Alcohol Availability

1
Alcohol Availability Alcohol ConsumptionNew
Evidence from Sunday Sales Restrictions

• Kitt Carpenter (UC Irvine)
• Daniel Eisenberg (University of Michigan)
• Comments welcome (kittc_at_uci.edu)

2
Motivation

• Large body of evidence in economics and public
  health that links alcohol availability and
  alcohol consumption.
• Availability as measured by
• Prices, taxes
• Age restrictions
• Proximity to liquor stores

3
Our Paper Sunday Sales Policies

• All states allow alcohol to be purchased on
  Sundays for on-premise consumption (e.g. at bars
  restaurants).
• States and provinces vary as to whether alcohol
  can be purchased on Sundays for off-premise
  consumption (e.g. at home).
• Some have no restrictions
• Some prohibit entirely
• Some allow localities to decide

4
Sunday Sales Map 2006 (APIS)
5
Background

• Blue laws have been around since before the
  Revolutionary war.
• Prohibited shopping, working, or consuming
  alcohol on Sundays.
• Enforcement unclear, but strong support around
  Prohibition era.

6
Policy Relevance

• Since 2002, 12 states have repealed their bans on
  off-premise Sunday sales, trying to increase
  state tax revenues.
• Possible unintended effects
• What if Sunday sales restrictions do not affect
  overall sales/consumption? (e.g. if they simply
  shift the within-week distribution?)
• What if bar goers substitute toward home
  drinking?
• If consumption does increase on Sundays, what if
  there are negative externalities? (e.g.
  fatalities)

7
Our QuestionDo Blue Laws Affect Drinking?

• We are aware of no empirical evidence on the
  effects of these restrictions on alcohol
  consumption per se.
• This is surprising, since the restrictions
• are widespread (16 states prohibit SS)
• are nontrivial (14 of hours of sale)
• have directly testable implications

8
Related Literature - Fatalities

• McMillan et. al. (2005 AJPH)
• Considers New Mexicos 1995 repeal of its Sunday
  sales ban.
• Finds extremely large fatality increases (42)
• Smith (78, 87, 88a, 88b, 90)
• Uses Australian city/state introduction of Sunday
  trading hours, controls for changes in outcomes
  on other days of the week.
• Finds extremely large fatality increases
  (32-100).

9
Data Requirements For This Study

• Geographic identifiers day-specific alcohol
  consumption outcomes.
• In US NLAES 1992 (N 40,000)
• In Canada NPHS 1994-99 (N 57,000)
• We use the data as repeated cross-sections to
  take advantage of large Ontario buy-in in 1996/97

10
Cross-Section Drinking Model OLS

• Day-specific drinking outcome ? ?1X
  ?2(Sunday sales allowed) ?3Z Jd ?
• X includes race, education, sex, marital status,
  veteran status, employment dummies
• Jd are Census division indicators (US only)
• Z is state ACCRA real beer, wine, and spirits
  prices (US only)

11
Coefficient is Sunday Sales OKUS NLAES 1992
Overall Wed Thur Fri Sat Sun Mon Tues
Any drinks .011 (.018) .014 (.010) .011 (.010) -.002 (.018) -.0004 (.018) .030 (.014) .013 (.008) .015 (.009)
Among current drinkers Among current drinkers Among current drinkers Among current drinkers Among current drinkers Among current drinkers Among current drinkers Among current drinkers Among current drinkers
Any drinks -- .017 (.015) .015 (.01) -.015 (.022) -.015 (.020) .047 (.021) .016 (.012) .020 (.013)
days drank in month -- .055 (.037) .041 (.036) .023 (.080) .037 (.067) .161 (.061) .057 (.033) .063 (.032)
12
Alternative Model Any Drinks
Overall Wed Thur Fri Sat Sun Mon Tues
Sunday sales partly restricted .002 (.025) .003 (.012) .002 (.012) -.021 (.026) -.019 (.025) .022 (.019) .006 (.010) .007 (.011)
Sunday sales not at all restricted .020 (.021) .025 (.012) .022 (.013) .019 (.020) .021 (.019) .038 (.016) .021 (.010) .024 (.011)
13
Coefficient is Sunday Sales OKCanadian NPHS
1994-1999
Overall Wed Thur Fri Sat Sun Mon Tues
Any drinks .022 (.006) .004 (.004) .004 (.004) .007 (.005) .011 (.005) .031 (.005) .008 (.004) .011 (.004)
Among past week drinkers Among past week drinkers Among past week drinkers Among past week drinkers Among past week drinkers Among past week drinkers Among past week drinkers Among past week drinkers Among past week drinkers
Any drinks -- -.003 (.008) -.004 (.008) -.004 (.009) -.004 (.009) .049 (.049) .005 (.008) .011 (.008)
drinks -.206 (.146) -.009 (.026) .011 (.032) -.094 (.041) -.261 (.048) .102 (.035) .016 (.027) .028 (.026)
14
Interpreting Cross-Section Results

