Goodman-Bacon decomposition of the TWFE DiD estimator — morie_did_bacon

Native Goodman-Bacon (2021) decomposition: the two-way fixed-effects DiD estimate is decomposed into a weighted average of all 2x2 timing comparisons, with each pair's weight derived from the Frisch-Waugh-Lovell identity (subsample size squared times the variance of the demeaned treatment). Reproduces bacondecomp::bacon exactly (see tests/cross/).

Usage

morie_did_bacon_decomposition(data, outcome, treatment, unit, time)

Arguments

data: Balanced panel data.
outcome: Outcome column.
treatment: Binary treatment indicator that turns on at onset.
unit: Unit identifier.
time: Time period.

Value

A list with components (data frame) and overall_estimate.

References

Goodman-Bacon, A. (2021). Difference-in-differences with variation in treatment timing. Journal of Econometrics, 225(2), 254–277.

Examples

set.seed(1)
df <- expand.grid(unit = 1:30, time = 1:6)
df$treat_time <- ifelse(df$unit <= 15, 4, Inf)
df$d <- as.integer(df$time >= df$treat_time)
df$y <- 0.1 * df$time + 0.7 * df$d + rnorm(nrow(df), sd = 0.4)
out <- morie_did_bacon_decomposition(df, "y", "d", "unit", "time")
out$overall_estimate
#> [1] 0.6266604