Content uploaded by Emil O. W. Kirkegaard
Author content
All content in this area was uploaded by Emil O. W. Kirkegaard on Aug 24, 2016
Content may be subject to copyright.
MANKIND QUARTERLY 2015 56:2 113-135
113
IQ and Socioeconomic Variables in French Departements:
Reanalysis and New Data
Emil O. W. Kirkegaard
University of Aarhus, Denmark
Email for correspondence: emil@emilkirkegaard.dk
Two sets of socioeconomic data for 90-96 French departements
were analyzed. One dataset was found in Lynn (1980) and contained
four socioeconomic variables. Mixed results were found for this
dataset, both with regards to the factor structure and the relationship
to cognitive ability. Another dataset with 53 variables was created by
compiling variables from the official French statistics bureau (Insee).
This dataset contained an impure general socioeconomic (S) factor
(some undesirable variables loaded positively), but after controlling for
the presence of immigrants, the S factor became purer. This was
especially salient for crime, unemployment and poverty variables. The
two S factors correlated at r = 0.66 [CI95: 0.52 – 0.76; N = 88]. The IQ
scores from the 1950s dataset correlated at 0.33 [CI95: 0.13 – 0.51,
N = 88] with the S factor from the 2010-2015 dataset.
Key words: S factor, general socioeconomic factor, France, French
departements, IQ, intelligence, cognitive sociology
When desirable outcomes generally correlate positively with other desirable
outcomes, and undesirable outcomes with other undesirable outcomes, then one
can extract a general factor (general socioeconomic factor; S factor) such that the
desirable outcomes have positive loadings and the undesirable outcomes have
negative loadings.
1
1
Sometimes it may need to be reversed by multiplying all values by -1. This happens if
the dataset contains more indicators of undesirable than desirable outcomes.
MANKIND QUARTERLY 2015 56:2
114
The S factor has been found in numerous studies in the past two years at
many different levels of analysis: between countries (Kirkegaard, 2014b),
between regions within a country (Kirkegaard, 2015b, c, d, f, g, h, i, j, k), between
districts within a city (Kirkegaard, 2015a), between people grouped by country of
origin (Kirkegaard, 2014a; Kirkegaard & Fuerst, 2014) and between first names
(Kirkegaard & Tranberg, 2015b). Substantial correlations with cognitive ability and
demographic variables have often been reported as well.
France is a fairly large country in Europe with a population of about 67
millions. It has a number of administrative divisions: 27 regions, 96 departements,
335 arrondissements, 2,054 cantons and 36,658 communes as of 2015, which
are respectively first, second, third, fourth and fifth-level divisions (Insee, 2015).
Lynn (1980) published a study of four socioeconomic correlates of IQ across
90 French departements (Metropolitan France, i.e. the part located in Europe,
only). He reported several correlations in the expected direction for IQ, but also
one in the “wrong” direction: unemployment had a correlation of .20 with IQ. Since
Lynn reported all the raw data, the data can be reused here 35 years later.
2
Because a much greater amount of data is now available, I gathered a large
new dataset of socioeconomic variables for French departements. Thus, study 1
is a reanalysis of Lynn's data and study 2 is based on new data. A map of the
current departements of France is in Figure 1 for reference.
Study 1 – Reanalysis of Lynn (1980)
Data Sources
The data were contained in Lynn (1980). He reports their origin as follows:
1. Mean IQ. This was copied from Montmollin (1958) and was based on
testing of conscripts in the mid-1950s.
2. Intellectual Achievement. “The criterion of intellectual achievement used
was membership of the Institut de France for the year 1975. The Institut de
France consists of five academies, namely the Académie française,
Académie des Inscriptions et Belles-Lettres, Académie des Sciences,
2
This is unusual for social science. Generally, authors are reluctant to share their data.
Worse, there is some evidence that those who refuse to share data make more errors
than those who share (Krawczyk & Reuben, 2012; Savage & Vickers, 2009; Wicherts,
Bakker & Molenaar, 2011).
KIRKEGAARD, E.O.W. IQ AND SOCIOECONOMIC VARIABLES IN FRANCE
115
Figure 1. A map of French departements in 2015. (“Departments of France”,
2015)
Académie des Beaux-Arts and Académie des Sciences Morales et
Politiques. The Institut de France had 253 French members in 1975. For 250
of these place of birth was ascertained. The subjects were allocated to the
departement in which they were born and a departement achievement
quotient calculated by expressing the numbers of Institut members per
million of the departement’s population in 1974.”
3. Mean Earnings, 1971. “These were from Statistiques et Indicateurs des
Regions Francais, Institut National Statistiques et Etudes Economiques,
Paris, 1975. These are normalized to 100.”
