This paper extends previous analyses of aggregate political popularity (par
tisanship) data by Box-Steffensmeier and Smith (Box-Steffensmeier, J.M., Sm
ith, R.M., 1996. The dynamics of aggregate partisanship. American Political
Science Review 90 (September), 567-580) for the US, and Byers et al. (Byer
s, D., Davidson, J., Peel, D.A., 1997. Modelling political popularity: an a
nalysis of long-range dependence in opinion poll series. Journal of the Roy
al Statistical Society Series A, 160, 471-490) for the UK. These studies in
dependently found that the time series of poll ratings are well modelled by
fractionally integrated processes. Here, the analysis is conducted for 26
political parties in eight different countries, and the results obtained ar
e on the whole closely in line with the ones cited above. As in the earlier
studies, we find in many of our cases that the estimated fractional integr
ation parameter d is close to 0.7. This implies that popularity is highly p
ersistent and a nonstationary process, but that it is also mean-reverting e
ventually. Most of the time series are also found to be pure fractional noi
se, effectively uncorrelated after fractional differencing, so that the d p
arameter alone accounts for the dependence. As well as offering added suppo
rt for theories of political allegiance based on a certain distribution of
the attributes of commitment and pragmatism in the voting population, these
findings have important implications for the explanation of political supp
ort using time series data. (C) 1999 Elsevier Science Ltd. All rights reser
ved.