The Economics of Substance Use

1
The Economics of Substance Use

• Don Kenkel
• Cornell University NBER
• Economics of Health and Healthcare
• July 12 18, 2007, European Science Days
• Steyr, Austria

2
Outline Part I

• 1. Introduction
• 2. Theory scientific (positive) models of
  consumer behavior
• 3. Theory welfare economics
• 4. Empirical studies of consumer demand
• Data issues
• Identification

3
Outline Part II

• 5. Empirical studies of the consequences of
  substance use
• 6. Supply side of legal substances
• Pricing of legal substances/ tax pass-thru
• Advertising of legal substances
• 7. Illegal substance markets
• 8. Future directions
• Implications for obesity research

4
Reference Sources

• Grossman, Individual Behaviors and Substance
  Use The Role of Price
• Many references I respect his opinions but dont
  always agree.
• Kenkel, Health Behaviors Among Young People
• Selected references I tend to agree with him
• Other chapters in the Elgar Companion to Health
  Economics
• Cawley Kenkel, The Economics of Health
  Behaviors, Edward Elgar volume of collected
  papers, forthcoming

5
Goals

• Introduce economic models of substance use
• Provide overview of some interesting questions
  being asked in empirical studies
• Focus on research challenges
• Not a round-up / meta-analysis of empirical
  results
• Policy-relevant summaries exist, but should be
  more critical
• My review is incomplete

6
1. Introduction

• Reasons to study the economics of substance use
• Use is widespread
• Short- and long-term health consequences
• Policies prohibitions, taxation, regulation
• Substance use is interesting economics!
• These are the ingredients of an interesting
  important study ( good dissertations
  publications!)

7
World extent of psychoactive substance use
8
Digression How do we define substance use and
abuse

• Psychoactive properties
• Addiction harm to self others?
• Common legal (alcohol tobacco) and illegal
  (marijuana, cocaine, )
• Legal status varies over time place
• What about caffeine?
• What about betel, khat?
• What about use of pharmaceutical products while
  under physicians care?

9
Defining substance abuse

• Abuse is a foreign concept to economic models
  that assume utility maximization
• Psychiatric criteria
• Example DSM IV defines alcohol abuse as
  drinking despite recurrent social,
  interpersonal, and legal problems as a result of
  alcohol use.
• Alcohol dependence involves tolerance and
  withdrawal
• When does use of other substances become abuse?

10
(No Transcript)
11
WHO Regions disease burden in 2000 attributable
to selected risk factors
12
Policies (the tions)

• Health information campaigns
• School-based
• Mass media
• Regulation
• Advertising restrictions
• License/ monopolize sales
• Taxation
• Litigation
• US 1998 settlement with tobacco industry

13
1998 Master Settlement Agreement

• Could you please explain the recent historic
  tobacco settlement?
• Sure. Basically, the tobacco industry has
  admitted that it is killing people by the
  millions, and has agreed that from now on it will
  do this under the strict supervision of the
  federal government.
• Dave Barry

14
Prohibitions

• Marijuana, cocaine, prohibited in most
  countries
• US Noble Experiment prohibited alcohol from
  1920 1933 (also in other countries)
• Legal drinking age partial prohibition
• Between 1895 1921, 14 states in the US banned
  cigarette sales
• Currently, cigarette sales to minors are banned
• Many countries ban smoking in public places
• Worksites
• Restaurants

15
Interesting economic issues

• Simple concepts like price-elasticity of demand
  take on policy-relevance
• Rationality of consumers
• Labor market consequences ? issues at
  intersection of health labor economics
• Health consequences ? issues for private health
  insurance markets (moral hazard, adverse
  selection) and design of public insurance programs

16
2. Theory Consumer Behavior

• Simple model of substance use as health-related
  consumption
• Rational addiction
• Quasi-hyberbolic discounting
• Cue-triggered addiction

17
Simple Model

• Maximize U (C, H, Y) subject to
• Budget constraint
• Health production function H H (C, .)
• First order condition
• UC ? pC UH HC
• Marginal benefit marginal cost, where marginal
  cost reflects utility value of monetary cost of C
  and the health costs (HC)

