Structural Equation Models and Direction of Causality with Longitudinal Data - An Application to Subjective Well-Being. When two variables are correlated, the researcher is often confronted with the question of the direction of causality. This substanttve question is, however, difficult to answer, especially when use of an expertimental design is impossible. In such situations, longitudinal data furnish some invaluable information: the temporal order of the variables. Unfortunately, the temporal order is not sufficient to answer questions of causality. The well-known technique of cross-lagged correlations has been criticized because of its unrealistic assumptions. The major purpose of this paper is to illustrate the use of structural equations models to help answer questions regarding the direction of causality in longitudinal data. After a brief presentation of structural equations models, focalizing on the LISREL model, we stress their advantages over more traditional approaches. An empirical illustration is presented which uses the data on subjective well-being published by Headey et aL (1991). We study the direction of causality between marriage satisfaction and general satisfaction. The different models tested seem to infirm the conclusions of Headey et al.