PISA Test Score Gap Closing

More evidence against Richard Lynn's Italian "IQ" study. The achievement test scores he used to "measure" intelligence have since improved dramatically in Southern Italy, which is rapidly catching up to the North.

Recent results of international assessment programs (e.g., PISA) have shown a large difference in high school students' performance between northern and southern Italy. On this basis, it has been argued that the discrepancy reflects differences in average intelligence of the inhabitants of regions and is associated with genetic factors (Lynn, 2010a, 2012). This paper provides evidence in contrast to this conclusion by arguing that the use of PISA data to make inferences about regional differences in intelligence is questionable, and in any case, both PISA and other recent surveys on achievement of North and South Italy students offer some results that do not support Lynn's conclusions. In particular, a 2006-2009 PISA data comparison shows a relevant decrease in the North-South difference in only three years, particularly evident in the case of a single region (Apulia). Other large surveys (including INVALSI-2011) offer different results; age differences suggest that schooling could have an important role.

Cornoldi et al. "Problems in deriving Italian regional differences in intelligence from 2009 PISA data". Intelligence, 2013.

Another Richard Lynn Refutation

They just keep coming. The others can be found here.

As stated at the outset, the aim of this analysis is not to demonstrate that IQ in Southern Italy is the same or higher than that in the North. On the other hand, a simple statistical exercise, based on correlations among variables, such as that by Lynn, is far from conclusive. Recent studies have shown a significant statistical relationship (p-value = 0.008) between the presence of storks in the European continent and the birth rate (Matthews, 2000); an association which seems particularly remarkable in the case of Germany (Höfer et al. 2004). We know how hard it is to explain causality by means of statistical exercises and certainly causal relationships are not captured by simple statistical correlations.

Our previous discussion suggests a different relationship among the variables analyzed by R. Lynn and those collected in the present article. Allowing that school tests are representative of differences in IQ, we have seen that:

• they clearly show how the North-South difference in cognitive ability does not exist at the age of 7. The correlation between regional educational achievement (INVALSI) and regional per capita income becomes positive (0.44) at the age of 10, but is, however, considerably lower than that found through PISA test scores;
• the IQ-average income relationship did not exist in the past, although if it had derived from a genetic difference, it would consequently have done so;
• knowledge of genetic differences in Italy does not support Lynn's opinion that peoples from North Africa and the Near East strongly influenced the genetic structure of the Southern Italian population. Genetically, the influence of the Phoenicians and the Near East populations accounts for a very small fraction, while the predominant genetic influence derives from the long phase of Greek colonization.

The existence of differences in IQ, as revealed by school tests and other tests (supposing that these actually reveal "fundamental" diversities in intelligence), seems much better explained as being socially, economically and historically influenced rather than being genetically determined. Capabilities in problem solving are enhanced by a developing and stimulating environment, according to the so-called "Flynn effect".

In the past, the Southern Italian economy has at times been more advanced than the Northern one; for example during both the Roman antiquity and the high Middle Ages. Perhaps the North and the Centre were more advanced than the South in the late Middle Ages; although nothing certain can be said on the matter. The following decline of the Italian economy as a whole, from the late Middle Ages until the end of the 19th century, probably cancelled the existing economic differences. When per capita GDP diminishes and approaches the level of bare subsistence, differences among regions disappear. In the 19th century, Italy was a relatively backward country both in the North and the South. The statistical material available from the end of the 19th century onwards, does not actually indicate a deep North-South divide, in economic terms. The start of modern growth from then on affected the North much more than the South and economic disparity began to exist between the two parts of the country. In 1891 the North-South difference in per capita GDP was less than 10 percent; it was 20 percent on the eve of World War 1, and 45 per cent after World War 2. In 2010, per capita GDP in the South is about 60 percent that of the North. As ordinarily happens, relative backwardness implies the accumulation of adverse influences. While literacy and years of education grew in the North with respect to the South, infant mortality diminished much more quickly in the North than the South. Institutions, including families and schools, work much better in prosperity than backwardness. Estimates of IQ are likely to be higher where families, municipalities, provinces and regions invest more in education. Remarkable emigration from the South to the North, especially between 1950 and 1975, increased the North-South diversity since emigration, in Italy as elsewhere, is always selective. Since IQ registers education and years of schooling to a greater extent than intelligence, the relative position of the South compared to the North deteriorated in both the cultural environment and the economy more or less contemporaneously.

Daniele and Malanima. "Are people in the South less intelligent than in the North? IQ and the North-South disparity in Italy". Journal of Socio-Economics, 2011.

