A Theory of Parties and Electoral Systems

Appendix A. Sampling

APPENDIX A Sampling SAMPLING for the questionnaires used in this study was based on constituency, In all three countries, stratified random samples of constituencies, weighted according to the number of deputies returned, were drawn. The sampling strata were regions, In Great Britain, the regions were London, Englandother than London, Scotland, and Wales; Northern Ireland was excluded In Eire, the sampling strata were the four historic provinces of Ireland plus the borough of Dublin, The Italianstrata werethe North ofItaly, the Center, andthe Mezzogiorno, with Sicily, Sardegna, the Valle d'Aosta, and Trieste eliminated. The total British sample size was sixty constituencies. An attempt was made to send questionnaires to (or obtain interviews with)the candidates and agents of both major parties at the timeof the 1970 generalelection in each constituency selected. (While either a Labour or a Conservative candidate was elected from each of the sampled constituencies, in some cases the candidate of the other major party finished third,) Questionnaires for the defeated candidates and agents were sent to the constituency party headquarters, and in some cases were returned with the notation that the person to whomtheywere addressed had left the constituency with no forwarding address. This reduced the original sampleof 240 individualsto 196, Eventually 110 interviewsor completed questionnaires wereobtained foran overall response rate of 56 percent of the reduced sample. The Irish sample consisted of seven constituencies, including twenty-four deputies of Fianna Fail and Fine Gael, or about one-fifth of the total, as opposed to just under one-tenth in Britain, An attempt was made to interview personally all deputies of the two parties elected from these constituencies. Questionnaires were sent to the defeated candidates of the two parties at the addresses listed on their nominating forms. Twenty such questionnaires were sent Since the parties refused to provide the names and addresses of local party officials, each responding deputy and candidate was asked to supply the name and address of one branch secretary from his own constituency party. In all, fourteen did so, yeilding a total potential sample of fifty-eight Of these, twenty-eightresponded, giving a total response rate of 48 percent There werenineconstituencies inthe Italiansample. Questionnaires weresent to all PSI deputies, representing just under one-third of the total number of PSI members of the Chamber of Deputies, and to an equal number of DC deputies randomly selected from the same constituencies with samplingbased on factional affiliation. Questionnaires also were sent to one defeated candidate (when an address could be obtained) and one federation secretary of each party in each 725 126 Appendix A constituency. In all, seventy-eight questionnaires weresent Due tothe chaotic state ofthe Italian postal service at the time ofthisresearch, there is no way oftellinghow many of these questionnaires actually were delivered, but twenty-four completed forms eventually were received, for a response rate of 31 percent ofthe total sample. In each case, the response rate was approximately constant across parties and regions, and in each case the response rate was lowest for defeated candidates and highest among elected deputies. There is a slight underrepresentation of ministers and front bench opposition spokesmen in all three countries (especially among British ministers), but in general a surprisingly large number of prominent people responded. While the response rates are far from high and the samples were not large, the overall representativeness of the samples coupled with the overwhelming differences observed between countries appears to compensate forthis. As has been observed, these differences were shown to be statistically significant with tests designed to be used with samples of this size drawn from infinite populations^ and thus there seems to be noreason to doubt the validityofthe findings reported here on the grounds of sampling bias. ...