Skip past navigation to main part of page
 
Land and Environment : Agribusiness Assoc. of Australia
---

Agribusiness Review - Vol. 7 - 1999

Paper 10
ISSN 1442-6951


An Econometric Analysis of the Demand for Eggs in Australia*

Edward Oczkowski - School of Management,
Charles Sturt University –Riverina, PO Box 588, Wagga Wagga, NSW, 2678, Australia.
Phone: +61 2 69332521 Fax: +61 2 69332125 E-mail: eoczkowski@csu.edu.au

and Tom Murphy - Director, Western Research Institute,
c/o Charles Sturt University, Panorama Ave, Bathurst, NSW, 2795, Australia.
Phone: + 61 2 63384435 Fax: +61 2 63384699 Email: tom.murphy@wri.org.au

* Funding for this study was provided by the Rural Industries Research and Development Corporation (Australia). The assistance of Hugh McMaster is gratefully acknowledged for the collation egg industry data. Comments from a referee and the editors are also gratefully acknowledged.

Abstract

This paper provides the first comprehensive econometric analysis of Australian State egg demand behaviour. Explicit diagnostic testing of models is employed to help gain robust demand elasticities. Demand is found to be own price and income inelastic, with price elasticity being effectively zero for the majority of states. Prices of related products tend to have only a minor overall influence. The proportion of paid working females is statistically important for the majority of states. A worldwide cholesterol information index appears to capture the health concerns held about egg consumption. Interestingly, results for advertising expenditure are mixed with both significant and insignificant effects identified.

Keywords: egg demand, econometric modeling, demand elasticities, cholesterol health concerns, advertising expenditure.

Introduction

There is substantial worldwide interest in identifying the factors, which determine the demand for eggs. Recent studies have analysed egg demand for California (Schmit, Reberte, and Kaiser, 1997) , Canada (Chyc and Goddard, 1994) , and the United States (Brown and Schrader, 1990). The objective of this paper is to add to this international literature by examining the demand for eggs in Australia using econometric methods. The most recent study of Australian-wide egg demand was performed by Hickman (1979) , his study was narrowly egg price and income focused and is now somewhat dated. Our paper updates these elasticity estimates, but more importantly extends the examination of demand determinants to include: prices of related foods, advertising expenditure, population structure and dietary concerns.

The Australian egg industry is well established, currently there are over 900 producers selling a gross production value of AUD$280 million p.a. Historically, up until 1989 commercial egg production and distribution was controlled by independent state marketing boards. Over the 1989/90 period commercial egg markets were deregulated in four of the five mainland states. Regulations remain in Western Australia. For an illuminating description of the Australian egg industry, see Australian Egg Marketing Council (1989) . The focus in our paper rests with the state demand for domestic shell egg sales in each of the five mainland states. This is necessary given the independence of state marketing, distribution, policy and legislative arrangements. In total, our focus on Australian mainland shell egg sales accounts for approximately over 80 per cent of total Australian egg production. The bulk of the remaining 20 per cent of egg production is pulped and exported.

This study focuses purely upon the demand for eggs and therefore possesses all the limitations of demand studies, which ignore issues pertaining to the interaction between demand and supply. Clearly, an egg industry model examining demand, supply and trade determinants would be required for a full comprehensive analysis. In any event, attempts have been made to mitigate any adverse effects of supply interaction on our demand estimates. First, structural shifts in demand estimates which may have occurred because of supply deregulation were examined and found to be statistically unimportant. Second, the employed dependent variable is shell egg sales rather than shell egg production. Finally, retail egg prices rather than farm or wholesale egg prices, are employed throughout the analysis.

In the next section we review both the Australian and international egg demand literature. Then the general model specification, data and regression modelling strategy are described. In the section thereafter, econometric results for the egg demand functions are outlined, followed by a discussion of their economic implications. The final section provides some concluding comments and a summary of the major findings. For a fuller discussion of all these issues see Oczkowski and Murphy (1998) .

Literature Review

In reviewing the egg demand literature six classifications for the variables seeking to explain demand for eggs are employed: (i) price of eggs; (ii) household income; (iii) prices of substitutes and complements for eggs; (iv) demographic variables; (v) advertising expenditure; and (vi) cholesterol and changing consumer taste variables. We report on only four available Australian econometric egg demand studies, Banks and Mauldon (1966) , Gruen et. al. (1967) , Hickman (1979) and Collard, et.al . (1982) . All these studies are now dated, employing pre-1980s data. Reported Australian egg price elasticities of demand range from -0.009 to -0.398 reflecting a low own price elasticity of demand for eggs in Australia. The estimates do point to some differences in elasticities between the Australian states. From the international egg demand literature we report on studies by: Chavas and Johnson (1981) ; Burney and Akmal (1991) ; Wu, Li and Samuel (1995) ; Brown and Schrader (1990) ; Chyc and Goddard (1994) ; and Schmit, Reberte and Kaiser (1997) . For non-Australian countries egg price elasticities appear to be generally more elastic than Australian estimates, ranging from -0.15 to -1.7.

