by Professor Peter Saunders

The Australian Institute of Family Studies is frequently asked to provide figures on what it costs Australian families to raise their children. The answers we give can have a big impact on people's lives, for they feed into the formulation of social policies designed to support families with children, and they are also used by lawyers and Family Court judges to determine how much absent parents need to pay to help maintain their children.

The significance attaching to these figures means that there is a heavy responsibility on anybody calculating the costs of children to make sure they get it right. The problem, however, is that any set of calculations inevitably reflects a set of starting assumptions, and these assumptions are inherently debatable. In short, there can be no single, authoritative answer to the question of what it costs to raise children in Australia today.

Problems in calculating the costs of children

For some years, the Institute published in its magazine Family Matters two sets of figures, regularly updated, which were based on two different approaches to calculating the costs of children.

Because of these problems, the Institute decided in 1999 to stop updating and publishing costs of children figures. At the same time, however, we also invited researchers based at three eminent institutions to write fresh articles for Family Matters explaining how they thought the costs of children should be calculated, and setting out the estimates which followed from the approach they had adopted.

This Guide reproduces these three recent articles, as well as an earlier paper explaining the two original approaches, so that readers can make their own evaluations of the different approaches and can, if they wish, update the estimates contained in each.

The original 'basket of goods' and 'survey expenditure' approaches

The two approaches which underpinned the figures which appeared regularly in Family Matters until 1999 are outlined and explained in the article we reproduce here by Peter McDonald, which was originally published in Family Matters in November 1990. This article distinguishes the 'basket of goods' approach, developed by Kerry Lovering in 1983, and the 'expenditure survey' approach, developed by Donald Lee in 1989. We also include an explanation of how each set of figures can be updated.

Lovering drew up two lists of what she thought parents needed to buy in order to raise children of varying ages, and she then worked out how much the items on each list would cost. One list was said to be 'basic' and was assumed to apply to low income families; the other was more generous and applied to middle income families. Neither list included items shared by the whole family, such as cars and housing, and both were inevitably arbitrary, for who is to judge what is necessary and what is not?

Lee's alternative approach was to use the Household Expenditure Survey to determine what parents in different income brackets, raising children of different ages, actually spent. His estimates were much higher than Lovering's (for example, at 1990 prices, Lovering calculated the weekly costs of a two year-old child to a middle income family as $40.91 while Lee put it at $136.06). This partly reflects the fact that Lee included factors like medical and educational costs omitted by Lovering.

Three later alternatives

This Guide also reproduces the three recent articles published in Family Matters, in August 1999 and December 1999.

The article by Peter Saunders, of the Social Policy Research Centre, at the University of New South Wales, is essentially a revised basket-of-goods approach (he calls it a 'budget standards' method) which takes account of the cost of the children's imputed share of family-owned items, such as a car. His estimates come out higher than Lovering's (but still lower than Lee's), and he recommends that his calculations should replace the older Lovering figures.

The article by Rebecca Valenzuela, of the Melbourne Institute of Applied Economic and Social Research, at the University of Melbourne, is, by contrast, a revised 'expenditure survey' approach in which she calculates for each additional child the percentage increase in spending which a family must incur in order to maintain the standard of living it enjoyed before the children were born. This approach produces relatively low estimates of costs, particularly as regards the additional costs incurred as a result of having a third, or subsequent, child.

Finally and most recently, the piece by Ann Harding and Richard Percival, of the National Centre for Social and Economic Modelling, at the University of Canberra, is also based on expenditure survey data, but it tries to compare the actual level of spending of families at the same standard of living but with different numbers of children. Their estimates come out higher than Valenzuela's, but are still considerably lower than those derived from basket-of-goods approaches.

Comparing approaches

The Australian Institute of Family Studies makes no judgement as between the relative merits of these different approaches and the calculations derived from them. These estimates vary widely from each other, and it is for readers to judge how useful or appropriate any one of these approaches might be.

by Ann Harding and Richard Percival

How much do children actually cost? Many people would like an accurate answer to this question! The answer is important for courts considering child support rules and for governments setting payments to foster parents and to families with children. The answer is also important to both parents and potential parents, as they try to assess the impact of children upon their financial position.

But unfortunately, as in so many other areas of economics, there is no single 'right' answer to this question. How much children are estimated to cost is in part a function of the method used to estimate those costs.

In the last issue of Family Matters, two different estimates of the costs of children were presented, based on two different methodologies (see Three Approaches to Measuring the Costs of Children, below). This article presents a third set of estimates, based upon yet another methodology.

