NotPoliticallyCorrect

Home » Race Realism » Worldwide IQ estimates based on education data

Worldwide IQ estimates based on education data

Enter your email address to follow this blog and receive notifications of new posts by email.

Join 112 other followers

Follow me on Twitter

JP Rushton

Richard Lynn

L:inda Gottfredson

Goodreads

By Afrosapiens, 2851 words

One of the leading theories to explain differences in cognitive test performance between time and place is that intelligence as measured by such tests is affected by exposure to formal schooling and the cognitive demands of a high-technology society (D. Marks, JR. Flynn). One of the strongest evidence for such an effect of schooling on IQ comes from a reform in the Norwegian school system in which an expansion of compulsory schooling was associated with a 3.7 points increase in IQ per additional year of education between pre-reform and post-reform cohorts. In order to test this relationship between years of schooling and commonly reported national IQ averages, I used data from the United Nation’s Development Program to estimate the average IQ of each country’s adult and school-age population. Adult IQs were estimated from mean years of schooling completed by adults aged 25 and older whereas School-population IQs were estimated based on the expected years of schooling that a student is supposed to complete if the enrollment ratios from primary through tertiary education remain constant. All variables were reported in year 2015. Great Britain was chosen as the reference country and assigned a default value of 100 on both variables. For each country, a difference of one year in completed or expected schooling added or removed 3.7 IQ points. Adult IQ and School-age population IQ were averaged to estimate the most probable mean IQ that would be found by randomly reviewing literature without controlling for the age or the health and socio-economic profile of the sampled individuals.

Results

Country Main ancestry School age-Adult IQ average School age IQ Adult IQ
Australia West-European 107 115 100
Denmark West-European 104 111 98
New Zealand West-European 104 111 97
Iceland West-European 103 110 96
Ireland West-European 102 109 96
Norway West-European 101 105 98
Germany West-European 101 103 100
Netherlands West-European 101 107 95
United States West-European 100 101 100
United Kingdom West-European 100 100 100
Switzerland West-European 100 99 100
Canada West-European 100 100 99
Slovenia East-European 100 104 96
Lithuania East-European 99 101 98
Czech Republic East-European 99 102 96
Estonia East-European 99 101 97
South Korea North-East Asian 99 101 96
Israel West and Central Asian, North African 99 99 98
Sweden West-European 98 99 96
Poland East-European 98 100 95
Finland East-European 97 103 92
France West-European 97 100 94
Japan North-East Asian 97 96 97
Latvia East-European 96 99 94
Belarus East-European 96 98 95
Greece East-European 96 103 90
Hungary East-European 96 97 95
Spain West-European 96 105 87
Hong Kong North-East Asian 96 98 94
Austria West-European 96 99 93
Italy West-European 96 100 91
Slovakia East-European 96 95 96
Argentina West-European 95 104 87
Singapore North-East Asian 95 97 94
Liechtenstein West-European 95 94 97
Russia East-European 95 95 95
Kazakhstan West and Central Asian, North African 95 95 94
Ukraine East-European 94 96 93
Palau South-East Asian and Polynesian 94 93 96
Croatia East-European 94 96 92
Montenegro East-European 94 96 93
Chile West-European 94 100 87
Georgia West and Central Asian, North African 94 91 96
Cyprus East-European 93 93 94
Luxembourg West-European 93 91 95
Malta West-European 93 94 93
Bulgaria East-European 93 95 91
Barbados Black African 93 96 90
Fiji South-East Asian and Polynesian

South Asian

93 96 90
Cuba West-European 93 91 94
Saudi Arabia West and Central Asian, North African 93 99 86
Portugal West-European 92 101 84
Romania East-European 92 94 91
Tonga South-East Asian and Polynesian 92 93 92
Serbia East-European 92 93 91
Belgium West-European 91 90 93
Sri Lanka South Asian 91 91 91
Mongolia North-East Asian 91 91 87
Grenada Black African 90 98 83
Mauritius South Asian 90 96 84
Uzbekistan West and Central Asian, North African 90 85 95
Uruguay West-European 90 97 83
Armenia West and Central Asian, North African 90 87 93
Brunei South-East Asian and Polynesian 89 95 84
Azerbaijan West and Central Asian, North African 89 87 92
Bahrain West and Central Asian, North African 89 93 86
Andorra West-European 89 90 89
Kyrgyzstan West and Central Asian, North African 89 88 91
Albania East-European 89 92 86
Moldova East-European 89 83 95
Venezuela West-European 89 93 86
Trinidad and Tobago Black African

South Asian

89 87 91
Bahamas Black African 89 87 91
Iran West and Central Asian, North African 89 94 83
Seychelles Black African

South Asian

West-European

89 92 86
Belize Black African

Native American

88 87 90
South Africa Black African 88 88 89
Malaysia South-East Asian and Polynesian 88 88 88
Bosnia East-European 88 92 84
Samoa South-East Asian and Polynesian 88 87 89
Jordan West and Central Asian, North African 88 88 88
Qatar West and Central Asian, North African 88 89 87
Brazil West-European 88 96 79
Costa Rica West-European 88 92 83
Panama Native American 88 88 87
United Arab Emirates West and Central Asian, North African 87 89 86
Turkey West and Central Asian, North African 87 94 80
Peru Native American 87 89 84
Saint Lucia Black African 87 88 85
Jamaica Black African 87 87 86
Macedonia East-European 86 87 86
Ecuador Native American 86 91 82
Algeria West and Central Asian, North African 86 93 82
Saint-Kitts and Nevis Black African 86 90 82
Bolivia Native American 86 91 81
Mexico West-European 86 89 83
Saint Vincent and the Grenadines Black African 86 89 83
Lebanon West and Central Asian, North African 86 89 83
Oman West and Central Asian, North African 86 90 81
Botswana Black African 86 86 85
Palestine West and Central Asian, North African 85 87 84
Tajikistan West and Central Asian, North African 85 82 89
Tunisia West and Central Asian, North African 85 94 77
Thailand South-East Asian and Polynesian 85 90 80
Micronesia South-East Asian and Polynesian 85 83 87
Colombia West-European 84 90 79
China North-East Asian 84 90 79
Philippines South-East Asian and Polynesian 84 83 85
Suriname South-East Asian and Polynesian

Black African

South Asian

84 87 82
Dominican Republic Black African 84 89 79
Indonesia South-East Asian and Polynesian 84 87 80
Dominica Black African 84 87 80
Gabon Black African 84 86 81
Libya West and Central Asian, North African 84 89 79
Turkmenistan West and Central Asian, North African 84 80 87
Kuwait West and Central Asian, North African 83 89 78
Vietnam South-East Asian and Polynesian 83 86 80
Paraguay Native American 83 85 81
Egypt West and Central Asian, North African 82 88 77
Kiribati Melanesian 82 84 80
El Salvador Native American 81 89 75
Zambia Black African 80 86 76
Maldives South Asian 80 87 74
Guyana Black African

South Asian

79 78 82
Namibia Black African 79 83 76
Ghana Black African 79 82 76
Cabo Verde Black African 79 90 69
Nicaragua Native American 79 83 75
Swaziland Black African 79 82 76
India South Asian 79 83 74
Zimbabwe Black African 79 79 79
Vanuatu Melanesian 78 80 76
Honduras Native American 77 81 74
Congo Black African 77 81 74
Kenya Black African 77 81 74
Sao Tome and Principe Black African 77 84 70
Morocco West and Central Asian, North African 77 84 69
Guatemala Native American 77 79 74
Timor-Leste Melanesian 76 86 67
Lesotho Black African 76 79 73
Togo Black African 76 84 68
Iraq West and Central Asian, North African 76 77 75
Cameroon Black African 76 78 73
Angola Black African 76 82 69
Madagascar South-East Asian and Polynesian

Black African

76 78 73
Nepal South Asian 75 85 66
Laos South-East Asian and Polynesian 75 80 70
Nigeria Black African 75 77 73
Comoros Black African 75 81 69
DR Congo Black African 75 76 73
Uganda Black African 74 77 72
Bhutan South Asian 74 86 62
Cambodia South-East Asian and Polynesian 74 80 68
Bangladesh South Asian 74 77 70
Malawi Black African 73 80 67
Solomon Islands Melanesian 73 75 70
Equatorial Guinea Black African 72 74 71
Tanzania Black African 72 73 72
Rwanda Black African 72 80 65
Haiti Black African 72 73 70
Liberia Black African 72 76 67
Benin Black African 72 79 64
Papua New Guinea Melanesian 72 76 67
Syria West and Central Asian, North African 71 73 70
Cote d’Ivoire Black African 71 73 69
Myanmar South-East Asian and Polynesian 71 73 68
Afghanistan West and Central Asian, North African 70 77 64
Burundi Black African 70 79 62
Pakistan South Asian 70 70 70
Mauritania West and Central Asian, North African

Black African

69 71 67
Sierra Leone Black African 69 75 63
Mozambique Black African 69 73 64
Senegal Black African 68 75 61
Gambia Black African 68 73 63
Guinea-Bissau Black African 68 74 62
Yemen West and Central Asian, North African 67 73 62
Guinea Black African 66 72 60
Central African Republic Black African 66 66 66
Ethiopia North-East African 66 71 60
Mali Black African 65 71 59
Sudan North-East African 65 66 64
Djibouti Black African 64 63 66
South Sudan Black African 63 58 69
Chad Black African 63 67 59
Burkina-Faso Black African 62 68 56
Eritrea Northeast-African 62 58 65
Niger Black African 58 60

57

The values were rounded to the nearest unit.

