Household factors associated with older Australian's purchasing a varied diet: results from household expenditure data.

Abstract

Objective: To examine the household factors associated with older Australian's (aged 55 years and over) purchasing a varied diet.

Design and methods: National cross-sectional survey.

Setting: All states and territories of Australia.

Subjects: A total of 1898 households headed by adults aged 55 years and over.

Main outcome measures: Food diversity was measured using the Dietary Variety Score (DVS). The DVS is the total number of food items purchased within a two-week period, from a possible 110 individual food expenditure items.

Data analysis: Negative binomial regression was used to examine the association between the respondent's individual and household characteristics with the DVS, after controlling for all factors in the model.

Results: Among the demographic factors, living arrangements, age and education are found to be associated with dietary diversity. Lone males, lone females, and households headed by people with no secondary school education, or aged over 75 years were most likely to experience lower levels of dietary diversity.

Conclusions: The finding that both poor education and low income in old age are associated with dietary diversity suggests the need for early life-course interventions to improve nutritional outcomes. Nutrition education programs that also focus on those with poor education and low income, as well as lone person households and those aged over 75 years, may improve nutritional wellbeing, and thereby reduce public health costs.

Key words: ageing, dietary guidelines, food diversity, household expenditure.

INTRODUCTION

Given that approximately four of every five older Australians have chronic diseases that are influenced by diet, (1) public health costs could be reduced considerably with improvements in diet. (2) It is well recognised that dietary-related illness among older Australians is a considerable driver of public health expenses. (3) As population ageing continues in Australia, more information is required on the dietary adequacy and dietary behaviour more generally of older people that contribute to public health care costs.

Research shows that an important measure of nutritional health is dietary diversity. It is well understood that dietary variety increases the likelihood of nutrient adequacy of older people. (4-7) Moreover, consuming a diverse diet has been shown to result in lower mortality (8) and a lower incidence of macrovascular disease, obesity, diabetes (9) and cardiovascular risk factors. (10) In recognition of these important health consequences, the Commonwealth Department of Health and Aged Care funded the establishment of a set of dietary guidelines for older Australians. (11) The first dietary goal outlined in this guide is the need to 'enjoy a wide variety of nutritious foods'.

Although there is growing recognition of the need to increase dietary diversity among older people, little is known specifically of the socio-demographic characteristics of older people who consume a diverse diet. The present paper examines one particular aspect of dietary diversity, diversity in food purchasing. More specifically the purpose of the present research is to use nationally representative data to examine the household and individual factors associated with higher diversity in food purchasing.

MEASUREMENT OF DIETARY VARIETY

Prior research has employed a range of dietary diversity measures. For example, Finke and Huston used the Healthy Eating Index to examine differences in dietary variety between the elderly and non-elderly in the USA. (12) In another study, Thiele and Weiss's analysis of German households used both the Berry Index and Entropy Index to examine the relationship between household composition and food diversity. (13)

Following research from both the economic and nutrition literature, the Dietary Variety Score (DVS) is used to measure diversity in food purchasing in the present study. (14,15) The DVS measures the total number of food items purchased during the reference period of the survey. This measure has the advantage of eliminating cultural and other forms of bias in assigning scores, or grouping foods to create an index. (14) Furthermore, the DVS has been used in prior studies that have used household expenditure, rather than nutrition data sets. (15,16)

One means of validating this measure of dietary variety is through a comparison of the DVS and expenditure on selected necessities. In research on low-income earners food-purchasing choices, Turrell hypothesises that food expenditures are reduced to cover non-discretionary costs such as rent and electricity. (17) Drawing upon other research, the author suggests that this may force low-income households to purchase a less varied diet, in smaller quantities and of lesser quality However, these data did not enable the issue to be resolved. In the following analysis, Turrell's proposition is tested.

METHODS

Data for the present study are from the 1998/99 Household Expenditure Survey (HES), collected by the Australian Bureau of Statistics. These data were made available to the author through an agreement between the Australian Bureau of Statistics and the Australian Vice Chancellors Committee.