Why is it so difficult to work out how much children cost? It seems easy to compare the weekly spending of a couple with one child and a couple without children and assume that the difference between their total spending represents the costs of the child.

But there are at least two problems here. First, what proportion of the spending of the couple with a child on 'indivisible' goods - housing, cars, refrigerators - should actually be attributed to the child? And, second, the total spending of the couple with a child doesn't tell us about the financial sacrifices that the couple are making for their child - and thus about what the child is really costing them.

For example, before they had their first child, a couple might have been spending a significant amount each week on fine wines and restaurant meals. Now that they have a baby, the same couple are likely to have reluctantly abandoned their wine cellar for disposable nappies and baby foods, while leisurely restaurant meals have become a faint memory! So if we simply compare the total spending of this couple before and after they had their baby, we will reach entirely the wrong conclusions about how much the child is actually costing them each week.

How can we measure the costs of children?

Economists have devoted many hours to trying to get around these problems. One ingenious method, a variant of which is used in this study, is to compare the total spending of two couples that are believed have the same standard of living but differ because one couple has a child and the other does not. So how then does one measure 'standard of living'?

Many studies of the costs of children have used the Engel estimator to gauge 'standard of living', as first proposed by Ernst Engel more than one hundred years ago. The central concept is simple: as family income falls, a family devotes a greater proportion of its total weekly spending to food. So Engel suggested that the proportion of a family's total spending that was devoted to food could be considered a reliable proxy for a family's standard of living. Thus, suppose we had a couple without children who were spending $500 a week and devoting 30 per cent of this to food. And then we found an otherwise similar couple who had one child and were also spending 30 per cent of their total weekly expenditure upon food, but whose total weekly spending was $600 a week. Then Engel suggested that this would mean that the cost of the child was $100 a week ($600 minus $500).

This method has an intuitive appeal. In the real world around us it seems clear that rich families spend less of their budget on food than poor families, so that the proportion of total spending devoted to food might appear a reasonable and practical way of working out a family's standard of living.

But the method has been extensively criticised for over-stating the real costs of children, because there are fewer economies of scale in food consumption than in many of the other goods and services that children consume. For example, it is not possible to feed to a second child a hamburger already consumed by a first child! But it is possible for a second child to ride a bicycle that a first child has outgrown.

As a result, subsequent studies have often extended beyond food and used a wider basket of goods and services as a proxy for the standard of living of a family. And that is what we have done here. So our indicator of the standard of living of a family is the proportion of total expenditure devoted to:

• food at home;
• fuel and power;
• household non-durables for use inside the home (for example. disposable nappies!) ;
• postal, telephone and telegram charges;
• and personal care products and services (for example, shampoo).

To estimate the total expenditure and standard of living in different households, we used a sample of 2658 households from the 1993-94 Household Expenditure Survey unit record tape issued by the Australian Bureau of Statistics. To improve the comparability of the results for couples with and without children, we removed couples where the spouse was not aged between 25 and 54 years. Thus, we did not compare the spending of an age pensioner couple with a working age couple with a young child. And we also excluded other couples that were significantly different from the norm (for example, the self-employed).

Costs of children by age

So how much do children cost? Table 1 shows the estimated costs by age of the child and family income. The bottom quintile income of $410 represents the average income of the poorest 20 per cent of couples with children in our sample, ranked by the gross income of the family.

As expected, the direct costs of children increased with the age of the child and with the level of family income. The lowest direct costs of $52 a week were estimated for children aged between zero and four years living in families in the bottom quintile (with average gross incomes of $410 a week). The highest costs of $280 a week were estimated for 15 to 17 year old children living in top quintile families. Wealthier families were found to spend a greater amount on their children, whatever their ages. Generally speaking, families in the top quintile spent more than twice as much on their children as families in the bottom quintile.

Source: Percival and Harding, 1999.

While there was a steady rise in the cost of a child as family income increased, when total costs were considered as a proportion of family income, there was a fall as incomes rose. However, the rate of decline was reduced as family income rose (Figure 1).

For families, the costs of a child as a proportion of their combined income ranged between a little over 9 per cent (for a child aged 0 to 4 years in a top quintile family with a gross income of $1700) to 31 per cent (for a child aged 15 to 17 years in a bottom quintile family with an income of $410 per week).