MANKIND QUARTERLY 2015 56:2
116
4. Unemployment. “Numbers registered unemployed at the end of January
1972 calculated per 1000 population. Source: Ministere du Travail, de
l’Emploi et de la Population.”
5. Infant Mortality. “Infant deaths under 1 year per thousand live births,
average for 1970-72. Source: Statistiques et Indicateurs des Regions
Français, Institut National Statistiques et Etudes Economiques, Paris, 1975.”
6. Migration Flows, 1801-1954. “These were estimated by calculation of the
average annual increase in population from 1801 to 1954. It is assumed that
rates of natural increase of population are approximately constant across the
departements, and hence that population increases largely reflect net
migration flows. Population data were supplied by the Institut National
Statistique et Etudes Economiques in Paris.”
Lynn notes that the IQ data were from the mid-1950s, whereas the
socioeconomic data were from the early 1970s. A delay of about 20 years may
have changed things due to internal migration.
Factor Analysis
Factor loadings were extracted using least squares and scored with Bartlett's
method (Revelle, 2015). Figure 2 shows the loadings.
Figure 1. S factor loadings in the socioeconomic dataset from the 1970s.
KIRKEGAARD, E.O.W. IQ AND SOCIOECONOMIC VARIABLES IN FRANCE
117
We see that the factor (“standard”) is impure in the sense that unemployment
loaded in the wrong direction. We also see that the factor was practically identical
to earnings (loading = .997). This near-identity of the factor with a single variable
can happen when there is only a small number of variables. This was also found
in the study of first names in Denmark (Kirkegaard & Tranberg, 2015b).
The scores and loadings were stable across different factor extraction and
scoring methods (all correlations >.997).
Mixedness
When extracting factors from a dataset, one might find cases that poorly fit the
factor structure of the data. Such cases are said to be highly mixed, but could
also be called structural outliers (Kirkegaard, 2015e). Often when analyzing
socioeconomic datasets for within country regions, the capital region is found to
be a strongly mixed case. For instance, it may have a high mean income and a
high level of educational attainment, but also have a high crime rate and high
unemployment rate e.g. as with London in an analysis of regions of the UK
(Kirkegaard, 2015g). The two methods for examining mixedness developed by
Kirkegaard (2015e) were used on the dataset. Figure 3 shows the scatterplot.
Figure 2. Mixedness in the socioeconomic dataset from the 1970s.
MANKIND QUARTERLY 2015 56:2
118
We see that Indre is an outlier according to one method, but not so much
according to the other. Notice that the change in factor size (CFS) value is
negative, which means that it is an outlier that inflates the correlations in the
dataset, thus giving rise to a larger general factor. I'm not quite sure why this is
so. The region does not have any large cities or otherwise seem special. Perhaps
it's a fluke due to only relying on 4 variables. Maybe someone more familiar with
this region can shed some light on this.
Relationship Between S and IQ
Lynn (1980) did not plot the data and so it remains possible that his results
were strongly affected by outliers. Figure 4 shows the scatterplot of IQ and S.
Clearly, the two Parisian departements of Seine-et-Oise and Seine were inflating
the correlation. Without the two outliers, the correlation was r = .45 [CI95: .27 –
.60].
Figure 3. Scatterplot of the IQ scores from the 1950s and socioeconomic (S)
factor extracted from the 1970s data.
KIRKEGAARD, E.O.W. IQ AND SOCIOECONOMIC VARIABLES IN FRANCE
119
To see whether the two cases were having a large effect on the factor
analysis, two further factor analyses were run: one without the two outlier cases
and one using rank-ordered data. Figure 2 shows the loadings. The effect of
excluding the outliers was that the S factor changed to become near-identical
(loading >.999) with infant mortality instead. It was still correlated in the right
direction with IQ, but only at .23 [CI95: .02 – .42]. The analysis using ranks did
not result in a factor that was practically identical with one variable, but
unemployment still had a slight positive loading. The correlation of this factor with
IQ was .54 [CI95: .38 to .68], shown in Figure 5. Given the mixed results, limited
number of socioeconomic indicators, and the age of the data, there was a clear
need to gather new data and analyze inequality in France further.
Figure 5. Scatterplot of IQ scores (Cognitive_ability) and socioeconomic (S)
factor extracted from a ranked version of the 1970s data (S_rank).
MANKIND QUARTERLY 2015 56:2
120
Study 2 – New data
Data Sources
Data was compiled from the French-language website of Institut national de
la statistique et des études économiques (Insee), the official French statistics
bureau: http://www.insee.fr.
