Press question mark to see available shortcut keys

Intelligence: A Unifying Construct for the Social Sciences, Lynn & Vanhanen 2012; Book excerpts:

Although the contribution of intelligence to the understanding of many social phenomena has been known for several decades, such is the compartmentalization of the social sciences that this has rarely been recognized. Thus, in sociology, James Coleman's Foundations of Social Theory (1990) has been described as "the most important book in social theory in a long time" by the Nobel prize- winning economist Gary Becker, yet in its 950 pages no mention is made of intelligence. In criminology, the significance of low intelligence as a factor in crime has been largely ignored. Wilson and Herrnstein (1985, p.155) observed a quarter of a century ago "Despite over forty years of confirmation, the correlation between intelligence and crime has yet to penetrate most of the textbooks or the conventional wisdom of criminology". Nothing had changed in the 1,246 page Oxford Handbook of Criminology that contains no mention of intelligence (Maguire, Morgan and Reiner, 1994). In epidemiology, numerous studies have shown socio-economic correlates of health such as mortality, obesity, accidents, lung cancer and stroke, but there has been virtually no recognition that much of these can be explained by intelligence until the recent work by a group of psychologists led by Ian Deary (Deary, Whalley and Starr, 2009).

Ten years ago we began our research program for the investigation of how far differences in intelligence can explain the differences in economic development between nations in our book IQ and the Wealth of Nations (2002). Our starting point was that it has been established that intelligence is a determinant of earnings among individuals, and hence that this association should also be present across nations. We searched for studies throughout the world in which intelligence tests had been administered, and found usable data for 81 nations. We calculated the results by setting the IQ in Britain at 100 (standard deviation =15) and the IQs of other nations were expressed on this metric. The results showed that there are huge differences in the average IQs of nations, ranging from approximately 70 in sub-Saharan Africa, to approximately 100 in most of Europe and the countries colonized by Europeans in the last few centuries (the United States, Canada, Australia, New Zealand, Argentina, Chile and Uruguay), to approximately 110 in China, Japan, Korea, Singapore and Taiwan. We then showed that national IQs were correlated with per capita income (measured as real GDP, gross national product, per capita) at 0.73 (Lynn and Vanhanen, 2002, p. 89). This showed that 53 per cent of the variance in per capita in this group of nations is attributable to differences in intelligence. We then used the measured IQ of the 81 nations to estimate the IQs for a further 104 nations that were ethnically similar to those for which we had measured IQs. For example, we assumed that the IQ in Luxembourg would be the same as in the Netherlands and Belgium. This gave us IQs for all 185 nations in the world with populations over 50,000. We showed that for these 185 nations, IQs were correlated with per capita income (measured as real Gross Domestic Product, per capita) at 0.62. This is lower than the correlation for 81 nations, probably because there was some error in the estimated IQs. Nevertheless, the correlation is highly significant and shows that 38 per cent of the variance in per capita income in the nations of the world is attributable to differences in intelligence. To establish the validity of these national IQs, we showed that they are correlated at 0.88 with national scores on tests of mathematics and at 0.87 with national scores on tests of science. In 2006 we published further evidence for this theory in our book JQ and Global Inequality. In this we presented measured IQs for an additional 32 nations, bringing the total number of nations for which we had measured IQs to 113. We showed that these were correlated with per capita income (measured as real GNI, gross national income) at 0.68 (Lynn and Vanhanen, 2006, p. 102). Following the method in our first study, we used the measured IQ of the 113 nations to estimate the IQs for an additional 79 nations that were ethnically similar to those for which we had measured IQs. This gave us a total of 192 nations, comprising all the nations in the world with populations over 40,000. We found a correlation of0.68 between national IQ and per capita income in the 113 nations for which they had measured IQs, and a correlation of0.60 between national IQ and per capita income in the 192 nations. Once again, the correlation for the 113 nations' measured IQs is a little higher than for the larger 192 nation data set, and probably for the same reason that measured national IQs are more valid than estimated national IQs. In our 2006 book we extended the analysis beyond economic development and showed that national IQs explain substantial percentages of the variance in national differences a number of other phenomena including literacy, life expectancy, and the presence of democratic institutions.

The complete list of several hundred studies of national IQ is given in Appendix 1. These have been calculated in relation to a British mean of 100 and standard deviation of 15. IQs have been increasing in all economically developed countries for which information is available from 1918 up to around the year 2000 and also in at least some of the less developed countries including Brazil, Sudan and Dominica (Colom, Flores-Mendoza and Abad, 2006; Khaleefa, Sulman and Lynn, 2009; Meisenberg, Lawless, Lambert and Newton, 2006). These increases have become known as the Flynn effect. We have dealt with this by adjusting national IQs for the year in which they were obtained. The adjustments made for these secular increases in IQ are 3 IQ points per decade calculated for the United States by Flynn (1984) for all tests except the Progressive Matrices, for which they are 2 IQ points per decade reported for Britain for the years 1938-1979 calculated by Lynn and Hampson (1986) and for the years 1979-2008 for children up to the age of 13 years, but not for those older than this, calculated by Lynn (2009). For example, if a study carried out in 1969 with the Progressive Matrices gives a country a British IQ of 90 on the 1979 British standardization, 2 IQ points are added because the British IQ in 1969 was 2 IQ points lower than 1979.

The school achievement data are on an approximate 500/100 scale, as shown in the column labeled SchAch. Two methods were used to convert these scores to the IQ metric. The column labeled SA direct is a direct transformation to the IQ metric that brings the score of the United Kingdom to 100 and its within- country standard deviation to 15. For the 87 countries having both IQ and school achievement scores, the between-countries standard deviation of these direct-transformed scores is 28% higher than the standard deviation of IQ. This means that school achievement is more "culturally biased" than IQ in the sense that between-country differences, relative to within-country differences, are larger for school achievement than for IQ. For this reason, school achievement scores scaled directly to the IQ metric are not suitable for creating a combined measure of a Final IQ calculated from school achievement and intelligence test results. Such a measure would give an artefactual advantage to low-scoring countries that have only IQ data relative to low- scoring countries that have only school achievement data.