Interestingly, as Figure 1 illustrates, there is much less variation in the direct costs of children by quintile for older children which, in turn, means that the proportion of family income devoted to spending on older children falls sharply as family income increases. Top quintile families with a child aged 15 to 17 spent only a little more than twice as much as bottom quintile families on comparable children. As a result, expenditure on older children aged 15 to 17 years fell sharply from 31 per cent of gross family income for bottom quintile families to 17 per cent of income for top quintile families.

In contrast, top quintile families with a young child aged 0-4 were estimated to spend about three times as much on their child as bottom quintile families (Table 1). As a result, expenditure on younger children as a percentage of income showed relatively little variation between the quintiles, ranging from 9 to 13 per cent of gross income.

Figure 1 Estimated average costs of children as a proportion of total family income, by age of child and gross family income quintile, Australia 1993-94

Costs of children by number of children in household

What if we look at the costs of children by the number of children in the family, rather than the age of the children? Again the cost of each child was found to rise with family incomes (Table 2). Bottom quintile families with one child were estimated to spend $70 a week on that child, while top quintile families with one child spent $191 a week.

Source: Percival and Harding, 1999.

As Figure 2 shows, the cost of a single child amounted on average to between 11 and 17 per cent of family income, for two children 20 to 33 per cent of family income and, for three children, about 27 to 48 per cent. These results contrast with those recorded by Valenzuela (1999) using the same 1993-94 Household Expenditure Survey data, but using the extended linear expenditure system. Valenzuela found that the estimated cost ratios appeared stable across the income distribution - for example, she estimated that spending on the first child amounted to a uniform 18 per cent of gross family income and that spending on three children remained at about 34 per cent of income irrespective of income level. In contrast, we have found much greater variation in child costs as a percentage of gross income for those at different points within the income spectrum (Figure 2).

Figure 2. Estimated average costs of children as a proportion of income, by number of children and gross family income quintile, Australia, 1993-94

Our results suggested that the marginal cost of a child decreased as the number of children increased. Thus, the cost of the first child was the greatest, ranging from $70 a week for low income families to $191 a week for top quintile families (Table 3). The cost of the second child was lower, ranging between $66 and $147 a week for bottom and top quintile families. And the cost of the third child was lower again. For example, for middle income couple families with average incomes of $810 a week, the first child was estimated to cost $110 a week, the second $94, and the third child only $78 a week. This reduction in the average cost of each additional child is a result of both the expenditure constraints and the economies of scale that families experience as their size increases.

Comparison of costs with other studies

How do our results compare with those produced using the other two methods and published in the last issue of Family Matters? Table 4 attempts to compare the results that the three methods would achieve for low and modest income families in 1993-94, with weekly expenditure at the levels shown in the Table. It should be noted that this is only an indicative comparison. For example, the budget standards were originally estimated for February 1997 and we have simply deflated the levels back to 1993-94 dollars using the Consumer Price Index, in order to compare the results with the 1993-94 estimates generated by the other two methods. Also note that our earlier results were presented for families at a given income level. In contrast, the results below are for families at a given expenditure level.

* The budget standards estimate for one child was derived by averaging the costs for a girl aged 6 and a boy aged 14, while the comparable cost using the Iso-Prop methodology was derived by averaging the costs for a child aged 5 to 9 and a child aged 10 to 14. For two and three child couples the age of the child using the budget standards methodology is as shown in Saunders (1999, p.69), while in the Iso-Prop methodology the children falling within the same age ranges as the budget standards estimates were calculated. The ELES estimates do not appear to vary by the age of the child.

The results suggest that our estimates lie between the budget standards estimates (Saunders 1999) and Valenzuela's (1999) ELES estimates (Figure 3 and Table 4). In particular, both our estimates and those produced using the budget standards methodology suggest much higher costs for two and three child families than does the ELES method. The ELES estimates suggest, for example, that a second child costs $42 a week for a low income family. Both our estimates and those of the budget standards approach put the cost of that second child at about $105 a week.

Figure 3. Indicative Comparison of Costs of Children Using Three Different Methods for Low Income Families, 1993-94

For families living at what the budget standards methodology terms a 'modest but adequate' standard of living, the estimates of the costs of children using the ELES method are substantially lower than those derived from the other two methods. For example, the ELES method suggests that three children cost their parents $227 a week, out of a total family expenditure of $895 a week (Table 4). This means that the three children absorb only 25 per cent of total family expenditure, with the remaining 75 per cent presumably being consumed by the parents. In contrast, our results suggest that these three children cost their parents $378 a week on average, while the budget standards approach puts the cost at $425 a week. Once again, therefore, the results using our Iso-Prop method fall between the other two methods, but are generally closer to the budget standards estimates.