A diverse collection of 46 variables was selected. The selection procedure
was the same as for previous studies, namely, that variables should concern
important socioeconomic aspects of society and that they should not be overly
dependent on local climate or natural environment (e.g. presence of a large body
of water for fishing). The following variables were selected:
1. Property.crime - Property crime rate per 100k inhabitants.
2. Violent.crime - Violent crime rate per 100k inhabitants.
3. Economic.financial.crime - Economic crime rate per 100k inhabitants.
4. Cinema rooms per 100k - Cinema rooms per 100k inhabitants.
5. Armchairs per 100k - Cinema seats per 100k inhabitants.
6. Youth.reading.difficulties - Percent of children aged 16-17 who have difficulty
reading.
7. Voter.turnout.2012 - Voter turnout in the first round of the French legislative
election, 2012.
8. Edu.employee.per.100k - Educational employees per 1000 children aged 3
to 19 years.
9. Percent.higher.degree - Percent with a higher degree (Diplôme supérieur).
10. Start.up.rate - Creation rate of new companies.
11. Farmer.pct - Percent of the population aged ≥15 who is a farmer.
12. Upper.profession.pct - Percent of the population aged ≥15 who is in the
intellectual professions (profession intellectuelle supérieure).
13. Non.working.pct - Percent of the population aged ≥15 who is not working and
not retired.
14. Population.under.25 - Percent of population below age 25.
15. Births.outside.marriage - Percent births outside of marriage.
16. Total.fertility.per.100 - Total fertility per 100 women.
17. Life.expect.men - Life expectancy at birth, men.
18. Life.expect.women - Life expectancy at birth, women.
19. Children.unemployed.parents - Percent of children living with unemployed
parents.
20. Children.unemployed.single.parent - Like above, but for those living with a
single parent.
KIRKEGAARD, E.O.W. IQ AND SOCIOECONOMIC VARIABLES IN FRANCE
121
21. Income.men - Mean annual wage, in Euros, men.
22. Income.women - Mean annual wage, in Euros, women.
23. Benefits.use.AAH - Recipients of AAH benefits per 100 persons aged 20-64.
24. Benefits.use.PCH - Recipients of PCH benefits per 100 persons aged 20-64.
25. Housing.aid - Rate per 1000 households receiving housing aid.
26. Disabled.adults.allowance - Rate per 1000 persons receiving disability
benefits.
27. Disabled.children.allowance - Rate per 1000 persons receiving benefits for
having a disabled child.
28. Total.poverty - Total poverty rate, all ages.
29. Poverty.intensity - Poverty intensity.
30. Median.income - Median income, Euros, both genders
31. Med.institi.per.1000 - Medical institutions per 1000 inhabitants.
32. Med.workforce.per.1000 - Medical work force per 1000 inhabitants.
33. Ambulances.per.10k - Ambulances per 10k inhabitants.
34. Radiology.specialists.per.10k - Radiology specialists per 10k inhabitants.
35. GPs.per.100k - General practitioners per 100k inhabitants.
36. Special.doc.per.100k - Specialist doctors per 100k inhabitants.
37. Cancer.death - Cancer death rate per 100k inhabitants.
38. Circulatory.death - Circulatory death rate per 100k inhabitants.
39. Respitory.death - Respitory death rate per 100k inhabitants.
40. Digestive.death - Digestive death rate per 100k inhabitants.
41. Infection.death - Infectious disease and parasites death rate per 100k
inhabitants.
42. AIDS.death - AIDS death rate per 100k inhabitants.
43. Suicides.death - Suicide rate per 100k inhabitants.
44. Recycling.rate - Percent of households that recycle.
45. Employment.rate.25.54 - Employment rate, age 25-54. Furthermore, the
following variables were selected for demographic analysis:
46. Pop.growth - Mean % annual population change rate 2007-2014.
47. Population - Total population (Jan. 2014).
48. Foreigner.pct - Percent of total population of foreign origin.
49. EU.origin - Foreigners with EU origin, percent.
50. Algerian.origin - Foreigners with Algerian origin, percent.
51. Morrocan.origin - Foreigners with Moroccan origin, percent.
52. Tunesian.origin - Foreigners with Tunisian origin, percent.
53. Turkish.origin - Foreigners with Turkish origin, percent.
MANKIND QUARTERLY 2015 56:2
122
Note that variables 49-53 do not sum to 100, meaning that some immigrants
were from other countries than those listed above. Two new variables were
created from these: percent EU-foreigners, percent non-EU foreigners.