Reported Australian income elasticities of demand range from 0.00 to 1.43. In general, low but positive income elasticities appear to be most common for the Australian states. The main features of non-Australian estimates are that all income elasticities are positive with the majority of overseas estimates being less than 0.6, indicating a low income elasticity, which is broadly consistent with the Australian estimates. In previous research only sausages have been found to be an important substitute for egg demand in Australia. While in non-Australian research red meat prices have been shown to have important substitute effects on U.S. egg demand.

Our literature review did not find any Australian studies assessing issues relating to demographic, advertising or consumer taste variables. As a consequence only results from non-Australian studies are cited. The review of literature identifies five factors in international studies which may be important demographic determinants of egg consumption: (i) women in the paid labour force; (ii) age (older people may consume fewer eggs); (iii) income (high income families may consume less); (iv) education (more highly educated families may consume less); and (v) location (more urbanisation less egg consumption). In assessing the impact of advertising expenditure on egg demand Hallam (1986) alludes to U.K. evidence which found no relation between advertising expenditure and egg consumption during the 1970s. However, Chyc and Goddard (1994) for Canada and Schmit, Reberte and Kaiser (1997) for California find some statistically important advertising elasticities for egg demand. In examining changing consumer tastes, Brown and Schrader (1990) found for the U.S. that both a declining time trend and a cholesterol information index to measure health concerns, to be important.

Model Specification, Data and Modelling Strategy

For numerous reasons we follow the regression modelling approach attributed to Hendry (1995) , but with modifications. Modelling from the general-to-specific is pursued. That is, all the independent variables hypothesised to be theoretically important are initially included in the model and then the model is reduced using tests of significance, diagnostic checks, model selection statistics and checks with theory consistency. This process leads to a more parsimonious model to be chosen as the preferred specification. The preferred model is then subject to a battery of diagnostic tests to ensure that the preferred equation adequately conforms to the assumptions, which underpin regression analysis. As a further check of the preferred model we attempted to re-include the previously omitted variables to investigate whether these omitted variables significantly alter any conclusions reached. Preliminarily modelling indicated a superior performance of linear over logarithm functional forms and hence linear models are presented throughout.

In developing the preferred regression estimates diagnostic tests were conducted to check for excessive degrees of non-normality, heteroscedasticity, specification error and autocorrelation. The specific diagnostic regression test statistics employed are described in Beggs (1988) . The following tests were employed: Jarque-Bera test for non-normality; Breusch, Pagan and Godfrey test for heteroscedasticity; Ramsey's RESET test for specification error; Durbin-Watson test for 1st order auto-correlation and the residual correlogram tests for autocorrelation of orders from 1 thru 6.

Rather than let the entire modelling process be influenced by the nature of the data alone we impose some theoretical expectations explicitly on the modelling process. Irrespective of their statistical significance in the regression equations, the egg price, household income and cholesterol information index variables will always be retained in all specifications. Standard consumer demand theory presents an overwhelming case for retaining the former two variables and the literature on egg consumption patterns worldwide establishes a compelling case for always retaining the cholesterol index. It appears that no other variables can mount similarly strong cases for unconditional inclusion in the preferred specification. With regard to the other hypothesised demand factors, these are only included if they contribute importantly to the explanation of variations in egg demand according to theoretical expectations. In other words, statistically significant results which are counter to theoretical expectations are ignored and discounted as due to data measurement, specification error or spurious correlation causes. In general our modelling approach therefore consists of both theoretical and empirically driven considerations.

The emphasis on diagnostic checking in the Hendry approach is particularly appealing. By confronting the preferred specification to a wide variety of checks we are trying to ensure that the results obtained and their economic implications are reliable. Effectively only robust models will pass these series of checks and hence may be of some use to practitioners.

To facilitate this process the most general model specification was initially identified by collecting data on all potentially important demand determinants. In particular, based on the literature review, discussion with industry representatives and the availability of suitable data, the demand determinants to be employed fall into six major groups: (i) price of eggs, (ii) prices of related products, (iii) household income, (iv) demographic variables, (v) advertising expenditure, and (vi) changing taste variables.

The dependent variable employed is commercial shell egg sales per head of population, measured in dozens. The use of shell egg sales per capita helps avoid econometric problems such as excessive multicollineaity (as both population and income would become independent variables) and heteroscedasticity (as the error variance typically gets larger for higher absolute consumption levels). It should be noted that since the late 1980s after deregulation, direct shell egg sales figures are not available. As a consequence, a disposal series based on production, import and export data was employed for most states during the 1990s.