Several critics of the national IQs given in our previous studies have asserted that the IQs obtained in different studies from the same countries are inconsistent and therefore that the IQ figures have poor reliability. For instance, Astrid Ervik (2003, p. 408) wrote that there are "large disparities in test scores for the same country" and "the authors fail to establish the reliability of intelligence (IQ) test scores". A similar criticism has been made by Susan Barnett and Wendy Williams (2004): "When more than one sample is used to estimate a national IQ, it is unsettling how great the variability often is between samples from the same country". The reliability of a psychometric test means the extent to which the score it provides can be replicated in a further study. The reliability of a test is best assessed by making two measurements of an individual or set of individuals and examining the extent to which the two measurements give the same results. Where the two measurements are made on a set of individuals the correlation between the two scores gives a measure of the degree to which they are consistent and is called the reliability coefficient. In our IQ and the Wealth of Nations we examined the reliability of the measures by taking 45 countries in which the intelligence of the population has been measured in two or more investigations. This is the same procedure that is used to examine the reliability of tests given to sets of individuals. We reported that the correlation between two measures of national IQs is 0.94, showing that the measures give high consistent results and have high reliability. This reliability coefficient is closely similar to that of tests of the intelligence of individuals, which typically lies in the range between 0.85 and 0.90 (Mackintosh, 1998, p. 56). In our IQ and Global Inequality we examined the consistency of the IQs for 65 countries for which there were two or more scores. The correlation between the two extreme IQs (i.e. the highest and lowest) was 0.93 and is highly statistically significant. This method underestimates the true reliability because it uses the two extreme values. As an alternative method we excluded the two extreme scores and used the next lowest and highest scores. There were 13 countries for which we had five or more IQ scores (China, Congo-Zaire, Germany, Hong Kong, India, Israel, Jamaica, Japan, Kenya, Morocco, South Africa-blacks, South Africa-Indians, and Taiwan). Using this method, the correlation between the two scores was 0.95. These figures established that the national IQs used in our earlier work had high reliability.

Critics have also asserted that our national IQs lack validity. For instance, Ervik (2003) has written that we fail to establish the cross-cultural comparability (i.e. validity) of intelligence and Barnett and Williams (2004) who argues that the tests are not valid measures of the intelligence of peoples in many economically developing nations. More recently, Hunt (2011, p. 439) has written that "Lynn and Vanhanen disregard any question about the validity of various intelligence tests across different countries and cultures". Contrary to these assertions, we have gone to considerable trouble to demonstrate that our national IQs are valid. The validity of an intelligence test is the extent to which it measures what it purports to measure and is established by showing that it is highly correlated with other measures of cognitive ability. Foremost among these is educational attainment. As noted in section 2 above, at the level of individuals, intelligence and educational attainment are typically correlated at between 0.5 and 0.8. We have demonstrated that our national IQs are valid by showing that this association is also present at the national level. In our first book, we showed that our national IQs are correlated with national scores on mathematics at 0.881 and with national scores on science at 0.868 (Lynn and Vanhanen, 2002, p. 71). In our second book, we showed that our national IQs are correlated with national scores on mathematics scores obtained by 15 year old school students in PISA 2000 at 0.876 and with national scores on science obtained in PISA 2000 at 0.833 (Lynn and Vanhanen, 2006, p. 69). We have confirmed these high correlations in subsequent studies with larger data sets and shown the correlations between the results of national IQ tests and scholastic assessments are in the vicinity of 0.9 (Lynn and Mikk, 2007; Lynn, Meisenberg, Mikk and Williams, 2007). These results have been confirmed by Rindermann (2007). In a later study of 108 nations, we have shown that national scores aggregated from the PISA and TIMSS studies are perfectly correlated with national IQs (r=1.0) (Lynn and Meisenberg, 2010). To examine further the association between national IQs and school achievement scores, the correlation between these (given in Table 2.1) is 0.907 for the 87 countries having both measures, as noted in Section 4. This confirms our numerous previous studies showing that national IQs and school achievement scores are measures of the same latent construct of cognitive ability of intelligence.

It will be seen that all the correlations between intelligence and subsequent educational attainment are substantial and lie in the range between 0.40 and 0.74. The median of the eight studies is 0.61. It has sometimes been argued that the correlation between intelligence and educational attainment is not a causal one but arises through the common effects of the socio-economic status of the family on both intelligence and educational attainment. Thus, middle class families produce children with high intelligence, either through genetic transmission or by providing environmental advantages, or both, and also ensure that their children have a good education. This explanation cannot be correct because the correlation between parental socio-economic status and their children's educational attainment obtained from a meta-analysis of almost 200 studies is only 0.22 (White, 1982). Such a low correlation could not account for much of the higher association between children's IQs and their educational attainment. In addition, it has been found that among pairs of brothers brought up in the same family, there is a correlation of approximately 0.3 between IQ and educational attainment (Jencks, 1972). This shows that, even when family effects are controlled, the correlation between IQ and educational attainment remains, although it is reduced. The only reasonable explanation of the correlations shown in Table 3.1 is that intelligence has a direct causal effect on educational attainment.