In comparison to the mean national IQs mainly reported by Richard Lynn, 65 countries differed by less than 5 IQ points using the present methodology. It can be said that such small differences validate Lynn’s estimates since it is unlikely that years of education have the same cognitive value in every country and likewise, averaging adult IQ and school-age population IQ without controlling for a country’s age structure somewhat weaken the representativeness of my findings. Differences larger than 5 points were found for 30 countries, and in these cases, I suspect it is due to Lynn manipulating the data to fit racial patterns, Sub-Saharan African countries have been systematically under-estimated and East-Asian ones have been systematically over-estimated by Lynn, also, Some nations in Europe, the Middle-East, South-Asia and Latin America seem to have their scores manipulated in order to appear closer to what they would be based on their racial composition.

Such inconsistencies result in incoherences between the reported IQs and the educational and socio-economic outcomes (regardless of which variable influences the other) of the affected countries and support the accusations of racially-motivated fraud in Richard Lynn’s data collection. In the same way, estimating the mean IQs of countries for which direct data is missing by averaging the figures of neighboring countries of similar ethno-racial composition is unwarranted as race does not seem to play a role in a country’s cognitive performance.

In spite of all the deserved criticism that Lynn’s data met, it can be said that most of the commonly cited mean IQs out of Africa and East-Asia are reliable and that a strong relationship between human capital and human development exists whether we measure it by IQ or years of education. The causes of international variation in school quality and enrollment are well-known and come down to school and student characteristics. Schools in developing countries face numerous challenges: lack of basic amenities such as electricity, potable water, air-conditioning and heating, like of educational supplies (school rarely have enough textbooks and rely on chalk and blackboards), high student to teacher ratios (primary school classes with more than 50 students are common low-income countries), chronic teacher absenteeism (teachers usually have a business on the side), obsolete pedagogy, outdated or irrelevant curricula, multilingualism, exam-corruption, low public funding, misguided policies, gender and ethnic discrimination. Pupils are held back by poor health and nutrition resulting in developmental delays, tuition fees and supplies that poor families can’t afford, war, population displacement, absent educational resources at home, low parental education, lack of transportation, child labor, excessive use of grade repetition, mismatch between school curricula and daily life demands and many other factors. Differences in human capital have large implication in terms of workforce qualification and social behavior, which contribute for a large part to a country’s socio-economic development. The present findings provide evidence for large international inequalities in inter-generational change in educational outcomes which are probably the driving cause of the Flynn effect.

Intergenerational change in cognitive performance.

This slideshow requires JavaScript.

Estimating IQs from the current school enrollment rates and the mean educational attainment of adults provide insights regarding intergenerational differences in cognitive performance. We can see from these figures that the countries that developed the fastest show large intergenerational differences in education/IQ favoring the younger cohorts, these countries are concentrated in South America, Southern Europe, West Africa, the Middle-East and Oceania, Ethiopia and China also show trends that are in line with their recent economic success. On the other hand, many ex-USSR countries, as well as Japan, Cuba, South Africa, Zimbabwe and the Philippines have been stagnant or even declining relative to the United Kingdom and this is also reflect in their poor socio-economic performance in the last decades. War-torn South Sudan and the Central African Republic experience alarming declines in their educational performance that expose them to grave humanitarian crises in the future. Although there is a clear relationship between socio-economic progress and gains in cognitive performance, a country’s ability to capitalize on its intellectual potential remains highly dependent on its leadership and the odds of the world-market, that’s why theories claiming that IQ is the main driver of global inequalities are not tenable in the light of the current evidence.

 

Update 09/07/2017 – Detailed comparison with Lynn’s Data

 

To test the predictive power of my estimates in comparison to Lynn’s, I decided to focus only on the world’s 20 most populous countries. The reason for that is that these countries are home to 70% of the world’s population and the law of large numbers says that they are likely more representative of whatever they could represent. On the other hand, the 100+ other countries are home to only 30% of humanity. They are a source of statistical noise due to extreme outlying values and differences in regional political fragmentation that would hide or weaken general trends better evidenced by considering large countries.

Data:

ranking

 

Correlations and averages:

correlations

 

Noticing an abnormal 22 points gap between Sub-Saharan African IQs and the world average on Lynn data, Suspecting that extremely low values would flaw the correlations, I tested if my estimates and Lynn’s would retain the same predictive power with the African IQs excluded. My assumption was that a strong causal relationship would leave the correlations unchanged no matter which countries were included whereas any change in predictive power resulting from excluding some countries would cast doubt on the accuracy of the reported data.

IQ-HDI correlation:

Similarly to my previous calculation including all the countries for which data were available, I found a 0.96 correlation between my estimates and HDI, Lynn’s estimates correlation with HDI was higher (+0.06) than with the worldwide data, but still largely inferior to mine. Removing African countries, the predictive power of my estimates remained the same (+02) whereas Lynn’s significantly decreased (-0.13) and left a predictive gap of 0.24 favoring my estimates. However, given the fact that my estimates are based on variables that are included in the calculation of HDI, such a high predictive power as to be met with caution.

IQ-GDP per capita correlation:

My previous calculation from the worldwide data yielded a correlation coefficient of 0.65 between my IQ estimates and GDP/capita and 0.60 for Lynn’s. Among the 20 most populated countries, the correlation rose by 0.24 points to 0.89 with my estimates and by 0.12 points to 0.72 with Lynn’s. Excluding Sub-Saharan African countries did not affect the predictive power of my estimates (+0.01) and further weakened Lynn’s by 0.04 points, resulting in a 0.22 gap in predictive power favoring my estimates again. This correlation of 0.89 between my IQ estimates and GDP per capita within the world’s population top 20 countries likely is the highest correlate of IQ ever reported in the psychological science and gives strong support to the relationship between schooling, economic development and cognitive ability.

IQ-Life expectancy correlation:

Compared with the worldwide database, the correlation between my IQ estimates and life expectancy was down 0.04 points within the world’s top 20 to 0.76, Lynn’s went up by o.o5 points to 0.84. However, removing Sub-Saharan Africa left the predictive power of my estimates unchanged whereas Lynn’s fell by 0.13 points to 0.71. My estimates again predicted life expectancy better by a small 0.4 points this time.

IQ-Homicide correlation:

Not estimated previously, my data finds an non-existent relationship between IQ and homicide rate (-0.01) and excluding Sub-Saharan Africa confirmed a null relationship between homicide rates and IQ in the rest of the world. Lynn’s estimates showed a low negative correlation between IQ and homicide (-0.35) and the exclusion of African countries further lowered the correlation (-0.25). Lynn’s estimates had a better predictive power which still remained in the range of low statistical significance.

IQ-Fertility correlation:

Adding a new variable, I found a negative correlation of -0.69 between my IQ estimates and Fertility, the correlation remained the same (-0.68) with the African countries excluded. The correlation between Lynn’s IQs and fertility was stronger (-0.84), but removing African data decreased it by 0.18 points to 0.66. My estimates ended up with a slightly stronger predictive power (+0.02).

General patterns:

In addition to having a stronger and globally consistent predictive power, my estimates reveal how Richard Lynn manipulates the data to fit desired racial patterns.

As expected from the 0.96 correlation between my IQ data and HDI, the ranking of countries by cognitive ability shows a perfect gradient from high to low development status. Moreover, the highest gap between two following countries is the 6 points separating Russia and Iran, showing a marked difference between the developed and the developing world.

Ranking countries by Lynn-estimated IQs results in a whole other pattern in which a country’s dominant ancestry seems to be the only variable that matters. East-Asians are on top, followed by Western Europeans, then Eastern-Europeans, South-East Asians, fair skinned Middle-Easterners (Turkey and Iran) and Latin Americans, Austronesians (Indonesia and the Philippines), South Asians and Arabs, and finally Sub-Saharan Africans far below, with a huge 10 points gap (the largest between two following countries in his dataset) separating Bangladesh from Nigeria.