Two real retail price of eggs measures are employed: (a) using the standard weight as defined by the various Australian Bureau of Statistics (ABS) surveys (labelled ABS); and (b) standardising all prices to a 55gm weight for all states and years (labelled 55gm). Even though the latter measure may technically be more appropriate as it standardises across states and time, the former measure may be useful as it may better reflect what was actually purchased at a given point in time in a given state. Both prices are CPI deflated (1989/90 prices) and final choice will be made on the basis of model selection statistics. For related products six CPI deflated retail prices are employed. For substitutes we will consider breakfast cereal, potatoes and sausages. For complements we will consider bacon, flour and sugar. Real per capita household disposable income is employed as the income variable.

Data availability permitted the examination of three demographic variables. The percentage of employed females to the female working population will be employed to capture the changing role of women in society over time and how that may affect egg consumption. Two opposing arguments have led to its consideration in demand for egg equations, see Chyc and Goddard (1994) and Schmit, Reberte and Kaiser (1997) . First, as more women go into the labour force they have less time for home egg cooking and baking and hence demand falls. Second, as more women go into the labour force more take-out breakfasts are consumed thereby increasing overall egg consumption. The median age of the population is employed to capture any ageing effects on consumption. To measure any urbanisation effects we employed the percentage of people in the capital city of a state to the state population.

Even though some state advertising was performed by all states over our sample period, consistent state egg advertising expenditure data is not available for all states. Only long-time series advertising data exists for Victoria (Vic) and Western Australia (WA). To this extent a thorough analysis of the impact of advertising on egg sales cannot be performed for New South Wales (NSW), Queensland (Qld) and South Australia (SA). Beyond state egg advertising expenditure two other related variables are employed in the analysis. During a three-year period (1986/7-88/9) the Australian Egg Marketing Council undertook some national egg advertising and promotion, this national advertising will be employed for all states in the regression analysis. Further, as with related product prices the impact of substitute products through advertising should also be examined. To this extent domestic advertising expenditure on red meat products committed by the Australian Meat and Livestock Corporation (AMLC) will also be employed in the data analysis. All advertising expenditure variables will be converted to a per capita basis (using the relevant population) and deflated to real terms using the relevant CPI. Even though the use of real per capita advertising variables is consistent with previous egg studies (Chyc and Goddard 1994 , and Schmit, Reberte and Kaiser 1997) , a counter argument exists which suggests that advertising can be viewed as a public good and hence specified in non-per capita terms. This counter-view should be considered when interpreting our per capita demand advertising estimates.

The previously employed approaches to incorporating consumer taste changes will be investigated. First, various time trends and structural breaks in these trends will be examined. Secondly, following Brown and Schrader (1990) we have employed the Medline medical database to construct a cholesterol information index. Two indexes are considered, a worldwide index using all relevant articles from the database and an Australian index using articles only with an Australian place of publication. These indexes are proxies for the general public's awareness of health issues pertaining to egg consumption. A priori the worldwide index is preferred given the small number and circulation of Australian medical publications and that major research breakthroughs are typically reported in non-Australian based publications. It is clear that Australian doctors, health professionals and dietary publication editors have access to literature from all over the world. For definitions of all employed variables see the data appendix. For a more comprehensive description of all the data and their construction see Oczkowski and Murphy (1998) .

Before examining the econometric results it is useful to consider some key features of the nature of the data. Annual data is employed examining the period 1962/3 - 1995/6, except for Victoria which covers 1962/3-1992/3 due to advertising expenditure data availability. All state shell eggs sales per capita variables tend to follow a flat inverted U shape, see Figure 1 (Note: not all states are presented because of disclosure clauses). In general sales increased up until the mid-1970s and have flattened and/or generally fallen ever since. Real retail egg prices have fallen significantly over time with falls of up to 50 per cent over the sample period, see Figure 2 . For substitutes, prices have generally risen in real terms, this compares to the substantial egg price falls. For complements prices have tended to fall like egg prices. In most cases these falls have not matched egg price falls. In all the remaining variables: income (1.1 per cent average growth), female workforce (1.1 per cent average growth), median age (0.7 per cent average growth), capital city population (0.2 per cent average growth), advertising, and cholesterol indexes (20.3 per cent average growth); general rising time trends of different degrees exist.

Figure 1 – Domestic Shell Egg sales per capita

Figure 1

Figure 2 – Real Egg Prices

Figure 2

To discuss issues of excessive multicollinearity, which may have an impact on the regression results, we make some comment on the pattern of high correlations between the demand determinants. First, the worldwide and Australian cholesterol indexes are nearly perfectly correlated (r = 0.995), so only one index is needed for the analysis and this will be determined empirically. Second, for all states high correlations appear to group the following variables: age, cholesterol index, female workforce, cereal prices, egg prices and AMLC advertising. The highest correlations occur for: age and cholesterol (over 0.95), age and AMLC advertising (over 0.95), cholesterol and AMLC advertising (0.92), female workforce and egg prices (over –0.87). Of all these variables the greatest concern is with is the age variable, we suspect that the regression analysis may not be able to accurately distinguish the individual impact of age on egg sales.

