Segmented Regressions and Causality (with applications to macroeconomic time series)

The segmented regression approach described in Bianchi (1993a,b) is applied in this paper to pairs of time series, with the aim of establishing empirical relationships and directions of causality between variables. The approach is based on the asumption of regime shifts in univariate time series occurring infrequently over time. Cause-effect relationships are established on the basis of the timing of the regime shifts in the series. Applications to the wage-price, the tax-spending, and the money-price relations in the United States are reported.