Comparison Table

There is certainly many more nuances to each estimator, hence this is a simplified version of the table.

Event Study Estimators
Estimator R function Type Comparison group Main Assumptions Covers use of pre-treatment covariates Uniform Inference
Gardner (2021) `did2s` Imputes Y(0) Not-yet- and Never-treated
  • Correct specification of Y(0)
  • Parallel Trends for all units
  • Limited anticipation*
Borusyak, Jaravel, and Spiess (2021) `didimputation` Imputes Y(0) Not-yet- and Never-treated
  • Correct specification of Y(0)
  • Parallel Trends for all units
  • Limited anticipation*
Callaway and Sant'Anna (2021) `did` 2x2 Aggregation Either Not-yet- or Never-treated
  • Correct specification of evolution of Y(0) or Propensity Score
  • Parallel Trends for Not-yet- or Never-treated
  • Limited anticipation*
Roth and Sant'Anna (2021) `staggered` 2x2 Aggregation Not-yet-treated
  • Treatment timing is random
Sun and Abraham (2020) `fixest::sunab` Interacted TWFE Not-yet- and Never-treated
  • Parallel Trends for all units
  • No Anticipation
* Anticipation can be accounted for by adjusting initial treatment day back x number of years

** Misclassifications are my own fault.