January 25, 2017

IQs of Races in the United States

Here I’m going to go over the median IQs of the main races in the United States. There are 4 sources for this.

  1. The IQ Studies
  2. SAT scores
  3. ACT scores
  4. NAEP scores
  5. TIMMS and PISA scores

And in this we will be looking at the scores for these racial categories – whites, blacks, hispanics, asians – and pacific islanders and american indians where available.

The SAT, ACT and NAEP tests are not IQ tests, but the SAT correlates very well with them, and the others are more like “pinch tests”, in that we can see if they align with the IQ test data.

One thing I do is convert the standardized tests that aren’t IQ tests to “IQ notation”, which simply means converting the scores to a scale where the mean and median is 100, and has standard deviations of 15.

I am not claiming that these (SAT, ACT, NAEP, TIMMS and PISA) are IQ tests, I am simply showing the scores on them as they would be presented on an IQ scale using the standard deviations on those tests.

For example, the TIMMS data in the US has a standard deviation of 70 and a mean / median of 541.5 for US whites. This means that a TIMMS score  of 471.5 would equate to an IQ of 85 if we used US whites as the “norm”, and since US whites scored 541.5, this means that a score of 541.5 would equate to an IQ of 100, while a score of 401.5 would equate to an IQ of 70.

If we were to “norm based off of the world average”, we would set the world average score which is 500 (which is by design) to an IQ of 100, and the global standard deviation of 100 (again by design) to 15. This would mean a score of 500 would equate to an IQ of 100, a score of 400 would equate to an IQ of 85, and a score of 300 would equate to an IQ of 70.

So there can be HUGE differences in interpreting test scores depending on who is defined as “the norm”, not only in terms of what the median / mean is set to, but also how many points you have to deviate from the mean in raw score to deviate in IQ notation equivalent.

And since we’re talking about the United States, I’m always going to set the “norm” as US Whites – which means that whatever US Whites score, that is defined as 100.

IQ studies

The median for these 20 studies is 89, and the mean is 88.9.

If you exclude Puerto Rico, the median is 89.5 and the mean is also 89.5.

Slicing this data doesn’t change much. Even if you only include the two studies from the 2000’s, that yields a score of 90.5. And there is no noticeable trend.

Roth’s meta analysis, from 2001, had 105 studies with 6,246,729 participants, and it found the median IQ of “hispanics” across a variety of tests to center around 89.

Regarding African IQ, the most comprehensive analysis I have found is from “Chuck” at Humanvarieties.org. He looked at 57 studies on the black-white IQ gap, and gave these as the gaps for each decade:

Pre-1960s          17.1

1960s                 17.55

1970s                 18.3

1980s                14.55

1990s                14.55

2000s               15.0

Chuck’s full analysis with list of studies is available here:


One thing you should keep in mind is that the very first IQ test of blacks was done in 1917, and this was to screen for possible officers for the US Army. The median black IQ on that test was 83. In the 10 most recent studies in Chuck’s list, which were all after the year 2000, the average black IQ is 85.7 in those studies, and the median is 85. This was also found by Richard Lynn.

Richard Lynn did similar analyses for East Asians, Amerindians and Pacific Islanders in the United States. I will not copy them down here like I did with the hispanics, but will describe them here.

For the East Asians, Lynn looked at 27 studies of East Asians in the United States (compared to 20 for hispanics as seen above). The mean of these studies was 101.3, and the median was 101.

For Pacific Islanders, Lynn had 4 studies of Pacific Islanders living in Hawaii and the mean IQ of those studies was 86.5, median 87.

For the Amerindians, Lynn looked at 21 studies. The mean was 86 and the median was 90.

SAT Scores

The SAT correlates with an IQ test at 0.86, almost the same as an IQ test correlates with itself. For this reason, we can very reliably take SAT scores and convert them to IQ scores.

The big question with SAT score data is that it uses elite samples. Is one group more or less elite than another? Well, for whites, blacks and hispanics not compared to each other, but the “asians” are far less elite than any of those 3.

What this means is that blacks and whites take the SAT at almost exactly same rate for their 18-19 year old population, which means that the SAT scores are *probably* representative relative to each other.

Now it could be that the whites who take the SAT are a dumber slice of the white population than the blacks who take the SAT are, or vice-versa. But the error-minimizing assumption is that they are “similar slices” from each group.

The “hispanic” scores are only *likely to be* slightly elite, while the “Asian” scores represent a much broader cross-section of their number, and they’re still on top.

This black-white SAT score gap has been persistent at least since 1986. From 1986 to 2013, the average math + verbal SAT for whites has been 1058.40, while the average for blacks has been 856.47, remarkably close to the 2015 scores of 1063 to 859, which roughly translate to IQs of 113 and 98.

And we know that, at least in 2015, the white test-takers and black test-takers both probably represented a similarly “elite” cut of their racial groups, and both each year regularly score almost exactly 13 points above the “conventional IQs” of their blacks and whites, which are 85 and 100 respectively.

Based on SAT score data, the median IQ of African-Americans is probably 85, “hispanics” are probably 88, 90 if you just use math.

It’s also important to distinguish the math and verbal scores to show that blacks actually do relatively better on verbal than on math, which is one of numerous disproofs of “cultural bias”.