The purpose of the HES was to collect information on the income, expenditure and other aspects of wellbeing in the Australian population. A total of 13 964 individuals responded to the survey, from 6892 households. People living in non-private dwellings, such as nursing and convalescent homes were not included in the survey.

The present paper examines the food-purchasing behaviour of a subsample of 1898 households headed by Australians aged 55 years and over. Apart from collecting information on a broad range of socio-demographic characteristics, the HES includes information on the weekly equivalent household expenditure for 110 individual food items, collected using a self-completed diary.

In order to examine the factors associated with purchasing a varied diet, Ordinary Least Squares regression is inappropriate. As is well established, Ordinary Least Squares regression produces inefficient, inconsistent and biased parameter coefficients when used with a count-dependent variable. (18) This source of bias is eliminated using negative binomial regression. (19)

Prior to proceeding to results from the multivariate analysis, descriptive statistics are presented. All analyses are conducted using STATA for Windows (STATA 8.2, Stata Corporation, College Station, TX, USA, 2004).

RESULTS

Table 1 displays the mean DVS classified by key socio-demographic characteristics of the sample. Significant mean differences (MD) are identified using Sidak's adjusted multiple comparison test. As hypothesised, purchasing diversity differs by both economic and demographic factors.

People living in couple-only households purchased approximately 34 individual food items over a two-week period. This is significantly higher than the reported average of 24 and 19 items purchased by lone females (MD = -10.29) and lone males (MD = -14.6), respectively. In contrast, couple-only households purchased significantly lower numbers of food items when compared with couples with children (MD = 7.15) items and people living with others (MD = 4.35).

The mean DVS index differed also according to age. Whereas those in the oldest age group (75+) purchased about 26 food items, those aged 65-74 and 55-64 years scored significantly higher on the DVS index, 30.72 and 33.31, respectively (MD = -6.43; MD = -2.58).

People with a university education purchased about 35 food items, significantly higher than people with no post-secondary education who purchased about 28 (MD = -6.3). The DVS does not differ by a person's country of birth, or whether the household contains a person with a health condition or disability (Note: A disability or health condition is characterised as: a sight problem not corrected by glasses, hearing problems, speech problems, blackouts, fits or loss of consciousness, slowness at learning or understanding, limited use of arms, difficulty gripping, limited use of legs or feet, any condition that restricts physical activity, shortness of breath or difficulty breathing, chronic or recurring pain, disfigurement or deformity, mental illness, long term effects as a result of head injury, stroke etc. or a long term condition requiring treatment or medication.) (MD= -0.14 and MD = 0.799, respectively).

Apart from the DVS index differing by demographic characteristics, economic factors appear to be important also. Significant differences in the number of food items purchased are found between labour force status, income and housing tenure. People in the labour force purchased about 35 food items, compared with 29 for retired people (MD = -6.18). Similarly, people within the second and higher quartiles of household income scored significantly higher on the DVS index than those in the lowest quartile of household income (MD = 3.5, MD = 8.9, MD = 13.6). Finally, people who had reached old age, without the resources to purchase their own home scored significantly lower on the DVS index (MD = -7.16, MD = -6.48).

Although these descriptive findings offer some evidence that food-purchasing diversity differs by both demographic and economic factors, results may be confounded by other variables. These factors are controlled for in the negative binomial regression model.

Table 2 displays results from a negative binomial regression model that estimates the importance of economic and demographic factors in explaining dietary diversity, after controlling for all other factors in the model. The regression coefficients (b) along with the SE are included to measure the direction and uncertainty surrounding the estimated relationship. The incidence rate ratio (IRR) is included along with 95% CI of this estimate. The IRR is a measure of the expected change in the DVS given a one-unit change in an independent variable. That is, how the number of food items purchased changes in response to changes in the independent variable.

