Replication of Card & Krueger (1994), "Minimum Wages and Employment: A Case Study of the Fast Food Industry in New Jersey and Pennsylvania."
minwage-did/
├── data/
│ ├── public.dat # Raw data (Card & Krueger)
│ └── clean/
│ ├── ck1994_wide.csv # One row per store
│ └── ck1994_long.csv # Long format for regression
├── R/
│ ├── 01_clean.R # Data ingestion and FTE construction
│ ├── 02_analysis.R # DiD regression (fixest)
│ └── 03_plots.R # Visualization
├── output/
│ ├── descriptive_stats.csv
│ ├── regression_table.tex
│ └── figures/
│ └── did_plot.png
└── README.md
Did New Jersey's April 1992 minimum wage increase (from $4.25 to $5.05/hr) reduce fast food employment relative to Pennsylvania, where wages were unchanged?
This project uses a difference-in-differences (DiD) design:
- Treatment group: Fast food stores in New Jersey (NJ)
- Control group: Fast food stores in Pennsylvania (PA)
- Before: February/March 1992 (pre-policy)
- After: November/December 1992 (post-policy)
The DiD estimator compares the change in employment in NJ relative to the change in PA, netting out any time trends common to both states.
The key identifying assumption is that, absent the policy change, employment trends in New Jersey and Pennsylvania would have evolved similarly. Because the data contain only one pre-treatment period, this assumption cannot be directly tested. Its plausibility instead rests on the geographic proximity of the two states, the similarity of the fast-food industry across the border, and the use of comparable chain restaurants in both markets.
- Source: Card, D. & Krueger, A.B. (1994). Original data retrieved from https://davidcard.berkeley.edu/data_sets/njmin.zip
- Sample: The public-use dataset contains 410 fast-food stores. The main estimation sample is a balanced sample of 384 stores with non-missing FTE employment in both periods. This complete-case restriction may introduce a selection concern if missing post-treatment employment is related to store closure or other treatment-relevant outcomes. This replication does not implement a formal attrition analysis, so the estimates should be interpreted as applying to the balanced sample.
- FTE definition: Full-time employees + 0.5 × part-time employees + managers
| NJ (Treated) | PA (Control) | |
|---|---|---|
| Before | 20.43 | 23.38 |
| After | 20.90 | 21.10 |
| Change | +0.47 | −2.28 |
DiD estimate: +2.75 (SE = 1.338, p < 0.05).* Employment in New Jersey rose relative to Pennsylvania after the minimum wage increase. A two-way fixed-effects specification with store and wave fixed effects yields a nearly identical estimate (coef = 2.764*, SE = 1.326), as expected in a two-period DiD setting.
These estimates should be interpreted as applying to the balanced sample of stores with non-missing FTE employment in both periods.
- Parallel trends cannot be directly tested, because the dataset includes only one pre-treatment period.
- Balanced-sample estimation may be sensitive to sample selection, especially if post-treatment missingness reflects store exit, temporary shutdown, or other economically meaningful outcomes. This replication does not formally test whether attrition is systematically related to treatment status.
install.packages(c("tidyverse", "fixest", "here"))
source("R/01_clean.R")
source("R/02_analysis.R")
source("R/03_plots.R")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.