Previous analyses have identified a strong inverse correlation between Indigenous ancestry and both academic achievement and socioeconomic outcomes in Canada. Similar patterns are anticipated in Alaska and Greenland. While evidence supports this trend in Alaska, the situation in Greenland remains unclear.
In the previous post, we noted that across Canada—and within individual provinces and territories—Métis, First Nations, and Inuit populations tend to underperform relative to non-Indigenous Canadians. To illustrate this, Table 1 presents average PIAAC literacy and numeracy d-values comparing Aboriginal and non-Aboriginal individuals whose mother tongue matches the language of the test (English or French), disaggregated by province and territory where data are available. Higher positive d-values indicate worse performance among Aboriginal groups. While the magnitude of the gaps varies by region, the disparities are consistently present.
Demographic Change and Cognitive Variation
Canada is rapidly diversifying as a result of relaxed immigration policies. According to the 2021 Census, individuals of European ancestry now constitute approximately 67% of the population, down from 83% two decades earlier. The largest non-European groups include East Asians (9%), South Asians (7%), Aboriginal peoples (6%), and Black Canadians (4%). The pace of demographic change has outstripped genetic survey estimates. For instance, Ancestry.com still reported over 90% European ancestry in Canada as recently as 2017, whereas the true proportion is now likely under 70%. Despite this shift, geographic variation in socioeconomic and cognitive outcomes continues to correlate strongly with European and Amerindian ancestry proportions. This post provides an overview of Aboriginal demographic distributions and cognitive performance in Canada.
In Canada, longstanding differences in cognitive ability and academic achievement exist between the European-origin majority and several traditional minority ethnic groups established prior to the mid-20th century. In particular, Aboriginal populations (First Nations, Inuit, and Métis) and the historically rooted Black Nova Scotian population tend to score below the national average on various educational metrics. Meng Hu and I recently published findings based on data from 2013 to 2022 indicating that Black Nova Scotians score, on average, d = –0.46 below the non-Black Nova Scotian mean (Fuerst & Hu, 2024).
Less is known about the performance of more recent immigrant-origin groups, as academic performance data disaggregated by race/ethnicity is generally not collected at the national level in Canada. However, relatively high academic performance among Chinese and other Northeast Asian Canadians has been documented since Peter Sandiford’s research on Vancouver’s Chinese population in the 1920s (Sandiford & Kerr, 1926), with further evidence summarized by Vernon (1982). More recent national and provincial achievement data corroborate these findings (Bacic & Zheng, 2024; Barber et al., 2021). By contrast, average performance among other ethnic groups—such as Asian Indians, Filipinos, and Hispanics—remains less clear. The same applies to the children of Black African immigrants, many of whom come from highly selected backgrounds — and whom are highly educated. While I have previously reported ethnic performance data from the Toronto District School Board, these findings should not be assumed to generalize nationally or even across Ontario.
Recently, I was able to compile a decade’s worth of Canadian Advanced Placement (AP) data, as reported by College Board (years: 2009, 2011–2019) and to convert AP threshold pass rates into d-scores using the method of thresholds. Unfortunately, post-George Floyd, College Board changed its policy and ceased reporting Canadian scores by race/ethnicity; as a result, data after 2019 is not available.
Note, for these analyses, I dropped the three studio art scores along with all of the ‘language & culture’ scores, as Warne (2016) found that these tests had low correlations with PSAT scores. (Both the original and modified datafiles are attached; thus if readers wish they can modify the analyses as desired.) Moreover, I computed d-values with respect to the total mean; since Asians score higher, self-identifying White Canadians scored below average. Additionally, I dropped the American Indian group since there were too few individuals to allow for reliable estimates. While Hispanic/Latino is a “visible minority” group in Canada, the number reporting as Hispanic in the Canadian AP datasets seems excessive. This group may include Iberians in addition to Latin Americans.
Summary results are presented below, with the final row reporting weighted-average ACHQ scores. Unlike in the United States, AP exams are not widely taken in Canada, so these results should be interpreted with caution — they represent a few more data points.