A final modelling issue relates to employing a panel/pooled data approach for the five states or performing separate state-by-state analyses. The latter approach is preferred for the following reasons. First, the previously cited Australian literature points to substantial elasticity differences between states. Relatedly, survey evidence on egg consumption patterns points to state differences. Second, data on a potentially important variable 'egg advertising expenditure' was not consistently available for all states and hence by definition could not be employed in any pooled/panel approach. Finally, our empirical results clearly point to substantial elasticity differences and included/omitted variable differences with regard to advertising, age, female workforce and prices of related products.

Econometric Results

Three general comments, which apply to all states, can be made about the preferred demand estimates. First, the preliminary modelling clearly points to the better performance of the worldwide cholesterol index over the Australian cholesterol index. As a consequence the worldwide index is used and retained throughout for all states. Second, for the changing consumer taste variables, time trends and structural changes in these time trends proved to be theoretically and empirically unsuccessful. As a consequence these time trend variables are omitted from all equations. Third, structural change however, is reflected via a change in the cholesterol impact on egg sales. Three structural breaks were considered, at 1980/1, 1985/6 and 1990/1. Only the latter structural break in cholesterol's impact on egg demand proved to be important.

In Table 1 is presented the estimates of demand for eggs from regressions for New South Wales. When interpreting these results recall that data on state advertising expenditure is not available for NSW. In terms of the diagnostic tests no statistics are significant at the 5 per cent level and thus the model appears to be relatively free of non-normality, heteroscedasticity, specification error and auto-correlation problems. The model with 55gm egg prices outperformed the model with ABS egg prices in terms of model selection statistics described by Ramanathan (1998) . Even though the age variable may appear to be mildly insignificant (p value = 0.163), it is retained because of its better performance in model selection statistics compared to a model with its exclusion.

TABLE: 1 New South Wales: Regression Estimates for Shell Egg Sales Per Capita

Variable Coefficient* T- Ratio P-Value Mean
Constant 10.332 1.53 0.138
Egg Price (55gm) 0.0018 0.80 0.430 276.76
Income 0.257* 3.73 0.001 12.547
Female Workforce 0.124* 2.28 0.030 40.851
Age -0.239 -1.44 0.163 30.62
Cholesterol Index -0.00089* -2.47 0.020 1035.0
Cholesterol*(90/1-95/6) 0.00023 1.81 0.081 619.38

* Denotes significant at a 5 per cent level of significance. Results based on 1962/3–95/6, N= 34. R2 = 0.931

Demand estimates suggest for NSW that egg prices are unimportant. Income and the female workforce have strong positive influences. The result for the female workforce variable reflects the increased demand for take-out-foods which use eggs. The ageing of the population appears to have some minor negative impact. The cholesterol index points to substantial health concerns in reducing egg demand. The structural change in this variable indicates however, that from 1990/1 the marginal impact of the cholesterol index has fallen significantly. Interestingly, none of the prices of related products proved to be important for NSW egg demand.

In Table 2 is presented the estimates of demand for eggs from regressions for Victoria. Recall for Victoria, state advertising expenditure is available but for a slightly shorter time period, 1962/3-1992/3. In terms of the diagnostic tests no statistics are significant at the 5 per cent level and thus the model appears to be relatively free of non-normality, heteroscedasticity, specification error and auto-correlation problems.

TABLE: 2 - Victoria: Regression Estimates for Shell Egg Sales Per Capita

Variable Coefficient* T- Ratio P-Value Mean
Constant 2.0488 0.83 0.414
Egg Price (ABS) 0.0014 0.51 0.612 294.83
Income 0.608* 7.78 0.000 12.42
Sausages Price 0.0029 1.47 0.154 393.71
Flour Price -0.0095 -1.86 0.075 256.54
State Advertising 3.306* 6.31 0.000 0.453
Cholesterol Index -0.00101* -7.01 0.000 740.13

Denotes significant at a 5 per cent level of significance. Results based on 1962/3-92/3, N = 31. R2 = 0.933

In contrast to NSW, the model with ABS egg prices outperformed the model with 55gm egg prices in terms of model selection statistics. Even though the prices of flour (p value = 0.075) and sausages (p value = 0.154) appear to be mildly insignificant their exclusion worsened model selection statistics and diagnostic tests, for example, specification error becomes a problem. As a consequence both variables are retained. Further, the AMLC and national egg advertising expenditure variables either individually or jointly proved to be unimportant in the model.

Demand estimates for Victoria suggest that egg prices are unimportant. Income and state advertising have strong positive influences. Sausages appear to be mildly important substitutes, while flour appears to be an important complement. The cholesterol index points to substantial health concerns in reducing egg demand.

