THE NIQ-DATASET V1.3 – A Summary (Part – II)

This is a continuation. Enumerations of tables and figures continuous from Part – I. Explanations for abbreviations etc. could already be in Part – I. For Part – I and – II, the same dataset was used.

Correlation analysis

Simply comparing means does not let us analyze important deviations. Therefore, this section will discuss results from correlation analyses, at first in Table 3 between the different estimations at the cross-national level. The national IQs [DB] show strong positive correlations with all those from Lynn and Vanhanen, with the highest value in case of L&V12 (r=.89; N=108; p<.001). A look in the scatterplot in Fig.3 shows that this relationship is linear. Using a second grade polynomial formula increases r by <.01. IQs from Lynn and Vanhanen are a bit higher for low-IQ nations and lower for high-IQ nations. This might be due to DB’s intensified use of IQ conversions between different Raven’s Matrices forms, which would also explain the higher standard deviation of IQ[DB] relative to IQ[L&V12] in Table 1 and 2 (Part – I).

Table 3: Correlation matrix between national IQs by different estimations. Underlined relations in scatterplots. Variable “DB & L&V12” is from unweighted means of both estimates or from the only available estimate when there was only one present. For all correlations, p<.001.

Fig.3:   Scatterplot between two estimations of national IQs (DB × L&V12; r(linear)=.89; N=108; p<.001).

Table 4: Correlation matrix between national IQs[DB] from named tests only. Underlined relations in scatterplots. *p<.001 (N varies). RPM=Raven’s Progressive Matrices (all); SPM=Raven’s Standard Progressive Matrices; SPM+=Raven’s Standard Progressive Matrices Plus; CPM=Raven’s Coloured Progressive Matrices; APM=Raven’s Advanced Progressive Matrices; WIS= Wechsler Intelligence Scales (all); WISC-R=Wechsler Intelligence Scale for Children revisited ed.; WISC-III=Wechsler Intelligence Scale for Children 3rd ed.; CFT=Culture Fair Test.

The correlations between the national IQs [DB] from different tests are strongly positive, especially between tests from the same type (Raven’s Matrices or Wechsler Scales). But between different types, they decrease to a low of .26 (APM × SPM+; N=7; p=.571) or .15 (CPM × CFT; N=10; p<.676). This might be due to error as a consequence of low sample sizes of nations in which estimates from these different tests or test-types are available. Between the two most frequently used test-types, RPM and WIS, the correlation stays strong with r=.76 (N=23; p<.001), also shown in Fig.2. But larger discrepancies between the RPM and WIS can be seen in some cases; the most marked is Egypt (EGY), the red datapoint on Fig. 4. The low WIS score of 43.54 came from a study by Wachs et al. (1995) from a sample of rural children with low SES. This source was rechecked during the writing of this blogpost but no errors in the calculation of IQs could be found. The same source gives a CPM raw score of 14.18 equivalent to a corrected IQ of 56.87 for 8.50y olds, a bit higher than the IQ from WISC-R but also very low for a North African or Middle Eastern nation. The WIS score was the only available such score for Egypt; however, there were other available RPM scores, and the RPM estimate for Egypt as a whole is 82.65. Such outliers will be discussed later, in a special blogpost.

Fig.4:   Scatterplot between estimations of national IQs[DB] from two different test-types (RPM × WIS; r(linear)=.76; N=23; p<.001). Red is strongest outlier EGY (M=82.65 vs. 43.54).

A correlation analysis at the cross-sample level allows one to observe relations between IQs from samples and some other properties of the samples. Line one in Table 5 shows that IQs are not significantly related to mean age or size of the samples, and also not to the amount of test-time adjustment (r(linear)=.05; N=566; p=0.435). Such adjustments are necessary to correct estimations of IQs for the FLynn-Effect in the country the used test was standardized (more in later posts). It is one of the most error-prone parts of the methods and criticized by other scientists (Wicherts, Dolan & van der Maas, 2010; Wicherts et al., 2010). But the weak correlation suggests that the adjustments lead to very little error.

Table 5: Correlation matrix between sample IQs and sample characteristics. Underlined relations in scatterplots. *p<.001.

