Contested ethnic categorisations in policy research
Is it time to abandon the use of ethnic categories in policy research?
Statistics according to ethnic background
‘Labour market participation of women with a Surinamese background is higher than for women with a Turkish background’. Dutch media consumers are used to reading such findings. There is a tradition of gathering this kind of information, which is broken down according to broad ethnic categories in order to develop social policies and to monitor their implementation.
Increasingly, this categorisation practice is coming under review. It no longer fits the complex realities in society. Urban populations, in particular, have become highly diverse. Rather than identifiable migrant groups, populations are now far more complex and layered. These changes have an impact on self-identification; young people feel pushed into categories that have very little to do with how they see themselves. These changes also have legal and political effects. Migrants with full residence rights are living alongside migrants who have a very different legal status. Moreover, mobility patterns have become more fluid. All these developments give rise to the question: is it time to get rid of the current practice of ethnic categorisation in policy research? I was recently invited by one the Dutch ministries to participate in a dialogue on this issue and was inspired by our findings to write this blog.
Reasons to continue
Keeping track of the social position of inhabitants is part and parcel of a welfare state that wants to fight inequalities. It is telling that the European Commission against Racism and Intolerance (ECRI), for instance, has been advocating the collection of ‘ethnic data’ in order to assess the effectiveness of policies targeting ethnic minority groups. This data could be based on self-identification, but that methodology is much more sensitive to changes in societal perceptions which would result in unreliable comparisons over time. Abandoning ethnic categorisation might lead to the same issues being studied but with less reliable data. Social scientists would then perhaps come up with proxies (using surnames etc.) as is done in France, where ethnic categorisation is deemed unacceptable. But this only makes matters worse.
Reasons to change
First, inequalities that are now presented along the lines of ethnic categories are related to gender, age, social-economic status, and neighbourhood etc. There is often very little added value from ethnic background variables. Second, most policies do not actually differentiate per ethnic group. A decade ago, the Dutch government decided to mainstream policies and put an end to specifically targeted policies for certain categories of people. And even before then, very few policy measures were specifically adapted or targeted. Finally, categories that were originally designed to fight inequalities can actually result in discrimination. This is particularly true for crime statistics. Given the paucity of reliable information on offending rates, the police may rely on their own prior statistics in deciding priorities when allocating resources. This, in turn, may accelerate the imbalance in the correlation between ethnic background and registered criminality. In this respect, ethnic categorisation and ethnic profiling feed one another. This implies that crime statistics are not simply comparable to labour market statistics.
During the dialogue we looked at the issue from various perspectives and came to three conclusions. First, we agreed that routinely producing and processing data on the ethnicity of the population raises serious issues, in particular with respect to crime statistics. There must be a theoretical reason (at least common sense) to assume that ethnic background is relevant when deciding whether ethnic categories are needed, but this is often lacking. Information on educational level or socio-economic status may be more difficult to get hold of than readily available ethnic categorisations. Second, in many cases presenting descriptive statistics of group differences is distorting. In particular when using crime statistics, agencies and research institutes should put more energy into further analysis including controls for other relevant background characteristics (as increasingly appears to be done). This will more aptly elucidate patterns in social outcomes. A third suggestion was to have the use of ethnic categories in policy research periodically reviewed by an external committee. Mixed committees, also comprising younger people who view these practices with a fresh perspective, could critically assess whether there are sufficient and substantiated reasons to use these categories in the way and for the aims they are used.
If the bold move of abandoning the use of ethnic categorisation in Dutch policy research seems a step too far, these steps might help avoid spurious interpretations of effects of ethnic categories and make researchers more aware of the problematic sides of their research. Making use of these categories should never be routine procedure; a sense of responsibility is required, since political and societal framings are inevitably linked to the statistical categories that support them.