Illiteracy Rates by Region

Historically, literacy has been much higher in Northern Italy than in the more rural South, but that gap has been narrowing steadily since the 19th century and is now almost completely closed, with the younger generations everywhere being >98% literate like the rest of the developed world.*

In the table below, column one shows regional illiteracy rates for people under the age of 65, which are nearly the same throughout Italy. Column two shows people 65 and over, where the gap still persists. And column three shows the total population. The data was compiled by Istat for the 2001 census.


Vincenzo D'Aprile. "Analfabetismo: Italia — Censimento 2001: Cinque Grandi Ripartizioni Geografiche e Venti Regioni". Educazione&Scuola, 2005.

---------------
*According to Felice and Giugliano (2011): "The reason literacy was so low in the Southern regions was that, until 1861, they formed a different state, the Kingdom of the Two Sicilies, which did not promote compulsory education, unlike the pre-unitarian states of Northern Italy. Once compulsory education was extended to the South (from 1861) and after adequate financial resources were spent for its provision by the Italian State (from 1911), the Southern regions converged in literacy."

Legacy and Value of Latin

Although Latin — an Indo-European language at its height during the Roman Empire — is nobody's native tongue these days, it certainly remains a topic of conversation. The usual point of debate? Whether learning Latin is valuable for modern-day students.

"I don't think people know what they mean when they dismiss Latin as a dead language," said Paul Harvey, associate professor of classics at Penn State. "Of course it is not spoken in many places, save for the Vatican and a few classics departments. But whether a language is currently spoken is irrelevant to the continuing value of learning it and to the value of literature written in that language."

Even if you never read Virgil or Cicero in the original, explains Harvey, "Latin is the root language from which variations developed into today's modern Spanish, Portuguese, French, Italian and a few less-spoken European tongues. On a practical level, it is far easier for those with a firm foundation in Latin to learn a modern Romance language."

Whether you've studied Latin or not, most people already use it constantly, Harvey adds. The ancient Latin/Roman alphabet is the most widely used writing system in the world, and is the alphabet of English and almost all other European languages. Up to 60 percent of modern English words derive from Latin, directly or indirectly.

Latin is our foundational language, explains Harvey, and that awareness may be a factor in the current revival of interest in Latin in our schools. One indicator is the number of students taking the AP Latin exam has doubled in the last decade. "From what I've seen and read," Harvey said, "the revival in schools — including inner city schools — has less to do with a renewed interest in the classical past than a realization that the more Latin students learn, the higher their SAT verbal and analytical scores." In fact, students of Latin had notably higher mean SAT Verbal scores than students of Spanish, French or German.

"Furthermore," Harvey notes, "there is a continuing appreciation that studying Greek and Latin — the classical languages of Western civilization — demonstrably enhances the ability to write cogently in English." The reason for this is "rather straightforward," believes Harvey: "The successful study of a highly inflected language forces students to understand better the grammar and syntax of their own native language and that, in turn, encourages clarity of expression and analytical thought."

In today's era of online chat acronyms and text-messaging abbreviations, the ability to write well in English may be an increasingly rare and valuable career asset. Future attorneys, doctors and scientists would be well advised to study Latin to get a jump on the professional jargon, says Harvey. "And students wishing to study practically any aspect of ancient, medieval, and Renaissance literature, art, music, and history are at a serious disadvantage if they do not know Latin well."

Melissa Beattie-Moss. "Probing Question: Is Latin valuable for today's students?". Research/Penn State, April 27, 2011.

Richard Lynn Further Refuted

I had recently criticized Richard Lynn's flawed study on IQ in Northern and Southern Italy, and now Italian researchers Cornoldi et al. (2010) have published their own critique, which confirms what I said and predicted:

Working with data from the PISA study (OECD, 2007), Lynn (2010) has argued that individuals from South Italy average an IQ approximately 10 points lower than individuals from North Italy, and has gone on to put forward a series of conclusions on the relationship between average IQ, latitude, average stature, income, etc. The present paper criticizes these conclusions and the robustness of the data from which Lynn (2010) derived the IQ scores. In particular, on the basis of recent Italian studies and our databank, we observe that : 1) school measures should be used for deriving IQ indices only in cases where contextual variables are not crucial: there is evidence that partialling out the role of contextual variables may lead to reduction or even elimination of PISA differences; in particular, schooling effects are shown through different sets of data obtained for younger grades; 2) in the case of South Italy, the PISA data may have exaggerated the differences, since data obtained with tasks similar to the PISA tasks (MT-advanced) show smaller differences; 3) national official data, obtained by INVALSI (2009a) on large numbers of primary school children, support these conclusions, suggesting that schooling may have a critical role; 4) purer measures of IQ obtained during the standardisation of Raven's Progressive Coloured Matrices also show no significant differences in IQ between children from South and North Italy.