The IRR is calculated as the exponent of b. For example, the negative b coefficient measuring membership of the 75+ group indicates that after controlling for all other factors in the model, they purchase a less varied diet when compared with the comparison age group, 55-64 year olds (b = -0.069). Taking the exponent of b, IRR = exp{-0.069} = 0.933, shows that a household headed by a 75-year-old person scored 6.7% lower on the DVS, ((0.933 - 1) x 100), when compared with a person aged 55-64 years.

Consistent with results from the univariate analysis, both demographic and economic factors are associated with the number of food items purchased by an elderly household. Among the demographic factors, living arrangements, age and education are all found to be associated with food diversity as measured by the DVS, even after controlling for economic factors in the model.

Belonging to either a lone female or lone male household is found to be associated with purchasing less food items, when compared with couple-only households. Lone males scored approximately 40% lower (IRR =0.597, CI = 0.564, 0.631), and lone females about 25% lower on the DVS index (IRR = 0.752, CI = 0.719, 0.786) when compared with couple-only households. That is, they purchased 40% and 25% less food items, respectively.

In contrast, living with co-resident kin, or others during the later life course is associated with higher food diversity. The DVS for people in these living arrangements was approximately 1.13-fold that of couple-only households (IRR= 1.127, CI = 1.070, 1.186).

Age is also an important factor in determining food diversity. As discussed above, households headed by people aged 75 years and over scored approximately 6.7% (IRR = 0.933, CI = 0.889, 0.980) lower on the DVS, when compared with the youngest age group (55-64 years). Although the parameter estimate for those aged 65-74 years is negative, suggesting lower purchasing diversity, the coefficient does not reach statistical significance (b = -0.031, SE = 0.020).

Finally, of the demographic factors low education is highly significant in predicting dietary diversity. Households headed by people who did not obtain post-secondary school education scored approximately 7.8% (IRR = 0.922, CI = 0.872, 0.975) lower on the DVS than households headed by a person with university education. No significant difference was found between those with vocational education and the university-educated (b = -0.03, SE = 0.029). Once more, consistent with the univariate results, having a person with a disability or health condition in the household is not associated with the level of food diversity (b = 0.017, SE = 0.018).

Even with controls for demographic effects, economic factors remained important in predicting food diversity.

As expected, a positive relationship was found between income and the DVS. Income is measured as the natural logarithm of weekly household income. For a one-unit increase in income, the number of food items purchased increases by 4.7% (IRR = 1.047, CI = 1.021, 1.074) holding all other factors constant.

As an indicator of wealth, owning a dwelling in old age is associated with higher food diversity. Older people who remain in rental accommodation or other forms of tenure experience considerably lower levels of food diversity when compared with home owners. For example, households living in a rented dwelling scored about 7.2% (IRR = 0.928, CI = 0.883, 0.975) lower on the DVS when compared with homeowners. Similarly, households living in other tenure accommodation scored about 14% (IRR =0.859, CI = 0.771, 0.0957) lower on the DVS compared with homeowners.

Contradicting results from the univariate analysis, once appropriate demographic controls are included, labour force status is not associated with food diversity (b = 0.004, SE = 0.023). Potential associations between labour force status and other variables were tested using both variance inflation and matrix condition number tests. (20) Both tests indicated that multicollinearity was not present in the model (mean variance inflation = 1.59 < 10; condition number = 15.14 < 30).

As indicated earlier, Turrell hypothesised that housing costs may limit expenditure on food, thereby constraining dietary variety. (17) The HES data enable the housing and power costs of elderly households to be compared with their DVS index. Table 3 displays the weekly-median housing and power expenditure, by four major categories, classified by the DVS. For comparative purposes, the DVS has been divided into three equal groups: <26 (33.4%), 26-36 (32.8%) and >36 (33.8%) individual food items purchased. Median weekly expenditures are displayed rather than means due to non-normality in these distributions. The 95% CI around the median is estimated using a non-parametric bootstrap, based on the bias-corrected method. (21)

As shown in Table 3, older households with the lowest DVS spent significantly less on both housing and power costs. Whereas a household with the lowest DVS index spent $31.77 per week on housing, households with a DVS higher than 36 spent approximately $37.50. Importantly, households with the lowest DVS actually devoted higher proportions of their budget to housing costs, almost 14% (95% CI = 29.68, 34.67) when compared with households in the middle (median = 8.84; 95% CI = 8.41, 9.27) and the highest DVS categories who spent under 7% of their weekly budget on housing costs (median = 6.45; 95% CI = 6.18, 7.03).

