1. No category

An Ex Ante Evaluation of the Effects of Banning
Advertising on Junk Food Markets
Pierre Dubois, Rachel Griffith and Martin O’Connell
Toulouse School of Economics
University of Manchester and IFS
UCL and IFS
May 2015
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
May 2015
1 / 60
Introduction
Motivation
Growing obesity and diet-related disease across developed world
Many policies proposed/introduced
Education and information campaigns
Fiscal measures
Regulations
One discussed policy consists in banning junk food advertising
UK currently bans advertising of foods high in fat, salt or sugar
during children’s programs (partial ban)
Need ex ante evaluation
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
May 2015
2 / 60
Introduction
Motivation
There are calls for restriction of advertising in junk food markets such
as a complete ban
But ex ante we don’t know what will be the impact on markets which
depends on
How the demand shape changes with advertising
Whether advertising is expansionary or pure business stealing across
brands
Strategic response of firms: price equilibrium
Need counterfactual evaluation of supply and demand
As well as questioning on evaluating welfare effects
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
May 2015
3 / 60
Introduction
Contribution
Develop model of consumer demand and oligopoly supply with
multi-product firms competing in price and advertising
Allow advertising to impact demand in a flexible way
Allow past advertising to impact current demand meaning firms play
a dynamic game
Estimate the model on the typical junk food market in the UK
(potato chips)
Simulate the impact of advertising ban on equilibrium outcomes
(prices, expenditures, quantities, nutrition)
Consider welfare evaluation depending on whether advertising distorts
consumer’s choices
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
May 2015
4 / 60
Consumers Demand
Outline of talk
1
2
3
4
5
6
Consumers Demand
Firms Behavior
Data
Estimates
Counterfactual Equilibrium
Welfare Effects
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
May 2015
5 / 60
Consumers Demand
Advertising in consumer demand model
Consumer demand model such that:
Allow cooperative or rival effects of advertising such that increase in
advertising of one brand may:
Increase demand for another brand (cooperative)
Decrease demand for another brand (predatory)
Lead to expansion or contraction of market
Allow dynamic effects of advertising on demand:
Advertising state vector for brand b: abt can depend current and past
advertising expenditures
Denote
abt = f (ebt , ebt −1 , ebt −2 , ..., eb0 )
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
May 2015
6 / 60
Consumers Demand
Consumer discrete choice model
Random utility for consumer i, brand b, pack size s, time (market) t
v¯ibst = αi (abt , pbst ) + ψi (abt , xb ) + γbi (at ) + ηi (zbs , ξ b ) + eibst
where:
pbst is price
xb is nutrient score
abt is advertising states for brand b; at = (a1t , ..., aBt )
zbs are functions of pack size
ξ b is an unobserved brand characteristic
eibst individual deviation that may contain some product speficic time
varying unobservables
With outside good : v¯i00t = ζ d0t + ei00t
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
May 2015
7 / 60
Consumers Demand
Consumer discrete choice model
A flexible specification:
αi (abt , pbst ) = (α0i + α1i abt ) pbst
ψi (abt , xb ) = (ψ0i + ψ1i abt ) xb
γbi (at ) =λi abt + ρi ∑l 6=b alt
2
ηi (zbs , ξ b ) =η1i zbs + η2i zbs
+ ηi ξ b
where πiu = (α0i , λi , ρi , ηi ) such that πiu = π0u + π1u di + υi di with
υi ∼ N (0, Σπ ) and πio = (α1i , ψ1i , η1i , η2i ) with πio = π0o + π1o di
Coefficients differ by demographics (di ) and purchase occasion
All advertising coefficients allow potential shift with consumer’s
observed and unobserved exposure to advertising. Exposure measure
is based on how much TV the consumer reports watching.