ACT Scores

There is no direct source I know of that converts ACT to IQ test scores. However, the ACT to SAT equivalents are well known, and the SAT equivalent can be converted to an IQ score, giving us an “ACTIQ”.

The differences in relative populations of races taking the ACT may be a function of regionality more than the eliteness of the test-takers.

IQ estimates from the ACT by group with the white score set to 100:


The NAEP is a government-issued achievement tests. While it is not a psychometric test, it is designed to have a standard deviation of 35. Since IQ tests have a standard deviation of 15, we can make an “NAEPIQ” equivalent by setting the white score to 100, and converting the other races’ scores to IQ based on a standard deviation of 15.

This “NAEPIQ” doesn’t necessarily mean IQ, as NAEP scores are not the same as an IQ test. Converting it to “IQ score notation” is just a different way of presenting the exact same data so long as you know the standard deviations.

If you wanted, you could take 40-yard dash times, find the standard deviation of the times, and come up with a “40-yard dash IQ”. That’s literally all I’m doing, but with the NAEP data.

The NAEP by design has a standard deviation of 35. With the white score set to 100, we can come up with “NAEPIQ” by race and year:

The NAEP shows a much smaller black-white gap in terms of standard deviations than the SAT, ACT or the IQ studies, regardless of the year. One explanation for this is that, because funding is often tied to scores, black schools fudge exclude a large proportion of their students from the test.

No sample for a study is ever perfect, but there is reason to believe that low-performing (“urban”) school districts would cheat, and there is evidence that they have in fact done so. I am not making a moral statement, in fact I support them and believe standards should be based on racial composition of the school, I am merely saying that this would bias the data to lower the black-white gap.

However, the NAEP isn’t an IQ test anyway, and in theory tests “achievement”, and since that should be more under the influence of environment than “native intelligence” is, a heredetarian would expect the black-white gaps to be smaller on achievement tests than on tests of native intelligence.

Which is what we see here.

The B-W Gap over time

The SAT, NAEP and IQ studies all show a similar pattern: a narrowing of black-white IQ scores from 1973 to 1986, and then flatlining from that point. In fact there is IQ data that goes back to 1917 where the US Army issued a nation-wide standardized test, and found a black IQ of 83 to the white 100.

The NAEP only deviates from the SAT and IQ studies in its “final resting place” for black IQ, which is roughly 89, whereas the IQ studies and SAT put it at 85.

The IQs of racial groups in the United States based on all of the data presented here:

At the moment this is taking into account whatever genetic makeup goes into these categories, especially “Hispanics”. For now, we are just looking at the IQs of these as categories, agnostic as to their genetic distinctiveness.

TIMSS and PISA data

TIMSS stands for “Trends in International Math and Science Study”, and gave a standard test to 42 countries on math and science. PISA stands for “Program for International Student Assessment” and gave a test for reading, math and science for 65 countries.

The IQ conversion for PISA and TIMMS scores were based on the US white standard deviation of 100, as that is the global norm.

*However, when normed based on US Whites (yes the US white SD happened to be a round number), the standard deviation used is 70 for TIMMS and 98 for PISA, which are lower than the global PISA and TIMMS standard deviations of 100. However I just used a SD of 100 for the PISA scores, it will make the racial gaps slightly smaller, but the math was easier.

When comparing racial groups in the US in the context of other standardized tests WITHIN the US, using US standard deviations is appropriate.

If you used global standard deviations, which are much larger, this would reduce the racial gaps in the standardized-tests to IQ conversion, unless of course we changed the standard deviation of IQ to what it actually is outside of the United States (which would be something like 20 instead of 15).


This part is where I give my opinion.

My estimate for the median IQs of various races in the US and the simple average of all 6 data sources next to it., based on all of the above data plus some stuff in other posts on this site, leads me to these estimates:

Asians          103.3 (103.26)
Whites         100 (100)
Hispanics    90  (89.75)
Blacks          85 (85.8)

Pacific Islanders        90
American Indians     92
Indian-Americans    106

For the first four I am very confident. For the next three I am less confident as it is based on less data.

I estimate the black IQ at 85 and not 86 because I discount the NAEP data. Removing the NAEP data gives a raw black number of 85.3, and the median of the 6 sources of data gives a black number of 85.495. But this is just my opinion, all the data is sourced.

I was so thorough with this, because unless you literally shove the readers’ face in this data, he will not believe how shockingly consistent the racial gaps are on all kinds of cognitive tests, and how little it matters what year it’s done and what test it is.

It’s vital to have a very good understanding of what the basic data actually is, so that when we start arguing causality, you have an idea of what “gaps” we’re talking about and how overwhelmingly well-established they are.

Cheating on the NAEP:

NAEP scores:
Page 1 – Long-term trend math or reading, age 17, long-term trend scale, national, all years
Page 2 – Race / ethnicity (6 categories), all years

SAT Scores:

Sources used for SAT to IQ conversion:

US High school population:

2015 Test taker population:

Race and SAT to 1986:

ACT Data:

ACT to IQ was done by converting ACT to equivalent SAT scores, and then using the SAT to IQ conversion table above:

Lynn’s “Race Differences in Intelligence”:

TIMSS and PISA data and standard deviations:

Facebook Comments