The manipulation is quite apparent, Lynn largely over-estimated China (+22), Japan (+7) to make East-Asians cluster on top, thus protecting himself from accusations of nordicism and giving support to the inter-cultural validity of the IQs that he cherry-picked. The western European and Russian  data remained mostly unchanged. Vietnam (+11) and Thailand (+5) were given a bonus for their genetic proximity to North-East Asia that is supposed to make them score in the low 90s despite their lack of development. Little changes were brought to the scores of the Latin American, Middle-Eastern and Austronesian countries usually scoring in the mid-80s. Major fraud (+14 in Pakistan, +7 in Bangladesh) was done to lift up South-Asian countries out of the 70s range and excluding Sub-Saharan Africa as the only region scoring 70 or below and downgrading Nigeria (-4) and the DR. Congo (-7) in the process.

By pointing this out, I’m warning honest researchers and laymen about the dangers of relying on data resulting from undisclosed, unsystematic and un-replicable methodology. And although my estimates do not result from any actual IQ measurement beyond the relationship between IQ and schooling evidenced in Norwegian cohorts, my method uses a single, universal conversion factor applied to representative official data collected by professional demographers whereas Lynn’s and the likes’ cherry-picking of samples is only the hobby of a dozen scholars and pseudo-scholars. This is how I found out strong, consistent and meaningful correlations between IQ and various development variables.

Although they are likely more representative of the worldwide distribution of cognitive ability, my estimates still provide evidence that a large part (the largest part, actually) of the world’s population scoring below one standard deviation on Western-normed IQ tests, which is the case for 11 of the world’s 20 most populated countries. Although this may sound alarming, with Pakistan and Ethiopia scoring in the range of mental disability (70 and 66 respectively), I think this effect comes from using Western populations as a reference for standardization.

In fact, another picture emerges when we compare countries with the world’s average, replacing the eurocentric British Greenwich IQ of 100 by an universal IQ of 84 and thus giving a more accurate idea of what is normal cognitive ability by the standards of humanity. In this sample, China, the Philippines and Indonesia are representative of the top of the bell curve whereas Ethiopia, the United States and Germany are the only outliers left with respective Universal IQs of 81.6, 115.6 and 116.6. For this reason, I recommend the use of Chinese or South-East Asian normalization samples in international IQ comparisons.

 

 

 

 

Advertisements

60 Comments

  1. Peter says:

    Afro, Where Did You Get Your Data?

    Like

  2. Name says:

    I’ll play devil’s advocate:
    My father, woodworker, high-school drop out (he had to work for his father, they were poor), scored 142 points.
    I, graduated in Civil Engineering, scored 134.
    My mother, primary school teacher, scored 101.

    Really, the sheer amount of Educated idiots and Uneducated geniuses I’ve found in life is outstanding. But I know, I know, “bias”, “small sampling”, etc

    Also, what I happened to see from your findings is that: Countries in which a meritocratic-system is applying, successful, often more intelligent people, are having more children.
    In places like Japan, where everybody stopped having kids, it’s the lower classes, often also with low IQ, who have the many, specially due to immense social stability and high quality of living with little competition struggle.

    The Flynn effect, meh, it’s just a phenotype of a genetic trend. “Every society selects for something”, right?

    Like

    • Afrosapiens 🇫🇷🇪🇺 says:

      It’s not that I don’t believe your story but individual anecdotes do not disprove general trends, I think we both know that. Someone who grew up undernourished can be very tall, it doesn’t contradicts the fact that the average height of a population increases as health and nutrition improve. The same has to be expected for IQ which in this case seems to be strongly affected by educational opportunities. But since it’s a highly heritable trait, people receiving the same schooling will still vastly differ in their cognitive outcomes.

      Countries in which a meritocratic-system is applying, successful, often more intelligent people, are having more children.

      I fail to see this pattern, I rather think that countries that offer little educational opportunities to their masses are cognitively held down whereas countries like Spain, Greece and Portugal which came from being authoritarian, kleptocratic societies to modern, more egalitarian ones made huge inter-generational improvements.

      In places like Japan, where everybody stopped having kids, it’s the lower classes, often also with low IQ, who have the many,

      I really don’t know, I’m not aware that Japan has generous child-related benefits. Japan has been stagnating or declining since the 1990s, mainly for macro-economic reasons I suspect but I believe social trends played a role. I can’t exactly tell which ones though.

      The Flynn effect, meh, it’s just a phenotype of a genetic trend. “Every society selects for something”, right?

      No, I don’t agree, there is no genetic correlate of the Flynn effect to my knowledge.

      Like

  3. ian smith says:

    according to peepee’s map, australia and nz have the smartest school age population. how’d that happen? there are still too few asian immigrants to make that big a difference.

    Like

  4. ian smith says:

    oh i see. peepee just pulled that out of her butt. australians have the highest expected IQ scores based on anticipated years of schooling completed.

    Like

    • Afrosapiens 🇫🇷🇪🇺 says:

      Please, we both know peepee is unable to write something so thorough.

      Like

    • GondwanaMan says:

      Afro’s whole post doesn’t even make sense to me. So you’re estimating IQ based off of years of schooling? Which proves years of schooling improves IQ?

      That just sounds like perfectly circular logic. Anyhow, back to sleep.

      Like

    • Afrosapiens 🇫🇷🇪🇺 says:

      Lol, yeah, go back to sleep. No, I’m not proving that years of schooling improves IQ. I’m just trying to see how close to Lynn’s estimates I can get by estimating IQs based on years of schooling. And I actually come very close except for some obvious cases in which Lynn manipulated the data to fit a racial discourse.

      Like

    • No, I’m not proving that years of schooling improves IQ. I’m just trying to see how close to Lynn’s estimates I can get by estimating IQs based on years of schooling.

      But what’s the point of showing Lynn’s figures can be estimated from the causal effect of school on IQ, unless you’re trying to prove the IQ gaps between countries are caused by schooling gaps?

      But you didn’t prove that because most of Lynn’s data comes from kids still in school, so years of schooling is already largely controlled.

      You should have used the causal effect of parental schooling on IQ which can be estimated from adoption studies.

      Like

    • Afrosapiens says:

      But what’s the point of showing Lynn’s figures can be estimated from the causal effect of school on IQ

      The point is seeing whether I find differences between Lynn’s IQs and what has to be expected from years of schooling. Seeing what are the patterns of those differences, like some regions being systematically over-estimated, others being systematically under-estimated. Which is a pattern I found. My figures also show the inter-generational variation in each country which allows me to estimate how strong the Flynn effect is or is supposed to be. Finally, I can see if those new figures correlate better and more invariably with other variables that have been claimed to be related to national IQ and they actually yield stronger and consistent correlations.

      unless you’re trying to prove the IQ gaps between countries are caused by schooling gaps?

      I’m not trying to prove anything. I’m just converting years of education into IQs using a rate of 3.7 points per year of schooling and then expressing it in relation to a British standard of 100.

      But you didn’t prove that because most of Lynn’s data comes from kids still in school, so years of schooling is already largely controlled.

      no, Lynn’s data come from samples of different eras, with different characteristics without a rigorous inclusion methodology that meets the standards of demographic analysis.

      You should have used the causal effect of parental schooling on IQ which can be estimated from adoption studies.

      I need data to create a conversion factor, and I have no such thing on an international level, neither do I have a rate to apply and a standard to compare. All I know about the Parent-child IQ correlation is that it’s only 0.42 when they live together and 0.22 when they live apart. That’s small.

      Like

    • Afrosapiens 🇫🇷🇪🇺 says:

      By the way, I can’t enlarge the charts, so here they are.


      Like

    • ian smith says:

      speaking of how profound an effect environment can have long term.

      i’ve had 6 dogs.

      3 died at age 10, 8, 7.

      3 died at age 13, 13, 13.

      the first 3 were adopted fully grown.

      the second 3 were raised from 6 weeks.

      AND they were all the same breeds. irish setter, airedale terrier, standard poodle.

      Like

    • The point is seeing whether I find differences between Lynn’s IQs and what has to be expected from years of schooling.

      But you’re not looking at what’s expected from years of schooling, you’re looking at what’s expected from years of schooling IF schooling were the ONLY cause of IQ. The 3.7 per year figure comes from a Norwegian study where average schooling increased while other factors stayed constant. That was the point of the study. They wanted to know the effect of schooling on IQ independent of any effect IQ has on schooling so they compared the IQ gap before and after the government intervened to increase the mean level of schooling, thus controlling for most of the confounds that usually separate schooled and less schooled samples (such as different initial IQ and social class background).

