An Econometric Analysis of Wine Consumption in Australia

E.A. Selvanathan and Saroja Selvanathan

[email protected] Griffith Business School Griffith University Nathan, Queensland 4111 AUSTRALIA

December 2009 Paper for the pre-AARES conference workshop on The World’s Wine Markets by 2030: Terroir, Climate Change, R&D and Globalization, Adelaide Convention Centre, Adelaide, South Australia, 79 February 2010.

An Econometric Analysis of Wine Consumption in Australia E.A. Selvanathan and Saroja Selvanathan

Abstract: According to some media reports that global wine consumption, especially in Europe’s major wine-producing and consuming countries such as France, Italy and Germany, is on the decline, due to the current global economic crisis. However, the same media also reports that wine consumption in a number of other countries, including Australia, Canada, the US and the UK, are still on the increase. For example, domestic wine sales in Australia have increased by 6 percent between April 2009 and May 2009 and the white table wine sales have increased by 9.2 percent during the same period. Also, the domestic wine sales over the 12-month period between May 2008 and May 2009 has increased by 4.7 percent and the white table wine sales has increased by 7.1 percent during the same period. In light of these observations, this paper analyses the factors that influence the demand for wine in Australia and compare the situation with three other countries, Canada, the UK and the US. We use a system-wide framework to analyse the demand for beer, wine and spirits with more focus on the demand for wine in Australia. Keywords: alcohol consumption, econometric analysis, simulations

1.

Introduction

During the last two decades the demand for alcohol as a whole in most countries is on the decline. It is believed that such a fall in alcohol consumption is mainly due to the campaign made by the social welfare groups and health profession on the problems associated with excessive alcohol consumption and the control policies on misuse/abuse of alcohol introduced by a number of governments around the world. There is also increasing pressure by the society on alcohol consumers not to drink and drive to the extent of the introduction of zero tolerance limits on the roads in a number of countries. As a result, a significant proportion of alcohol drinkers were forced to shift their habit of drinking from high level alcohol content drinks to low level alcohol content drinks or drink alcohol from home. This could have resulted in a shift in peoples’ habit of having daily social drinks such as beer after a hard working day at workplace to having a complementary meal time drink such as wine at home. In this paper we analyse the consumption patterns of beer, wine and spirits in Australia over the last five decades with more emphasis on wine and compare the Australian situation with Canada, the UK and the US. According to alcohol statistics published by the OECD (see Figure 1, below), in 2007, Australians consumed 9.9 litres of pure alcohol per adult just above the OECD average of 9.7 litres of pure alcohol per adult. It can also be seen from Figure 1 that alcohol consumption shows a negative growth during 1980-2007 in Australia (-23%) as well as in Canada (-24%) and the US (-17%) while showing a positive growth in the UK (19%). Alcohol consumption, population aged 15 years and over, 2007 (or latest year available)

Turkey Mexico Norway Sweden Iceland Japan Korea Canada Italy United States Slovak Republic Greece New Zealand Netherlands OECD Australia Germany Poland Switzerland Finland Belgium United Kingdom Portugal Spain Czech Republic Denmark Austria France Hungary Ireland Luxembourg

1.2 4.6 6.6 6.9 7.5 7.7 8.0 8.1 8.1 8.6 8.9 9.0 9.2 9.6 9.7 9.9 9.9 10.3 10.4 10.5 10.7 11.2 11.4 11.7 12.1 12.1 12.9 13.0 13.2 13.4 15.5 20

15

10

5

Change in alcohol consumption per capita, population aged 15 years and over, 1980-2007

0

-33 39 25 3 74 8 n.a. -24 -50 -17 -39 -32 -22 -17 -13 -23 -30 18 -23 33 -21 19 -23 -36 3 3 -11 -33 -11 40 16 -75

-25

25

75

Figure 1: Alcohol consumption in 2007 and Change in alcohol consumption between 1980 and 2007, OECD countries

Table 1 presents the trend in alcohol consumption in Australia, Canada, UK, USA and the OECD average from 1980 to 2007. As can be seen, per adult Australian alcohol consumption is higher than that for Canada and the US. The UK alcohol consumption was lower than that for Australia in the 1980s and early 1990s but is higher than the Australian consumption since 1996. Overall, there is a declining trend in pure per adult alcohol consumption in Australia, Canada and the US while there is an increasing trend in the UK. The level of alcohol consumption in the US and Canada generally appear to be lower than that of Australia.

Table 1: Trends in alcohol consumption, selected OECD countries, 1980 to 2007 Year Australia Canada 1980 12.9 10.7 1981 13.0 10.8 1982 12.6 10.5 1983 12.3 10.2 1984 11.6 9.9 1985 11.7 9.8 1986 11.2 9.6 1987 11.3 9.6 1988 11.2 9.4 1989 10.9 7.5 1990 10.5 7.4 1991 10.0 7.3 1992 9.7 7.5 1993 10.0 7.4 1994 9.8 7.3 1995 9.6 7.4 1996 9.8 7.3 1997 9.9 7.3 1998 9.8 7.5 1999 9.6 7.6 2000 9.8 7.6 2001 9.6 7.7 2002 10.0 7.7 2003 9.8 7.8 2004 9.9 7.8 2005 9.9 8 2006 9.9 8.1 2007 Source: OECD Health Data 2009.