This effect is apparent for power costs also. Older households in the lowest DVS range, spent only $10.82 per week on power, compared with $17.64 by households with the highest DVS. Yet, households with the highest category devoted significantly less of their weekly budget to power (2.82%) when compared with households with a DVS less than 26 (4.41%).

DISCUSSION

The present paper has found that both demographic and economic characteristics of the household are associated with diversity in the food-purchasing practices of older Australians.

The finding that people living alone purchase a less varied diet, when compared with couples is consistent with international findings that show the presence of a spouse is strongly related to dietary variety. (4,7) More generally, living alone is an important risk factor for poor nutrition. (22,23) The social aspect of eating, difficulty in food preparation for one person and depression-induced appetite loss have all been suggested as reasons for the poorer nutritional status of lone people. (24) A further explanation is the diseconomies of scale that occur with living alone. As lone person households are unable to share fixed costs, for example housing and power bills, it is relatively more expensive to meet these expenses. All other things being equal, lone people must devote larger proportions of their budget to fixed expenses before allocating money to food.

With advancing age, the body's need for total energy decreases, but its requirement for micro- and macronutrients increases. (11) The finding that people in the oldest age group purchase a less varied diet is important as it shows that people in greatest need of a varied diet are at risk of meeting nutritional requirements. It may be that with advancing age, older people return to consuming food that they are familiar with, (25) thereby reducing dietary variety. However, a physiological effect also plays an important role. Lower dietary variety in old age may be due to a combination of reduced appetite, lower taste acuity and a reduction in the physical ability to prepare food. (26)

Apart from living arrangements and age being associated with dietary variety education was found to be important too. Low education was associated with lower dietary diversity even after controlling for the effect of demographic and economic factors. Prior research, specifically on the elderly, has suggested that the importance of education on food behaviour occurs as the more highly educated are better informed about nutritional health, (27) and are better able to understand nutritional advice. (24) It may also be, as suggested by Tucker, that as education in old age represents an accumulation of advantages over the life course, that those with higher education have the economic resources to act upon nutritional advice. (7)

Apart from demographic factors, economic factors were also important in predicting variety in food purchasing of older Australians. The positive relationship between income and food-purchasing variety is consistent with prior findings. (13) This result is expected, as income is the primary means by which a household could increase the total number of food items that it is able to purchase.

Once controls were added, labour force status was not associated with the DVS. This finding is likely due to the inadequate way in which food consumed away from home is recorded in the. HES. Several studies examining the expenditure behaviour of older Americans have found that, employed people devote a larger portion of income to food away from home. (28,29) Out of the possible 110 food expenditure items in the HES, only two measure food consumed outside of the home, which may explain why food-purchasing variety is not significantly higher for employed older people. A further explanation, offered by McKie's qualitative study is that older people dedicate more time to prepare food in a way that they can manage, in order to maintain control over food preparation and nutrition more generally. (30)

The importance of tenure in explaining dietary diversity is likely related to two effects. First, ownership of a dwelling is a proxy measure of wealth. As shown by Kelly, the value of an older Australians dwelling is a good proxy for the households total wealth. (31) As such, those people who were unable to save enough over the life course to purchase a house have significantly lower levels of wealth. Wealth irrespective of the effect of income is related to the dietary behaviour of elderly people. (32) Tenure may also be related to dietary diversity as owning a home frees up expenditure on rent and mortgages, enabling a greater portion of income to be devoted to food. Lending support to this argument, the present study found that households with the lowest DVS, although spending similar amounts on housing and power, had to devote much larger proportions of their weekly budgets to these costs, when compared with older households with the highest DVS. This provides strong support for Turrell's hypothesis that housing costs are an important component of food expenditure and may therefore influence nutritional status. (17)

CONCLUSIONS

The present study has sought to examine the association between characteristics of older households and dietary variety, as measured by the DVS. Supporting the hypothesis of the present research, both economic and demographic characteristics of an older persons household were found to be associated with purchasing a varied diet.

