Testing Reciprocal Relations by Nonrecursive Structural Equation Models Using Cross-Sectional Data

Structural equation modeling has become a common technique to test the reciprocal relations between two constructs using cross-sectional data in management research. Although there are pros and cons to this application, it appears clear that sometimes it is necessary to use a cross-sectional nonrecursive model as an approximation of the cross-lagged reciprocal effects. Several factors (i.e., proper specification of the nonrecursive model, effects of instrumental variables, time interval, true cross-lagged effects, and sample size) that may affect the adequacy of the nonrecursive model are identified from the literature. Two examples using realistic parameters found in the literature are used to explore the potential effects of these factors in the application of a nonrecursive model. The article ends with a discussion of future research implications and recommendations of studies designed to test possible reciprocal relations between constructs.