Content uploaded by Emil O. W. Kirkegaard
Author content
All content in this area was uploaded by Emil O. W. Kirkegaard on Dec 22, 2020
Content may be subject to copyright.
MANKIND QUARTERLY 2020 61:2 293-323
293
What Happened to Brussels? The Big Decline and Muslim
Immigration
Emil O. W. Kirkegaard*
Ulster Institute for Social Research, London, UK
Baptiste Dumoulin**
Independent Researcher, Belgium
* Email: emil@emilkirkegaard.dk
**Email: BaptisteDumoulin@hotmail.com
We examined regional inequality in Belgium, both in the 19
communes of Brussels and in the country as a whole (n = 589
communes). We find very strong relationships between Muslim% of
the population and a variety of social outcomes such as crime rate,
educational attainment, and median income. For the 19 communes
of Brussels, we find a correlation of -.94 between Muslim% and a
general factor of socioeconomic variables (S factor) based on 22
diverse indicators. The slope for this relationship is -7.52, meaning
that a change in S going from 0% to 100% Muslim corresponds to a
worsening of overall social well-being by 7.52 (commune-level)
standard deviations. For the entire country, we have data for 8
measures of social inequality. Analysis of the indicators shows an S
factor which is very similar to the one from the Brussels data only
based on the full set of indicators (r’s = .98). In the full dataset, the
correlation between S and Muslim% is -.52, with a slope of -8.05.
Adding covariates for age, population density, and spatial
autocorrelation changes this slope to -8.77. Thus, the expected
change going from a 0% to 100% Muslim population is -8.77
standard deviations in general social well-being. We discuss our
findings in relation to other research on immigration and social
inequality, with a focus on the causal influence of intelligence on life
outcomes in general.
Key words: Belgium, Brussels, Inequality, Immigration, Muslim,
Islam, Income, Crime, S factor, Spatial autocorrelation, Intelligence
MANKIND QUARTERLY 2020 61:2
294
Northwest European countries have received large numbers of non-
European immigrants beginning approximately in the 1960s. In every country,
some immigrant groups have fared poorly, and this has resulted in strong public
opposition to immigration among the natives such that most elections in the past
two decades have been strongly focused on the immigration question (Kaufman,
2019; D. Murray, 2017; Roth, 2010; Sanandaji, 2017; Sarrazin, 2012). It has been
noted many times, however, that immigrants are not all alike, but outcomes differ
by country of origin (Adsera & Chiswick, 2006; Beenstock et al., 2001; Borjas,
2016; Boyd & Thomas, 2002; Jones & Schneider, 2010; Osili & Paulson, 2008;
Vinogradov & Kolvereid, 2010). Many previous studies have examined the
relative social performance or well-being — measured as income, educational
attainment, crime rate, unemployment etc. — of immigrant groups according to
country of origin (Jones & Schneider, 2010; Kirkegaard, 2015a; Kirkegaard &
Becker, 2017; Kirkegaard & Fuerst, 2014; Vinogradov & Kolvereid, 2010). These
studies show that immigrant populations usually, but not always, perform below
the native population. Furthermore, country of origin characteristics such as
Muslim percentage of the population and average intelligence in the country
(measured as IQ or as scores on scholastic achievement tests such as PISA)
strongly predict outcomes (mean absolute correlation around .60) and plausibly
serve as causal variables that country of origin is a proxy of (Kirkegaard & de
Kuijper, 2020).
In the Belgian context, the capital of Brussels (Bruxelles) saw the relative
wealth of its inhabitants reduced dramatically in about 50 years, shown in Figure
1 (Ashworth et al., 2003; Van Hamme, 2015). In this same period, the population
composition of the city changed drastically from mostly natives, or at least
persons of European ancestry, to now being approximately 25% Muslim of mostly
non-European ancestry. The wealth index (the ratio of average earnings in a
given unit in Belgium compared to the Belgian mean) in the Belgian capital
declined from about 150 in the early 1970s to 79 in 2015, meaning that the relative
prosperity of the Brussels region is today half of what it was 50 years before. This
decline in well-being is very noticeable to the inhabitants and is strongly
concentrated in neighborhoods with more immigrants. However, we are not
aware of any published academic work statistically examining the relationships
between immigrant populations, or Muslims in particular, and well-being in
Brussels or Belgium at large, though it has been mentioned occasionally (Corijn
& Ven, 2013, Chapter 5). The purpose of the present study is to fill this gap in the
literature.
KIRKEGAARD, E.O.W. & DUMOULIN, B. WHAT HAPPENED TO BRUSSELS?
295
Figure 1. Timeline of Brussels wealth index compared to surrounding areas,
1967-2011 (data from Van Hamme, 2015).
Data
We compiled data from a number of sources for this study since a collection
of suitable data was not available at a single source. Some variables were
available for many years, and others only for a single year. For each variable, we
attempted to pick a single year of data closest to the mean of a given dataset, or
averaged surrounding years of data if the data were noisy. The details of the data
sources and their years are given in the supplementary materials. Generally
speaking, the data sources were governmental sources such as national or local
government web portals. As such, they are considered to have high validity.
One data source, however, warrants some discussion here. Government
population records give the number of persons by citizenship in each municipality
from 1989 to 2019. We estimated Muslim% for each municipality using the Muslim
prevalence in origin countries in 2010 as estimated by Pew Research (Pew
Research Center, 2011). However, we found that this method did not work as
expected because many immigrants change their citizenship, and their children
may be born with Belgian citizenship, thus being counted as natives and non-
Muslims in these data. This problem has been noted by others as well e.g.
MANKIND QUARTERLY 2020 61:2
296
“Official statistics do not provide a good reflection of people’s ethnic origins in
Brussels.” (Corijn & Ven, 2013, Chapter 5) and was confirmed by a government
employee by email. This results in many estimates of Muslim% actually
decreasing while in reality they are increasing. Figure 2 shows the estimated
Muslim% using this method.
Figure 2. Muslim% in Belgium as estimated by citizenship and country of origin
Muslim%. Each black line is a commune (n = 589). The red line shows the entire
country.
Because of this issue, we sought a different source of Muslim% estimates.
We found public estimates by Belgian sociologist Jan Hertogen (Dutch speaker)
who runs the website http://www.npdata.be/ (“Non-Profit Data”). He has published
estimated Muslim% for the years 2011, 2013, 2015-2017 (Hertogen, 2017). His
KIRKEGAARD, E.O.W. & DUMOULIN, B. WHAT HAPPENED TO BRUSSELS?
297
values are based on the same citizenship data we obtained, but he makes
adjustments to correct for the people who changed their citizenship, as well as
their children. According to his 2011 estimates, Belgium was 6.3% Muslim while
for Brussels it was 22.4%. Pew Research estimated the country value at 6.0% for
2010 (Pew Research Center, 2011), so the estimates are closely in line. Figure 3
shows his estimates, used in this study. All data used in the study are available
for reuse in the supplementary materials at https://osf.io/ja6ce/.
Figure 3. Muslim% in Belgian communes as estimated by Jan Hertogen. Each
black line is a commune (n = 589). The red line shows the entire country.
Analyses
Analyses are first presented for the Brussels region where we have the most
detailed data. The units are relatively homogenous in terms of population density
MANKIND QUARTERLY 2020 61:2
298
and geographical factors that might distort the results. After this we expand our
scope to the entirety of Belgium and assess whether the results are in line with
those from the capital region. Figure 4 shows a reference map of Belgium with
provinces and communes.
