Alcohol Availability and Public Health

Causal Evidence from a Post-Rationing Reform

Tomas Reivinger

05 June, 2025

1 Introduction

  • Well established relationship between alcohol and (adverse) effects on health.

  • Big picture: higher consumption of alcohol is associated with worse health.

  • Causal evidence is scarce

  • Quasi-experimental designs are a natural, potential, solution

2 Roadmap

  1. Introduction
  2. Roadmap
  3. Historical Background
  4. Institutional setting
  5. Data
  6. Empirical strategy
  7. Results
  8. Further tests (Robustness)
  9. Discussion

3 Historical Background

Before Bratt:

  • High levels of alcohol consumption (vodka belt)
  • 1922 prohibition referendum

The Bratt system: a system of individual control (approx. 1920-1955)

  • Rationing booklet
  • Additional controls such as individual sanctions

After Bratt: Systembolaget

  • Established in october, 1955.

4 Institutional setting

Guidelines:

  • 1955-82: Main town to nearest store > 40 km; Population > 5000
  • After 1982: Main town to nearest store > 30km
  • Stated goal of ‘a couple’ new store locations (municipalities) per year1

4.1 A snapshot from 1955

  • No delivery-point sales on the municipality \(\times\) year level

5 Data

  • All cause mortality (1968-2023)
    • Municipality of residence
  • Motor vehicle accidents (1985-2023)
    • Municipality of accident
  • Sales in Liters of (1978-2008)
    • 100° alcohol

5.1 Municipality reforms and sample restrictions

  • 900 municipalities in 1968, 288 in 1996 (Source: REGINA, SCB)

  • Mortality data aggregated to 1995 municipal borders

  • Mortality sample restricted to 1968-1996

6 Empirical strategy and design

The population equation of interest is

\[\begin{equation} Y_{mt} = \gamma_m + \lambda_t + \beta_{mt} \text{store}_{mt} + \epsilon_{mt}\, , \label{eq:pop-eq-of-interest} \end{equation}\]

  • Callaway and Sant’Anna (2021)
  • Long differences with base-period \(t=-1\)
  • Time-invariant controls (population size)

7 Results

  1. Sales
  2. Mortality
  3. Motor vehicle accidents

7.1 Staggered adoption schedule (sales)

  • 1st store test uses municipalities with no store as the control group

  • 2nd store test uses municipalities with only one store as the control group1

7.2 1st and 2nd store results

  • 1st store test: positive increase1

  • 2nd store test ATT: 33%2

7.3 Staggered adoption schedule (mortality)

7.4 All-cause mortality

7.5 Population size

7.6 Taking population into account

  1. \(y_{mt} = \frac{\#Deaths}{\#Population}\)
  2. \(\frac{1}{\#Population}\)
  3. Including lagged values of population size

Instead: Callaway and Sant’Anna (2021)

7.7 All-cause mortality adjusting for population size

7.8 MVs

8 SUTVA: definition

Concern: cross-boundary spillover

Test: define municipalities with at least one store as treated when the neighboring municipality opens a store.

  • Comparison group has at least one store but the neighbor does not open a store.

If SUTVA is violated, there should be a negative effect on sales.

8.1 SUTVA: results

There are indeed violations of the SUTVA assumption (ATT = 5%) for the full sample.

Using the 2nd store test sample, there is no detectable violations against SUTVA.

9 Summing up

Caveats:

  • Substitution effect?
    • Moonshine
    • Pickup points1
    • Other sources of alcohol
    • Is all of it substitution or just some?

*\(0.035 \times 6 = 0.21\), compared to the \(0.25\) from the 2nd store test (column 5).

9.1 HonestDiD: Intuition

  • Relative Magnitudes
  • Smoothness Restriction?

Source: https://github.com/Mixtape-Sessions/Advanced-DID/blob/main/Slides/03-violations.pdf

9.2 HonestDiD: Results