PAUL BRACE, RICE UNIVERSITY KEVIN ARCENEAUX, TEMPLE UNIVERSITY MARTIN JOHNSON, UNIVERSITY OF CALIFORNIA, RIVERSIDE STACY G. ULBIG, MISSOURI STATE UNIVERSITY We recently became aware of a computation error in correlations appearing in footnote 9 on page 536 in our recent article in this journal.1 In searching for the cause of these errors, we became aware that Berry, Ringquist, Fording, and Hanson have produced a series of corrections of their ideology measure, which we use for these correlations and other analysis. The corrected versions of their measures are archived at http://webapp. icpsr.umich.edu/cocoon/ICPSR-STUDY/01208.xml. The primary source of the correlation errors we discovered was a file merging error on our part restricted to the dataset designed to produce the correlations between the Berry et al. measure (1998) and the other ideology measures reported in footnote 9. We recomputed the correlation using their revised measure and correcting our merge procedure with the other measures of ideology to produce the following Spearman's correlations: The Berry et al. (1998) measure is more highly correlated with the Erikson, Wright, and Mclver (Wright 2001) measure and our out-of-sample GSS/ANES state ideology timeseries than we originally reported in our footnote 9. We gratefully acknowledge Evan Ringquist for correspondence that led us to reconsider the correlations in this footnote and the corrected Berry et al. data. Concerned that the corrected Berry et al. measure could produce fundamentally different results concerning intrastate ideological change (one of our primary interests in the paper), we re-analyzed these data to reconsider hypotheses concerning linear, curvilinear, or cyclic patterns of changes in political ideology at the state level. In revisiting our analysis of Berry, et al, data, we discovered an estimation error which affected coefficients on the original table, but not the inferences we drew from it. We replace Table 3 using the updated BRFH data and correcting our estimation. Reanalysis of the recently updated Berry et al. measure does not change our fundamental points: There is more cross-sectional than longitudinal variation in the three measures of state ideology and researchers should be cautious not to confuse the effects of cross-sectional differences with longitudinal changes. Replicating the ANOVA estimation we discuss in the original paper using the corrected Berry et al. data, we similarly find that 5.76 percent of the variation in their new data is attributable to longitudinal differences while 77.3 percent of the measures variance is attributable to differences among states. A small number of states (California and West Virginia) have seen systematic changes in state-level political ideology identified using all three measures of political ideology. We regret error in our original article but appreciate the opportunity to reconsider our conclusions in light of corrections Berry et al. have made to their measure. The fundamental point of our paper does not change. Updated versions of other tables from the article are available at http:// CORRECTION URL. REFERENCES Berry, William D., Evan J. Ringquist, Richard C. Fording, and Russell L. Hanson. 1998. Measuring Citizen and Government Ideology in the American States, 1960-93. American Journal of Political Science 42: 327-48. Wright, Gerald C. 2001. Zip File of the CBS/New York Times National Polls, Ideology Party Identification, 1977-1998. pbrace@rice.edu DUMMY TEXT A Comment on Correction: Does State Political Ideology Change Over Time? WILLIAM D. BERRY, FLORIDA STATE UNIVERSITY EVAN J. RINGQUIST, INDIANA UNIVERSITY BLOOMINGTON RICHARD C. FORDING, UNIVERSITY OF KENTUCKY RUSSELL L. HANSON, INDIANA UNIVERSITY BLOOMINGTON Brace et al. (2004: 537) assert that mass political ideology in the American states is overwhelming[ly] stable, based on their analysis of three longitudinal measures of citizen ideology: Berry et al. …
Read full abstract