districts and wards. For the 2003 elections, the BNP’s percentage of the white vote
appears to increase in association with the relative size of a district’s
Pakistani/Bangladeshi population, as predicted by racial threat theory. However, none of
the other ethnicity variables appears to have much of an effect on the BNP’s vote share,
as none of the other estimated coefficients are statistically signficant.
Model 4 adds independent variables that measure aspects of the socio-economic
composition of the districts and wards. Because the number of observations is so small,
only the variables that were found to have the greatest effects were included in the final
specification this model. Once these other variables are included, the percent
Pakistani/Bangladeshi in a district no longer has a statistically significant effect on the
BNP’s percentage of the white vote in 2003, and its estimated coefficient becomes
negative. Similarly, the estimated effect of the relative size of a district’s Indian
population, which displayed a weak, positive relationship to the BNP’s vote share in
Model 3 for 2003, is found to have a statistically significant, negative effect on the BNP
vote in Model 4. None of the other ethnic composition variables is found to have a
significant effect.
The difference in the estimated effects of the percent Indian and percent
Pakistani/Bangladeshi on the BNP’s 2003 vote share between Models 3 and 4 is probably
due to the inclusion of the percentage of schoolchildren who are not native English-
speakers in the latter model. The BNP appears to have been most successful in 2003 in
those districts that are most culturally diverse, though this variable does not appear to
have any effect on the 2002 results. Also at the district level, the percentage of the local
workforce employed in the manufacturing sector has a strong positive relationship with
the BNP’s vote share. Districts dominated by manufacturing seem to be fertile territory
29
Convention