¹ Studies of elite background are too numerous to cite at length. The seminal influence is in many respects Matthews, D., The Social Background of Political Decision-Makers (New York: Random House, 1954).Google Scholar Useful collections which include citations to large numbers of other studies are contained in the Hoover Elite series, the introduction to which is Lasswell, H. D., Lerner, D. and Rothwell, C., The Comparative Study of Elites (Stanford: Hoover Institute, 1952).Google Scholar Other useful collections are Marvick, D., ed., Political Decision-Makers (Glencoe: Free Press, 1961)Google Scholar and Maynaud, J., Decisions and Decision-Makers in the Modern State (Paris: Unesco, 1967).Google Scholar For Britain specifically see Guttsman, W., The British Political Elite (London: MacGibbon and Kee, 1963)Google Scholar; Ranney, A., Pathways to Parliament (Madison, Wis.: Wisconsin University Press, 1965)Google Scholar; Finer, S., Berrington, H., Bartholomew, D., Backbench Opinion in the House of Commons 1955–1959 (London: Pergamon, 1961).Google Scholar

² Schwarz, David C., ‘Toward a Theory of Political Recruitment’, Western Political Science Quarterly, XXII (1969), 552–71.CrossRefGoogle Scholar The paradigm derives from Jacob, H., ‘Initial Recruitment of Elected Officials in the U.S. – a Model’, Journal of Politics, XXIV (1962), 703–16CrossRefGoogle Scholar; Barber, J. D., The Lawmakers (New Haven: Yale, 1965).Google Scholar A collection of relevant and interesting studies which discuss some of these points is contained in Annals of the American Academy of Political and Social Science, Vol. 361, 1965.Google Scholar

³ Schwarz, , ‘Political Recruitment’, p. 552Google Scholar; dots indicate omissions from Schwarz’ text.

⁴ For psychological factors (stronger motivation for achievement and power, and weaker motivation for affiliation) linked to participation, see Browning, R. and Jacob, H., ‘Power Motivation and the Political Personality’, Public Opinion Quarterly, XXVIII (1964), 75–90CrossRefGoogle Scholar; also Browning, R., ‘The Interaction of Personality and Political System in Decisions to Run for Office’, Journal of Social Issues, XXIV (1968), 98–109.Google Scholar On the political side, activists’ own recollections emphasize the influence of parties and internal party factions: cf. Schwarz, , ‘Political Recruitment’, pp. 563–6Google Scholar; Budge, Ian and O'leary, C., Belfast: Approach to Crisis (London: Macmillan, 1973). Chap. 10CrossRefGoogle Scholar; Bowman, L. and Boynton, G. R., ‘Recruitment Patterns among Local Party Officials’, American Political Science Review, LX (1966), 667–76CrossRefGoogle Scholar; Seligman, L. G., ‘Political Recruitment and Party Structure’, American Political Science Review, LV (1961), 77–88CrossRefGoogle Scholar; Snowiss, L. M., ‘Congressional Recruitment and Representation’, American Political Science Review, LX (1966), 627–39CrossRefGoogle Scholar; Kim, Y. C., Political Recruitment: the Case of Japanese Prefectural Assemblymen’, American Political Science Review, LXI (1967), 1056–68.Google Scholar Structural constraints have been studied in relation to political careers: cf. Schlesinger, J. A., Ambition and Politics (Chicago: Rand McNally, 1966)Google Scholar; this has been elaborated into a rational choice model in Black, G. S., ‘A Theory of Political Ambition’, American Political Science Review, LXVI (1972) 144–59.CrossRefGoogle Scholar The effect of partisan or non-partisan elections and of the intensity of party competition on recruitment is studied in Kaplan, H., Urban Political Systems: a Functional Analysis of Metro-Toronto (New York: Columbia University Press, 1967), pp. 202–6Google Scholar, which suggests that a non-partisan system favours middle-class candidates. For a case in which non-partisanship favoured the American Democrats, see Wildavsky, A., Leadership in a Small Town (Totowa, N. J.: Bedminster Press, 1964), pp. 47–51.Google Scholar