UK 9.4 9.1 8.8 9.1 9.2 9.3 9.3 9.5 9.8 9.8 9.8 9.4 9.3 9.3 9.6 9.4 9.8 10.0 9.8 10.3 10.4 10.7 11.0 11.2 11.5 11.4 11.0 11.2

USA 10.4 10.5 10.5 10.3 10.2 10.0 9.9 9.8 9.6 9.4 9.2 9.3 8.7 8.7 8.4 8.3 8.1 8.2 8.1 8.1 8.2 8.3 8.3 8.3 8.4 8.4 8.6

OECD 11.2 11.1 11.0 11.0 10.9 10.8 10.6 10.5 10.4 10.4 10.5 10.3 10.1 10.0 9.8 9.7 9.7 9.8 9.7 9.7 9.8 9.7 9.7 9.7 9.7 9.7 9.7 9.8

Figure 2 presents the per adult pure alcohol consumption in the four countries between 1960 and 2007. As can be seen, alcohol consumption peaked in Australia, Canada, UK and the US during the late 1970’s. Since early 1980s, alcohol consumption declined and became almost to a constant level in the three countries Australia, Canada and the US, but in the UK, after an intial decline it has continued to increase. The OECD statistics on liver related diseases and alcohol consumption, displayed in Figure 3, also shows a positive relationship which has been used to support the case for the 2

11 10 9 8 7 6 5

2005

2002

1999

1996

1993

1990

1987

1984

1981

1978

1975

1972

1969

1966

1963

4

1960

Liters of pure alcohol per adult

14 13 12

Year Australia

Canada

United Kingdom

United States

Figure 2: Per adult pure alcohol consumption, 4 Countries, 1960-2007

45 40

Liver diseases

35 30 y = 1.2563x - 2.8008 R2 = 0.2421

25 20 15 10 5 0 4

6

8

10

12

14

16

18

Alcohol conaum ption (liters pure alcohol per adult)

Figure 3: Liver diseases vs Alcohol consumption, 26 OECD countries

health profession which has forced several governments to take firm action to introduce new policies and educational measures to curb excessive alcohol drinking. In Section 2, we present a preliminary data analysis of the Australian alcohol consumption data and compare them with the alcohol consumption in Canada, the UK and the USA.

3

2.

Basic Data

Columns 2-4 of Table 2 and Figure 4 present the annual per capita Australian consumption in litres (qit) of beer, wine and spirits for the period 1955-2005. These data are obtained from Australian Bureau of Statistics, Cat nos. 4306.0 and 4315.0. As can be seen, the per capita consumption of beer has peaked during mid 70’s to about 140 litres per capita and steadily declined below the 1950’s level during the 2000 to 2005 to about 86 litres per capita. The wine consumption has steadily increased from 5 litres per capita in 1955 to 23 litres per capita in 2005. The spirits consumption has increased from 2 litres per capita to 4 litres per capita during the same period. Columns 5-7 of Table 2 and Figure 5 present the annual price indices (pit) for the three beverages with base period 1980=100. The price data are obtained from ABARE (2009). As can be seen, the prices of the three beverages increased at a slower rate until the mid 70’s and increased at a faster rate from then onwards. Among the three, beer and spirits prices increased at a faster rate than the wine prices. Table 2: Per capita annual consumption and price index (1980=100), beer, wine and spirits, Selected years Year

Beer (2) 105.98 110.39 137.39 115.47 95.30 86.40

(1) 1955 1965 1975 1985 1995 2005

Consumption Wine Spirits (3) (4) 5.34 2.00 6.08 2.09 13.01 2.87 21.60 3.05 18.30 3.18 22.57 4.33

Price Index (1980=100) Beer Wine Spirits (5) (6) (7) 21.90 20.60 23.10 30.00 33.70 31.80 63.60 75.40 60.40 162.40 136.40 147.10 273.19 214.89 273.67 393.03 262.63 355.08

130 110 90 70 50 30

2005

2003

2001

1999

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

1977

1975

1973

1971

1969

1967

1965

1963

1961

1959

-10

1957

10

1955

Consumption (liters/capita)

150

Year

Beer

Wine

Spirits

Figure 4: Per capita consumption of beer, wine and spirits, Australia, 1955-2005 4

Price Index (1980=100)

450

360 270

180 90

2005

2003

2001

1999

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

1977

1975

1973

1971

1969

1967

1965

1963

1961

1959

1957

1955

0

Year

Beer

Sprits

Wine

Figure 5: Price of beer, wine and spirits ((1980=100), Australia, 1955-2005 Table 3 presents the quantity and price log-changes for the three beverages for selected years. The log-changes in quantity and prices from year t-1 to t are calculated as Dqit = ln qit – ln qit-1 and Dpit = ln pit – ln pit-1, where qit and pit are the per capita consumption and prices of beverage i (=beer, wine and spirits) in year t. The last row of the table presents the averages of these variables over the sample period. As can be seen from the table, the average annual growth in per capita consumption of beer has declined at a rate of 0.4 percent per annum while wine and spirits consumption has increased at an annual growth rate of 2.9 and 1.6 percent. On average, the prices of beer, wine and spirits have increased at a rate of 5.8, 5.1 and 5.5 percent per annum. Figures 6 and 7 display the quantity and price log-changes for beer, wine and spirits.