While the ELES methodology indicates that a low income couple with three children require only 33 per cent more income than a couple without children to achieve the same material standard of living, our results suggest that the same three-child couple requires 65 per cent more income than the couple without children. The budget standards approach suggests that a low income couple with three children requires 73 per cent more income than a couple with no children to achieve the same standard of living (Table 5).

* Note: A couple without children has an equivalence scale value of 1.0.

How do these equivalence scale estimates compare with those implicit within the Australian cash transfer system for low income families with children? For this comparison, we have used children of the same age as the budget standards methodology (which is important because family payments vary with the age of the child). The Australian income support system in July 1999 paid a couple with one child 19 per cent more income than a couple with no children. Couples with two children were paid 39 per cent more, and couples with three children 61 per cent more than couples without children dependent upon social security. As a comparison with Table 5 shows, these implicit social security equivalence scales are more generous than those Valenzuela's ELES method suggests are needed. And they are less generous than those suggested by the budget standards method. But they are reasonably close to the equivalences suggested by our Iso-Prop estimates, although still a little lower.

Concluding comments

In this study we have estimated the direct costs of children in Australian two parent families, with the costs of children being defined as parental expenditures on children up to 17 years of age. The level of expenditure was determined by comparing the expenditures of couple families with and without children at the same 'material standard of living'. The measure of the material standard of living was the proportion of total expenditure spent upon a basket of goods that included food consumed at home and fuel and power. It is important to note that no account was taken of the indirect costs of children (for example, the forgone earnings due to mothers working part time rather than full-time).

We found that younger children cost less than older children, and that high income families spent more on their children each week than low income families.

But when we looked at spending as a proportion of total family income, then parental spending on children declined as family income increased.

Our estimates were lower than those derived using the budget standards methodology in the previous issue of Family Matters. But, especially for two and three child families, they were markedly higher than those derived using the Extended Linear Expenditure System. The ELES estimates suggested that in a modest income family consisting of two parents and three children, only 25 per cent of total family expenditure would be devoted to the children's needs and the remaining 75 per cent would be devoted to the parents. Parents will have to decide for themselves whether that seems a realistic proportion.

This just underlines the important conclusion that there is no 'right' answer about how much children actually cost their parents. In large part, the methodology used to answer this question affects the estimates achieved.

References

• Merz, J., Garner, T., Smeeding, T., Faik, J. & Johnson, D. (1993), 'Two scales, one methodology: Expenditure Equivalence Scales for the United States and Germany', All-University Gerontology Center, Syracuse University, Syracuse, New York.
• McHugh, M. (1999), 'The costs of children: Budget Standards Estimates and the Child Support Scheme', Discussion Paper No 103, Social Policy Research Centre, University of NSW, Sydney.
• Percival, R., Harding, A. & McDonald, P. (1999), 'Estimates of the costs of children in Australian Families 1993-94', Report prepared for the Department of Family and Community Services, National Centre for Social and Economic Modelling, Canberra, May.
• Percival, R. & Harding, A. (1999), 'The public and private costs of children in Australia, 1993-94', Discussion Paper, National Centre for Social and Economic Modelling, University of Canberra (forthcoming).
• Saunders, P. (1999), 'Budget Standards and the costs of children', Family Matters, no 53, Winter, pp. 62-70.
• Valenzuela, M.R. (1999), 'Costs of children in Australian households', Family Matters, no 53, Winter, pp. 71-76.
• Whiteford, P. (1985), A Family's Needs: Equivalence Scales, Poverty and Social Security, Research Paper no. 27, Development Division, Department of Social Security, Canberra.

Three Approaches to Measuring the Costs of Children

The last issue of Family Matters contained two separate articles on the costs of children.

The first article, by Peter Saunders (1999), of the Social Policy Research Centre, used the budget standards method to estimate the costs of children. A budget standard is calculated by specifying what is needed (in terms of the goods and services that contribute to material consumption) by particular households living in a particular place at a particular time in order to achieve a specified standard of living. After each item has been identified, it is then costed and summed to arrive at the total budget required to reach the given standard.

This method thus involves experts trying to identify what children need, rather than what parents actually spend on their children. Generally, the budget standards method appears to result in higher estimated costs of children than many other methods.