All the variables above concern the most recent data. This was always from
somewhere between 2010 and 2015. The supplementary material (data.ods) has
links to the exact Insee source tables for each variable.
Missing Data
There were 4 missing datapoints out of a total of 5088 possible (96 cases,
53 variables). These were imputed using the irmi() function without noise from the
VIM package (Templ et al., 2015).
Redundant Variables
When compiling large datasets like the one described above, one might
obtain variables that are duplicates of each other, reverse-coded duplicates or
near-duplicates. For instance, one variable may concern percent youth who have
not completed secondary education, while another may concern youth who have
completed secondary education (reverse-coded duplicate). Using such variables
in a factor analysis may skew the extracted general factor towards group factors
in the dataset (Jensen, 1998, p. 85). For this reason, all pairs of socioeconomic
variables that correlated at r≥|.90| were identified. For each such pair, one
variable was removed. The following pairs were found:
1. Disabled.adults.allowance and Benefits.use.AAH r = .972
2. Upper.profession.pct and Percent.higher.degree r = .968
3. Income.women and Income.men r = .946
4. Circulatory.death and Cancer.death r = .927
5. Employment.rate.25.54 and Children.unemployed.parents r = -.920
For each pair, the first variable was excluded from further analysis. This
reduced the dataset to 40 socioeconomic variables.
Mixedness
Mixedness in the present dataset was examined using the two methods
mentioned earlier. Figure 6 shows the scatterplot. We see that Paris and Seine-
Saint-Denis are far outliers. But unlike London in Kirkegaard (2015g), Paris does
not decrease but increase the factor size. A new dataset was created without the
two outlying cases.
KIRKEGAARD, E.O.W. IQ AND SOCIOECONOMIC VARIABLES IN FRANCE
123
Figure 6. Mixedness in the 2010-2015 socioeconomic dataset without redundant
variables.
Factor Analysis - Effects of Removing Redundant Variables and Outliers
To see if and how removing the two outliers affected the factor analysis, three
different datasets were analyzed: 1) with the structural outliers, 2) without the
structural outliers, 3) with the structural outliers using ranked data. The last was
used in an attempt to correct for non-linearity and other undetected lack of
normality in the data. As before, the analysis was done using least squares and
Bartlett's method. Figure 7 shows the loadings.
Generally, loadings were similar across the variations. The congruence
coefficients were between .97 and .99 (Lorenzo-Seva & Ten Berge, 2006).
MANKIND QUARTERLY 2015 56:2
124
Figure 7. S factor loadings in the new (2010-2015) dataset.
Factor Analysis - Effects of Controlling for Immigrants
Immigrants tend to move into areas inside the large cities. Since large cities
are usually higher S than the country areas, this can have the effect of creating
mixedness as well as socioeconomic inequality (Kirkegaard and Tranberg,
2015a; Meisenberg, 2007, 2008). One can attempt to statistically correct for this
by partialing out the effect of immigrants on the variables before subjecting them
to factor analysis.
Three different ways to control for the presence of immigrants were used: 1)
partialing out immigrant%, 2) partialing out EU immigrant%, and 3) partialing out
non-EU immigrant%. Figure 8 shows the loadings of the three analyses along
with the standard analysis from before for comparison.
It can be seen that controlling for immigrant% or non-EU immigrant% had
about the same effect, while controlling for EU immigrant% had a weaker effect
KIRKEGAARD, E.O.W. IQ AND SOCIOECONOMIC VARIABLES IN FRANCE
125
(purple dots are close to the red dots in the figure). This is in line with
expectations, as the immigrants from the EU are, as a group, expected to function
largely equivalently to the native population while the non-EU foreigners are not.
This pattern has also been found in Denmark (Indvandrere i Danmark 2014,
2014).
Figure 8. S factor loadings in the 2010-2015 dataset with and without controlling
for immigrants.
Table 1 shows the loadings for the original analysis and the analysis
controlled for non-EU immigrants as well as the change in the loading.
MANKIND QUARTERLY 2015 56:2
126
Table 1. Factor loadings for the S factor in the new dataset before and after
controlling for non-EU immigrants.