In Table 3 is presented the estimates of demand for eggs from regressions for Queensland. When interpreting these results recall that data on state advertising expenditure is not available for Qld. In terms of the diagnostic tests the lack of state advertising data appears to have profound effects. No matter what combinations of variables were employed in modelling Qld egg demand, significant specification error problems were encountered with resulting highly significant RESET test statistics. The fact that the national advertising variable proved to be important in the preferred specification implies that maybe the omission of a state advertising variable is the cause of the specification error problems for Qld. As a consequence all the results for Qld should be interpreted with some degree of caution. Even though all three RESET test statistics are significant at the 5 per cent level, the preferred model appears to be relatively free of the other problems of non-normality, heteroscedasticity and auto-correlation.

TABLE: 3 - Queensland: Regression Estimates for Shell Egg Sales Per Capita

Variable Coefficient* T- Ratio P-Value Mean
Constant 4.476 1.50 0.146
Egg Price (55gm) -0.0051 -1.93 0.064 278.86
Income 0.739* 11.25 0.000 11.093
Flour Price -0.022* -4.39 0.000 210.33
Female Workforce 0.087 1.85 0.075 39.12
National Advertising 5.965 1.93 0.064 0.047
Cholesterol Index -0.0011* -7.59 0.000 1035.0

The model with 55gm egg prices outperformed the model with ABS egg prices in terms of model selection statistics. Even though the female workforce (p value = 0.075) and national advertising (p value = 0.064) variables may appear to be mildly insignificant, they are retained because of their better performance in model selection statistics.

Unlike NSW and Victoria, for Qld demand estimates suggest that egg prices are important. Income, female workforce and national egg advertising have strong positive influences. Again the result for the female workforce variable reflects the increased demand for take-out-foods which use eggs. The price of flour indicates significant complementary with egg demand. The cholesterol index points to substantial health concerns in reducing egg demand.

In Table 4 is presented the estimates of demand for eggs from regressions for South Australia. When interpreting these results recall that data on state advertising expenditure is not available for SA. In terms of the diagnostic tests the lack of state advertising data appears not to have any serious impact on the reliability of regression results. The preferred model appears to be relatively free from the problems of non-normality, specification error, heteroscedasticity and auto-correlation. The model with ABS egg prices outperformed the model with 55gm egg prices in terms of model selection statistics.

TABLE: 4 - South Australia: Regression Estimates for Shell Egg Sales Per Capita

Variable Coefficient* T- Ratio P-Value Mean
Constant -0.508 -0.17 0.870
Egg Price (ABS) -0.0070 -1.73 0.095 287.60
Income 0.287* 2.59 0.015 11.85
Sausages Price 0.0068* 3.18 0.004 363.82
Female Workforce 0.157* 2.88 0.008 41.37
Cholesterol Index -0.0017* -6.44 0.000 1035.0
Cholesterol*(90/1-95/6) 0.00058* 3.47 0.002 619.38

Similar to only Qld, SA demand estimates suggest that egg prices are important. Income and female workforce have strong positive influences. Again the result for the female workforce variable reflects the increased demand for take-out-foods which use eggs. The price of sausages indicates significant substitutability with egg demand. The cholesterol index points to substantial health concerns in reducing egg demand. The structural change in this variable indicates however, that from 1990/1 the marginal impact of the cholesterol index has fallen significantly.

In Table 5 is presented the estimates of demand for eggs from regressions for Western Australia. Recall, state advertising expenditure is available for WA for our entire period of analysis 1962/3 to 1995/6. In terms of the diagnostic tests no statistics are significant at the 5 per cent level and thus the model appears to be relatively free of non-normality, heteroscedasticity, specification error and auto-correlation problems.

TABLE: 5 - Western Australia: Regression Estimates for Shell Egg Sales Per Capita

Variable Coefficient* T- Ratio P-Value Mean
Constant -0.612 -0.32 0.754
Egg Price (ABS) 0.0021 0.75 0.460 284.00
Income 0.265* 4.24 0.000 12.34
Sugar Price -0.0067* -2.12 0.044 204.91
Female Workforce 0.222* 5.64 0.000 42.15
Cholesterol Index -0.0010* -4.69 0.000 1035.0
Cholesterol*(90/1-95/6) 0.0002 1.57 0.127 619.38

 The model with ABS egg prices outperformed the model with 55gm egg prices in terms of model selection statistics. Even though the cholesterol index structural change variable may appear to be mildly insignificant (p value = 0.127), it is retained because of its better performance in model selection statistics compared to a model with its exclusion. Interestingly, none of the three advertising variables were significant when either included individually or jointly. For example, the state advertising variable when re-entered into the preferred model had a t ratio of –0.83, which if anything implies a very weak negative influence on egg sales.

Demand estimates suggest that egg prices are unimportant. Income and female workforce have strong positive influences. Again the result for the female workforce variable reflects the increased demand for take-out-foods which use eggs. The price of sugar indicates significant complementary with egg demand. The cholesterol index points to substantial health concerns in reducing egg demand. The structural change in this variable indicates however, that from 1990/1 the marginal impact of the cholesterol index has fallen somewhat.