The second article, by Rebecca Valenzuela (1999), of the Melbourne Institute of Applied Economic and Social Research, used a method called the Extended Linear Expenditure System (ELES) - sometimes also known as the Barten-Gorman method. This method involves estimating a utility function and demand equations from a sample survey that contains details of the expenditures of families - in this case, the 1993-94 Household Expenditure Survey conducted by the Australian Bureau of Statistics. The equations attempt to estimate how much parents actually spend on their children, and families with the same level of utility are assumed to be equally well-off.

International research suggests that the estimates of the costs of children produced by the ELES method tend to be relatively low, especially for third and subsequent children. For example, some overseas studies using this method have found that four children cost their parents less than three children (Merz et.al 1993). Valenzuela also found that the marginal costs of the second and third child were quite low. She also concluded that parents devoted the same proportion of their income to their children, irrespective of whether they were rich or poor.

This study uses a third method, often termed the Iso-Prop method. Like the ELES method, this method also involves using the 1993-94 Household Expenditure Survey and econometrics to estimate how much children cost their parents. But instead of assuming that families with the same estimated level of utility have the same standard of living, this method assumes that families that devote the same proportion of their total expenditure to a specified basket of basic goods and services have the same standard of living.

This method appears to produce estimates that are lower than the budget standards estimates but higher than the ELES estimates. The methodology is described in more detail in this article (and in Percival, Harding and McDonald 1999).

There are two important differences between results obtained by using the basket-of- goods method and the expenditure survey method as presented in the accompanying Tables. First, the basket-of-goods approach provides only part of the cost of a child, while the expenditure survey measures the total amount spent on the child. Second, the basket-of-goods method indicates how much parents would spend on their children if the child was to enjoy the fruits of the basket specified by the researcher. In this sense, it pro- vides an ‘ideal’ or desirable costing. In contrast, the expenditure survey approach indicates how much parents actually spend on their children, even though the amount spent might be considered inadequate or excessive by the objective standards of the basket-of-goods method. For a full description of the two approaches see McDonald, P. (1990), ‘The costs of children: a review of methods and results’, Family Matters, no. 27, pp. 18–22.

Both sets of estimates are updated using Australian Bureau of Statistics figures contained in Australian Economic Indicators, ABS Catalogue No. 1350.0. The Lovering figures are updated by using the Consumer Price Index (weighted average of eight capital cities); the most recent CPI figure is divided by the previous quarter’s figure, then multiplied by the Lovering figures. The Lee figures are updated by using the Average Weekly Earnings (all males, total earnings); the most recent AWE figure is divided by the previous quarter’s figure, then mul- tiplied by the Lee figures.

Update: Basket-of-Goods Approach Based on Lovering 1983

Note: Included are food and clothing, fuel, household provisions, costs of schooling (not fees), gifts, pocket money and entertainment. NOT included are housing, transport, school fees or uniforms, child care, medical or dental expenses. Holidays are a component of the middle income figures only.
Source: Lovering, K. (1984), Cost of Children in Australia, Working Paper no.8, Australian Institute of Family Studies, Melbourne.

Update: Expenditure Survey Approach Based on Lee 1989

Notes: * Includes medical and dental costs, education costs and other miscellaneous costs. Costs of children vary according to the number of children in the family, the parents' incomes and whether one or both parents are working. Note:The figures in the table relate to a one-child, one-income family with an income of $722.80 gross per week. The Lee data show that two children cost about 55 per cent more than one child, while three children cost about twice the cost of one child. The dollar costs of children are relatively 'flat' compared with rises in family income: children in poor families cost proportionally more, and children in rich families proportionally less than those in middle income families.
Source: Lee, D. (1989), Calculations of the direct costs of children based on the 1984 ABS Household Expenditure Survey, Australian Institute of Family Studies, Melbourne.

by Peter Saunders

The costs of children is a topic of widespread community interest among policy makers and those involved in a broad range of public programs and family welfare issues. It is also of immediate interest to parents, who have to bear the bulk of these costs, and to couples considering when (or whether) to start a family.

A broad range of government programs provide assistance to those who have to pay the costs of children, including family assistance, child support and foster care allowances. If the benefits provided under these programs are to be adequate, they must reflect what it actually costs to raise children, and research is therefore required to identify these costs and estimate their magnitude.

The significance of the costs of children has emerged in the recent debate about tax reform and the GST, where it was argued before the Senate Tax Committee Hearings that the government's original compensation package treated larger families less generously than smaller families and was thus both inadequate and inequitable. The cost of children - or at least that component which parents have to pay - has also been highlighted as a possible factor underlying the decline in the fertility rate in the debate over population ageing and population policy more generally (McDonald 1997). Clearly, there is an ongoing need to ensure that the best possible estimates of the costs of children are available so as to inform public debate and policy formulation on these important issues.

