Instrumental Variable Estimation in Social Work Research: A Technique for Estimating Causal Effects in Nonrandomized Settings

Abstract

Social work researchers should become familiar with an econometric approach, known as instrumental variable estimation, to estimate causal effects when the variable of interest is not randomly assigned and is, therefore, nonignorable or endogenous. In this article, we introduce instrumental variable estimation and describe how this method is used in model estimation. We focus on the critical assumptions needed to support causal inferences using instrumental variable estimation, provide clear examples of instruments, and describe how these assumptions promote the validity of inference with instruments. We compare the instrumental variable approach to causal estimation with the path model, which is a familiar approach in most social work research, and provide a balanced appraisal of the advantages and disadvantages of each approach.

Family Business Growth Around the World

“…Our finding also contrasts with prior cross-sectional studies on family firm growth by Lee (2006) and Campopiano et al (2019), very likely because our panel data allow us to obtain more accurate estimates of model parameters by observing growth rates over a longer period and covering different stages of the business cycle and of firm development (Evert et al, 2016). In addition, our identification strategy reduces endogeneity concerns via strong and valid IVs (Rose & Stone, 2011), as confirmed by both Kleibergen-Paap and Hansen tests, and it also mitigates survivorship bias (Elton et al, 1996). Table 8 assesses the effect of institutional development on firm growth.…”

Social Work Research and Endogeneity Bias

“…Further, instrumental variable approaches do not appear to be widely used in the social work literature. These methods are discussed in depth in our other work (Rose & Stone, 2011).…”

“…Despite the frequent difficulty in meeting the requirements of the exclusion restriction, instrumental variable estimation represents a very powerful approach to causal inference. Given that instrumental variable approaches have not been as widely referenced in social work-specific journals, we discuss these techniques in depth elsewhere (Rose & Stone, 2011).…”

“…A companion article provides an indepth discussion of the instrumental variable approach, which is a method that has potential power for addressing multiple sources of endogeneity bias (Rose & Stone, 2011).…”

Firm Growth and Legal Environment

“…Size it , R&D ic,t , Lev ic,t and Cash ic,t were detected as endogenous by both tests. Therefore, our explanatory model was estimated using the instrumental variable (IV) estimator, namely 2SLS regressions, to alleviate the endogeneity problem that may influence estimates (Rose and Stone, ). Due to the absence of good external instruments, we adopted the second‐ and third‐year lagged values of all endogenous variables ( Size ic,t‐2 , Size ic,t‐3 , R&D ic,t‐2 , R&D ic,t‐3 , Lev ic,t ‐2 , Lev ic,t ‐3 , Cash ic,t‐2 and Cash ic,t‐ 3 ) following a common practice in the literature (Krafft et al ., ).…”