We now consider the contribution of intelligence to cognitive attainment defined as attainment for which a high IQ is a major necessary condition such as the publication of papers in academic journals, obtaining Nobel prices and so forth. Studies that have examined the relation of national IQs to a variety of measures of cognitive attainment are summarized in Table 3.3. Row 1 gives a correlation of 0.87 between national IQs and academic publications measured as numbers of papers per capita published in academic journals, based on 137 nations. Row 2 gives a correlation of 0.51 between national IQs and the patent index: measured as the number of patents granted in the USA per million population, based on 112 nations. The author of this study adopts the patent index as a measure of a nation's technological achievement, and "technological achievement mediates the relationship between IQ and wealth; in other words, high IQ nations generate more technical knowledge, which in turn leads to more wealth" (Gelade, 2008, p. 712). Row 3 gives a correlation of 0.63 between national IQs and "intellectual autonomy" based on 63 nations. This construct is defined as follows: "in cultures that emphasise intellectual autonomy individuals are encouraged to create and innovate, and to pursue their own ideals" (Gelade, 2008, p. 172). The author predicted that cultures that value intellectual autonomy should have high production of patents, which in turn promotes economic development. This prediction was confirmed by the correlation of 0.71 between intellectual autonomy and per capita income. Row 4 gives a correlation of 0.74 between national IQs and STEM, a measure of scientific and technological excellence, based on 90 nations. Row 5 gives a correlation of 0.40 between national IQs and patents per capita granted during 1960-2007, based on 76 nations. Rows 6, 7 and 8 give the correlation of between national IQs and Nobel prizes awarded per capita (1901-2004) for literature (0.13), peace (0.21) and science (0.34), based on 97 nations. It may be surprising that the correlation with literature is as low as 0.13 and is not statistically significant. The reason for this is that the Nobel Committee has not been good at picking works of literature that have endured. Who now reads or has even heard of the first literature Nobel prizewinners Sully Prudhomme (1901), Theodor Mommsen (1902), Frédéric Mistral (1904) and Giosuè Carducci (1906). Yet remarkably the prize was not awarded to Leo Tolstoy who did not die until 1910.

Row 9 gives a correlation of 0.61 between national IQs and the numbers of scientists and engineers working in research, per capita, based on 51 nations. Row 10 gives a correlation of 0.38 between national IQs and technology exports as percentage of all manufactured exports, 1997, based on 61 nations. Row 11 gives a correlation of 0.36 between national IQs and the cognitive ability of politicians 1990-2009 estimated from their educational qualifications, based on 90 nations.

However, there are serious shortages in the comparability of data in several cases. Some large deviations may be more due to significant differences in the criteria of tertiary [higher] education than to real differences in the extent of tertiary education. For example, it is highly improbable that tertiary education is two times more extensive in Greece (95%) than in Switzerland (46%). Unfortunately national differences in the criteria of tertiary education and the use of estimated data weaken the reliability of data on Tertiary-09. ...For 17 of these 20 countries, national IQ is above 90. The other five countries (Lebanon, Libya, Panama, St Lucia and Venezuela) are dispersed around the world without any common characteristics. The group of large negative outliers (residual -18.0 or higher) includes the following 17 countries: Andorra, Brunei, Cambodia, China, Hong Kong, Iraq, Laos, Madagascar, Mauritius, Myanmar, Pakistan, Papua New Guinea, Samoa, Suriname, Tonga, Vanuatu and Vietnam. Asian and Oceanian countries (12) dominate in the group of large negative outliers as clearly as European countries in the group of large positive outliers. Another clear difference between the two groups of large outliers is that for most positive outliers national IQ is higher than 90, whereas national IQ varies between 80 and 90 for most negative outliers. National IQ level of 90 seems to constitute a threshold above which tertiary education starts to rise. Three of these 17 countries are socialist or former socialist countries (China, Laos and Vietnam). China has the highest negative residual in the world (-46.0), which implies that human potential for the extension of tertiary education is enormous in China. Cambodia, Iraq, Myanmar and Pakistan have suffered from serious civil wars. The small size of population and/or isolated geographical position may have hampered the extension of tertiary education in countries like Andorra, Brunei, Mauritius, Samoa, Suriname, Tonga and Vanuatu. It is more difficult for very small countries to provide tertiary education than for more populous countries. This concerns particularly isolated small island states. Papua New Guinea and Madagascar are also island states.

Unfortunately data on R&D are available only from 97 countries, and countries with low national IQs (below 80) are underrepresented in the sample. The Pearson correlation between national IQ and R&D is 0.666 (N=97) and Spearman rank correlation considerably higher (0.828). Empirical evidence supports the hypothesis strongly. However, national IQ does not need to be the only factor which explains variation in the R&D variable. It can be assumed that per capita income, democratization, and the level of tertiary education are able to raise the explained part of variation in R&D independently from national IQ. When national IQ, PPP-GNI-08, ID-08, and Tertiary-9 are used to explain variation in R&D, the multiple correlation rises to 0.795 (N=96) and the explained part of variation to 63 per cent, which is 19 percentage points more than national IQ explains (44%). National IQ remains as the dominant explanatory factor, but the three environmental variables raise the explained part of variation significantly. The results of the regression analysis of R&D on national IQ given in Figure 3.2 clarify the relationship between the two variables at the level of single countries.

