Synthetic Difference-in-Differences (Arkhangelsky et al., 2021)

Native synthetic difference-in-differences estimator (the Arkhangelsky et al. 2021 algorithm: ridge-regularized unit weights, simplex time weights, weighted DiD) with bootstrap inference. The engine lives in R/synth_native.R; see also morie_synth_control for the classic Abadie synthetic control with placebo inference.

Usage

morie_did_synthetic(
  data,
  outcome,
  unit,
  time,
  treatment_time,
  treated_units = NULL,
  zeta = NULL,
  n_bootstrap = 200L,
  seed = 42L,
  alpha = 0.05
)

Arguments

data

Balanced panel.

outcome, unit, time, treatment_time

Column names.

treated_units

Optional explicit list of treated unit IDs.

zeta

Retained for back-compat; ignored (the engine derives the SDID regularisation from the data).

n_bootstrap

Bootstrap replications for the SE / CI.

seed

RNG seed.

alpha

Significance level.

Value

A result list; see morie_did_2x2.

References

Arkhangelsky, D., et al. (2021). Synthetic difference-in-differences. American Economic Review, 111(12), 4088–4118.

Examples

set.seed(3)
df <- expand.grid(unit = 1:30, time = 1:8)
df$treat_time <- ifelse(df$unit <= 5, 6, Inf)
df$d <- as.integer(df$time >= df$treat_time)
df$y <- 0.1 * df$time + 0.5 * df$d + rnorm(nrow(df), sd = 0.3)
out <- morie_did_synthetic(df, "y", "unit", "time", "treat_time",
                           n_bootstrap = 50L, seed = 3)
str(out, max.level = 1)
#> List of 10
#>  $ estimate : num 0.598
#>  $ std_error: num 0.0726
#>  $ t_stat   : num 8.25
#>  $ p_value  : num 1.65e-16
#>  $ ci_lower : num 0.456
#>  $ ci_upper : num 0.741
#>  $ n_treated: int 5
#>  $ n_control: int 25
#>  $ method   : chr "synthetic_did (rmorie native)"
#>  $ details  :List of 3