⁵ Bowman, and Boynton, , ‘Recruitment Patterns’, p. 669.Google Scholar

⁶ Barber, , Lawmakers, p. 15.Google Scholar

⁷ Ian Budge et al., Class, Religion, Politics: Glasgow (unpublished MS), Chap. 5, reports the high values of Goodman and Kruskall's tau (·77 and ·69) obtained for the relation between nine-variable combinations and the activist—elector distinction; values in Belfast (.82) for similar combinations were comparable, Budge and O'Leary, Belfast, Chap. 10.

⁸ Webb, Eugene J. et al. , Unobtrusive Measures (Chicago: Rand McNally, 1966), passim.Google Scholar As well as fitting inside the ‘sociological’ paradigm stressing resources, motivations and opportunities, our procedure is also compatible with the ‘economic’ approach stressing selective incentives and coercion, in the sense that both incentives and coercion are most readily applicable to those outside an organization who closely resemble those already inside it. Cf. Olson, Mancur, The Logic of Collective Action (Cambridge, Mass.: Harvard, 1965).Google Scholar

⁹ A definition which originated with Jacob, ‘Initial Recruitment’, p. 708; see also Schwarz, , ‘Political Recruitment’, p. 552.Google Scholar

¹⁰ This is of relevance to the question of whether there is an ordered hierarchy of political participation, raised in Milbrath, L. W., Political Participation (Chicago: Rand McNally, 1964).Google Scholar

¹¹ The studies cited in fn. 1 above and some of those cited in fn. 7 do use historical data on political careers. Such data are necessarily limited however and often do not allow clear decisions to be made between competing hypotheses. Detailed information is available at the moment only for cross-sectional comparisons.

¹² Detailed information about the Glasgow surveys is given in Appendix A of Budge, Ian, Brand, J. A., Margolis, M. and Smith, A. L. M., Political Stratification and Democracy (London: Macmillan, 1972).CrossRefGoogle Scholar

¹³ Full details about the Belfast surveys are reported in Appendix A of Budge and O'Leary, Belfast.

¹⁴ The Colchester and Maldon surveys were carried out under the supervision of Jean Blondel. Some details of the surveys are reported in Blondel, J. and Hall, R., ‘Conflict, Decision Making and the Perceptions of Local Councillors’, Political Studies, XV (1967), 338–50.Google Scholar

¹⁵ For details of the London surveys see Budge, Ian, Agreement and the Stability of Democracy (Chicago: Markham, 1970), Appendix D.Google Scholar All five data sets listed are held for public use by the British SSRC Survey Archive, University of Essex.

¹⁶ All population figures are approximate.

¹⁷ For a comparison of Glasgow and Belfast in these terms see Budge, Ian and O'leary, C., ‘Cross-cutting Cleavages, Agreement and Compromise’, Midwest Journal of Political Science, XV (1971), 1–30CrossRefGoogle Scholar; and ‘Background Characteristics and Cross-cutting’, Midwest Journal of Political Science, XVI (1972), 712–22.Google Scholar

¹⁸ The measures are described in Budge, Ian, Patterns of Democratic Agreement (Ann Arbor: University Micro-films, 1966), pp. 609–14.Google Scholar Sociability is a subset of questions from Milbrath, L. W., ‘Predispositions towards Political Contention’, Western Political Quarterly, XIII (1960), 5–18Google Scholar; anxiety a sub-set from Taylor, J. A., ‘A Personality Scale of Manifest Anxiety’, Journal of Abnormal and Social Psychology, XLVIII (1953), 285–90CrossRefGoogle Scholar; ‘Authoritarianism’ a sub-set from Adorno, T. W. and associates, The Authoritarian Personality (New York: Harper, 1950).Google Scholar Value orientations are estimated through first choices out of eight personality types based on Lasswell's typology of values — power, affection, respect, rectitude, wealth, well being, skill, enlightenment; Lasswell, H. D. and Kaplan, A., Power and Society (New Haven: Yale, 1950), pp. 55–6.Google Scholar None of these subsets has been tested for scalability with the London groups; they must be regarded more as summarizing patterns of response to particular groups of items than as scales in a true sense.