Figure 4. Map of Belgium showing the division into provinces and communes.
Brussels
We factor analyzed the available 22 indicators of well-being. A very strong
(76% of variance) general factor was found. We used extensive method variation
to check if this was robust to choice of factor analytic method and factor scoring
method, and we found that it was (all method variations resulted in factor scores
KIRKEGAARD, E.O.W. & DUMOULIN, B. WHAT HAPPENED TO BRUSSELS?
299
that correlated near 1.00), see statistical output for details. Figure 5 shows the
estimated factor loadings across 3 weighing method variations.
Figure 5. General socioeconomic factor (S) loadings across three methods of
estimation.
In the figure, standard is the default settings used by fa() from the psych
package (Revelle, 2020). This is the combination of unweighted least squares for
the factor loadings estimation, and the regression method for scoring. The
weights analysis was done using the square root of population size as weights,
and finally the rank analysis was done using rank-transformed data. Inspection of
the factor loadings showed that they conformed to the usual interpretation:
desirable indicators had positive loadings and undesirable indicators negative
loadings (Pesta et al., 2010). This allows for a simple interpretation of the resulting
scores as a clean measure of social well-being or generalized socioeconomic
status, termed the S factor (see e.g. Kirkegaard, 2014a; Kirkegaard & Fuerst,
2017 for other examples of such factor analyses). Figure 6 shows the scatterplot
between Muslim% and the S factor, while Figure 7 shows a map of Brussels with
the S factor score. The supplementary materials have maps of Brussels with each
MANKIND QUARTERLY 2020 61:2
300
variable. For comparison, Figure 8 shows a similar map of Brussels with
percentage of Muslims.
Figure 6. Scatterplot of communes of Brussels showing Muslim% and general
socioeconomic factor (S). Weighted by the square root of population size.
Figure 7. Map of Brussels showing variation in general socioeconomic factor (S)
score.
KIRKEGAARD, E.O.W. & DUMOULIN, B. WHAT HAPPENED TO BRUSSELS?
301
Figure 8. Map of Brussels showing variation in Muslim%.
The scatter plot shows an extremely strong negative correlation such that the
greater the Muslim% in a city district, the worse the social well-being. The
relationship between Muslim% and every indicator of S was very strong, as
shown in Table 1. The relationship to the S factor was ‘S-loaded’ in that the social
indicators that were more strongly related to the S factor also had a stronger
relationship to Muslim%, r = .92, shown in Figure 9.
Table 1. Correlation between Muslim% and well-being indicators. Weighted by
the square root of population.
Variable
Correlation with Muslim
%
Wealth index
-0.94
Wage index
-0.94
Share of households in demand for social housing % (2012)
0.84
Unemployment rate (2012)
0.96
Median income in Euro (2012)
-0.90
18-64 year old beneficiaries of social integration income % (2012)
0.90
Delinquent population aged <18 years %
0.89
Delinquent population aged 18-25 years %
0.91
University degree %
-0.80
Low birth weight %
0.50
Life expectancy at birth
-0.73
MANKIND QUARTERLY 2020 61:2
302
Variable
Correlation with Muslim
%
Children born in a family with no income from work %
0.96
Median house price in Euro
-0.65
Fraudulent insurance declarations %
0.68
Rate of borrowers with bad credit
0.89
"Mothers <20 years % (2009-2013)
0.83
Infant mortality (2009-2013)
0.62
Proportion two years late in school (2013-2014)
0.93
Special education mental deficiency % (2013-2014)
0.80
Relative risk of mortality by cardiovascular diseases, men (2009-2013)
0.56
General humanities to technical and professional humanities ratio in
education
-0.86
Blood donor %
-0.85
General socioeconomic factor (S)
-0.94
Figure 9. Jensen’s method used on Muslim% and general socioeconomic status
(S factor). Note that indicators with negative loadings were reversed to avoid
variance inflation (Kirkegaard, 2016b), and are marked with “_r”.
KIRKEGAARD, E.O.W. & DUMOULIN, B. WHAT HAPPENED TO BRUSSELS?
303
This Jensen pattern1 suggests that whatever it is Muslim% relates to, it is
something that relates chiefly to the joint variation of a broad variety of indicators
of well-being. This pattern has previously been found for immigrant group
differences in Denmark (Kirkegaard, 2014b), and for a variety of predictors at the
national level (Kirkegaard, 2014a).
Instead of looking at the S factor, which is a unitless metric, we can also
examine specific indicators. Table 2 shows regression results for selected
indicators, as well as S itself. Because all the units are part of Brussels, we
considered it unnecessary to control for e.g. population density.
Table 2. Summary of regression results for general socioeconomic status (S
factor), median income, educational attainment, and unemployment rate.
Weighted by the square root of population. * = p <.01, ** = p < .005, *** = p < .001.
B coefficients with standard errors in parentheses.
Predictor/Model
Median
income
Unemployment
rate
Uni degree S factor
Intercept
23106***
(553)
11.82***
(0.850)
14.56***
(1.27)
1.48***
(0.16)
Muslim%
-19357***
(2304)
48.36***
(3.542)
-29.01***
(5.30)
-7.52***
(0.65)
R2 adj.
0.795
0.912
0.616
0.881
N
19
19
19
19
Belgium
In the analyses above, we saw that the association of social outcomes with
Muslim% in Brussels is very strong. However, it is not certain that it will hold up
for the country at large. To examine this, we computed a general socioeconomic
factor score from 8 available indicators for each commune: median income
reported (2016), mean income reported (2016), income inequality (2016), total
crime rate (average of 2009-2016), violent crime rate (average of 2009-2016), life
expectancy (2014), proportion of population with a university degree (2011), and
unemployment rate (average of 2009-2016). The general factor showed very little
method variance, accounted for 47% of the variance, and showed a similar
loading pattern as that found for Brussels. The factor scores derived from the
Brussels area communes correlated .98 with those for the same units derived
from the full country analysis, and .98 with those derived from the full set of 22
1 The concept was named Jensen effect by Phil Rushton for the analogous finding where
g-loadings of intelligence tests relate to the strength of tests’ relationship with other
variables (Rushton, 1998), but we have generalized it here to any factor.
MANKIND QUARTERLY 2020 61:2
304
indicators used in the analysis above. Figure 10 shows the scatterplot of Muslim%
and the S factor score for all communes. Figures 11 and 12 show maps of
Belgium colored by S factor score and Muslim%. The supplementary materials
contain maps for every other variable. It can be seen that for Belgium as a whole,
Muslim% is correlated with worse outcomes, though not as strongly as for just the
municipalities of Brussels. Table 3 shows the correlation matrix between the
primary variables.
Figure 10. Scatterplot of Muslim% and general socioeconomic factor (S factor).
Weighted by the square root of population.
KIRKEGAARD, E.O.W. & DUMOULIN, B. WHAT HAPPENED TO BRUSSELS?
305
Figure 11. Map of Belgian communes showing variation in general
socioeconomic status (S) factor scores.
Figure 12. Map of Belgian communes showing variation in Muslim %.
MANKIND QUARTERLY 2020 61:2
306
Table 3. Correlation matrix for primary variables for Belgian municipalities, n =
561-589. Values below diagonal are weighted by the square root of population.
S = general social well-being.
