Information technology outsourcing and firm productivity: eliminating bias from selective missingness in the dependent variable

Abstract

Summary Missing values are a major problem in all econometric applications based on survey data. A standard approach assumes data are missing at random and uses imputation methods or even listwise deletion. This approach is justified if item nonresponse does not depend on the potentially missing variables’ realization. However, assuming missingness at random may introduce bias if nonresponse is, in fact, selective. Relevant applications range from financial or strategic firm-level data to individual-level data on income or privacy-sensitive behaviors. In this paper, we propose a novel approach to deal with selective item nonresponse in the model’s dependent variable. Our approach is based on instrumental variables that affect selection only through a partially observed outcome variable. In addition, we allow for endogenous regressors. We establish identification of the structural parameter and propose a simple two-step estimation procedure for it. Our estimator is consistent and robust against biases that would prevail when assuming missingness at random. We implement the estimation procedure using firm-level survey data and a binary instrumental variable to estimate the effect of outsourcing on productivity.

Early child development and parents' labor supply

“…In our correction approach, we assume that income is the main driver for selection and not the outcome variable–hours worked. This is different to the work by Breunig et al (2020), Breunig et al (2018), and d'Haultfoeuille (2010) who use the selection variable also as outcome. Therefore, we add in addition the following assumption:…”

“…Similar assumptions have been used in the literature dealing with nonignorable nonresponse, such as Breunig, Kummer, Ohnemus, and Viete (2020), Chen (2001), Ramalho and Smith (2013), and Tang, Little, and Raghunathan (2003). Fricke, Frolich, Huber, and Lechner (2020) also deal with endogeneity of the treatment and nonresponse at the same time.…”

Time-Series Imputation with Wasserstein Interpolation for Optimal Look-Ahead-Bias and Variance Tradeoff

“…[14] found that earnings data is not missing at random, and hence there is a bias (which they call attrition bias) in estimates of earnings properties. [4] proposed an approach when the random missing pattern depends on observable variables. [19,26,31] use logistic regression models to model the missing data distribution.…”

“…The sum of the two components ECMSE achieves a minimum for some non-trivial selection of δ for the last two cases, showing that our method has the potential to outperform the naive posteriors (i.e., the posteriors corresponding to δ = 0 or δ = δ max ). 4 Retrieved on Jan. 24th. 2021 from https://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html…”