Estimate the crop-choice selection equation and compute the Inverse Mills Ratio

Step 1 of Koundouri & Nauges (2005). Fits a probit on a binary selection indicator (e.g., 1 = farmer grew vegetables, 0 = otherwise) and returns the Inverse Mills Ratio (IMR) for every observation in `data`, ready to be plugged into the Step-2 mean function.

Usage

estimate_selection(data, selection_var, covariates)

Arguments

data: A data frame containing the selection indicator and the covariates listed in `covariates`.
selection_var: Character. Name of the binary 0/1 selection variable.
covariates: Character vector of explanatory variable names.

Value

A list with:

model: The fitted probit (a `glm` object).
imr: Inverse Mills Ratio \(\phi(x'\beta)/\Phi(x'\beta)\) per row.
probabilities: Predicted probability of selection per row.

Examples

farms <- simulate_kiti_data(seed = 1)
sel <- estimate_selection(farms, "vegetables",
                          c("rainfall","irrigated","dist_town",
                            "dist_coast","experience"))
head(sel$imr)
#>         1         2         3         4         5         6 
#> 1.6969300 0.9725054 0.8323618 0.6767589 0.9008406 1.0675290