S
Muslim%
Median income
Mean income
Income inequality
S
-0.41
0.88
0.88
0.48
Muslim %
-0.52
-0.37
-0.24
0.05
Median income
0.89
-0.51
0.85
0.26
Mean income
0.87
-0.32
0.83
0.59
Income inequality
0.40
0.13
0.19
0.58
Crime rate
-0.65
0.63
-0.50
-0.27
0.07
Violent crime rate
-0.81
0.49
-0.58
-0.45
-0.16
Life expectancy
0.64
-0.15
0.49
0.53
0.32
Higher educ. %
0.59
-0.06
0.45
0.77
0.86
Unemployment
-0.81
0.72
-0.74
-0.51
0.04
Table 3. Correlation matrix for primary variables for Belgian municipalities, n =
561-589. Values below diagonal are weighted by the square root of population.
S = general social well-being.
Crime rate
Violent crimes
Life expect.
Higher edu. %
Unempl.
S
-0.61
-0.79
0.61
0.65
-0.78
Muslim %
0.56
0.40
-0.10
-0.08
0.62
Median income
-0.43
-0.54
0.45
0.50
-0.67
Mean income
-0.27
-0.45
0.47
0.78
-0.48
Income inequality
-0.03
-0.24
0.28
0.86
-0.05
Crime rate
0.85
-0.28
-0.10
0.68
Violent crime rate
0.88
-0.47
-0.31
0.73
Life expectancy
-0.28
-0.47
0.35
-0.53
Higher educ. %
-0.02
-0.24
0.41
-0.19
Unemployment
0.70
0.72
-0.53
-0.15
To understand whether the pattern was plausibly causal, we fit a series of
regression models adding potential confounders. In the second model step, we
add the traditional controls: age, and population density. We used natural splines
to adjust for any nonlinear effects of these (Harrell, 2015, 2019). For the third
step, we add the spatial lag, i.e. the outcome variable as predicted by spatial data.
Concretely, we used spatial k nearest neighbor regression by averaging the
values of the three nearest neighboring units (Anselin & Bera, 1998; Kirkegaard,
2015b; Pesta et al., in review see spatial statistics supplement). The position of
each commune was calculated by the centroid of their polygons. The
autocorrelations for the main variables are shown in Table 4 (each variable
correlated with its own lag).
KIRKEGAARD, E.O.W. & DUMOULIN, B. WHAT HAPPENED TO BRUSSELS?
307
Table 4. Spatial autocorrelation (SAC r) for the main variables in the country-wide
dataset. Unweighted correlations.
Variable
SAC r
General socioeconomic factor
0.74
Life expectancy
0.58
Mean income
0.79
Median income
0.77
Income inequality
0.76
Muslim%
0.73
Crime rate
0.47
Violent crime rate
0.50
Unemployment
0.83
Population density
0.82
Age 65 and above %
0.71
Age 0-19 %
0.79
The autocorrelation results confirm the usual expectations: everything close
to each other is more similar (spatially autocorrelated, SAC), and the
mean/median SAC are .70/.75. This has come to be called Tobler’s first law of
geography (for historical review, see Tobler, 2004; and for a powerful modern
illustration, see Li et al., 2014), though it relates back to Francis Galton, and the
problem is known also as Galton’s problem (Eff, 2004). As many have noted
(Gelade, 2008; Hassall & Sherratt, 2011), when SAC is present, datapoints are
not fully independent and then residuals from models will usually also be SAC
which violates the assumption of most regression methods. The regression
results are shown in Table 5.
Across the three outcomes, we see that there are large effects of Muslim%.
For S factor, the simple model suggests a -8.05 SD decrease in general well-
being by increasing the Muslim% from zero to 100%. Adding age and population
density controls (betas are not shown because nonlinear) increases this estimate
to a 12.80 SD decrease. However, adding the spatial lag in Model 3 reduces it to
an 8.77 SD decrease. For income, we see weaker results compared to those from
Table 2, based on just the 19 communes of Brussels. This is not surprising
because much income variation is linked to population density — cities and city-
adjacent areas are richer on average (cf. Figure 11). However, mass immigration
has primarily been to the cities as well. This migration pattern results in cities now
being a mix of above average natives (and other Europeans), and generally low-
status immigrants. This creates a particularly large contrast among city
MANKIND QUARTERLY 2020 61:2
308
communes, and thus the large slope seen. Consistent with this, the slope
increases in Model 2 when we add covariates which include a control for
population density. Finally, in Model 3, the slope shrinks to -6.8k. Thus, if we
imagined two counterfactual Belgiums, one with 0% and one with 100% Muslims,
the one with 100% would be €6,800 lower in median income. For crime rate, we
see that Model 1 has a slope of 3.76, thus an increase in Muslim% to 100% from
0% would result in a 376%points increase in crime rate (the intercept is 0.65, so
this would be an increase of 478%). This value is reduced to 224%points in Model
2, and 198%points in Model 3. Surprisingly, the lag variable is weaker for the
crime model than for the other two, despite the fact that one might expect
contagious effects of crime.
Table 5. Regression models for general socioeconomic status (S factor) for
Belgian municipalities. Weighted by the square root of population. * p < .01, ** p
< .005, *** p < .001. Standard errors in parentheses.
Predictor/Model
1
2
3
outcome = S
Intercept
0.27*** (0.046)
3.13 (1.540)
3.89*** (1.135)
Muslim%
-8.05*** (0.55)
-12.80*** (0.95)
-8.77*** (0.72)
S lag
0.73*** (0.033)
age & population density
no
yes
yes
R2 adj.
0.268
0.384
0.666
N
589
589
589
outcome = median income
Intercept
18699*** (58)
16681*** (1065)
4478*** (905)
Muslim%
-9909*** (694)
-12608*** (1250)
-6788*** (914)
median income lag
0.77*** (0.032)
age & population density
no
yes
yes
R2 adj.
0.256
0.378
0.691
N
589
589
589
outcome = crime rate relative rate (country = 1)
Intercept
0.65*** (0.016)
1.37*** (0.283)
0.71 (0.292)
Muslim%
3.76*** (0.194)
2.24*** (0.335)
1.98*** (0.326)
crime rate RR lag
0.38*** (0.058)
age & population density
no
yes
yes
R
2
adj.
0.393
0.537
0.568
N
582
582
582
KIRKEGAARD, E.O.W. & DUMOULIN, B. WHAT HAPPENED TO BRUSSELS?
309
As a robustness test, we residualized the S indicator variables for age and
population density before conducting the factor analysis. This would theoretically
allow one to leave out the age and population density variables in the regression
models, and might affect the structure of the S factor. However, we found that the
S factor scores from this approach correlated r = .90 with those from the standard
approach, and we did not investigate it further.
Finally, we examined whether the Jensen pattern was also present for the
country as a whole, as it was for the Brussels region (cf. Figure 9). This analysis
presents the difficulty that we know that age distribution and population density
confound the pattern. For this reason, we introduce a new variant on the method
by using metrics from regression models. Specifically, we fit a regression model
for each S indicator as the outcome with and without Muslim%, along with the age
and density controls. We saved the model R2 gain from adding the Muslim%
predictor, as well as the slope of this predictor. Finally, we reverse the negative
S indicators as normally done. Table 6 shows the resulting values from this
approach, and Table 7 shows their correlations.
Table 6. Expanded Jensen method results for Belgian general socioeconomic
status indicators, and Muslim% as predictor.