In Australia, the Australian Institute of Family Studies has played a leading role in undertaking and commissioning research on the costs of children, and in updating and publishing the cost estimates in Family Matters, thereby bringing the significance of the issues to the public's attention.

Two sets of estimates of the costs of children have been maintained by the Institute. The first was originally derived by Institute researcher Kerry Lovering in the 1980s and have been updated since then by movements in consumer prices (Lovering 1984). The Lovering study was the first attempt to bring together in an integrated framework a range of Australian evidence and research on the costs of children. It adopted a limited variant of the 'basket of goods' (or what is now called the budget standards) approach in which the goods and services needed by children in different families are identified and costed.

The second set of estimates was derived by Donald Lee in a research study commissioned by the Institute in 1989, partly in response to concerns over the incomplete coverage of the Lovering estimates. Building on research undertaken in the United States (Espenshade 1984), the method developed by Lee used household expenditure data to estimate the costs of children from comparisons of the expenditures of otherwise similar households with and without children. (Research recently commissioned by the Department of Family and Community Services is reviewing the advantages and limitations of this second approach and applying it to more recent expenditure data for 1993-94).

Neither of these two approaches are without its critics - the former because of its incomplete coverage of costs, and the latter because of the difficulty of deducing what a child actually costs from data that describes what households choose to spend on their children. These limitations were described in detail in an article published some time ago in Family Matters by Peter McDonald, the Institute's Deputy Director at the time (McDonald 1990:18-22). The article began by noting that the Institute receives many inquiries about the costs of children, but went on to note that while the 'expectation of most of those who inquire is that the cost of a child is a fairly straight-forward matter . . . this is far from being the case; estimating the cost of children is a complex and highly imprecise exercise. Therefore, estimates of the cost of children should be used with caution' (italics added).

In response to a concern that the limitations of the cost estimates published by the Institute were making them increasingly inaccurate (possibly even misleading) as the length of time since they were originally derived increased, the Institute announced in the Autumn 1995 issue of Family Matters that it was planning to convene a small expert group to consider whether new methodologies and data should be used to produce a revised set of estimates of the costs of children. However, the workshop was postponed following the announcement in August 1995 by the Minister for Social Security that the Social Policy Research Centre (SPRC) had been commissioned to develop a set of indicative budget standards for a range of households that would, among other things, 'examine the costs of children in different family circumstances'.

This article describes that research and its findings in relation to the costs of children, and explains current research being undertaken within the SPRC to extend what has already been done.

Budget standards

The budget standards method dates back to the turn of the century to when British social researcher Seebohm Rowntree first attempted to estimate the extent of poverty in England by identifying what different families needed to 'maintain merely physical efficiency' and then costing a basket of goods that was adequate to meet those needs (Rowntree 1901; Saunders 1998a). The approach has often been referred to as the 'basket of goods method' and is in fact the basis of the research undertaken by Kerry Lovering.

After Rowntree's studies, budget standards research spent a long time in the doldrums, but it has recently generated renewed interest - partly because it provides a framework for identifying needs as a basis for deriving a poverty line and thus estimating poverty, but also because the availability of better data and more sophisticated methods for analysing them have made the application of the method more of a practical reality. It has also become clear that the method has wider applicability than just to the identification of minimum needs and poverty. Indeed, the method can, in principle, be used to develop budgets that correspond to any standard of living (Bradshaw 1993).

A budget standard is derived by specifying what is needed, in terms of the goods and services that contribute to material consumption, by particular households living in a particular place at a particular time in order to achieve a specified standard of living. Having identified each item, they are then costed and summed to arrive at the total budget required to reach the standard. In the SPRC study, budgets were developed for a range of different households at two separate standards - a modest but adequate standard and a low cost standard (Saunders, Chalmers, McHugh, Murray, Bittman and Bradbury 1998).

The modest but adequate standard is one which affords full opportunity to participate in contemporary Australian society and the basic options it offers, lying between the standards of survival and decency and those of luxury. It attempts to describe the situation of households whose standard of living falls somewhere around the median standard experienced in the Australian community as a whole.

The low cost standard, in contrast, is seen as one which may require frugal and careful management of resources but still allow social and economic participation consistent with community standards, and enable the individual to fulfil community expectations in the workplace, home and in the community. Whilst not seen as a minimum standard, the low cost standard is one below which it would become increasingly difficult to maintain an acceptable standard of living because of the increased risk of deprivation and disadvantage. (For a comparison between the low cost budget standards and the Henderson poverty line, see Saunders 1998b.)