• Patterns of coefficients support a causal effect
  of Sunday sales restrictions on point in time
  consumption.
• Modest evidence of Monday/Tuesday spillovers
  Friday/Saturday substitution.
• Estimates consistent with a small but nontrivial
  effect of Sunday sales on overall population
  drinking.

15
Remaining Unobservables?

• What if unobserved characteristics about states
  are correlated not only with overall alcohol
  consumption but also day-specific consumption?
• Religiosity may be correlated both with the
  presence of a Sunday sales restriction and lower
  drinking on Sundays.
• Goal isolate a plausibly exogenous change in
  Sunday-specific availability.

16
Ontarios policy change

• Alcohol sales are heavily regulated by the
  Canadian provincial governments.
• In Ontario, off-premise alcohol sales only
  available at LCBOs (not at supermarkets).
• Prior to 1997, allowed some Sunday sales at a few
  Nov/Dec holidays.
• After 1997, Sunday sales ok.
• No other province changed Sunday alcohol sales
  policy over this period.

17
Diff-in-Diff Drinking Model OLS

• Day-specific drinking outcome ? ?1X
  ?2(After 1997) ?3(Ontario) ?4(After 1997
  Ontario) ?
• X includes race, education, sex, marital status,
  employment dummies
• Robust standard errors clustered on province.

18
Coefficient is Ontario After 97Canadian NPHS
1994-1999
Overall Wed Thur Fri Sat Sun Mon Tues
Any drinks .012 (.013) -.00002 (.010) -.005 (.009) .002 (.011) .004 (.012) .021 (.011) .008 (.009) .014 (.009)
Among past week drinkers Among past week drinkers Among past week drinkers Among past week drinkers Among past week drinkers Among past week drinkers Among past week drinkers Among past week drinkers Among past week drinkers
Any drinks -- -.010 (.018) -.018 (.017) -.006 (.020) -.008 (.019) .031 (.020) .008 (.018) .021 (.018)
drinks .083 (.315) -.068 (.057) -.035 (.061) -.032 (.090) -.035 (.106) .148 (.086) -.007 (.058) .113 (.050)
19
Interpreting the DD Results

• Consistent with a causal effect of Sunday sales
  restrictions on Sunday alcohol consumption
• Modest evidence of effects on overall population
  drinking
• Effect sizes slightly smaller than those implied
  by cross-sectional results
• Relevant subsamples are significant at 5 (prime
  age adults, females)

20
Implications

• We have not evaluated the overall costs/benefits
  of liberalizing Sunday sales policies.
• Main benefits are reductions in inconvenience
  costs.
• Modest consumption effects suggest health costs
  are unlikely to be severe, though this requires
  more research.

21
Next Steps

• Canadian Community Health Survey (2001 and 2003),
  very large samples (100K each)
• Fatal Accident Reporting System (FARS)
• 1992 Do day-specific fatalities mirror our
  day-specific consumption patterns?
• More recently, do repeals of Sunday sales bans
  affect day-specific fatalities?
• Comments welcome (kittc_at_uci.edu)