic status because it specifies something which happened very suddenly in the 50s and 60s but which was not a continual process throughout the entire century.

Moreover, this explanation tells us why White cognitive ability declined during this time period while others do not. It can also explain why there were no significant cognitive gains among elite Blacks: they likely already went to schools with high quality student populations. Further still, this explains why there was no narrowing of the gap among 3-year-olds, who were too young to go to school, and why it narrowed more among high school students than elementary schoolers, since high schoolers would be exposed to beneficial changes in education for longer than elementary schoolers would be. This also explains why the gap narrowed more on school tests than on formal IQ tests and why more G loaded IQ sub tests were more resistant to the convergence.

In short, this hypothesis is far from proven but it fits all the data and makes a few novel predictions which seem to match up with reality.

Implications

The data I’ve reviewed here has some important implications about the causes of racial intelligence differences. Throughout the twentieth and twenty-first century, racial differences in education and income have diminished. Racial gaps in single mother hood have grown, especially in the late 20th century. Racism has profoundly declined.

None of these things are likely to be large causes of the Black/White IQ gap for the simple reason that changes in them over time do not coincide at all with the changes in the IQ gap over time which I have examined in this article.

Moreover, the only decline that has occurred in the B/W IQ gap has been relatively small. The gap today seems to be around 1 SD compared to something like 1.15 SD in the 1950s. In light of the fact that the environment has been made more equal in so many regards and, in fact, has been shifted to favor Blacks in many regards (private and public affirmative action) this makes such a slight convergence in IQ scores somewhat astounding from an egalitarian perspective.

 

