of individual attitudes toward immigration,19 we also control for it by using
a dummy variable that is equal to 1 if a respondent was unemployed and
wanting a job in the last seven days and 0 otherwise.20 We call this variable
unemploy.
Finally, we control for race/ethnicity-based discrimination attitudes. To
do this, we use the responses to the following two questions:
• To what extent do you think [the country where the respondent was
questioned] should allow people of the same race or ethnic group as
most of the people in the country to come and live here?
• To what extent do you think [the country where the respondent was
questioned] should allow people of a race or ethnic group different from
most of the people in the country to come and live here?
These questions ask the same except the race or ethnic group of migrants.
Therefore, any difference between the responses within a respondent should
pick up his/her discrimination based on race or ethnicity. For each question,
a response is one of “none (coded as 0)”, “a few (1)”, “some (2)” or “many
(3)”. We subtract the response to the second question from that to the first
question for each respondent. Hence the higher the number we obtain, the
more discriminating against different races and ethnic groups.21 We call this
variable racist.
19See for example O’Rourke and Sinnott (2006).
20We included both those who indicated that they were actively looking for a job
(uempla =1 in ESS Question F8a) and those who indicated that they were not actively
jobhunting but wanting a job (uempli =1in the same question). We are not primarily
interested in attitudes of the unemployed, but we include this variable to increase the
goodness of fit.
21This way of taking the difference is inappropriate, for the response categories are only
ordered. In other words, the magnitude of an estimated coefficient is not meaningful. Our
14