There is a large research literature showing the positive effect of intelligence on earnings among individuals. The classical study is Christopher Jencks' Inequality (1972) in which he synthesized American research and estimated that the correlation between intelligence and earnings is 0.31 (corrected for attenuation to 0.35). He also estimated that IQ has a heritability of about 50 per cent, and therefore that genetic factors contribute to income differences. Jencks' estimate has proved remarkably accurate in the light of later studies reported for a number of countries and summarized in Table 4.1. This gives the ages at which the IQs were measured and the age at which the earnings were obtained. Thus, the first two rows give the results for Britain for a national sample whose intelligence was measured at the age of 8 years and whose income was obtained at the age of 43 years. The correlations between IQ and income were 0.37 for men and 0.32 for women. These results are typical of the rest of the studies. It will be noted that the correlation in the British study is a little lower for women than for men, and this is also present in the Netherlands given in rows 3 and 5, where the correlation is 0.17 for men and 0.03 for women. The explanation for this is probably that a number of high IQ women take out time to rear children and this reduces their earnings. It will also be noted that the correlations are lower for those in their twenties, for which the median is 0.21, than for those aged thirty and over, for which the median is 0.33. The explanation for this is that those with high IQs frequently do not earn much more in their twenties than those with low IQs, but from their thirties their earnings' advantage increases. The last row gives the results of a meta-analysis of studies in the United States, Britain, Norway, New Zealand, Australia, Estonia, Sweden and the Netherlands, and reports a correlation of 0.23 between IQ and subsequent earnings. ... It might be supposed that the family environment is the common cause of children's intelligence and their subsequent adult earnings, but this is improbable because it has been shown by Duncan, Featherman and Duncan (1972) and by Jencks (1972) that the positive relation between childhood IQ and adult income is present when parental socio-economic status is controlled. Furthermore, among pairs of brothers who have been raised in the same family and have experienced the same environment, the brother with the higher IQ in childhood has the greater earnings in adulthood (Jencks, 1972; Murray, 1998; Waller, 1971). ...Row 1 gives data for a sample in Kalamazoo whose IQs were obtained in sixth grade between 1928 and 1952, and whose earnings were obtained as adults of various ages. His estimate was that an increase of one standard deviation of intelligence produces a 15 percentage increase in earnings. It will be seen that the effect of intelligence on earnings is greater among older people. For instance, rows 3 and 4 give the results of a national sample of the NLSY (National Longitudinal Study of Youth) that was born between 1961 and 1964 and intelligence tested between the ages of 15-18 with the AFQT (Armed Forces Qualification Test). The results show that a one standard deviation advantage in IQ produces a 17 per cent increase in earnings for men at the age of 19 to 32 and 23 per cent increase in earnings of women. The data given in rows 6 and 7 confirm those shown in Table 4.1 that intelligence has a greater positive effect on earnings among older people. The results show that a one standard deviation advantage in IQ produces an 11 per cent increase in earnings for men at the age of 35 and a 22 per cent increase in earnings at the age of 53.

people with high IQs typically obtain advanced education in which they acquire more complex skills, such as those required for professional and executive occupations, that command higher earnings. This has been shown by Hunter and Hunter (1984) in a meta-analysis of 425 American studies through in which jobs were categorized into high, medium and low complexity. The results were that intelligence is correlated with trainability for high complexity occupations at 0.58, for medium complexity occupations at -0.40, and for low complexity occupations at 0.25. These results have been confirmed in a meta-analysis of 69 European studies by Salgado et al. (2003) that reported that intelligence is correlated more highly with trainability for high and medium complexity occupations (0.29) than for low complexity occupations (0.23), although the correlations in the European studies are rather lower than in the United States. The second explanation for the positive association between IQ and earnings is that people with high IQs work more proficiently than those with low IQs. This makes them more productive and able to secure higher earnings. This has been shown by Ghiselli (1966) and by Hunter and Hunter (1984) in their meta-analysis of 425 American studies. In more recent work, Schmidt and Hunter (1998) have published a synthesis of American studies reported from the 1920s through the mid-1990s showing an overall correlation of 0.51 between IQ and job proficiency. They conclude that "the conclusion from this research is that for hiring employees without previous experience in the job the most valid predictor of future performance is general mental ability". Similar results have been reported in a meta-analysis of 69 European studies by Salgado et al. (2003) who conclude that intelligence is positively correlated at 0.25 with job proficiency.

Rows 13 through 18 give six correlations between national IQs and various measures of per capita income reported. The author analyzed further the relationship by fitting linear, quadratic and exponential curves to the data for 81 and 185 nations and found that fitting exponential curves gave the best results. His interpretation was that "a given increment in IQ, anywhere along the IQ scale, results in a given percentage in GDP, rather than a given dollar increase as linear fitting would predict" (Dickerson, 2006, p. 291). He suggests that
> "exponential fitting of GDP to IQ is logically meaningful as well as mathematically valid. It is inherently reasonable that a given increment of IQ should improve GDP by the same proportional ratio, not the same number of dollars. An increase of GDP from $500 to $600 is a much more significant change than is a linear increase from $20,000 to $20,100. The same proportional change would increase $20,000 to $24,000. These data tell us that the influence of increasing IQ is a proportional effect, not an absolute one (p. 294)."
The author noted that his correlations were consistently higher for the 81 nation sample than for the 185 nation sample and suggested that this is attributable to more errors in the 1985 nation sample.

The principal conclusion to be drawn from these studies is that national IQs predict economic growth rates over very long periods, such as 1500-2000 given in row 10, for which the correlation is 0. 71. Over shorter time periods such as 1950-1990 given in row 14, the correlation is lower at 0.44. Over very short time periods such as 1990-2002 the correlation is zero (-0.06). The explanation for this is that various shocks such wars, large increases in the price of oil and so on, reduce the growth rate of some countries in the sort term, but over the long term these have little effect and national IQ emerges as the major determinant of economic growth rates. This conclusion may be surprising to economists because theoretically it would be expected that low IQ countries would have faster economic growth rates than high IQ countries because of what Weede and Kämpf (2002) call "the advantage of backwardness". This advantage should be present because of the potential of poor countries to adopt the technologies and management practices of wealthier countries, whereas wealthier countries depend on innovation. However, the studies summarized in this section show that this is not so, and that the correlation between national IQs and economic growth over the long period is positive. Meisenberg (2011) discuss this question and suggests that the explanation may be that a high IQ population is more likely to establish effective economic institutions that favor economic growth.

Rows 7 and 8 give negative correlations of -0.71 and -0.72 between national IQs and the percentage of the labor force engaged in agriculture. The author suggests "the most parsimonious explanation is that the lower level of education received in agricultural societies means that there is less opportunity for academic ability to develop. As countries become economically developed and as the importance of agricultural labor declines, parents produce fewer offspring and invest more in their education and cognitive development" (Barber, 2005, p. 280). It may be doubted whether this is the correct explanation because of the weight of evidence indicating that family size has no causal relation to IQ (Abdel-Khalek and Lynn, 2008; Rogers, Cleveland,van den Ord and Rowe, 2000). Row 9 gives a correlation of 0.61 between national IQs and investments as the average ratio of investment to GDP over the years 1960-85.