Economic Implications

Table 6 summarises the main results for all states in terms of a priori and statistically important variables and their implied estimated elasticities. Given the demand focus of the study these elasticities are interpreted in the conventional ceteris paribus sense. At the outset it is important to reiterate that these results may be greatly influenced by the availability of state advertising expenditure data for only Victoria and WA. Since we found that state advertising may (Victoria) or may not (WA) have an important influence on egg demand, the omission of this data for the other states may or may not seriously affect results. Even though the diagnostic test statistics suggest that results for NSW and SA appear not to be seriously affected by this data omission, Qld results indicate that specification error had probably resulted from this omission.

TABLE: 6 - A Priori and Statistically Important Elasticities of Shell Egg Sales

Variable NSW VIC QLD SA WA
Price Elasticity 0.046 0.042 -0.159 -0.236 0.057
Income Elasticity 0.292 0.799 0.912 0.400 0.319
Sausages Price 0.120 0.292
Flour Price -0.258 -0.506
Sugar Price -0.134
Female Workforce 0.460 0.379 0.764 0.910
Age -0.663
Cholesterol Index -0.076 -0.079 -0.125 -0.171 -0.092
Advertising 0.159 (state) 0.029 (national)

Elasticities based on sample averages of all data except national advertising, which is based on three years of non-zero data only.

Even though our regression modelling strategy was always to include the household income variable for theoretical reasons, its inclusion in all states was also guaranteed for statistical reasons. The coefficient for income was strongly statistically significant and positive for all states. At the sample means of all data, all estimates are inelastic. The state estimates can be broadly grouped into low elasticity (NSW 0.29, SA 0.40 and WA 0.32) and high elasticity (Victoria 0.80 and Qld 0.91). These estimates differ somewhat from results from previous studies (both Australian and international) where the expectation was for low elasticities no greater than 0.6.

In terms of the price of related products, of the six prices considered only three prices proved to be important and not consistently for all states. Only one substitute proved to be important, the price of sausages was important for Victoria and SA. The result for Victoria is consistent with the Collard et. al. (1982) study. Two complements were identified to be important. The price of flour was important for both Victoria and Qld, while the price of sugar proved to be important for WA only. For NSW, prices of related products were always insignificant. Victoria exhibited the greatest relationship with related products with two important variables. In general, these results imply that there appears to be only a minor degree of substitutability and complementary between egg and related products.

The female workforce variable was strongly significant and positive for all states except Victoria. The positive findings suggest that as more women go into paid employment more take-out foods are consumed which use eggs and this outweighs any negative effect due to women having less time for home egg cooking and baking. The positive impact is consistent with some findings from the U.S., (Schmit, Reberte, and Kaiser, 1997) . The insignificant result for Victoria is interesting, and we point out that Victoria has the slowest growth rate amongst all states for the female workforce percentage.

The median age variable proved to be important for NSW only. It is not obviously clear why this variable should be important only for NSW. It could reflect the multicollinearity problems with age, as age was found to be highly correlated with other variables, especially the important cholesterol index.

One of the most important identified demand determinants is the impact of health concerns on egg demand. These health concerns have been proxied by a cholesterol information index. The estimated impact of these concerns on egg sales has been quantified for three time periods: (i) the entire sample, (ii) since 1975/6 when the publication index started to grow significantly, and (iii) over the last ten years. The results are presented in Table 7 .

TABLE: 7 - Cholesterol Impact on Shell Egg Sales Per Capita (%)

Variable 1962/3-95/6 1975/6-95/6 1986/7-95/6
New South Wales -7.6% -11.5% -20.0%
Victoria (upto 92/3 only) -7.9% -12.7% -24.5%
Queensland -12.5% -19.3% -34.9%
South Australia -17.1% -27.4% -45.2%
Western Australia -9.2% -14.4% -24.4%

Average cholesterol impact on shell egg sales per capita as a percentage of shell egg sales.

There are substantial differences between the states with three meaningful state groupings: (i) SA clearly has the greatest impact, (ii) Qld falls in the middle, and (iii) NSW, Victoria and WA have similar and the smallest impacts.

The differences are substantial, for example health concerns in SA have more than double the impact on egg demand than they have in NSW. In absolute terms the percentage impacts appear to be large. However, they are also reasonably consistent with results from the U.S. (Brown and Schrader, 1990) which estimate -16 per cent for 1955-1987 and -25 per cent for 1966-1987.

Our estimated health concerns impacts increase substantially when we consider the more recent time periods only. For example, for 1962/3-95/6 the NSW impact is –7.6 per cent of sales, but for the last ten years 1976/7-95/6 it increases to –20 per cent of sales. Even though the estimated marginal impact of the cholesterol index falls for three states from 1990/91, the ever rising cholesterol information index still ensures that the percentage impact on sales increases further in recent years.

