Estimating the Nonfatal Injury Undercount in Agriculture from 2004 to 2019

Government estimates of nonfatal occupational injuries and illnesses understate the hazards in U.S. agriculture. From 2004 to 2019, government estimates only captured 13% to 26% of the true number of nonfatal injuries. Estimates of nonfatal injuries are more accurate for crop production than for animal production. Estimates are more accurate over time, with a decline in self-employed and unpaid family labor in agriculture.

The U.S. Bureau of Labor Statistics provides annual estimates of nonfatal occupational injuries and illnesses by U.S. industry sector. We performed a series of corrections to these estimates for each year from 2004 through 2019 to account for institutional and behavioral drivers of the undercount in the sample used to construct these estimates for the U.S. agricultural industry. Institutional factors consisted of the exclusion of small farms and self-employed and family workers, as well as the employment undercount due to the highly seasonal nature of agricultural work. Behavioral factors consisted of willful and negligent underreporting by employers. We updated the estimates using information on the number of people employed in the excluded portions of the agricultural industry and estimates of the underreporting rate from prior work. Over this period, we show that the government estimates only captured 13% to 26% of the true number of nonfatal injuries and illnesses, missing 74% to 87% of the true case counts each year. The government estimates were more accurate for crop production, missing an average of 77% of cases, than for animal production, missing an average of 83% of cases. Willful and negligent underreporting was the largest contributor to the undercount, followed by the exclusion of self-employed and unpaid family workers.