Variable
original
controlled
change
Violent.crime
0.58
-0.21
0.79
Non.working.pct
0.37
-0.31
0.67
Poverty.intensity
0.28
-0.36
0.64
Property.crime
0.72
0.17
0.55
Start.up.rate
0.56
0.02
0.54
Children.unemployed.parents
-0.31
-0.80
0.49
Economic.financial.crime
0.67
0.25
0.42
Total.poverty
-0.48
-0.90
0.42
Housing.aid
0.15
-0.24
0.39
AIDS.death
0.32
-0.07
0.38
Population.under.25
0.61
0.26
0.35
Special.doc.per.100k
0.74
0.43
0.30
Children.unemployed.single.parent
-0.54
-0.80
0.26
Edu.employee.per.100k
0.58
0.35
0.23
Radiology.specialists.per.10k
0.30
0.09
0.22
Total.fertility.per.100
-0.03
-0.22
0.19
Percent.higher.degree
0.91
0.73
0.18
Income.men
0.90
0.74
0.16
Med.institi.per.1000
0.24
0.13
0.11
GPs.per.100k
0.29
0.20
0.09
Youth.reading.difficulties
-0.60
-0.67
0.07
Life.expect.men
0.67
0.62
0.05
Med.workforce.per.1000
0.10
0.07
0.03
Disabled.children.allowance
-0.18
-0.16
-0.02
Median.income
0.91
0.94
-0.03
Life.expect.women
0.50
0.57
-0.07
Benefits.use.PCH
-0.36
-0.29
-0.08
Cinema.seats.per.100k
0.16
0.28
-0.12
Infection.death
-0.55
-0.42
-0.13
Benefits.use.AAH
-0.71
-0.57
-0.14
Ambulances.per.10k
-0.53
-0.38
-0.15
Cinema.rooms.per.100k
0.09
0.26
-0.17
Digestive.death
-0.70
-0.52
-0.18
Cancer.death
-0.81
-0.58
-0.23
Births.outside.marriage
-0.80
-0.56
-0.24
KIRKEGAARD, E.O.W. IQ AND SOCIOECONOMIC VARIABLES IN FRANCE
127
Variable
original
controlled
change
Respitory.death
-0.75
-0.51
-0.24
Suicides.death
-0.72
-0.40
-0.32
Recycling.rate
-0.28
0.08
-0.36
Farmer.pct
-0.63
-0.23
-0.41
Voter.turnout.2012
-0.43
0.03
-0.46
Some of the changes were quite drastic and some loadings that were in the
'wrong' direction according to the S factor model were reversed or strongly
reduced after the control:
Violent crime had a loading of .58 but changed to -.21
Non-working percent from .37 to -.31
Property crime from .72 to .17
Poverty intensity from .28 to -.36
Economic and financial crime from .67 to .25
Voter turnout from -.43 to .03
Other variables had changes that are more difficult to interpret: start-up rate
(.56 to .02) and farmer% (-.63 to -.23). In Denmark, it is known that non-Western
immigrants have a high start-up rate which is closely connected to being self-
employed (Indvandrere i Danmark 2014, 2014, p. 50). This is due to them owning
many small businesses, usually family enterprises. Probably something similar is
happening in France.
The farmer% is puzzling. In the Brazilian study (Kirkegaard, 2015i), farmers%
had a strong negative loading of about -.75. However, Brazilian farmers are quite
unlike French farmers. In Brazil, probably the majority of farmers are subsistence
farmers, while in France, this would generally be rare (non-existent?) and many
farmers are actually viticulturists. There is no particular expectation that
viticulturists should be below average S.
Relationship Between New S, Old S and Cognitive Ability
Even though the old S factor scores do not seem very reliable, it may
nonetheless be useful to compare them to the new ones. The general trouble is
that the departements don't match up perfectly. In some cases, this is due to
name changes, but in other cases a departement has been split. This presents
the problem of how to match them up for comparison. One possibility is that if
departement D1 has been split into D1a and D1b, one can duplicate D1 and
match it with both D1a and D1b. Another is to take the average of D1a and D1b
and match that with D1. The first method will keep the sample size as it is in the
MANKIND QUARTERLY 2015 56:2
128
larger dataset by duplicating some cases, which could be seen as improper, while
the second method will reduce the sample size to the size of the smaller dataset.
Of the two methods, the second is the more conservative so it was chosen.
Details about the matching procedure can be found in the supplementary
material.
Figure 9 shows the scatterplot of old (S_rank) and new S (standard) scores.
This combination was chosen because the ranked analysis in the old dataset
produced more sensible loadings but in the new dataset, there was little change
and interval data is generally to be preferred.
Figure 9. Scatterplot of old (1970s) and new (2010-2015) S scores.
Thus, the relative socioeconomic performance of French departements
seems to have been relatively stable through a period of 4 decades. Note that
because Paris was split up, the Parisian departements do not have proper
counterparts and so are not included. Some of these are very high S (e.g. Hauts-
de-Seine and Yvelines both have S scores >3) and so this probably has the effect
KIRKEGAARD, E.O.W. IQ AND SOCIOECONOMIC VARIABLES IN FRANCE
129
of reducing the correlation because variance is reduced (Hunter & Schmidt,
2004). Table 2 shows the S scores.
