EMSS: New EM-type algorithms for the Heckman selection model in R

Abstract

When investigators observe non-random samples from populations, sample selectivity problems may occur. The Heckman selection model is widely used to deal with selectivity problems. Based on the EM algorithm, Zhao et al. ( 2020) developed three algorithms, namely, ECM, ECM(NR), and ECME(NR), which also have the EM algorithm's main advantages: stability and ease of implementation. This paper provides the implementation of these three new EM-type algorithms in the package EMSS and illustrates the usage of the package on several simulated and real data examples. The comparison between the maximum likelihood estimation method (MLE) and three new EM-type algorithms in robustness issues is further discussed.

Workplace Injuries and Receipt of Benefits From Workers’ Compensation and Social Security Disability Insurance

“…Moreover, there is other indirect evidence supporting the notion that there is scope for structuring WC settlements to avoid offsets—as noted earlier, most likely in lump-sum cases. A recent WCRI study 24 provides evidence on attorney involvement, which could potentially reflect when attorneys are more involved in these types of settlements. The study notes that of the states covered in their data, the one with the lowest worker attorney involvement (Wisconsin) has limits on lump-sum settlements, and in another state with among the lowest worker attorney involvement (Texas), most lump-sum settlements are prohibited, which the study suggests may have held down attorney involvement.…”

“… o Attorneys may simultaneously help workers get permanent WC benefits and apply for SSDI. The study by Yang et al 24 provided evidence on how common attorney involvement is and found anywhere from 12% to 50% of claims with more than 7 days of lost time worker attorneys had, with the percentage varying across states.…”