Note on choosing between competing interpretations of cross-lagged panel correlations.

Abstract

Campbell's cross-lagged correlation technique (Campbell & Stanley, 1963) has recently been discovered to offer competing interpretations (Rozelle & Campbell, 1969; Yee & Gage, 1968a, 1968b). Specifically, the technique in itself does not discriminate between congruent effects of one of the two variables involved and incongruent effects of the other. As Rozelle and Campbell (1969) note, This greater equivocality must be accepted as a definite reduction over previous estimates of the utility of the method [p. 76|. Yee and Gage (1968a, 1968b) proposed a method to identify congruent and incongruent changes, which is essentially an extension of Campbell's method of analysis. The idea is to identify congruent changes with increases and incongruent changes with decreases in synchronous correlations. The causal direction is then inferred as in Campbell's method. However, it would seem that this could not be unconditionally true, depending as it must be on converging and diverging changes in the variables themselves. Moreover, as Duncan (1969) recently has pointed out, causal inferences will always be underdetermined by 2W2V data. However, by a simple and reasonable assumption and a modification of the operational procedures, a possible solution for identifying congruent and incongruent causal relationships might be found. The assumption is that the effects of a causal state or event decrease (vanish, dissipate) over time. In fact, this is a common-