The definition of each of these two standards have emerged over the last two decades from research on budget standards conducted in an increasing number of countries. Most countries - certainly most industrial countries - that have developed their own budget standards have adopted one or both of the low cost and modest but adequate concepts, which originated in the post-war years in the United States and were later given prominence by the Expert Committee on Family Budget Revisions (Watts 1980). They have been used as the basis for developing budget standards in Canada and a number of European countries, including Denmark, Norway, Sweden and the United Kingdom, as well as in countries nearer to home such as Hong Kong and Malaysia. Much of the practical experience gained in budget standards research undertaken in these countries proved invaluable in guiding the SPRC study.

Having articulated in words a standard of living (low cost or modest but adequate), the main research task is to identify the goods and services required to achieve that standard and then cost them. This is a complex, time consuming and tedious process that has been described by Jonathan Bradshaw (1993) as 'a ghastly chore' - a characterisation that is probably somewhat on the generous side! It involves identifying each and every item that a family needs to attain and maintain a particular standard of living - everything from the 'big ticket' items like the house, car and furniture to small items like toothpaste, teaspoons and toilet paper.

It is worth noting at this stage that, in practice, some of the most costly items in fact add rather little to the calculated budgets. This reflects the fact that many expensive items last for a long time and their cost is averaged over the assumed lifetime in deriving their contribution to the weekly household budget. For example, the low cost budget standard for a couple with two children (a 6-year-old girl and a 14-year-old boy) includes a refrigerator costing $1429. However, the fridge is assumed to last for 15 years, so that its contribution to the household's weekly budget is $1.83 ($1429 divided by the number of weeks in 15 years). This works out to be considerably less than $5.87 that the family spends on milk each week. Similarly, although the family is assumed to own a car, it is a 12 year-old Corolla and the depreciation cost of the vehicle itself is well below the weekly cost of petrol and insurance, and even lower than the cost of registration (Saunders et al. 1998, Appendix 12.A).

In determining which items to include in the budgets, a combination of both normative and behavioural criteria were used. The starting point, where they exist was a set of normative principles which specify the needs of households as determined by experts in each budget area. Some of these norms have an official or quasi-official status, being endorsed by the relevant bodies. The food budgets, for example, were developed from the nutritional guidelines of the National Health and Medical Research Council (NH& MRC), although these were modified to reflect actual Australian dietary patterns.

There were many areas where there are no normative standards from where to begin and here greater reliance was made on behavioural data - what families actually do. This makes the budgets better grounded in the reality of everyday experience and custom, but at the same time weakens the claim that the standards represent a measure of need that is independent of the resources available to meet those needs. The budget standards thus reflect a myriad of judgements and assumptions, many of them a compromise between the ideal and the practical. As a consequence, the budgets reflect a series of informed judgments regarding which items correspond to each standard, their quality and price, and how long they are assumed to last.

To help us meet the challenge this posed, the SPRC research benefited from the views provided by a Steering Committee of relevant experts in many fields including nutrition, health economics, housing, clothing needs, consumer behaviour and household budgeting. The Committee included individuals from the Australian Bureau of Statistics, the Australian Consumers Association, the Australian Institute of Health and Welfare, the Australian Institute of Family Studies, CSIRO, the Department of Social Security, ACOSS, and the Brotherhood of St Laurence. Preliminary budget estimates were discussed by a series of focus groups who provided valuable advice on where revisions were necessary, and comparisons were made with data on actual expenditure patterns using data from the latest Household Expenditure Survey to ensure that they did not deviate substantially from actual spending patterns.

The standards are thus not perfect, as this is an unobtainable objective. What can be claimed for them is that they reflect the best of the existing Australian research, expertise, data and judgments in each budget area and to this extent they are at the leading edge of what is currently possible. There is certainly scope for further refinement of the standards as new data or methods come to light, or as those who study them are prompted to think of better ways to develop them. While the budget standards are only indicative, they are still far ahead of anything else that has been developed along similar lines for Australia.

Budgets were estimated at either the low cost or modest but adequate standards for a range of household types that differed according to household size, the age and sex of children, housing tenure and the labour force status of the parent(s). A total of 46 separate budgets were derived. The magnitude of the task can be gauged from the fact that just one of these budgets (the low cost budget for a couple with two children) contains no less than 840 separate items - each of which had to be identified, priced and assigned a lifetime. The budgets were costed in Sydney (which is very important to remember when considering the housing budgets), using prices existing in major retail stores in February 1997.