      But for most populations that differ in schooling, the causation works in both directions: IQ inequality causes schooling inequality which in turn expands the IQ inequality. Your 3.7 figure only captures the effect of schooling but not the cause and thus underestimates the expected IQ gap between the more or less schooled (though at the extremes you may have overestimated because the IQ schooling relationship is negatively accelerated, not linear as you assumed), but putting that aside, schooling’s a bad way to compare IQs internationally because of the differences in education budgets and academic standards.

      Seeing what are the patterns of those differences, like some regions being systematically over-estimated, others being systematically under-estimated.

      Just because a country has a higher or lower IQ than education predicts, doesn’t mean they’re being systematically underestimated. If one people has a higher mean IQ than another, we’d expect them to score higher even controlling for education. For example, in the U.S., nearly the full black-white IQ gap is seen at every level of schooling. See page 374 of Coming Apart by Charles Murray. Rushton found large racial differences even among university students taking the same classes at the same school.

      Finally, I can see if those new figures correlate better and more invariably with other variables that have been claimed to be related to national IQ and they actually yield stronger and consistent correlations.

      If a country can afford to educate its people it’s not surprising it’s doing well.

      I’m not trying to prove anything. I’m just converting years of education into IQs using a rate of 3.7 points per year of schooling and then expressing it in relation to a British standard of 100.

      You’re not converting anything to anything. You’re misapplying the Norwegian data which meant to show the expected IQ gap for a year of schooling holding all other variables constant.

      And even if you were converting education to IQ, it still makes no sense because they’re two very different variables, like converting weight into height. Samoans weigh more than whites despite being shorter, so let’s convert weight into height to prove they’re actually taller and that the height data is being manipulated by racists.

      #think like Afro

      Like

    • Afrosapiens says:

      But you’re not looking at what’s expected from years of schooling, you’re looking at what’s expected from years of schooling IF schooling were the ONLY cause of IQ.

      Which is likely the case given the previous literature:

      A body of data on IQ collected over 50 years has revealed that average population IQ varies across time, race, and nationality. An explanation for these differences may be that intelligence test performance requires literacy skills not present in all people to the same extent. In eight analyses, population mean full scale IQ and literacy scores yielded correlations ranging from .79 to .99. In cohort studies, significantly larger improvements in IQ occurred in the lower half of the IQ distribution, affecting the distribution variance and skewness in the predicted manner. In addition, three Verbal subscales on the WAIS show the largest Flynn effect sizes and all four Verbal subscales are among those showing the highest racial IQ differences. This pattern of findings supports the hypothesis that both secular and racial differences in intelligence test scores have an environmental explanation: secular and racial differences in IQ are an artifact of variation in literacy skills. These findings suggest that racial IQ distributions will converge if opportunities are equalized for different population groups to achieve the same high level of literacy skills. Social justice requires more effective implementation of policies and programs designed to eliminate inequities in IQ and literacy.

      https://openpsych.net/forum/attachment.php?aid=473 (quoted at the beginning).

      The 3.7 per year figure comes from a Norwegian study where average schooling increased while other factors stayed constant. That was the point of the study. They wanted to know the effect of schooling on IQ independent of any effect IQ has on schooling so they compared the IQ gap before and after the government intervened to increase the mean level of schooling, thus controlling for most of the confounds that usually separate schooled and less schooled samples (such as different initial IQ and social class background).

      Which doesn’t disprove the causal role of schooling on international differences in IQ. It must be assumed that if any country implements the same policies as Norway, they can increase their IQ scores likewise.

      But for most populations that differ in schooling, the causation works in both directions: IQ inequality causes schooling inequality which in turn expands the IQ inequality.

      I sufficiently detail the causes of differences in schooling across the world in the post. However, if a country decides to make schooling a priority in their budget, it is assumed that they can expect results similar to Norway.

      Your 3.7 figure only captures the effect of schooling but not the cause and thus underestimates the expected IQ gap between the more or less schooled (though at the extremes you may have overestimated because the IQ schooling relationship is negatively accelerated, not linear as you assumed), but putting that aside, schooling’s a bad way to compare IQs internationally because of the differences in education budgets and academic standards.

      I actually don’t get extreme results, everything is in the range of previously claimed variation 57-107. Niger’s IQ57 doesn’t surprise me, it’s a very poor rural country where most people live in their multicentury-old ways. And that’s the reason why I don’t even think these numbers measure an intellectual phenotype. Instead I think they measure paper-and-pencil ability. People’s scores reflect how good they are at taking tests, including IQ tests, independently of skills and knowledge. And more schooling mechanically leads to better IQ test performance. It explains the Flynn effect and inter-group variation without explaining raw differences in cognitive ability. Which is, people in all countries can similarly learn things and skills that are relevant to their lifestyle.

      Just because a country has a higher or lower IQ than education predicts, doesn’t mean they’re being systematically underestimated. If one people has a higher mean IQ than another, we’d expect them to score higher even controlling for education. For example, in the U.S., nearly the full black-white IQ gap is seen at every level of schooling.

      That’s because blacks, at every level of income are much more likely to attend urban schools where more baby-sitting than teaching is going. But I doubt such differences show up in internal comparisons since pupils drop out nearly as early as classes become to difficult for them to catch up, so the enrollment rate virtually tells the intellectual ability of the school-age population. On the contrary, schools in the developed world keep pupils if school and graduate them even if they’re barely learning anything.

      Rushton found large racial differences even among university students taking the same classes at the same school.

      Tells nothing about what types of schools they attended before college.

      If a country can afford to educate its people it’s not surprising it’s doing well.

      I’m not finding correlations with government school budgets, I’m finding correlations with GDP. Which is different and the causal direction is easy to understand: more education/IQ gives a country the ability to add more value-add to its economy (GDP = value added). But the most important part is that my correlations with all measures I tested (life expectancy, fertility, GDP/capita, HDI, homicide) aren’t only stronger, they are the same no matter what country I include in my dataset. So whatever I measured, it has more explanatory power for the main measures of socio-economic development than Lynn’s IQs and debunks his claim that his IQs are the most important cause of variation in national prosperity.

      You’re not converting anything to anything. You’re misapplying the Norwegian data which meant to show the expected IQ gap for a year of schooling holding all other variables constant.

      I’m correctly applying the Norwegian data in the most straightforward sense, and I find better associations with all aspects of development than Lynn does, irrespective of the causal direction.

      And even if you were converting education to IQ, it still makes no sense because they’re two very different variables, like converting weight into height. Samoans weigh more than whites despite being shorter, so let’s convert weight into height to prove they’re actually taller and that the height data is being manipulated by racists.

      No, IQ test performance is for a large part an artifact of education instead of an independent real biological phenotype. And the IQ/education correlation is much higher than the height/weight one, which means that at every level of education, the variation in IQ is lower than the variation in weight at each level of height. It’s easy to explain why Samoans are heavier than expected from their height: http://edition.cnn.com/2015/05/01/health/pacific-islands-obesity/index.html

      What you could not explain however is how Barbados, Jamaica, Gabon, Botswana or the Bahamas manage to get well developed economies and high quality of life in spite of IQs in the range of mental retardation according to Lynn. Lynn who, in case you don’t know, is a self-declared racist of the most vicious type.

      #think like Afro

      And that’s a lot better than making up a theory on why Bushmen did not make the leap to agriculture in the Kalahari desert. So try thinking like me, you’d probably have a much higher quality blog than what you have now.

      Like

    • RaceRealist says:

      Re IQ tests:

      See this article by Ken Richardson: What IQ Tests Test:

      The controversies and debates that result are well known. This paper brings together results and theory rarely considered (at least in conjunction with one another) in the IQ literature. It suggests that all of the population variance in IQ scores can be described in terms of a nexus of sociocognitive-affective factors that differentially prepares individuals for the cognitive, affective and performance demands of the test—in effect that the test is a measure of social class background, and not one of the ability for complex cognition as such. (p. 283)

      On the height/weight example:

      A correlation between test scores does not necessarily mean that they are measuring the same thing. As Raven et al. (1993) put it, ‘height and weight are correlated to much the same extent as “academic abilities”—yet height and weight are clearly not the same thing’ (p. G8) (p. 300)

      Also see Does IQ Really Predict Job Performance?

      See also Ken Richardson’s book review of Jensen’s The g Factor: The Science of Mental Ability:

      DEMYSTIFYING G: Book Review of Jensen on Intelligence-g-Factor

      Jensen also writes on p. 48 of The g Factor:

      My study of these two symposia and of many other equally serious attempts to define “intelligence” in purely verbal terms has convinced me that psychologists are incapable of reaching a consensus on its definition. It has proved to be a hopeless quest. Therefore, the term “intelligence” should be discarded altogether in scientific psychology, just as it discarded “animal magnetism” as the science of chemistry discarded “phlogiston.” “Intelligence” will continue, of course, in popular parlance and in literary usage, where it may serve a purpose only because it can mean anything the user intends, and where a precise and operational definition is not important.