Table 2. Socioeconomic (S factor) scores in the old (1970s) and new (2010-
2015) datasets.
Departement
Old S
New S
Departement
Old S
New S
Ain
0.27
1.10
Maine-et-
Loire
-0.58
0.28
Aisne
0.18
-1.09
Manche
-0.63
-0.69
Allier
0.77
-1.10
Marne
0.30
0.32
Alpes-de-Haute-
Provence
0.46
-0.37
Haute-Marne
-0.27
-0.99
Hautes-Alpes
-0.04
-0.06
Mayenne
-0.35
-0.17
Alpes-Maritimes
0.57
0.74
Meurthe-et-
Moselle
0.61
0.31
Ardèche
-0.56
-0.46
Meuse
-0.21
-0.76
Ardennes
0.38
-1.08
Morbihan
-0.14
-0.33
Ariège
-0.37
-1.04
Moselle
0.18
0.16
Aube
-0.01
-0.35
Nièvre
-0.07
-1.62
Aude
-0.47
-1.05
Nord
0.43
-0.20
Aveyron
-0.99
-0.94
Oise
0.68
0.58
Territoire de Belfort
0.13
0.51
Orne
-0.80
-0.91
Bouches-du-Rhône
0.87
0.71
Pas-de-
Calais
-0.12
-1.12
Calvados
0.41
0.12
Puy-de-
Dôme
0.30
0.43
Cantal
-0.54
-1.39
Pyrénées-
Atlantiques
0.40
0.27
Charente
-0.31
-0.67
Hautes-
Pyrénées
-0.40
-0.75
Charente-Maritime
-0.18
-0.52
Pyrénées-
Orientales
-0.66
-0.90
Cher
-0.46
-0.63
Bas-Rhin
0.44
1.05
Corrèze
-0.48
-0.84
Haut-Rhin
0.38
0.78
Corse
-1.07
-0.15
Rhône
0.84
1.80
Côte-d'Or
0.24
0.74
Haute-Saône
-0.49
-0.59
Côtes-d'Armor
-0.47
-0.77
Saône-et-
Loire
-0.26
-0.59
MANKIND QUARTERLY 2015 56:2
130
Departement
Old S
New S
Departement
Old S
New S
Creuse
-1.03
-2.14
Sarthe
0.09
-0.19
Dordogne
0.61
-1.05
Savoie
0.56
0.93
Doubs
0.21
0.68
Haute-Savoie
0.58
1.93
Drôme
0.32
0.11
Seine
1.37
Eure
0.24
0.14
Seine-
Maritime
0.84
0.08
Eure-et-Loir
0.4
0.52
Seine-et-
Marne
1.14
1.68
Finistère
0.13
-0.19
Seine-et-
Oise
1.37
Gard
0.38
-0.19
Deux-Sèvres
-0.79
-0.61
Haute-Garonne
0.6
1.77
Somme
-0.68
-0.67
Gers
-0.5
-0.79
Tarn
-0.36
-0.45
Gironde
0.67
0.86
Tarn-et-
Garonne
-0.85
-0.45
Hérault
0.5
0.25
Var
-0.16
0.16
Ille-et-Vilaine
0.14
0.87
Vaucluse
0.50
-0.15
Indre
-0.49
-1.15
Vendée
-0.61
-0.18
Indre-et-Loire
-0.01
0.57
Vienne
-0.48
0.00
Isère
0.62
1.41
Haute-
Vienne
0.18
-0.34
Jura
-0.43
-0.3
Vosges
-0.43
-0.83
Landes
-0.64
-0.11
Yonne
0.07
-0.68
Loir-et-Cher
-0.47
-0.1
Corse-du-
Sud
-0.18
Loire
0.12
-0.02
Haute-Corse
-0.12
Haute-Loire
-0.56
-0.88
Yvelines
3.24
Loire-Atlantique
0.65
0.92
Essonne
2.19
Loiret
0.31
0.77
Hauts-de-
Seine
3.75
Lot
-0.74
-0.79
Val-de-Marne
2.09
Lot-et-Garonne
-0.65
-0.75
Val-d'Oise
1.70
Lozère
-0.62
-1.21
One might wonder whether the IQ scores from the 1950s still predict
socioeconomic development in France 60 years later. Figure 10 shows the
scatterplot. We see a correlation even after the exclusion of the cases that were
KIRKEGAARD, E.O.W. IQ AND SOCIOECONOMIC VARIABLES IN FRANCE
131
likely to increase the correlation (high S Parisian departements). The correlation
is not impressive however, and the 95% confidence interval reaches close to 0.
