Machine Labor

Abstract

Walters, and seminar participants at Columbia and MIT for helpful discussions and comments. They are not to blame for any of our mistakes or conclusions. This paper is dedicated to the memory of Alan Krueger. We would have liked to have his comments. The views expressed herein are those of the authors and do not necessarily reflect the views of the National Bureau of Economic Research. NBER working papers are circulated for discussion and comment purposes. They have not been peerreviewed or been subject to the review by the NBER Board of Directors that accompanies official NBER publications.

Misdemeanor Prosecution

“…Columns (3) and ( 6) split the sample into 8 subgroups by quartiles of this propensity score and by whether the defendant has a previous complaint, an important source of hetereogeneity. ***p < 0.01,**p < 0.05 ,*p < 0.10. as instruments (Bekker, 1994;Chao and Swanson, 2005;Angrist and Frandsen, 2020). The coefficient in this model is -0.27, closer to our main leave-out mean 2SLS estimate than the estimate in Column (2) with all the ADA dummy variables.…”

“…49 In practice implemented via the user-written package ivlasso in Stata (Ahrens, Hansen and Schaffer, 2019), using the post-lasso results and using the ivlasso defaults with a plug-in penalty. The procedure retains three out of 315 instruments (similarly in the Angrist and Frandsen (2020) implementation of the plug-in penalty, lasso retains two instruments out of 180 in a re-estimation of the Angrist and Keueger (1991) QOB study). We also implemented a version of ivlasso with a cross-validated penalty; see Angrist and Frandsen (2020) for details on implementation.…”

“…The procedure retains three out of 315 instruments (similarly in the Angrist and Frandsen (2020) implementation of the plug-in penalty, lasso retains two instruments out of 180 in a re-estimation of the Angrist and Keueger (1991) QOB study). We also implemented a version of ivlasso with a cross-validated penalty; see Angrist and Frandsen (2020) for details on implementation. The algorithm with the CV penalty chooses more instruments, namely 173 out of the 315 in our case.…”

Promoting Parental Involvement in Schools: Evidence from Two Randomized Experiments

“…Recent studies have shown that machine-learning tools can be useful for principled variable selection (Goller, Lechner, Moczall, & Wolff, 2020;B. K. Lee, Lessler, & Stuart, 2010;Urminsky, Hansen, & Chernozhukov, 2016) and we show that our results are robust to a range of specifications, as suggested by Angrist and Frandsen (2019). We take caution in interpreting the treatment effect of the standard grant amount, given the strong unconfoundedness assumption required for identification.…”

Decomposition of the gender wage gap using the LASSO estimator

“…For example,Bach, Chernozhukov and Spindler (2018) analyze the gender wage gap using data from the 2016 American Community Survey and use the double LASSO method to select among up to 4,382 regressors. See alsoAngrist and Frandsen (2019).3Miller (1984) discusses different algorithms for the subset selection technique. The algorithms either evaluate all subsets of the set of explanatory variables or use a heuristic for which subsets to evaluate.…”