Table 1. Canadian 2009-2019 Advance Placement Results by Self-reported race/ethnicity
| Asian | Black | Hispanic | White | |||||||
|---|---|---|---|---|---|---|---|---|---|---|
| N | d | N | d | N | d | N | d | |||
| 2019 | 14061 | 0.172 | 575 | -0.565 | 730 | -0.254 | 10516 | -0.180 | ||
| 2018 | 13282 | 0.181 | 504 | -0.385 | 557 | -0.285 | 10362 | -0.175 | ||
| 2017 | 12258 | 0.189 | 471 | -0.486 | 673 | -0.320 | 10856 | -0.146 | ||
| 2016 | 11291 | 0.172 | 519 | -0.490 | 590 | -0.330 | 10649 | -0.128 | ||
| 2015 | 10037 | 0.179 | 473 | -0.545 | 274 | -0.160 | 10359 | -0.094 | ||
| 2014 | 9619 | 0.221 | 440 | -0.584 | 286 | -0.240 | 10479 | -0.130 | ||
| 2013 | 9680 | 0.210 | 419 | -0.423 | 236 | -0.351 | 10901 | -0.135 | ||
| 2012 | 8887 | 0.213 | 380 | -0.584 | 188 | -0.277 | 10649 | -0.141 | ||
| 2011 | 8152 | 0.218 | 318 | -0.783 | 155 | -0.212 | 10920 | -0.125 | ||
| 2010 | 7172 | 0.233 | 204 | -0.794 | 202 | -0.094 | 10324 | -0.136 | ||
| 2009 | 6772 | 0.205 | 211 | -0.898 | 87 | -0.466 | 9602 | -0.126 | ||
| Ave d. | 0.196 | -0.557 | -0.275 | -0.138 | ||||||
| ACHQ M | 105.01 | 93.72 | 97.95 | 100.00 | ||||||
| N | 111211 | 4514 | 3978 | 115617 | ||||||
For comparison, Warne (2016) reports that, in the United States, the weighted mean AP group differences in 2015 were: d = 0.774 (White–Black), d = 0.484 (White–Hispanic), and d = –0.156 (White–Asian). Thus, relative to the U.S., all non-White groups perform better on AP tests in Canada. For other national-level comparisons using Canadian datasets, see here. As I’ve previously argued, it would be worthwhile to collect more robust data on ethnic differences across a range of countries prior to hypothesizing about causes, rather than relying solely on U.S. data.
The R code and data files can be found here.
References
- Bacic, R., & Zheng, A. (2024). Race and the income‐achievement gap. Economic Inquiry, 62(1), 5–23.
- Barber, M., & Jones, M. E. (2021). Inequalities in test scores between Indigenous and non-Indigenous youth in Canada. Economics of Education Review, 83, 102139.
- Sandiford, P., & Kerr, R. (1926). Intelligence of Chinese and Japanese children. Journal of Educational Psychology, 17(6), 361.
- Vernon, P., (1982). The abilities and achievements of orientals in North America. Academic Press.
- Warne, R. T. (2016). Testing Spearman’s hypothesis with Advanced Placement examination data. Intelligence, 57, 87–95.
The Cayman Islands, a British Overseas Territory, boasts one of the world’s highest standards of living and per capita GDP, largely due to its status as a major tax haven. Its average Human Development Index (HDI) from 2010, 2015, and 2020 was 0.876 (Economic and Statistics Office, 2021), compared to the United Kingdom’s 0.919 over the same period—both rated as very high human development. The 2021 census recorded a population of 68,811, with 53% identified as Caymanian residents, of whom only 62% were born on the island. These demographic shifts, detailed in Figure 1, complicate efforts to estimate ancestry proportions.
Although my colleagues are international, their research predominantly focuses on differences within the USA, a tendency I find perplexing. Years ago, at an LCI conference, a British colleague asserted that further research on ethnic or racial differences was unnecessary, claiming the matter was settled. Yet, within a year, he contacted me seeking data on the performance of various ethnic groups in the UK. Similarly, a French colleague, whose work almost exclusively examines differences in the USA, argued that such data was unavailable in France. However, after just ten minutes of searching, I located several French government reports detailing the performance of first- and second-generation immigrants by region of origin. Evidently, he didn’t try too hard. In my view, a logical approach, if one were to pursue this topic seriously, would involve first collecting a broad international dataset—potentially requiring effort such as submitting freedom of information requests—then identifying patterns, and finally hypothesizing about possible causes. After all, the USA is not the world, and we should not assume it reflects global trends. The lack of such comprehensive data is why I remain largely agnostic about the existence of worldwide race differences in cognitive ability, and even more so about their causes
. . .
France governs three overseas departments in the Americas—French Guiana, Guadeloupe, and Martinique—along with three overseas American collectivities: Saint Barthélemy (St. Barth), Saint Martin, and Saint Pierre and Miquelon. The populations of these territories display a rich diversity of ancestries, making them an interesting case study.
