Pesticide use in California agriculture with information on pest control advisors
Data files
Dec 24, 2025 version files 363.21 MB
-
PCAstudy2_metadata-for_anonymized_data.csv
3.09 KB
-
PCAstudyPart1-anonymized.csv
129.59 MB
-
PCAstudyPart2-anonymized.csv
233.58 MB
-
PesticideModels.R
33.11 KB
-
README.md
8.17 KB
Abstract
Pesticides support crop production, enhancing global food security, but are associated with serious environmental and health risks. Factors that promote overuse of pesticides are therefore of great concern. Pest control advisors (PCAs) are agricultural professionals who scout fields for pests and may recommend pesticide applications. We test two hypotheses regarding the influence of PCAs on pesticide use by California farmers, contrasting four groups: independent PCAs, sales PCAs, farm-staff PCAs, and farmer PCAs. The long-discussed conflict of interest hypothesis posits that sales commissions earned by PCAs who work for agricultural chemical retailers (“sales PCAs”) incentivize pesticide use; it predicts elevated use of all pesticides by farmers advised by sales PCAs. The risk aversion hypothesis posits that the risk of damaging pest outbreaks incentivizes pesticide use; it predicts elevated pesticide use when targeting pests that can exhibit outbreaks (arthropods and plant pathogens) but not when targeting non-outbreak pests (weeds). We assembled a dataset of pesticide use on nearly 600,000 crop-years grown in California from 2012 to 2021 by farmers advised by different types of PCAs. Our analysis provides little to no support for the conflict of interest hypothesis; farmers advised by sales PCAs used slightly fewer pesticides than farmers advised by independent PCAs (who receive no sales commission). Instead, our analysis reveals pesticide use consistent with the risk aversion hypothesis, with elevated use of pesticides by one group of PCAs (farm-staff) when managing arthropods and pathogens, but not when managing weeds. Risk aversion, rather than sales commissions, may be shaping pesticide use in California.
Dataset DOI: 10.5061/dryad.8pk0p2p0h
Description of the data and file structure
This dataset, which is taken entirely from publicly available data, describes pesticide use on approximately 1.4 million crop-years grown in California from 2012 to 2021 by farmers advised by different types of pest control advisors (PCAs). Records are for pesticide use in outdoor, commercial agriculture. The dataset was assembled to test whether conflicts of interest among sales PCAs, who work for agricultural chemical distributors and who earn commissions on pesticide sales, are associated with elevated use of pesticides by their client farmers. The values from permit_number and site_code fields were removed to avoid the risk of re-identification.
Files and variables
File: PCAstudy2_metadata-for_anonymized_data.csv
Description: This file defines each of the variables (columns) in the associated data files
File: PCAstudyPart1-anonymized.csv
Description: The overall dataset is too large (ca. 1.4 million rows) to be handled readily by Excel. So, I split the dataset into two parts: part1 contains dataframe rows 1-500,000; part2 contains the remaining rows (500,001 - 1,412,419).
Variables: (please see attached metadata file for all variable definitions)
- log_all_targets: log base 10 of the number of pesticide products applied against all pest targets
- log_weeds: log base 10 of the number of pesticide products applied against weed targets
- log_arthropods: log base 10 of the number of pesticide products applied against arthropod targets
- log_pathogens: log base 10 of the number of pesticide products applied against plant pathogen targets
- log_adjuvants: log base 10 of the total number of adjuvants applied
- log_adjuvants_per_event: log base 10 of the total number of adjuvants applied per spray event (discrete date and time when a pesticide application was made)
- prop_single_reg: proportion of all pesticide products applied that were marketed by a single registrant (these are nearly always patent-protected materials)
- all_target_psyc_n_appl: total number of pesticide products applied against all pest targets (not log transformed)
- psyc_n_appl_plant_growth_regulator: total number of plant growth regulators applied (not log transformed)
- farm_org_staff: proportion of pest control advisor team associated with a given farmer (permit holder) that was farm-staff PCAs
- all_indep: proportion of pest control advisor team associated with a given farmer (permit holder) that was independent PCAs
- sales: proportion of pest control advisor team associated with a given farmer (permit holder) that was sales PCAs
- farmer: proportion of pest control advisor team associated with a given farmer (permit holder) that was farmer PCAs
- coname: county name
- Year: year
- area_planted: size of the field (hectares)
- farm_size: size of the farm (hectares)
- is_organic: the field is being managed conventionally (is_organic = 0) or organically (is_organic = 1)