Table 3: Quantity and Price log-changes, beer, wine and spirits, Australia, Selected years (1) 1955 1965 1975 1985 1995 2005

Quantity log-changes Beer Wine Spirits (2) (3) (4) -5.36 -2.46 -10.54 0.05 8.76 -11.73 -2.18 5.77 -3.42 0.88 1.63 1.65 -1.56 -0.54 -0.63 -0.25 -0.15 1.62

Mean

-0.41

Year

2.88

1.55

Price log-changes Beer Wine Spirits (5) (6) (7) 14.43 6.12 0.00 8.34 7.39 13.80 22.67 18.60 11.94 8.48 2.49 9.18 5.32 5.06 2.70 4.53 0.79 2.77 5.77

5.09

5.47

All entries are to be divided by 100.

5

20 15 Percentage change

10 5

2004

2002

2000

1998

1996

1994

1992

1990

1988

1986

1984

1982

1980

1978

1976

1974

1972

1970

1968

1966

1964

1962

1960

1958

-5

1956

0

-10 -15 -20 -25 Year

Beer

Spirits

Wine

Figure 6: Beer, wine and spirits consumption in log-changes, Australia, 1955-2005

35

25 20 15 10 5

2004

2002

2000

1998

1996

1994

1992

1990

1988

1986

1984

1982

1980

1978

1976

1974

1972

1970

1968

1966

1964

1962

1960

-5

1958

0

1956

Percentage change

30

-10 Year

Beer

Wine

Spirits

Figure 7: Beer, wine and spirits prices in log-changes, Australia, 1955-2005 Columns 2-4 of Table 4 present the unconditional budget share of the three beverages (wit, i=1,2,3) and in the total consumer expenditure (Mt), where wit = pitqit/Mt and Mt = ∑ in=1 pit qit , and column 5 presents the total budget share of alcohol (Wgt), where Wgt = ∑ 3i =1 wit . The consumer expenditure allocation to alcohol on their total budget has almost halved from 6.1 percent in 1955 to 3.6 percent in 2005. During 1955 to 2005, beer share on the total budget has fallen from 4.6 percent to 1.9 percent, wine share has increased from 0.6 percent in 1955 to 1.31 percent in 1985 and has fallen to 0.9 percent in 2005, and spirits share has been stable around 0.7-0.9 percent. Columns 6-8 of Table 4 present the conditional budget share within alcohol ( wit' , i = 1,2,3), where wit' = pitqit/Mgt with Mgt = ∑ 3i =1 pit qit . Within alcohol, beer share declined

6

from 75.9 percent in 1955 to 52.4 percent in 2005, wine has increased from 9.9 percent to 25.2 percent and spirits has also increased from 14.2 percent to 22.4 percent during the same period. On average, over the 5 decades, consumers allocated about 5 percent of their income on alcohol with 3.3 percent to beer, 1 percent to wine and 0.7 percent to spirits. Within alcohol, for every dollar the Australians spent on alcohol, on average, he/she allocated 65 cents on beer, 20 cents on wine and the remaining 15 cents on spirits. Figures 8 and 9 plot these unconditional and conditional budget shares for beer, wine and spirits over the 5 decades.

Table 4: Unconditional and conditional budget shares, beer, wine and spirits, Australia, Selected years Unconditional budget shares Beer Wine Spirits (2) (3) (4) 4.63 0.60 0.87 4.05 0.69 0.77 3.94 1.22 0.74 3.03 1.31 0.68 2.31 0.96 0.73 1.90 0.91 0.81

Year (1) 1955 1965 1975 1985 1995 2005 Mean

3.28

0.98

Conditional budget shares Beer Wine Spirits (6) (7) (8) 75.86 9.91 14.23 73.54 12.55 13.91 66.83 20.68 12.49 60.30 26.11 13.59 57.76 24.05 18.19 52.41 25.21 22.37

Total (5) 6.10 5.51 5.90 5.03 4.00 3.63

0.73

4.99

64.60

20.26

15.14

7 6

Percentage

5 4 3 2 1

2005

2003

2001

1999

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

1977

1975

1973

1971

1969

1967

1965

1963

1961

1959

1957

1955

0

Year

Beer

Wine

Spirits

Total

Figure 8: Unconditional budget shares of beer, wine and spirits and total alcohol share, Australia, 1955-2005

7

90 80

Percentage

70 60 50 40 30 20 10

2005

2003

2001

1999

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

1977

1975

1973

1971

1969

1967

1965

1963

1961

1959

1957

1955

0

Year

Beer

Wine

Spirits

Figure 9: Conditional budget shares of beer, wine and spirits, Australia, 1955-2005

We now summarise the price and quantity data for the alcoholic beverages group (Sg) in the form of Divisia price index, 3

DPgt

=

∑w

' it

i =1

Dpit ,

Divisia quantity index, 3

DQgt

=

∑w

' it

i =1

Dqit ,

the corresponding Divisia price variance, Πgt

3

=

∑w

' it

i =1

( Dpit − DPgt ) 2 ,

the Divisia quantity variance, 3

Kgt

=

∑w i =1

' it

( Dqit − DQ gt ) 2 ,

and the associated Divisia price-quantity covariance,

Γgt

3

=

∑w i =1

' it

( Dqit − DQ gt )( Dpit − DPgt ),

and Divisia price-quantity correlation coefficient,

ρgt

=

Γgt Π gt K gt

.

where wit/ = ½ ( wit' + wit' −1 ) is the arithmetic average of the conditional budget shares of beverage i in periods t and t-1. 8

