Metadata-Version: 2.2
Name: diffindiff
Version: 1.2.2
Summary: diffindiff: Python library for convenient Difference-in-Differences Analyses
Author: Thomas Wieland
Author-email: geowieland@googlemail.com
Description-Content-Type: text/markdown
Requires-Dist: numpy
Requires-Dist: pandas
Requires-Dist: statsmodels
Requires-Dist: matplotlib
Requires-Dist: datetime
Requires-Dist: scikit-learn
Requires-Dist: xgboost
Requires-Dist: lightgbm
Dynamic: author
Dynamic: author-email
Dynamic: description
Dynamic: description-content-type
Dynamic: requires-dist
Dynamic: summary

# diffindiff: Difference-in-Differences (DiD) Analysis Python Library

This Python library is designed for performing Difference-in-Differences (DiD) analyses in a convenient way. It allows users to construct datasets, define treatment and control groups, and set treatment periods. DiD model analyses may be conducted with both datasets created by built-in functions and ready-to-use external datasets. Both simultaneous and staggered adoption are supported. The library allows for various extensions, such as two-way fixed effects models, group- or individual-specific effects, and post-treatment periods. Additionally, it includes functions for visualizing results, such as plotting DiD coefficients with confidence intervals and illustrating the temporal evolution of staggered treatments.

The diffindiff package is currently being extensively revised (v2.0.0 and following).


## Author

Thomas Wieland [ORCID](https://orcid.org/0000-0001-5168-9846) [EMail](mailto:geowieland@googlemail.com) 


## Features

- **Data preparation and pre-analysis**: 
  - Define custom treatment and control groups as well as treatment periods
  - Create ready-to-fit DiD data objects
  - Create predictive counterfactuals
- **DiD analysis**: 
  - Perfom standard DiD analysis  
  - Model Extensions:
    - Staggered adoption
    - Two-way fixed effects models
    - Group- or individual-specific treatment effects
    - Group- or individual-specific time trends
    - Including covariates
    - After-treatment period
    - Triple Difference (DDD)
    - Multiple treatments
    - Own counterfactuals
- **Visualization**: e.g.
  - Plot observed and expected time course of treatment and control group
  - Plot expected time course of treatment group and counterfactual
  - Plot model coefficients with confidence intervals
  - Plot individual or group-specific treatment effects with confidence intervals
  - Visualize the temporal evolution of staggered treatments
- **Diagnosis tools**: e.g.
  - Test for control conditions
  - Test for type of adoption
  - Test whether the panel dataset is balanced
  - Test for parallel trend assumption
  - Placebo test


