Arretral. J . Stuttkt., 14 (3). 1972, 240-255

HOUSEHOLD CONSUMPTION OF FOOD IN AUSTRALIA'

N. PODDER* The Vaivmsity of New Xouth Wales

This paper is the second (see the first reference) in the series of studies entitled " Household Consumption and Income in AnstraJia '?

undertaken by the author to analyse data from an AustraJkwide eonsumption survey (Drane, Edwards and Gates, 1969). Whercm the first study wm an investigation into the over@ mmumption patterns, the present study is devoted to analysing intensively the patterns of household food consumption. Food is especially important since it is the h g m t single consumption item in the family budget. In thk paper the analysis is extended in two main directions. The first is the disaggregation of total food consumption into its com- ponents. The degree of disegregation, however, is limited by the avdhbility of detailed information. Acoorclingly, only five com- ponents of expenditure on food could be effectively considered. The second direotion is the investigation of the poseible differences in food consumption pafterns among variou~ eociologid and demographic group8. To that end multivarhte shthticafl hhniqum have been employed.

The first section of this paper gives a brief description of the scope and limitations of the study and sets out Werent hypotheses that muld be tested. The second seotion discusses the approprhte statistical techniques and the estimation methods that would be nseful, while the results and their exposition have been presented in the third section. The final section contains a short comparative study and a few concluding remark#.

I It is a common phenomenon, even in an apBuent economy, that

the level of real expenditure per head on food incremes with addi t iod economic prosperity of the people. An increased expenditure on food is mually accompanied by a corresponding increase in the physical volume of food. But, perhaps, more important changes occur in the q d t y and composition of food with aa increased spending on mazketing services. A change in the quality could be conceived 88 the substitution of a commodity by the same commodity of a superior quality, whereas the change in the composition of food could be regarded a,a the substitution of one commodity for another. However, due to the absence of information about quality vaziations and m k e t i n g services analysis must be limited to money expenditure

lManuscript received Augast 2, 1971 ; revised June 9. 1972. *I em indebted to Profeseor N. C. gBkwsni of the University of New South

Weles for hie comments on 8x1 earlier draft of this pper. The essietance of Mr. P. Yong, a post-graduate student of wonomics, in the computational works is ptefillly acknowledged.

HOUSEHOLD CONSUlVIPTION OF FOOD IN AUSTRALLA 241

only, which is the payment for quality, quantity, and marketing services together. From the differences in family expenditures on different food items it is not diillcult to obtain a fairly accurate idea of the Merences in food composition.

Another related phenomenon that needs to be explained is the inter-family variations of expenditure on food, which are not attribut- able to differences in income8 j i.e., why do two families with the same income have different food habits or why do they respond differently as regards food purchases to the same increase in income. The answer to these questions may be found in the socio-demographic characteristics of the families. Hence, variations in income elasticities among various sociological groups will be examined. Account should be taken of influences of the differences of social status, cultural origin and the stage of the family life cycle. Social class has been defined by the occupational status of the head of the household using a simple blue collar/white collar dichotomy* and again by using the magnitude of family income. Cultural differences of the families me accounted for by the country of origin of the immigrants in Australia. Thus, four different immigrant groups and one indigenous group have been studied. British and New Zealand migrants have been grouped with families of Aust,ralia,n origin because of their cultural similarity. The other four groups are : West Europeans, East Europeans, South Europeans, and Asians and dl others, respectively.

The data available make i t possible to define nine stages of family life cycle:

1. 2. 3. 4. 5.

6. 7. 8. 9.

Male or female head with no spouse or prior issue (P.T.). Male head with wife but no children or P.I. Male or female head with children 1-5 years only. Male or female head with children 1-5 and 6-17 years. Male or female head with children 6-17 years, and over 17 years, 1-5 years, 6-17 years and over 1 7 years. Male or female head with children 6-17 years only. Male or female head with children over 1 7 years only. Male head with wife and P.I. only. Male or female head with no spouse and P.I.

II The common explanatory variables in the Engel relationship are

total expenditure and family size. As these variables are essentially quantitative, they render no difficulty in the estimation of the equations. Difficulties creep in as one proceeds to introduce other sociological and demographic variables. Take, for example, occupa- tion, which is essentially qualitative in character. Any attempt at quantikation of occupation is likely to be frustrated. The statistical problems involved can be avoided by classifying the families into a number of demographically and sociologically homogeneous groups and then estimating the Engel relationship for each group. The amount of work involved in this procedure does not always make it a practicable proposition. Moreover, some groups may consist of only a few families, in which case estimation may not be precise. As a matter of fact, there exists an alternative. Each qualitative variable

a This olessification is discussed in the Appendix.