Fig.5:   Scatterplot between estimations of sample IQs [DB] and test-time adjustments; r(linear)=.05; N=566; p=.435).

There is a moderate and significant correlation (r=-.21; N=566; p<.001) between sample IQs and years of measurement, together with a weak but also significant correlation (r=-.13; N=566; p<.001) between sample IQs and the year of test standardization. Both sample properties are in turn strongly positively related (r=.48; N=566; p<.001), possibly because more recent studies mostly use more recent standardized tests. A closer look at the correlation between sample IQs and the year of measurement (Fig.6) shows that also an U-shaped regression-line would not increase the correlation noticeably (r(square)=.22).

Fig.6:   Scatterplot between estimations of sample IQs [DB] and years of measurement; r(linear)=-.21; N=578; p=.435).

Once again, IQs from different estimations [DB; L&V] are very strongly linearly related, as shown in Fig.7 (r(linear)=.87; N=566; p<.001). One extreme outlier was found in this analysis, which is a sample for India (IND) with an IQ[DB] of 120.22 but an IQ[L&V] of 77.00, and the second matches much better to other scores found for this nation (M=81.94 than the first. The sample is from a publication of Gandhi-Kingdon (1996), consisting of government aided children of around 13.50y from urban areas, and the IQs were once again checked as a result of this finding. The source gives two additional samples with IQs [DB] of 70.08 (N=252) and 74.78 (N=290). The third sample, named “PUA” (private unaided) within the source, differs from both others by having much higher scores for education (self and family) and SES, which makes higher scores in Raven’s Tests expectable. The mean age of the PUA sample is 13.60 and the CPM raw score 36.03, converted to a SPM raw score of 52.91 because the age range of the CPM norms (4.00 to 11.00y) does not extend to the PUA samples’ age. This is equivalent to a corrected IQ of 120.22. An error could not be detected and is eventually hidden in inaccuracies of the CPM to SPM conversionLynn and Vanhanen did not split the samples and reported the score of 77.00 for all 928 individuals. By using the same procedure, a corrected IQ [DB] of 88.36 was estimated, which is still 11.36 points above their estimate, but at least closer. As mentioned before, this outlier will be discussed together with some others in a special blogpost later.

Fig.7:   Scatterplot between two estimations of sample IQs [DB × L&V]; r(linear)=.87; N=566; p<0.001).

The last scatterplot in Fig.8 shows the relationship between sample IQs [DB] and the absolute value differences between both estimates of sample IQs [DB; L&V] (r=-.26; N=566; p<.001). The differences were much greater for low-IQ samples and a U-shaped regression line would show a correlation (r=.42) twice stronger than a linear one. This might be due to data quality or to the use of extrapolated IQs from conversion formulas. So far, it is the only pattern for a systematic deviation between both estimations and definitely requires further investigation.

Fig.8:   Scatterplot between sample IQs [DB] and absolute differences between two estimations of sample IQs [DB × L&V]; r(linear)=.87; N=566; p<.001).

What’s next?

The further course of this project will follow three major goals. At first, analyses should be done to uncover the causes of deviations between the different estimations of national IQ as well as to find eventual methodological variables that might be responsible for variations in IQs. The second aim is a detailed explanation of the methods used to estimate national and sample IQs (DB), to make all steps and decisions easier to understand and transparent for critics. And last, a sample quality index should be created to evaluate the goodness and validity of the national IQ dataset (both in its entirety and for specific samples), and to evaluate the impact of data quality on estimated national and sample IQs.

 

Sources and References (Part – I and II)

Central Intelligence Agency (CIA) (2017). The World Fact Book. From: https://www.cia.gov/library/publications/the-world-factbook/

Gandhi-Kingdon, G. (1996). The quality and efficiency of private and public education: A case-study of urban India. Oxford Bulettin of Economics and Statistics, 58, 57-82. doi:10.1111/j.1468-0084.1996.mp58001004.x

Lynn, R., & Vanhanen, T. (2002). IQ and the Wealth of Nations. Westport, CT: Praeger.

Lynn, R., & Vanhanen, T. (2012). Intelligence: A Unifying Construct for the Social Sciences. London: Ulster Institute for Social Research.

