moderndid 0.0.3

Modern difference-in-differences estimators.

moderndid is a unified Python implementation of modern difference-in-differences (DiD) methodologies, bringing together the fragmented landscape of DiD estimators into a single, coherent framework. This package consolidates methods from leading econometric research and various R packages into one comprehensive Python library with a consistent API.

[!WARNING] This package is currently in active development with core estimators and some sensitivity analysis implemented. The API is subject to change.

Installation

uv pip install moderndid

Or install from source:

uv pip install git+https://github.com/jordandeklerk/moderndid.git

Features

• Multiple DiD estimators — Staggered adoption, doubly robust, continuous treatment, triple difference, and sensitivity analysis
• Fast computation — Polars for data wrangling, NumPy vectorization, and Numba JIT compilation for performance-critical paths
• Native plots — Built on plotnine with full customization support
• Robust inference — Analytical standard errors, bootstrap (weighted and multiplier), and simultaneous confidence bands
• Documentation — https://moderndid.readthedocs.io/en/latest/index.html

Consistent API

All estimators share a unified interface, making it easy to switch between methods:

# Staggered DiD
result = did.att_gt(data, yname="y", tname="t", idname="id", gname="g", ...)

# Triple DiD
result = did.ddd(data, yname="y", tname="t", idname="id", gname="g", pname="p", ...)

# Continuous DiD
result = did.cont_did(data, yname="y", tname="t", idname="id", gname="g", dname="dose", ...)

# Doubly robust 2-period DiD
result = did.drdid(data, yname="y", tname="t", idname="id", treatname="treat", ...)

Example Datasets

Several classic datasets from the DiD literature are included for learning and experimentation:

did.load_mpdta()  # County teen employment (staggered adoption)
did.load_nsw()    # NSW job training program (2-period panel)
did.load_ehec()   # Medicaid expansion (sensitivity analysis)
did.load_engel()  # Household expenditure (continuous treatment)

Quick Start

This example uses county-level teen employment data to estimate the effect of minimum wage increases. States adopted higher minimum wages at different times (2004, 2006, or 2007), making this a staggered adoption design.

import moderndid as did

# County teen employment data
data = did.load_mpdta()

# Estimate group-time average treatment effects
result = did.att_gt(
    data=data,
    yname="lemp",
    tname="year",
    idname="countyreal",
    gname="first.treat",
    est_method="dr",
)
print(result)

The output shows treatment effects for each group (defined by when they were first treated) at each time period:

Reference: Callaway and Sant'Anna (2021)

Group-Time Average Treatment Effects:
  Group   Time   ATT(g,t)   Std. Error    [95% Simult.  Conf. Band]
   2004   2004    -0.0105       0.0232    [ -0.0743,   0.0533]
   2004   2005    -0.0704       0.0319    [ -0.1582,   0.0173]
   2004   2006    -0.1373       0.0356    [ -0.2352,  -0.0394] *
   2004   2007    -0.1008       0.0331    [ -0.1918,  -0.0098] *
   2006   2004     0.0065       0.0241    [ -0.0597,   0.0727]
   2006   2005    -0.0028       0.0202    [ -0.0582,   0.0527]
   2006   2006    -0.0046       0.0179    [ -0.0538,   0.0446]
   2006   2007    -0.0412       0.0211    [ -0.0992,   0.0167]
   2007   2004     0.0305       0.0145    [ -0.0095,   0.0705]
   2007   2005    -0.0027       0.0173    [ -0.0502,   0.0448]
   2007   2006    -0.0311       0.0190    [ -0.0833,   0.0211]
   2007   2007    -0.0261       0.0168    [ -0.0721,   0.0200]
---
Signif. codes: '*' confidence band does not cover 0

P-value for pre-test of parallel trends assumption:  0.1681

Control Group:  Never Treated,
Anticipation Periods:  0
Estimation Method:  Doubly Robust

These group-time effects can be aggregated into an event study to see how effects evolve relative to treatment:

event_study = did.aggte(result, type="dynamic")
print(event_study)

==============================================================================
Aggregate Treatment Effects (Event Study)
==============================================================================

Call:
  aggte(MP, type='dynamic')

Overall summary of ATT's based on event-study/dynamic aggregation:

      ATT      Std. Error     [95% Conf. Interval]
  -0.0772          0.0214     [-0.1191, -0.0353] *


Dynamic Effects:

    Event time   Estimate   Std. Error   [95% Simult. Conf. Band]
            -3     0.0305       0.0151   [-0.0084,  0.0694]
            -2    -0.0006       0.0132   [-0.0346,  0.0335]
            -1    -0.0245       0.0139   [-0.0602,  0.0113]
             0    -0.0199       0.0120   [-0.0508,  0.0109]
             1    -0.0510       0.0172   [-0.0951, -0.0068] *
             2    -0.1373       0.0371   [-0.2326, -0.0419] *
             3    -0.1008       0.0352   [-0.1912, -0.0104] *

------------------------------------------------------------------------------
Signif. codes: '*' confidence band does not cover 0

Control Group: Never Treated
Anticipation Periods: 0
Estimation Method: Doubly Robust
==============================================================================

We can also use built-in plotting functionality to plot the event study results:

did.plot_event_study(event_study)

Available Methods

Each core module includes a dedicated walkthrough covering methodology background, API usage, and guidance on interpreting results.

Core Implementations

Module	Description	Reference
moderndid.did	Staggered DiD with group-time effects	Callaway & Sant'Anna (2021)
moderndid.drdid	Doubly robust 2-period estimators	Sant'Anna & Zhao (2020)
moderndid.didhonest	Sensitivity analysis for parallel trends	Rambachan & Roth (2023)
moderndid.didcont	Continuous/multi-valued treatments	Callaway et al. (2024)
moderndid.didtriple	Triple difference-in-differences	Ortiz-Villavicencio & Sant'Anna (2025)

Planned Development

Module	Description	Reference
moderndid.didinter	Intertemporal DiD with non-absorbing treatment	Chaisemartin & D'Haultfœuille (2024)
moderndid.didml	Machine learning approaches to DiD	Hatamyar et al. (2023)
moderndid.drdidweak	Robust to weak overlap	Ma et al. (2023)
moderndid.didcomp	Compositional changes in repeated cross-sections	Sant'Anna & Xu (2025)
moderndid.didimpute	Imputation-based estimators	Borusyak, Jaravel, & Spiess (2024)
moderndid.didbacon	Goodman-Bacon decomposition	Goodman-Bacon (2019)
moderndid.didlocal	Local projections DiD	Dube et al. (2025)
moderndid.did2s	Two-stage DiD	Gardner (2021)
moderndid.etwfe	Extended two-way fixed effects	Wooldridge (2021), Wooldridge (2023)
moderndid.functional	Specification tests	Roth & Sant'Anna (2023)