Table 5 presents these Divisia summary measures. The last row of the table presents the average of each column. Entries in columns 2 and 3 indicate that, on average, the price of alcohol has increased by 5.5 percent per annum and the per capita alcohol consumption has increased by only 0.6 percent per annum. Comparing columns 4 and 5, we see that, on average, quantity variance is greater than the corresponding price variance. (It is worth noting that this relationship holds for the individual variances for a majority of years throughout the sample period.). A plot of the quantity variances against the price variances presented in Figure 10 reveals that most of the points lie above the 450 degree line through the origin indicating that on average, quantity variance exceeds the price variance. This pattern agrees well with the results of a number of previous studies, e.g., Clements (1982), Meisner (1979), Theil and Suhm (1981), S. Selvanathan (1993) and Selvanathan and Selvanathan (1994, 2003, 2005, 2007). The pricequantity correlations presented in column 7 of the table are mostly negative with an average of -0.4. This reflects the tendency of the drinkers to move away from those beverages having above average price increases. Under the assumption of unitary income elasticity of each alcoholic beverage, the demand equation for a beverage i can be written as (Dqit – DQgt) = φ(Dpit – DPgt)

i=1,2,3 and t=1,2, .., T.

Therefore, if we plot the relative consumption changes (Dqit – DQgt) against the relative price changes (Dpit – DPgt), we would expect the points to be scattered around a straight line with slope φ. This slope coefficient φ is known as the income flexibility (which is the inverse of the income elasticity of the marginal utility of income). Figure 11 presents the plots for the three

Quantity variance

50 40

y=x

30 20 10 0 0

5

10

15 20 Price variance

25

30

Figure 10: Divisia quantity variance vs price variance of alcohol, Australia, 1956-2005 9

Table 5

Divisia price and quantity summary measures

Year (1) 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005

Divisia price index, DPgt (2) 11.72 1.40 0.90 1.39 3.39 0.90 1.27 1.88 1.76 9.01 3.10 4.00 4.94 3.83 6.47 4.56 3.59 8.51 14.96 20.43 8.11 4.99 14.19 7.34 7.38 8.77 11.25 8.35 7.41 6.98 9.31 8.01 4.09 6.10 6.67 3.69 2.40 3.43 3.50 4.78 2.68 1.15 1.52 1.50 7.25 2.79 2.24 4.30 3.39 3.18

Mean

5.50

Divisia price- Divisia pricequantity quantity Divisia quantity Divisia price Divisia quantity variance, Πgt variance, Kgt covariance, Γgt correlation, ρgt index, DQgt (3) (4) (5) (6) (7) -5.76 26.80 4.19 6.39 0.60 0.22 1.48 0.19 -0.49 -0.92 1.04 2.85 2.00 -0.33 -0.14 2.49 4.51 5.36 -1.83 -0.37 -0.60 6.88 0.65 -1.35 -0.64 -0.20 0.08 0.14 0.10 0.91 0.94 3.36 2.16 -2.17 -0.80 4.10 0.36 2.16 -0.80 -0.92 3.39 2.33 2.84 -1.10 -0.43 -0.59 3.95 28.66 -9.81 -0.92 3.20 9.53 11.74 5.38 0.51 5.40 3.39 12.03 -1.06 -0.17 3.39 11.36 7.17 5.49 0.61 4.55 0.97 9.06 0.95 0.32 0.78 2.95 3.25 -2.13 -0.69 0.96 2.71 3.75 -2.60 -0.82 5.46 6.28 13.62 -8.77 -0.95 6.98 24.80 9.02 -14.62 -0.98 2.23 11.34 21.96 -6.10 -0.39 -0.74 13.54 10.91 -1.54 -0.13 1.64 2.82 11.07 -5.38 -0.96 0.66 0.65 8.01 -0.24 -0.11 -1.32 49.65 107.83 -72.43 -0.99 1.39 5.19 9.01 -6.36 -0.93 0.83 2.57 17.43 -6.46 -0.96 1.45 2.69 6.70 -4.06 -0.96 -2.48 5.79 15.94 -9.45 -0.98 -1.74 0.72 8.98 -1.52 -0.60 0.29 0.34 15.59 -2.32 -1.00 1.18 7.37 0.14 -0.67 -0.67 -3.69 0.71 0.10 -0.19 -0.72 1.66 1.39 2.69 1.44 0.74 0.07 9.09 18.80 -6.82 -0.52 -1.93 2.28 1.63 1.53 0.79 -3.98 10.26 3.44 -1.65 -0.28 -3.52 0.97 20.72 -1.85 -0.41 -2.41 0.32 10.47 -1.45 -0.79 2.36 0.38 41.70 -1.90 -0.48 -2.28 0.70 5.08 0.27 0.14 -1.15 0.98 0.24 -0.29 -0.60 0.29 0.31 6.49 -1.41 -0.99 3.41 1.81 38.99 -1.95 -0.23 -0.77 0.02 1.19 -0.16 -0.94 -1.00 0.58 11.33 -1.25 -0.49 2.23 2.17 12.75 -2.26 -0.43 -0.89 0.02 5.79 0.32 0.98 3.15 0.89 5.63 -0.90 -0.40 -0.70 4.34 10.86 -6.69 -0.98 0.37 1.09 5.34 -2.29 -0.95 0.19 2.45 0.58 -0.24 -0.20 0.61

