28Each year, millions of kilograms of insecticides are applied to crops in the US. While insecticide 29 use supports food, fuel, and fiber production, it can also threaten non-target organisms, a concern 30 underscored by mounting evidence of widespread insect decline. Nevertheless, answers to basic 31 questions about the spatiotemporal patterns of insecticide use remain elusive, due in part to the 32 inherent complexity of insecticide use, and exacerbated by the dispersed nature of the relevant 33 data, divided between several government repositories. Here, we integrate these public datasets 34 to generate county-level annual estimates of total 'insect toxic load' (honey bee lethal doses) for 35 insecticides applied in the US between 1997-2012, calculated separately for oral and contact 36 toxicity. To explore the underlying drivers of the observed changes, we divide insect toxic load 37 into the components of extent (area treated) and intensity (application rate x potency). We show 38 that while contact-based insect toxic load remained relatively steady over the period of our 39 analysis, oral-based insect toxic load increased roughly 9-fold, with reductions in application rate 40 outweighed by disproportionate increases in potency (toxicity/kg) and increases in extent. This 41 pattern varied markedly by region, with the greatest increases seen in Heartland and Northern 42Great Plains regions, likely driven by use of neonicotinoid seed treatments in corn and soybean. 43In this "potency paradox," US farmland has become more hazardous to insects despite lower 44 volumes of insecticides applied, raising serious concerns about insect conservation and 45 highlighting the importance of integrative approaches to pesticide use monitoring. 46 47 48 the potency (toxicity/kg) of insecticides applied and in the area treated; the volume of 56 insecticides applied declined. Toxic load increased most dramatically in regions where 57 neonicotinoid seed treatments for field crops are commonly used. 58 59 395