Dr Topiwala raises two additional points which we have discussed in our paper. The original EMPIRIC survey omitted a number of ethnic minority groups. ^{Reference Sproston and Nazroo1} We were therefore unable to analyse associations for these groups as we did not have the data. As we mention in the ‘Limitations’ section of our paper, ^{Reference Das-Munshi, Bécares, Boydell, Dewey, Morgan and Stansfeld2} since this survey (2002) there have been a number of other new migrant groups to Britain, for example from Eastern Europe. Therefore we caution that the findings should not be generalised outside of the ethnic minority groups surveyed in our study.

The issue of non-response has been discussed in the Method under ‘Survey design’. ^{Reference Das-Munshi, Bécares, Boydell, Dewey, Morgan and Stansfeld2} Weights to account for non-response bias were derived using stepwise logistic regression techniques using data which were available from previous Health Surveys for England. ^{Reference Sproston and Nazroo1} A number of important demographic predictors of non-response were included in the models (including individual and household predictors as well as primary sampling units). ^{Reference Sproston and Nazroo1} Interactions by ethnicity were also included in the models, in order to account for the differing probabilities of response by ethnic group. ^{Reference Sproston and Nazroo1} As mentioned in our paper, wherever possible we used these survey weights to correct for non-response in our analyses. ^{Reference Das-Munshi, Bécares, Boydell, Dewey, Morgan and Stansfeld2}