242 N. PODDEXZ

may be represented by a set of dummy varirtbles. The technique of duxnmy variitbles has been wed extensively in thi8 paper. Therefore, it is worth while to give a description of the method. Let t ~ 8 formulate, say, the demand relationship for expenditure on food. If it is assumed to be a function of total family expenditure, f d y size and the occupation of the head of the household, the function can be repre- sented symbolically as

where X is the expenditure on food, P F8 the total family expenditure, 8 is the size of the family, and W is the occupation of the head. 'phe error term has been omitted for simplicity. Obviously W is 84 " attribute " vaxiable, whereas the others are ordinary quantitative variables. Therefore equation (1) can be written a8 the snm of two functions

where P is a step function of W which can be represented by a set of dummy vaxiables. Each dummy vaziable will represent a particnlar occupation group. If k fl occupation8 are distinguished, the number of dummy varbbles should be one less than the number of oocupatiom, i.e. k, otherwise the moment matrix wil l be singulrtr, and in that w e its inverse would not exist. Let the dummy v d b l e wi repwent the ith occupation. It will take value 1 if a family belongs to the ith occupation, otherwise it will take value 0. Needless to say, if any one dummy variable takes value 1, others will nwsazily assume d u e 0. Equation (2) can be written more explicitly if i t is a s ~ ~ m e d that the relation ie of the constant elasticity type, which is

where the Greek letters am the relevant mficienfs. Similarly, if more quaJitative variables are present in the equation,

each of them can be represented by a set of dummy variables. The most important advantage of the dummy variables teohnique is that the effect of the quaJitative variable is not for& to be linear. For example, a quantitative vaziable such 8s age can &o be represented by a set of dummy variables, and if treated in thh m y it may give a non-lineax effect, whereas if treated as an ordinary single variable it will always give either a positive or negative linear effect on food consumption. The other advantage of the dummy vaxiablea technique is that interactions of the qualitative variables can &o be isolated by using an extra set of dummy varbbles. However, for the sake of simplicity we have not introduced the interactions dummy variables in what follows.

The estimation of equation (3) is eqllivalent to estimating a set of equations corresponding to the occupation groups. This wi l l be evident if we break down the equation to obtain the following :

(1) X=.W,4W)

(2) X=qJW,4 +mv

(3) Log x=?o+@i Log p+pa hog B+ylW,+= * +ykwk

Log X=P,+P1 Log P + P a Log 8 Log x=(Bo+Yl)+Pl Log Y+P* Log fJ

Log x=(pO+yk)+Pl Log P + P e Log 8. ................................

The &st equation is for the &st occupation group, the second for the second occupation group, etc. For convenienw of exposition, we shall refer to the first occupation group as the reference group. Now it is obvious that in the above equations all regression coeficients

HOUSEHOLD CONSTIMPTION OF FOOD IN AUSTRALIA 243

except for the intercept terms are the same in each equation. The interpretation of this phenomenon is that at a p m t i d m level of total expenditure and for a particular family size each group spends a different amount on food, but the elasticities with respect to total expenditme and f d y size remain the same for each group. Although this method is quite useful in the analysis of inter-group differences, in the level of expenditure on a commodity, it does not explore the possibility of variations of the elasticities. To ascertain if there are differences in the elasticities among the groups, we may introduce another set of variables which are not exactly dummies. The extended equation wiU be: (4) Logx=po+piLog P + ~ B Log fi+y1'%f- * *+ykWk+h%+. - -

+hamk+pln,+. . - - h w k

where m,=w, . Log P and ni=wi . Log S .

This equation again can be broken down into Log x=po+p1 Log p+pz Log f l

Log~=(po+Y1)+(P1+~l) Log P+(P,+lll) LogE .......................................... Logx=(po+yk)+(pi+~k) Log P + ( p , + P k ) Log 8

Therefore the estimation of equation (4) is equivalent to the estimation of k + l distinct equations. It is obvious that the A and the p coefficients are the differences of income and family size elasticities of the group being considered from those of the reference group. For example, the difference in income elasticity between the reference group and the ith group is

where the second subscript of p refers to the relevant group. Similarly, the difference in f d y size elasticity is

B1-Pli=Ai-l

PS -Pei = ~ i - 1 . After the income and family size elasticities for various family

groups have been computed i t is necessary t o test the si@cance of the difEsrences of elasticities between groups. Needles8 to say, the null hypothesis in this case is that there is no difference, say, of income elasticity between the i+ l th group and the j + l t h group. Symbolically

Similarly, if the difference in f d y size elasticity is to be tested between the same groups the null hypothesis wil l be

Let Ai and Aj be the least squares estimators of A, and Aj respectively. Then under the null hypothesis that Ai =Aj, the difference (ii -ij) will have zero mean and variance

xi-?. .