Indicator Reversed Loading r (indicator *
Muslim%)
r (indicator
*
Muslim%),
weighted
Partial
r2 Δ r2 Slope
Median income
no
0.83
-0.37
-0.51
0.13
0.13
-10.38
Mean income
no
0.85
-0.24
-0.32
0.19
0.19
-12.06
Income inequality
no
0.46
0.05
0.13
0.10
0.09
-8.52
Crime rate
yes
0.57
-0.56
-0.63
0.08
0.08
-9.73
Violent crime rate
yes
0.76
-0.40
-0.49
0.11
0.11
-10.46
Life expectancy
no
0.59
-0.10
-0.15
0.04
0.03
-4.46
Higher education
%
no 0.62 -0.08 -0.06 0.16 0.17 -11.43
Unemployment %
yes
0.74
-0.62
-0.71
0.07
0.08
-8.72
Table 7. Correlations among variables from Table 6. Pearson correlations above,
and Spearman correlations below the diagonal.
Loading
r (indicator
*
Muslim%)
r (indicator
*
Muslim%), weighted
Partial
r
2
Δ r2 Slope
Loading
-0.46
-0.56
0.51
0.53
-0.50
r (indicator x
Muslim%)
-0.31 0.99 0.21 0.15 0.18
r (indicator x
Muslim%),
weighted
-0.33 0.98 0.17 0.12 0.18
Partial r2
0.60
0.29
0.26
1.00
-0.87
Δ r2
0.60
0.29
0.26
1.00
-0.90
Slope
-0.69
0.05
0.02
-0.90
-0.90
MANKIND QUARTERLY 2020 61:2
310
Though we only have 8 S indicators, we see evidence of values in the
expected direction. The correlations vary somewhat, but average |.50| across
methods. Note that since the correlations and slopes are signed, negative values
are expected, i.e., the variables with stronger loadings, have a stronger negative
relationship with Muslim %.
Discussion
We studied regional (subnational) inequality in Belgium using a rich dataset
of communes (n = 589). Our main goal was to examine to what degree inequality
between communes could be explained by the population proportions of Muslims.
In particular, we were interested in the explanation for the relative decline of
Brussels. It has generally been found in regional research that capital cities or
areas are unusually high in general well-being (Fuerst & Kirkegaard, 2016).
However, in Belgium the exact opposite is the case, raising the question of why
this is so (Ashworth et al., 2003).
In all models, we find large, statistically certain (low p values) effects of
Muslim% of the population. For the models of Brussels, we did not utilize any
controls since the communes are already very similar in plausible exogenous
confounders we could control for. For the regressions using the complete dataset,
our addition of a spatial lag control is of note. A spatial lag controls for
unmeasured confounders. These are variables that cause S and that are
correlated with Muslim% but are not entirely caused by it (Rohrer, 2018), to the
extent that they are spatially autocorrelated (Baller et al., 2001; Jerrett et al., 2003;
Kühn, 2007; Pesta et al., in review, spatial statistics supplement). As we saw in
Table 4, most of our variables are very highly spatially autocorrelated (mean r =
.70). We can probably expect plausible unobserved or partially observed
confounders to be so as well, thus being at least partially accounted for by our
spatial lag control. The remaining estimates in the full models are still quite large
and socially significant. For instance, going from a 0% to 100% Muslim population
would yield an estimated increase in the crime rate of about 200%points, cf. Table
5, column 3.2 However, the slopes were often weaker when we added spatial lag
control compared to only the classic controls. The interpretation of this is tricky.
Decreases in slopes from adding spatial lags can indicate a successful control for
unmeasured confounding variables, but it can also indicate measurement error in
2 The association of Muslim immigrants with crime is long-standing and of great interest
to the public according to survey data. While associations in the USA are weaker, they
are quite strong in European countries (Ahlberg, 1996; Carl, 2016; Junger & Polder,
1992; Kirkegaard, 2014b; Kirkegaard & Becker, 2017; Kirkegaard & de Kuijper, 2020;
Skardhamar et al., 2014).
KIRKEGAARD, E.O.W. & DUMOULIN, B. WHAT HAPPENED TO BRUSSELS?
311
the predictor. Our estimates of Muslim% are certainly not without error, especially
when comparing units close in proximity, so by this reason alone we expected
some decrease in the slope. Furthermore, the sizable betas of some of the spatial
lag variables (e.g. .73 for S factor in Table 5, Model 3) could also suggest that
there are contagion effects, i.e. that neighboring units affect each other’s well-
being. For instance, having a neighboring commune or city district with many
criminals probably means at least some of them will cause crime in your
neighborhood, which causes an assortment of problems downstream, such as
decreasing house prices and increased spending on local police or private
guards. These effects would eventually show up in the various well-being
indicators and thus the S factor. However, when we actually model the crime rate
itself, it shows only a moderately strong beta for the spatial lag (.38) which
suggests this particular contagion effect is not strong, at least for our data. In the
same vein, crime rate actually has the weakest SAC of our variables, .47, cf.
Table 4. These results seem to speak against the spatial contagiousness or
roaming criminals models (Loftin, 1986; Mennis & Harris, 2013; Osorio, 2015).
The main focus of this investigation was our overall measure of general social
well-being, or general socioeconomic factor. However, most research focuses on
particular indicators of this general factor, and it is worth reviewing some of this
to see whether it is congruent with our findings. Among the indicators used for
Brussels, we have the rate of fraudulent insurance declarations, which shows a
correlation to Muslim% of .68. This finding is in line with results by Carl (2017)
who found that in the United Kingdom, the population proportion of Pakistanis and
Bangladeshis (both Muslim populations) was a strong predictor of electoral fraud.
He interpreted this in line with a history of cousin marriages promoting
ethnocentric behavior resulting in relatively stronger distrust and hostility towards
outsiders, including scamming behaviors (Schulz et al., 2019; Woodley & Bell,
2013).
Lynn (2020) reviewed findings on race differences in blood donation, as a
measure of altruism. He finds that overall European-descent (‘whites’) have more
blood donors per capita than the 5 other racial/ancestry groups considered,
including East Asians. He did not find a study that included Middle Easterners or
Muslims in particular, but one German study compared immigrants to natives and
found that the native blood donor rate was slightly less than twice that of
immigrants, perhaps a majority of which are Muslims (Boenigk et al., 2015). Thus
the rank order does not follow the usual IQ order entirely, as the results suggest
that East Asians are less altruistic than Europeans despite superior average
intelligence. In fact, inspection of country-level data about foreign aid suggests
the same. The largest 10 donor countries in terms of % of GNI to foreign aid are
all European, mainly Northern European (see appendix for details). These various
MANKIND QUARTERLY 2020 61:2
312
results fit well with our finding of a negative correlation between blood donations
and Muslim% of -.85 among the Brussels communes (Table 1).
Among the Brussels communes we found that the proportion of students in
special education for the mentally retarded or who are otherwise educationally
backwards has a strong relationship to Muslim%, r = .80 (Table 1).3 This result fits
well with cognitive testing results from Belgium. PISA studies provide us with
data, shown in Tables 8A-C (Jacobs et al., 2007; Lafontaine et al., 2017).
Table 8A. Scholastic ability results from PISA 2003 by origin group: natives
versus first and second generation immigrants. The standard deviation for natives
is around 90.
Origin Science Reading Math Problem solving
Natives 524 523 545 540
1st 416 407 437 447
2nd 435 439 454 445
Table 8B. Scholastic ability results from PISA 2003 by national origin.
