Global IQ: 1950–2050

Global IQ

Chart:    Non-JavaScript Chart Browser

Doesn't it seem like the world is getting dumber with every passing year? Well, maybe it is!

In IQ and the Wealth of Nations, Lynn and Vanhanen[1] report large differences, amounting to more than two standard deviations, in the mean IQ of the populations of different countries around the world, and find that these mean population IQ scores correlate more strongly with economic development as measured by gross domestic product (GDP) per capita and long term economic growth than any other single factor.

It has been widely observed that the birthrate of countries tends to fall as they become more wealthy. Most countries in Western Europe now have birthrates below the replacement rate; in the absence of immigration, their populations can be expected to fall in the future.

Putting these two pieces of information together, one might expect that since low IQ countries tend to be less wealthy, they should also be expected to have higher birthrates than countries with high IQ. If population IQ and wealth remain constant, the average IQ of the world should then fall over time, since a larger portion of population growth will occur in low IQ countries.

  Year     Population×109     Mean IQ  
1950 2.55 91.64
1975 4.08 90.80
2000 6.07 89.20
2025 7.82 87.81
2050 9.06 86.32

There are a lot of assumptions going into this conclusion, starting out with what IQ measures and what, if anything, it means. See the “Quarrels, Questions, and Answers” section below for discussion of some of these issues.

The animation above (you can view the chart for a given year by selecting it from the Chart box below the image) shows the global histogram of IQ and global mean IQ for the hundred year period from 1950 through 2050. Mean population IQ is taken from Lynn and Vanhanen's[1] figures and yearly population estimates for each country from the U.S. Census Bureau International Data Base 2003[3]. Taking these figures at face value, we find that world population and mean IQ evolve at 25 year intervals over the century as given in the table at the right.


The mean IQ of 185 countries, measured and estimated in Lynn and Vanhanen[1], were taken as the invariant IQ of each country over the 1950–2050 time period. (The figures are given in terms of countries existing as of the year 2000. For countries which came into being in the preceding 50 years due to decolonisation, breakup of the Soviet Union, etc., years prior to independence refer to the territory with borders identical to the present-day country. The list of countries includes Hong Kong and Taiwan, considered in some sense provinces of China, but with large populations, well measured demographically, and economic performance distinctly different from that of the People's Republic; and Puerto Rico, a United States territory with different demographics than the parent country. The remaining 182 countries include all independent countries with populations greater than 50,000 with the exception of Bosnia and Herzegovina, for which no data were available due to the conflict throughout most of the 1990s.)

The 100 year population history and forecast for the 185 countries with measured or estimated mean IQ was obtained from the U.S. Census Bureau International Data Base 2003[3], using the mid-year population estimate or projection for each year.

For each year in the hundred year period, each country's estimated population for that year was apportioned into bins of 5 IQ points using a normal distribution with the mean IQ for the country from Lynn and Vanhanen and the 15 point standard deviation defined for IQ scores. These country histogram bins were summed to create a global histogram for each year. Global mean IQ was computed by an average of country IQs weighted by their population.

Quarrels, Questions, and Answers

Don't differing IQ figures for various countries simply measure cultural bias in the tests?
This is a possibility, and in certain cases undoubtedly plays a factor. Yet tests carefully designed to exclude cultural bias (for example, spatial relationship tests based entirely on pictures, memorisation of digit sequences, and pure eye-hand reaction time) produce results comparable to those of traditional IQ tests. Further, if IQ tests embody cultural biases of the largely U.K. and U.S. creators of the tests, it's odd that populations of East Asian countries, with a variety of very different cultures, all test higher than those of the test makers.
Won't economic development reduce the rate of population growth in the low-IQ countries?
Future population estimates for countries in the Census Bureau database already take this into account. These are, of course, consensus estimates which do not take into effect such impossible-to-forecast circumstances as environmental crises, plague, bad asteroid days, or, on the other hand, technological breakthroughs which accelerate economic development in third world countries. Looking 50 years ahead, only rapid demographic shifts in the near term will have much impact on the figures for 2050, since the parents of adults of that year are already mostly alive today.
Lynn and Vanhanen only actually have IQ data for 81 countries and they've estimated the rest. How reliable are those estimates?
I don't know. In most cases their estimates were made by averaging known IQs of adjacent countries with similar demographic mix. In the few cases of countries with ethnically diverse populations, they estimated IQ based on a weighted average of IQs of the country of origin of each group. They tested this process by using it to estimate IQ of several countries with known IQ and the results correspond well with the measured IQs of those countries. Still, one should bear in mind that 56% of the country IQ figures are estimated, and not based on any actual in-country measurement at all.
And those 81 countries they have IQ data for—there seem to be an awful lot of fudge factors used in computing the numbers they cite in the tables. How trustworthy are they?
Fudge factors? Indeed…more than 25 pages are devoted to explaining the “adjustments”, “corrections”, “calibrations”, and “weightings” which go into that table of 81 numbers. The state of the raw data is more or less hideous. There is no regular, standardised measurement of IQ in nations of the world. One is forced to use sporadic studies, published at widely spaced intervals, using a variety of tests with more or less cultural bias, on populations which may exhibit a variety of selection effects. (For example, if you only test high school children in a country where 75% of children do not attend high school, you can't expect your results to be representative of the population as a whole.) Still, if you want to do this research, you have work with the data at hand.