Saint Pierre and Miquelon, located off the coast of Canada, has a population of approximately 5,819 and is the most European-influenced of France’s overseas territories. Its residents are primarily descendants of French settlers from Normandy, Brittany, and other regions, supplemented by recent migrants from metropolitan France. Saint Barthélemy, a small Caribbean island that separated from Guadeloupe in 2007, has a population of around 10,000. Although France does not officially collect ethnic or racial data, visual observations of children in local primary schools and at festivals suggest the population is roughly 80% of European ancestry (mostly French), 15% of African ancestry, and 5% of other origins. This aligns with historical settlement records, which indicate a majority of French descent.
Jason Malloy previously wrote lengthy blog posts summarizing IQ and admixture data for both Puerto Rico and the U.S. Virgin Islands. The achievement data, which is the focus of this series of posts, was outdated. Here, we provide an update.
Puerto Rico
Puerto Rico is a predominantly Spanish-speaking U.S. territory. Its residents have full citizenship—allowing free movement to the mainland. Currently, more Puerto Ricans live on the mainland (5.8 million in 2023) than on the island (3.2 million). Based on the average of 16 samples, individuals on the island of Puerto Rico have an average European, African, and Amerindian ancestry of 66.67%, 19.80%, and 13.53%, respectively. The samples are summarized in Table 1 below.
According to Matthews, Tabery, & Turkheimer (2025), our Admixture in the Americas project is “abhorrent” because nothing good could come from investigating “alarming hypotheses regarding the genomic basis of differences in cognitive abilities between racial and ethnic groups.” While Eric Turkheimer’s position is predictable given his well-known motivations, it is surprising to see philosophers like James Tabery also resorting to sloppy, moralistic reasoning. Notably, in the two Admixture in the Americas papers cited by Matthews et al. (2024), we explicitly stated that our research does not test a genetic model. Instead, drawing on Putterman and Weil’s Post-1500 Population Flows and the Long-Run Determinants of Economic Growth and Inequality and a large body of “deep roots” economic literature, we examined a genealogical model, recognizing that various factors could explain the intergenerational transmission of differences. In The Genealogy of Differences (2016), responding to a critique from Ibarra (2016), we elaborated:
According to [our model], intergenerationally transmitted factors such as genes, epigenes and culture code for individual-level traits related to individuals’ ability to acquire knowledge and to develop better societies (e.g., a cultural appreciation of education and learning affecting the development of cognitive abilities). By this model, BGA acts as a crude index of the lines of descent along which the individual-level traits, the true causal factors, are passed… to deny a priori the possibility of our model, Ibarra equates an intergenerational model with a behavioral genetic one and then, incredibly, adopts a Blank Slate position. This is, of course, a doubly absurd argument. First, we stipulated that our model was an intergenerational, or genealogical, transmission, and not necessarily a behavioral genetic one. We made that point in three separate sections of the [original] paper.
Emil Kirkegaard requested an update to the Admixture in the Americas analyses, a complex task requiring, among other things, the computation of new national cognitive scores for countries and territories in the Americas. Three existing datasets are available for this purpose:
- Antilok et al.’s Harmonized Test Score (HTS) Dataset: Used to create the World Bank’s Harmonized Test Scores (HTS), this dataset covers all but three sovereign countries—Barbados, the Bahamas, and Suriname. Scores were averaged over 5-year intervals from 2000 to 2020, except for Bolivia, where data from 1995–2000 was used.
- Becker’s 2023 National IQ Scores: These measure a construct distinct from achievement scores and rely primarily on convenience samples.
- GMAT-Based Scores: Previously utilized in an earlier paper, these scores have been updated. They cover 2000–2020 data, except for small territories (e.g., Aruba, British Virgin Islands, Guadeloupe, Martinique, Montserrat, and Netherlands Antilles), where 1980s–2020 data was included due to limited 2000–2020 sample sizes and the absence of a clear secular trend.
Academic papers’ abstracts are sometimes manipulated to display what is seemingly a good result. Alternatively, abstract conclusions are affected by faulty statistical analyses. The large profits in online publishing, especially the open-access (OA) type, are largely (but unfairly) criticized, with endless calls for regulation. In reality, the problem of monopolistic prices and quasi-monopolistic status of the most prestigious online publishers is not due to lack of regulation, but due to a combination of copyright law, OA mandates, and moral hazard driven by subsidies.