The negative correlations between national IQs and income inequality is predictable from studies showing that among individuals, intelligence is associated with liberalism defined as genuine concern for the welfare of genetically unrelated others and the willingness to contribute larger proportions of private resources for the welfare of such others. In the modern political and economic context, this willingness usually translates into paying higher taxes toward government and its welfare programs (Kanazawa, 2010, p. 286). It has been shown that those who identify themselves as very liberal in early adulthood had a childhood IQ of 106.4, while those who identify themselves as very conservative in early adulthood had a childhood IQ of 94.8 (Kanazawa, 2010, p. 286). It follows from this that national populations with high IQs would be more liberal and favor greater equality of incomes.

Row 17 gives a correlation of 0.48 between national IQs and the savings rate calculated from the ratio of the holdings of US treasury bonds to nominal GDP over the years 1980-2005. The authors argue that this is predictable from the positive association of IQ with a lower time preference and a greater propensity to postpone immediate gratification for future benefits among individuals. Row 18 gives a correlation of 0.49 between national IQs and the rate of self-employment among 117 immigrant groups in Norway. The author notes that this is consistent with results at the individual level showing that the self-employed have above average IQs reported by De Wit and Winden (1989).

Bahrain, Brunei, Equatorial Guinea, Kuwait, Qatar, Saudi Arabia and the United Arab Emirates are oil exporting countries in which the level of per capita income has risen much higher than expected on the basis of their national IQs. In these countries foreign investments, technologies, and management have had a crucial role in their oil industries and these explain the exceptionally high level of per capita income in these countries. The fact that residuals are negative for most neighboring countries without significant oil resources supports this conclusion. Our interpretation is that the existence of exceptional natural resources combined with western technologies has raised per capita income in these eight countries much higher than expected.

The relationship between national IQ and rates of unemployment has not been examined hitherto and is considered in this section. At the individual, within-country level, several studies have shown a robust association between low intelligence and unemployment. Toppen (1971) reported a sample of the unemployed in the United States had an average IQ of 81, more than a standard deviation (15 IQ points) below the U.S. mean IQ of approximately 100. Lynn, Hampson and Magee (1984) reported that a sample of the unemployed in Northern Ireland had an average IQ of 92, again below the national mean. Herrnstein and Murray (1994) reported that in a sample in the United States, 14 per cent of those with IQs below 74 had been unemployed for one month or longer during the preceding year, and the percentages of the unemployed declined in successively higher IQ groups to 4 per cent among those with IQs above 126. Thus, low- IQ individuals make up a disproportionate share of unemployed. Mroz and Savage (2006), using the National Longitudinal Survey of Youth, found that lower IQ predicted higher probability of unemployment within the last year, higher average weeks of unemployment, and higher probability of job change, even after controlling for years of education, ethnicity, parental education, whether the person's childhood home received periodicals, and a rich variety of additional covariates. Thus, both the rate of job destruction and the length of job search are higher for workers with lower IQ. ...The average of the two periods yielded unemployment data for 107 nations for which national IQ data exist. The correlation between the unemployment estimate based on this equation and national IQ is -0.66 (107 nations) and therefore national IQ explains 43.5% of the variance in unemployment. The negative correlations show that unemployment is lower in high IQ nations. The correlation can be corrected for unreliability of both variables. The reliability of the average unemployment figures taken as the correlation between the unemployment figures in the two periods is 0.81. The reliability of national IQs given in Chapter 2 is 0.91. Corrected for unreliability, the correlation between national IQ and unemployment is -0.76 and 57 per cent of variance in the rate of unemployment across nations is explained by national IQ. Thus the relationship between low IQ and high rates of unemployment that is present among individuals also holds across nations.

Studies of the correlations of national IQ and political institutions are summarized in Table 5.1. Row 1 gives a correlation of 0.47 between national IQs and "big government" defined as government expenditure as percentage of GDP, 1980 89. The negative correlation indicates that high IQ nations have less "big government". Row 2 gives a correlation of 0.64 between national IQ and the efficiency of bureaucracy measured as quality and speed of decisions made by public officials. Rows 3 through 10 give eight negative correlations ranging from -0.27 to -0.68 between national IQs and the amount of corruption measured as the Corruption Perception Index (CPI). The negative correlations show that there is less corruption in high IQ countries. The explanation for this proposed by Potrafke (2012, p. 109) is that "intelligent people have longer time horizons" and can understand that corruption is likely to have negative effects over the long term. Rows 11 through 16 give six correlations ranging from 0.53 to 0.79 between national IQs and the amount of democracy measured as the extent to which countries have established democracies. We have proposed that the explanation for this is that "people in countries with low national IQs are not as able to organize themselves, to take part in national politics, and to defend their rights against those in power as people in countries with higher national IQs" (Vanhanen, 2009, p. 270). Rows 17 and 18 confirm these positive correlations (0.57 and 0.58) using a different measure of democracy defined as the averaged scores of political rights and civil liberties and based on 126 and 82 nations. Row 19 gives a correlation of -0.58 between national IQs and the Failed State Index, a measure of state vulnerability to political breakdown. Row 20 gives a correlation of 0.72 between national IQs and institutional quality measured by the Doing Business Index, a measure of the easy of conducting business transactions in 21 Asian countries. Rows 21 through 25 give five correlations ranging from 0.49 to 0.77 between national IQs and the amount of political freedom and citizens' legal rights. Row 26 gives a correlation of 0.75 between national IQs and "Power Resources" defined as an index of the equality of the distribution of important intellectual and economic power resources. The positive correlation shows that countries with higher IQs have a more equal distribution of this power. Row 27 gives a correlation of 0.17 between national IQs and property rights measured as security of property rights and includes efficiency of government bureaucracy. The correlation is quite low and only statistically significant at p<.10. Rows 28 through 30 give correlations ranging from 0.62 to 0.82 between national IQs and the rule of law defined as an index of the independence of the judiciary and the ability of the citizen to enforce contracts in courts of law.

