Singleton groups—groups with only one observation—are common in regressions with multiple levels of fixed effects, such as in the work of Carneiro, Guimarães, and Portugal (2012), who estimate linear regressions that feature fixed effects for each worker, firm, and job title. For
instance, in their employer-employee matched dataset, 27% of all fixed effects were singletons (1.5 million singleton groups out of a total of 5.6 million fixed effects). The consequences of having such a large occurrence of singleton group has not been studied in practice, as models
with many levels of fixed effects were not feasible until recently due to the lack of practical estimators (Abowd, Creecy, and Kramarz 2002; Guimarães and Portugal 2010; Gaure 2013; Correia 2015).