Alcohol Consumption and

1
Alcohol Consumption and Body Weight
Johanna Catherine Maclean, University of
Miami Edward C. Norton, UNC at Chapel
Hill Michael T. French, University of
Miami Funding NIAAA (R01-AA13167)
2
Weighty Problem

• Majority of adults in the U.S. are overweight or
  obese
• Rates of overweight and obese have increased over
  the past century
• Increases in past few decades have occurred while
  weight levels at or above healthy levels
• Economic explanations
• Real price of calories has decreased
• Real price of exercise has increased

3
Role of Alcohol

• Alcoholic drinks are high in calories
• Alcohol interferes with metabolism
• Inhibits fat burning
• Alcohol is addictive
• Alcohol impairs judgment
• Note Drinking levels have been stable over the
  past several decades and therefore alcohol cannot
  explain upward trend in average weight status

4
Health and Economic Outcomes

• Published literature associates obesity and
  alcohol use with health and economic outcomes
• Obesity
• Health Heart disease, diabetes, stroke, high
  blood pressure, and some cancers
• Labor Market Worse labor market performance
  (e.g., employment, wages)
• Alcohol
• Health/Labor Market
• Non-linear effects
• Moderate drinkers more productive and earn higher
  wages
• Heavy/problematic drinkers perform worse

5
Does Government Have a Role?

• Economists suggest that government has limited
  role in fight against obesity if consumers are
  well informed
• Are they?
• Nutritional information on food products?
• Relationship between calories in/calories out
  well understood?
• Taxing calories is difficult
• Hard to tax only the calories that make you fat
• Distributional issues
• If alcohol impairs judgment and/or is addictive,
  larger role for government may exist
• Alcohol may affect consumer rationality

6
Conceptual Framework (1)

• Alcoholic beverages are high in calories
• 12-oz Can of Beer 145 calories
• 4-oz Glass of Wine 108 calories
• 1.5-oz Shot of Spirits 130 calories
• 12-oz Can of Soda 144 calories
• Alcohol inhibits fat burning
• Ceteris paribus, higher alcohol consumption
  should increase BMI

7
Conceptual Framework (2)

• Alcohol may be a compliment for activities that
  promote weight gain (e.g., watching television)
• Increase weight
• May crowd out other calories
• Decrease weight
• May impair judgment
• Increase/decrease weight
• Relationship is theoretically indeterminate
• Requires empirical investigation

8
Conceptual Framework (3)

• Males
• More likely to consume alcohol
• Drink more if they do consume
• Cause more social problems as a result of their
  drinking
• Females
• Biological reasons for greater alcohol effect
• Physically smaller on average
• Labor market effects of obesity more severe
• Higher rates of extreme body weights (at both
  tails)
• Age
• Alcohol may have a cumulative effect over time

9
Endogeneity of Alcohol Use

• Omitted variables
• If alcohol use is strictly exogenous in WS
  equations ? single-equation models can generate
  consistent estimates
• If (unobservable or hard to measure) variables
  are omitted from WS equations that are correlated
  with both alcohol use and WS ? single-equation
  models may generate biased estimates
• Examples financial stress, professional
  ambition, all sources of income
• Direction of bias theoretically indeterminate ?
  depends on nature of omitted variable(s) and on
  correlations among the covariates
• Potential solution ? estimate WS and alcohol use
  equations simultaneously

10
Data

• Wave 1 of the National Epidemiological Survey of
  Alcohol and Related Conditions (2001/2002)
• 43,093 respondents
• Overall survey response rate 81 percent ?
  comparable to other co-morbidity surveys
• Comprehensive alcohol use measures
• Self-reported weight and height
• Geographic identifiers
• Sample
• Nationally representative of civilians
• Full analysis sample 30,438
• Age range 21-65 years