18
Simple Model

• First order condition implicitly defines demand
  for C as a function of its price, perceived
  health costs, income,
• Price-elasticity, income-elasticity
• Shifters of perceptions of health costs
• Econometric evidence that cigarette demand
  responded to information shocks about the
  health consequences of smoking
• Different people have different information

19
Ex Evolution of the Schooling-Smoking Gradient

• Schooling-smoking gradient only emerged after the
  health hazards were established
• Doesnt prove causality could still be third
  factor like time preference
• Persistence and growth of gradient ? hard to
  reconcile that it can be traced back to
  differences in perceptions of HC

20
(No Transcript)
21
Rational Addiction

• Adjacent complementarity marginal utility of
  consuming substance at time t depends on how much
  was consumed at time t-1
• Myopic addict just looks backwards
• Rational forward-looking addict
• Takes into account that choice of consumption
  today will affect marginal utility of consuming
  the substance next period
• Addictions can be healthy (exercise) or unhealthy
  (cocaine) or unrelated to health (opera)

22
Rational Addiction

• Lifetime utility discounted at rate r
• ? ßt-1 U (Ct , Ct-1 , Yt , et)
• where ß 1 /(1r)
• Assume quadratic utility function, solving first
  order conditions for C implies first-difference
  equation or structural demand function
• Ct ? Ct-1 ß ? Ct1 ?1 Pt ?2 et ?3 et1

23
Rational Addiction, cont.

• Model can explain many features of addictive
  consumption, and generates new predictions
• Dynamics involve two unstable steady states with
  low and high consumption
• Life cycle shock can move consumer from low to
  high state
• Cold turkey quitting can be optimal
• Model extended to incorporate learning about
  propensity to become addicted

24
Empirical Estimation of Rational Addiction Model

• Structural demand eq. include Ct-1 , Ct1
• Ct ? Ct-1 ß ? Ct1 ?1 Pt ?2 et ?3 et1
• Estimate by two-stage least squares, use Pt-1 ,
  Pt1 as IVs for Ct-1 , Ct1
• Reduced-form demand equation
• Ct a0 a1 Pt a2 Pt-1 a3 Pt1 .. .

25
Empirical Content of Rational Addiction Model

• Ct ? Ct-1 ß ? Ct1 ?1 Pt ?2 et ?3 et1
• Prediction that ß ? gt 0 provides test of rational
  addiction versus myopic addiction
• Evidence for rational addiction to cigarettes,
  alcohol, cocaine, and coffee
• But same approach also suggests rational
  addiction to milk, eggs, and oranges, so approach
  (as implemented) is problematic
• Estimated coefficients on first two terms yields
  estimate of discount rate r
• Often, implied discount rates are implausible

26
Empirical Content of Rational Addiction Model,
cont.

• Long-run price elast. gt short-run price elast.
• Temporary and permanent price hikes have
  different impacts
• War on drugs ? temporary price hikes
• Anticipated and unanticipated price hikes have
  different impacts
• Announced tax hikes ? anticipated
• Compared to non-addicts, rational addicts more
  price-responsive to anticipated, permanent price
  hikes
• ? Price of current consumption and ? price of
  its complement (future consumption)

27
Addiction with Quasi-hyperbolic Discounting

• Time-inconsistent taste for immediate
  gratification
• Lifetime utility at t ut d ? ßt-1 ut
• The parameter d lt 1 represents the taste for
  immediate gratification at any given moment,
  the person has an extra bias for the present over
  the future.
• Inconsistency Marginal rate of substitution
  between periods t1 and t2
• From perspective of time t, only reflects ß
• From perspective of time t1, will reflect dß

28
Evidence for quasi-hyperbolic discounting

• Behavioral economics research
• Psychological experiments
• In experiment given choices between a delayed
  reward of 1,000 and immediate rewards ranging
  from 1 to 1000 undergraduate subjects revealed
• Discount rate 60 per year for year 1
• Discount rate 16 per year for years 2 - 5

29
Empirical Content of Addiction Model with
quasi-hyperbolic discounting

• Some of the same testable predictions as rational
  addiction
• We find that this model also generates the
  prediction that future prices matter for today's
  consumption indeed, they matter in ways that are
  sufficiently similar to the Becker-Murphy model
  that we are unable to empirically distinguish the
  two with our data.