The finding that poor education and low income are associated with poor dietary diversity indicates the need for early life-course interventions. The potential for improved socioeconomic wellbeing in retirement may well result in improved nutritional outcomes, and thereby help reduce public level health costs. Further, the finding that the oldest-old, people in lone person households and those who live in rented accommodation in later life tend to purchase a less varied diet, suggests that nutrition education programs may help improve the nutrition status of these groups.

Although the present paper has provided interesting findings on the association between household characteristics and variety in food purchasing, further analysis is required on the association between access to the food supply and purchasing a varied diet. For example, appropriate local transport services and distance to shopping centres are likely to be associated with accessing a varied diet. Moreover, the local food supply itself may operate as a constraint upon obtaining a varied diet. Unfortunately, the data set used in this analysis did not contain variables measuring access to the food supply or variables measuring the quality of the food supply.

Despite the shortcomings in the data set used in the present study, the HES provides an important source of data to examine among other things food diversity, food security and food expenditure. The HES is collected every five years by the Australian Bureau of Statistics and includes a comprehensive number of food items coded in the survey. This provides the opportunity to monitor dietary variety and food expenditure among key subgroups of the Australian population cross-sectionaly at five-year intervals.

The ability to maintain access to a varied diet in the population is important because of its association with other nutritional outcomes. As shown in the recently established 'Dietary Guidelines for Older Australians', obtaining adequate dietary diversity is linked with the other guidelines of keeping active and maintaining a healthy body weight, eating at least three meals a day eating plenty of vegetables, cereals, breads, pasta and foods high in calcium, and limiting consumption of foods high in saturated fat, salt and sugars. Consuming a diverse diet and achieving nutritional wellbeing more generally are important components of 'successful aging'. (2)

ACKNOWLEDGEMENTS

The author wishes to thank Dr Don Rowland and two anonymous referees for comments on an earlier draft. The 1998/99 Household Expenditure Survey was made available through the Australian Bureau of Statistics (ABS) and Australian Vice Chancellors Committee (AVCC) agreement. The opinions expressed herein do not represent those of the ABS or AVCC.

REFERENCES

1 Volker D, Caterson I. Nutrition and senescence: healthy ageing. Aust J Ageing 2001; 20: 114-23.

2 Rowe J, Kahn R. Successful Aging. New York: Pantheon Books, 1998.

3 Bennet S, Magnus P, Gibson D. Obesity Trends in Older Australians. Bulletin no. 12. AIHW cat. No. AUS 42. Canberra: AIHW, 2004.

4 Bernstein M, Tucker K, Ryan N et al. Higher dietary variety is associated with better nutritional status in frail elderly people. J Am Diet Assoc 2002; 102: 1096-104.

5 Marshall T, Stumbo P, Warren J, Xie X. Inadequate nutrient intakes are common and are associated with low dietary variety in rural, cmmunity-dwelling elderly. J Nutr 2001; 131: 2192-6.

6 Foote J, Murphy S, Wikens L, Basiotis P, Carlson A. Dietary varity increases the probability of nutrient adequacy among adults. J Nutr 2004; 134: 1779-85.

7 Tucker K, Spiro A, Weiss S. Variation in food and nutrition intakes among men: age and other socio-demographic factors. Nutr Res 1995; 15: 161-76.

8 Kant A, Schatzkin A, Harris T, Ziegler R, Block G. Dietary diversity and subsequent mortality in the first national health and nutrition examination survey epidemiologic follow-up study. Am J Clin Nutr 1993; 57: 434-40.

