Lambcast

Montgomery County Bag Tax: A Difference-in-Differences Analysis

In January 2012, Montgomery County, Maryland imposed a $0.05 tax on store-provided disposable bags -- the same tax Washington, D.C. had implemented two years earlier. Arlington County, Virginia debated but never passed similar legislation. This project uses a difference-in-differences design to estimate the causal effect of the tax on consumer bag-use behavior, exploiting the natural experiment created by the policy's staggered adoption across the DMV region.

The Economic Problem

Disposable bags impose environmental and waste-management costs not reflected in their market price -- a classic negative externality. A Pigovian tax corrects this by raising the private cost of the externality-generating behavior, ideally by the amount of external damage caused. The $0.05 tax is small relative to most purchases, which raises an immediate theoretical question: is it large enough to change behavior?

Utility Theory: When Does the Tax Work?

Consider a shopper deciding whether to use a store-provided bag or bring their own reusable bag b. Let c be the personal cost of bringing a reusable bag (negative if inconvenient, positive if it generates satisfaction), and w be wealth. Without any financial incentive, the shopper maximizes:

U_N(w, b) = u(w) − b·c

Under risk-neutrality -- a reasonable approximation when the tax is small relative to wealth -- the utility function is linear: u(w) = γw. With a tax t on store-provided bags:

U_T(w, b) = u(w − (1−b)·t) − b·c

The shopper brings a reusable bag when U(b=1) ≥ U(b=0), which under risk-neutrality simplifies to c ≤ γt. The tax works by expanding the set of consumers for whom reusable bags are utility-maximizing -- everyone whose personal cost of bringing a bag falls below the tax value. A subsidy produces the same threshold through a different mechanism: raising the payoff to bringing a bag rather than raising the cost of not doing so. Whether $0.05 is large enough to produce measurable aggregate effects is an empirical question.

Data

Survey data from 16 stores across Montgomery County (MD), Washington, D.C., and Arlington County (VA) for October 2011 through March 2012 -- three months before and three months after the Maryland tax took effect. The dataset contains 16,251 shopper-level observations recording bag type used, quantity, and demographic characteristics. Three outcomes are measured: the extensive margin (whether a shopper used each bag type), the intensive margin (how many bags conditional on using any), and per-person demand (total bags per shopper, combining both margins).

* Import data
import excel "DMV bag tax-2.xls", sheet("countdata") firstrow

* Summary statistics
estpost summarize
esttab using "summary.rtf", cells("count mean sd min max") replace

* Create categorical state variable
generate state = .
replace state = 1 if dc == 1
replace state = 2 if md == 1
replace state = 3 if va == 1
label define statelab 1 "DC" 2 "MD" 3 "VA"
label values state statelab

* Extensive margin means by state and period
tabstat useplastic usereuse nobags, by(post) stat(mean) nototal
tabstat useplastic usereuse nobags if md==1, by(post) stat(mean) nototal
tabstat useplastic usereuse nobags if va==1, by(post) stat(mean) nototal

Identification Strategy

The difference-in-differences estimator compares the pre-to-post change in Maryland (treatment) to the pre-to-post change in the control group, netting out time trends common to both. Washington, D.C. -- already taxed and in behavioral equilibrium throughout the study period -- serves as the primary reference group. Virginia provides an additional untreated control. The core specification is:

y = β₀ + β₁·MD + β₂·VA + β₃·Post + β₄·(Post × MD) + Xβ + ε

The coefficient of interest is β₄ on the interaction term postXmd, capturing Maryland's post-tax change relative to D.C.'s change over the same period. Covariates include race, gender, time-of-day, organic store indicator, and whether a bonus rebate was offered. The critical assumption is parallel trends: in the absence of the tax, Maryland would have followed the same trajectory as D.C. Virginia's near-zero changes across all outcomes over the same period supports this assumption.

Descriptive Evidence

Virginia shows no meaningful movement over the same period: plastic use falls from 82.2% to 80.8%, reusable use rises from 16.3% to 17.2% -- changes well within normal variation. D.C. is similarly flat. Maryland's sharp behavioral shift is therefore unlikely to reflect a time trend common to all three jurisdictions.

Regression Results

Extensive margin — probability of using each bag type

reg useplastic postXmd post md va black othrace female aft night organic bonus
reg usereuse   postXmd post md va black othrace female aft night organic bonus
reg nobags     postXmd post md va black othrace female aft night organic bonus

Intensive margin — bags used, conditional on using any

reg plastic_int postXmd post md va black othrace female aft night organic bonus
reg reuse_int   postXmd post md va black othrace female aft night organic bonus

Per-person demand — total bags per shopper

reg plastic postXmd post md va black othrace female aft night organic bonus
reg reuse   postXmd post md va black othrace female aft night organic bonus

The tax reduced per-person plastic bag demand by roughly one full bag per shopping trip relative to D.C. -- a large effect given that pre-tax demand in Maryland averaged about 1.9 bags per shopper. Reusable bag demand increased by 0.58 bags per trip. Both estimates are highly significant and consistent across the extensive and intensive margin results.

Interpretation

The results are consistent across all three margins and with Pigovian tax theory: a small per-unit tax on an externality-generating good produces measurable behavioral change even when the nominal amount is low. The mechanism operates on both margins simultaneously -- fewer shoppers use plastic bags at all, and those who continue do so less intensively. Substitution toward reusable bags and no bags suggests the tax redirected behavior rather than simply extracting revenue.

Demographic covariates behave as expected: female shoppers substitute more readily toward reusable bags, store rebate incentives shift behavior at the margin, and organic stores show markedly lower plastic demand -- likely reflecting both shopper selection and smaller basket sizes. These patterns add credibility to the identification by confirming the controls capture real behavioral heterogeneity rather than noise.

The primary limitation is the parallel trends assumption. While Virginia's stability provides supporting evidence, the three jurisdictions differ demographically and in baseline bag-use rates. A longer pre-treatment panel would allow an event study specification to formally test whether pre-trends were parallel -- a natural extension with richer data.