fcipDemand — Tools to estimate Federal Crop Insurance Program (FCIP) demand

• 📖 Introduction
• ✨ Key Features
• 📦 Installation
• 🚀 Quick Start
• 🧮 Example 1: Full Sample Estimation
• 🧩 Example 2: Subsample Estimation
• 📉 Example 3: Subsample Estimation with Non-Positive Elasticities
• 🧠 Example 4: Estimation with User-Added Variables
• 📚 Citation
• 🤝 Contributing
• 📬 Contact

📖 Introduction

rfcipDemand provides a reproducible pipeline for analyzing U.S. Federal Crop Insurance Program (FCIP) demand.
It enables flexible modeling of participation, coverage, and elasticity across counties, crops, and insurance plans.

Its empirical foundations build on:

• Tsiboe, F., & Turner, D. (2023). The crop insurance demand response to premium subsidies: Evidence from U.S. Agriculture. Food Policy, 119(3)
• Tsiboe, F., & Turner, D. (2023). Econometric identification of crop insurance participation. Agricultural and Resource Economics Review

---

✨ Key Features

• Builds harmonized FCIP panel datasets across crops, years, and insurance plans.
• Estimates multi-equation demand systems with endogenous premium rates.
• Supports heterogeneity and calibration to maintain economic consistency.
• Provides diagnostics for instrument validity and elasticity bounds.
• Integrates easily with other ARPC and NIFA-supported research pipelines.

Disclaimer: This package uses USDA data but is not endorsed by or affiliated with USDA or the Federal Government.
See LICENSE for terms.

---

📦 Installation

# Install from GitHub
if (!requireNamespace("devtools", quietly = TRUE)) install.packages("devtools")
devtools::install_github("ftsiboe/rfcipDemand", force = TRUE, upgrade = "never")

---

🚀 Quick Start

The three core functions are:

• fcip_demand_data_dispatcher() → assemble the modeling data
• fcip_demand_sys_estimate() → estimate demand equations
• calibrate_fcip_demand_elasticities() → estimate with non-positive elasticities

---

🧮 Example 1: Full Sample Estimation

In this example, we estimate a baseline FCIP demand system to evaluate how demand changes with percentage changes in the premium rate.
This estimation uses the full sample, following the structure in Tsiboe & Turner (2023) but with updated RMA data.

📦 Click to expand setup code

# library(rfcipDemand)

# 1) Identify fields for panel building
FCIP_INSURANCE_POOL <- c("state_code","county_code","commodity_code","type_code","practice_code")

# 2) Build data (example years - keep short so examples are fast)
df <- fcip_demand_data_dispatcher(
  study_years = 2001:2024,
  identifiers = c("commodity_year", FCIP_INSURANCE_POOL, "insurance_plan_code", "unit_structure_code")
)

# Set price to 1 for crops with no RP/RP-HPE options [NEW]
df[insurance_plan_code %in% c(1L, 90L), insurance_plan_code := 1L]
df[insurance_plan_code %in% c(44L, 2L), insurance_plan_code := 2L]
df[insurance_plan_code %in% c(25L, 42L, 3L), insurance_plan_code := 3L]
df[, rp_eligible := max(as.numeric(insurance_plan_code %in% 2:3)), by = "commodity_code"]
df[rp_eligible == 0, price := 1]
  
# 3) Prep variables 
data                <- as.data.frame(df)
data$net_reporting_level_amount <- log(data$net_reporting_level_amount/10000)
data$coverage_level_percent_aggregate <- log(data$coverage_level_percent_aggregate)
data$rate           <- log(data$premium_per_liability*(1-data$subsidy_per_premium))
data$county_acreage <- log(data$county_acreage/10000)
data$rent           <- log(data$rent/1000)
data$price          <- log(data$price)
data$tauS0          <- log(data$tau*(1-((data$subsidy_rate_65+data$subsidy_rate_75)/2)))
data$trend          <- data$commodity_year - min(data$commodity_year, na.rm=TRUE)

for(i in unique(data$commodity_code)){ data[,paste0("Crop_",i)] <- ifelse(data$commodity_code %in% i,1,0)*data$trend }
for(i in unique(data$commodity_year)){ data[,paste0("year_",i)] <- ifelse(data$commodity_year %in% i,1,0) }
data <- data[names(data)[!names(data) %in% c(paste0("year_",max(data$commodity_year,na.rm=T)),"Crop_41")]]