- log_weeds_kg: kg of formulated pesticide product applied that target weeds
- log_arthropods_kg: kg of formulated pesticide product applied that target arthropods
- log_pathogens_kg: kg of formulated pesticide product applied that target plant pathogens
- log_all_targets_kg: kg of formulated pesticide product applied that target any pest group
- log_weeds_ai: kg of active ingredients (AI) of pesticide product applied that target weeds
- log_arthropods_ai: kg of active ingredients (AI) of pesticide product applied that target arthropods
- log_pathogens_ai: kg of active ingredients (AI) of pesticide product applied that target plant pathogens
- log_all_targets_ai: kg of active ingredients (AI) of pesticide product applied that target any pest group
- log_adjuvants_kg: log base 10 kg of formulated adjuvants applied
- total_kg_ha_adj: kg of formulated adjuvants applied
- log_adjuvants_kg_per_event: log base 10 kg of formulated adjuvants applied per spray event (discrete date and time when a pesticide application was made)
- total_kg_ha_pgr: kg of formulated plant growth regulators applied
- total_kg_ha_ai_pgr: kg of active ingredients (AI) of plant growth regulators applied
File: PCAstudyPart2-anonymized.csv
Description: The overall dataset is too large (ca. 1.4 million rows) to be handled readily by Excel. So, I split the dataset into two parts: part1 contains dataframe rows 1-500,000; part2 contains the remaining rows (500,001 - 1,412,419). Variables and definitions are the same as for Part 1.
Variables
- log_all_targets: log base 10 of the number of pesticide products applied against all pest targets
- log_weeds: log base 10 of the number of pesticide products applied against weed targets
- log_arthropods: log base 10 of the number of pesticide products applied against arthropod targets
- log_pathogens: log base 10 of the number of pesticide products applied against plant pathogen targets
- log_adjuvants: log base 10 of the total number of adjuvants applied
- log_adjuvants_per_event: log base 10 of the total number of adjuvants applied per spray event (discrete date and time when a pesticide application was made)
- prop_single_reg: proportion of all pesticide products applied that were marketed by a single registrant (these are nearly always patent-protected materials)
- all_target_psyc_n_appl: total number of pesticide products applied against all pest targets (not log transformed)
- psyc_n_appl_plant_growth_regulator: total number of plant growth regulators applied (not log transformed)
- farm_org_staff: proportion of pest control advisor team associated with a given farmer (permit holder) that was farm-staff PCAs
- all_indep: proportion of pest control advisor team associated with a given farmer (permit holder) that was independent PCAs
- sales: proportion of pest control advisor team associated with a given farmer (permit holder) that was sales PCAs
- farmer: proportion of pest control advisor team associated with a given farmer (permit holder) that was farmer PCAs
- coname: county name
- Year: year
- area_planted: size of the field (hectares)
- farm_size: size of the farm (hectares)
- is_organic: the field is being managed conventionally (is_organic = 0) or organically (is_organic = 1)
- log_weeds_kg: kg of formulated pesticide product applied that target weeds
- log_arthropods_kg: kg of formulated pesticide product applied that target arthropods
- log_pathogens_kg: kg of formulated pesticide product applied that target plant pathogens
- log_all_targets_kg: kg of formulated pesticide product applied that target any pest group
- log_weeds_ai: kg of active ingredients (AI) of pesticide product applied that target weeds
- log_arthropods_ai: kg of active ingredients (AI) of pesticide product applied that target arthropods
- log_pathogens_ai: kg of active ingredients (AI) of pesticide product applied that target plant pathogens
- log_all_targets_ai: kg of active ingredients (AI) of pesticide product applied that target any pest group
- log_adjuvants_kg: log base 10 kg of formulated adjuvants applied
- total_kg_ha_adj: kg of formulated adjuvants applied
- log_adjuvants_kg_per_event: log base 10 kg of formulated adjuvants applied per spray event (discrete date and time when a pesticide application was made)
- total_kg_ha_pgr: kg of formulated plant growth regulators applied
- total_kg_ha_ai_pgr: kg of active ingredients (AI) of plant growth regulators applied
Code/software
This R script (PesticideModels.R) describes the analyses performed in the paper. The script systematically evaluates how PCA team composition and farm characteristics influence pesticide use patterns across different crops, pesticide types, and metrics (number of applications, kg product, kg AI, adjuvants, PGRs).
Access information
Other publicly accessible locations of the data:
- California Department of Pesticide Regulation, California Association of County Agricultural Commissioners and Sealers Association,and diverse public websites