Improving instrumental variable estimators with poststratification

被引:0
作者
Pashley, Nicole E. [1 ]
Keele, Luke [2 ,3 ]
Miratrix, Luke W. [4 ]
机构
[1] Rutgers State Univ, Dept Stat, 501 Hill Ctr,110 Frelinghuysen Rd, Piscataway, NJ 08854 USA
[2] Univ Penn, Dept Surg, Philadelphia, PA USA
[3] Univ Penn, Dept Biostat, Philadelphia, PA USA
[4] Harvard Univ, Grad Sch Educ, Cambridge, MA USA
关键词
blocking; compliance; instrumental variables; poststratification; randomization inference; voter mobilization; PRINCIPAL STRATIFICATION; VOTER TURNOUT; CAUSAL; MOBILIZATION; INFERENCE; CALLS; MAIL;
D O I
10.1093/jrsssa/qnae073
中图分类号
O1 [数学]; C [社会科学总论];
学科分类号
03 ; 0303 ; 0701 ; 070101 ;
摘要
Experiments studying get-out-the-vote (GOTV) efforts estimate causal effects of various mobilization efforts on voter turnout. However, there is often substantial noncompliance in these studies. A usual approach is to use an instrumental variable (IV) analysis to estimate impacts for compliers, here being those actually contacted by the investigators. Unfortunately, popular IV estimators can be unstable in studies with a small fraction of compliers. We explore poststratifying the data (e.g. taking a weighted average of IV estimates within each stratum) using variables that predict complier status (and, potentially, the outcome) to mitigate this. We present the benefits of poststratification in terms of bias, variance, and improved standard error estimates, and provide a finite-sample asymptotic variance formula. We also compare the performance of different IV approaches and discuss the advantages of our design-based poststratification approach over incorporating compliance-predictive covariates into the two-stage least squares (2SLS) estimator. In the end, we show that covariates predictive of compliance can increase precision, but only if one is willing to make a bias-variance trade-off by down-weighting or dropping strata with few compliers. By contrast, standard approaches such as 2SLS fail to use such information. We finally examine the benefits of our approach in two GOTV applications.
引用
收藏
页数:26
相关论文
共 53 条
[11]   Principal stratification analysis using principal scores [J].
Ding, Peng ;
Lu, Jiannan .
JOURNAL OF THE ROYAL STATISTICAL SOCIETY SERIES B-STATISTICAL METHODOLOGY, 2017, 79 (03) :757-777
[12]   Principal Score Methods: Assumptions, Extensions, and Practical Considerations [J].
Feller, Avi ;
Mealli, Fabrizia ;
Miratrix, Luke .
JOURNAL OF EDUCATIONAL AND BEHAVIORAL STATISTICS, 2017, 42 (06) :726-758
[13]   Principal stratification in causal inference [J].
Frangakis, CE ;
Rubin, DB .
BIOMETRICS, 2002, 58 (01) :21-29
[14]  
Gerber A. S, 2012, Field experiments: Design, analysis, and interpretation
[15]   Social pressure and voter turnout: Evidence from a large-scale field experiment [J].
Gerber, Alan S. ;
Green, Donald P. ;
Larimer, Christopher W. .
AMERICAN POLITICAL SCIENCE REVIEW, 2008, 102 (01) :33-48
[16]   The effects of canvassing, telephone calls, and direct mail on voter turnout: A field experiment [J].
Gerber, AS ;
Green, DP .
AMERICAN POLITICAL SCIENCE REVIEW, 2000, 94 (03) :653-663
[17]  
Green D. P., 2003, Replication materials for: Getting out the vote in local elections: Results from six door-to-door canvassing experiments
[18]  
Green D.P., 2017, Journal of Experimental Political Science, V4, P107, DOI DOI 10.1017/XPS.2017.5
[19]   Voter Mobilization, Experimentation, and Translational Social Science [J].
Green, Donald P. ;
Gerber, Alan S. .
PERSPECTIVES ON POLITICS, 2016, 14 (03) :738-749
[20]   Field Experiments and the Study of Voter Turnout [J].
Green, Donald P. ;
Mcgrath, Mary C. ;
Aronow, Peter M. .
JOURNAL OF ELECTIONS PUBLIC OPINION AND PARTIES, 2013, 23 (01) :27-48