5.16

11.11

-3.46

-0.42

10

Relative Quantity changes

Beer 3 y = -0.4192x - 0.9019

2 1 0 -3

-2

-1

-1

0

1

2

3

-2 -3 -4 -5 Relative price changes

Wine Relative quantity changes

20 y = -0.5204x + 2.0615

15 10 5 0 -10

-7

-4

-1

2

5

8

-5 -10 Relative price changes

Spirits 15

Relative quantity changes

y = -0.6537x + 0.9154 5

-15

-10

-5

-5

0

5

10

15

-15

-25 Relative price changes

Figure 11: Relative quantity changes vs relative price changes, Beer, Wine and Spirits, Australia, 1956-2005

beverages. As can be seen, the estimated value of income flexibilities are -0.42 for beer, -0.52 for wine and 0.65 for spirits, giving an average value of -0.53. This estimate of φ is very close to the value obtained in a number of consumption studies (eg, see, Chen, 2001; Selvanathan, 1993; Theil and Clements, 1987; Theil and Suhm, 1981; and Selvanathan and Selvanathan, 2003). 11

3.

The Demand System for Beer, Wine and Spirits

The absolute price version of the conditional demand equation for beverage i can be written as (e.g., see Theil and Clements 1987, pp.242-243) 3

wit' Dqit = θ i' DQ gt + ∑ π ij' Dp jt + ε it , j =1

i=1,2,3,

(3.1)

where wit' , Dqit, DQgt and Dpit are as defined in Section 2. The coefficient θ i' is the conditional marginal share of beverage i and π ij' is the Slutsky price coefficient. The parameters of equations (3.1) satisfy the adding-up restrictions 3

' ∑θi = 1

i =1

3

' ∑ π ij = 0, j=1,2,3.

and

i =1

(3.2)

The conditional marginal share θ i' = ∂(pitqit)/∂Mgt, answers the question that, if an additional income of $1 is allocated to alcohol, how much of this is spent on beverage i. If the marginal share for a beverage is positive, then it is classified as a normal good and if it is negative then the beverage is classified as inferior. The ratio of conditional marginal share to conditional budget share is called the conditional income elasticity of good i. That is,

η i' =

∂ (log qi ) θ' = i , ∂ (log M g ) wi'

i=1,2,3.

(3.3)

The conditional income elasticity of beverage i, η i' , answers the question, if expenditure on alcohol increases by 1 percent with prices held constant, what is the percentage change in consumption of beverage i? Beverages with income elasticity less than one are called necessities, while those with income elasticity greater than one are called luxuries. If the income elasticity is negative, the beverage is said to be inferior as its consumption falls with increasing income. The Slutsky coefficient measures the total substitution effect on the demand for beverage i of a change in the jth price. These Slutsky coefficients are symmetric in i and j (also known as Slutsky symmetry), π'ij = π' ji ,

i, j =1,2,3

(3.4)

and satisfy demand homogeneity

12

3

' ∑ π ij = 0,

j =1

i=1,2,3.

(3.5)

Demand homogeneity postulate that a proportionate change in all prices has no effect on the demand for any good when real income remains constant. The n × n Slutsky matrix (here we have n=3) formed by the Slutsky coefficients is a symmetric negative semi-definite n × n matrix with rank n-1. The price elasticity of demand for beverage i with respect to the price of beverage j is given by ' d (log qi ) π ij ' = η ij = d (log p j ) wij'

(3.6)

The price elasticity η ij' measures the percentage change in qi resulting from a one percent change in price pj, assuming income and other prices remain constant.

4.

Empirical Results

We estimated the demand equations with constant terms added. These constant terms take account of autonomous trend in consumption. As one of the three estimating equations is redundant, we estimate only the first two equations (see Barten, 1969). We take total consumption of alcohol and the prices to be predetermined and use the DEMMOD (Barten et.al, 1989) program for estimation. We estimated the system of equations (3.1) for beer, wine and spirits, tested demand homogeneity and Slutsky symmetry and found both hypotheses were acceptable by the data. Table 6 presents the estimated coefficients of the system of equations with homogeneity and symmetry restrictions imposed. As can be seen, most of the estimated coefficients are statistically significant at the 5 percent significant level. The estimated constant terms show that there is an autonomous trend out of beer into wine and spirits. The estimated marginal shares indicate that if expenditure on alcohol increases by one dollar, 55 cents of that increase will be spent on beer, 19 cents on wine and the remaining 26 cents on spirits. All the own-Slutsky (diagonal) coefficients are negative as expected and are also statistically significant. All the cross-Slutsky coefficients are positive indicating that the three beverages are pair-wise substitutes. Table 7 presents the income and price elasticities calculated at sample mean budget shares using equations (3.3) and (3.6) implied by Table 6 estimates. The estimated income 13

alasticities of beer, wine and spirits are 0.78, 0.82 and 1.92, respectively. These income elasticity estimates indicate that beer and wine are necessities (η i' < 1) , and spirits is a luxury (η i' > 1) . All the own-price elasticities are less than one in absolute value, indicating that demand for all the three beverages are price-elastistic. Furthermore, all the cross-price elasticities are positive implying that the three beverages are pairwise substitutes.

