TY - JOUR
T1 - The effects of language bias and cultural bias estimated using the method of correlated vectors on a large database of IQ comparisons between native Dutch and ethnic minority immigrants from non-Western countries
AU - te Nijenhuis, J.
AU - Willigers, D.
AU - Dragt, J.
AU - van der Flier, H.
PY - 2015
Y1 - 2015
N2 - Cultural factors and language bias are often used to explain the large differences in mean intelligence test scores between ethnic groups, for instance majority Dutch compared to ethnic minorities. A competing explanation comes from Spearman's hypothesis, which states that these group differences in intelligence scores are a function of the cognitive complexity of the intelligence tasks, and that the regression of group differences on IQ subtests (d) on g loadings of subtests (g) leads to a regression line with only a modest amount of scatter. However, when testing Spearman's hypothesis there are two findings that have to be taken into consideration. First, subtests of an IQ battery with a substantial language component partially measure the extent of proficiency of ethnic minorities in the Dutch language, therefore underestimate the level of g of the tested nonnative speakers, which negatively influences the hypothesized correlation between d and g. Second, ethnic minorities score relatively lower on subtests of spatial visualization and relatively better on subtests of short-term memory, and having these 'outliers' in an IQ battery also attenuates the outcome of tests of Spearman's hypothesis, and additionally tilts the regression line of d on g. Spearman's hypothesis was tested using 40 datasets, comparing native Dutch to Turkish, Moroccan, Netherlands Antillean, and Surinamese immigrants. We expected to find a strong positive correlation between group differences on the subtests of an IQ battery on the one hand, and the vector of g loadings on the other hand. We then estimated the influence of language bias and the influence of cultural bias, expecting at most small effects.First, we applied bare-bones psychometric meta-analytical techniques to all Dutch datasets in our non-meta-analytical database to test Spearman's hypothesis. Second, the effect of language bias was estimated by measuring the difference of subtests with a substantial language component to the regression line of d on g. Third, the effect of cultural bias was estimated by leaving out both language-biased subtests and the two kinds of 'outliers', computing the regression line of d on g on the remaining subtests, and checking where that more stable regression line intersected the y-axis. Fourth, we checked for the presence of publication bias, among others, by comparing the outcomes of published studies with the outcomes of unpublished studies, and by using standard publication bias analyses.First, Spearman's hypothesis is not so strongly supported in the present study with a sample size-weighted mean of only r= .32 (K= 40, total N= 5504). Leaving out the language-biased subtests substantially increased the value of the correlation between group differences and g loadings to r= .43. Second, a mean underestimation was found of 1.63 IQ points for all studies due to language-biased subtests. This is a small effect. Third, there was no effect of cultural bias as measured by examining the distance of the regression line from the zero point before and after removing subtests on which ethnic minorities tend to perform relatively higher or lower than native Dutch. Fourth, there is no clear proof of publication bias concerning confirmation of Spearman's hypothesis.It is concluded that mean differences in intelligence between native Dutch and ethnic minority immigrants from non-Western/non-Westernized countries in the Netherlands can be very strongly explained by the complexity of the subtests in an IQ battery, with a small effect of language bias, and very little place left for cultural bias.
AB - Cultural factors and language bias are often used to explain the large differences in mean intelligence test scores between ethnic groups, for instance majority Dutch compared to ethnic minorities. A competing explanation comes from Spearman's hypothesis, which states that these group differences in intelligence scores are a function of the cognitive complexity of the intelligence tasks, and that the regression of group differences on IQ subtests (d) on g loadings of subtests (g) leads to a regression line with only a modest amount of scatter. However, when testing Spearman's hypothesis there are two findings that have to be taken into consideration. First, subtests of an IQ battery with a substantial language component partially measure the extent of proficiency of ethnic minorities in the Dutch language, therefore underestimate the level of g of the tested nonnative speakers, which negatively influences the hypothesized correlation between d and g. Second, ethnic minorities score relatively lower on subtests of spatial visualization and relatively better on subtests of short-term memory, and having these 'outliers' in an IQ battery also attenuates the outcome of tests of Spearman's hypothesis, and additionally tilts the regression line of d on g. Spearman's hypothesis was tested using 40 datasets, comparing native Dutch to Turkish, Moroccan, Netherlands Antillean, and Surinamese immigrants. We expected to find a strong positive correlation between group differences on the subtests of an IQ battery on the one hand, and the vector of g loadings on the other hand. We then estimated the influence of language bias and the influence of cultural bias, expecting at most small effects.First, we applied bare-bones psychometric meta-analytical techniques to all Dutch datasets in our non-meta-analytical database to test Spearman's hypothesis. Second, the effect of language bias was estimated by measuring the difference of subtests with a substantial language component to the regression line of d on g. Third, the effect of cultural bias was estimated by leaving out both language-biased subtests and the two kinds of 'outliers', computing the regression line of d on g on the remaining subtests, and checking where that more stable regression line intersected the y-axis. Fourth, we checked for the presence of publication bias, among others, by comparing the outcomes of published studies with the outcomes of unpublished studies, and by using standard publication bias analyses.First, Spearman's hypothesis is not so strongly supported in the present study with a sample size-weighted mean of only r= .32 (K= 40, total N= 5504). Leaving out the language-biased subtests substantially increased the value of the correlation between group differences and g loadings to r= .43. Second, a mean underestimation was found of 1.63 IQ points for all studies due to language-biased subtests. This is a small effect. Third, there was no effect of cultural bias as measured by examining the distance of the regression line from the zero point before and after removing subtests on which ethnic minorities tend to perform relatively higher or lower than native Dutch. Fourth, there is no clear proof of publication bias concerning confirmation of Spearman's hypothesis.It is concluded that mean differences in intelligence between native Dutch and ethnic minority immigrants from non-Western/non-Westernized countries in the Netherlands can be very strongly explained by the complexity of the subtests in an IQ battery, with a small effect of language bias, and very little place left for cultural bias.
U2 - 10.1016/j.intell.2015.12.003
DO - 10.1016/j.intell.2015.12.003
M3 - Article
SN - 0160-2896
VL - 2016
SP - 117
EP - 135
JO - Intelligence
JF - Intelligence
IS - 54
ER -