The Best of Two Worlds? Merging Psychometrics and Econometrics

• In Event: The New Kids on the Methods Block: Using New Methods in Existing Data to Get New Answers

Abstract

Non-experimental researchers in developmental psychology and education have traditionally been concerned with measuring and modelling the key constructs precisely, often applying psychometric techniques such as factor models. Recently, the field has become increasingly focused on causal inferences, and hence, on applications of econometric techniques such as instrumental variables, difference-in-difference, and regression discontinuity. Each of these approaches represent some clear advantages. Psychometrics provides a framework for conseptualizing measured contructs and a tool kit for reducing bias and impresision in these measures. Econometrics provides a framework for conseptualizing causal relations and a tool kit for estimating causal effects. By focusing primarily on psychometrics, there is a potential risk of misidentifying causal effects due to unmeasured confounders. By ignoring psychometrics in causal estimations, there is a risk of imprecise, biased, or wrong estimates, due to lack of attention to multidimensionality, measurement invariance, or other construct model misspecifications.
In this presentation, we bring the best of these two worlds together. Using a Structural Equation Model framework, we demonstrate how measurement models can be merged with causal identification models in one analytical framework. We show examples applying instrumental variable and difference-in-difference models (including standard fixed-effects) on the X-side of the equation, with confirmatory factor models on the Y-side. We show how the model can be generalized to include measurement models on the X-side as well. In our examples, we use simulated and real data to identify conditions under which combining psychometric and econometric techniques provides provides more correct or precise estimates, and provide specific recommendations for application.

Authors

• Henrik D. Zachrisson, University of Oslo
  Presenting Author

• Nicolai Topstad Borgen, Universitetet i Oslo
  Non-Presenting Author

• Astrid Marie Jorde Sandsør, Universitetet i Oslo
  Non-Presenting Author

• Eric Dearing, Boston College
  Non-Presenting Author

• Gulbin Sengul-Inal, University of Oslo
  Non-Presenting Author

• Lynn A Karoly, RAND - DC
  Non-Presenting Author

• Eivind Ystrøm, University of Oslo
  Non-Presenting Author

• Torkild Hovde Lyngstad, University of Oslo
  Non-Presenting Author