Table 6: Estimated coefficients and standard errors Rotterdam model: Symmetry-constrained with constants 3

wit' Dqit = αi + θ i' DQgt + ∑ π ij' Dp jt + εit j =1

(Standard errors are in parentheses) _____________________________________________________________________________ Commodity

Intercept αi×100

Conditional marginal share

Slutsky price coefficients ______________________________________

θ i'

i

πi'1

πi' 2

πi' 3

____________________________________________________________________________________________ Beer (1)

-0.420 (0.130)

0.496 (0.047)

-0.133 (0.039)

0.057 (0.033)

0.076 (0.025)

Wine (2)

0.348 (0.100)

0.147 (0.036)

0.059 (0.030)

-0.086 (0.025)

0.028 (0.019)

Spirits (3)

0.072 0.357 0.074 -0.029 -0.103 (0.137) (0.050) (0.041) (0.035) (0.026) _____________________________________________________________________________________________ Note: Coefficient estimates are given in line 1 and standard errors in line 2.

Table 7: Estimated income and price elasticities _____________________________________________________________________________ Price Commodity Income ___________________________ Beer Wine Spirits _____________________________________________________________________________ Beer

0.780

-0.209

0.093

0.116

Wine

0.824

0.319

-0.484

0.164

Spirits

1.923

0.408

0.148

-0.556

14

5.

What caused wine and spirits consumption to grow and beer consumption to fall?

Earlier in the paper we noticed that while beer consumption has fallen, wine and spirits consumption has increased at a faster rate. On average, beer consumption has fallen at a rate of 0.4 percent per annum, the wine consumption has increased at a rate of 3 percent per annum and spirits consumption has increased at a rate of 1.5 percent per annum. What caused wine and spirits consumption to grow and beer consumption to fall? In this section we investigate the reasons for the growth in consumption of wine and spirits and fall in the consumption of beer using the estimation results given in Tables 6 and 7. In order to make such investigation, we decompose the growth in consumption of beer, wine and spirits in terms of autonomous trend, income, own price and cross-prices. If we divide both sides of demand equation (3.1), with a constant term added, by the budget share wit' to give

Dqit =

αi wit'

+

θ i' wit'

3

DQ gt + ∑

π ij'

' j =1 wit

ε Dp jt + it , wit'

i=1,2,3.

(5.1)

That is, 3

Dqit = α i* + η i' DQ gt + ∑ η ij' Dp jt + ε it' , j =1

i=1,2,3.

(5.2)

where α i* = α i / wit' is the autonomous trend in consumption of beverage i and η i' and η ij' are income and price elasticities. Therefore,

Total growth in consumption of beverage of i

(5.3)

= Autonomous trend component (α i* ) + income component (ηi) + own-price component (πii) + cross-price component (πij) + Residual component(εit)

Table 8 presents the components in the above equations for beer, wine and spirits calculated based on the estimates presented in Tables 6 and 7. The last row of the table gives the averages over the whole sample period. As can be seen from the beer columns, the beer consumption in Australia has fallen in majority of the years with an average growth rate of -0.4 percent per annum. This fall in beer consumption growth is made up as follows: shift in preferences or the autonomous trend contributed, on average, a shift away from beer by -0.66 percent per annum, the growth income has contributed positively to the beer consumption growth at a rate of 0.48 15

16

percent per annum, own-price of beer contributed negatively at a rate of -1.21 percent per annum, cross-prices of wine and spirits contributed positively together at a rate of

1.11

(=0.46+0.65) percent per annum and residual -0.13 percent per annum. Overall, on the beer consumption, the positive contribution of income and cross-prices were offset by the negative contribution of shift in preferences and own-price, resulting in a negative growth. As can be seen from the wine columns, on average, the wine consumption grew at a rate of 2.9 percent per annum. This positive growth is due to the shift in preference towards wine (1.91 percent per annum), income growth (0.5 percent per annum), and cross-price increase of beer and spirits combined 2.74 percent per annum (=1.84+0.90). Own price of wine contributed -2.46 percent per annum to its own growth. Overall, shift in preferences, income and crossprices of beer and spirits contributed positively to the growth in wine consumption, which were offset by its own-price but resulted in a positive overall growth. The spirits columns reveal that spirits consumption has increased at a rate of 1.6 percent per annum about half that of wine growth. Even though shift in consumer preferences played a positive role with regard to spirits against the negative contribution in the case of beer, shift in preferences contributed to wine almost 4 times that of spirits. The major factors that contributed towards the positive growth in spirits are the income (1.17 percent per annum) and the crossprices of beer and wine 3.11 (=2.36+0.75) percent per annum which is strong enough to offset spirits’ own-price contribution of -3.04 to give an overall positive growth in spirits consumption. Thus the conclusion of the analysis in this section is that, the strong shift in preferences towards wine and spirits and against beer are the reasons for the positive growth in wine and spirits and negative growth in beer consumption. Income also plays a significant positive role in the growth in consumption of all three beverages but at an almost equal rate. For all three beverages, the positive growth due to cross-prices has been almost offset by their negative ownprice contribution.

6.

Simulation of alcohol Consumption

Now we use the demand model (5.2) to simulate alcohol consumption under various scenarios. First, we investigate what happens to the consumption of beer, wine and spirits (1) if total alcohol consumption remains at the same level. That is, zero growth rate in total alcohol cnsumption. (2) If no constant terms in the demand equations. That is, no shift in consumer preferences.

17

What if total alcohol consumption doesn’t grow? It is interesting to know what happens to the level of beer, wine and spirits consumption if total alcohol consumption doesn’t grow. This condition is equivalent to setting DQgt = 0 in the demand equation (5.1). If Dqits is the simulated consumption log-change, then we can easily show that when DQgt = 0,

Dqits = Dqit − η i' DQ gt .