9 Hsu-Hage B, Wahlqvist M. Food variety of adult Melbourne Chinese: a case study of population in transition. World Rev Nut Diet 1996; 79: 53-69.

10 Wahlqvist M, Lo C, Myers K. Food variety is associated with less macrovascular disease in those with type II diabetes and their healthy controls. J AM Coll Nutr 1989; 8: 515-23.

11 National Health and Medical Research Council (NHMRC). Dietary Guidelines for Older Australians. Canberra: Commonwealth of Australia, 1999.

12 Finke M, Huston S. Healthy index scores and the elderly. Fam Econ Nutr Rev 2003; 15: 67-73.

13 Thiele S, Weiss C. Consumer demand for food diversity: evidence for Germany. Food Policy 2003; 28: 99-115.

14 Drewnowski A, Henderson S, Shore A, Fischler C, Preziosi P, Herberg S. Diet quality and dietary diversity in France: implications for the French paradox. J Am Diet Assoc 1996; 96: 663-9.

15 Lee J, Brown M. Consumer demand for food diversity. Southern J Agr Econ 1989; 21: 47-53.

16 Shonkwiler J, Lee J, Taylor T. An empirical model for the demand for a varied diet. Appl Econ 1987; 19: 1403-10.

17 Turrell G. Structural, material and economic influences on the food-purchasing choices of socioeconomic groups. Aust NZJ Public Health 1996; 20: 611-17.

18 Long J. Regression Models for Categorical and Limited Dependent Variables. Thousand Oaks: Sage Publications, 1997.

19 Lee J. The demand for varied diet with econometric models for count data. Am J Agr Econ 1987; 69: 687-92.

20 Belsley D, Kuh E, Welsch K. Regression Diagnostics: Identifying Influential Data and Sources of Collinearity. New York: Wiley, 1980.

21 Mooney C, Duval R. Bootstrapping: A Nonparametric Approach to Statistical Inference. Newbury Park: Sage Publications, 1993.

22 Lilley J. Food choice in later life. Nutr Food Sci 2004; 34: 4-7.

23 Walker D, Beauchene R. The relationship of loneliness, social isolation and physical health to dietary adequacy of independently living elderly. J Am Diet Assoc 1997; 91: 300-304.

24 Rainey C, Mayo R, Zitlin V, Kemper K, Cason K. Nutritional beliefs, attitudes and practices of elderly, rural, southern women. J Nutr Elderly 2000; 20: 3-27.

25 Horwath C, Worsley A. Dietary supplement use in a randomly selected group of elderly Australians: results from a large nutrition and health survey. J Am Geriatr Soc 1989; 37: 689-96.

26 Johnson A, Donkin A, Morgan K, Neale R, Page R, Silburn R. Fruit and vegetable consumption in later life. Age Ageing 1998; 27: 723-8.

27 Lokken S, Byrd S, Hope K. Assessing nutritional risk and sociodemographic characteristics of low-income older adults living in Mississippi. J Nutr Elderly 2002; 21: 21-37.

28 Wang H, Mok C, Fox J. Food expenditure patterns of elderly consumers: a question time allocation. Family Economics and Resource Management Biennial 1995; 1: 185-91.

29 Rubin R, Nieswiadomy M. Expenditure patterns of retired and nonretired persons. Monthly Labor Rev 1994; 117: 10-21.

30 McKie L. Older people and food: independence, locality and diet. Br Food J 1999; 101: 528-36.

31 Kelly S. Wealth and retirement: latest estimates for Australia. Paper presented at the 9th Annual Colloquium of Superannuation Researchers, University of New South Wales, 2001.

32 Quandt S, Rao P. Hunger and food security among older adults in a rural community. Hum Organ 1999; 58: 28-35.

Jeromey B. TEMPLE

Demography and Sociology Program, Research School of Social Sciences, Institute of Advanced Studies, The Australian National University, Canberra, Australian Capital Territory, Australia

J.B. Temple, PhD candidate, Research Assistant

Correspondence: J.B. Temple, Demography and Sociology Program, Research School of Social Sciences, Institute of Advanced Studies, The Australian National University, Canberra, ACT 0200, Australia.