Wachs, T. D., Bishry, Z., Moussa, W., Yunis, F., McCabe, G., Harrison, G., … & Shaheen, F. (1995). Nutritional intake and conext as predictors of cognition and adaptive behavior of Egyptian school children. International Journal of Behavioral Development, 18, 425-450. doi:10.1177/016502549501800303

Wicherts, J. M., Dolan, C. V., & van der Maas, H. L. J. (2010). A systematic literature review of the average IQ of sub-Saharan Africans. Intelligence, 38, 1-20. doi:10.1016/j.intell.2009.05.002

Wicherts, J. M., Dolan, C. V., Carlson, J. S., & van der Maas, H. L. J. (2010). Raven’s test performance of sub-Saharan Africans: Average performance, psychometric properties, and the Flynn Effect. Learning and Individual Differences, 20, 135-151. doi:10.1016/j.lindif.2009.12.001

THE NIQ-DATASET V1.3 – A Summary (Part – I)

A new version of the THE NIQ-DATASET (V1.3) is nearing completion. It will be based largely on the same data that already underlie V1.2, but will contain a large amount of new statistical analysis. This blog post summarizes the most important findings. It presents comparisons between different estimates,––such as between Lynn and Vanhanen’s 2002 and 2012 estimates and between the combined results of different IQ tests –– and addresses the most striking outliers. However, a deeper investigation of the reasons for deviations or relationships between the different estimates is planned for later posts. Please keep also in mind that due to some new data and changes in methods, the numbers and findings presented here may differ from those in V1.2.

Some explanations in advance: Abbreviations are used to specify different estimations and tests. The abbreviation “L&V” results from the surnames of Richard Lynn and Tatu Vanhanen, and the numbers “02” and “12” represent the years in which Lynn and Vanhanen’s estimates were published (Thus, L&V02 indicates the national IQ estimates published by Lynn and Vanhanen in 2002).  “DB” indicates estimates made by me, David Becker. A “+” always marks estimations which include national IQs calculated not only directly from psychometric measurements but from geographic mean IQs from neighboring nations. Thus, L&V02+ indicates the national IQ estimates published by Lynn and Vanhanen in 2002, including nations for which IQs were calculated by averaging the IQs of neighboring nations. Two levels are observed in the analyses: The cross-national level and the cross-sample level, by which numbers at the the cross-national level based on calculations (mostly means) from data at the cross-sample level.

Database

At the time of writing, the database consists of IQs for 203 nations [DB+], of which 125 were calculated from psychometric measurements of intelligence [DB] from 325 scientific sources (see Table 1), and 78 were calculated from geographic means when no data was available for the nations themselves. The World Fact Book of the CIA (2017) currently lists 267 “world entities”; the [DB] estimates include 46.82% of these, and the [DB+] estimates include 76.03%. This includes also such entities as Hong Kong, the Gaza Strip, Puerto Rico, Greenland, and others; these are not “nations” in the sense of wholly autonomous political entities, but they are often listed separately in cross-national comparisons and their IQs should therefore also be of interest.

Sources are mostly publications in scientific periodicals, reports or books. Sources with samples of doubtful origin or methods were not used; neither were sources that were missing necessary data, such as sample size (N of individuals), mean age, geographic origin of a sample and year of test-administration. A huge amount of additional data was collected from the sources for more accurate tracking of methods and eventually necessary corrections. For example, nearly all of the IQs from Raven’s Progressive Matrices were recalculated from raw scores to avoid irregularities caused by different standardizations. Some of the sources gave more than one sample, and some samples were split into more homogeneous sub-samples to allow specific analyses.

Overall, there are 566 samples included (one overall IQ estimate per sample) encompassing 627,098 individuals. The number may seems high; however, it only accounts for <0.01% of the world’s total population of 7,405,107,650 (CIA, 2017). Possible multiple inclusion of individuals in different samples cannot be ruled out. However, care was taken to not include the same sample twice if published in more than one source, and the geographical and temporal distances between the different test-administrations should make such overlaps unlikely.

Table 1: Sample sizes in V1.3 on the national and individual levels. *Number of world entities and world’s total population from CIA (2017).