Ho : A i d j .

a, : p i = p j . A CI

A h A h 2 CTA A =v(At -Aj) =*("A,) +*(^Aj) -2 COT (Ai,Aj).

Assuming that the disturbance term E~ follows normal distribution with zero mean and constant variance 02, the (&-Aj) will also be normdly distributed. observations on 3k+3, explanatory variables as given in the right-hand side of equation (4), the variance-covariance matrix of all the least squares estimators of (3k +3) coefficients is given by C T ~ M - ~ where M is the moment matrix.

A h

If P is the matrix of

244 N. PODDEP

The variances and covariances of individual coeffiaients can be written a8

A h

v(&) = C ~ U ~ , I ~ , and cov (Ai&) =&Ai)ri,

where ~ A ~ A ~ is an element of dl-l corresponding to mi and mr vasiables. Let s2 be the residual variance estimator of the equation (4). Then

82 (n -3k -3)- C2

follows x2 distribution with (12-3k-3) degrees of freedom. It follows that

A h

4 -1, 4/S2(Ukj) i i +UAjkj --2UA,hj)

1=

is distributed as Student's " 1 " with (n-3k-3) degrees of freedom. Equations of types (3) and (4) have been estimated by using the

method of ordinary least sqwes. The explanatory variable here is family expenditure, and not income, the rawon of which is explained in the first article. There it ie shown that lea& squares estimators of the coefficients would be biased. But it is found that the bias in the LS estimator of the income elagtjcity for food is very small. Also, the high varianoe of the instrumental variables estimator justifies the use of LS estimator in this cme.

rn The five components of expenditure on food that have been

1. Expenditure on groceries. 2. Expenditure on fruit and vegetables. 3. Expenditure on meat, &h, and poultry-fresh. 4. Expenditure on m a t , fish, and poultry-frozen. 5. Expenditure on food away from home. The results of the application of the s b t i ~ t i d methods discnssed

in the previous section we presented in this section. It should be mentioned that some discnssion of the regional dispaxities of family food consumption may be found in the flmt article of this series. It W&B found that the difFerences among the regions are ineignif5cant.

TABLE 1

considered are:

Average Percentage Pementage 1 - Family I of Total 1 of Total Expenditure Expenditure Food

Grooeriea .. .. Meet, etc. : Fruit and vegetebies . .

h h .. .. Frozen .. ..

Food SW837 from home . .

646 160

277 23 79

18- 50 3-83

1-08 0.59 2.02

54-93 12-76

23-55 1.96 6-72

~~ ~~

Tom food .. .. I 1,176 I 30.04 I 100*00

Therefore, we shall not pursue the analysis of regional disparities any further in this paper. The starting point of this section is a table which will give a preliminary idea of the level and composition of the food consumption expenditure of the Arustfalian familiee. Table 1

HOUSEHOLD CONSTJMPTION OF FOOD IN AUSTRALIA 245

shows average family expenditure on each of the five food items, percentage of total expenditure spent on each item, and the com- position of food in terms of the share of each item in total food consumption.

It is seen that the largest proportion of food expenditure is made on groceries. This is quite reasonable considering that groceries are a bundle of heterogeneous commodities, all of which are necessities. Nest in order of the proportion of expenditure are meat-fish-poultry, fruit and vegetables, and food away from home. It should be noted that the amount spent on frozen meat is extremely small and insi,@€icant. This shows a definite preference for fresh meat to frozen meat.

A set of Engel equations for the five items and for total food is presented below. These are simple relationships of the constant elasticity type where total expenditure and family size are the only two explanatory variables. Groceries . . LogX= 1*1895+0.3626L0g Y+0.5616 L o g s R*=0-667

(0.0139) (0.0148) Fruit md veget- 8 b h . . . . Log X= 0*3263+0.4266 Log Y +0-4472 Log S R8=0-422

M a t , &&.--fresh Log x= 0*0756+0.5892 Log Ys-O.3567 Log S Ra=O-466 (0- 0261) (0.0280)

(0.0247) (0.0266)

(0 * 0530) (0- 0569) Meat, etc.-frozen LogX=-l-3389+0.5225Log P+0.0308 Logs R"0.177

Food away from home .. . . LogX=-5~4657+1~9020LogY+O~2024Log~ R*=0.456

(0.0580) (0.0623)

(0.0099) (0.0107) Totd food . . LogX= 1.4073+0.4883 Log P-0.3860 L o g s R'=O.776