(6.1)

We then convert the simulated changes to level form as

qits = qits −1 exp( Dqits )

(6.2)

We first evaluate the values of Dqits from equation (6.1) using estimates from Table 6 and 7 and then input those values into equation (6.2) to get the simulated consumption values, qits . We set

qis1955 = qi1955 , i=1,2,3. Columns 5-7 of Table 9 give the simulated consumption of beer, wine and spirits assuming that the level of total alcohol consumption remains at the same level as in 1955. Columns 2-4 of Table 9 gives the actual observed consumption. Last row of the table gives the sample averages. As can be seen, the simulated per capita consumption for the year 2005 for beer is 68.1 litres, wine is 17.6 litres and spirits is 4.10 litres. Accordingly, if the total alcohol consumption did not grow from 1955, it would have caused the beer consumption to be about (86.40-68.09)/86.40 = 21.2% lower, wine to be about (22.57-17.55)/22.57 = 22.2% lower and spirits to be (4.33-4.10)/4.33 = 5.2 percent lower, than otherwise.

What if there weren’t any shift in consumer preferences? It is also interesting to know what happens to the level of beer, wine and spirits consumption if there is no change in the shift of the consumer preferences. This condition is equivalent to setting each constant term α i* = 0 in the demand equation (5.2), for i=1,2,3. If Dqits is the simulated consumption log-change, we can easily show that when α i* =0, i=1,2,3, Dqits

= Dqit −

α i* wit'

.

(6.3)

We then convert the changes to level form as

qits = qits −1 exp( Dqits )

(6.4)

18

Table 9 Actual and simulated consumption of alcoholic beverages, Australia, 1955-2005

Year (1) 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 Mean

Actual consumption Beer Wine Spirits (2) (3) (4) 105.98 5.34 3.41 100.45 5.21 1.80 100.66 5.17 1.82 101.14 5.21 1.91 102.91 5.25 2.08 102.61 5.10 2.07 102.19 5.12 2.08 103.46 5.28 2.03 106.99 5.53 2.19 110.34 5.57 2.35 110.39 6.08 2.09 113.13 6.82 2.07 116.93 7.57 2.33 120.53 8.25 2.30 123.72 8.95 2.55 125.84 8.68 2.58 125.67 8.85 2.73 129.51 9.79 3.08 138.94 10.98 3.10 140.42 12.28 2.97 137.39 13.01 2.87 136.70 13.54 3.14 134.80 14.16 3.29 130.81 16.36 2.67 132.29 17.28 2.54 129.34 18.19 2.76 128.63 19.07 2.90 121.68 19.72 2.93 117.76 20.38 2.80 114.46 21.25 3.00 115.47 21.60 3.05 111.00 20.90 2.95 113.00 20.80 3.10 115.40 19.30 3.23 113.90 18.50 3.20 110.60 17.80 2.95 104.00 18.70 2.80 99.50 18.30 2.93 98.00 18.60 3.43 96.80 18.40 3.20 95.30 18.30 3.18 95.50 19.00 3.05 94.68 19.74 3.57 93.60 19.94 3.50 92.61 20.62 3.27 92.89 20.82 3.60 90.40 20.73 3.73 91.64 21.43 4.02 88.22 22.05 4.13 86.62 22.60 4.26 86.40 22.57 4.33 110.90

14.59

2.89

Simulated consumption with zero total alcohol consumption growth Beer Wine Spirits (5) (6) (7) 105.98 5.34 3.41 105.06 5.46 3.43 105.10 5.41 3.45 104.75 5.41 3.55 104.54 5.34 3.69 104.72 5.21 3.71 104.46 5.24 3.74 104.99 5.36 3.59 105.15 5.43 3.58 105.62 5.32 3.60 106.15 5.83 3.23 106.11 6.37 3.01 105.15 6.77 3.06 105.57 7.17 2.83 104.58 7.49 2.87 105.73 7.22 2.86 104.80 7.31 2.97 103.50 7.73 3.02 105.15 8.18 2.66 104.44 8.98 2.44 102.78 9.57 2.39 100.96 9.83 2.54 99.04 10.22 2.62 97.11 11.94 2.18 97.15 12.47 2.02 94.37 13.04 2.16 92.80 13.51 2.21 89.50 14.26 2.34 87.80 14.95 2.31 85.15 15.55 2.47 85.11 15.65 2.45 84.21 15.61 2.54 84.62 15.32 2.59 86.37 14.21 2.70 86.54 13.84 2.77 86.68 13.76 2.76 83.78 14.88 2.80 81.68 14.86 3.07 78.98 14.81 3.43 79.41 14.93 3.35 78.88 14.99 3.40 78.87 15.52 3.24 76.14 15.68 3.56 75.73 15.94 3.54 75.51 16.62 3.37 74.43 16.47 3.55 72.93 16.52 3.75 72.14 16.64 3.80 69.83 17.23 3.96 68.37 17.60 4.05 68.09 17.55 4.10 91.81

