Last spring when I heard that we (the Swiss) would have to vote over the anti-minaret, I was surprised by such an initiative, because I believed that such a stupid proposal would not even be supported by all members of the Swiss People’s Party (UDC/SVP). Obviously, I was wrong. Although the latest polls still predict a stable majority of 53% will reject the initiative, the proportion of respondents in favor of a constitutional ban of minaret construction has grown to more than a third (35%).
Anyway, as I was working with Swiss data last spring, I made a little exercise that is not really original. I computed the probability to be employed in Switzerland with usual controls (age, number of years of education, civil status, type of permit, number of children, dummies for the linguistic region, type of communes and employment region) and dummies for region of origin for migrants of the first and second generation. Then I did the same thing, but with the additional distinction between Muslims and non-Muslims… the marginal effects of the dummies are plotted in the graphs below (sorry if they are in French, but that should not be a problem in our interdisciplinary and multikulti Institute :). Of course, the marginal effects are not equivalent to discrimination, but I don’t see a plausible alternative explanation for why all points are below the 45 degrees line. Any idea? Can Swiss employers have the benefit of the doubt?
The data used is from the census 2000. There is no issue of sample selection. The data is a little old, but after 9/11, it would be doubtful to claim that Muslims are enjoying specific sympathy in the Western countries…
9 comments:
whats the difference between the two sets of graphs?
Think, the title of your post is a bit exaggerated.
I'm no expert on these things, but maybe there is an above average crime rate prevalence among muslims (thinking about those young guys coming from france...).
Criminal hisotry clearly has a negative impact on your chances of getting a job. This will get picked up by the muslim dummy. (and you'll incorrectly conclude discrimination).
maybe "standard" controls are not good enough...
The model is linear?
If so:
For instance # of children might not enter linearly (having 4 children has more than twice the negative effect of having two children - actually think that's definitely the case). Muslim mothers have more children, the linear model does not pick up everything, part of the effect is being attributed to the muslim dummy, leading to its overestimation...
TO sum up, I think there is some level of discrimination, but also think your model overestimates it...
Hi Katya,
I did two regressions on the male and female population. The marginal effects for the 1st and 2nd generation (left and right graphs) are computed using the Swiss natives as reference group. As you can see, the differences are larger for the 1st than for the 2nd generation, and for men than for women.
Hi Tomas,
Sorry I forgot to put the question mark in the title... basically, I reproduced the results of published study (Lorenz Widmer, "Population étrangère et chômage", 2005, OFS), adding the religious dimension. The model is linear. Of course, standard controls are never satisfying, but that's all that there is in the census.
I don't completely buy your arguments in the sense that the sample is limited to the active population, and women with more children would generally decide to stay at home (and be part of the inactive population). In the data, women that have most children are from the TurkeyMiddleeastMaghreb group and they don't display the largest difference. Anyway, this could be tested.
For the criminal records, I think the argument kinds of illustrates how discrimination might work: "Because the young French guys from Arab origin that come to Geneva on Friday night have a higher probability of having a criminal record, Muslims in general have a higher probability of being criminals". First, the young French guys that come from Lyon or other French cities are not part of the census data, and if I look at drug dealers in Geneva they rather look like Christians guys from West Arica than like Muslims from Ethiopia, and I am not sure there is a point in asserting the average crime rate is more prevalent among muslims. There is also the problem of seeing Muslims as criminals and/or religious rigorists if not fundamentalists, as if they would suffer from community schizophrenia.
My point is rather, that a guy who thinks (1) "Well, Switzerland is a Christian country, and I think it is natural to keep it this way; therefore, let's ban minaret construction", this guy, when confronted to the choice of hiring a guy he thinks/knows is Muslim and another non-Muslims guy will be inclined to say (2) "There are enough Muslims in Switzerland, let's not encourage them to come and hire the non-Muslim one". This is the soft version... Apparently there are quite a few people in Switzerland who support who adopt the first part of the statement, and my guess is most of them would also adopt the second statement, which would explain the differentials. But of course, not everything is in the data, and you might be right. Thanks a lot for the comments.
Sorry Muslims from Somalia... not Ethiopia.
Hey Pierre,
Women: I don't think that women who has four children is confined to stay home for the rest of her life. Eventually, she will search for a job too. Having four children in the past is more than twice as bad for her chances of getting employed as having two children. (Children don't enter linearly, think it's the case...)
Since muslim women tend to have more children, this negative effect will get picked up by the dummy (repeating myslef :-)
The crime rates: Was first thing that came to my mind and you are probably absolutely right that muslims living in CH will not have higher crime rates than is the average in the pop.
I think the story of young 2nd generation french immigrants you are describing is exactly right.
The resulting "discrimination" is a natural behavioral bias. If U see young arabic guy trying to steal someone's purse (like me this summer)+ your friends tell you similar story you'll conclude that there is a high crime prevalence in that population.
A lot of people in the general public can't make a distinction between arabic guy that come to geneva from france vs. a muslim guy who actually lives in geneva. Wrongly attributing the crimes to muslim from geneva. But it ain't DISCRIMINATION it's just a mistake we make!
Even people who know that the crime they saw was very likely committed by a french guy will "discriminate" young swiss muslims.
If a young swiss muslim tries to get a job the employer will extrapolate the experience he had with the french guy. The french muslim seems to be in to some extent representative of the swiss muslim. french muslims in geneva steals=>the swiss one is likely to steal...
(this is done automatically with no conscious control over the thought process)
I'm not saying this is the right way to do things. I'm just saying that discrimination can partly result from behavioral biases over which people have NO or very limited CONTROL...
Moreover, are the standard covariates sufficient covariates?
Plus I think there actually is some conscious discrimination (Way smaller than the data would lead you to conclude though...)
Hi Tomas,
The inclusion of a squared term for the number of children doesn't change the value of the dummies coefficients plotted in the graphs, although the coefficent is significant (negative for men, positive for women). The squared term smoothens the effect related to the number of children (positive for men, negative for women) rather than amplifying it. If you want I can send you the details.
"Unconscious behavioral biases" might explain some of the discrimination, but for me it doesn't really matter if discrimination is conscious/deliberate or unconscious. This is not a legal exercise, and discrimination can have a broader meaning. What matters is the result: lower probability to be employed, other things equal. Every third Swiss who will vote in favor of a popular initiative to ban minaret in the constitution will do so consciously (most of them would not say they are making an act of discrimination, but it is not because they don't perceive it that way that the popular initiative is not discriminating). So it doesn't surprise me if this ends up in a higher employment penalty for Muslims.
Pierre
hey Pierre,
conscious vs. unconscious discrimination, are they the same? I guess that's a philosophical question... In my opinion, if it's unconscious its only implication is loss of efficiency, if it's conscious it also implies that something is wrong with the society, that's far worse...
I agree, the proposed ban is ridiculous...
TH
After the vote (58% in favor of banning minarets), Le Temps proposed I present and analyze the results. They only had room for a massacred version of the first graph and cut two footnotes... The article is available online:
http://www.letemps.ch/Facet/print/Uuid/0b384098-e442-11de-8d9d-6e1e9c52b424/Les_signes_avant-coureurs_de_lislamophobie_en_Suisse
http://www.letemps.ch/Page/Uuid/0b384098-e442-11de-8d9d-6e1e9c52b424/Les_signes_avant-coureurs_de_lislamophobie_en_Suisse
Post a Comment