Descriptive statistics

Fig. 1 and 2 survey the geographic distribution of all national IQ scores [DB; DB+], which pattern is similar to that found by Lynn and Vanhanen (2002; 2012) and described in the earlier post. The purpose of Table 2 is to give a more detailed look into the data at the cross-national level. According to the dataset, the mean IQ weighted by national populations is 88.17 [DB] or 88.16 [DB+]. These means are similar to those found by Lynn and Vanhanen in 2002 (89.30 or 88.00) and 2012 (83.71 or 89.73), and the standard deviations are also similar.

Fig.1: National IQs [DB] from THE NIQ-DATASET (V1.3) (M=84.74; SD=12.79; N=125). Data from psychometric tests only.

Fig.2: National IQs [DB+] from THE NIQ-DATASET (V1.3) (M=82.57; SD=12.50; N=199). Data from psychometric tests and geographic means.

For IQ[DB] and IQ[DB+], the nation with the lowest estimated score is Nicaragua (NIC) with 53.69, an implausible value that should be treated with caution. It does not make sense considering the IQs of nearby and culturally or ethnically similar nations; nor does it fit Nicaragua’s economic and educational status. Unfortunately, Nicaragua is neither included as a case with psychometric measured IQ in one of the L&V estimations, nor into the international school assessment tests PISA or TIMSS, and can therefore not be validated. The highest IQ score for Nicaragua found within a sample is 63.63, which seems more realistic but more should not be speculated here. In the L&V estimations, the nation with the lowest estimated IQ is Equatorial Guinea (GNQ) with 59.00 in 2002 or Malawi (MWI) with 60.10 in 2012. It should be noted, that the score of GNQ was wrongly attributed to GNQ and the value thus a mistake, as noted by Wicherts, Dolan & van der Maas (2010). On the other hand, the East Asian political entities Hong Kong (HKG) and Singapore (SGP) obtained the highest IQ scores in both the present database and the L&V12 database. For Singapore, these are 113.67 [DB; DB+] and 107.10 [L&V12; L&V12+]. For Hong Kong, these are 105.90 [DB; DB+] and … [L&V12; L&V12+]. In 2002, Lynn and Vanhanen estimated an IQ for Hong Kong of 107.00, which is very close to DB’s estimate of 105.90. The last line of Table 2 shows a combination of the two estimations DB and L&V12, for which a simple mean was calculated if scores were available for both variables; otherwise, the only available score was taken.

In lines 2 to 10 a breakdown of IQ[DB] is made for different tests. These are the Raven’s Progressive Matrices (RPM) in their standard (SPM), standard-plus (SPM+), colored (CPM) and advanced (APM) version. For Wechsler Intelligence Scales (WIS), only the two specific tests for children WISC-R (Wechsler Intelligence Scale for Children––Revised, the second edition) and (WISC-III) (the third edition) are listed. At least, the Culture Fair Test (CFT) is also separated. Results from many other tests like the Kaufman Assessment Battery for Children (K-ABC), the Stanford-Binet Intelligence Scales (SBIS), or other Wechsler tests (including the WAIS, the WPPSI, and other editions of the WISC) are included in the dataset but not shown below due to the low number of estimates that use these tests. Overall, separated results are very close to each other, except those from the CFT, which are around 10 IQ points higher. Deviations might be caused by case selection. There are 26 national IQs measured by the CFT from which 22 are for Western or European nations plus Hong Kong. For this reason, the standard deviation of the CFT scores is comparatively low.

Table 2: Descriptive statistics for national IQs by different estimations at the cross-national level. Country-codes: ISO 3166-1 ALPHA-3; *Unweighted means of both estimates or the only available estimate. Only tests with N(nations)≥10 separated: RPM=Raven’s Progressive Matrices (all); SPM=Raven’s Standard Progressive Matrices; SPM+=Raven’s Standard Progressive Matrices Plus; CPM=Raven’s Coloured Progressive Matrices; APM=Raven’s Advanced Progressive Matrices; WIS= Wechsler Intelligence Scales (all); WISC-R=Wechsler Intelligence Scale for Children revisited ed.; WISC-III=Wechsler Intelligence Scale for Children 3rd ed.; CFT=Culture Fair Test.