Table 5.5 shows that women's representation in parliaments is statistically significantly but only weakly associated with national IQ, which explains only 11 percent of the variation in Women-08 in the total group of 187 countries. Correlations in the two other groups of countries are equally weak, and Spearman rank correlations are only slightly stronger. Approximately 90 percent of the variation in Women-08 seems to be due to various local, cultural, institutional, and accidental factors. What might those other factors be? Could IPR raise the explained part of variation in Women-08 significantly. When national IQ and IPR are taken together to explain variation in Women-08, the multiple correlation rises to 0.332 in the group of 171 countries and the explained part of variation in Women-08 remains in 11 percent. It is evident that the variation in women's representation is nearly completely independent from national IQ and resource distribution (IPR). So we come to the conclusion that women's representation in parliaments is only slightly related to national IQ and to some other explanatory variables. It is not possible to predict the level of women's representation in parliaments on the basis of national IQ to any significant extent, although the relationship between them is positive. Consequently, national IQ does not constrain women's representation to any significant extent and it does not prevent a rise of women's representation in countries with low national IQs, but there may be other factors (cultural and institutional) which maintain great global differences in women's representation.

For instance, Anstey, Low and Sachdev (2009) have shown that the intelligence is associated with higher levels of physical activity, greater likelihood of taking vitamins, and reduced likelihood of smoking, all of which promote good health. Several studies have reported that low birth weight is associated with low IQ in childhood and adolescence, e.g. Bhutta, Cleves, Case, Cradock and Anand, 2002; Deary, Whalley and Starr (2009, pp. 193-195). Infant mortality (infant deaths in the first year of life) is associated with low IQ mothers. This was first shown by Savage (1946) who reported that the mothers of infants who died in their first year had below average intelligence. This was confirmed by Herrnstein and Murray (1994, p. 218) who showed that the mothers of infants who had died in their first year had an average IQ of 94, compared with 100 of the mothers of infants who had not died in their first year. These results are understandable, because mothers with low IQs would be less competent in taking care of the health of their infants. Mothers with higher IQs would be better at anticipating possible accidents and preventing them happening, judging whether illnesses are sufficiently serious to justify seeing a physician, and giving medications that are prescribed. Several studies have found that intelligence is a determinant of life expectancy. This was shown first in Australia by O'Toole and Stankov (1992) in a study of 2,309 men who were conscripted into the military and intelligence tested at the age of 18, between 1965 and 1971. They were followed up in 1982, when they were aged between 22 and 40, and it was found that 523 had died. These had an IQ 4 points lower than those who remained alive. The commonest cause of death was accidents of various kinds (389), of which motor vehicle accidents (217) were the most frequent. It seems probable that the explanation for this association is that those with lower IQs make more misjudgments. Some of these misjudgments result in accidents and some of these are fatal. Gottfredson (2004) has reviewed a number of subsequent studies confirming the association of low intelligence with high mortality, and this has also been found in Sweden (Hemmingsson, 2009). An extensive research program in Scotland examining the relation of IQ measured at the age of 11 to mortality (i.e. age of death) has been summarized by Deary, Whalley and Starr (2009, pp. 50-52). They confirm that low intelligence predicts high mortality and have found that low intelligence is associated with several specific causes of death. Low intelligence is associated with smoking and death from lung cancer and other smoking- related cancers, namely mouth, pharynx, esophagus, larynx, pancreas and bladder cancers. Low intelligence is also associated with death from all cardiovascular diseases, coronary heart disease, stroke, and respiratory disease. They suggest four explanations for these associations. First, childhood IQ might be a record of bodily insults including illness, poor nutrition, and injuries. Second, childhood IQ might be a marker for genetic bodily system integrity. Third, people with higher IQs may be better at avoiding risks and at preserving their health, for instance by eating sensible foods, avoiding smoking, recognizing symptoms that might be injurious to health, consulting physicians, and complying with prescribed treatments. This theory implies that intelligence differences are causal to mortality. Fourth, people with higher IQs may tend to work in occupations where there is less risk of death. ...Barber (2005) was the first to report this negative correlation based on infant mortality rates averaged for 1978-1980 and suggests that this arises because "infant mortality is affected by the prevalence of infection as well as infant nutritional status and is considered a sensitive indicator of infant health for a population" (p. 278). ...Kanazawa (2006) reports a negative correlation of -0.84 based on 126 countries and notes that "the unstandardized regression coefficient of 22.5816 for national IQ . . . means that each additional point in the mean IQ of a population saves more than two and half infants from death per 1,000 live births."

Rows 26 through 34 give nine studies showing positive correlations ranging from 0.37 to 0.70 between national IQ and suicide rates. The positive correlations show that suicide rates are higher in high IQ nations. The evidence on the relation between suicide rates and intelligence among individuals is conflicting. Four studies have reported that suicide is associated with higher IQ (De Hert, McKenzie and Peuskens, 2001; Fenton, 2000; Webb, Långström, Runeson, Lichtenstein and Fazel, 2011; Westermeyer, Harrow and Marengo, 1991). On the other hand a study in Sweden has shown that suicide is associated with low IQ among males, although not among females (Andersson, Allebeck, Gustafsson and Gunnel, 2008). Other studies have shown that suicide is associated with poor educational attainment in Australia, Norway, Denmark and Finland (Gunnell, Lofving, Gustafsson and Allebeck, 2011). In the United States, university students who have higher than average IQs, have lower suicide rates than non-students of the same age, where the percentages of deaths due to suicide are 14.4% for students and 16.7% non-students (Stack, 2011). A theory that assumes there is a positive association between suicide and intelligence among individuals and across nations has been proposed by Voracek (2009a), who suggests that a certain level of intelligence is required to understand that a person's kin would benefit from one's death, and therefore that suicide can increase a person's inclusive fitness. A possible alternative or additional explanation is that depression is less prevalent in the low IQ countries of sub-Saharan Africa. This was noted in the early 1950s by Carothers (1953, p. 144), a medical officer at the mental hospital in Nairobi, who recorded that among 1,508 patients admitted over the years 1939-48, only 24 suffered from depression, amounting to 1.6 per cent of admissions. He contrasted this with 22 per cent of admissions of European patients admitted to the same hospital diagnosed as depressives. He wrote that "there is no doubt that classical psychotic depression of any type is relatively rare in the African" (p. 145). The low prevalence of depression among sub-Saharan Africans has been confirmed in the United States by Gonzalez, Neighbors, Nesse, Sweetman and Jackson (2007).

