A Python replication of Card & Krueger's seminal 1994 paper, "Minimum Wages and Employment: A Case Study of the Fast-Food Industry in New Jersey and Pennsylvania".
For the full write-up with figures, equations, and regression output, see the GitHub Pages site or the corresponding Substack post.
Prior to the early 1990s, over 90% of economists agreed that minimum wage laws reduced employment among low-skilled workers (Kearl et al., 1979). Card & Krueger's natural experiment challenged this consensus by surveying 410 fast-food restaurants in New Jersey and eastern Pennsylvania before and after New Jersey raised its minimum wage from $4.25 to $5.05 on April 1, 1992.
Their key finding: no statistically significant negative effect on employment from the minimum wage increase, contrary to standard competitive labour market predictions. This replication reproduces the core tables and regressions from the paper using Python.
| Paper Section | Description | Method |
|---|---|---|
| Table 2 | Distribution of starting wage rates (Feb & Nov 1992) | Grouped bar charts comparing NJ vs PA |
| Table 3 | Mean FTE employment before/after, DiD estimator | Sample means, balanced & unbalanced |
| Table 4 | Reduced-form models for change in employment (Models 1-5) | OLS with NJ dummy and GAP variable, with/without controls |
| Table 5 | Specification tests (alternative FTE measures) | OLS with reweighted part-time employment |
| DiD Visual | Difference-in-differences graph with counterfactual | Plotted from Table 3 sample means |
- Difference-in-Differences (DiD): Compares employment changes in NJ (treatment) vs PA (control) to estimate the causal effect of the minimum wage increase, relying on the parallel trends assumption.
- OLS Regression: Models range from a simple DiD specification to models with controls for chain, ownership, and region. The GAP variable captures within-NJ variation based on the initial wage distance to the new minimum.
- Specification Tests: Robustness checks using alternative FTE employment definitions (excluding managers, reweighting part-time at 0.4x and 0.6x).
The original dataset is automatically downloaded at runtime from David Card's website. It contains survey data on 410 fast-food restaurants across NJ and eastern PA from two waves (February 1992 and November 1992).
git clone https://github.com/AnthonyPuggs/CardKrueger1994Replication.git
cd CardKrueger1994Replication
pip install -r requirements.txt- Python 3.8+
- pandas
- numpy
- scikit-learn
- statsmodels
- requests
- matplotlib
- seaborn
- scipy
python CardKrueger1994Replication.pyThis will download the data, run all analyses, print regression output to the console, and save the following figures:
Distribution of Starting Wage Rates Feb 1992.pngDistribution of Starting Wage Rates Nov 1992.pngDID.png
- The DiD estimator for change in mean FTE employment is +2.75, indicating NJ stores gained employment relative to PA stores after the minimum wage increase.
- Regression analysis confirms this: the interaction term (state x time) is positive across all specifications.
- Adding controls for chain, ownership, and region does not meaningfully alter the result.
- The GAP variable models show similar positive effects, though Model 5 (with region dummies) attenuates the coefficient.
The replication uses 351 observations in Table 4 compared to 357 in the original paper, due to differences in how missing employment data is handled. See the full write-up for details.
- Card, D. & Krueger, A.B. (1994). Minimum Wages and Employment: A Case Study of the Fast-Food Industry in New Jersey and Pennsylvania. American Economic Review, 84(4), 772-793.
- Hill, R.C., Griffiths, W.E. & Lim, G.C. (2016). Principles of Econometrics, 5th Edition.
- PyMC DiD Example
- BiomedSciAI Card-Krueger Notebook
- Aaronmams R Replication
- Alopatina R Replication