Needless to say, the coefficients of the above equations are the income and f d y size elasticities. Now we are in a position to compare the income elasticities of the various items. It is seen that of all the items groceries have the lowest income elasticity and the highest family sbe elasticity. This shows the importance of groceries as the most necessary item in the food list. The next item, fruit and vegetables, has a slightly higher income elasticity and a slightly lower family size elasticity. Income elasticity of demand for meat again is slightly higher and the family size elasticity still lower. This obviously shows that the importance of meat as a necessary food is, technically speaking, less than the previous two items. It should be noted that the expenditure on frozen meat is hardly explained by its equation, which may suggest that buying frozen meat is almost a random phenomenon. Because of the ins imcant proportion spent on frozen meat and the poor explanatory power of the corres- ponding Engel equation, there will be no further treatment of this category in the following analysis. Therefore, from now on the fourth item will be food away from home and the fifth item will represent total food. Of the above equations, the one for food away *om home is of particular interest. In this case the income elasticity is almost equal to 2, which meam that eating-out is a luxury and it is becoming increasingly popular in Australia. The coefficients of determination show that the equation for to td food has the best fit.

The additional explanatory variables will now be introduced. The first set of variables is introduced to take account of whether the

D

246 N. PODDm

families in the higher income class have a different food habit from that of the families in the lower income class. InitiaJly, one dummy variable is introduced to change the intercept term for families with income more than $4,000. It will be denoted by DI1. The eqnations are presented below according to the order mentioned in the previous PaJ=#PPh-

1. Log X= 1*0699+0.5616 Log Y+0.4012 Logs-0.0385 DI, R*=0.676

2. L*X= 0.2949+0*4367 Log Y+0.4473 Logs-0~0101 DI, R"0.425

3. Log X= 0.0978+0*5820 Log Y+0*3568 Log 8+0.0071 DI, R'=0*473

(0-0148) (0.0154) (0- 0068)

(0 * 0291) (0.0280) (0 - 0128)

(0.0275) (0.0660) (0.0121)

Fruit FlWh and Meat

class \ Groceries Vegetables

4. Log X=-4*6325+1.6332 Log Y-0.2031 Log S+0*2680 DI, R'=0*466

6. LogX= 1.0843f0.4957 Log Y+0.3860 Logs-0.0074 DI, R'=0*766 (0.0640) (0.0683) (0.0282)

(0- 011 1) (0.0107) (0 * 0048)

Food Away Total from Food Home

Low income Y .. s ..

High income Y .. s ..

0.436 0.475 0 * 594 1.544 0.607 0.532 0 * 384 0-347 - O * l O a 0.373 -- 0.347 0.405 0.563 1-762 0.481 0.699 0.533 0.368 -0.345 0 * 406

Table 2 shows t h t there is a systematic pattern of vrtriations of income and family size ela8ticities between the two income ClaElses. For all the item with the singulm exception of food away from home, the income ek~ticity is higher and the f d y size elasticify is lower for the low income chss. In the w e of food away from home the situation is exactly the opposite. This means that the high income group wi l l respond with a higher percentage increase in expenditure on eating-out than the low income group to the same percentage increase in income. As for the significance of the dXerences of income

HOUSEHOLD CONSUMPTION O F FOOD IN AUSTRALIA 247

Class

elasticities, t-statistics are computed according to the formula given in Section 11.

1-461, 2.131*, 2-844*, 1 -077 , and 1.843* respectively for the five items. The figures with an asterisk are significant at 95% confidence interval, implying that the differences are significant only for fruit and vegetables, and for fresh meat, while it is also significant for overall food consumption expenditure.

Next, we shall analyse the difference of food consumption between the middle class and the working class. As it is mentioned in the previous section, this class distinction is made on the basis of the occupation of the head of the family, and hence to a great extent arbitrary. As an introduction to the analysis of dietary differences, a composite table will be presented. Table 3 shows number of families, average family expenditure, average percentage of expenditure on food, and percentage of total food expenditure on the four items

The computed t-values are:

Number of

Families

Middle

Aver5g0 Expend.

iture (8

4,765

3,532

TDLE 3

Average Percentage of Total Food

centage Per- Expenditure on

from on Food ables Home

26.46 52.89 12.37 23.24 9.36

32.19 1 56.11 1 13.01 1 23.66 1 5.36

From Table 3 a preliminary idea of the food consumption patterns of the two social classes may be obtained. While there is a si@cant difference in the proportion of total family expenditure spent on food by each of the two classes, the differences in the proportions of total food expenditure on each of the four items appear to be insig- nificant except, perhaps, for food away from home. However, further statistical analysis is needed before any firm conclusion is drawn. It is to be noticed that the absolute difference of average family expenditure between the two classes is quite sizeable.

A refined measure of the income and family size elasticities may be obtained when the differences between the two social classes are taken account of. This is done in the following set of equations by introducing the dummy variable DO1, representing the working class. The order of the equations is the same as before.