Figure 10. Scatterplot of old (1950s) IQ scores and new (2010-2015)
socioeconomic (S) scores.
Discussion and Conclusion
Results from the old dataset were only mixed with regards to their support
for the S factor model. However, results from the new dataset were generally in
line with the S factor model, especially once the effects of non-EU immigrants
were statistically controlled.
As mentioned in the introduction, France has other administrative divisions.
Perhaps one could obtain similar data as those presented in this paper for the
arrondissements and see how the S factor works at that level. I was not able to
find any data, but they may exist. Data are readily available for the 27 regions at
the Insee website, but I did not collect it for the purpose of the present study due
to their limited number.
The fact that the two mixedness detection methods failed to identify the two
MANKIND QUARTERLY 2015 56:2
132
outliers in the old dataset is a good reason to see if one could develop better
methods for identifying mixed cases.
Limitations
The cognitive ability dataset was very old (1950s). It would be good to
acquire PISA scores by departement to see how they relate to S factor
scores. However, I was unable to locate such data, and the person behind
the French human biodiversity science blog (http://thosewhocansee.
blogspot.com/) was unable to find any either.
The old dataset had only 4 socioeconomic variables, which makes the
results difficult to interpret. It is possible to obtain more variables from that
time. This would enable a detailed study of the S factor in France before
the recent mass immigration.
The splitting and reforming of some departements introduce uncertainty
into the longitudinal comparisons.
The limited data concerning which immigrants are present in the
departements limit more detailed analyses of the effect of immigrants on
socioeconomic development. There was also no information regarding
their generation. Usually it is found that second generation immigrants
perform better in the country than first generation, except with regards to
crime (Indvandrere i Danmark 2014, 2014).
The French-language only Insee site makes it difficult for non-French
speakers to research socioeconomic dispersion in France. Some
important variables may have been overlooked for this reason.
There were only data for one level, so it was not possible to examine how
S factor loadings change across levels.
Supplementary material
Data files, R source code and high quality figures are available at the OSF
repository (https://osf.io/g85ap).
Extensive comments on the comparison of the departements in the old
dataset and the new can be found in the French_departments.ods file.
KIRKEGAARD, E.O.W. IQ AND SOCIOECONOMIC VARIABLES IN FRANCE
133
References
Departments of France (2015, June 30). In Wikipedia, the free encyclopedia. Retrieved
from
https://en.wikipedia.org/w/index.php?title=Departments_of_France&oldid=669397801
Indvandrere i Danmark 2014. (2014). København: Danmarks Statistik.
Insee (2015). Circonscriptions administratives au 1er janvier 2015: comparaisons
départementales. Institut national de la statistique et des études économiques. Retrieved
from
http://www.insee.fr/fr/themes/tableau.asp?reg_id=99&ref_id=t_0203D
Jensen, A.R. (1998). The g Factor: The Science of Mental Ability. Westport, Conn.:
Praeger.
Kirkegaard, E.O.W. (2014a). Crime, income, educational attainment and employment
among immigrant groups in Norway and Finland. Open Differential Psychology. Retrieved
from http://openpsych.net/ODP/2014/10/crime-income-educational-attainment-and-
employment-among-immigrant-groups-in-norway-and-finland/
Kirkegaard, E.O.W. (2014b). The international general socioeconomic factor: Factor
analyzing international rankings. Open Differential Psychology. Retrieved from
http://openpsych.net/ODP/2014/09/the-international-general-socioeconomic-factor-
factor-analyzing-international-rankings/
Kirkegaard, E.O.W. (2015a). An S factor among census tracts of Boston. The Winnower.
Retrieved from https://thewinnower.com/papers/an-s-factor-among-census-tracts-of-
boston
Kirkegaard, E.O.W. (2015b). A replication of the S factor among US states using a new
and larger dataset. The Winnower, Retrieved from https://thewinnower.com/papers/1418-
a-replication-of-the-s-factor-among-us-states-using-a-new-and-larger-dataset
Kirkegaard, E.O.W. (2015c). Examining the S factor in Mexican states. The Winnower.
Retrieved from https://thewinnower.com/papers/examining-the-s-factor-in-mexican-
states
Kirkegaard, E.O.W. (2015d). Examining the S factor in US states. The Winnower.