Mother’s origin
French area
Dutch area
Natives
514
567
France
409
Netherlands
521
Other EU countries
476
469
Eastern Europe
508
479
Turkey
430
414
Maghreb
434
452
Africa (non-Maghreb)
429
454
Other countries
437
456
3 To be sure, we checked the criteria used for this classification
(https://www.vaph.be/professionelen/mdt/mdv/modules/verstandelijke-handicap). The
diagnoses are primarily based on low IQ scores on standardized tests (Wechsler, SON-
R and others), but also based on behavioral and learning problems in school (Deblonde
et al., 2020). Race differences in rates of mental retardation are well known and relate
to the difference in average intelligence and thus also the proportion that are below a
given threshold (Jensen, 1969; Yeargin-Allsopp et al., 1995).
KIRKEGAARD, E.O.W. & DUMOULIN, B. WHAT HAPPENED TO BRUSSELS?
313
Table 8C. Scholastic ability results from PISA 2015 by origin group: natives
versus first and second generation immigrants.
Group
Science
Reading
Math
French area and Brussels
Natives
499
496
502
1st
438
440
443
2nd
459
465
465
Dutch area
Natives
529
523
534
1st
419
427
425
2nd
448
427
425
German area
Natives
511
508
510
1st
457
455
456
2nd
472
470
470
We see that substantial cognitive gaps exist, whether the data concern
reading comprehension, mathematics, science or problem solving; whether it is
in French, Dutch, or German speaking areas; and whether it was measured in
2003 or 2015. Additionally, Klein et al. (2007) tested 69 sub-Saharan Africans
living in Belgium on a reduced form of the Cattell test as part of a stereotype threat
study (study participation rate was 15%). They obtained an IQ score of 81.
However, when the necessary adjustment for the Flynn effect is applied, this
results in an estimate of roughly 70 (see the appendix for details of calculation).
Rindermann & Thompson (2016) calculated an overall IQ metric score of 92 for
Belgian immigrants relative to natives’ 100 based on various waves of scholastic
testing data. The standard deviation for the PISA scale among natives is
approximately 90, so the gaps we see in Table 8 are congruent with Rindermann
and Thompson’s results.
The presence of a substantial cognitive gap is important because of the
evidence that intelligence gaps are the primary cause of scholastic ability gaps
and of later social inequality. Evidence for this claim comes from many research
designs including prospective studies, which rule out reverse causation, sibling
studies, which rule out any confounder that differs between families (Frisell et al.,
2012; Hegelund et al., 2018, 2019; Herrnstein & Murray, 1994; C. Murray, 2002),
and GWASs that allow for functional analysis of genetic causes of intelligence
and outcomes, as well as studies using polygenic scores constructed from them
(Belsky et al., 2016; Hill et al., 2019; Lee et al., 2018). Thus, generally, one would
MANKIND QUARTERLY 2020 61:2
314
expect cognitive gaps to result in social gaps. The various estimates of cognitive
gaps summarized above are therefore a likely cause of immigrant-native gaps in
social outcomes. Unfortunately, evidence to directly test this mediation is
generally lacking. One exception is the study by Nordin & Rooth (2007), which
examines income gaps between natives and various second generation
immigrants in Sweden. They used IQ data from the Swedish military draft test
taken at age 18, and income data from the national register. They found that 66%
of the native income advantage over those from “outside Europe” could be
explained by the test score gap (their Table 2, column 2). There will be some
measurement error in the IQ test, so if this were accounted for, the true mediation
% will be higher. Nevertheless, it is a plausible lower bound estimate. In their
model that includes family background (a composite of parental years of
schooling and annual income), they found that 88% of the gap could be explained.
However, including family background in a regression is causally ambiguous
because of its genetic correlations with offspring personality traits. It would be
informative to replicate this study in other countries. The most likely candidates
would be other Nordic countries with similar existing datasets. Preferably, one
would include measures of Muslim religious beliefs to gauge their potential causal
effects, controlling for intelligence. Using a sibling design would allow for a natural
adjustment for any parental effects. It is unfortunate that Hegelund et al. (2019),
who conducted a sibling study for IQ and social outcomes in Denmark, did not
examine the question about immigrants more closely.
Our results from Jensen’s method, i.e., the positive relations between S
factor loadings and measures of strength of association with Muslim%, adds
plausibility to the above interpretation of causality mediated by intelligence. This
is because the same analysis has been done with intelligence as the predictor in
other studies, and very strong Jensen patterns were also found for this variable
(Kirkegaard, 2014a, 2016a). This essentially serves as a kind of fingerprint of the
causal variable involved. This kind of interpretation has also been advocated with
regard to genetic causation of other group gaps (Metzen, 2012; Rushton &
Jensen, 2010).
Limitations
Perhaps the most important confounding variable is that many Muslims are
first generation and relatively recent immigrants. It is well known that it takes a
number of years for immigrants to ‘catch-up’ to their maximal social performance
level (Boyd & Thomas, 2002; Hu, 2000; Husted et al., 2001). Unfortunately, we
do not have any good estimates of this confounding factor (e.g. mean duration of
stay) or any suitable proxy such as proportion of Muslims who are first generation
in each commune. The causal route of this factor would probably mainly be
KIRKEGAARD, E.O.W. & DUMOULIN, B. WHAT HAPPENED TO BRUSSELS?
315
language proficiency, unless the migrant happens to speak the same language
in his home country, e.g. due to colonial heritage. It will probably also relate to
broad cultural knowledge, and perhaps education in the host country (especially
for younger immigrants).
Our estimates of Muslim% of the population were less than ideal. However,
one cannot argue that our findings are questionable because the Muslim
estimates were produced by political opponents of immigration of Muslims. The
sociologist who produced them is clearly left-leaning. His website features a quote
by Karl Marx, and he often is critical towards claims by Belgian nationalists (who
are critical of Muslims). Thus, our results cannot be explained by the estimates
being purposefully or inadvertently produced to make Muslims look bad, quite the
opposite. This being said, future research should attempt to find ways to estimate
population proportions more effectively, especially when the governments seem
uninterested or actively hostile to doing so (e.g. in France where such statistics
are illegal for the government to collect).
One promising route for Belgium is to use data from first names. The Belgian
statistics agency (StatBel) publishes counts of first names for newborns and the
total population for every commune (https://statbel.fgov.be/en/open-
data?category=214). With these, one could assign every name to a probable
origin using databases of known names (such as
https://www.behindthename.com/), perhaps augmented by machine learning
methods. Better yet would be to convince or pay for StatBel to compute ancestries
for each commune. This can be done accurately using a recursive method that
relies on having an accurate population register. With this method, one begins
with every person alive, and asks if one has data about their parents or not. If yes,
then one asks whether one has data about the parents’ parents or not. At some
point, no further data are available about parents. Then one asks whether the
person was born in Belgium, and if not, where they were born. Then one assigns
the ancestry of the origin to the person this way. Finally, one averages the
ancestry of all ancestors to estimate the ancestry of living persons. The result will
be a quite accurate estimate of the population level ancestry, insofar as country
of origin is concerned. This method has been used in Denmark recently, with
results forthcoming.
Thus to answer our question in the title of the paper. The big decline of
Brussels is mainly due to large-scale low-intelligence immigration to the capital
region. The immigrants have generally not been selected for their ability to
contribute to Belgian society, and originated from second- and third-world
countries. Furthermore, they are mainly Muslims, and across the world, adherents
of this faith generally perform relatively poorly compared to other groups, both in
Muslim majority and non-Muslim majority countries (Kuran, 1997).