Similar pattern can be seen in Table 3 for the cross-sample level. Means and standard deviations are close to those from the cross-national level. The differences between the global means from DB and L&V (available for 450 samples up to 2016) is only 0.03, and the mean of the absolute value differences between both estimations is only 5.00 with a standard deviation of 5.15. This means that a majority of 64% of the DB estimations are less than 5 IQ points away from the L&V estimations.

Sources and References (Part – I and II)

Central Intelligence Agency (CIA) (2017). The World Fact Book. From: https://www.cia.gov/library/publications/the-world-factbook/

Gandhi-Kingdon, G. (1996). The quality and efficiency of private and public education: A case-study of urban India. Oxford Bulettin of Economics and Statistics, 58, 57-82. doi:10.1111/j.1468-0084.1996.mp58001004.x

Lynn, R., & Vanhanen, T. (2002). IQ and the Wealth of Nations. Westport, CT: Praeger.

Lynn, R., & Vanhanen, T. (2012). Intelligence: A Unifying Construct for the Social Sciences. London: Ulster Institute for Social Research.

Wachs, T. D., Bishry, Z., Moussa, W., Yunis, F., McCabe, G., Harrison, G., … & Shaheen, F. (1995). Nutritional intake and conext as predictors of cognition and adaptive behavior of Egyptian school children. International Journal of Behavioral Development, 18, 425-450. doi:10.1177/016502549501800303

Wicherts, J. M., Dolan, C. V., & van der Maas, H. L. J. (2010). A systematic literature review of the average IQ of sub-Saharan Africans. Intelligence, 38, 1-20. doi:10.1016/j.intell.2009.05.002

Wicherts, J. M., Dolan, C. V., Carlson, J. S., & van der Maas, H. L. J. (2010). Raven’s test performance of sub-Saharan Africans: Average performance, psychometric properties, and the Flynn Effect. Learning and Individual Differences, 20, 135-151. doi:10.1016/j.lindif.2009.12.001

A Warm Welcome

Scientific conferences often start with a greeting to the participants and audience who have traveled long distances around the globe to communicate with, or just listen to, their colleagues. I would like to adhere to this tradition, even if the Internet has significantly shortened the travel time, because this blog has as its aim to communicate with interested people from around the globe about a topic that has a global scale: The IQ of the nations of the world.

IQ is a parameter for intelligence, which can be defined as a “very general mental capability that, among other things, involves the ability to reason, plan, solve problems, think abstractly, comprehend complex ideas, learn quickly and learn from experience.” It “can be measured, and intelligence tests measure it well” and it is of “great practical and social importance” (Gottfredson, 1994).

A trait so important for a huge variety of individual and social phenomena might be able to explain differences existing in many social, political, economic, cultural, and/or biological areas between nations and geographic regions. However, to observe such associations requires the existence of valid figures for the IQs of the world’s nations. This blog aims to present work on national IQ to an unlimited audience.

Ground already covered

In 2002, a book was published that had both a considerable impact on intelligence research and a substantial public reception. The name of this book was “IQ and the Wealth of Nations“, written by the British psychologist Richard Lynn, Professor Emeritus at the University of Ulster, UK, and the Finnish professor of political science at the University of Tampere, Finland, Tatu Vanhanen (†2015), well known for creating the Vanhanen Index of Democratization (Lynn & Vanhanen, 2002; Lynn, 2017; Vanhanen, 2003). One highlight of this book was a list of average national IQs for 185 nations, of which 81 national IQs were collected and calculated directly from studies, the remaining 104 national IQs being estimated from the IQs of nearby and ethnically comparable countries. This list was expanded and improved over time, first in “Race Differences in Intelligence: An Evolutionary Analysis” (Lynn, 2006) next in “The Global Bell Curve: Race, IQ, and Inequality Worldwide” (Lynn, 2008), and finally in “Intelligence: A Unifying Construct for the Social Sciences” (Lynn & Vanhanen, 2012a).

To calculate the national IQs, Lynn and Vanhanen collected IQ test results from samples that were as representative or applicable as possible of/to their respective national populations, and set these results on a scale with the IQ of the UK at 100. This comparison IQ is often called the Greenwich IQ Mean, after Greenwich Mean Time. Some corrections were needed, in the main to adjust for the FLynn Effect, the increase of IQs ascending over time during the 20th century in many nations (Lynn, 1982; Flynn, 1984; Pietschnig & Voracek, 2015). If more than one measurement of IQ was available for a nation, the mean or median was calculated. IQs for other nations for which no measurements could be found were calculated as means of their adjacent nations.

