Skip to contents

Prepare Chilean Agricultural Census (CAF 2021) data for [jp_fit()]

Source: R/prepare_caf_data.R
prepare_caf_data.Rd

Reads the raw CAF 2021 microdata files as INE actually ships them and returns a single farm-level (`GUID`) data frame ready for [jp_fit()].

Usage

prepare_caf_data(
  data_dir,
  crop_group_map = caf2021_g2_groups(),
  include_household = TRUE,
  include_crop_surface = TRUE,
  verbose = TRUE
)

Arguments

data_dir

Path to the folder that contains the two database subfolders, e.g. `"ChileCensusAg/"`.

crop_group_map

Named list mapping a broad group name to the numeric `G2` codes (default [caf2021_g2_groups()]).

include_household

Logical. Merge the Hogar Agricola tables (covers natural-person producers only – legal-person producers will get NAs for those columns). Default TRUE.

include_crop_surface

Logical. Aggregate the `seccion_9_*` parcel files into farm-level area-per-crop variables. Default TRUE. Skip to read fewer files (`prepare_caf_data` is dominated by I/O).

verbose

Logical. Print progress.

Value

A data frame with one row per `GUID`.

Real file layout (as of 2024 release)

INE distributes two parallel databases. Point `data_dir` at the folder that contains them:


  data_dir/
  |-- Actividad Silvoagropecuaria/
  |   |-- seccion_1.csv               # UPA-level core (G1-G4, SUP_UPA)
  |   |-- seccion_5.csv               # Administrator (ID39, ID40)
  |   |-- seccion_9_cereales.csv      # Surface by crop (one file per group)
  |   |-- seccion_9_hortalizas.csv
  |   |-- seccion_9_frutales.csv
  |   |-- seccion_9_vinas.csv
  |   |-- seccion_11.csv              # Practices: PM213-PM227 (fert, pest)
  |   |-- seccion_12.csv              # Irrigation: AR228-AR229
  |   |-- seccion_13_activos.csv      # AC242 (infrastructure)
  |   |-- seccion_13_maquinaria.csv   # AC230-AC238 (machinery + value)
  |   `-- ...
  `-- Hogar agricola/                 # Natural-person producers only
      |-- gestion.csv                 # TR244, TR250, HP280_2
      |-- actividad_agricola.csv      # US61_*, GA* (land use, livestock)
      |-- seccion_15_hogar.csv        # HP261-HP276 (household chars)
      `-- seccion_15_hogar_oa.csv

CSVs use `;` as separator and UTF-8 with BOM; `read.csv2()` is used internally.

What the census does NOT contain

The CAF is structural. Key gaps for a Just-Pope application:

  • No crop-level yield or production quantity.

  • No farm-level prices.

  • No labour counts in the productive database (TR244 and TR250 in `gestion.csv` are only yes/no indicators for having permanent / temporary workers).

  • No fertilizer/pesticide expenditure – only categorical use indicators (PM213-PM214).

  • HP280_1 (sales band) is not included in the public release; only HP280_2 (band type, 1 = annual / 2 = monthly) is shipped. The function fills `sales_band_code` from HP280_2 but cannot map it to a peso amount without HP280_1.

The closest output proxy available within the census alone is `machinery_value_clp` (a capital stock). For a true yield-based replication, merge external ODEPA data by `CUT_COMUNA`.

Examples

if (FALSE) { # \dontrun{
caf <- prepare_caf_data("ChileCensusAg/")
table(caf$crop_group)

fit <- jp_fit(
  data                 = caf,
  selection_var        = "hortalizas",
  selection_covariates = c("CUT_REGION","admin_age","admin_female",
                           "irrigation_river","irrigation_well"),
  output_var           = "machinery_value_band",       # capital-stock proxy
  input_vars           = c("total_surface_ha","machinery_count",
                           "infrastructure_count","irrigated_ha"),
  shifter_vars         = c("fertilizer_use","pesticide_use","CUT_REGION"),
  bootstrap_reps       = 200
)
} # }