Contemporaneous endogenous regressors make concurrent-effect estimates unreliable, favoring lagged specifications

Evidence

We restrict the contemporaneous coefficient to zero in our analysis for the following reasons. First, as mentioned, interpreting contemporaneous coefficients requires the assumption of contemporaneous exogeneity. We expect the explanatory variables to be contemporaneously endogenous to the outcome variables, so the inclusion of contemporaneous terms leads to the problem of overcontrol. Second, our theoretical model does not suggest any industry-wide instantaneous effect of the independent variables on the dependent variables. Third, with the sample size we have, including the contemporaneous term would either limit the number of the control variables in the model or induce a collinearity problem. Finally, when we include the contemporaneous terms, most contemporaneous coefficients are statistically nonsignificant. — Ken-Hou Lin & Donald Tomaskovic-Devey, Financialization and U.S. Income…, 116

Categorization

• XREF: Connects to the methodological debate over simultaneous vs. lagged effects in time-series analysis, and to econometric warnings about overcontrol bias when endogenous covariates are included.
• QUESTION: Worth noting that if contemporaneous coefficients are mostly insignificant anyway, lagged ECM is a safety check rather than a substantive choice — could be a review angle on robustness in financialization studies.