      Like

    • RaceRealist says:

      What is most important, in my opinion, for individual differences in IQ is a theory of individual intelligence differences. As far as I know, there is no such theory. Why not? Jensen and Deary state that there is no theory of why individuals differ in g or IQ tests. I think that’s a huge problem. A quote from Richardson’s new book Genes, Brains, and Human Potential: The Science and Ideology of Intelligence (p. 104):

      Intelligence is viewed as the most important ingredient of human potential. But there is no generally accepted theoretical model of what it is (in the way that we have such models for other organic functions). Instead, psychologists have adopted physical metaphors: mental speed, energy, power, strength, and so on, together with simple genetic models of how it is distributed in society. The IQ test was invented to create scores that correspond with such metaphors, with the distribution—who is more or less intelligent—already presumed.

      This circularity in IQ testing must not be forgotten or overlooked. IQ tests do not have what is called “construct” validity, in the way that breathalyzer is calibrated against a model of the passage of alcohol in the bloodstream. They are constructed on the basis of prior beliefs of who is or is not intelligent. But by creating a numerical surrogate of a social class system, they make that system appear to spring from biological rather than social forces. Such ideas are dangerous, because they demean the real mental abilities and true potential of most people in everyday social situations.

      Ken Richardson has constructed a theoretical model of intelligence, the basis of which are intelligent cells and intelligent physiology. His dynamic/intelligent systems/physiology theory is great, and could explain the emergence of intelligence, as well as the evolution of new species.

      Whatever the case may be, there is no hard theory for individual differences in g (whatever that is), and no agreed-upon definition intelligence. Without getting past these two hurdles, the “IQ research community” has a lot of ground to cover.

      Like

    • In eight analyses, population mean full scale IQ and literacy scores yielded correlations ranging from .79 to .99.

      Literacy score != years of education.  And group level correlations tend to be higher than individual level correlations.

      It must be assumed that if any country implements the same policies as Norway, they can increase their IQ scores likewise.

      Your analysis assumes that differences in schooling are the only cause of international IQ gaps which is debunked by the fact that these IQ gaps occur at ages when kids are still in school.   In one of the largest African IQ studies ever done (Owen, 1992) 1,093 black South Africans drawn from 28 schools, had a mean Raven IQ below 70.  This very low score can not be caused by the kids leaving school early because the sample was still in school so your numbers explain nothing. 

      Now you could argue it’s PARTLY caused by their parents leaving school early, but then you’d have to use the independent effect of parental IQ on one’s IQ; you used the independent effect of one’s own schooling on one’s IQ which makes no sense given how early in life international IQ gaps appear.

      I’m not finding correlations with government school budgets, I’m finding correlations with GDP.

      One would expect GDP to be related to school budgets.

       Which is different and the causal direction is easy to understand: more education/IQ gives a country the ability to add more value-add to its economy (GDP = value added).

      One would expect the causation to flow in BOTH directions, hence the high correlations.

      But the most important part is that my correlations with all measures I tested (life expectancy, fertility, GDP/capita, HDI, homicide) aren’t only stronger, they are the same no matter what country I include in my dataset. So whatever I measured, it has more explanatory power for the main measures of socio-economic development than Lynn’s IQs and debunks his claim that his IQs are the most important cause of variation in national prosperity.

      Even if Lynn did claim that, you haven’t debunked him because you haven’t proved education is the cause of these high correlations.  It could be GDP + IQ causing education causing more GDP etc.  Rich smart countries get educated and get richer.  Shocking!

       and I find better associations with all aspects of development than Lynn does, irrespective of the causal direction.

      Lynn’s data is not that reliable for individual countries since it’s just based on whatever studies he could find in the literature,  so it doesn’t surprise me that authoritative data from the United Nations is more predictive, especially since education reflects not only the national IQ, but also the skills, work ethic, resources, and values of the country.

      And yet your own matrix shows Lynn’s data better predicting life span than your data (0.84 vs 0.76) and homicide (-0.35 vs -0.01).

      You also may have tried to manipulate the data by including only the most populous countries.

      And the IQ/education correlation is much higher than the height/weight one,

      Not much higher.  The IQ-education correlation is about 0.57 for full-scale IQ and 0.47 for performance IQ (see table 4.6):

      https://books.google.ca/books?id=ee4KTFdIabAC&pg=PA117&lpg=PA117&dq=IQ+WAIS+years+of+education&source=bl&ots=4S0ceFrmV8&sig=EGZkh11cwByFrlfkXopbcNnP4ZI&hl=en&sa=X&ei=U8KiVZ3sJtjToAStoJyQDQ&ved=0CEgQ6AEwBw#v=onepage&q=IQ%20WAIS%20years%20of%20education&f=false

      By contrast the correlation between weight and squared height is about 0.45 in adults and 0.81 in kids.

      http://ocean.sci-hub.bz/129237bbed1564b53a62f51e3d2bedb6/10.1016%40j.endonu.2013.06.001.pdf

      In adult athletes, the correlation is 0.78:

      http://www.nzqa.govt.nz/assets/qualifications-and-standards/qualifications/ncea/NCEA-subject-resources/Mathematics/91581/91581-EXP-student4-001.pdf

      What you could not explain however is how Barbados, Jamaica, Gabon, Botswana or the Bahamas manage to get well developed economies and high quality of life in spite of IQs in the range of mental retardation according to Lynn.
       
      Lynn’s data is not reliable at the country level because he only has a few studies of questionable quality per country.  Where his data might be a lot more reliable is at the regional level (i.e. sub-Saharan Africa, East Asia, Northwest Europe etc) because by averaging many countries in a region, the errors for individual countries cancel out.
       
      Having said that, I agree that mental retardation for entire countries is absurd but that doesn’t necessarily mean Lynn manipulated the data, it could just mean paper-pencil IQ tests are culturally biased for people in less developed countries because as Nell (2000) argued “they are less test-wise, less interested, more anxious, work less efficiently, or give up sooner on items they find difficult”

      Like

    • Afrosapiens 🇫🇷🇪🇺 says:

      Literacy score != years of education.

      Indeed, education and the time spent in school obviously is the main thing that improves literacy scores in life and it probably explains most of the Flynn effect. Then individual reading, commitment to schoolwork and teaching standards can causes differences in the way individuals acquire those literacy skills but I think years of education is the best proxy we have for literacy. It’s better than literacy rate because it makes no difference between different levels of mastery beyond the ability to read and write simple sentences.

      There is a reason why literacy, years education and IQ are all interrelated and good proxies for each other. If you look at the WAIS-IV’s subtests g-loadings, the verbal parts are more g-loaded than the performance ones. The highest being vocabulary.

      https://books.google.fr/books?id=vMR9b7dshrQC&pg=PA49&lpg=PA49&dq=subtest+g+loading&source=bl&ots=7FikGNWy2q&sig=o2WEo8mFeUTdxpaLt-T52l9oEFY&hl=en&sa=X&ved=0ahUKEwjuqOCvlK_WAhVQJ1AKHQbWBwEQ6AEIYzAL#v=onepage&q=subtest%20g%20loading&f=false

      Likewise, reading proficiency is an important determinant of school achievement for easily understandable reasons and it improves brain function.

      http://www.ala.org/aasl/sites/ala.org.aasl/files/content/aaslpubsandjournals/slr/vol3/SLMR_IndependentReading_V3.pdf
      https://www.theguardian.com/books/2014/jan/23/can-reading-make-you-smarter

      So everything adds up. “g” is a measure of literacy, which is under socio-cultural influence. It explains group differences and the Flynn effect better than any other model.

      Your analysis assumes that differences in schooling are the only cause of international IQ gaps which is debunked by the fact that these IQ gaps occur at ages when kids are still in school.

      They do not attend the same schools, the 1992 study you quote was from apartheid South Africa. Lol! how can you claim to control for anything in apartheid South Africa? Blacks and whites lived on different planets. And all the other studies you can mention are based on samples that do not meet the representativeness of professional demographic analysis.

      Now you could argue it’s PARTLY caused by their parents leaving school early, but then you’d have to use the independent effect of parental IQ on one’s IQ;

      That’s what I partly do by averaging adult and children schooling characteristics.


      you used the independent effect of one’s own schooling on one’s IQ which makes no sense given how early in life international IQ gaps appear.

      It makes more sense than anything you wrote on your blog. My measure of children performance is school life expectancy, which is not exactly one’s schooling, it’s the time they’re expected to stay in school based on current enrollment rates. It means that when you claim Lynn’s samples show gaps appearing early, it only means that even if still in school, a large part of the schooled children are at risk of leaving school early because they’re not learning. And that translates in lower school life expectancy.

      One would expect GDP to be related to school budgets.