¹⁹ We define movement in terms of logarithms of the likelihood so that constant ratios between likelihood are expressed as constant differences between logarithms. If p _i is the probability of the i ^th response for a particular party then the movement in the right direction for the i ^th response is logp _i and the average movement in the right direction is There is a simple lemma that states that is maximized, for variation of x _i such that when then we are discussing average movement in the wrong direction and this is less than the average movement in the correct direction. For a proof see Khinchine, A., Foundations of Information Theory (New York: Dover, 1960).Google Scholar

²⁰ Strictly speaking, as we make clear, we are not dealing in true Bayesian prior and posterior probabilities but in standardized likelihoods. A compact derivation of Bayes’ Theorem from general reasoning about probability is found in Wonnacott, and Wonnacott, , Introductory Statistics (New York: Wiley, 1969)Google Scholar, Chap. 5.

²¹ The computer program LiRaS (for Likelihood Ratio Space) in which these procedures are encapsulated, is currently operating on the ICL 1900 and the linked ICL 1900-PDP10 at the University of Essex. All enquiries should be addressed to Dennis Farlie at the Department of Mathematics, University of Essex.

²² Summary statistics might be employed to measure the success of our assignments. We do not report any because none of those commonly used are well adapted to estimate the assignment success of our procedure without bias. Kendall's tau, which would respect the ordinal nature of the scores, is a measure of association rather than of predictive success. Goodman and Kruskall's measures lambda and tau do estimate predictive success but impose a crude dichotomization of scores at ·5, and are also unduly affected by (arbitrary) differences in the size of the population samples.

²³ As noted above we except characteristics very highly correlated with each other in terms of standard measures of association (such as chi-square). Reckoning their effects would in effect be double counting under our procedure, the equivalent of multicollinearity in regression analysis, except that its effect is less awkward in a computational sense. Highly correlated characteristics are therefore represented by only one of the pair, in our analysis, which avoids giving too much weight to one type of variable.

²⁴ Candidates are combined with MPs in London because both are concerned with elective office and differences between them are not great. Party workers are not combined with councillors in Glasgow because they are not concerned with elective office.

²⁵ For the influence of religion on voting in Glasgow, see Budge et al., Class, Religion, Chap. 6. For its influence in Belfast, see Budge and O'Leary, Belfast, Chap. 7.

²⁶ Budge and O'Leary, Belfast, Chap. 10.

²⁷ Budge and O'Leary, Belfast, Chap. 7.

²⁸ Butler, David and Stokes, D. E., Political Change in Britain (Harmondsworth, Middx.: Penguin, 1971), pp. 174 ff.CrossRefGoogle Scholar

²⁹ Budge et al., Class, Religion, Chap. 2.

³⁰ The assumption that dominates discussion in Milbrath's, propositional inventory of the literature on participation, Political Participation, p. 16 ff.Google Scholar; for an application see Budge et al. Political Stratification, Chap. 1.

³¹ Cf. D. Marvick and C. Nixon, ‘Recruitment Contrasts in Rival Campaign Groups’ in Marvick, ed., Political Decision-Makers.

³² Parliamentary candidates are readily traced through the Times Guide to the House of Commons for 1964. Belfast councillors’ dropout is traced from municipal election data reported in local newspapers for 1967.

³³ The eight common characteristics are used instead of the full set of original characteristics for each locality because we wish to generalize our results as far as possible to the whole of Britain. Little difference emerges between the likelihood distribution whether we use the full or limited sets of background characteristics.

³⁴ Variation in such categories did not affect the generalization of scores to other localitiesreported in Tables 5 and 6, because it affects the least populous categories and hence few individuals in each case. The most populous categories display only limited variation between localities, since changes in a few individuals’ position naturally affect them less.