Run Rigobon (2003) Regime-Based Estimation

Implements Rigobon's heteroskedasticity-based identification using discrete regime indicators. This is a wrapper around the Lewbel estimation framework that automatically constructs instruments from regime dummies.

Usage

run_rigobon_estimation(
  data,
  endog_var = "Y2",
  exog_vars = "Xk",
  regime_var = "regime",
  df_adjust = "asymptotic",
  return_diagnostics = FALSE
)

Arguments

data

Data.frame. Must contain Y1, Y2, X variables, and regime indicator.

endog_var

Character. Name of endogenous variable (default: "Y2").

exog_vars

Character vector. Names of exogenous variables (default: "Xk").

regime_var

Character. Name of regime indicator variable (default: "regime").

df_adjust

Character. Degrees of freedom adjustment: "asymptotic" or "finite" (default: "asymptotic").

return_diagnostics

Logical. Whether to return additional diagnostic information (default: FALSE).

Value

If return_diagnostics = FALSE: A list with:

• ols: OLS estimates and standard errors

• tsls: 2SLS estimates and standard errors using Rigobon instruments

• first_stage_F: Vector of F-statistics for each instrument

If return_diagnostics = TRUE: Additionally returns:

• instruments: Matrix of generated instruments

• regime_props: Proportion of observations in each regime

• heteroskedasticity_test: Test for regime-based heteroskedasticity

Details

The function:

1. Creates centered dummy variables for each regime

2. Estimates first-stage residuals

3. Constructs instruments as (centered regime dummy) * (first-stage residual)

4. Performs 2SLS estimation using all regime-based instruments

This implements the procedure described in Rigobon (2003) for identification through heteroskedasticity across discrete regimes.

References

Rigobon, R. (2003). Identification through heteroskedasticity. Review of Economics and Statistics, 85(4), 777-792. doi:10.1162/003465303772815727

See also

generate_rigobon_data for generating regime-based data

Examples

if (FALSE) { # \dontrun{
# Generate Rigobon-style data
params <- list(
  beta1_0 = 0.5, beta1_1 = 1.5, gamma1 = -0.8,
  beta2_0 = 1.0, beta2_1 = -1.0,
  alpha1 = -0.5, alpha2 = 1.0,
  regime_probs = c(0.4, 0.6),
  sigma2_regimes = c(1.0, 2.5)
)
data <- generate_rigobon_data(1000, params)

# Run Rigobon estimation
results <- run_rigobon_estimation(data)
print(results$tsls$estimates)

# With diagnostics
results_diag <- run_rigobon_estimation(data, return_diagnostics = TRUE)
print(results_diag$heteroskedasticity_test)
} # }