Finally, consider the advertising expenditure impact as it relates to national egg advertising expenditure in Qld and state advertising expenditure in Victoria. The average annual impact of national advertising on Qld egg sales was 2.9 per cent of sales (95 per cent confidence interval: -0.2 per cent to 6.0 per cent of sales). In terms of total gross retail revenue (retail price * sales) this implies an average annual revenue of 62.4 cents per capita in 1989/90 dollars (95 per cent interval: -$0.04 to $1.29), and compares to the average advertising expenditure of 4.6 cents per capita. More generally, the national advertising Qld results imply that every extra cent of expenditure per capita, generates 13 cents of total gross retail revenue per capita (95 per cent interval: -0.93 cents to 27 cents).

We now turn to the state advertising impact on egg sales for Victoria. The average annual impact of state advertising on Victoria egg sales was 15.9 per cent of sales (95 per cent interval: 4.4 per cent to 27.4 per cent of sales). In terms of total gross retail revenue this implies an average annual revenue of $4.42 per capita in 1989/90 dollars (95 per cent interval: $2.47 to $7.61), and compares to the average advertising expenditure of 45 cents per capita. More generally, the state advertising Victoria results imply that every extra cent of expenditure per capita generates 9.7 cents of total gross retail revenue per capita (95 per cent interval: 6.6 cents to 12.9 cents).

Interestingly, the marginal impact of advertising expenditure is larger with national advertising for Qld (0.06 dozen for each cent) compared to state advertising for Victoria (0.03 dozen for each cent). However, because the average annual expenditure was far greater for Victoria (45 cents) than for Qld (5 cents), then the actual Victoria advertising impact (15.9 per cent ) far exceeds that for Qld (2.9 per cent ).

Conclusion and Summary

This study is the first study of Australian egg demand that has systematically considered a variety of demand determinants in addition to standard price and income effects. The explicit modelling of demographic, advertising and health concerns provides new insights into egg consumption patterns in the five mainland states. Moreover, the employed regression modelling strategy has attempted to ensure that results are robust to violations in regression assumptions and the vagaries of subjective model building. Importantly, when only price and income variables are used in demand specifications for our up-dated data set, as per the Hickman (1979) study, substantial RESET specification error and auto-correlation problems exist for all models and significant positive price elasticities are estimated for some states.

The most important empirical findings of the study are now summarised. Egg prices represent an important factor for only Qld and SA, and even for these states the price elasticities are highly inelastic, not exceeding 0.24 in absolute terms. The strategic implication of this finding, is that future price changes of the magnitude experienced over the last thirty years will not significantly alter egg demand. Household disposable income is an important egg demand determinant for all states, but particularly for Victoria and Qld where income elasticities are 0.8 or higher. Consistent with previous studies, the prices of related products tend to have only a minor overall and spasmodic influence on egg demand. Of the demographic factors two variables were found to be important, the female workforce for all states except Victoria and the age variable only for NSW. The worldwide cholesterol index appears to capture the health concerns the population has about cholesterol and egg consumption. This variable proved to be very important for all states.

State advertising was analysed only for Victoria and WA. It was important for egg sales in Victoria only. Reasons why the advertising variable in the WA regressions proved to be unimportant are unclear. In any event taken together our results do not suggest that universally, advertising expenditure will always have a positive effect on egg sales. As argued by the U.K. study of Hallam (1986) it may be the case that how an advertising campaign is pursued may significantly influence its effectiveness on sales. On the other hand, our results do also imply that advertising can have a significant influence on sales and this is particularly important given the health concerns the population appears to have about egg consumption in general.

Interestingly, some of the potentially important demand determinants proved to be unimportant in the final analysis. The egg substitutes of cereal and potatoes were unimportant, as was the egg complement bacon. The age variable was only important for NSW, this though probably reflects the fact that it was ‘over-powered' by the highly correlated cholesterol index. Essentially, for most states the cholesterol index appears to be also capturing the fact that the ageing of the population is leading to fewer egg sales. The capital city population variable was clearly unimportant, there is probably too little time series variation in this variable to have an important influence. A cross-section study maybe more useful for capturing this urbanisation effect. Advertising expenditure for the egg substitute of red meat also proved to be statistically insignificant for all states. This result may however be influenced by the lack of contrasting state egg advertising data in three states. Finally, the Australian cholesterol index was clearly outperformed and swamped by the worldwide cholesterol index.

The main limitations of the study relate to the quality of the shell eggs sales data employed for the 1990s and the lack of comprehensive state egg advertising expenditure data for all states. Only data from WA was directly available for the entire data analysis time period, given that it still operates under a regulated environment. However, for the remaining states and recent years (from the late 1980s) a disposal series based on ABS agricultural census data was employed. Various egg industry representatives have questioned the reliability of this data. Advertising expenditure data was only available for Victoria and WA. For the remaining states this implies that results may be unreliable given this data omission and the positive advertising results for Victoria. However, our RESET diagnostic test results indicate that estimates for NSW and SA are reasonably valid. The results for Qld may be less valid given these RESET results and the fact that national advertising expenditure was found to be important for Qld.