Retrieved from https://thewinnower.com/papers/examining-the-s-factor-in-us-states
MANKIND QUARTERLY 2015 56:2
134
Kirkegaard, E.O.W. (2015e). Finding mixed cases in exploratory factor analysis. The
Winnower, Retrieved from https://thewinnower.com/papers/finding-mixed-cases-in-
exploratory-factor-analysis
Kirkegaard, E.O.W. (2015f). Indian states: G and S factors. The Winnower. Retrieved
from https://thewinnower.com/papers/indian-states-g-and-s-factors
Kirkegaard, E.O.W. (2015g). IQ and socioeconomic development across regions of the
UK: a reanalysis. The Winnower, Retrieved from https://thewinnower.com/papers/1419-
iq-and-socioeconomic-development-across-regions-of-the-uk-a-reanalysis
Kirkegaard, E.O.W. (2015h). S and G in Italian regions: Re-analysis of Lynn’s data and
new data. The Winnower. Retrieved from https://thewinnower.com/papers/s-and-g-in-
italian-regions-re-analysis-of-lynn-s-data-and-new-data
Kirkegaard, E.O.W. (2015i). The S factor in Brazilian states. The Winnower, Retrieved
from https://thewinnower.com/papers/the-s-factor-in-brazilian-states
Kirkegaard, E.O.W. (2015j). The S factor in China. The Winnower. Retrieved from
https://thewinnower.com/papers/the-s-factor-in-china
Kirkegaard, E.O.W. (2015k). The S factor in the British Isles: A reanalysis of Lynn (1979).
The Winnower. Retrieved from https://thewinnower.com/papers/the-s-factor-in-the-
british-isles-a-reanalysis-of-lynn-1979
Kirkegaard, E.O.W. & Fuerst, J. (2014). Educational attainment, income, use of social
benefits, crime rate and the general socioeconomic factor among 71 immigrant groups in
Denmark. Open Differential Psychology. Retrieved from
http://openpsych.net/ODP/2014/05/educational-attainment-income-use-of-social-
benefits-crime-rate-and-the-general-socioeconomic-factor-among-71-immmigrant-
groups-in-denmark/
Kirkegaard, E.O.W. & Tranberg, B. (2015a). Increasing inequality in general intelligence
and socioeconomic status as a result of immigration in Denmark 1980-2014. Retrieved
from
http://openpsych.net/ODP/2015/03/increasing-inequality-in-general-intelligence-and-
socioeconomic-status-as-a-result-of-immigration-in-denmark-1980-2014/
Kirkegaard, E.O.W. & Tranberg, B. (2015b). What is a good name? The S factor in
Denmark at the name-level. The Winnower, Retrieved from
https://thewinnower.com/papers/what-is-a-good-name-the-s-factor-in-denmark-at-the-
name-level
KIRKEGAARD, E.O.W. IQ AND SOCIOECONOMIC VARIABLES IN FRANCE
135
Krawczyk, M. & Reuben, E. (2012). (Un)available upon request: Field experiment on
researchers’ willingness to share supplementary materials. Accountability in Research
19(3): 175–186. http://doi.org/10.1080/08989621.2012.678688
Lorenzo-Seva, U. & Ten Berge, J.M. (2006). Tucker’s congruence coefficient as a
meaningful index of factor similarity. Methodology 2(2): 57–64.
Lynn, R. (1980). The social ecology of intelligence in France. British Journal of Social and
Clinical Psychology 19(4): 325–331. http://doi.org/10.1111/j.2044-8260.1980.tb00360.x
Meisenberg, G. (2007). Does multiculturalism promote income inequality? Mankind
Quarterly 46: 395-433.
Meisenberg, G. (2008). How does racial diversity raise income inequality? Journal of
Social, Political and Economic Studies 33: 3-26.
Montmollin, M.D. (1958). Le niveau intellectuel des recrues du contingent. Population
13(2): 259–268. http://doi.org/10.2307/1524820
Revelle, W. (2015). psych: Procedures for Psychological, Psychometric, and Personality
Research (Version 1.5.4).
Retrieved from http://cran.r-project.org/web/packages/psych/index.html
Savage, C.J. & Vickers, A.J. (2009). Empirical study of data sharing by authors publishing
in PLoS journals. PLoS ONE 4(9): e7078. http://doi.org/10.1371/journal.pone.0007078
Templ, M., Alfons, A., Kowarik, A. & Prantner, B. (2015, February 19). VIM: Visualization
and Imputation of Missing Values. CRAN. Retrieved from http://cran.r-
project.org/web/packages/VIM/index.html
Wicherts, J.M., Bakker, M. & Molenaar, D. (2011). Willingness to share research data is
related to the strength of the evidence and the quality of reporting of statistical results.
PLoS ONE 6(11): e26828.
http://doi.org/10.1371/journal.pone.0026828