30
Empirical Content of Addiction Model with
quasi-hyperbolic discounting, cont.

• Avoids problems of implausible implied discount
  rates
• Predicts consumer demand for self-control or
  commitment devices
• Empirical evidence that former smokers support
  smoking bans, possibly as a commitment device

31
Cue-Triggered Addiction

• Individual operates in two modes
• Cold mode properly functioning decision
  processes lead to selection of most preferred
  alternatives
• Hot mode dysfunctional, decisions and
  preferences may diverge results in use of
  substance
• Probability of entering hot mode depends on
  history of use, choice of lifestyle, random
  events
• Addicts know they make bad decisions while in hot
  mode, and choose lifestyles accordingly

32
3. Theory Welfare Economics

• Welfare economic analysis of substance use/
  policies differs depending which model of
  consumer behavior is right
• Rational addiction
• Quasi-hyperbolic discounting
• Cue-triggered addiction
• This is troubling, because empirical implications
  are often similar
• Applied welfare economics

33
Welfare Economics Rational Addiction

• Rational addicts maximize lifetime utility?
• Revealed preference for substance use
• Limiting substance use reduces their welfare
• Necessary condition before policy to reduce
  substance use will increase social welfare one
  or more neoclassical market failures
• Example Taxing cigarettes justified based on
  externalities from secondhand smoke and costs
  borne by third party payors

34
Evidence on Market Failures Externalities

• Controversial hard to
• Empirically estimate causal treatment effects of
  substance use on outcomes
• Conceptually distinguish internal and external
  costs
• the risks to passengers in impaired (drunk or
  drugged) drivers cars ?
• the costs addicts impose on their families?

35
Evidence on Market Failures Consumer Information

• Controversial hard to quantify extent of lack
  of information
• Example health risks of smoking
• Over 90 of US population aware of major health
  risks due to smoking
• At least some (maybe majority) over-estimate
  risks of smoking for the average smoker
• However, on average, smokers report that their
  personal risks are lower than the average smoker,
  and only slightly above the risks of non-smokers

36
Welfare Economics Quasi-hyperbolic discounting

• Quasi-hyperbolic discounting ? time-inconsistent
  choices ? internalities
• while the rational addiction model implies that
  the optimal tax on addictive bads should depend
  only on the externalities that their use imposes
  on society, the time-inconsistent alternative
  suggests a much higher tax that depends also on
  the internalities that users impose on
  themselves. At standard values of a life, these
  internalities are on the order of 30 per pack of
  cigarettes, which is 100 times the size of the
  estimated externalities from smoking.

37
Welfare Economics Cue-Triggered Addiction.

• Limitations of informational policies Even a
  highly knowledgeable addict may make uninformed
  choice in hot mode
• Advantage of criminalization difficult to
  obtain substance on short notice, while in hot
  state
• But in hot state, addicts may engage in costly
  dangerous search

38
Welfare Economics Cue-Triggered Addiction, cont.

• Taxes in excess of externalities likely to
  distort cold state choices without reducing
  problematic hot state usage
• Optimal rate of taxation with cue-triggered
  addiction may be lower than implied by
  externalities
• Cue-based policies
• Ban cigarette ads to help smoking cessation
• Provide counter-cues

39
(No Transcript)
40
Applied Welfare Economics

• Methods used in the economic evaluation of
  substance use/abuse policies
• Cost of illness/ burden of disease
• Cost-effectiveness analysis
• Cost-utility analysis
• Cost-benefit analysis

41
Cost of illness/ burden of disease

• Cost of illness direct medical costs indirect
  costs of foregone earnings
• Burden of disease DALYs lost
• These are not economic evaluations because they
  do not involve a comparative analysis of
  alternative courses of action in terms of both
  their costs and consequences.
• Estimates of total costs or burden only relevant
  for counter-factual world without the substance
• Assumes substance use disease that does not
  provide utility

42
Cost-Effectiveness Analysis of Interventions

• Incremental costs of intervention C1 C0
• Incremental effectiveness E1 E0
• Natural units addict treated, successful
  smoking cessation
• Utility-based measure such as QALY requires
  additional steps such as an epidemiologic model
  to predict the number of QALYs gained per
  successful smoking cessation
• Assumes substance use disease
• Reasonable perspective for substance abuse
  treatment
• How can the CEA capture societal gains unrelated
  to health (e.g. reductions in addict-related
  crime)?