Estimate model

# Specify the system
model <- list(
  name       = "demo_sys",
  FE         = TRUE,
  outcome    = c("net_reporting_level_amount","coverage_level_percent_aggregate"),
  endogenous = "rate",
  excluded   = "tauS0",
  partial    = c("trend", names(data)[grepl("Crop_", names(data))],
                 names(data)[grepl("year_", names(data))]),
  disag      = NULL,
  included   = c("county_acreage","price","rent")
)

# Estimate demand system
res <- fcip_demand_sys_estimate(model = model, data = data)

# Save results
write.csv(res, "data-raw/examples/example1.csv", row.names = FALSE)

Discussion:
The premium rate enters with a negative sign as expected, indicating that higher producer-paid premiums reduce insurance demand.
County acreage remains the strongest driver of participation across all margins. For more detailes on the resulst see Example 1 — Discussion of Results —

🧩 Example 2: Subsample Estimation

This example explores heterogeneity in demand response across subgroups—specifically by commodity using the variable CROP.

# Specify the system
model <- list(
  name       = "demo_sys",
  FE         = TRUE,
  outcome    = c("net_reporting_level_amount","coverage_level_percent_aggregate"),
  endogenous = "rate",
  excluded   = "tauS0",
  partial    = c("trend", names(data)[grepl("year_", names(data))]),
  disag      = "CROP",
  included   = c("county_acreage","price","rent")
)

# Estimate demand system
res <- fcip_demand_sys_estimate(model = model, data = data)

# Save results
write.csv(res, "data-raw/examples/example2.csv", row.names = FALSE)

Note: This setup allows estimation of separate elasticities and participation effects for each commodity group.

---

📉 Example 3: Subsample Estimation with Non-Positive Elasticities

This example estimates demand heterogeneity while constraining elasticities to be non-positive, ensuring consistency with the law of demand.

# Prep data
df2 <- fcip_demand_data_dispatcher(
  study_years = 2001:2024,
  identifiers = c("commodity_year", FCIP_INSURANCE_POOL, 
                  "insurance_plan_code", "unit_structure_code")
)

# Specify and Estimate demand system
res <- calibrate_fcip_demand_elasticities(
  calibration_year  = 2024,
  estimation_window = 5,
  data              = df2,
  disaggregate      = "CROP"
)

# Save results
write.csv(res, "data-raw/examples/example3.csv", row.names = FALSE)

Insight:
Capping elasticities between –2 and 0 ensures stable simulation behavior and adherence to demand theory.

---

🧠 Example 4: Estimation with User-Added Variables

fcip_demand_data_dispatcher() generates a standard analysis-ready dataset.
Users can merge additional variables of interest (e.g., external climate or policy data) by matching on identifiers such as state_code, county_code, and commodity_year.

If a new variable (ramble) is: - Exogenous → add under included - Endogenous → add under endogenous and specify instruments under excluded

# Prep data
exampleX_data <- unique(data[c("state_code", "commodity_year", "county_code")])
set.seed(123)
exampleX_data$ramble <- runif(nrow(exampleX_data), min = 0, max = 1)

data4 <- dplyr::inner_join(data, exampleX_data, by = c("state_code", "commodity_year", "county_code"))

# Specify the system
model <- list(
  name       = "demo_sys",
  FE         = TRUE,
  outcome    = c("net_reporting_level_amount", "coverage_level_percent_aggregate"),
  endogenous = "rate",
  excluded   = "tauS0",
  partial    = c("trend",
                 names(data4)[grepl("Crop_", names(data4))],
                 names(data4)[grepl("year_", names(data4))]),
  disag      = NULL,
  included   = c("county_acreage","price","rent","ramble")
)

# Estimate demand system
res <- fcip_demand_sys_estimate(model = model, data = data4)

# Save results
write.csv(res, "data-raw/examples/example4.csv", row.names = FALSE)

Tip: Adding stochastic variables like ramble helps test robustness and explore model sensitivity under simulated shocks.

---

📚 Citation

If you use rfcipDemand in your research, please cite:

• Tsiboe, F., & Turner, D. (2023). The crop insurance demand response to premium subsidies: Evidence from U.S. Agriculture.
  Food Policy, 119(3). https://doi.org/10.1016/j.foodpol.2023.102505

• Tsiboe, F., & Turner, D. (2023). Econometric identification of crop insurance participation.
  Agricultural and Resource Economics Review. https://doi.org/10.1017/age.2023.13

---

🤝 Contributing

Contributions, issues, and feature requests are welcome.
Please see the Code of Conduct.

---

📬 Contact

Questions or collaboration ideas?
Email Francis Tsiboe at ftsiboe@hotmail.com.
⭐ Star this repository if you find it useful!