This negative association has become known as dysgenic fertility and has been extensively investigated in the United States. The American studies reporting this negative association are summarized in Table 7.1. Row 1 gives a correlation of -0.49 between intelligence and fertility derived from data on the IQs of all children aged 10 to 14 in Georgia, numbering approximately a quarter of a million. The correlation was calculated for the average IQs of children in each of 159 counties and the fertility rates of women aged 15 to 49. Because this is group data, the correlation is higher than would be expected on individual data. Rows 2, 3 and 4 give negative correlations between intelligence and fertility based on a nationally representative American sample showing that the negative correlation is higher for white women than for white men, and higher for white women than for black women. This study is not wholly satisfactory because the age of the sample was 25 to 34 years and many of them would not have completed their fertility. To overcome this problem, Vining (1995) published data on the fertility of his female sample of the ages between 35 and 44, which can be regarded as close to completed fertility. The results are given in rows 4 and 5 for white and black women and show that the correlations between intelligence and fertility are still significantly negative and are higher for black women (-0.226) than for white women (-0.062). These correlations are probably underestimates because the samples excluded high-school dropouts, who were about 14 per cent of whites and 26 per cent of blacks at this time, and who likely had low IQs and high average fertility. Rows 8 and 9 give negative correlations between intelligence and fertility for approximately 17,000 white and 19,000 black babies born in the late 1970s. The IQs of the mothers correlated negatively with number of children at -0.22 for blacks and -0.12 for whites. Rows 10 through 13 show negative correlations for national samples born between 1900-1949 and with completed fertility. As in previous studies, the negative correlations are higher for women than for men among both blacks and whites, and are higher for blacks than for whites. Rows 14 through 17 give updated data for show further negative correlations for national samples aged between 30 and 47 years and therefore with virtually completed fertility. The results provide further confirmation that the negative correlations are higher for women than for men among both blacks and whites, and are higher for blacks than for whites. All the studies summarized in Table 7.1 show that dysgenic fertility for intelligence has been present in the United States during the twentieth century. ...It has been shown by Meisenberg and Kaul (2010, p. 177) that the lower fertility of intelligent women is not due to a lack of desire for children. ...Apart from the United States, the evidence on intelligence and fertility elsewhere is remarkably sparse. There are studies in England, Scotland, Sweden, and Greece showing negative associations between intelligence and fertility, that we have summarized in Lynn (2011). There is also a study showing dysgenic fertility in the Caribbean island of Dominica reported by Meisenberg, Lawless, Lambert and Newton (2005). They found that the correlation between IQ and numbers of children was slightly positive (r = 0.06) for men, while for women it was negative (r = -0.163). These correlations show that fertility is eugenic for men, but more strongly dysgenic for women, as in the American studies summarized in Table 7.1. A study reporting dysgenic fertility for intelligence in Sudan has been published by Khaleefa, Haroon and Abdulradi (2011). They report a negative association between intelligence and the number of siblings in a sample of 5,215 school students. Thus lower IQ children had larger numbers of siblings, and it can be inferred from this that lower IQ parents had larger numbers of children. They calculate a decline of genotypic intelligence at 0.66 IQ points a generation.

In a review of these studies, Wilson and Herrnstein (1985, p. 159) wrote that "For four decades, large bodies of evidence have consistently shown about a ten IQ point gap between the average offender and the average non-offender in Great Britain and the United States". This conclusion has subsequently been confirmed by Ellis and Walsh (2003) in a summary of more than a hundred studies from all over the world. The influence of socio-economic status and family environment on crime has been controlled in a Danish study of pairs of brothers that has shown that the brother with a criminal record scored an average of 15 IQ points lower than the law-abiding sibling (Kandel, Mednick and Kirkegaard-Sorensen, 1988). ...The first study showing that this is so was published by Maller (1933a, 1933b) in an analysis of average IQs and crime rates in 310 districts of New York City. He found that the correlation between the average IQ of ten year olds and the rates of juvenile delinquency was 0.57. The relation between intelligence and crime among populations has also been investigated by Bartels, Ryan, Urban and Glass (2010) in a study of the IQs of American states and crime rates. They report that crime rates are higher in states with lower IQ and that these negative correlations are higher for violent crime (-0.58) than for non-violent crime, including motor vehicle theft and other theft (-0.29).

We now consider some cognitive expressions of intelligence that are correlated with national IQs. Studies of this kind are summarized in Table 9.3. Row 1 gives a negative correlation of -.55 between national IQ and "acquiescence" defined as agreement with statements presented in opinion surveys. The negative correlation shows that people in low IQ countries are more likely to acquiesce. Meisenberg and Williams (2008) report that acquiescence is associated at the individual level with low IQ, predict that the same association should be present across nations, and demonstrate that this is the case.