43
Cost-Benefit Analysis

• Neoclassical market failures ? policies to reduce
  substance use might yield social benefits gt
  social costs
• Estimate willingness to pay to avoid social costs
  of substance use, compare to consumers surplus
  from substance use
• Example Optimal tax on alcohol trades off gains
  from reducing socially costly heavy drinking
  drunk driving versus losses imposed on moderate
  drinkers

44
4. Empirical Studies of Consumer Demand for
Substances

• Data Issues
• Identification
• Beyond Price Effects
• Econometrics (brief intro.)

45
Types of Data

• Aggregate sales data national time series
• Aggregate sales data pooled time series of
  cross-sections
• US states
• EU countries?
• Micro data single cross-sectional survey
• Micro data repeated cross-sections
• Micro data longitudinal panel survey data
• Claims data

46
Measuring Substance Use

• Sales data apparent consumption
• Stockpiling, cross-border sales
• Unavailable or unreliable data on sales of
  illegal substances
• Self-reported substance use
• Quantity x frequency, measure of heavy use,
  binges
• Self-reported alcohol, tobacco consumption dont
  add up to sales
• Validation with objective measures
• Item non-response

47
(No Transcript)
48
Measuring Substance Abuse

• Psychiatric criteria for abuse/ dependence
• Self-reported for DSM IV criteria for alcohol
  abuse/ dependence
• Tobacco addiction measures
• Claims data ? substance abuse treatment
• Untreated ? no problem
• Psychiatric measure might not measure what
  economists want to measure

49
Data Issues the Sample

• Household surveys miss
• institutionalized population (prisoners)
• homeless populations
• School-based surveys also miss
• Non-attendees
• Dropouts
• Refusals at the school-level
• As many as 50 of schools chosen to participate
  in the Monitoring the Future surveys refuse

50
Empirical Demand Model

• C a0 a1 P a2 B a3 I a4 X e
• Suppress sub-scripts depend on data (time
  series, cross-sectional, panel)
• Estimate
• a1 gives impact of Price
• a2 gives impact of Ban (or other policy)

51
Identification Validity

• First question is about research design What is
  the natural, quasi-experiment in the data that
  provides identifying variation in P and B?
• This is not just a statistical question just as
  relevant if observe population instead of sample
• This is not just an econometric problem when
  using instrumental variables (IVs)

52
Invalid Sources of Identifying Variation

• Across political jurisdictions, taxes policies
  reflect public sentiment
• Example Cross-state differences in cigarette
  taxes correlated with tobacco production,
  sentiment
• Bias estimates away from zero
• Policies enacted in reaction to problem
• Example Increase drinking age in response to
  high rate of traffic deaths
• Bias estimates towards or past zero
• Multiple policies enacted simultaneously
• Example interdiction of illegal drugs (drives
  up P) and prevention campaign (shifts demand)
• Hard to dis-entangle

53
Identification Strength

• Statistical issue is there adequate valid
  variation to yield sufficiently precise
  estimates?
• Not different from zero not necessarily an
  interesting result
• Can you rule out that the policy has a large
  effect?

54
(No Transcript)
55
Dilemma Valid or Strong?

• Common example use state fixed effects to
  control for sentiment, year fixed effects to
  control for general trends
• Identification within-state variation, off
  national trend
• Sometimes, even add state-specific time trends
• Policy variable, state year fixed effects
  highly multicollinear

56
Dilemma, cont.

• False solution drop the controls (Occams
  hatchet)
• Unfortunately, the inclusion of state-fixed
  effects in conjunction with the time-fixed
  effects eliminated virtually all the independent
  variation in cigarettes prices .which implies
  that there is not enough variation in cigarette
  prices within states over time to justify the
  inclusion of state-fixed effects in the model.
• Emphasis added this is where the author went
  wrong and decided to use the hatchet.