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
May 2015
8 / 60
Consumers Demand
Consumer discrete choice model
From choice set Ωκ consumer chooses (b, s ) if:
v¯ibst ≥ v¯ib0 s 0 t for all (b 0 , s 0 ) ∈ Ωκ
Probability of purchasing (b, s ) is sibs (pt , at , ζ t ) equal to:
exp(ζ d0t ) +
exp[αi (abt , pbst ) + ψi (abt , xb ) + γbi (at ) + ηi (zbs , ξ b )]
exp [αi (ab0 t , pb 0 s 0 t ) + ψi (ab0 t , xb 0 ) + γbi (at ) + ηi (zb0 s0 , ξ b 0 )]
∑
(b 0 ,s 0 )∈Ωκ
Then aggregate demand is:
sbs (pt , at , ζ t ) =
Dubois, Griffith, O’Connell (TSE)
Z
sibs (pt , at )dF (υi , di )
Banning Junk Food Advertising
May 2015
9 / 60
Consumers Demand
Impact of advertising on demand is flexible
Brand advertising can be (even at individual level):
predatory with respect to some products and cooperative with respect
to others
market expanding or contracting
∂sibst
=sibst
∂abt
∂sibst
=sibst
∂ab 0 t
λ˜ ibst − ρi (1 − si00t ) −
∑
!
(λ˜ ibs 0 t − ρi )sibs 0 t
s 0 ∈ Kb
ρi si00t −
∂si00t
= − si00t
∂ab 0 t
∑
!
(λ˜ ib 0 s 0 t − ρi )sib 0 s 0 t
s 0 ∈ Kb 0
ρi (1 − si00t ) +
∑
!
(λ˜ ib 0 s 0 t − ρi )sib 0 s 0 t
s 0 ∈Kb 0
where
λ˜ ibst = λi + α1i psbt + ψ1i xb
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
May 2015
10 / 60
Consumers Demand
Willingness to pay for reduction in nutrient score
Advertsising affects willingness to pay:
∂v¯ibst /∂xb
∂v¯ibst /∂pbst
ψ0i + ψ1i abt
=
α0i + α1i abt
WTPibt =
Increases or decreases with abt depending on the sign of
ψ1i α0i − ψ0i α1i
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
May 2015
11 / 60
Firms Behavior
Outline of talk
1
2
3
4
5
6
Consumers Demand
Firms Behavior
Data
Estimates
Counterfactual Equilibrium
Welfare Effects
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
May 2015
12 / 60
Firms Behavior
Supply overview
Multi-product firms compete by setting simultaneously two strategic
instruments to maximize profits
prices and advertising expenditures
Firms’ problem is dynamic because
advertising today affects future demand and hence profits
Estimation is innocuous to product entry and exit firm optimization
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
May 2015
13 / 60
Firms Behavior
Profit
Multi-product firm j chooses (pbst , ebt ) to maximize intertemporal
profit:


∞
∑ βt  ∑
t =0
(pbst − cbst ) sbs (pt , at , ζ t ) Mt −
(b,s )∈Njbs
∑
ebt 
b ∈Njb
where
abt = f (ebt , ebt −1 , ebt −2 , ..., eb0 )
Njbs : set of products owned by firm j
Njb : set of brands owned by firm j
cbst : constant marginal cost
Mt : size of the potential market
ebt : advertising expenditure
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
May 2015
14 / 60
Firms Behavior
Structure of game
Timing of the game:
at beginning of each period firms observe state variables (at −1 , θt )
play their strategy σj mapping states to decisions (pbst , ebt )((b,s )∈N bs )
j
individual demand shocks are realized and firms get current profit
We assume:
cost and market size follow independent Markov processes such that
Et [cbst +1 ] = cbst and Et [Mt +1 ] = Mt
firms play strategies which depend only on payoff relevant state
variables, so σj (at −1 , θt ) (where θt ≡ (ct , Mt , ζ t ))
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
May 2015
15 / 60
Firms Behavior
Markov perfect equilibrium
Firm j makes an assumption on competitive strategy profile
σ−j = (σ1 , ..., σj −1 , σj +1 , ..., σJ ) and chooses its own strategy σj
Value function πj∗ (., .) from Bellman equation conditional on specific
strategy profile σ−j :
πj∗ (at −1 , θt ) =
max
∑ (pbst − cbst )sbs (pt , at , ζ t )Mt
σj =(pbst ,ebt )∈Njbs
−
∑
ebt +
(b,s )∈Njbs
βE [πj∗
(at , θt +1 )]
b ∈Njb
A Markov perfect equilibrium is a list of strategies σ∗ = (σ1∗ , ..., σJ∗ )
such that no firm has an incentive to deviate from the action
prescribed by σj∗ in the subgame that starts from the state (at −1 , θt )
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
May 2015
16 / 60
Firms Behavior
Markov perfect equilibrium
Assume existence of a subgame perfect Markov equilibrium, and
restrict attention to Markov Perfect Equilibrium in pure strategies
(Maskin and Tirole, 2001)
Ericson and Pakes (1995), Doraszelski and Satterthwaite (2003) give
general conditions for the existence of equilibria in similar games
Each solution of Bellman equation πj∗ corresponds to each MPE of
dynamic game.