Fig. 1 is based on the 2012 version and shows how the national IQ levels form geographic clusters, with the lowest scores, between 60 and 80, in Sub-Saharan Africa; middling scores, between 80 and 95, mostly in Arabian or Middle Eastern, South Asian, Oceanian and South American nations; high scores, between 95 and 102, in European nations or those with a population descended mostly from Europe; and finally very high scores, above 102, in Northeast Asian nations.

Fig. 1:  National IQs as calculated or estimated by Lynn and Vanhanen (2012).

Such a non-random global pattern suggests that national IQs might be related to other unequally distributed factors; indeed, a large amount of significant correlates, summarized by Lynn and Vanhanen (2012b) have been found between national IQs and other international variables. Statistical relations were found between national IQs and variables in the fields of education (e.g. results from school assessment, literacy, etc.); cognitive outputs (e.g. scientific publications, patents, Nobel prizes, etc.); economics (e.g. GDP/c in varying designs, economic growth, income, etc.); politics (e.g. democracy, freedom, corruption, etc.); social behavior (e.g. crime, suicide rate, happiness, modernization, etc.); medicine (e.g. infant mortality, HIV rates, fertility, etc.); geography (temperature, latitude); and anthropology (skin color).

In most cases, correlations were positive for those variables for which higher scores are usually considered desirable, and negative for variables for which lower scores are usually considered undesirable. On average, nations with higher IQs had stronger economies; were healthier, wealthier, and more educated; and were more modernized, politically stable, and peaceful (national suicide rates are an exception: suicides are less common in low-IQ nations). As the map suggests, there are strong positive correlations between national IQs and average skin brightness of a nation’s population (.69 to .92), and between national IQ and a nation’s mean latitude (.68 to .72). More recently, variables from population genetics joined in, such as population frequencies of intelligence-associated SNPs (e.g. Piffer, 2013, 2015), Y-chromosomal haplogroups (Rindermann, Woodley, & Stratford, 2012) or genetic distances in general (León & Burga-León, 2015; Becker & Rindermann, 2016).

Hunt and Sternberg (2006) called the list of national IQs “far from ideal” in a review of Templer and Arikawa’s (2006) paper of correlations between national IQs, skin color, and income. They (Hunt and Sternberg) criticized the non-representativeness of the included samples; Lynn and Vanhanen’s failure to weight the studies for sample size; and the practice of estimating some countries’ IQs from the IQs of nearby and ethnically comparable countries.

Malloy (2013-2014) addressed Lynn and Vanhanen’s work in great detail at the Human Varieties blog. A huge amount of different measurements of IQ were collected for 15 nations or insular areas in Southeast Asia or the Caribbean. Some national means were very close to those from Lynn and Vanhanen; others showed large differences, especially Cambodia. I compared the national IQs from both sources (some numbers listed in the Malloy column were calculated by me from Malloy’s data) in Table 1 and got a correlation of .47 (= 12; p > 0.050) with and .79 (= 11; p < 0.001) without Cambodia. On the one hand, these numbers suggest considerable reliability of Lynn and Vanhanen’s results; on the other hand, some ambiguities, errors, and deviations were reported by Malloy, suggesting that Lynn and Vanhanen’s IQs are not as reliable as have sometimes been believed.

Nation IQ Dif.
Lynn and Vanhanen (2012) Malloy (2013-2014)
Bahamas, The 84.00 92.52 -8.52
Bermuda 90.00 92.00 -2.00
Burma 85.00 97.00 -12.00
Cayman Islands n.d. 75.00 n.d.
Cuba 85.00 90.00 -5.00
Dominican Republic 82.00 73.14 8.86
Haiti 67.00 68.00 -1.00
Jamaica 71.00 79.00 -8.00
Laos 89.00 90.60 -1.60
Puerto Rico 83.50 84.60 -1.10
Thailand 89.90 93.90 -4.00
Turks and Caicos Isl. n.d. 89.40 n.d.
U.S. Virgin Islands n.d. 82.20 n.d.
Vietnam 94.00 91.40 2.60
Cambodia 92.00 66.56 25.44
r 0.47
p >0.050
r (KHM* excluded) 0.79
p (KHM* excluded) <0.001

Table 1: National IQs from Lynn and Vanhanen (2012) and Malloy (2013-2014) compared. *KHM = Cambodia.