      Partly, but not exactly, countries with the same GDP/capita differ vastly in the share of national income that is spent on education.

      One would expect the causation to flow in BOTH directions, hence the high correlations.

      Education increasing value added makes more sense than value added increasing education. Because value added wont be spent on education to the same extent in all countries.

      Even if Lynn did claim that, you haven’t debunked him because you haven’t proved education is the cause of these high correlations. It could be GDP + IQ causing education causing more GDP etc. Rich smart countries get educated and get richer. Shocking!

      I did prove that because my correlations are not just stronger, they’re invariant.

      Lynn’s data is not that reliable for individual countries since it’s just based on whatever studies he could find in the literature

      You mean every study that could make his point.

      And yet your own matrix shows Lynn’s data better predicting life span than your data (0.84 vs 0.76) and homicide (-0.35 vs -0.01).

      Again, what matters the most here is invariance, international consistency of correlation coefficients. Lynn’s data losing predictive power once Africa is removed prove that that his correlations are only artificially higher due to the fake manipulated African scores. My correlations maintaining their predictive power with and without level prove worldwide reliability of my estimates.

      You also may have tried to manipulate the data by including only the most populous countries.

      No, these correlations are still worth for 70% of the world’s population and it removes oil-rich countries with very high GDP per capita and other extreme outliers, it also gives world regions the same weight in the correlations. Because when there are 50 African countries and only 5 in East Asia, it gives way too much weight to the African data in the calculation of the correlation coefficient.

      Not much higher. The IQ-education correlation is about 0.57 for full-scale IQ and 0.47 for performance IQ (see table 4.6):

      It explains 50% more variance, and IIRC, the correlation you cite is about school grades, not years of schooling which is considered a good proxy for IQ in GWAS studies. One would never say height genes are a good proxy for weight genes.

      Lynn’s data is not reliable at the country level because he only has a few studies of questionable quality per country. Where his data might be a lot more reliable is at the regional level (i.e. sub-Saharan Africa, East Asia, Northwest Europe etc) because by averaging many countries in a region, the errors for individual countries cancel out.

      Lol! there is no such thing, no other international data is calculated this way. No one would say GDP figures are more reliable on the continental level than on the country level. Because there is high variance in GDP and GDP-influencing factors within a continent so it would be absurd to say since two countries have the same racial majority and are neighbors, they must be just as rich. The Dominican Republic is richer than Haiti (which is black), but it’s much poorer than Barbados (which is black too) So you need to use national data instead of a “mean Caribbean GDP” that would not reflect the large differences within the region. Saying continental level data is more reliable than country data is assuming that continental ancestry significantly influences country average. And the errors you mention are not errors, they are intentional selection of unrepresentative data. Don’t get it twisted, when serious analysts have no reliable data, they just don’t report estimates.


      Having said that, I agree that mental retardation for entire countries is absurd but that doesn’t necessarily mean Lynn manipulated the data it could just mean paper-pencil IQ tests are culturally biased for people in less developed countries because as Nell (2000) argued “they are less test-wise, less interested, more anxious, work less efficiently, or give up sooner on items they find difficult

      IQ scores in the mental retardation range not absurd, they’re only absurd if we assume that it truly reflects actual functioning instead of only test scores. But it’s truly absurd to believe that a country like Jamaica is run by retards or has a school system that doesn’t accustom its population to test-taking skills that are reflected in IQ scores higher than 80. By estimating advanced black countries in the Caribbean in the 70s and by estimating under-developed China above 100, Lynn had no intent to give a cultural-bias explanation to account for this. You’re too gullible.

      Like

    • RR, Jensen was making progress towards a theory of individual differences before he died. See his book Clocking the Mind

      Like

    • Indeed, education and the time spent in school obviously is the main thing that improves literacy scores in life and it probably explains most of the Flynn effect.

      They’re related but I repeat, literacy scores != years of education so citing the high correlation between literacy scores and IQ does not prove your nonsense claim that years of schooling is the only cause of national IQ gaps.  Even among people with the same years of completed schooling, literacy scores differ enormously.

      They do not attend the same schools, the 1992 study you quote was from apartheid South Africa. Lol! how can you claim to control for anything in apartheid South Africa?

      Which shows the absurdity of you claiming years of education is the only cause of national IQ gaps.  Countries can differ in all kinds of ways that affect IQ, beyond just mean schooling, with apartheid being an example.

      That’s what I partly do by averaging adult and children schooling characteristics.

      All you’re doing is averaging the expected IQ of adults with the expected future IQ of kids based on the wrong assumption that years of schooling is the only cause of IQ.  

      My measure of children performance is school life expectancy, which is not exactly one’s schooling, it’s the time they’re expected to stay in school based on current enrollment rates. It means that when you claim Lynn’s samples show gaps appearing early, it only means that even if still in school, a large part of the schooled children are at risk of leaving school early because they’re not learning. And that translates in lower school life expectancy.

      So now that your claim that years of schooling is the only cause of national IQ differences has been debunked, you’re now claiming that even while still in school, countries that are X numbers of years less schooled than Great Britain are also X number of years behind in actual learning even before they dropout.

      If so, by the time they’re adults they are effectively 2X years behind Great Britain in schooling, so by your method, all the IQ gaps should be double by adulthood.  So when you claim Africa’s mean IQ is 72, because they average 7.57 years less schooling than Great Britain,  and each missed year deducts 3.7 points,  you’re actually claiming that by adulthood, they’re effectively 15.14 years less schooled than Great Britain.

      Great Britain IQ – 15.14(3.7) = Adult African IQ of 44

      Absurd!

      Partly, but not exactly, countries with the same GDP/capita differ vastly in the share of national income that is spent on education.

      But the two are correlated so the correlation between national education and GDP is partly just rich countries being able to afford to educate the masses.

       Education increasing value added makes more sense than value added increasing education.

      They’re not mutually exclusive.  The causation would work in both directions thus making the correlation extra high.

      Again, what matters the most here is invariance, international consistency of correlation coefficients. Lynn’s data losing predictive power once Africa is removed prove that that his correlations are only artificially higher due to the fake manipulated African scores. My correlations maintaining their predictive power with and without level prove worldwide reliability of my estimates.

      Actually it may show just the opposite.  Correlations are expected to decline when you restrict the range of scores which is what you did by removing African countries:

      http://davidmlane.com/hyperstat/A68809.html

      Lynn’s declining predictive power is exactly as expected.  It’s your numbers that are behaving suspiciously.

      No, these correlations are still worth for 70% of the world’s population and it removes oil-rich countries with very high GDP per capita and other extreme outliers, it also gives world regions the same weight in the correlations. Because when there are 50 African countries and only 5 in East Asia, it gives way too much weight to the African data in the calculation of the correlation coefficient.

      Such arbitrary decisions on your part create the appearance of data manipulation.

      It explains 50% more variance,

      Height explains FAR MORE of the variance in childhood weight than IQ explains variances in adult education.

      and IIRC, the correlation you cite is about school grades, not years of schooling

      No table 4.6 specifically says “years of education”

      which is considered a good proxy for IQ in GWAS studies. One would never say height genes are a good proxy for weight genes.

      That’s because it’s assumed that the non-IQ components of education are non-genetic, yet the non-height component of weight is still seen as genetic.  The former assumption is false however:

      https://www.pnas.org/content/111/42/15273.full.pdf

      Lol! there is no such thing, no other international data is calculated this way. No one would say GDP figures are more reliable on the continental level than on the country level.

      That’s because for other international data points we have excellent country by country data points.

      Saying continental level data is more reliable than country data is assuming that continental ancestry significantly influences country average.

      No, it assumes nothing about cause.  It’s just the well known principle of aggregation:

      https://link.springer.com/article/10.3758/BF03208007

      Don’t get it twisted, when serious analysts have no reliable data, they just don’t report estimates.

      Actually they do, particularly in fields where they don’t have the luxury of reliable data, for example in anthropology they might estimate the brain size of an extinct hominin based on just a couple skulls of unknown representativeness.  These are just considered the best estimates we can make at the time, and as more and better data comes in, they’re revised.

      By estimating advanced black countries in the Caribbean in the 70s and by estimating under-developed China above 100, Lynn had no intent to give a cultural-bias explanation to account for this.

      I know Lynn’s numbers seem very wrong for Australian aboriginals so I would not be surprised if he’s wrong or biased in a lot of other areas too, but your article didn’t land a single punch.

      You’re too gullible.

      No gullible is you arguing the average adult African IQ is 44 and not even realizing you’re arguing that.

      Like

    • Afrosapiens 🇫🇷🇪🇺 says:

      They’re related but I repeat, literacy scores != years of education so citing the high correlation between literacy scores and IQ does not prove your nonsense claim that years of schooling is the only cause of national IQ gaps.

      The relation is not coincidental.

