Hostname: page-component-cd9895bd7-gxg78 Total loading time: 0 Render date: 2024-12-25T07:39:52.251Z Has data issue: false hasContentIssue false

Do Critics Always Matter? Or, A Comedy of Errors

Published online by Cambridge University Press:  27 January 2009

Extract

On one level, it is something of a pleasure when one's colleagues take sufficient interest in a paper to write a rejoinder. Furthermore, Roweth et al. are to be complimented on writing their rejoinder in a systematic manner. It seems only reasonable therefore, that we should respond to each of their charges, which they label our ‘shortcomings’, by following their format.

Type
Notes and Comments
Copyright
Copyright © Cambridge University Press 1980

Access options

Get access to the full version of this content by using one of the access options below. (Log in options will check for institutional or personal access. Content may require purchase if you do not have access.)

References

1 One major collaborative and international research endeavour is currently being conducted under the auspices of the European Consortium for Political Research. For a relevant and early extract from this study, see Castles, Francis G., ‘Patterns of Politics and Patterns of Public Expenditure’ (paper presented to ECPR Joint Sessions, Florence, 03 1980).Google Scholar

2 These three components have relatively low positive correlations with each other (transfer–education ·34; transfer–(reciprocal) mortality ·13; education-mortality ·31). Each also has a quite high negative correlation with vote for the Right (transfers–·62, education –·49, mortality –·50). Since the inter-component correlations are all lower than each component's individual correlation with vote for the Right, then of course collinearity is relatively low and we find a higher correlation between the composite index and vote for the Right (–·75). The important point to note, however, is that the inter-component correlations have consistent signs. In this respect we are dealing with a uni-dimensional phenomenon. If we had included more components, then the correlation between this enlarged composite index and vote for the Right would actually have increased as long as the extra dimensions did not have significant negative correlations with the existing set. Since public welfare is widely conceptualized as a composite measure, then it does seem sensible to use a composite indicator as an operational measure as long as the components are consistent (which in our case they are). This seems to be one of the several occasions on which Roweth et al. make rather pedantic points based, it seems to us, on a failure to distinguish between (or unawareness of) conceptual and operational definitions.

3 There are, of course, other techniques. For example, we could have run a principal components analysis, and used factor loadings as the coefficients. Apart from a number of technical reservations about factor analysis in general, we are also of the opinion that in practice interpretation of factor analysis is often quite arbitrary and also that the weighting problem is solved not theoretically but in a data-dredging manner. The case of public welfare does seem to us to be a relatively straightforward one, and therefore our preference (and of course it is very clearly a personal preference) is to go for a more simple and explicit technique.

4 The most knowledgable of all are presumably the electorate of each particular country. It is, therefore, of some interest that our classification closely corresponds with survey evidence of voters' perceptions of the parties' relative positions on national Right-Left scales. See Inglehart, Ron and Klingemann, Hans, ‘Party Identification, Ideological Preference and Right-Left Dimension among Western Mass Publics’, in Budge, Ian, Crewe, Ivor and Farlie, Dennis, eds., Party Identification and Beyond (London: Wiley, 1976), pp. 243–73.Google Scholar

5 See Gold, David, ‘Statistical Tests and Substantive Significance’, American Sociologist, IV (1969), 42–6Google Scholar; Kish, Leslie, ‘Some Statistical Problems in Research Design’, American Sociological Review, XXIV (1959), 328–38CrossRefGoogle Scholar; Selvin, H. C., ‘A Critique of Tests of Significance in Survey Research’, American Sociological Review, XXII (1957), 519–27CrossRefGoogle Scholar; Winch, R. F. and Campbell, D. T., ‘Proof? No. Evidence? Yes. The Significance of Tests of Significance’, American Sociologisl, IV (1969), 140–3.Google Scholar

6 Glejser, H., ‘A New Test for Heteroscedasticity’, Journal of the American Statistical Association, LXIV (1969), 316–23CrossRefGoogle Scholar. For a recent very interesting discussion of a positive approach to this problem, see Downs, George W. and Rocke, David M., ‘Interpreting Heteroscedasticity’, American Journal of Political Science, XXIII (1979), 816–28.CrossRefGoogle Scholar

7 For the benefit of Roweth et al., we are well aware that the traditional response to heteroscedasticity is to use either weighted least squares or some variable transformation. Since there is no such problem, then of course we are not obliged to use either of these techniques. Roweth et al. may be interested to know, however, that a log-log regression of welfare on vote for the Right gives an R2 value of 61 (compared with the one of 56 that we report). Since there is neither a technical nor a theoretical reason why we should not use a non-linear solution, we would prefer to remain with our linear result.