1. Log X= 1*0396+0.3959 Log Y+0*5529 LogS+O.O539 DO, RP=0*672

2. LOgX= 0*1825+0.4585 Log P+0.4382 Log S+0.0517 DO, R2=0.425

3. L0gX=-0.0309+0.6128 Log Y+0.3501 L0gS+0.0383DOI R'=0-474

(0.0144) (0.0149) (0.0065)

(0.0271) (0.0266) (0.0123)

10.0257) 10.0266) 10.01 17) 4. Log X= -4- 7918+1* 7525'Log Y-'O.2425'LOg Sf'O.2325'00, R'=O. 468

5. Log X= 1.0308+0.5052 Log Y+0.3812 Log 8+0-0275 DO, RB=0-767 (0.0599) (0.0621) (0.0272)

Apart from the more accurate measure of the elasticities, the equations again confirm the conclusions drawn from Table 3.

248 N. PODDER

country origin Of

Auetrslie, Great Britain and New Zeslend

south Evope We& Europe Eaet Europe.. Mi and othm

Now we shall show how income and family size elasticities vary between the two social classes. This has been done by introducing two extra dummy variables in the above equations. The results me presented in Table 4.

TABLE 4 Inunne and Fami& Sdze Ehticitties for th.e Two Classes

Number F$ Expend.

4,699 3,862

4,239 242 4,790

4,695

\

29-86 34.89 29.12 29.78 26-42 L

\ I I Vegetables I

Fruit Gro- and wries Veget-

ables

55.32 12.65 53-76 13-86 56-52 13-68 51-16 13.24 50.12 13-70

wor?w:.. I 0.430 I 0.490 I 0.658 I 1-717 I 0.529 8 .. 0.507 0.405 0-290 -0.053 0.370

Food Fresh Away Meat from

Home ----

23.21 6-81 26-04 4-59 23-24 6-85 25.79 7-56 24-33 9-99

Middle : Y .. 1 0-324 1 0.384 1 0.518 1 1.759 1 0-432 8 .. 0.043 0.502 0.472 -0.408 0-394

Table 4 shows that for aJ1 items excepting food, away from home income elasticities me higher asd f d y she elssticities are lower for the working class than those for the middle &as. This implies that the working clam hm more unsatisfied for the b s i c food items, and hence for total food.

A8 for the significance of the Werence of income elaaticitim between the two classes, the t-atatistics a m given below in order.

2*838*, 1*644*, 29100*, 0.371, 3*627*. The t-values are s igdcant for aJ1 the items excepting expenditure on food away from home. Overd, it crta be concluded that there is a statistically aigniflcamt difference of behaviour of the two classes with respect t o food consumption.

The effects of another sociological variriable, the country of origin of the families, will now be considered. Again, as an introduotion t o the analysis of the differences of food consumption of various cultural groups Table 6 is presented, which is similar to Table 3. The first column of the table now represents the country of origin.

Per- oentage

of Expend

iture on

Food

HOUSEHOLD CONSUMPTION OF FOOD IN AUSTRALIA 849

' 0.0097 (0.0158)

~ (KE) 0.0904

-0'1538

0.0462

' (0.0280)

1 (0.0657)

, (0.0112)

There are a number of interesting points that can be noted from Table 5. The South Europeans spend a significantly higher pro- portion of expenditure on food than any of the other groups, although the average family expenditure of the former group is not the lowest. One explanation of this phenomenon may be that the South Europeans have a bigger family size on the average. Asians, by far, spend the smallest proportion of family expenditure on food and the= average family expenditure is one of the highest. If the proportion spent on food is an index of the standard of living, it could be concluded that the Asians as a demographic group has the highest standard of living in Australia. It should also be noted that the Asians spend a higher proportion on food away from home than any other group. Except for food away from home, inter-group variations in the proportions spent on other food items do not seem to be quite significant.

When the effects of cultural background of the families are taken moun t of, a more refined set of income and family-size elasticities may be obtained. To that end a set of dummy variables representing the countries of origin are introduced in the Engel equations. These variables will be denoted by DC,, DC,, DC,, DC, for South Europe, West Europe, East Europe and Asia and others respectively. Needless to say, the Australians are the reference group. The equations are presented in Table 6 in the same order as before.