If population mean IQ indeed correlates strongly with economic performance, then measuring IQ figures for developing countries and studying ways to increase IQ could play an important rôle in development assistance. A UNESCO program to regularly measure IQ of, say, 16 year olds in all countries could provide hard data and, potentially, by permitting assessment of the effectiveness of programs such as nutrition aid for mothers and infants, educational initiatives, etc., do a world of good. Alas, this entire topic is so politically radioactive there is little likelihood of this ever happening.

You're assuming the mean IQ of countries won't change over the hundred year period. How valid is that assumption?
Apart from the Flynn effect (discussed below), which doesn't seem to have much effect on the relative IQs of countries, in cases where the data are available, national mean IQ does not seem to have varied much over the last 50 years. As long as the population makeup and general circumstances of a country don't change, it's reasonable to expect the mean IQ for a given country to remain much the same over the next 50 years. Population migration, however, can have substantial effects and is not taken into account in these data. The Census Bureau population estimates include migration, but the assumption of constant mean IQ may be invalid when the population of a given country consists of a large fraction of immigrants from regions with different mean IQ. This is particularly the case for Western Europe, where the indigenous population has fertility below the replacement rate, and the population includes an increasing proportion of immigrants predominantly from regions with lower mean IQ. To the extent immigrants have more children per family than the original population, the effect is magnified. Whether immigrant populations converge toward the original IQ of their new country as they assimilate is an open question. In all, since most present day and anticipated future population migration is from lower IQ to higher IQ countries, assuming constant IQ probably biases the global mean forecasts toward the high end.
Won't the Flynn effect compensate for the downward demographic shift in IQ?
The Flynn effect is an undisputed yet enigmatic aspect of IQ testing. Shortly after the first IQ tests were standardised, it was observed that the scores of those taking them tended to rise from year to year, as much as 15 points (one standard deviation) per generation. To maintain a mean score of 100 for the population on which IQ tests were standardised, test makers were forced to make their tests increasingly difficult over the years. In other words, to get the same IQ score as your father, you must perform equally well on a substantially tougher test than he took.

If, for whatever reason, everybody were getting smarter, this would be wonderful news indeed. But a glance at the numbers shows that something very curious must be going on here. If IQ were, in fact, rising at a rate of 15 points per generation then, if the mean IQ of today is 100, that of our grandparents' generation would have been about 70—generally considered the threshold of mental retardation. Clearly, anybody who's spent time with their grandparents and other folks of that generation knows that's utter nonsense.

The literature and music of a century or more ago is clearly not the work of marginally retarded minds, and its abundance indicates those who wrote it were not rare exceptions in a generally dull population. Consider genius in the past. Most people considered geniuses have IQs in the vicinity of 150, or 3 1/3 standard deviations above the mean IQ of 100. In a population with a mean IQ of 100, individuals with IQs of 150 occur with a frequency of about one in 2300 people—they're rare, but every medium-sized town has one or more, and even a small country with a population of one million has more than 425 such geniuses.

Now, in a population with a mean IQ of 70, which naïve interpretation of the Flynn effect would deem our grandparents to have had, genius-level IQs of 150 would be 5 1/3 standard deviations above the mean and occur, on average, in only one out of 20,396,324 people. If we take the Flynn effect as 3 IQ points per decade, then we'd expect a mean IQ of 70 around the year 1900. In 1900, the world population was about 1.7 thousand million, which would imply there were only 80 people with genius-level IQs in the entire world of 1900. The merest glance at the history of that era will reveal how ridiculous a supposition this is.

Adults, whatever their opinion may be of “what's the matter with kids today”, are most unlikely to cite “they're just too doggone smart!” So, the Flynn effect is a conundrum: a wide variety of tests which agree with one another and reliably predict outcomes we identify with “intelligence” all indicate that the general population is becoming more intelligent at an almost dizzying rate, while other evidence for this (for example, individuals with Einstein-calibre intelligence being almost 10,000 times more common than a hundred years ago) is notably absent. There is no shortage of hypotheses for what's going on, but little evidence to support any of them. Flynn himself believes that IQ tests measure test-taking and problem-solving ability, not genuine intelligence, and that this has risen over time as more and more children receive compulsory education and are subjected to ever more tests. Improved nutrition over the 20th century is often cited as a factor, as well as the introduction of egalitarian welfare state systems in developed countries tending to reduce poverty. But all of these are factors which one would expect to eventually reach a plateau, and that doesn't seem to have happened, at least so far.