Finally, concern may be raised about the small insignificant positive own price elasticities for NSW, VICTORIA and WA. to avoid this problem an alternative Bayesian estimation framework, which can impose negativity constraints, could be employed. However, such a procedure is at variance with all previous literature on egg demand (which employs OLS), and it would also invalidate the use of standard classical diagnostic testing procedures. Moreover, such restrictions would be expected to have little if any effect on results given that the offending own price coefficients are numerically very small and highly insignificant.

References

Australian Egg Marketing Council (1989) ‘Egg Industry.' pp163-167 in J. Cribb (ed) Australian Agriculture, National Farmers Federation: Camberwell.

Banks E.L. and R.G. Mauldon (1966) ‘Effects of Price Decisions of a Statutory Marketing Board.’ Australian Journal of Agricultural Economics,10, 1-13.

Beggs, J.J. (1998) ‘Diagnostic Testing in Applied Econometrics.’ Economic Record, 64, 81-101.

Brown, D.J. and L.F. Schrader (1990) ‘Cholesterol Information and Shell Egg Consumption.’ American Journal of Agricultural Economics, 72, 548-555.

Burney, N.A., and M. Akmal (1991) ‘Food Demand in Pakistan: An Application of the Extended Linear Expenditure System.’ Journal of Agricultural Economics, 42, 185-195.

Chavas, J.P., and S.R. Johnson (1981) ‘An Econometric Model of the US Egg Industry.’ Applied Economics ,13, 321-335.

Chyc, K.M. and E.W. Goddard (1994) ‘Optimal Investment in Generic Advertising and Research: The Case of the Canadian Supply-Managed Egg Market.’ Agribusiness, 10, 145-166.

Collard, A.K., Ryan, T.J., and J.M. Alston (1982) A Monthly Demand Model for Eggs: Victoria. Research Project Series no. 106, Department of Agriculture, Victoria.

Gruen, F.H. et. al. (1967) Long-Term Agricultural Supply and Demand Projections, Australia, 1965 to 1980 Monash University, Department of Economics.

Hallam, D. (1986) ‘The Eggs Authority: A Critical Appraisal.’ Journal of Agricultural Economics, 37, 185-191.

Hendry, D.F. (1995) Dynamic Econometrics. Oxford University Press: Oxford.

Hickman, R. (1979) ‘The Effects of Price and Income on Domestic Egg Consumption.’ BAE, Eggs: Situation and Outlook, pp11-22, AGPS: Canberra.

Oczkowski, E. and T. Murphy (1998) Modelling the Determinants of Domestic Egg Demand. Report for the Rural Industries Research and Development Corporation, RIRDC project no. UCS-17A, Canberra.

Ramanathan, R. (1998) Introductory Econometrics with Applications. 4th ed., Dryden: Forth Worth.

Schmit, T.M., Reberte, J.C., and H.M. Kaiser (1997) ‘An Economic Analysis of Generic Egg Advertising in California, 1985-1995.’ Agribusiness, 13, 365-373.

Wu Y., Li E. and N. Samuel (1995) ‘Food Consumption in Urban China: An Empirical Analysis.’ Applied Economics, 27, 509-515.

DATA APPENDIX

This appendix defines the variables employed in the preferred regression equations. Annual data covering 1962/3-1995/6 is used for all sates except Victoria which covers 1962/3-1992/3.

Egg Sales: Shell egg sales in Australia by commercial producers. Dozen per head of population.
Egg Price (ABS) Average retail price of the ABS standard dozen eggs. Cents per dozen, 1989/90 constant CPI deflated prices.
Egg Price (55gm) Average retail price of a 55gm (size) dozen eggs. Cents per dozen, 1989/90 constant CPI deflated prices.
Income Household disposable income per head of population, $1,000s in 1989/90 constant CPI deflated prices.
Sausages Price: Average retail price of 1kg of sausages. Cents per dozen, 1989/90 constant CPI deflated prices.
Flour Price Average retail price of 2kgs of self-raising flour. Cents per dozen, 1989/90 constant CPI deflated prices.
Sugar Price Average retail price of 2kgs of sugar. Cents per dozen, 1989/90 constant CPI deflated prices.
Female Workforce Females in paid employment as a percentage of the civilian female population over 15 years of age.
Age Median age of all persons.
State Advertising Sales promotion and marketing costs. Cents per head of population, 1989/90 constant CPI deflated prices.
National Advertising Advertising expenditure and promotion expenses committed by the Australian Egg Marketing Council for the years 1986/7-88/9. Cents per head of the Australian population, 1989/90 constant CPI deflated prices.
Cholesterol Index The cumulative number of articles in the Medline medical database which have the words cholesterol and heart disease or arteriosclerosis in either the title or abstract. Lagged six months.

 

top of pagetop of page

Contact us

Contact the University : Disclaimer & Copyright : Privacy : Accessibility