-0.0511 (0.0205)

(o":::::, -0.0099

0.1664

-0.0054

(0.0346)

(0.0855)

(0.0146)

Equa- tion

1

2

3

4

5

1-1842

0.3257

0.0833

-5.4714

1.1115

0.3646 (0.0139) 0.4264

(0.0262) 0.5862 (0- 0247) 1.5051

(0.0581) 0.4867

(0.0099)

0.5586 (0.0150) 0.4386

(0.0282) 0.3466

(0.0266)

(0.0625) 0.3829

(0.0172)

- 0 * 1755

0.0239 (0.0140) 0.0511

(0.0263) 0.1157

(0.0245)

(0.0585) 0.0441

- 0.2163

(0~0100)

- 0.0146

-0.0056 (0.0161)

(0.0302)

(0.0286)

(0.0671)

(0.01 15)

-0.0125

-0 1808

- 0.0258

It can be observed that in at least two equations all the coefficients of the dummy variables are statistically significant. These two equations are for food away from home and for total food. In the other equations only a few coefficients of the dummies are significant, e.g. in the case of meat the dummies for South Europe and East Europe are significant. The interpretation is that these two migrant groups have a sigmficantly different meat eating habit from that of the reference group. However, to explore the possibility of inter-group variations of elasticities two sets of additional dummy variables have been introduced in each equation. The results have been presented in Table 7, which is similar to Tables 2 and 4. It is obvious from Table 7 that there are wide variations in both income and family size elasticities among the cultural groups for any single item of food. It mn be said that the lower the income elasticity for any item, the more saturated the group is in the consumption of that item. Thus, one codd safely conclude that the East Europeans are more saturated

250 N. PODDEB

Fruit and

Vegetables Groceries

Origin

with groceries than any other group. Consequently, they are likely to spend most of their additional food expenditure on items other than groceries. However, for the statist id signiscance of the

Food A m y Totel

Meat from Food Home

Auetralia, Great Britain and New Zealand : Y . . .. s .. ..

0.356 0.421 0.565 1.941 0-477 0.558 0-423 0.357 -0.142 0.383

South Europe : Y .. 0.418 0-462 0.797 0.570 s .. :: 1 0-469 1 0-271 1 0.357 1 --A:::: 1 0.348

Asia and others : Y .. .. s .. ..

West Europe : Y .. .. I 0.609 I 0.628 I 0.885 1 1-443 I 0.714

0.424 0.351 0.718 1.689 0.545 0.594 0.774 0.280 -1.244 0.318

I -0-366 1 s .. 0.666 I 0.681 I 0.415

East Europe : Y .. .. I 0.297 1 0.507 1 0.738 I 1-546 1 0.459 s .. . . 0.604 0-633 0.414 -0.215 0.384

difference of income elasticities between any two cultural groups t-statistia have been computad. As five different family groups are comidered, t-values for each item am presented in the form of a matrix. As the matrix is always symmetric, the lower off-diagonal elements have been omitted. To examine the significance of the Merence of income elasticities between the ith and the jth cultural groups, the relevant t-value is the element corresponding to the ith row and the j th column of the matrix. The six matrices are presented below.

The values with an asterisk are signiffcant at 95% confidence interval. It is interesting to note that for food away from home the differences of income elasticities among cultural groups are statistically insignificant. Aa regmds f i t and vegetables, the only difference that is significant is the one between the third and the B th group, while for meat the significant difference is between the South Emopeam and the reference group. For groceriea and for total food, most of the t-values of the differences of elasticities are significant. The South Europeans and the West Europeans, on the whole, have significantly different income elasticities for food than the Australisns. It should be noted that the homogeneity of the kmigmnt groups does not necessdy meaa that they have the same food conerumption pattern. It has already been found that with respect to the level of consumption of food items the differences among the groups are quite significant.

HOUSEHOLD CONSUMPTION OF FOOD IN AUSTBALIA 251

The last demographic characteristic to be considered is the stage of the family life cycle. AB has already been mentioned, nine different stages of the family life cycle have been distinguished, and their effects on food consumption analysed. Table 8 would give us a preliminary idea of the composition of food of the families at different stages of their life cycle. The average family expenditure systematically

Groceries Fruit and Vegetables

1 2 3 4 5 1 2 3 4 5

1 0 1-247 3-038* 0.741 0.927 1 0 0.295 1.321 0.581 0.507 2 0 1.667* 1-358 0.241 2 0 0.929 0.218 0.772 3 0 3.463* 1*821* 3 0 0.644 1.972* 4 0 1.118 4 0 0.933 5 0 5 0

Fresh Meat Food Away from Home

1 2 3 4 5 1 2 3 4 6

1 0 1-770* 0.807 1.230 1.165 1 0 1.067 1-432 1.201 0.823 2 0 0.611 0.392 0.512 2 0 0.472 0.332 0.215 3 0 0.327 0.299 3 0 0.344 0-266 4 0 0-254 4 0 0.300 5 0 5 0

Total Food

1 2 3 4 5

1 0 1*768* 3.988' 0.320 1-290 2 0 1*955* 1.533 0-401 3 0 2*622* 2*011* 4 0 1-094 6 0

rises and then falls over the life cycle. Absolute expenditure on food also varies in the same way. The table also shows that the families in the first stage of the life cycle spend the lowest percentage of total expenditure on food, whereas the families in the last stage spend the highest percentage on food. This may be due to change in the attitude of the family over the life cycle. In this case it may be wrong to conclude that a family is better off in the first stage than in the last stage of its life cycle. As for the composition of food, it is =cult to find any significant variation over the family life cycle. However, there are some differences in the composition of food at the first and the last stage. The only item that shows a fluctuation is food away from home. It is interesting to note that the families in the first stage spend a greater percentage of food expenditure on eating out than any other group. The same group spends the lowest percentage on groceries. The families in the last stage spend proportionately more on fruit and vegetables and less on eating-out than any other group.

