The published county link between neonicotinoid use and US bird declines in 2008–2014 comes from national time trends, and BBS routes in 1994–2014 reject declines of that size after adoption

Li, Miao and Khanna (2020, Nature Sustainability) reported that 100 kg more neonicotinoid use per US county decreased grassland bird counts by 2.2% in 2008–2014. We did two tests. Test A used the published data and code. We reproduced the estimate (−18.07 birds per 1000 kg, −2.15% per 100 kg). The two instruments (national price indices) have the same value in every county in a given year, so only seven national annual means identify the effect. When we shuffled neonicotinoid use among counties in each year, the estimate stayed near the published value (median −16.6; all 200 shuffles z < −1.96). A random upward series gave −16.0. With county and year fixed effects, the estimates were near zero in all four bird groups (grassland −0.009% per 100 kg, SE 0.14%). Test B was an independent negative-control test on 41,050 standard Breeding Bird Survey runs (3,337 routes, 1,827 counties, 1994–2014) with USGS EPest county estimates. National use in 1994–2003 was about 2% of the 2012–2014 level. On routes in counties with high later use, grassland birds already declined faster before adoption (−6.5% per decade per SD of later use, SE 1.7%). This differential became smaller after 2004 (−1.7%, SE 1.4%), not larger. Insectivores and all species showed no differential change. When we added the published effect to the real counts, the test detected it clearly, and the observed data rejected it (z = 10.7 to 17.1). With use per county, the effects consistent with the data were −0.5% to +0.3% per 100 kg. With use per km², the all-species estimate was a 0.20% decline per 100 kg (95% CI 0.04 to 0.37), about one tenth of the published size. Limitations: EPest values are modelled county estimates, and each route is assigned only to the county of its start point. The break test assumes no other change in differential trends at 2004. Effects below about 0.5% per 100 kg are not excluded.

neonicotinoids · Breeding Bird Survey · grassland birds · insectivorous birds · pesticides · instrumental variables · negative control · event study · reanalysis · EPest · difference-in-differences · bird decline