## Literature

  - Baker AC, Larcker DF, Wang CCY (2022) How much should we trust staggered difference-in-differences estimates? *Journal of Financial Economics* 144(2): 370-395. [10.1016/j.jfineco.2022.01.004](https://doi.org/10.1016/j.jfineco.2022.01.004)
  - Card D, Krueger AD (1994) Minimum Wages and Employment: A Case Study of the Fast Food Industry in New Jersey and Pennsylvania. *The American Economic Review* 84(4): 772-793. [JSTOR](https://www.jstor.org/stable/2677856)
  - de Haas S, Götz G, Heim S (2022) Measuring the effect of COVID‑19‑related night curfews in a bundled intervention within Germany. *Scientific Reports* 12: 19732. [10.1038/s41598-022-24086-9](https://doi.org/10.1038/s41598-022-24086-9)
  - Goodman-Bacon A (2021) Difference-in-differences with variation in treatment timing. *Journal of Econometrics* 225(2): 254-277. [10.1016/j.jeconom.2021.03.014](https://doi.org/10.1016/j.jeconom.2021.03.014)
  - Greene WH (2012) *Econometric Analysis*. Chapter 6.2.5.
  - Goldfarb A, Tucker C, Wang Y (2022) Conducting Research in Marketing with Quasi-Experiments. *Journal of Marketing* 86(3): 1-19. [10.1177/00222429221082977](https://doi.org/10.1177/00222429221082977)
  - Isporhing IE, Lipfert M, Pestel N (2021) Does re-opening schools contribute to the spread of SARS-CoV-2? Evidence from staggered summer breaks in Germany. *Journal of Public Economics* 198: 104426. [10.1016/j.jpubeco.2021.104426](https://doi.org/10.1016/j.jpubeco.2021.104426)
  - Li KT, Luo L, Pattabhiramaiah A (2024) Causal Inference with Quasi-Experimental Data. *IMPACT at JMR* November 13, 2024. [AMA](https://www.ama.org/marketing-news/causal-inference-with-quasi-experimental-data/)
  - Olden A, Moen J (2022) The triple difference estimator. *The Econometrics Journal* 25(3): 531-553. [10.1093/ectj/utac010](https://doi.org/10.1093/ectj/utac010)
  - Villa JM (2016) diff: Simplifying the estimation of difference-in-differences treatment effects. *The Stata Journal* 16(1): 52-71. [10.1177/1536867X1601600108](https://doi.org/10.1177/1536867X1601600108)
  - von Bismarck-Osten C, Borusyak K, Schönberg U (2022) The role of schools in transmission of the SARS-CoV-2 virus: quasi-experimental evidence from Germany. *Economic Policy* 37(109): 87–130. [10.1093/epolic/eiac001](https://doi.org/10.1093/epolic/eiac001)
  - Wieland T (2024) Assessing the effectiveness of non-pharmaceutical interventions in the SARS-CoV-2 pandemic: results of a natural experiment regarding Baden-Württemberg (Germany) and Switzerland in the second infection wave. *Journal of Public Health: From Theory to Practice*. [10.1007/s10389-024-02218-x](https://doi.org/10.1007/s10389-024-02218-x)
  - Wooldridge JM (2012) *Introductory Econometrics. A Modern Approach*. Chapter 13.2.


## Examples

```python
curfew_DE=pd.read_csv("data/curfew_DE.csv", sep=";", decimal=",")
# Test dataset: Daily and cumulative COVID-19 infections in German counties

curfew_data=create_data(
    outcome_data=curfew_DE,
    unit_id_col="county",
    time_col="infection_date",
    outcome_col="infections_cum_per100000",
    treatment_group= 
        curfew_DE.loc[curfew_DE["Bundesland"].isin([9,10,14])]["county"],
    control_group= 
        curfew_DE.loc[~curfew_DE["Bundesland"].isin([9,10,14])]["county"],
    study_period=["2020-03-01", "2020-05-15"],
    treatment_period=["2020-03-21", "2020-05-05"],
    freq="D"
    )
# Creating DiD dataset by defining groups and treatment time

curfew_data.summary()
# Summary of created treatment data

curfew_model = curfew_data.analysis()
# Model analysis of created data

curfew_model.summary()
# Model summary

curfew_model.plot(
    y_label="Cumulative infections per 100,000",
    plot_title="Curfew effectiveness - Groups over time",
    plot_observed=True
    )
# Plot observed vs. predicted (means) separated by group (treatment and control)

curfew_model.plot_effects(
    x_label="Coefficients with 95% CI",
    plot_title="Curfew effectiveness - DiD effects"
    )
# plot effects

counties_DE=pd.read_csv("data/counties_DE.csv", sep=";", decimal=",", encoding='latin1')
# Dataset with German county data

curfew_data_withgroups = curfew_data.add_covariates(
    additional_df=counties_DE, 
    unit_col="county",
    time_col=None, 
    variables=["BL"])
# Adding federal state column as covariate

curfew_model_withgroups = curfew_data_withgroups.analysis(
    GTE=True,
    group_by="BL")
# Model analysis of created data

curfew_model_withgroups.summary()
# Model summary

curfew_model_withgroups.plot_group_treatment_effects(
    treatment_group_only=True
    )
# Plot of group-specific treatment effects
```

See the /tests directory for usage examples of most of the included functions.


## Installation

To install the package, use `pip`:

```bash
pip install diffindiff