Consistent with Frazer's analysis, it has been found in a number of studies of individuals within nations that there is a negative relationship between intelligence and religious belief. This negative relationship was first reported in the United States in the 1920s by Howells (1928) and Sinclair (1928), who both reported studies showing negative correlations between intelligence and religious belief among college students of -0.27 to -0.36 (using different measures of religious belief). A number of subsequent studies confirmed these early results, and a review of 43 of these studies by Bell (2002) found that all but four found a negative correlation. To these can be added a study in the Netherlands of a nationally representative sample (total N=1,538) that reported that agnostics scored 4 IQs higher than believers (Verhage, 1964). In a more recent study Kanazawa (2010) has analyzed the data of the American National Longitudinal Study of Adolescent Health, a national sample initially tested for intelligence with the PPVT (Peabody Picture Vocabulary Test) as adolescents and interviewed as young adults in 2001-2 (N=14,277). At this interview they were asked: "To what extent are you a religious person?" The responses were coded "not religious at all", "slightly religious", "moderately religious", and "very religious". The results showed that the "not religious at all" group had the highest IQ (103.09), followed in descending order by the other three groups (IQs = 99.34, 98.28, 97.14). The negative relationship between IQ and religious belief is highly statistically significant. These studies are confirmed by evidence showing that the percentages of religious believers among intelligence elites are lower than in the general population. This was shown as early as 1921 in a survey of the religious beliefs of eminent American scientists and scholars that reported that 39 per cent stated that they believed in God (with a range of 48 per cent among historians to 24 per cent among psychologists) (Leuba, 1921). It was later reported by Roe (1965) that among a group of 64 eminent scientists, 61 were "indifferent to religion", while only three were religious believers. These are much lower than the percentage religious believers in the population among whom 95.5 per cent in the United States stated that they believed in God in a 1948 Gallup Poll (Argyle, 1958). In the 1990s a study of members of the American National Academy of Sciences reported that 7 per cent believed in the existence of God, as compared with approximately 90 per cent found in a poll of the general population (Larsen and Withham, 1998). In Britain, it has been reported that 3.3 per cent of Fellows of the Royal Society believed in the existence of God, while 78.8 per cent did not believe (the remainder being undecided) (Dawkins, 2006). At the same time a poll showed that 68.5 per cent of the general population believed in the existence of God. Further evidence for a negative correlation between intelligence and religious belief is the decline in religious belief during adolescence and into adulthood as cognitive ability increases. This has been found in the United States for the age range of 12-18 year olds by Kuhlen and Arnold (1944) who reported that among 12 year olds 94 per cent endorsed the statement "I believe there is a God", while among 18 year olds this had fallen to 78 per cent. Similarly, in England Francis (1989) has found a decline in religious belief over the age range 5-16 years. Religious belief was measured by a scale consisting of questions like "God means a lot for me" and "I think that people who pray are stupid", etc. The results were that among 5-6 year olds 87.9 per cent of boys and 96.0 per cent of girls held religious belief, but at the age of 15-16, these percentages had fallen to 55.7 of boys and 70.4 of girls. Finally, in several economically developed countries there has been a decline of religious belief during the course of the last 150 or so years, while at the same time the intelligence of the population has increased. For instance, in England self-reported weekly attendance at church services reported in census returns declined from 40 per cent of the population in 1850, to 35 per cent in 1900, to 20 per cent in 1950, and to 10 per cent in 1990 (Giddens, 1997, p. 460). Church of England Easter week communicants declined from 9 per cent of the population in 1900 to 5 per cent in 1970 (Argyle and Beit-Hallahmi, 1975). The attendance of children at Sunday schools declined from 30 per cent of the child population in 1900 to 13 per cent in 1960 (Goldman, 1965). In Gallup Polls 72 per cent of the population stated in 1950 that they believed in God (Argyle, 1958), but by 2004 this had fallen to 58.5 per cent (Zuckerman, 2006). There has also been some decline of religious belief during the course of the last century in the United States. Hoge (1974) has reviewed several surveys that have found a decline of religious belief in college students. For instance, students at Bryn Mawr were asked whether they believed in a God who answered prayers. Positive responses were given by 42 per cent of students in 1894, 31 per cent in 1933, and 19 per cent in 1968. Students enrolling at the University of Michigan were invited to provide a "religious preference". In 1896, 86 per cent of students did so; in 1930 this had dropped to 70 per cent, and in 1968 it had dropped to 44 per cent. At Harvard, Radcliffe, Williams and Los Angles City College the percentages of students who believed in God, prayed daily or fairly frequently, and attended church about once a week all declined from 1946 to 1966. Heath (1969) has also reported a decline in belief in God among college students from 79 per cent in 1948 to 58 per cent in 1968. Among the general population, Gallup Polls have found that 95.5 per cent stated that they believed in God in 1948 (Argyle, 1958), but by 2004 this had fallen to 89.5 per cent (Zuckerman, 2006).

Table 10.4 shows that empirical evidence supports the hypothesis on the negative impact of socioeconomic development on religious beliefs. All correlations are negative as hypothesized, but they are weak or only moderate and clearly weaker than corresponding correlations between national IQ and the four indicators of religious beliefs (cf. Table 10.3). The results show that national IQ explains much more of the variation in the dependent variables than the four indicators of socioeconomic development, education, and democratization.

The results of empirical analyses indicate that the degree of religiosity tends to decline when the level of national IQ rises, but until now this has clearly occurred only at higher levels of national IQ (above 85). There seems to be little variation in the level of religiosity below national IQ of 85. The problem is why the decline accelerates above national IQ level of 85 and turns the relationship partly curvilinear. The results of regression analysis provided two additional and important explanatory factors, which help to solve this problem. One is the anti-religious Communist heritage in contemporary and former socialist countries and another concerns secularization in many socioeconomically highly developed societies. Because of anti-religious state policies in socialist countries and especially in the former Soviet Union, the level of religiosity declined in most of those countries more than expected on the basis of national IQ, and residuals became highly negative for most of them as indicated in previous sections. This illustrates the hostile impact of political and cultural climate on religiosity, although there has been some religious revival in some of these countries after the collapse of socialist systems (see Norris and Inglehart, 2004, pp. 111-132). In fact, residuals based on Religiosity are clearly positive for some former socialist countries like Croatia, Armenia, Bosnia & Herzegovina, Poland and Romania. In Poland, Yugoslavia and Romania, religious communities retained their independence much better than in the Soviet Union during the Communist period. Secularization in Western Europe constitutes another factor which helps to explain the curvilinear relationship between national IQ and measures of religious beliefs. The exceptionally strong impact of secularization has reduced the level of religiosity in most Western European countries. The same has occurred in Japan. However, secularization has not yet spread equally to all countries of high national IQs. There are some highly religious countries like Cyprus, Malta, Poland and the United States with large positive residuals. In the end, only two additional factors - the heritage of the Communist political culture and the exceptionally strong level of secularization - are needed to explain most large negative deviations from the regression line.
Shared publiclyView activity