      -verbal tasks are the most g-loaded
      -reading improves cognition
      -schooling teaches reading
      -IQ improves as schooling expands

      Logical conclusion: IQ is a proxy for literacy

      This makes more sense than all your estimations of IQ from “number of splits” or brain size. The former being based on nothing, the latter being based on a much weaker correlation coefficient than literacy or education.

      Even among people with the same years of completed schooling, literacy scores differ enormously.

      Not enormously, but it’s due to the fact that in Western countries, students are kept in school even if they’re not learning due to legal school-leaving age and lowering standards for students to graduate. There is no such thing in the developing world where students only stay in school if they’re learning, or if they can afford to stay in school (which is correlated).

      Which shows the absurdity of you claiming years of education is the only cause of national IQ gaps. Countries can differ in all kinds of ways that affect IQ, beyond just mean schooling, with apartheid being an example.

      I’m not exactly making this claim, I’m claiming that IQ estimates make much more sense if they’re calculated as if years of schooling are treated as the only cause of variance. I accept an error margin of 5 points, but not an error margin of 20 points as seen in China or Jamaica. And the only reasons to explain Lynn’s senseless estimates is that he wanted to make it seem like race is the only predictor of IQ. Otherwise, you need to explain why the Chinese have such poor educational and socio-economic indicators and despite a supposedly superior IQ which is claimed to be the best predictor of those things. And you have to explain why Jamaica is in the opposite situation.

      So now that your claim that years of schooling is the only cause of national IQ differences has been debunked, you’re now claiming that even while still in school, countries that are X numbers of years less schooled than Great Britain are also X number of years behind in actual learning even before they dropout.

      No, you don’t get it. Those who remain in school are learning something, those who persist into college probably have learned more than Britons who have watered down diplomas. You’re forgetting that many in the developing world are schooled in excellent private schools or have private instructors to help them. Those who drop out are mostly poor, rural and attending underfunded public schools. School life expectancy roughly estimates the size of this population that brings the national average down.

      If so, by the time they’re adults they are effectively 2X years behind Great Britain in schooling, so by your method, all the IQ gaps should be double by adulthood. So when you claim Africa’s mean IQ is 72, because they average 7.57 years less schooling than Great Britain, and each missed year deducts 3.7 points, you’re actually claiming that by adulthood, they’re effectively 15.14 years less schooled than Great Britain.

      Great Britain IQ – 15.14(3.7) = Adult African IQ of 44

      Absurd!

      What’s truly absurd is the strawman you’re attacking. I make no claim on an average African IQ to begin with. The African IQs that I estimate go from 58 to the high 80s. Secondly, your assumption that I’m implying that years of education could be twice less effective in improving IQ in Africa is absurd, which is why neither me or anyone else estimates an IQ of 44 for African adults.

      Now, would you care to explain which data point looks anomalous to you, and why?

      But the two are correlated so the correlation between national education and GDP is partly just rich countries being able to afford to educate the masses.

      Partly, but international income variation is more subtle than just rich versus poor, and countries of similar GDP/capita can have vastly different school expenditure.

      They’re not mutually exclusive. The causation would work in both directions thus making the correlation extra high.

      No, oil rich countries trump the correlation for instance, and their lower IQ than expected from GDP reflects their severe lack of education spending.

      Actually it may show just the opposite. Correlations are expected to decline when you restrict the range of scores which is what you did by removing African countries:

      http://davidmlane.com/hyperstat/A68809.html

      Lynn’s declining predictive power is exactly as expected. It’s your numbers that are behaving suspiciously.

      No, for instance, Lynn’s correlations are higher within the top 20 than in the whole list. So range restriction does not affect Lynn’s correlations.

      Such arbitrary decisions on your part create the appearance of data manipulation.

      It’s a well justified decision, I wanted to remove outliers like countries at war, resource rich countries, and not having Luxembourg given the same weight as China in my calculation. So I chose to include only the top 20 countries for my data to still be representative of 70% of the world’s population. Lynn’s correlations did not suffer from my choice, and my correlations were stronger in both lists.

      Anyway, you’re forced to acknowledge that years of education are a better predictor of every outcome that Lynn claims to be caused by IQ than IQs estimated by Lynn himself.

      Height explains FAR MORE of the variance in childhood weight than IQ explains variances in adult education.

      Who cares? You know full well that obesity becomes higher in adulthood. And what we’re discussing is the analogy you made with Samoans, which made no sense. The cause of their obesity is known, what is unknown is the weight/height correlation among them. I use years of schooling as a good proxy for IQ, which makes much more sense than you using brain size and just saying “some exceptions have to be expected” when things do not add up. Again, tell me which country estimates do not add up in my chart.

      That’s because for other international data points we have excellent country by country data points.

      My estimates are based on such excellent country by country data points. As a result, all the correlations I find are stronger and invariant.

      No, it assumes nothing about cause. It’s just the well known principle of aggregation:

      A principle that you’re misinterpreting.

      The principle of aggregation states that the sum of a set of multiple measurements is a more stable and representative estimator than any single measurement.

      It applies when you’re measuring the same thing. It’s absurd to estimate an “African IQ” unless you assume that a Nigerian sample can be treated as representative of a Gabonese sample. If you say they’re representative, you must justify it, and Lynn’s justification is that countries of similar ancestry have the same IQ.

      Actually they do, particularly in fields where they don’t have the luxury of reliable data, for example in anthropology they might estimate the brain size of an extinct hominin based on just a couple skulls of unknown representativeness.

      Anthropology more often presents these characteristics as a range, instead of a confirmed average in the way Lynn says the average IQ of Africa is 68 without mentioning the possibility of high variation between countries.

      Anyway, anthropology is a thing, demographics and economics statistics do not publish data of insufficient reliability, neither do they write books about it.

      I know Lynn’s numbers seem very wrong for Australian aboriginals so I would not be surprised if he’s wrong or biased in a lot of other areas too, but your article didn’t land a single punch.

      It did, it did prove that whatever the meaning of my estimates, they’re a better predictor of everything Lynn claims to be caused by his national IQ.

      No gullible is you arguing the average adult African IQ is 44 and not even realizing you’re arguing that.

      Can you tell me where I’m estimating an average adult African IQ? You’re the only one making things nonsensical here. Look at your blog, it’s a complete disaster.

      Like

  5. rw95 says:

    I just cannot understand how South Asian IQ is so low. How can Indian IQ be lower than Amerindian IQ when India was vastly more important than south America in world history, as well as contributing much more in terms of science and culture? It just doesn’t make sense to me. And if Indian IQ is basically African-tier, how is it Sri Lanka scores in the low 90s?

    Like

    • Afrosapiens 🇫🇷🇪🇺 says:

      Because they are poor countries with low education. IQ is not about race

      Like

    • Because his numbers are not the IQs of those countries, they’re the IQs countries would have if years of schooling were the only cause of national IQ differences.

      Like

    • Afrosapiens 🇫🇷🇪🇺 says:

      Because his numbers are not the IQs of those countries

      Neither are Lynn’s, they’re the unweighted average of un-professionally selected samples.

      they’re the IQs countries would have if years of schooling were the only cause of national IQ differences.

      Which is likely the case.

      Like

    • pithom says:

      @Afrosapiens It’s obviously not the case. Compare Saudi Arabia to Finland, Vietnam to South Africa, China to Egypt.

      India may have had a higher environment-adjusted avg IQ in earlier times, or, more likely, the difference is just due to different population densities, which was a more important variable than avg IQ in ancient times.

      Like

    • Afrosapiens 🇫🇷🇪🇺 says:

      South Africa is more developed than Vietnam, China slightly more than Egypt, Saudi Arabian citizens are richer than Finns.

      or, more likely, the difference is just due to different population densities, which was a more important variable than avg IQ in ancient times.

      Yes, IQ makes no sense in historical comparisons.

      Like

    • pithom says:

      We’re not talking about development here, but average national IQ. South Africa obviously has a lower avg IQ than Vietnam, Saudi Arabia much lower than Finland, Egypt much lower than China. This is clear from the test score data, as well as from the bulk of anecdotal evidence.

      Like

    • Drake says:

      Pithom, what about dismissing Afrosapiens’s data and correlation to make your point. You say that his IQ average can’t be true without any correct arguments so go ahead, destroy his correlations

      Like

  6. more on twitch fiber bullshit.

    as people get older their fiber composition changes. the type ii are converted to type i.

    then there’s kim collins. the fastest time he ever ran he ran at age 40.

