GitHub

Calibration

JCGECalibrate converts SAM and IO data into calibrated parameters and starting values. The calibration process is model-agnostic but assumes a canonical input schema. The goal is a consistent benchmark where accounting identities and zero-profit conditions hold.

Inputs

  • sam.csv: square SAM matrix with labeled rows/columns
  • sets.csv: canonical lists of goods, activities, factors, and institutions
  • Optional: subsets.csv, labels.csv, mappings.csv, params.csv

The Imports guide describes the canonical schema and how to build these files from raw sources.

Outputs

Calibration produces:

  • Parameter tables for production, demand, and tax structures.
  • Starting values for levels and prices.
  • Consistent mappings that align data with model sets.

Typical workflow

using JCGECalibrate
sam = load_sam_table("path/to/sam.csv"; goods=..., factors=...)
start = compute_starting_values(sam)
params = compute_calibration_params(sam, start)

Calibration is deterministic and should produce a consistent benchmark that satisfies model accounting identities.

Validating calibration

Before building the model, check that:

  • All accounts balance.
  • Mappings cover all required sets.
  • Calibration outputs match the expected model inputs.

If a balance check fails, fix the SAM or mappings before proceeding.

Common pitfalls

  • Misaligned labels between sam.csv and sets.csv.
  • Missing or duplicated accounts in the SAM.
  • Implicit zeros that should be explicit.

Next steps