We agree that job satisfaction is a multidimensional construct, and we measured several features of job-related satisfaction in addition to satisfaction with one's job itself. Although the measures suggested by Dr Kader would have been appropriate, some of their items overlapped with the Karasek Job Content Questionnaire (Reference KarasekKarasek, 1979), and we were keen to avoid such duplication and overburdening respondents. As we were interested in the relationship between satisfaction with one's job and other indicators of job-related satisfaction such as feelings about pay, operational and policy contexts (which were and remain topical because of proposed changes to the Mental Health Act 1983) and feeling valued, it would have been inappropriate to use a multi-faceted job satisfaction scale as a dependent variable. All of the scales used in the survey are well known and have established reliability and validity.

The adjusted response rate of 49%, although low in comparison with experimental studies, is very reasonable for social surveys of this type. We agree that it would have been helpful to know how non-respondents compared with respondents in terms of demographic and other details, but the methodology meant that was not achievable. Nevertheless, we do know that our sample was very similar, demographically and in terms of tenure, length of experience, approved social worker status, etc., to another recent study of mental health social workers (ADSS Cymru, 2005). Therefore we have no reason to believe that these data are not representative.

Finally, although it might have been interesting to present a stepwise regression model, we opted for an ‘enter’ model in the interests of brevity. Subsequent analyses have shown that a stepwise approach offers little added value.

Like Dr Kader, we hope that the results of our survey are an eye-opener for employers.