Abstract :
While usingdetailedfirm-leveldatafromtheprivatebusinesssector,thisstudy
identifiestwoempiricalpuzzles:(i)returns-to-scale(RTS)parameterestimatesriseat
higherlevelsofdataaggregationand(ii)estimatesfromthefirmlevelsuggest
decreasingreturnstoscale.Theanalysisshowsthat,althoughconsistentwithrising
estimates,neitherentry/exitnortheBasu–Fernald[ReturnstoscaleinU.S.production:
estimatesandimplications.JournalofPoliticalEconomy105,249–283)aggregation-
bias effectdrivesthisresult.Rather,risingandtoolowRTSestimatesseemtoreflecta
mixtureofrandomerrorsinfactorinputsatthefirmlevel.Itturnsout,infact,thata
7.5–10percenterrorinlabor(hoursworked)canexplainbothpuzzles.
