¹ÏÌïºÚÁÏ

The case of restaurant hygiene grading in Los Angeles occupies a pivotal role in the scholarship on information disclosure. Comparing Los Angeles, which enacted grading in 1998, with the rest of California from 1995-99, Jin and Leslie (2003) (J&L) found that grading caused a 20% reduction in foodborne illness hospitalizations. These dramatic findings influenced a large number of policy reforms and persuaded adherents and detractors alike that simple mandated information disclosure works. Expanding hospitalization data by more than fivefold to cover 1983-2009 and collecting new, richer data on mandatorily reported illnesses from 1990-2015 and salmonella from 1964-2015, we show that the original finding is erroneous. First, raw data reveal that one of the largest salmonella outbreaks in state history hit Southern California in the years preceding Los Angeles's implementation of grading. Foodborne illnesses and hospitalizations drop sharply in 1998 across Southern California counties. Placebo tests detect the same treatment effects for these counties, none of which changed restaurant grading in that time period. Second, we show that J&L's positive result stems from a misspecfication of triple differences and omission of the most prevalent foodborne pathogen next to salmonella. Third, the short observation window yields high false positive rates (Type I error), because a small number of outbreaks are responsible for the vast majority of hospitalizations. Improved designs that control Type I error lack the power to detect effects of substantial magnitude.