At What Level Should One Cluster Standard Errors in Paired and Small-Strata Experiments? / Clément de Chaisemartin, Jaime Ramirez-Cuellar.

Format:

Book

Author/Creator:

de Chaisemartin, Clément.

Contributor:

National Bureau of Economic Research.

Ramirez-Cuellar, Jaime.

Series:

Working Paper Series (National Bureau of Economic Research) no. w27609.

NBER working paper series no. w27609

Language:

English

Physical Description:

1 online resource: illustrations (black and white);

Place of Publication:

Cambridge, Mass. National Bureau of Economic Research 2020.

Summary:

In clustered paired experiments, randomization units, say villages, are matched into pairs, and one unit of each pair is randomly assigned to treatment. To estimate the treatment effect, researchers often regress their outcome on the treatment and pair fixed effects, clustering standard errors at the unit-of-randomization level. We show that the variance estimator in this regression may be severely downward biased: under constant treatment effect, its expectation equals 1/2 of the true variance. Instead, researchers should cluster at the pair level. Using simulations, we show that those results extend to clustered stratified experiments with few units per strata.

Notes:

Print version record

July 2020.