This isn't a document about the Flynn effect (although it risks becoming one unless I wind this up rather soon), and since no solution to this long-standing puzzle is at hand, one can only speculate on what it really means. Since correction for the Flynn effect is substantial in Lynn and Vanhanen's national IQ estimates, and can be expected to strongly influence IQ scores published in the future, it is essential one bear it in mind in any analysis of population intelligence trends.

IQ scores are normalised for a mean of 100 and standard deviation of 15 in the populations for which they were originally developed. Is the standard deviation the same in populations with higher or lower mean IQ?
I don't know. This is a fascinating question about which I have found no research whatsoever. Absent any information to the contrary, in computing the global IQ histogram in the charts at the top of this document, I assume a standard deviation of 15 points regardless of the mean. Note that this assumption only affects the shape of the histogram; the global mean is independent of the variance of individual country populations.


Data Sources for this Page

  1. Lynn, Richard and Tatu Vanhanen. IQ and the Wealth of Nations. Westport, CT: Praeger, 2002. ISBN 0-275-97510-X.
  2. Lynn, Richard and Tatu Vanhanen. “Intelligence and the Wealth and Poverty of Nations”. 2000–2002.
  3. U.S. Census Bureau. International Data Base 2003.
  4. U.S. Central Intelligence Agency. The World Factbook 2003.

IQ Phenomenology: The Flynn Effect

  1. Neisser, Ulric, ed. The Rising Curve: Long-Term Gains in IQ and Related Measures. Washington: American Psychological Association, 1998. ISBN 1-55798-503-0.
  2. Flynn, J. R. “The Mean IQ of Americans: Massive gains 1932 to 1978”, Psychological Bull. 95, 29 (1984).
  3. Flynn, J. R. “Massive IQ gains in 14 nations: What IQ tests really measure”, Psychological Bull. 101, 171 (1987).
  4. Flynn, J. R. “IQ gains over time”, in Sternberg, Robert J.. ed. Encyclopedia of Human Intelligence. New York: Macmillan, 1994. ISBN 0-028-97407-7. (pp. 617–623)

“IQ Exists and Matters” Arguments

  1. Herrnstein, Richard J. and Charles Murray. The Bell Curve. New York: The Free Press, [1994] 1996. ISBN 0-684-82429-9.
  2. Jensen, Arthur R. The g Factor: The Science of Mental Ability. Westport, CT: Praeger, 1998. ISBN 0-275-96103-6.
  3. Rushton, J. Philippe. “Race, Intelligence, and the Brain: The Errors and Omissions of the `Revised' Edition of S. J. Gould's The Mismeasure of Man (1996)”, Person. individ. Diff. 23, 169 (1997).

“IQ Doesn't Exist and/or Matter” Arguments

  1. Gould, Stephen Jay. The Mismeasure of Man. New York: W. W. Norton, 1996. ISBN 0-393-31425-1.
  2. Devlin, Bernie et al., eds. Intelligence, Genes, and Success: Scientists Respond to The Bell Curve. New York: Copernicus, 1997. ISBN 0-387-94986-0.
  3. Fraser, Steven. The Bell Curve Wars. New York: Basic Books, 1995. ISBN 0-465-00693-0.

Economic Development and Population Trends

  1. Todd, Emmanuel. Après l'Empire. Paris: Gallimard, 2002. ISBN 2-07-076710-8. English translation: After the Empire. New York: Columbia University Press, 2004. ISBN 0-231-13102-X.

Download Database and Analysis Programs

The data and software used to produce the individual frames and animation at the top of the page may be downloaded from the following links. The population and IQ data are extracted from the Global Composite Country Database in CSV (Comma-Separated Value) format, which contains a variety of other data assembled from resources on the Web. CSV files can be loaded into most spreadsheet and database programs. Lines which begin with a sharp sign (“#”) are comments—if your software doesn't ignore them, you should manually delete them before or after loading the database. The first non-comment record gives the field names in the data which follow. The contents of the fields and the sources from which they were obtained is given in the following comments. Fields for which no data are available or are inapplicable are left blank.

The analysis and image generation was performed by a purpose-built Perl program which you can download from the second link below. This program contains little or no documentation and is utterly unsupported—you are entirely on your own. Histograms are plotted with the Gnuplot package and post-processed with utilities from the Netpbm image processing toolkit. The animation was assembled with the Gifsicle utility. These are all free software which may be downloaded from the respective links.

by John Walker
April, 2004

Introduction to Probability and Statistics

Fourmilab Home Page