Fractional Integration Methods in Political Science
Box-Steffensmeier, Janet M.
Controversies in researching political time series often revolve around the best
characterization of the series, i.e., whether a series is stationary or integrated.
By using a fractional integration approach, one can avoid this controversy. Fractionally
integrated series are mean-reverting, but decay at different rates than a stationary series.
Theoretical reasons may also lead one to expect a fractionally integrated series.
Estimation of the d parameter in an ARFIMA (p, d, q) model is no longer difficult and
multivariate extensions are proving useful. Using fractionally integrated methods can
lead to substantive and methodological insights about political processes. We estimate
d for congressional approval and economic expectations data from Durr, Gilmour, and
Wolbrecht (1997) and test for fractional cointegration.
Document ID Number