2SLS versus 2SRI: Appropriate methods for rare outcomes and/or rare exposures

Abstract

This study used Monte Carlo simulations to examine the ability of the two-stage least squares (2SLS) estimator and two-stage residual inclusion (2SRI) estimators with varying forms of residuals to estimate the local average and population average treatment effect parameters in models with binary outcome, endogenous binary treatment, and single binary instrument. The rarity of the outcome and the treatment was varied across simulation scenarios. Results showed that 2SLS generated consistent estimates of the local average treatment effects (LATE) and biased estimates of the average treatment effects (ATE) across all scenarios. 2SRI approaches, in general, produced biased estimates of both LATE and ATE under all scenarios. 2SRI using generalized residuals minimized the bias in ATE estimates. Use of 2SLS and 2SRI is illustrated in an empirical application estimating the effects of long-term care insurance on a variety of binary health care utilization outcomes among the near-elderly using the Health and Retirement Study.

Mother’s mental health after childbirth: Does the delivery method matter?

“…The bivariate probit model assumes that the outcome and treatment are each determined by latent linear index models with jointly normal error terms. The two-stage residual inclusion (2SRI) estimation is a semi-parametric approach that uses the residuals from the first stage to control for endogeneity in the second stage (Terza et al, 2008;Basu et al, 2018). The estimates derived from a bivariate probit are interpreted as average treatment effects (ATE), as bivariate probit and other models of this sort can be used to estimate unconditional average causal effects and/or effects on the treated.…”

“…When choosing between linear IV and 2SRI, Terza et al (2008) argued that in cases of binary dependent variables, we should rely on nonlinear 2SRI as these are the only ones that produce consistent estimates of ATE. Basu et al (2018) analysed the case of binary outcome, a binary treatment and binary instrument and found that the 2SLS method with binary IV produce consistent estimates of LATE, while 2SRI estimate of ATE and LATE are generally bias. Among 2SRI the type of residual form that minimises the bias for the ATE are the generalised residuals.…”

Outcomes of primary care delivery by nurse practitioners: Utilization, cost, and quality of care

“…36 For IV models with dichotomous treatment and outcome variable, the bivariate probit model has been shown to minimize bias when estimating average treatment effects. 37 For continuous outcomes, IV models were implemented using the two-stage residual inclusion estimator. 38 In the first stage, we modeled NP reassignment as a function of the IV and control variables using a GLM with a binomial distribution and logit link.…”

“…Instrument strength was assessed by comparing the partial F‐statistic for the IV with a cutoff of 10, defined by previous studies as a minimum threshold . For IV models with dichotomous treatment and outcome variable, the bivariate probit model has been shown to minimize bias when estimating average treatment effects …”

No Clue About Bioplastics

“…A series of papers has demonstrated that 2SLS generates consistent estimates even when a non-linear model seems more applicable (cf. Angrist, 2001;Basu et al, 2018).…”