MANKIND QUARTERLY 2020 61:2
316
Supplementary material and acknowledgments
Supplementary materials including code, high quality figures and data can
be found at https://osf.io/ja6ce.
References
Adsera, A. & Chiswick, B.R. (2006). Are there gender and country of origin differences
in immigrant labor market outcomes across European destinations? Journal of
Population Economics 20: 495. https://doi.org/10.1007/s00148-006-0082-y
Ahlberg, J. (1996). Invandrares och invandrares barns brottslighet: En statistisk analys
[Immigrants’ and immigrants’ children’s crime: A statistical analysis].
Brottsförebyggande rådet (BRÅ).
Anselin, L. & Bera, A.K. (1998). Spatial dependence in linear regression models with an
introduction to spatial econometrics. In: A. Ullah (ed.), Handbook of Applied Economic
Statistics.
Ashworth, J., Geys, B. & Heyndels, B. (2003). Income tax base evolution in Brussels
1980-1999: The budgetary value of the rich. In: E. Witte, A. Alen, H. Dumont, P.
Vandernoot & R. De Groof (eds.), De Brusselse negentien gemeenten en het Brussels
model.
Baller, R.D., Anselin, L., Messner, S.F., Deane, G. & Hawkins, D.F. (2001). Structural
covariates of U.S. county homicide rates: Incorporating spatial effects. Criminology 39:
561-588. https://doi.org/10.1111/j.1745-9125.2001.tb00933.x
Beenstock, M., Chiswick, B.R. & Repetto, G.L. (2001). The effect of linguistic distance
and country of origin on immigrant language skills: Application to Israel. International
Migration 39(3): 33-60. https://doi.org/10.1111/1468-2435.00155
Belsky, D.W., Moffitt, T.E., Corcoran, D.L., Domingue, B., Harrington, H., Hogan, S.,
Houts, R., Ramrakha, S., Sugden, K., Williams, B.S., Poulton, R. & Caspi, A. (2016).
The genetics of success: How single-nucleotide polymorphisms associated with
educational attainment relate to life-course development. Psychological Science 27:
957-972. https://doi.org/10.1177/0956797616643070
Boenigk, S., Mews, M. & de Kort, W. (2015). Missing minorities: Explaining low Migrant
blood donation participation and developing recruitment tactics. VOLUNTAS:
International Journal of Voluntary and Nonprofit Organizations 26(4): 1240-1260.
https://doi.org/10.1007/s11266-014-9477-7
Borjas, G.J. (2016). We Wanted Workers: Unraveling the Immigration Narrative, 1st
edition. W.W. Norton & Co.
Boyd, M. & Thomas, D. (2002). Skilled immigrant labour: Country of origin and the
occupational locations of male engineers. Canadian Studies in Population 29: 71-99.
https://doi.org/10.25336/P6X60F
KIRKEGAARD, E.O.W. & DUMOULIN, B. WHAT HAPPENED TO BRUSSELS?
317
Carl, N. (2016). Net opposition to immigrants of different nationalities correlates strongly
with their arrest rates in the UK. Open Quantitative Sociology & Political Science.
https://openpsych.net/paper/48
Carl, N. (2017). Ethnicity and electoral fraud in Britain. Electoral Studies 50: 128-136.
https://doi.org/10.1016/j.electstud.2017.09.011
Corijn, E. & van de Ven, J. (eds.) (2013). The Brussels Reader (Urban Notebooks): A
Small World City to Become the Capital of Europe. ASP.
Deblonde, N., Gheysen, T., Gysen, E., Lebeer, J., Maes, B., Schittekatte, M. & Verlinde,
L. (2020). Classificerend Diagnostisch Protocol Verstandelijke Beperking.
https://portaal.kwaliteitscentrumdiagnostiek.be/wp-
content/uploads/2020/05/CDP_Verstandelijke_Beperking_FINAAL_v25.pdf
Eff, E.A. (2004). Does Mr. Galton still have a problem? Autocorrelation in the Standard
Cross-Cultural Sample. World Cultures.
http://capone.mtsu.edu/eaeff/downloads/SAS2001.pdf
Frisell, T., Pawitan, Y. & Långström, N. (2012). Is the association between general
cognitive ability and violent crime caused by family-level confounders? PLoS ONE 7(7):
e41783. https://doi.org/10.1371/journal.pone.0041783
Fuerst, J. & Kirkegaard, E.O.W. (2016). Admixture in the Americas: Regional and
national differences. Mankind Quarterly 56: 255-373.
https://doi.org/10.46469/mq.2016.56.3.2
Gelade, G.A. (2008). The geography of IQ. Intelligence 36: 495-501.
https://doi.org/10.1016/j.intell.2008.01.004
Harrell, F.E. (2015). Regression Modeling Strategies: With Applications to Linear
Models, Logistic and Ordinal Regression, and Survival Analysis, 2nd ed. Springer.
Harrell, F.E. (2019). rms: Regression Modeling Strategies (5.1-3.1) [Computer
software]. https://CRAN.R-project.org/package=rms
Hassall, C. & Sherratt, T.N. (2011). Statistical inference and spatial patterns in
correlates of IQ. Intelligence 39: 303-310. https://doi.org/10.1016/j.intell.2011.05.001
Hegelund, E.R., Flensborg-Madsen, T., Dammeyer, J. & Mortensen, E.L. (2018). Low
IQ as a predictor of unsuccessful educational and occupational achievement: A register-
based study of 1,098,742 men in Denmark 1968-2016. Intelligence 71: 46-53.
https://doi.org/10.1016/j.intell.2018.10.002
Hegelund, E.R., Flensborg-Madsen, T., Dammeyer, J., Mortensen, L.H. & Mortensen,
E.L. (2019). The influence of familial factors on the association between IQ and
educational and occupational achievement: A sibling approach. Personality and
Individual Differences 149: 100-107. https://doi.org/10.1016/j.paid.2019.05.045
MANKIND QUARTERLY 2020 61:2
318
Herrnstein, R.J. & Murray, C.A. (1994). The Bell Curve: Intelligence and Class Structure
in American Life. Simon & Schuster.
Hertogen, J. (2017). PEW-onderzoek: % moslims in 2050 (BuG 374).
http://www.npdata.be/BuG/374-Moslims/Moslims.htm
Hill, W.D., Davies, N.M., Ritchie, S.J., Skene, N.G., Bryois, J., Bell, S., Angelantonio,
E.D., Roberts, D., … & Deary, I.J. (2019). Genome-wide analysis identifies molecular
systems and 149 genetic loci associated with income. Nature Communications 10(1): 1-
16. https://doi.org/10.1038/s41467-019-13585-5
Hu, W.-Y. (2000). Immigrant earnings assimilation: Estimates from longitudinal data.
American Economic Review 90(2): 368-372. https://doi.org/10.1257/aer.90.2.368
Husted, L., Skyt Nielsen, H., Rosholm, M. & Smith, N. (2001). Employment and wage
assimilation of male first‐generation immigrants in Denmark. International Journal of
Manpower 22(1/2): 39-71. https://doi.org/10.1108/01437720110386377
Jacobs, D., Rea, A. & Hanquinet, L. (2007). Performances des élèves issus de
l’immigration en Belgique selon l’étude PISA: Une comparaison entre la Communauté
Française et la Communauté Flamande.