57
Dilemma, cont.

• tossing out collinear variates is quite
  generally neither a good nor a recommended
  solution to the collinearity problem. If
  collinearity can be shown to be rendering some a
  priori important effect insignificant, the
  appropriate conclusion is that the data lack the
  information needed to accomplish the statistical
  task at hand with precision.one must either get
  better data or introduce appropriate prior
  information.

58
Identification Freaks versus Geeks

• Many economists concluded that the path to
  knowledge lay in solid answers to modest
  questions. Henceforth, the emphasis would be on
  clean identification, on sorting out what
  caused what.
• Has clean identification become a fetish?
• The literature on treatment effects has given
  rise to a new language of economic policy
  analysis where the link to economic theory is
  often obscure and the economic policy questions
  being addressed are not always clearly stated.

59
Identification of Price Effects

• Consensus estimates based on what is arguably
  flawed identification strategy
• Same biased results from repeating the wrong
  experiment
• More compelling identification strategies
• World Trade Organization agreement caused big
  drop in Swiss liquor prices
• Fundamental market reforms created variation in
  Chinese Russian cigarette prices
• Cigarette price hike due to 1998 Master
  Settlement Agreement

60
Identification of the Impact of Bans, Other
Policies

• Bans (partial or complete), regulations, other
  policies ? full price of substances ?
• Time inconvenience costs
• Outlet density
• Restrictions on day, time, place of sales
• Illegality costs (probability of punishment) x
  (penalty if arrested, convicted)
• Again, hard to find good quasi-experiments that
  provide useful identifying variation

61
Beyond Prices Identifying the Impact of Peers
Parents

• Psycho-social research suggests peers parents
  have major influence on young peoples substance
  use
• Economic theory?
• Utility from peer acceptance, substance use a
  possible input into production of peer acceptance
• Household bargaining models ? parents influence

62
Identification of Peer Effects

• Reflection problem peer influences go both
  ways, so individuals behavior and those of his
  or her peers are simultaneously determined
• Peer groups endogenously chosen based on shared
  preferences over risk, time, social deviancy
• Parents choices determine set of possible peers
• Peers may share unobserved common environmental
  factors

63
Experimental Variation in Peers

• Random assignment of college roommates
• Random assignment of classmates
• Natural experiments
• When substance-using friend moves away
• Presence of older peers depends on school
  structure

64
Econometric Methods

• Could teach an econometrics course just using
  substance use examples
• Aggregate ? micro data
• Micro data allows research to address interesting
  new questions
• Micro data requires use of econometric methods
  for limited qualitative dependent variables

65
Decomposing the Price Elasticity(New questions)

• Distinguish impact of price on
• Participation (prevalence) any use
• Demand, conditional on use
• Price elasticity ? ?C/?P P/C
• C PrC gt0 C Cgt0
• ? ?P ?I

66
Elasticities, cont.

• ?P is a population concept 1 increase in price
  causes prevalence to fall by ?P
• To calculate ?P, replace Pr Cgt0 with fraction
  of population with C gt 0
• Comparing across population groups with much
  difference prevalence requires caution
• Suppose marginal effect is the same in two
  groups 1 increase in P causes prevalence to
  drop by ½ percentage point
• ?P is larger in group with lower prevalence

67
Corner Solutions

• 30 40 of population are non-drinkers
• 70 80 of population are non-smokers
• gt 95 are non-users of various other substances
• Also, heavy users account for large share of
  total consumption of alcohol, illegal drugs (less
  true for cigarettes)

68
Two part model

• The two part model is probably the standard
  approach for studies that use micro data on
  substance use
• First part Pr C gt 0
• Probit, logit, linear probability model
• Second part C C gt 0
• OLS, maybe after logging
• This may not be enough Mannings lectures will
  cover relevant methods

69
Dynamics (rational) addiction

• Standard approach only includes Pt
• Pt proxies for history of past prices
• Including Pt-1, Pt-2 multi-collinear with Pt
• Structural rational addiction model
• The functional form and exogeneity assumptions
  invoked in this literature the structural
  approach are sometimes controversial and the
  sources of identification of parameters of these
  models are often not clearly articulated.