252 N. PODDER

Number of

Families

418 609 776 787 584

1,001 487 522 259

The Engel relationships that contain the dummy varhbles for the stages of the f d y life cycle may now be presented. The number of dummy v h b l e e is obvioudy eight. Tbe reference group in thie w e ia the group of f d e s in the first stage of their life cycle. The

TABLE 8

Average Expend-

iture

2,805 3,699 3,846 4,228 5,369 4,284 4,547 2,997 1,462

8-0

cycle

Ill Family

Life

centage of

Expend- iture on

Food Groceries

Fruit and

Veget- 8blea

Fresh Mest

Expenditure on

24-35 26.25 29.15 33.09 30.68 31.42 28.63 30.76 33-93

48-75 51.59 58.62 59.42 53.61 55.79 50.38 53.69 52-62

13.62 12.36 11-78 12.57. 12-87 12.93 13-29 13-02 16.13

~~

Food A-Y from Home

12.74 9.27 6- 14 4-93 7-72 6-94 8-22 4-12 3-43

22.84 24-61 21.41 21.36 23-86 23.66 25-96 26.89 25.40

equationa are presented in Table 9. Judging from the standard errom of the coef€icienta of the dummy variebles, moat of the variablm am significant. It has &eady been noted that the fht group, which consists of families of young married couples without children or of unmarried persona only, have a Merent food habit from those of other groups. This being the reference group, it i~ hardly surprising that levela of food conaumption of the other groups are si@cantly Werent. Aa a matter of fact, groups 2-8 seem to have a similar composition of food.

Iv In the first paper of thie series a few comment8 have been made

with reepect to the di.fklcultie8 and problems involved in making intemtiond comparieone of income elasticities. In compa+ng income elasticities of the Austdian f d m with those of the famdies of other countries, the additional difEculQ ia that the food items coded in the A n s m survey are too aggregative, wheress the estimetes of income alasticities that are available for other countries am mostly for individual commodities. A6 a result the scope of a comprehensive international comparison is severely limited. However, an attempt haa been made for a comparison of the income elagticitiw of demand for two selected food items and of tota,l food. These two items am meat, and fruit and vegehbles. Even this compasison is only an approxi- mation due to the k t that oversw works considered frnit and vegetables aa two separate items, and @ expenditure on meat does not include expenditure on poultry and fish aa in om data. For the purpose of cornparisone, me ha6 been made of the Wticitia computed from aimple log-linear Engel eqnationa where total expanditnre and homehold size am the only explana;tory variables. The murce of the estimates for other countrie~ ia LiviataS (1964), who in turn obtained them from elsewhere. It should be noted that the data from which

Equ

atio

n N

umbe

r

0.10

04

(0.0

146)

0.06

23

(0.0

216)

0.21

78

(0.0

267)

0.03

72

(0.0

607)

0.07

64

(0.0

104)

Inte

mep

t T

erm

0-09

66

0.09

62

0.08

44

(0.0

179)

(0

.021

2)

(0.0

213)

0.01

06

0*06

87

0.07

42

(0 0

337)

(0

.039

9)

(0 - 04

01)

0.13

13

0,14

36

0.22

63

(0.0

316)

(0

.037

3)

(0.0

376)

0.05

62

0.14

68

0.23

73

(0.0

743)

(0

*088

0)

(0*0

886)

0.03

67

0*06

67

0.07

89

(0.0

127)

(0

.016

1)

(0.0

162)

1.07

62

0.22

71

0.01

28

- 6.0

430

1.07

578

Log

Y

0.37

13

(0.0

144)

0.43

30

(0.0

272)

0-66

89

(0,0266)

1 * 8

438

(0.0

600)

0.48

01

(0.0

103)

~~

Num

bers

in

brso

kets

are t

he

Log

8

0.63

28

(0- 02

96)

0.46

16

(0 - 06

66)

0.31

93

(0.0

620)

- 0.6

847

(0- 12

26)

0.37

40

(0.0

210)

anda

rd errors.