Do not need to assume equilibrium is unique
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
May 2015
17 / 60
Firms Behavior
Price first order conditions
Price first order conditions depend on Markov perfect equilibrium only
through observed goods and state vector (pt , at )
sbs (pt , at , ζ t ) +
∑
0 0
(pb0 s 0 t − cb0 s 0 t )
(b ,s )∈Nj
∂sb0 s 0 (pt , at , ζ t )
=0
∂pbst
... we can identify marginal costs without solving for the value
function πj∗
Optimality conditions of entry, exit and advertising decisions not
needed for identification of costs
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
May 2015
18 / 60
Firms Behavior
Advertising Ban
Simulate Counterfactual equilibrium with ban on advertising (at = 0)
New price equilibrium will be played and satisfy the following per
period Bertrand-Nash conditions, for all (b, s )
sbs (p, 0, ζ ) +
∑
0 0
(pb0 s 0 t − cb0 s 0 t )
(b ,s )∈Nj
where
sbs (p, 0, ζ ) =
Z
∂sb 0 s 0 (p, 0, ζ )
=0
∂pbs
sibs (p, 0, ζ )dF (υi , di )
is aggregate demand for product (b, s ) when advertising is banned
Can check exit decisions for all possible assortments
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
May 2015
19 / 60
Data
Outline of talk
1
2
3
4
5
6
Consumers Demand
Firms Behavior
Data
Estimates
Counterfactual Equilibrium
Welfare Effects
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
May 2015
20 / 60
Data
Purchase data
From Kantar/TNS Worldpanel
June 2009 - October 2010
Use information on a subset of households
all groceries brought into home by 2873 households (food at home),
161,513 transactions
all snacks bought for consumption outside the home by 2306
individuals (food on the go), 99,636 transactions
Observe all barcodes bought and transaction level prices
Plus demographics and product characteristics
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
May 2015
21 / 60
Data
Food at home - 26 products in total
Brand
Pringles:
Walkers Regular:
Walkers Sensations:
Walkers Doritos:
Walkers Other:
Golden Wonder:
Size
150-300g
300g+
150-300g
300g+
150-300g
300g+
150-300g
300g+
<150g
150-300g
300g+
<150g
150-300g
300g+
Purchase Share
1.34%
5.54%
1.77%
23.98%
0.43%
1.81%
1.30%
3.29%
0.69%
3.73%
8.66%
0.10%
0.25%
1.15%
Price (£)
1.10
2.63
1.25
2.77
1.26
2.52
1.21
2.47
1.24
1.77
3.17
1.28
1.35
2.70
...
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
May 2015
22 / 60
Data
Food on the go - 11 products in total
Brand
Walkers Regular
Walkers Sensations
Walkers Doritos
Walkers Other
KP
Golden Wonder:
Other
Size
34.5g
50g
35g
50g
<30g
30g+
35g
<40g
40g+
<40g
40g+
Purchase Share
27.16%
7.19%
2.04%
4.70%
4.34%
8.94%
0.83%
3.08%
1.09%
17.57%
20.01%
Price (£)
0.45
0.63
0.61
0.54
0.45
0.61
0.57
0.39
0.73
0.48
0.59
...
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
May 2015
23 / 60
Data
Nutrient score
Interested in nutrient characteristics of the products
Use nutrient profile model developed by Rayner et al (2005), used by
FSA and Ofcom
The nutrient profiling model aggregates all nutrient characteristics
into a single score
For potato chips, relevant nutrients are the amount of energy,
saturated fat and sodium
Score is the sum of points, 1 point for each 335kJ per 100g, 1 for each
1g of saturated fat per 100g, and 1 for each 90mg of sodium per 100g
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
May 2015
24 / 60
Data
Nutrient score
Brand
Pringles
Walkers
Walkers
Walkers
Walkers
KP
GW
Asda
Tesco
Other
Nutrient score
Reg
Sens
Dor
Oth
16
10
11
12
15
18
16
15
15
12
Energy
(kj per 100g)
2160
2164
2023
2095
2020
2158
2101
2125
2145
2084
Saturated fat
(g per 100g)
6.31
2.56
2.16
2.86
2.50
5.87
4.01
4.13
4.65
3.84
Sodium
(g per 100g)
0.62
0.59
0.71
0.66
0.82
0.85
0.92
0.75
0.77
0.70
Proposal is to ban adversiting for score above 4 (fiber and protein not shown)
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
May 2015
25 / 60
Data
Advertising data (AC Nielsen digest of advertising)
Advertising expenditure by brand and month from 2001 to 2010
Includes all potato chips advertising appearing on TV, in press, on
radio, on outside posters and internet
We compute the stock of advertising goodwill according to:
at = δat −1 + et
with δ = 0.75
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
May 2015
26 / 60
Data
Advertising Expenditures
Pringles
Walkers Regular
Walkers Sensations
Walkers Doritos
Walkers Other
KP
Golden Wonder
Asda
Tesco
Other
Dubois, Griffith, O’Connell (TSE)
Monthly expenditure
Mean Min
4.50
0.00
4.97
0.00
0.54
0.00
1.75
0.00
2.89
0.00
2.09
0.00
0.08
0.00
0.01
0.00
0.08
0.00
1.58
0.00
(£100,000)
Max
10.14
18.29
1.46
8.25
8.99
8.49
0.80
0.23
0.68
5.74
Banning Junk Food Advertising
Total
(06/09-10/10)
76.54
84.47
9.12
29.67
49.07
35.60
1.34
0.23
1.44
26.83
May 2015
27 / 60
Data
0 2000000
4000000
6000000
Pringles
Walkers_Regular
Walkers_Sen
4000000
0 2000000
6000000
Advertising Stocks and Flows by Brands (2000-2012)
Walkers_Dor
Walkers_Oth
KP
0
2/2004
4/2008
6/2012 0
2/2004
4/2008
6/2012 0
2/2004
4/2008
6/2012
date
Advert. Stock (.85)
Advert. Stock (.75)
Advert. Stock (.8)
Advert. Flow
Graphs by agg_brand
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
May 2015
28 / 60
Data
Advertising Stocks and Flows
4000000
0 1000000
2000000
3000000
5000000
Pringles
0
2/2004
4/2008
6/2012
date
Advert. Stock (.85)
Advert. Stock (.75)
Advert. Flow
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
Advert. Stock (.8)
Advert. Stock (.7)
May 2015
29 / 60
Data
Advertising Stocks and Flows
0
2000000
4000000
6000000
Walkers Regular
0
2/2004
4/2008
6/2012
date
Advert. Stock (.85)
Advert. Stock (.75)
Advert. Flow
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
Advert. Stock (.8)
Advert. Stock (.7)
May 2015
30 / 60
Data
Advertising Stocks and Flows
0
1000000200000030000004000000
Walkers Sensations
0
2/2004
4/2008
6/2012
date
Advert. Stock (.85)
Advert. Stock (.75)
Advert. Flow
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
Advert. Stock (.8)
Advert. Stock (.7)
May 2015
31 / 60
Data
Advertising Stocks and Flows
2000000
5000000
0 1000000
3000000
4000000
Walkers Doritos
0
2/2004
4/2008
6/2012
date
Advert. Stock (.85)
Advert. Stock (.75)
Advert. Flow
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
Advert. Stock (.8)
Advert. Stock (.7)
May 2015
32 / 60
Data
01000000
2000000
3000000
4000000
Pringles
Walkers_Regular
Walkers_Sen
3000000
01000000
2000000
4000000
Advertising Stocks and Flows by Brands (2009-2010)
Walkers_Dor
Walkers_Oth
KP
2/20097/2009
12/2009
5/2010
10/2010 2/20097/2009
12/2009
5/2010
10/2010 2/20097/2009
12/2009
5/2010
10/2010
date
Advert. Stock (.85)
Advert. Stock (.75)
Advert. Flow
Advert. Stock (.8)
Advert. Stock (.7)
Graphs by agg_brand
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
May 2015
33 / 60
Data
0
0
.5
1
1.5
2
2.5
3
3.5
4
500000 1000000150000020000002500000
Advertising Stocks, Flows and Prices
7/2009
12/2009
5/2010
10/2010
date
Advert. Stock (.75)
Price (Pringles:150-300g)
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
Advert. Flow
Price (Pringles:300g+)
May 2015
34 / 60
Data
0
0
.5
1
1.5
1000000
2
2.5
3
2000000
3.5
4
3000000
Advertising Stocks, Flows and Prices
7/2009
12/2009
5/2010
10/2010
date
Advert. Stock (.75)
Price (Walkers Regular:150-300g)
Dubois, Griffith, O’Connell (TSE)
Advert. Flow
Price (Walkers Regular:300g+)
Banning Junk Food Advertising
May 2015
35 / 60
Data
0
0
.5
1
200000
1.5
2
400000
2.5
3
600000
3.5
4
800000
Advertising Stocks, Flows and Prices
7/2009
12/2009
5/2010
10/2010
date
Advert. Stock (.75)
Price (Walkers Sens.:150-300g)
Dubois, Griffith, O’Connell (TSE)
Advert. Flow
Price (Walkers Sens.:300g+)
Banning Junk Food Advertising
May 2015
36 / 60
Data
0
0
.5
1
1.5
500000
2
2.5
3
1000000
3.5
4
1500000
Advertising Stocks, Flows and Prices
7/2009
12/2009
5/2010
10/2010
date
Advert. Stock (.75)
Price (Walkers Doritos:150-300g)
Dubois, Griffith, O’Connell (TSE)
Advert. Flow
Price (Walkers Doritos:300g+)
Banning Junk Food Advertising
May 2015
37 / 60
Data
Advertising and Price Strategies
Unconditionally, prices are uncorrelated with advertising stocks and
advertising flows
Conditional on product fixed effects:
Prices are uncorrelated with advertising stocks and advertising flows
Prices have brand specific within product correlations with (lagged)
advertising stocks (positive for Pringles, Golden Wonders, negative for
Walkers Sensations and not significant for others)
Evidence consistent with multiple equilibria, pulsing strategies
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
May 2015
38 / 60
Data
Consumers Descriptive Statistics
Demographic group
Composition
HH no children
skill level
high
low
Pensioners
HH children
high
low
Number of purchase occasions
food at home food on-the-go
income
high
medium
low
medium-high
low
high
medium
low
medium-high
low
Child purchase
22721
13178
13341
10187
16147
14384
20426
14292
7091
15349
14397
14371
8376
8219
6667
8559
6016
12786
8502
4494
9549
8932
3165
All parameters are allowed to vary across cells
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
May 2015
39 / 60
Estimates
Outline of talk
1
2
3
4
5
6
Consumers Demand
Firms Behavior
Data
Estimates
Counterfactual Equilibrium
Welfare Effects
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
May 2015
40 / 60
Estimates
Identification: price and advertising variations
Use individual transaction level data
Control for time and brand effects
Price variation
National pricing in the UK (no variation according to regional
unobserved heterogeneity)
Time series variation of prices within product due to promotions
Advertising variation
Exposure, demographics and random coefficients on advertising allow
heterogenous effects across consumers
Use control function approach (Blundell and Powell, 2004 and Petrin
and Train, 2010) because brand advertising flows may be correlated
with unobserved individual demand shocks. Use advertising
expenditure on ready-meals interacted with brand fixed effects are used
as instruments
Many robustness checks
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
May 2015
41 / 60
Estimates
Advertising effects on brand demand
% change in demand if advertising expenditure set to zero (ceteris paribus)
Walkers Regular
Pringles
KP
0.497
0.450
0.209
-2.77
1.39
0.63
[-4.30, -1.44]
[1.06, 1.72]
[0.50, 0.76]
Pringles
3.43
-19.53
0.25
[2.78, 4.10]
[-21.54, -17.97]
[0.11, 0.39]
KP
-0.35
0.03
-2.63
[-0.81, 0.11]
[-0.35, 0.39]
[-3.36, -1.99]
...
...
...
...
-1.15
-1.10
-0.42
[-1.46, -0.85]
[-1.41, -0.79]
[-0.53, -0.31]
Numbers are means across markets (i.e. months).
Adv exp (£m)
Walkers Regular
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
May 2015
42 / 60
Estimates
Effect of advertising on own price elasticities
Walkers Regular
Obs.
Zero
advert.
advert.
exp.
exp.
Pringles
Obs.
Zero
advert.
advert.
exp.
exp.
<150g
150-300g
-1.49
[-1.57, -1.44]
300g+
-2.20
[-2.32, -2.10]
Dubois, Griffith, O’Connell (TSE)
-1.62
[-1.69, -1.57]
-2.54
[-2.67, -2.44]
-1.40
[-1.46, -1.35]
-2.37
[-2.51, -2.26]
Banning Junk Food Advertising
-1.53
[-1.60, -1.49]
-2.74
[-2.88, -2.64]
KP
Obs.
advert.
exp.
-1.33
Zero
advert.
exp.
-1.37
[-1.38, -1.29]
[-1.42, -1.32]
-1.68
[-1.75, -1.63]
-2.77
[-2.89, -2.67]
May 2015
-1.74
[-1.80, -1.68]
-2.88
[-3.01, -2.79]
43 / 60
Estimates
Willingness to pay for one point reduction in nutrient score
Advertising:
Food at home
Willingness to pay in pence
None
Medium
High
5.3
[4.7, 5.8]
2.5
[2.3, 2.8]
3.5
[3.0, 3.9]
1.7
[1.5, 1.9]
0.6
[-0.4, 1.6]
0.3
[-0.2, 0.8]
0.9
[0.7, 1.1]
0.0
[-0.2, 0.1]
-0.8
[-0.9, -0.5]
1.7
-0.1
[1.3, 2.1] [-0.4, 0.3]
Numbers are median WTP in pence.
-1.5
[-1.8, -1.0]
% of mean price
Food on-the-go
Willingness to pay in pence
% of mean price
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
May 2015
44 / 60
Counterfactual Equilibrium
Outline of talk
1
2
3
4
5
6
Consumers Demand
Firms Behavior
Data
Estimates
Counterfactual Equilibrium
Welfare Effects
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
May 2015
45 / 60
Counterfactual Equilibrium
Counterfactual
Estimate marginal costs using supply model
Simulate counterfactuals
no pricing response
with pricing response
check deviations of Nash equilibrim in product exits
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
May 2015
46 / 60
Counterfactual Equilibrium
Advertising ban: pricing response
Banning advertising leads to toughening price competition
The average price in the market falls by 9%
Pricing response differs across firms and over products
The big advertisers (e.g. Walkers and Pringles) lower prices
For instance, Walkers reduces price of its most popular brand by the
most, 34p (or 28%) reduction for the 150-300g pack, and 56p (or
20%) for the 300g+ pack
Besides advertising ban, no products exit the market (keeping all
products is a Nash equilibrium)
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
May 2015
47 / 60
Counterfactual Equilibrium
Advertising ban
Pre ban
Expenditure (£m)
220.86
[217.17, 222.99]
% change
Quantity (mKg)
Mean pack size
condi. on purchase
% change
Dubois, Griffith, O’Connell (TSE)
[195.79, 216.42]
[201.23, 220.38]
-6.35
-3.92
[-10.88, -1.82]
[-8.13, 0.41]
32.14
29.18
34.95
[31.62, 32.47]
[27.62, 30.65]
[33.18, 36.28]
% change
Probability of selecting
potato chips
% change
Post ban
No firm response With firm response
206.82
212.19
-9.21
8.73
[-13.60, -4.29]
[3.39, 13.53]
0.37
0.37
0.38
[0.36, 0.37]
[0.35, 0.39]
[0.35, 0.39]
0.22
1.79
[-5.33, 6.16]
[-4.02, 7.11]
0.17
0.15
0.18
[0.17, 0.17]
[0.15, 0.16]
[0.17, 0.19]
-9.43
6.63
[-13.18, -5.65]
[2.23, 10.83]
Banning Junk Food Advertising
May 2015
48 / 60
Counterfactual Equilibrium
Advertising ban
Saturates (1000 kg)
Observed
equilibrium
1264.91
[1244.03, 1279.87]
% change
Salt (1000 kg)
% change
Dubois, Griffith, O’Connell (TSE)
-0.54
[-18.38, -9.31]
[-5.53, 4.26]
515.34
601.74
[487.81, 540.07]
[571.33, 625.24]
-10.41
4.61
[-14.79, -5.56]
[-0.42, 9.15]
13.80
13.33
13.05
[13.77, 13.83]
[13.25, 13.42]
[12.97, 13.15]
-3.36
-5.41
[-3.96, -2.75]
[-5.95, -4.66]
3.94
3.73
3.61
[3.92, 3.95]
[3.69, 3.76]
[3.58, 3.66]
% change
Salt intensity (g/100g)
[1191.95, 1313.92]
-14.08
575.20
% change
Saturates intensity (g/100g)
[1026.40, 1142.62]
[565.94, 581.14]
% change
Nutrient score
Advertising banned
no price response
with price response
1086.79
1258.06
-5.34
-8.28
[-6.17, -4.41]
[-9.08, -7.09]
1.79
1.77
1.73
[1.79, 1.79]
[1.76, 1.77]
[1.72, 1.73]
-1.32
-3.61
[-1.67, -0.93]
[-3.95, -3.07]
Banning Junk Food Advertising
May 2015
49 / 60
Counterfactual Equilibrium
Consumer welfare
But what about welfare?
Consumers may be hurt by advertising ban
How we measure welfare depends on whether we view advertising as:
Informative about prices/characteristics (Stigler, 1961; Nelson, 1995)
A characteristic that consumers value (Stigler and Becker, 1977)
Persuasive (Marshall, 1921; Robinson, 1933; Kaldor, 1950)
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
May 2015
50 / 60
Counterfactual Equilibrium
Consumer welfare: advertising as a characteristic
If advertising is a characteristic, the payoff function represents the
consumer’s (indirect) utility function; the consumer makes decisions
to maximize utility (standard revealed preference approach)
Expected utility is given by:
Wit (pt , at ) = E
max v¯ibst
(b,s )∈Ωκ
#
"
= ln
∑
exp [αi (abt , pbst ) + ψi (abt , xb ) + γbi (at ) + ηi (zbs , ξ b )]
(b,s )∈Ωκ
Issue: the normalization of common advertising effect on outside
good will affect welfare changes of the ban
Model with v¯ibst (at ) − γ(at ) and v¯i00t − γ(at ) is observationally
equivalent but welfare change depends on γ(at ) being on inside or
outside good
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
May 2015
51 / 60
Counterfactual Equilibrium
Consumer welfare: advertising distorts decisions
Persuasive view
Advertising can lead consumers to act as non-standard decision
makers, by providing environmental “cues” to consumers (Bernheim
and Rangel, 2005).
Bernheim and Rangel (2009): “choices made in the presence of those
cues are predicated on improperly processed information, and welfare
evaluations should be guided by choices made under other conditions”
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
May 2015
52 / 60
Counterfactual Equilibrium
Consumer welfare: advertising distorts decisions
If advertising is distorting, then consumer’s (“experience”) utility
(Kahneman et al. 1997) should be evaluated in the absence of
advertising :
vbibst = αi (0, pbst ) + ψi (0, xb ) + γbi (0) + ηi (zbs , ξ b ) + eibst
Expected “experience” utility from the choice made with different
“decision” utility is:
"
#
ci (at , pt ) =E vbarg max{v¯ }
W
(b,s )∈Ωκ
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
ibst
May 2015
53 / 60
Counterfactual Equilibrium
Consumer welfare: advertising distorts decisions
Expected “experience” utility from the choice made with different
“decision” utility:
#
"
ci (at , pt ) =E vbarg max{v¯ }
W
ibst
(b,s )∈Ωκ
=Wit (pt , at )
−
∑
sibst [ (αi (abt , pbst ) − αi (0, pbst ))
(b,s )∈Ωκ
+ (ψi (abt , xb ) − ψi (0, xb )) + (γbi (at ) − γbi (0)) ]
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
May 2015
54 / 60
Counterfactual Equilibrium
Consumer welfare: advertising distorts decisions
When advertising distorts decision making, welfare impact of
advertising evaluated under preferences in absence of advertising
Denote p0 a counterfactual price equilibrium with no advertising
Welfare difference between the post and pre advertising ban is:
ci (at , pt )
Wi 0, p0t − W
ci (at , pt )
= Wi (0, pt ) − W
+Wi 0, p0t − Wi (0, pt )
(choice distortion effect)
(price competition effect)
ci (0, p) = Wi (0, p)
where we use W
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
May 2015
55 / 60
Counterfactual Equilibrium
Typical Junk Food Advertising
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
May 2015
56 / 60
Counterfactual Equilibrium
Consumer welfare: advertising distorts decisions
Choice distortion effect (£m)
Price competition effect (£m)
Total compensating variation ($m)
Change in profits ($m)
Total change in welfare (£m)
Dubois, Griffith, O’Connell (TSE)
Advertising banned
No price resp. With price resp.
35.9
35.9
[34.7, 40.5]
[34.7, 40.5]
0.0
19.2
[15.9, 21.6]
35.9
55.1
[34.7, 40.5]
[52.8, 60.1]
0.2
0.2
[-4.9, 5.9]
[-4.8, 5.3]
36.1
55.3
[32.2, 43.7]
[49.9, 62.6]
Banning Junk Food Advertising
May 2015
57 / 60
Counterfactual Equilibrium
Consumer welfare: characteristic view
Characteristics effect (£m)
Price competition effect (£m)
Total compensating variation ($m)
Change in profits ($m)
Total change in welfare (£m)
Dubois, Griffith, O’Connell (TSE)
Advertising banned
No price resp. With price resp.
-30.6
-30.6
[-38.6, -20.4]
[-38.6, -20.4]
0.0
19.2
[15.9, 21.6]
-30.6
-11.4
[-38.6, -20.4]
[-20.4, -1.3]
0.2
0.2
[-4.9, 5.9]
[-4.8, 5.3]
-30.4
-11.2
[-43.6, -14.8]
[-25.2, 3.0]
Banning Junk Food Advertising
May 2015
58 / 60
Conclusions
Aggregate impact of ban
We find that in response to introduction of an advertising ban in potato
chips markets:
Advertising ban leads to substitution to healthier products (higher
WTP)
At constant prices, quantity of potato chips purchased would decrease
But stronger price competition leads to lower prices and thus increase
in quantity consumed and total calories but not significant changes in
salt or saturated fat
Lack of advertising isn’t enough to deter bad products
Profitability in the market is almost unchanged
If advertising is viewed as distorting prices, total welfare would rise
Welfare would decrease if advertising as a characteristic (assuming
advertising affects only inside goods)
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
May 2015
59 / 60
Conclusions
Further research
Current counterfactual considered does not need to solve fully for
firms’ equilibrium policy functions
But this is not the case for tax simulation
How would firms adjust their price and advertising strategies in
response to introduction of a nutrient tax? an advertising tax?
Raises issues around multiple equilibria in counterfactuals
Dubois, Griffith, O’Connell (TSE)
Banning Junk Food Advertising
May 2015
60 / 60