70
Participation, Initiation Cessation

• Decompose participation elasticity Users
  consist of starters and non-quitters
• Pr Ct gt 0 Pr Ct gt 0 Ct-1 0 PrCt-1
  0
• Pr Ct gt 0 Ct-1 gt 0 PrCt-1 gt 0
• ?P wI ?I wQ ?Q
• Stock ? Flows
• Weights reflect relative magnitudes of flows of
  starters and quitters

71
5. Empirical studies of the consequences of
substance use

• Theoretical Framework Substances have negative
  marginal product in some household production
  function
• Health f (substance use,)
• Human capital f (substance use, )
• Estimate structural production function, or more
  usually try to estimate the causal treatment
  effect of substance use

72
Theoretical framework, cont.

• Other behavioral functions conditional on
  substance use
• Labor supply f (substance use,)
• Criminal activity f (substance use, )
• Earnings f (substance use, )
• Sexual behavior f (substance use, )
• Observe equilibrium relationships
• Equilibrium of workers/ firms
• Equilibrium with potential sexual partners

73
Example of Empirical ChallengesDrug Use and
Crime

• Drug use causes crime
• Users commit crimes to finance addiction
• Intoxication makes crime more likely
• Involvement in illegal markets leads to other
  crime
• Crime causes drug use
• Initiation of drug use is through illegal
  markets most user-criminals were criminals
  before they were users
• Both drug use and crime caused by other factors
• Economic marginality, social isolation, family
  pathology, mental illness,

74
Empirical Approaches

• Reduced-form approach
• Outcome g (beer tax, )
• Focuses on tax or some other directly
  policy-manipulable variable
• Black box or chain of causality
• Tax ? Price ? Substance use ? Outcome
• From reduced-form estimates, can infer substance
  use has causal impact on outcome
• Check plausibility of implied structural
  parameters

75
Instrumental Variables Approach

• Need for valid, strong instrumental variables
  that predict substance use, can be excluded as
  direct determinant of outcome
• Prices, demand-side policies attractive from
  theoretical perspective
• But if invalid for demand models, probably
  invalid as IVs
• Weak IVs Even if prices are valid IVs, may not
  explain much variation in use

76
Weak IVs

• If excluded IVs do not explain much variation in
  endogenous regressor, IV estimates can have
  atrocious finite sample properties
• Even very small correlation between IVs and error
  term in equation of interest can cause large bias
• Cure can be worse than the disease (bias in IV
  estimate gt bias in OLS estimate)

77
Anti-Test of Identification

• Several studies estimates ? illegal drug use
  causes risky sexual behavior
• Similar identification strategy ? cigarette
  smoking causes risky sex
• Implausible result calls into question
  identification strategy
• Further tests suggest weak IVs, possibly invalid
  exclusion restrictions

78
Anti-Test of Identification, cont.

• Zero Tolerance laws make it illegal for underage
  drivers to have BAC gt 0
• Reduced-form results ? ZT laws ? gonorrhea rates
  among 15-19 year old white males
• No relationship between ZT laws and gonorrhea
  rates among slightly older males ? supports
  identification

79
Average Treatment Effects

• Heterogeneous treatment effects
• Hi di B a1 X1
• Focus on some average of di
• Can not estimate the effect of B ? every
  individual i experiences a different effect
• What does IV approach yield?
• IV estimator does not converge to mean di
• IV estimator converges to a weighted average of
  di , where the weights are larger for respondents
  who change their behavior most in response to the
  instruments

80
Average Treatment Effects, cont.

• Suppose estimate
• Wage d (Drug use)
• Use drug price as IV for drug use
• IV estimate of d is a weighted average of the
  effect of drug use, where weights depend on the
  elasticity of drug demand with respect to price
• Users with perfectly inelastic demand given 0
  weight
• Is the average treatment effect estimated by IV
  interesting and relevant?
• Be careful how interpret