Notwithstanding all criticism, Lynn and Vanhanen’s dataset remained the only comprehensive cross-national collection of psychometric IQ measurements. In 2015, a group of scientists corresponded about how to make it more accessible to a larger audience and preserve it for future times. As a participant in this correspondence, I welcomed this idea and saw a lot of potential to expand this intention to a comprehensive open access database of global IQs; however, at the time, I could not guess what scale the project would get. It all started with a simple search for digital copies of all the (or as many as possible of the) sources Lynn and Vanhanen had used, and the cataloguing of these digital copies. As we tried to get a deeper understanding of the methods, we obtained more and more IQ data, so that it became possible to re-estimate some of the IQs Lynn and Vanhanen had calculated.

In May 2017, the first version (0.9) of this project was presented at the London Conference on Intelligence (Becker, 2017). This revision included around 340 recalculations of IQs for 115 nations, around half of the nations in the original dataset. At the time, the revision included only IQs measured by Raven’s Matrices but the correlations to Lynn and Vanhanen’s IQs were .95 at the sample level and .83 at the cross-national level (Fig. 2). In the following months, data from other IQ tests such as the Wechsler tests for children and adults (WPPSI; WISC; WAIS; WASI); the Stanford-Binet Intelligence Scale (SBIS); the Cattell Cultural Fair Tests (CFT); and some others, were added to the dataset. On July 9th, 2017, the NIQ-DATASET version 1.0 was born, followed by version 1.1 on October 22nd with improved calculations and corrections.

Fig. 2:  Scatterplot between original national IQs (horizontal axis) and the new national IQs (vertical axis) (= .83; N(nat.) = 92; N(samp.) = 691(L&V)|347(rev.); p < 0.001). Red marks debatable cases.

New insights and old problems made the current version 1.2 necessary, which contains some structural changes. This version has calculated IQs for 124 nations (presented in Fig. 3), based on 543 datapoints. A comparison with the national IQs from Lynn and Vanhanen (2012), which unlike the IQs in the present dataset were partially based on international school assessment tests, showed great similarities but also differences. Using version 1.2, a correlation of .87 (= 445; < 0.001) was calculated between Lynn and Vanhanen’s estimated IQs and my estimated IQs. For the same datasets; a correlation of .86 (= 123; < 0.001) was calculated between Lynn and Vanhanen’s estimated IQs and my estimated IQs for the same nations. Furthermore, I calculated a mean IQ for 124 nations of 84.90 (SD = 12.72) or 88.12 if the nations’ IQs were weighted by population.

Since one of my intentions was to make all the methods for calculations and corrections comprehensible, I started to write a manual, which helped for understanding but turned out to be not suitable to go into all details, to include all ideas, or to help springboard discussions. So, when a post on James Thompson’s blog (Thompson, 2017) was published about this project and gained a lot of attention, praise, and criticism, it gave me the incentive to start a blog of my own. The blog aims to present, explain, and discuss this project in a more detailed way, and to document its progress step by step.

Fig.3:   National IQs from the NIQ-DATASET (V1.2).

About the Dataset and Blog

This blog is divided into two sections.

(1) The database section contains the current, as well as all earlier, versions of the National IQ dataset, starting with V1.1. The folders contain the main dataset in an Excel file together with a few notes about methods and updates, and also some additional and changing material, like maps, graphs, etc.

(2) The blog will be used to take a closer look at important issues, findings, and thematically related things, and to allow open discussion. Posts will be published irregularly to document the current state of the national IQ project.

Everything here should follow the principle of an open source project, and not a single number or calculation should be left in the “black box”. Another principle is the use of standardized methods. Special methods for individual cases should be avoided. All uploaded data are available for free use. All this is intended to guarantee clarity and traceability. I hope to meet the needs of interested people as much as possible, and I look forward to readers’ comments and discussions.