Email: [e-mail]

Table 1 Mean Dietary Variety Score (DVS) by socio-demographic
characteristics of the sample, Australians aged 55 years and over,
1998-1999

                                n     Mean items purchased (DVS)  SD

Demographic factors
Living arrangement
  Couple only                    878  34.4                         9.8
  Lone female                    507  24.1                         9.2
  Lone male                      219  19.8                         8.7
  Couple with children           212  41.5                        10.4
  Living with others              82  38.7                        10.9
Age (years)
  55-64                          790  33.3                        12.3
  65-74                          680  30.7                        11.4
  75+                            428  26.9                        10.5
Education
  University-educated            173  35.3                        12.2
  Skilled/vocational/diploma     616  33.2                        12.2
  No post-school education      1109  28.9                        11.2
Birth place
  Australia                     1304  31.0                        12.1
  Overseas                       594  30.9                        11.8
Disability or health condition
  Yes                           1328  31.2                        11.6
  No                             570  30.4                        12.4

Economic factors
In labour force
  Yes                            545  35.3                        12.3
  No                            1353  29.2                        11.2
Income per week
  <$196                          487  24.5                        10.5
  $196-354                       472  27.9                        10.2
  $355-690                       471  33.4                         9.9
  $692+                          468  38.1                        11.9
Tenure
  Owner/purchaser               1587  32.1                        11.7
  Renter                         270  24.9                        11.1
  Other                           41  25.6                         8.5

                                Mean difference (MD) (a)  P > |t|

Demographic factors
Living arrangement
  Couple only                    --                       --
  Lone female                   -10.3                     <0.001
  Lone male                     -14.6                     <0.001
  Couple with children            7.2                     <0.001
  Living with others              4.5                      0.001
Age (years)
  55-64                          --                       --
  65-74                          -2.6                     <0.001
  75+                            -6.4                     <0.001
Education
  University-educated            --                       --
  Skilled/vocational/diploma     -2.1                      0.100
  No post-school education       -6.3                     <0.001
Birth place
  Australia                      --                       --
  Overseas                       -0.1                      0.816
Disability or health condition
  Yes                            --                       --
  No                              0.8                      0.178

Economic factors
In labour force
  Yes                            --                       --
  No                             -6.2                     <0.001
Income per week
  <$196                          --                       --
  $196-354                        3.5                     <0.001
  $355-690                        8.9                     <0.001
  $692+                          13.6                     <0.001
Tenure
  Owner/purchaser                --                       --
  Renter                         -7.2                     <0.001
  Other                          -6.5                      0.001

(a) Difference in mean from the base comparison category.
--, Base comparison category.
Source: Australian Bureau of Statistics 1998/99 Household Expenditure
Survey.

Table 2 Negative binomial regression model of factors associated with
Dietary Variety Score, Australians aged 55 years and over, 1998-1999

                                        Coefficient
                                        b (a) (SE) (b)    P > |z|

Demographic factors
Age (years)
  55-64                                   --              --
  65-74                                   -0.031 (0.020)   0.135
  75+                                     -0.069 (0.025)   0.006
Living arrangement
  Couple only                             --              --
  Lone female                             -0.285 (0.023)  <0.001
  Lone male                               -0.516 (0.028)  <0.001
  Couple with co-resident kin              0.119 (0.026)  <0.001
  Living with others                       0.119 (0.038)   0.002
Education
  University education                    --              --
  Vocational education                    -0.030 (0.029)   0.310
  School education only                   -0.081 (0.026)   0.004
Birth place
  Australia                               --              --
  Overseas                                 0.010 (0.017)   0.555
Person in household with
  disability or health condition?
  No                                      --              --
  Yes                                      0.017 (0.018)   0.351

Economic factors
  Log of weekly income                     0.046 (0.013)  <0.001
  Owner/purchaser                         --              --
  Renter                                  -0.075 (0.025)   0.003
  Other                                   -0.152 (0.055)   0.006
  Region measure                           0.005 (0.003)   0.046
In labour force
  Yes                                     --              --
  No                                       0.004 (0.023)   0.859
Constant term                              3.295 (0.091)
Alpha (d)                                  0.075 (0.004)
Likelihood-ratio test of alpha = 0 (d)  1475.740          <0.001
Maximum likelihood [r.sup.2]               0.348

                                        Incidence rate
                                        ratio (c) (95% CI)

Demographic factors
Age (years)
  55-64                                 --
  65-74                                  0.969 (0.930, 1.009)
  75+                                    0.933 (0.889. 0.980)
Living arrangement
  Couple only                           --
  Lone female                            0.752 (0.719, 0.786)
  Lone male                              0.597 (0.564, 0.631)
  Couple with co-resident kin            1.127 (1.070, 1.186)
  Living with others                     1.127 (1.045, 1.216)
Education
  University education                  --
  Vocational education                   0.971 (0.917, 1.028)
  School education only                  0.922 (0.872, 0.975)
Birth place
  Australia                             --
  Overseas                               1.010 (0.977, 1.045)
Person in household with
  disability or health condition?
  No                                    --
  Yes                                    1.017 (0.982, 1.054)

Economic factors
  Log of weekly income                   1.047 (1.021, 1.074)
  Owner/purchaser                       --
  Renter                                 0.928 (0.883, 0.975)
  Other                                  0.859 (0.771, 0.957)
  Region measure                         1.005 (1.000, 1.011)
In labour force
  Yes                                   --
  No                                     1.004 (0.960, 1.050)
Constant term
Alpha (d)
Likelihood-ratio test of alpha = 0 (d)
Maximum likelihood [r.sup.2]

(a) Negative binomial regression coefficient.
(b) Standard error of b.
(c) Calculated as the exponent of b.
(d) Tests for unobserved heterogeneity. The significant test statistic
indicates that the negative binomial specification is preferable to a
Poisson specification.
--, Comparison category.
Source: Australian Bureau of Statistics 1998/99 Household Expenditure
Survey.

Table 3 Median housing and power expenditure by the Dietary Variety
Score, Australians aged 55 years and over, 1998-1999

                           Median weekly expenditure per person (95% CI)
Dietary Variety Score      0-25                   26-36

n =                        635 (33.4%)            622 (32.8%)
Housing costs ($) (a)       31.77 (29.68, 34.67)   30.53 (28.87, 32.54)
Housing costs (share) (b)   13.65 (12.21, 15.15)    8.84 (8.41, 9.27)
Power costs ($) (c)         10.82 (10.40, 11.66)   13.04 (12.50, 13.58)
Power costs (share) (d)      4.41 (4.20, 4.68)      3.56 (3.28, 3.76)
Total food ($) (e)          35.19 (34.21, 36.58)   49.35 (47.91, 50.56)

                           Median weekly expenditure per person (95% CI)
Dietary Variety Score      37+

n =                        641 (33.8%)
Housing costs ($) (a)       37.50 (36.16, 39.97)
Housing costs (share) (b)    6.45 (6.18, 7.03)
Power costs ($) (c)         17.64 (16.95, 18.64)
Power costs (share) (d)      2.82 (2.67, 2.98)
Total food ($) (e)          62.08 (60.62, 64.99)

(a) Weekly expenditure on rent, mortgages, rates, insurance and repairs.
(b) The share of housing costs out of total weekly expenditure on all
goods and services.
(c) Weekly expenditure on electricity, gas, oil and other household
fuels.
(d) The share of power costs out of total weekly expenditure on all
goods and services.
(e) Weekly expenditure per person on all 110 food items recorded in the
Household Expenditure Survey.
Source: Australian Bureau of Statistics 1998/99 Household Expenditure
Survey.