-- 08,

0.11

06

(0.0

183)

0.10

03

(0,0

346)

0 * 2

243

(0- 03

23 1)

0,17

36

(0.0

762)

0.07

64

(0.0

130)

08

,

0.09

94

(0 0

1 76

)

0.12

66

(0.0

330)

0.26

77

(0.0

309)

0.04

32

(0 * 0

727)

0.09

85

(0.0

126)

D8

,

0.14

16

(0.0

149)

0.12

44

(0.0

280)

0.29

13

(0.0

262)

-0.4

336

(0 0

61 8)

0.10

11

(0.0

106)

a g 1 0.

1188

(0

.017

6)

0-13

87

a (0

.033

0)

8 0.

1600

(0

* 030

9)

w 0

-0.2

478

0

(0.0

728)

u

0,06

73

2 (0

.012

6)

p w P 8 E

254 N. PODDER

Meet .. .. Fruit .. Vegetables . . Total food ..

the overseas estimates are made are those of pre-wrtr years. Only the European countries are considered for compaxbon. Table 10 presents the estimates.

TABLE 10 Incomc Elasticities

Average

countn'es

United of Kingdom Sweden Europem Australia

0.40 0.49 0.80 0.52 0.88 0.84 n*8.} 0-43 0.57 0-46 n.8. 0.53 0.53 0.52 0.49

From Table 10 it is seen that the income elasticities of the major food items in Austdia are very similar to those in the Enropean countries. While the income elasticity of demand for meat in Austrah is slightly lower than the European average, it is slightly higher than that in the United Kingdom and in Sweden. It is diBcuIt to draw any conclusion about fruit and vegetables, but one could guess that for this item the income elasticity is lower in AustraJia than that in the European countries. With respect to the elasticity of total food, the similaxity of Australia with Europe is obvious.

At this stage it is approprkte to make a few concluding remarks. Present-day consumption economics is p t l y infiuend by the works in the related fields of knowledge, e.g. sociology, psychology and marketing. The demand for a commodity is no longer deemed to be a function of prioe and income only. Motives, attitudes, expectations of the consumers me recognized to have an impact on the decision to buy. As none of these factors is quantifiable in any way, one naturally looks for the variables that are responsible for forming motives, etc., and andyse the effects of these variables on consumption. Some of these factors are considered in this study. It may be recdled that only one such factor wzts considered at a time. The prtECtim wfls to an extent imperative, due to the existence of high intercorrelation among variables.

The statiaticd model used in this study is fairly simple. The andysis may not be quite comprehensive. Yet it i6 the first study of its nature in Australia. A more comprehensive study may be canied out when detailed information about expenditures on individual commodities me available.

HOUSEHOLD CONSUMPTION OF FOOD IN AUSTRALIA 255

Appendix (A) Nine Merent occupations of the head of the family have been

coded in the survey. These are: 1. Self-employed. 2. Executive and managerial. 3. Professional and technical. 4. Vocational and semi-professional. 5. Clerical. 6. Sales. 7. Craftsmen and skilled technical. 8. Operative and semi-skilled. 9. Labourers, farmers, fishermen and unskilled.

Of these the first six categories me included in the group of white collar and the rest are included in the blue collar workers. It is obvious that this division is somewhat arbitrary. This was due to the lack of more detailed information.

(B) Table 11 has been added to give some idea of the food consumption patterns of the different age groups. It was not added in the text bemuse of the high correlation of the stage of the family life cycle and the age of the head of the family.

TABLE 11

Fresh Meat

-~ 20.15 21-27 21.25 22.51 23.12 23-43 24-81 26-17 26.28 25.58

Age of H 0 d

Fruit and

ables veget-

-~ 10.83 11.60 12.36 12.64 12.91 12.23 12.82 13.65 14.01 14.16

Under 25 years 26-29 . . .. 30-34 . . .. 35-39 . . . . 40-44. . . . 45-49.. .. 50-54 . . .. 55-59 . . .. 60-69 . . .. Over 69 ..

Numbex of

Families

286 55 1 604 634 726 630 510 388 675 439

4verege Expend iture

3,633 3,886 4,005 4,399 4,801 4,681 4,487 3,968 2,991 1,931

Per- I Percentage of Total Food xntage Expenditure on

of Expend iture on

Food

25.41 27.95 30.31 30.39 31.32 31.08 29.55 25-98 30.79 35.85

Gro- ceries

56.12 56.08 58.07 56.62 55.67 51.82 52.87 52.00 52.88 55-20

Food Away from Home

10-94 9.30 6.10 6-13 6-45 7 -22 7-54 6-35 4-67 3.03

References The other articles in this series are : (a) '' Patterns of household consumption

Expenditures." The Eummnk Rewrd, September, 1971. (a) "The estimation of an equivalent income scale.'' Awtralhn EcmwmiC Papem, December, 1971. (c) I' Distribution of household income in Australia." The Ewnumic Record, June, 1972. Liviaten, N. (1964). G o n e m H h Pattern in h a d . Jerusalem.