    9.93s

    While the 9.96 will remain as the M35 World Record, at age 40 years, 54 days, in a new age category, Collins improved his national record to 9.93 +1.9 at the NRW Gala in Bottrop, Germany.[4] He is the first man over age 40 to break the 10 second barrier. This yet again extended his own record as the oldest man to run a sub 10 second 100m, which continues with each sub-10 performance.

    voltaire is unpopular in france. french is now JEWISH . sad!

    voltaire said: I Disapprove of What You Say, But I Will Defend to the Death Your Right to Say It

    Like

    • RaceRealist says:

      Put this in the relevant threads on fiber typing. Thanks.

      Like

    • Phil78 says:

      “as people get older their fiber composition changes. the type ii are converted to type i.

      then there’s kim collins. the fastest time he ever ran he ran at age 40.”

      Was this finding consistent with people who trained or did this apply just for those that didn’t?

      Like

    • Afrosapiens 🇫🇷🇪🇺 says:

      Phil, please reply to off-topic comments on the other thread.

      Like

  7. Jm8 says:

    (approx) 90 for barbados certainly makes more sense than the 79 figure given by Lynn.
    I figured that that it was likely to be too/inaccurately low (not too surprising since it’s from Lynn), and that the real average was likely closer to the approximately 91-93 for the Bahamas, and the somewhat higher figure for Bermuda. I look forward to seeing more data from the region (and others).

    http://humanvarieties.org/2013/03/12/hvgiq-the-bahamas/

    http://humanvarieties.org/2013/05/03/hvgiq-bermuda/

    Like

    • Afrosapiens says:

      Yeah, the low 90s is what we must expect from Barbados’ socio-economic indicators. Contrary to Lynn’s estimates, there is no black nation in the Caribbean that can realistically score below 80, except Haiti for which I estimated an IQ of 73, which is still higher than the scores in the 60s that Lynn and even Malloy estimated for this country. Subsaharan Africa’s middle-income economies also score in the 80s.

      Lynn’s estimates are worthless.

      Like

    • Jm8 says:

      “Jm8 = peepee.”

      I’m definitely not PP, that’s for sure.

      (perhaps I’m misunderstanding something)

      Like

    • Afrosapiens says:

      Don’t mind mug of pee.

      Peepee shows signs of mental retardation that no one could fail to notice.

      Like

    • Jm8 says:

      Edit: “(perhaps I’m misunderstanding something—I most likely am)”

      Like

    • Jm8 says:

      “how much does learning to speak english increase a frenchman’s IQ? another very important question.”

      I wondered that too, just after reading your last post (on Anglophone vs francophone blacks.

      Like

    • Afrosapiens 🇫🇷🇪🇺 says:

      Don’t mind mug of pee, he’s trolling and since he uses my gravatar, his comments are now in the trash.

      Like

    • Afrosapiens 🇫🇷🇪🇺 says:

      “RealAfrosapiens” is mug of pee/mugabe/ian smith/ilovehitler/whatever stupid name.

      Now about African countries, first there aren’t real Francophone/Anglophone Africans, these countries have official languages that are inherited from colonialism but virtually none is spoken as a mother tongue by Black Africans.

      Secondly, I don’t see a real difference. The lowest of the lowest estimates are in the Sahelian belt and the horn of Africa. These countries are very rural and barely touched by modernity. But I can’t see a link with French colonialism, since countries like Benin, Togo, Cameroon, the Congo(s) and Gabon which are more modern also have higher estimates, the highest being Gabon: 84.

      Like

    • Jm8 says:

      Edit: “…reading your last comment (on anglophone vs. francophone…”

      Like

    • Jm8 says:

      That was him? Strange that I didn’t notice the different name (from yours). That’s weird, well don’t I feel dumb. It makes sense now that it was a troll

      Like

    • Afrosapiens 🇫🇷🇪🇺 says:

      LOL! Never mind.

      Like

    • Jm8 says:

      “Now about African countries, first there aren’t real Francophone/Anglophone Africans, these countries have official languages that are inherited from colonialism but virtually none is spoken as a mother tongue by Black Africans.

      Secondly, I don’t see a real difference. The lowest of the lowest estimates are in the Sahelian belt and the horn of Africa. These countries are very rural and barely touched by modernity. ”

      Yeah, that makes a lot more sense. I should have checked the data.

      Like

    • Jm8 says:

      Edit: “…Yeah, that makes a lot more sense. I should have checked the data—a bit of a mental lapse on my part I guess..

      Like

    • Afrosapiens 🇫🇷🇪🇺 says:

      It happens 😉

      Like

  8. Gary Nobelheim says:

    I can’t get behind some of these numbers. Spain adult average at 87? An 85 IQ precludes you from most basic work competency, according to neuroscientist, Richard Heiar. I can’t imagine half of Spain’s population being completely extraneous. These numbers have to be wrong.

    Like

    • Afrosapiens 🇫🇷🇪🇺 says:

      I was equally surprised by Spain’s adult IQ below 90. But here is how I explained it to myself:

      -Spain’s adult population is very old and grew up under a fascist dictatorship, when the country was poor, education was poorly funded and brain drain was high. We see the same low adult IQ in Portugal and Greece. These three southern European country have the same history and a large inter-generational gap, which is indicative of a large Flynn effect. I wonder how developed they would if they hadn’t joined the European Union which highly subsidized them. We have clues that Greece is not entirely able to manage a first world economy independently like Germany and France do.

      I don’t believe in IQ thresholds. 85 is the supposed IQ of African American adults, and obviously, much more than 50% of African adults can be competent workers, they’ve been working in factories since the 1920s. I don’t even know what is “basic work competency”. Even in Niger, where I found an adult IQ of 57, people engage in crafts, agriculture, animal husbandry and trade and support themselves independently. You don’t see a population of impotent adults there.

      My opinion on IQ tests is not that they reflect “intelligence” but IQ test competency and some aspects of logical thinking that depend on exposure to formal schooling.

      These numbers have to be wrong.

      I think your interpretation is wrong. Because whatever you think of these numbers, in the case of Spain, it’s the conversion of an average educational attainment equaling middle school dropout level.

      Like

    • Afrosapiens 🇫🇷🇪🇺 says:

      That’s what I found on education and skills among Spanish adults.

      http://www.oecd.org/skills/piaac/Country%20note%20-%20Spain.pdf
      https://www.oecd.org/education/Spain-EAG2014-Country-Note.pdf

      The OECD is the organization that manages the PISA ranking.

      Like

  9. James Dalon says:

    Wow, I’m really surprised to see a race realist allowing a black man to give an credible opinion on IQ.

    Don’t get me wrong, I’m not saying that race realist can’t be wrong but it’s nice to see people in this site actually trying to reach for an honest conversation.

    People like tara mcarthy and the alternative hypothesis (Ryan Faulk) are far to narcissic to argue honestly, everyone has bias but Ryan and Tara can’t minimize their bias, in contrary to the owner of this site.

    Kudos!!!!!

    Liked by 1 person

  10. Hey Afrosapiens! Great article but pumpkinperson posted a criticism of your article which you didn’t respond: https://notpoliticallycorrect.me/2017/09/05/worldwide-iq-estimates-based-on-education-data/#comment-4231

    I’m not claiming that you’re wrong and I’m not claiming that pumpkinperson is wrong, I’m just interested by your opinion on his comment.

    Thanks

    Like

  11. James Dalon says:

    By the way Afrosapiens, what do you think of the alternative hypothesis last video?

    It would be nice to have an great honest opposing view here.

    Like

    • Afrosapiens 🇫🇷🇪🇺 says:

      Hi, sorry I’ve been busy lately. Back to you soon.

      Like

    • Michel Dupont says:

      Take your time but be careful. The alternative hypothesis apparently edited the video he responded too to make it look like “shit”

      Here’s the original video he responded too. There is an ongoing war between the alternative hypothesis and black lightning.

      I’m just curious about your thoughts on their arguments. No need to be precise though.

      Like

    • Afrosapiens 🇫🇷🇪🇺 says:

      I played most of both videos. I completely side with black lightening. I will elaborate on my opinion in a more detailed reply later.

      Like

  12. Paprika says:

    Which IQ test are the most reliable then Afro? Since you proved that Lynn’s data is flawed.

    Like

    • Afrosapiens says:

      Sorry, your comments were marked as spam.

      Which IQ tests are the most reliable? I don’t think one is better than an other, Lynn’s data is biased because of confirmation bias in data selection.
      There are many tests on the market, some pretend to be culture-fair so they are more often administered to non-western populations. But even the culture-fairness assumption is disputed.

      Otherwise, when it comes to Westerners, the Welscher Adult Intelligence Scales is the most widely used test in psychology.

      Like

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

Please keep comments on topic.

Charles Murray

Arthur Jensen

Blog Stats

  • 215,729 hits
Follow NotPoliticallyCorrect on WordPress.com

suggestions, praises, criticisms

If you have any suggestions for future posts, criticisms or praises for me, email me at RaceRealist88@gmail.com
%d bloggers like this: