The demand for wine and substitute products: A survey of the - - PowerPoint PPT Presentation

The demand for wine and substitute products:

A survey of the literature A survey of the literature

James Fogarty Economics Program Economics Program The University of Western Australia

Key findings Key findings

Demand for alcoholic beverages is price inelastic

g p

Imported beverages are more elastic Trend for more elastic demand since 1958

Country effects are generally not statistically

different

Stigler and Becker (1977 p 76) “tastes neither change Stigler and Becker (1977, p. 76) tastes neither change

capriciously nor differ importantly between people”

Wine in France is an exception

Framework of analysis matters

Consider just elasticity point estimate -- OLS Consider the point estimate and the SE -- WLS

Data for the study Data for the study

102 papers provided elasticity estimates 102 papers provided elasticity estimates

From Stone (1945) to the present English speaking country bias English speaking country bias

Occasionally more than one country considered In some cases more than one type of estimate

Beer Wine Spirits Beer Wine Spirits 154 estimates 155 estimates 162 estimates

Standard data summary: wine Standard data summary: wine

Wine Own-Price Elasticity Frequency Distribution

F

50

No Frequency

Mean: -.65 Median: -.55 St d 51

30 40

• . Observa
• St. dev.: .51

Max: .82 Min: -3.00

10 20

ations

Obs: 155

• sitive

.00

• .20
• .40
• .60
• .80
• 1.00
• 1.20
• 1.40
• 1.60
• 1.80

wards po

• nw

Elasticity Value

Summary country details for wine Summary country details for wine

Country Est Mean S D Country Est Mean S D Country Est. Mean S.D Country Est. Mean S.D

Summary country details for wine Summary country details for wine

Country Est Mean S D Country Est Mean S D Country Est. Mean S.D Country Est. Mean S.D Australia 18

• .66

.67

Summary country details for wine Summary country details for wine

Country Est Mean S D Country Est Mean S D Country Est. Mean S.D Country Est. Mean S.D Australia 18

• .66

.67 Canada 33 80 39 Canada 33

• .80

.39

Summary country details for wine Summary country details for wine

Country Est Mean S D Country Est Mean S D Country Est. Mean S.D Country Est. Mean S.D Australia 18

• .66

.67 Canada 33 80 39 Canada 33

• .80

.39 Cyprus 2

• .40

.23 Denmark 2

• .61

.45 Finland 9

• 1.14

.63 France 3

• .07

.02 Germany 1

• .38
• Ireland

3

• 1.33

.46 Italy 1 1 00 Italy 1

• 1.00
• Japan

2

• .10

.05

Summary country details for wine Summary country details for wine

Country Est Mean S D Country Est Mean S D Country Est. Mean S.D Country Est. Mean S.D Australia 18

• .66

.67 N’lands 1

• .50
• Canada

33 80 39 N Z 8 56 28 Canada 33

• .80

.39

• N. Z.

8

• .56

.28 Cyprus 2

• .40

.23 Norway 7

• .37

.43 Denmark 2

• .61

.45 Poland 1 .82

• Finland

9

• 1.14

.63 Portugal 1

• .68
• France

3

• .07

.02 Spain 3

• .98

3 Germany 1

• .38
• Sweden

12

• .83

.41 Ireland 3

• 1.33

.46 U.K. 39

• .72

.56 Italy 1 1 00 U S 31 55 45 Italy 1

• 1.00
• U.S.

31

• .55

.45 Japan 2

• .10

.05

Meta-analysis framework Meta analysis framework

Meta-analysis question: Meta analysis question:

Is the observed variation in elasticity estimates

due to sampling error only? due to sampling error only?

Stepwise process of analysis

St id th fi d ff t d l

Step one: consider the fixed effects model Step two: consider the random effects model If both the fixed and random effects models

are rejected design a meta-regression

Meta-analysis approaches Meta analysis approaches

Fixed effects model Fixed effects model

Find the weighted mean where the weights

are the inverse of the estimate variance are the inverse of the estimate variance

Test statistic is based on the sum of the

weighted mean square differences g q

High values lead to rejection of null that the

reported elasticity estimates are from the p y same population

Meta-analysis approach continued Meta analysis approach continued

Random effects model Random effects model

Proceed as for fixed effects but reduce the

weight to very precise estimates weight to very precise estimates

Meta-regression approach

Ob ti b d t th

Observations can be grouped together

according to study characteristics Grouping are likely to be based around

Grouping are likely to be based around

country, estimation method, time period, data frequency, etc. q y,

Compensated wine estimates Compensated wine estimates

, ,

Compensated wine estimates

100

Compensated wine estimates

Est. ⎛ ⎞ ⎜ ⎟

100

SE ⎛ ⎞ ⎜ ⎟ ⎝ ⎠

75

, ,

50 25

• 2
• 1.5
• 1
• 0.5

0.5 1

Compensated wine estimates

100

Compensated wine estimates

Est. ⎛ ⎞ ⎜ ⎟

75 100

SE ⎛ ⎞ ⎜ ⎟ ⎝ ⎠

Unweighted mean: -.62

75

, ,

50 25

• 2
• 1.5
• 1
• 0.5

0.5 1

Compensated wine estimates

100

Compensated wine estimates

Est. ⎛ ⎞ ⎜ ⎟

75 100

SE ⎛ ⎞ ⎜ ⎟ ⎝ ⎠

Unweighted mean: -.62 Fixed effects mean: -.83

75 50 25

• 2
• 1.5
• 1
• 0.5

0.5 1

Compensated wine estimates

100

Compensated wine estimates

Est. ⎛ ⎞ ⎜ ⎟

75 100

Unweighted mean: -.62 Fixed effects mean: -.83 R d ff t 57

SE ⎛ ⎞ ⎜ ⎟ ⎝ ⎠

75

Random effects mean: -.57

50 25

• 2
• 1.5
• 1
• 0.5

0.5 1

Summary testing results Summary testing results

Model Held constant Result Model Held constant Result

Summary testing results Summary testing results

Model Held constant Result Model Held constant Result

Fixed Effects Beverage Always reject Beverage and country Always reject Beverage and country Always reject

Summary testing results Summary testing results

Model Held constant Result Model Held constant Result

Fixed Effects Beverage Always reject Beverage and country Always reject Beverage and country Always reject Random Effects Beverage Always reject B d t Al j t Beverage and country Always reject

Summary testing results Summary testing results

Model Held constant Result Model Held constant Result

Fixed Effects Beverage Always reject Beverage and country Always reject Beverage and country Always reject Random Effects Beverage Always reject B d t Al j t Beverage and country Always reject

So try meta-regression

WLS where weights are inverse variance

Interesting findings: Time Interesting findings: Time

The time trend variable

Enters as a quadratic, 1958 is the point of most inelastic demand The trend is gentle and between 1958 and 1994

the implied trend increase in elasticity is .13

OLS

between 1958 and 1994 more inelastic

OLS – between 1958 and 1994 more inelastic

A possible relationship with illicit substances

Marijuana Ecstasy Speed etc could be Marijuana, Ecstasy, Speed, etc. could be

substitutes

Speculative so other suggestions welcome

Interesting findings: Country effects Interesting findings: Country effects

Pair-wise testing – 66 comparisons per beverage

Pair wise testing 66 comparisons per beverage

Beer Wine Spirits Average Rejection Rates Beer Wine Spirits 12 percent 21 percent 12 percent

The main exceptions relate to wine:

Wine in France: 73 percent rejection rate (inelastic) Wine in France: 73 percent rejection rate (inelastic) Wine in UK: 45 percent rejection rate (elastic) Wine Canada: 45 percent rejection rate (elastic)

Wine Canada: 45 percent rejection rate (elastic)

Beer in NZ: 45 percent rejection rate (inelastic)

Final points of note Final points of note

Paper available with details and an appendix Paper available with details and an appendix

covering each paper

The approach could be a useful framework

pp for some of the hedonic literature on expert

• pinion etc.