11
Measures

• Weight status
• Body Mass Index (BMI) ? (weight in kg) (height
  in cm)2
• Overweight or obese ? BMI gt 25
• Obese ? BMI gt 30
• Alcohol use
• Weekly or more frequent alcohol use (frequency)
• Weekly or more frequent binge drinking (binging)
• DSM diagnosis of alcohol abuse and/or dependence
  (abuse)
• Standard covariate set
• Age, race, ethnicity, birth outside the U.S.,
  education, income, marital status, employment,
  state prevalence of overweight or obese

12
Descriptive Statistics
13
Hypotheses

• Alcohol use increases WS
• Effect of alcohol may not be consistent across
  all groups
• Stronger for
• Higher weight groups
• Women
• Older individuals (i.e., over age 40)
• Alcohol use endogenous in any model of WS

14
Methods (1)

• All models gender specific
• Three WS measures x three alcohol use measures x
  two genders
• 18 gender-specific models
• Estimate a sub-set of gender- and age-specific
  models
• 12 gender- and age-specific models
• Standard errors adjusted for clustering at the
  state level in all models
• Core regression model
• WS ?0 ?1A ?2X ?
• A alcohol use
• X vector of all other exogenous variables
• Function is probit or OLS

15
Methods (2)

• Estimates from core model biased if variables
  that influence WS and alcohol are omitted from WS
  equation ? estimate WS and alcohol equations
  simultaneously
• Recursive bivariate probit model (RBVP)
• Treatment-effects regression model (TERM)
• WS ?0 ?1A ?2X ?
• A ?0 ??IV ?2X ?
• A, X as before
• IV Instrument Variables
• State excise beer tax per 12 ounce drink
• State percentage of residents living in dry
  counties
• State population per alcohol outlet
• State prohibits the use of credit cards in
  off-premise alcohol outlets
• State bans consumption of alcoholic beverages in
  a motor vehicle
• Valid IVs must
• Jointly predict significant variation in alcohol
  use
• Correctly excludable from WS equations (Bollen,
  Guilkey and Mroz 1995 Rashad and Kaestner, 2004)

16
Instrumental Variables Males
17
Instrumental Variables Females
18
Single-Equation Estimation Results
19
Simultaneous-Equation Estimation Results
20
Age-Specific Estimation Results for BMI
21
Sensitivity Analysis

• Estimated parsimonious models
• Included only exogenous RHS
• i.e., age, gender, race, ethnicity, and birth
  outside the U.S.
• Alternative IV sets
• e.g., per capita alcohol sales, bans on sales of
  alcohol on Sundays
• Results qualitatively the same as those generated
  in core models

22
Summary

• First study to examine the alcohol-WS
  relationship using economic framework and
  econometric techniques
• Results generally confirmed our five hypotheses
• Among women and older men alcohol use positively
  associated with elevated WS after correcting for
  omitted variable bias
• Among younger men alcohol use protective of
  elevated WS
• Alcohol-WS relationship not consistent across
  weight spectrum, gender, or age
• Alcohol use endongenous in over half the
  estimated models

23
Limitations

• Height and weight self-reported
• Objective physical measurements preferable
• Individuals tend to over-estimate height and
  underestimate weight ? downwardly biased
  estimates of WS
• Systematic reporting bias ? individuals at the
  extreme tails of the WS spectrum
  disproportionately misreport or refuse to report
  height/weight
• Alcohol use self-reported
• Objective confirmation preferable
• Literature suggests self-reported estimates
  reliable for statistical analyses (e.g., Del Boca
  and Darkes, 2003 Friesema et al., 2004)

24
Discussion

• Alcohol a popular and high calorie food product
• Potential target area in fight against elevated
  WS
• Governments have used alcohol policies to combat
  other risks associated with alcohol use
• e.g., motor vehicle crashes
• This study identifies a strong relationship
  between alcohol use and elevated WS among women
  and older, but not younger men
• Any government intervention designed to lower
  average WS through reduced alcohol consumption
  should consider these findings

25

• If you are young and you drink a great deal it
  will spoil your health, slow your mind, make you
  fatin other words, turn you into an adult.
• P.J. O'Rourke