Weak Instruments in Instrumental Variables Regression: Theory and Practice

By Isaiah Andrews, James H. Stock, Liyang Sun. Annual Review of Economics, 2019

667 citations recorded by the source corpus.

When instruments are weakly correlated with endogenous regressors, conventional methods for instrumental variables (IV) estimation and inference become unreliable. A large literature in econometrics has developed procedures for detecting weak instruments and constructing robust confidence sets, but many of the results in this literature are limited to settings with independent and homoskedastic data, while data encountered in practice frequently violate these assumptions. We review the literature on weak instruments in linear IV regression with an emphasis on results for nonhomoskedastic (heteroskedastic, serially correlated, or clustered) data. To assess the practical importance of weak instruments, we also report tabulations and simulations based on a survey of papers published in the American Economic Review from 2014 to 2018 that use IV. These results suggest that weak instruments remain an important issue for empirical practice, and that there are simple steps that researchers can take to better handle weak instruments in applications.

Subfields: Economics and Econometrics, Statistics and Probability

View Paper: https://www.semanticscholar.org/paper/c9e830074084952064a139fd17e7ff55a8788e60

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Economics and Econometrics

In Statistics and Probability

Citations in the corpus

One Instrument to Rule Them All: The Bias and Coverage of Just-ID IV (2021, methodology): “This modeling framework parallels that used in Andrews et al. (2019) and earlier theoetical investigations of weak instrument problems. The normality of reduced-form estimates is justified by conventional symptotic reasoning, as well as by the local-to-zero asymptotic sequence used in Staiger and St”
Does Institutional Quality Foster Economic Complexity? (2019, positive): “Given that none of these bound estimates contains zero, the estimated effect of the plausibly exogenous component of EFW on ECI retains its statistical precision regardless of the strength of the instrument in the first-stage regression (Andrews et al., 2019).”
Misinformation During a Pandemic (2020, neutral): “25In independent work, Ash et al. (2020) also find survey evidence that Republican Hannity viewers adopt social distancing measures significantly later than Republicans who do not watch Hannity, while Republican Tucker Carlson Tonight viewers adopt social distancing measures significantly earlier th”
Determinants of Global Value Chain Participation: Cross-Country Evidence (2020, neutral): “The existing econometric theory focuses on cross-sectional IV estimation, with i.i.d. error terms as discussed in Andrews, Stock, and Sun (2019).”
The effect of media-linked directors on financing and external governance (2021, neutral): “As recommended by Andrews, Stock, and Sun (2019), we compute the Montiel Olea and Pflueger (2013) effective first stage F statistic and compare to it to a benchmark of 10.”
Methods Matter: p-Hacking and Publication Bias in Causal Analysis in Economics (2020, background): “3Our findings are broadly consistent with a growing literature discussing model misspecification for IV regressions (see, for instance, Andrews et al. (2019) for a discussion on weak instruments). Using 1,359 instrumental variables regressions from 31 published studies, Young (2020) show that more t”
The Causal Impact of Removing Children from Abusive and Neglectful Homes (2019, neutral): “As an additional robustness test, we follow the recommendation from Andrews, Stock, and Sun (2019) and compute Anderson-Rubin (AR) confidence intervals for the effects on test score and non–test score outcomes.”