Lynn's feeble reply, where he's basically forced to admit he's wrong:

Beraldo (this issue) and Cornoldi, Belacchi, Giofre, Martini, and Tressoldi (2010) (CBGMT) have eight criticisms of my paper (Lynn, 2010) claiming that the large north-south differences in per capita income in Italy are attributable to differences in the average levels of intelligence in the populations. CBGMT give results for seven data sets for IQs in the north and south of Italy. All of these show that IQs are higher in the north than in the south, although the differences are not as great as those I calculated. Other criticisms to the effect that the PISA tests are not measures of intelligence are refuted. The results of two further studies are given that confirm that IQs in the north of Italy are approximately 10 IQ points higher than in the south.

The other critique mentioned there is Sergio Beraldo (2010), which focuses mostly on Lynn's claims about economic differences.

Refuting Richard Lynn's Italian "IQ" Study

Controversial psychologist Richard Lynn, who looks at IQ and its correlates, has published a study claiming to show regional (North-South) differences in intelligence within Italy, which he attempts to correlate with achievement and attribute to admixture. The guy's been called just about every name in the book, and he can now add Padanian Nordicist to that list.

Intelligence

Generally speaking, Lynn is not to be trusted. He's been caught numerous times falsifying and manipulating data to fit his conclusions (e.g. here, here, here, here, here and here), and it looks like he's up to his old tricks again.

This time around, he's not even using actual IQ data, but the proxy of scores on reading, math and science tests administered to 15-year-olds (PISA 2006). So he's attempting to quantify innate general intelligence by looking at the academic performance of school kids, a measure that to a large extent involves learned knowledge and other factors. Indeed, while some researchers report a strong correlation between general intelligence and educational attainment, one of Lynn's own sources, Deary et al. (2007), addressing two of his other sources, suggests that caution should be exercised when attempting to equate the two:

There are various possible causes of the cognitive ability-educational achievement association. Bartels et al. (2002b) found a strong genetic correlation between cognitive ability (measured at 5, 7, 10, and 12 years) and educational achievement at age 12. In an overview, Petrill and Wilkerson (2000) concluded that genetics and shared and non-shared environmental factors all influence intelligence and education, with genetics being important in the correlation between them, and non-shared environment being important in discrepancies between intelligence and educational attainments.

Whereas the correlations indicate that around 50% to 60% of the variance in GCSE [General Certificate of Secondary Education] examination points score can be statistically explained by the prior g [general intelligence] factor, by the same token a large proportion of the variance is not accounted for by g. Some of the remaining variance in GCSE scores will be measurement error, but some will be systematic. Thus, non-g factors have a substantial impact on educational attainment. These may include: school attendance and engagement; pupils' personality traits, motivation and effort; the extent of parental support; and the provision of appropriate learning experiences, teaching quality, school ethos, and structure among other possible factors (Petrides, Chamorro-Premuzic, Frederickson, & Furnham, 2005; Strand, 2003).

But Lynn already knows the pitfalls of his approach. Finland had the highest score in Europe on the 2006 PISA tests, and using his method leads to a calculated IQ of 107, yet he reports Finns' IQ as being just 97. Romanians' PISA score is near the very bottom of Europe, leading to an estimate of 85, though their measured IQ is in fact 94 according to Lynn, just three points lower than that of Finns. With discrepancies like that, there's absolutely no reason to trust his calculated IQs of around 100 and 90 for Northern and Southern Italians. Clearly, PISA scores are not a good substitute for IQ.

Then, to try to prove that disparities in intelligence are long-standing, and therefore genetically based, he uses literacy as another (questionable) proxy for IQ. But whereas for other correlates like stature and infant mortality he includes data from the past and present to show that the North-South gap has remained fairly stable, for literacy he only includes data from 1880, when it was extremely large (55% vs. 20%, on average). Obviously, he wants to hide the fact that the gap has been closing steadily since then, and by the 21st century, literacy among Italians under the age of sixty-five was 99.7% in the North and 99% in the South (Istat 2001).