Sources

  1. This breakdown of Shuey’s data comes from table 6.1 on page 141 of “Race Differences in Intelligence” by Loehlin, Lindzey, and Spuhler (1975).
  2. This data comes fromtable 6.2 on page 143 “Race Differences in Intelligence” by Loehlin, Lindzey, and Spuhler (1975). Some authors (Jensen) have reported the size of the WW2 gap being only 1.25 SD. Others (Chuck) have reported it being still smaller. Regardless, there is agreement that the gap was larger in the second world war testing. Loehlin et al give a clear explanation of how they came to their figure and that is why I chose to use theirs instead of others.
  3. The magnitude and components of change in the black–white IQ difference from 1920 to 1991: A birth cohort analysis of the Woodcock–Johnson standardizations” by Charles Murray (2007).
  4. Notes and Shorter Communications Has the black–white intelligence difference in the United States been narrowing over time?” by Richard Lynn (1998)
  5. This is based on my own analysis of GSS data. Decades before 1900 and after 1980 were excluded because of the very low sample sizes for Blacks. For each decade shown, the sample size of each group was at least 70. Variables used were “cohort”, “race”, and “wordsum”. The 67% figure was derived from the table in this article showing the IQ gap by decade within commonly used IQ tests.
  6. 100 Years of Testing Negro Intelligence” by Chuck (2013).
  7. Wisc and AFQT data was taken from “100 Years of Testing Negro Intelligence” by Chuck (2013). SB data came from “Black Americans Reduce the Racial IQ Gap: Evidence from Standardization Samples” by Dickens and Flynn (2006).
  8. Data for the Wonderlic and NAAL was taken from from “100 Years of Testing Negro Intelligence” by Chuck (2013). Data on the Kuafman comes from “Black Americans Reduce the Racial IQ Gap: Evidence from Standardization Samples” by Dickens and Flynn (2006).
  9. Data on the WAIS 2000s standardization sample comes from table 4.8 on page 190 of “WAIS-IV, WMS-IV, and ACS: Advanced Clinical Interpretation” by Holdnack, Drozdick, Wiess, and Iverson (2013). Data for the 1990s standardization sample came from “100 Years of Testing Negro Intelligence” by Chuck (2013). Data for the WAIS 1980s standardization sample came from “Black/White IQ Differences: Does Age Make the Difference?” by Ken Vincent (1991).
  10. Scholastic Assessment or g? The Relationship Between the Scholastic Assessment Test and General Cognitive Ability” by Frey and Detterman (2003).
  11. SAT data, by race, for 1975-1988 was taken from “Digest of Education Statistics, 1990” by Synder and Hoffman (1991). Data for the years 1989-1993 were taken from “Digest of Education Statistics, 1995” by Synder and Hoffman (1995). Data from 1993 to 2013 was taken from “Table 226.10 SAT mean scores of college-bound seniors, by race/ethnicity: Selected years, 1986-87 through 2013-14” from the 2014 Digest of Education Statistics. SAT norming procedures changed in the mid 1990s. However, as can be seen, this seems to have had no effect on the racial gap.
  12. This data was taken from the NAEP Long Term Trend data explorer. The 4 way racial categorization was used.
  13. This data come from Table 226.5 of the 2015 Digest of Education Statistics.
  14. Taken from table 4.7 on page 189 of “WAIS-IV, WMS-IV, and ACS: Advanced Clinical Interpretation” by Holdnack, Drozdick, Wiess, and Iverson (2013).
  15. The magnitude and components of change in the black–white IQ difference from 1920 to 1991: A birth cohort analysis of the Woodcock–Johnson standardizations” by Charles Murray (2006).
  16. Changes over time in the black–white difference on mental tests: Evidence from the children of the 1979 cohort of the National Longitudinal Survey of Youth” by Charles Murray (2006).
  17. The Onset and Development of B-W Ability Differences: Early Infancy to Age 3 (Part 1)” by Jason Malloy (2013).
  18. Subgroup Differences on Cognitive Tests in Contexts Other Than Personnel Selection” by Sackett and Shen (2009).
  19. The analysis of AFQT is “Black Americans Reduce the Racial IQ Gap: Evidence from Standardization Samples” by Dickens and Flynn (2006). The analysis of NAEP data is “Ability rise in NAEP and narrowing ethnic gaps?” by Rindermann and Thomspon (2013).
  20. Black Americans Reduce the Racial IQ Gap: Evidence from Standardization Samples” by Dickens and Flynn (2006) and “The magnitude and components of change in the black–white IQ difference from 1920 to 1991: A birth cohort analysis of the Woodcock–Johnson standardizations” by Charles Murray (2006).
  21. Data from table P-4 from the Census’s “Historical Income Tables: People” page.
  22. Race and Human Capital” by James Smith (1984).
  23. Race of Prisoners Admitted to State and Federal Institutions, 1926-86” by Patrick Langan (1991).
  24. Black-White Test Score Convergence” by Hedges and Nowell (1998) found that the gap converged even after controlling for changes in SES. However, “Convergent Trends in Black-White Test-Score Differentials in the U.S: A Correction of Richard Lynn” by Huang and Hauser (2000) found that controlling for SES stopped Black scores for rising over time.
  25. School Composition and the Black–White Achievement Gap” by the National Center for Education Statistics (2015).
  26. How Ethnic Diversity Harms Students” by John Macintyre (2016).
  27. Why Has the Black-White Skill Convergence Stopped” by Derek Neal (2005)

 

Facebook Comments
  • WaterLeaf

    The obvious explanations for any closing in the IQ gap would be:

    #1 Increased European admixture in American Blacks

    #2 Increased Black Immigration from Africa. I assume the Blacks who come to the U.S from Africa tend to be more intelligent than the average Black currently in the U.S as like with the U.K.

    The fact both of these things have taken place since the 40’s (increased admixture and increased Black immigration) yet the gap hasn’t changed much is telling.

    Also, many tests for colleges and schools have indeed been made easier to make room for Blacks and Hispanics but I cba to get the sources for that right now. Any sort of a closing gap in which the test in question has changed over time is suspect until investigated.

  • John_HD

    With our massive immigration over the last 30 years, I’m wondering if Mestizos and Middle Easterners are lumped in the “white” category. That would be explanatory, too.

  • Audacious Epigone

    Great insight, definitely something to drill down further on.

    If you’re not already, I’d suggest always filtering Wordsum scores with BORN(1).

  • JustforBob

    Is there any evidence the IQ gap widens once blacks leave school?