Fig. 3: Lognormal hurdle models run separately for 5 commonly grown organic and conventional crops.
From: Identifying and characterizing pesticide use on 9,000 fields of organic agriculture
Across all five crops, organic fields have a lower probability of using any pesticide active ingredients (a). The effect of organic on pesticide use for fields that do spray is crop-dependent (b). For the second hurdle (b), percent change is calculated from the log-level model as \(100(e^{\beta }-1)\) and standard errors are derived using the delta-method implemented with the nlcom function in Stata. Symbols indicate point estimates (mean) and error bars represent the 95% CI. All models include heteroskedasticity robust standard errors. All models include covariates for field size, farm size, and soil quality, as well as year random intercepts. For the first hurdle, N = 4289 (Carrot), N = 8760 (Grape), N = 4654 (Orange), N = 2804 (Potato), N = 1126 (Onion). For the second hurdle, N = 2766 (Carrot), N = 7678 (Grape), N = 4316 (Orange), N = 2059 (Potato), N = 814 (Onion). Coefficient estimates for all covariates are provided in Supplementary Table 6.