Achievement

Lynn goes on to attempt to correlate his fake IQs with achievement. The primary measure he uses is per capita income, which is double in the North what it is in the South. His source is the Italian Statistical Office, but he should be aware that figures for the South's economic performance are greatly underestimated because the official statistics fail to take into account a large underground economy there, according to Burnett and Vaccara (1999):

But the third factor, somewhat alleviating the second, is the existence of a far vaster private sector than ever shows up in the economic statistics. The size of the lavoro nero sector and the black market in the South clearly exceeds that of any other EU region.... In Calabria, with its dire employment figures, 84 percent of the families own their own home. What such anomalies must mean is that real income in Calabria is far higher than what is "on the books." Many among the vast numbers of officially unemployed are, in fact, partly or fully employed. They are earning no social benefits, but they are earning the daily lire that keep their families afloat. [...] A very large part of the South's hidden labor is made up of entrepreneurs, sometimes also employing black labor, and existing themselves outside official recognition, taxation, protection, control, or counting. A recent analysis concludes that "there exists in several zones of the Mezzogiorno [Southern Italy] a whole fabric of small and very small businesses that escape every census, but that work and make profits, share among themselves a serious level of production, export to other regions [of Italy] and abroad." [...] This massive sector skews all the statistics. It means that the GDP for the Italian South (and for Italy as a whole) is far from accurate. And the unemployment figures do not reflect reality.

Then, with the same goal of establishing a pattern that extends far back in time, he also looks at historical achievement, citing Charles Murray's Human Accomplishment. But here again he plays fast and loose with the data. He adds up Murray's "significant figures" for Italy from 1400-1950 and divides them into "North", "Center" and "South" based on their origins. Almost all of them (187 out of 236) come from the "North". However, he explains that he uses the 42nd and 41st lines of latitude as borders, which are, respectively, just north of Rome and just north of Naples. So more than half of the country is put into the "North" category, while the "Center" gets flattened down and pushed into the territory of the "South" (click here to see what that looks like). This is a shameless and transparent ploy by Lynn to hugely inflate the number of significant figures from the "North" and reduce the number elsewhere in the country.

But perhaps even worse, he ignores the main finding of Murray's book, which is that achievement has been concentrated in just a few places, and Southern Italy is but one of many "low-achievement" areas throughout Europe, along with some of the northernmost regions in Italy:


As you can see — and as Murray points out in his book — the highest ranking region is Tuscany, which would normally be considered part of Central Italy. Ironically, Lynn doesn't even have PISA/IQ data for Tuscany, though its 1880 literacy rate isn't particularly high. So he's basically attributing the unique genius of that region to the brainpower of regions that have not produced the same level of genius.

Admixture

Finally, Lynn offers his "explanation" for the disparities in fake IQ, which, unsurprisingly, turns out to be admixture from the Middle East and North Africa, where (according to him) IQs are in the range of 80-84. As you might expect from someone with an agenda and little knowledge of genetics, he references a lot of old studies that use single or small numbers of loci and don't directly address the question of admixture. One of them has "Neolithic demic diffusion in Europe" in its title, yet he stupidly follows the citation with references to historical groups like Phoenicians and Arabs.

Recent genome-wide studies have been able to detect and quantify admixture like never before. Li et al. (2008), using more than 600,000 autosomal SNPs, identify seven global population clusters, including European, Middle Eastern and Central/South Asian. Contrary to Lynn's claims, it's actually the overachieving Tuscans who have a small amount of non-European admixture and not the underachieving Sardinians:


López Herráez et al. (2009) typed the same samples at close to 1 million SNPs and analyzed them in a Western Eurasian context, identifying a number of subclusters. This time, all of the European samples show some minor admixture. Among the Italians, Tuscany still has the most, and Sardinia has a bit too, but so does Lombardy (Bergamo), which is even farther north:

Conclusion

The bottom line is, we don't know the average IQs for different regions in Italy, which is why Richard Lynn had to resort to making them up. And while Southern Italians are likely to be a few points lower than Northern Italians — as the Irish and Scottish are a few points lower than the English — there's absolutely no reason to believe that North and South would be separated at their extremes by almost a full standard deviation. Lynn certainly hasn't proven anything of the kind with this ridiculous study, nor has he provided any valid explanations for such a disparity.

Updates

Since I wrote this article, a number of studies have been published that refute Lynn and confirm what I've said:

• Cornoldi et al. (2010) and Beraldo (2010)
• Felice and Giugliano (2011)
• Daniele and Malanima (2011)
• Robinson et al. (2011)
• D'Amico et al. (2012)
• Templer (2012)
• Cornoldi et al. (2013)

---------------
Richard Lynn. "In Italy, north-south differences in IQ predict differences in income, education, infant mortality, stature, and literacy". Intelligence, 2010.