With best regards,

David Becker

Sources and References

Becker, D. (2017). National IQs revisited: The first steps of a long-term project. London Conference on Intelligence, 12-14 May 2017.

Becker, D., & Rindermann, H. (2016). The relationship between cross-national genetic distances and IQ-differences. Personality and Individual Differences, 98, 300-310. doi:10.1016/j.paid.2016.03.050

Flynn, J. R. (1984). The mean IQ of Americans: Massive gains 1932 to 1978. Psychological Bulletin, 95, 29-51. doi:10.1037/0033-2909.95.1.29

Gottfredson, L. S. (1994). Mainstream science on intelligence: An editorial with 52 signatories, history, and bibliography. Intelligence, 24, 13-23. doi:10.1016/S0160-2896(97)90011-8

Hunt, E., & Sternberg, R. J. (2006). Sorry, wrong numbers: An analysis of a study of a correlation between skin color and IQ. Intelligence, 34, 131-137. doi:10.1016/j.intell.2005.04.004

León, F. R., & Burga-León, A. (2015). How geography influences complex cognitive ability. Intelligence, 50, 221-227. doi:10.1016/j.intell.2015.04.011

Lynn, R. (1982). IQ in Japan and the United States shows a growing disparity. Nature, 297, 222-223. doi:10.1038/297222a0

Lynn, R. (2006). Race Differences in Intelligence: An Evolutionary Analysis. Augusta, GA: Washington Summit Publishers.

Lynn, R. (2008). The Global Bell Curve: Race, IQ, and Inequality Worldwide. Augusta, GA: Washington Summit Publishers.

Lynn, R. (2017). Personal Homepage. From: http://www.rlynn.co.uk/

Lynn, R., & Vanhanen, T. (2002). IQ and the Wealth of Nations. Westport, CT: Praeger.

Lynn, R., & Vanhanen, T. (2012a). Intelligence: A Unifying Construct for the Social Sciences. London: Ulster Institute for Social Research.

Lynn, R., & Vanhanen, T. (2012b). National IQs: A review of their educational, cognitive, economic, political, demographic, sociological, epidemiological, geographic and climatic correlates. Intelligence, 40, 226-234. doi:10.1016/j.intell.2011.11.004

Malloy, J. (2013-2014). HVGIQ. From: http://humanvarieties.org/

Piffer, D. (2013). Factor analysis of population allele frequencies as a simple, novel method of detecting signals of recent polygenic selection: The example of educational attainment and IQ. The Mankind Quarterly, 65, 168-184.

Piffer, D. (2015). A review of intelligence GWAS hits: Their relationship to country IQ and the issue of spatial autocorrelation. Intelligence, 53, 43-50. doi:10.1016/j.intell.2015.08.008

Pietschnig, J., & Voracek, M. (2015). One century of global IQ gains: A formal meta-analysis of the Flynn Effect (1909–2013). Perspectives on Psychological Science, 10, 282-306. doi:10.1177/1745691615577701

Rindermann, H., Woodley, M. A., & Stratford, J. (2012). Haplogroups as evolutionary markers of cognitive ability. Intelligence, 40, 362-375. doi:10.1016/j.intell.2012.04.002

Templer, D. I., & Arikawa, H. (2016). Temperature, skin color, per capita income, and IQ: An international perspective. Intelligence, 34, 121-139. doi:10.1016/j.intell.2005.04.002

Thompson, J. (2017). The World’s IQ = 86. Test results of 550,492 individuals in 123 countries. From: http://www.unz.com/jthompson/the-worlds-iq-86/

Vanhanen, T. (2003). Democratization: A comparative analysis of 170 countries. London; New York: Routledge.

Wicherts, J. M., Dolan, C. V., & van der Maas, H. L. J. (2010). A systematic literature review of the average IQ of sub-Saharan Africans. Intelligence, 38, 1-20. doi:10.1016/j.intell.2009.05.002

Wicherts, J. M., Dolan, C. V., Carlson, J. S., & van der Maas, H. L. J. (2010). Raven’s test performance of sub-Saharan Africans: Average performance, psychometric properties, and the Flynn Effect. Learning and Individual Differences, 20, 135-151. doi:10.1016/j.lindif.2009.12.001