Jensen, A.R. (1969). How much can we boost IQ and scholastic achievement? Harvard
Educational Review 39: 1-123.
Jerrett, M., Burnett, R., Willis, A., Krewski, D., Goldberg, M., DeLuca, P. & Finkelstein,
N. (2003). Spatial analysis of the air pollution–mortality relationship in the context of
ecologic confounders. Journal of Toxicology and Environmental Health, Part A 66:
1735-1778. https://doi.org/10.1080/15287390306438
Jones, G. & Schneider, W.J. (2010). IQ in the production function: Evidence from
immigrant earnings. Economic Inquiry 48: 743-755. https://doi.org/10.1111/j.1465-
7295.2008.00206.x
Junger, M. & Polder, W. (1992). Some explanations of crime among four ethnic groups
in the Netherlands. Journal of Quantitative Criminology 8: 51-78.
https://doi.org/10.1007/BF01062759
Kaufman, E. (2019). Whiteshift: Populism, Immigration and the Future of White
Majorities. Penguin Books.
Kirkegaard, E.O.W. (2014a). The international general socioeconomic factor: Factor
analyzing international rankings. Open Differential Psychology.
https://doi.org/10.26775/ODP.2014.09.08
Kirkegaard, E.O.W. (2014b). Crime, income, educational attainment and employment
among immigrant groups in Norway and Finland. Open Differential Psychology.
http://openpsych.net/ODP/2014/10/crime-income-educational-attainment-and-
employment-among-immigrant-groups-in-norway-and-finland/
KIRKEGAARD, E.O.W. & DUMOULIN, B. WHAT HAPPENED TO BRUSSELS?
319
Kirkegaard, E.O.W. (2015a). Crime among Dutch immigrant groups is predictable from
country-level variables. Open Differential Psychology.
Kirkegaard, E.O.W. (2015b). Some methods for measuring and correcting for spatial
autocorrelation. The Winnower. https://thewinnower.com/papers/2847-some-methods-
for-measuring-and-correcting-for-spatial-autocorrelation
Kirkegaard, E.O.W. (2016a). Inequality across US counties: An S factor analysis. Open
Quantitative Sociology & Political Science.
http://openpsych.net/OQSPS/2016/05/inequality-across-us-counties-an-s-factor-
analysis/
Kirkegaard, E.O.W. (2016b). Some new methods for exploratory factor analysis of
socioeconomic data. Open Quantitative Sociology & Political Science 1(1).
https://openpsych.net/paper/47
Kirkegaard, E.O.W. & Becker, D. (2017). Immigrant crime in Germany 2012-2015. Open
Quantitative Sociology & Political Science 1(1).
https://doi.org/10.26775/OQSPS.2017.02.11
Kirkegaard, E.O.W. & de Kuijper, M. (2020). Public Preferences and Reality: Crime
Rates among 70 Immigrant Groups in the Netherlands. Mankind Quarterly 60: 327-351.
https://doi.org/10.46469/mq.2020.60.3.3
Kirkegaard, E.O.W. & Fuerst, J. (2014). Educational attainment, income, use of social
benefits, crime rate and the general socioeconomic factor among 70 immigrant groups
in Denmark. Open Differential Psychology. https://openpsych.net/paper/21
Kirkegaard, E.O.W. & Fuerst, J. (2017). Admixture in Argentina. Mankind Quarterly 57:
542-580. http://mankindquarterly.org/archive/issue/57-4/4
Klein, O., Pohl, S. & Ndagijimana, C. (2007). The influence of intergroup comparisons
on Africans’ intelligence test performance in a job selection context. Journal of
Psychology 141: 453-468. https://doi.org/10.3200/JRLP.141.5.453-468
Kühn, I. (2007). Incorporating spatial autocorrelation may invert observed patterns.
Diversity and Distributions 13: 66-69. https://doi.org/10.1111/j.1472-4642.2006.00293.x
Kuran, T. (1997). Islam and underdevelopment: An old puzzle revisited. Journal of
Institutional and Theoretical Economics 153: 41-71.
Lafontaine, D., Crépin, F. & Quittre, V. (2017). Les compétences des jeunes de 15 ans
en sciences, en mathématiques et en lecture. Résultats de l’enquête PISA 2015 en
Fédération Wallonie-Bruxelles.
Lee, J.J., Wedow, R., Okbay, A., Kong, E., Maghzian, O., Zacher, M., Nguyen-Viet,
T.A., Bowers, P., … & Cesarini, D. (2018). Gene discovery and polygenic prediction
from a 1.1-million-person GWAS of educational attainment. Nature Genetics 50: 1112-
1121. https://doi.org/10.1038/s41588-018-0147-3
MANKIND QUARTERLY 2020 61:2
320
Li, T.J.-J., Sen, S. & Hecht, B. (2014). Leveraging advances in natural language
processing to better understand Tobler’s first law of geography. Proceedings of the
22nd ACM SIGSPATIAL International Conference on Advances in Geographic
Information Systems - SIGSPATIAL ’14, 513–516.
https://doi.org/10.1145/2666310.2666493
Loftin, C. (1986). Assaultive violence as a contagious social process. Bulletin of the New
York Academy of Medicine 62: 550-555.
Lynn, R. (2020). Racial and ethnic differences in altruism assessed by blood donation.
Mankind Quarterly 60: 539-550. https://doi.org/10.46469/mq.2020.60.4.6
Mennis, J. & Harris, P. (2013). Spatial contagion of male juvenile drug offending across
socioeconomically homogeneous neighborhoods. In: M. Leitner (ed.), Crime Modeling
and Mapping Using Geospatial Technologies, pp. 227-248. Springer Netherlands.
https://doi.org/10.1007/978-94-007-4997-9_10
Metzen, D. (2012). The causes of group differences in intelligence studied using the
method of correlated vectors and psychometric meta-analysis. Masters Thesis, Univ. of
Amsterdam.
Murray, C. (2002). IQ and income inequality in a sample of sibling pairs from
advantaged family backgrounds. American Economic Review 92: 339-343.
Murray, D. (2017). The Strange Death of Europe: Immigration, Identity, Islam, 1st
edition. Bloomsbury Continuum.
Nordin, M. & Rooth, D.-O. (2007). Income gap between natives and second generation
immigrants in Sweden: Is skill the explanation? (SSRN Scholarly Paper ID 984423).
Social Science Research Network. https://papers.ssrn.com/abstract=984423
Osili, U.O. & Paulson, A. (2008). What can we learn about financial access from U.S.
immigrants? The role of country of origin institutions and immigrant beliefs. World Bank
Economic Review 22: 431-455. https://doi.org/10.1093/wber/lhn019
Osorio, J. (2015). The contagion of drug violence: Spatiotemporal dynamics of the
Mexican war on drugs. Journal of Conflict Resolution 59: 1403-1432.
https://doi.org/10.1177/0022002715587048
Pesta, B.J., Fuerst, J.G.R., Lasker, J. & Kirkegaard, E.O.W. (in review). County-level
USA: No robust relationship between geoclimatic variables and intelligence.
Pesta, B.J., McDaniel, M.A. & Bertsch, S. (2010). Toward an index of well-being for the
fifty U.S. states. Intelligence 38: 160-168. https://doi.org/10.1016/j.intell.2009.09.006
Pew Research Center (2011). Table: Muslim Population by Country. Pew Research
Center. http://www.pewforum.org/2011/01/27/table-muslim-population-by-country/
Revelle, W. (2020). psych: Procedures for Psychological, Psychometric, and Personality
Research (1.9.12.31) [Computer software]. https://CRAN.R-project.org/package=psych
KIRKEGAARD, E.O.W. & DUMOULIN, B. WHAT HAPPENED TO BRUSSELS?