81
6. Supply side of legal substances

• Legal substances
• Industrial organization of markets for alcoholic
  beverages, cigarettes
• Public finance studies of taxation
• Gains from trade between health economics other
  fields of economics

82
Tax Pass-Thru

• To what extent are taxes on alcohol cigarettes
  passed through to consumer prices?
• Perfect competition, constant marginal costs?
  passed through 11
• Monopoly, linear demand ? monopolist passes
  through exactly ½ the tax
• Oligopoly/ game theory ? wide range of
  possibilities (little guidance from theory)

83
Tax Pass Thru, cont.

• Cigarette excise taxes passed thru at rate gt 1
  (maybe 1.2)
• Preliminary evidence pass thru rate gt 1 for
  pack-buyers, lt 1 for carton-buyers
• Alcohol excise taxes also passed thru at rate gt 1
• Evidence from Alaska tax pass-thru rate of at
  least 2, rate tends to be higher (3 to 4) for
  on-premise wine, spirits

84
Advertising of legal substances two views

• Advertising informs consumers, lowers search
  costs, and makes markets more competitive
• This implies alcohol cigarette ads hurt public
  health
• Advertising is wasteful, creates meaningless
  product differentiation and barriers, and leads
  to market power
• This implies cigarette alcohol ads improves
  public health
• we are not interested in trying to persuade any
  nonsmokers to begin smoking or in persuading any
  smokers to not quit. Marketing Philosophy, R J
  Reynolds Tobacco Company, 2006.

85
1989
86
(No Transcript)
87
Profit-Maximizing Choice of Advertising Messages
(M)

• Profits PC S1 (M, M, PC) S2 (M, M, PC)
  S3 (M, M, PC) - C( ) - M PM
• (PC - MC) ?S1 /?M ?S2 / ?M ? S3 / ? M
• (PC - MC) (dM/dM) ?S1/?M ?S2/? M
  ?S3/?M
• PM

88
Questions about Advertising

• Does it expand size of market ?
• ? initiation ?
• ?cessation ?
• Or is it just about brand-switching?
• Conjectural variations what game(s) are tobacco
  companies playing?
• Welfare economics

89
7. Illegal substance markets

• Impact of illegal status on price
• Theoretical models of economics of crime ?
  certainty severity of punishment
• Empirical study (very difficult!) suggests
  illegal price of cocaine is 2 to 4 times higher
  than price that would prevail if legal
• Study of gangs suggests that most members dont
  earn much
• Search costs ? higher variance in prices, quality
  problems

90
Economies of Scale in Illegal Substance Markets

• Fixed costs of illegal distribution on supply
  side
• Larger birth cohorts ? thicker market for illegal
  drugs ? lower sales arrest risk informational
  economies ? lower prices
• Empirical evidence that these efficiency gains
  more important than explanations based on general
  strain on police resources, attitude transfers

91
8. Future Directions Implications for Obesity
Research

• (Many possible directions for future research
  obesity seems hot.)
• Obesity substance use seem to have some similar
  causes consequences
• Theory
• Empirical Studies

92
Obesity Research Theory

• Weight increases if
• Calories In gt Calories Out
• Do consumer decisions about food exercise share
  similarities with consumer demand for substances?
• Addictive stocks
• Rationality? Hyperbolic discounting? Cues?

93
Welfare Economics of Obesity

• Neoclassical market failures
• Internalities
• Are fat taxes welfare enhancing?
• Closest analogy from economics of substance use
  may be alcohol
• Many moderate eaters
• Some abusive eaters

94
Empirical studies of consumer demand for obesity

• Try to explain trend of rapidly increasing
  obesity (epidemic)
• Labor market trends ? higher opportunity cost of
  time ? shift towards calorically dense fast
  foods
• Technological change and the relative prices of
  different foods

95
Empirical studies of the consequences of obesity

• Labor market impacts
• Returns to beauty
• Use parental weight, genetic markers as IVs
• Impact on human capital acquisition