11.50

3.07

Simulated consumption with no constant terms in demand equations Beer Wine Spirits (8) (9) (10) 105.98 5.34 3.41 101.00 5.03 3.05 101.76 4.82 3.07 102.80 4.69 3.20 105.17 4.57 3.46 105.45 4.30 3.43 105.60 4.19 3.43 107.52 4.19 3.33 111.81 4.26 3.57 115.97 4.16 3.81 116.69 4.41 3.38 120.27 4.82 3.32 125.02 5.22 3.72 129.63 5.57 3.65 133.86 5.92 4.03 136.98 5.63 4.05 137.63 5.63 4.26 142.70 6.11 4.78 154.03 6.73 4.78 156.65 7.38 4.56 154.24 7.69 4.38 154.44 7.87 4.77 153.27 8.10 4.97 149.71 9.22 4.01 152.40 9.59 3.79 149.99 9.95 4.10 150.17 10.29 4.28 143.02 10.49 4.30 139.36 10.70 4.09 136.40 11.01 4.35 138.57 11.04 4.40 134.14 10.54 4.24 137.50 10.35 4.43 141.40 9.47 4.59 140.52 8.94 4.53 137.38 8.47 4.16 130.07 8.76 3.93 125.31 8.44 4.09 124.31 8.45 4.77 123.69 8.24 4.44 122.66 8.08 4.39 123.82 8.27 4.19 123.67 8.47 4.89 123.20 8.44 4.78 122.83 8.61 4.45 124.14 8.58 4.88 121.75 8.43 5.03 124.40 8.60 5.41 120.71 8.73 5.54 119.48 8.83 5.69 120.14 8.69 5.77 129.46

7.58

4.25

19

As before, we first evaluate the values of Dqits from equation (6.3) using estimates from Table 6 and 7 and then input those values into equation (6.4) to get the qits values. We set

qis1955 = qi1955 , i=1,2,3. Columns 8-10 of Table 9 give the simulated consumption of beer, wine and spirits assuming that there is no shift in consumer preferences. As can be seen, the simulated per capita consumption for the year 2005 for beer is 120.14 litres, wine is 8.69 litres and spirits is 5.77 litres. Accordingly, if there is no shift in the consumer preferences, it would have caused the beer consumption to be about (86.40-120.14)/86.40 = 39.1 percent higher, wine to be about (22.57-8.69)/22.57 = 61.5 percent lower and spirits to be (4.33-5.77)/4.33 = 33.1 percent higher, than otherwise. 7.

Concluding comments

In this paper, we first presented the trend in consumption pattern of alcohol during 1960-2007 in Australia and compared it with that of Canada, the United Kingdom and the United States. We then estimated the demand equations for beer, wine and spirits using Australian data for years 1955-2005 to investigate the reasons for the positive growth in wine and spirits consumption and the fall in beer consumption. The estimation results of the demand equations reveal that the income elasticities for beer, wine and spirits are 0.78, 0.82 and 1.92, respectively, implying that beer and wine are necessities and spirits is a luxury. The own-price elasticities of beer, wine and spirits are -0.2, -0.5 and -0.6, respectively, implying that the demand for these three beverages are price inelastic. Further investigation of the growth in consumption of beer, wine and spirits reveal that the strong shift in preferences toward wine and spirits, and against beer are the reasons for the positive growth in wine and spirits, and negative growth in beer consumption. Income also plays a significant positive role in the growth in consumption of all three beverages but at an almost equal rate. For all three beverages, the positive growth due to cross-prices has been almost offset by their negative own-price contribution. Simulation of alcohol consumptions reveals that (a) If total alcohol consumption remained at the 1955 level, then it would have caused beer consumption to be 21% lower, wine consumption to be 22% lower and spirits to be 5% lower, than it would be otherwise. (b) If there is no shift in consumer preference, beer consumption would have been 39% higher, wine 62% lower and spirits 33% higher than the current consumption.

20

References ABARE (2009). Australian Commodity Statistics 2009, Australian Bureau of Agricultural and Resources Economics, Canberra. Barten, A.P. (1969). ‘Maximum Likelihood Estimation of a Complete System of Demand Equations,’ European Economic Review 1: 7-73. Barten, A.P., L. Bettendorf, E. Meyermans and P. Zonderman (1989). Users’ Guide to DEMMOD-3. Kathlieke Universiteit Leuven, Belgium. Chen, D.L. (2001). World Consumption Economics. New Jersey, London, Singapore: World Scientific. Clements, K.W. (1982). ‘Divisia Moments in Australian Consumption,’ Economics Letters 9: 43-48. Meisner, J.F. (1979). ‘The Sad Fate of the Asymptotic Slutsky Symmetry Test for Large Systems,’ Economics Letters 2: 231-233. Theil, H. and F.E. Suhm (1981). International Consumption Comparisons: A System- Wide Approach. Amsterdam: North-Holland Publishing Company. Selvanathan, S. (1993). A System-Wide Analysis of International Consumption Patterns. Advanced Studies in Theoretical and Applied Econometrics, Boston: Kluwer Academic Publishers. Selvanathan, S. and E.A. Selvanathan (1994). Regional Consumption Patterns: A System-Wide Approach. London: Avebury Publishers. Selvanathan, E.A., and S. Selvanathan (2003). International Consumption Comparisons: OECD vs LDC. New Jersey, London, Singapore: World Scientific. Selvanathan, E.A., and S. Selvanathan (2005). The Demand for Alcohol, Tobacco and Marijuana: International Evidence. London: Ashgate Publishers. Selvanathan, E.A., and S. Selvanathan (2007). ‘Another Look at the Identical Tastes Hypothesis on the Analysis of Cross-Country Alcohol Data,’ Empirical Economics, 32(1) (2007): 185215. Theil, H. and K.W. Clements (1987). Applied Demand Analysis: Results from System-Wide Approaches. Cambridge, Mass.: Ballinger Publishing Company.

21