321
Rindermann, H. & Thompson, J. (2016). The cognitive competences of immigrant and
native students across the world: An analysis of gaps, possible causes and impact.
Journal of Biosocial Science 48: 66-93. https://doi.org/10.1017/S0021932014000480
Rohrer, J.M. (2018). Thinking clearly about correlations and causation: Graphical causal
models for observational data. Advances in Methods and Practices in Psychological
Science 1(1): 27-42. https://doi.org/10.1177/2515245917745629
Roth, B.M. (2010). The Perils of Diversity: Immigration and Human Nature, 1st edition.
Washington Summit.
Rushton, J.P. (1998). The “Jensen effect” and the “Spearman-Jensen hypothesis” of
Black-White IQ differences. Intelligence 26: 217-225. https://doi.org/10.1016/S0160-
2896(99)80004-X
Rushton, J.P. & Jensen, A.R. (2010). The rise and fall of the Flynn effect as a reason to
expect a narrowing of the Black–White IQ gap. Intelligence 38: 213-219.
https://doi.org/10.1016/j.intell.2009.12.002
Sanandaji, T. (2017). Massutmaning: Ekonomisk politik mot utanförskap och antisocialt
beteende. Kuhzad Media.
Sarrazin, T. (2012). Deutschland schafft sich ab: Wie wir unser Land aufs Spiel setzen,
1. Aufl. Dt. Verl.-Anst.
Schulz, J.F., Bahrami-Rad, D., Beauchamp, J.P. & Henrich, J. (2019). The Church,
intensive kinship, and global psychological variation. Science 366: 707.
https://doi.org/10.1126/science.aau5141
Skardhamar, T., Aaltonen, M. & Lehti, M. (2014). Immigrant crime in Norway and
Finland. Journal of Scandinavian Studies in Criminology and Crime Prevention 15: 107-
127. https://doi.org/10.1080/14043858.2014.926062
Tobler, W. (2004). On the First Law of Geography: A reply. Annals of the Association of
American Geographers 94: 304-310. https://doi.org/10.1111/j.1467-
8306.2004.09402009.x
Van Hamme, G. (2015). Inégalités et flux entre Bruxelles et sa périphérie.
https://docplayer.fr/62414320-Inegalites-et-flux-entre-bruxelles-et-sa-peripherie.html
Vinogradov, E. & Kolvereid, L. (2010). Home country national intelligence and self-
employment rates among immigrants in Norway. Intelligence 38: 151-159.
https://doi.org/10.1016/j.intell.2009.09.004
Woodley, M.A. & Bell, E. (2013). Consanguinity as a major predictor of levels of
democracy. A study of 70 nations. Journal of Cross-Cultural Psychology 44: 263-280.
https://doi.org/10.1177/0022022112443855
MANKIND QUARTERLY 2020 61:2
322
Yeargin-Allsopp, M., Drews, C.D., Decouflé, P. & Murphy, C.C. (1995). Mild mental
retardation in black and white children in metropolitan Atlanta: A case-control study.
American Journal of Public Health 85: 324-328. https://doi.org/10.2105/ajph.85.3.324
Appendix
Spatial units merged
One can see the old NIS on Esperanto Wiki (click the change language
function)
Nevele (44049) was merged into Deinze (44083)
Knesselare (44029) was merged into Aalter (44084)
Effective 1 January 2019, Waarschoot (44072), Lovendegem (44036)
and Zomergem (44085) were merged into the new municipality of
Lievegem (44085).
Klein et al. (2007) calculation
David Becker suggested the following calculation (email Jan 30, 2018). The
context is that Richard Lynn calculated an IQ of 70 for this in his Race Differences
in Intelligence book. He writes as a comment:
I only have German norms of the CFT-20 and CFT-3. Don't know which CFT
they have used and if I could apply norms from the German version on raw scores
from the other. But let's try it:
If assumed mean age is 32.30y and mean score is 19.94, CFT-20 would give
an IQ of 81 (10th German percentile). CFT-3 doesn't have norms for >19y olds.
Year of standardization is 1977, the year of measurement is ~2006 (or
earlier). Pietschnig & Voracek (2015) calculated a FLynn-Effect/y at the CFT-20
in Germany of 0.6 between 1977 and 1995. Therefore, the IQ inflation would be
between 17.4 (29y from 1977 to 2006) and 10.8 (18y from 1977 to 1995). Second
value is more likely because no FE-data is available in this case after 1995.
In addition, 1.20 IQ must be deducted because norms are from Germany and
Germany is 1.20 IQ below the UK (this part of the method is questionable).
Overall, 12 IQ (10.8+1.20) must be deducted from the score of 81, which would
give the sample a final IQ of 69 or 70.2 without country-correction. This is very
close to Richard's 70.
National foreign aid rates, % of GNI
These were computed based on OECD data (n = 44) at
https://data.oecd.org/oda/net-oda.htm. Country-years with missing data were
filled with 0%, i.e. no aid that year. The table below shows the averages across
KIRKEGAARD, E.O.W. & DUMOULIN, B. WHAT HAPPENED TO BRUSSELS?
323
years. In the last 10 years, a few Muslim countries (United Arab Emirates and
Turkey) have started donating.
Rank
Country ISO-3
1960-2017
Country ISO-3
2008-2017
1
NOR
0.78
SWE
1.05
2
SWE
0.75
NOR
1.01
3
NLD
0.74
LUX
1.00
4
DNK
0.70
DNK
0.83
5
FRA
0.55
NLD
0.72
6
BEL
0.49
ARE
0.69
7
GBR
0.42
GBR
0.61
8
AUS
0.40
FIN
0.51
9
DEU
0.38
BEL
0.50
10
LUX
0.38
CHE
0.46
11
CAN
0.35
DEU
0.46
12
DAC
0.33
IRL
0.44
13
FIN
0.32
LIE
0.44
14
CHE
0.28
FRA
0.42
15
NZL
0.26
TUR
0.39
16
USA
0.25
AUT
0.32
17
JPN
0.24
AUS
0.31
18
IRL
0.23
DAC
0.30
19
AUT
0.22
CAN
0.29
20
ITA
0.20
ESP
0.27
21
ESP
0.14
NZL
0.27
22
PRT
0.14
ISL
0.26
23
ARE
0.12
PRT
0.23
24
TUR
0.09
ITA
0.20
25
LIE
0.08
JPN
0.20
26
ISL
0.07
USA
0.19
27
GRC
0.06
MLT
0.18
28
KOR
0.04
GRC
0.15
29
CZE
0.03
SVN
0.14
30
MLT
0.03
EST
0.13
31
SVN
0.03
KOR
0.13
32
EST
0.03
CZE
0.13
33
CYP
0.03
LTU
0.12
34
HUN
0.03
CYP
0.11
35
LTU
0.03
HUN
0.11
36
SVK
0.02
SVK
0.10
37
ISR
0.02
POL
0.10
38
POL
0.02
ROU
0.09
39
TWN
0.02
ISR
0.08
40
LVA
0.02
LVA
0.08
41
ROU
0.02
BGR
0.08
42
BGR
0.01
TWN
0.08
43
RUS
0.01
RUS
0.04
44
THA
0.01
THA
0.02