The costs of children

Since a budget standard was derived for couple households with no, one, two, three and four children, and for sole parent households with one and two children, they provide a basis for estimating the costs of children.

There is no single agreed method for doing this, although three main approaches have been identified in the literature (Oldfield 1993). The methods differ primarily in the way they attempt to treat the cost of items that are consumed by the household as a whole rather than by any specific individuals (adults or children) within it. While it is relatively straightforward to decide which items of clothing or food are consumed by a child, for example, this is not so easy when it comes to the cost of housing, transport, consumer durables or furniture. What portion of the cost of a fridge should be assigned to the child or children who use it, for example?

The first approach to the treatment of these shared household costs is to ignore them. This results in a method which concentrates solely on those costs that can clearly be ascribed to children such as their food, clothing, health, personal care, schooling and leisure needs. This approach (sometimes called the individualised method) was adopted in the Lovering study, and while it produces only partial estimates of the total cost of the child, it has the advantage that the uncertainty surrounding the size of each cost components is kept to a minimum.

The second approach extends the first by also including estimates of the portion of the cost of each shared item the fridge, for example, that can be assigned to the child(ren). This method requires normative judgements to be made when allocating shared costs to individuals and is thus called the normative method. It has the advantage of being comprehensive in coverage, but is also difficult and time-consuming to apply in practice and because it requires judgements, will always be open to criticism.

The third method, called the difference method, estimates the costs of children by taking the difference in the budget standards for households with and without children, or with different numbers of children. Thus, for example, the cost of the first child is estimated by taking the difference between the budgets of a couple with one child and a couple with no children (or as the difference between a sole parent with one child and a single person). The great advantage of the difference method is simplicity; once the budgets have been derived it is straightforward to take differences to estimate the costs of children. The difference method also has the advantage that it does not include as part of the cost of the child those large cost components that would have been present even if there had been no children in the household.

Consider the case of housing. A couple (or sole parent) with one child requires one more bedroom than a couple without children (or a single adult) and they must thus incur higher housing costs. However, it is only the additional housing costs that relate to the presence of the child and only these should be included in the cost of the child. This is precisely what the difference method does. Similarly, in the case of the fridge referred to earlier, it is only the additional cost of the fridge in the larger family (due either to the fridge itself being bigger and hence more expensive, or to the fact that increased usage shortens its lifetime and thus raises the weekly cost) that enter into the calculation of the cost of the child. If the same fridge will suffice (without adversely affecting the family's standard of living) then the 'fridge costs' of the child are equal to zero.

However, there are other features of the difference method that are somewhat less attractive. One of these is that some of the cost differences may reflect changes in the costs of household members other than children. Here, the reliance on behavioural data can give rise to problems. If, for example, mothers with one child (particularly a young child) tend to leave the labour force, this will cause the mother's clothing budget to decline as fewer 'smart' clothes are needed. The difference method will include this decline as part of the costs of the child, causing them to be lower than otherwise. Does it make sense for the lower clothing budget of the mother to be included (negatively) as part of the cost of the child, or should the higher clothing costs of working women without children be counted as part of their work-related costs?

Another limitation of the difference method (which also applies to the normative method) relates to how housing costs vary with changes in household size and composition. The budget standards were developed on the normative judgement (modified in the light of feedback provided by the focus groups) that children under the age of five and those aged between five and ten of the same sex can share a bedroom, while others should have their own bedroom. This implies that whether a two-child family requires a two bedroom or three-bedroom dwelling will depend upon the age and sex combination of the two children. Since housing costs are so high - even when considering only the addition associated with the need for an extra bedroom - this means that the costs of a child of a given age and sex will depend critically upon the age and sex of other children in the household.

The example illustrates the point that there is no such thing as 'the' cost of a child, even a child of a given age and sex assumed to be living in a household at a given standard of living. The best that can be done is to produce estimates that are likely to be broadly applicable to most circumstances and to be clear about how they are derived and what limitations apply to them. These comments apply to the results that follow and should be borne in mind.

Some results: the lovering method

It has already been noted that the cost of children estimates produced by Kerry Lovering were based on a restricted application of the 'basket of goods' or budget standards method which only included the cost of those items that could be clearly associated with the individual needs of the child. A useful check on that research and on the robustness of the budget standards method itself, is to replicate the limited approach of Lovering using the modern budget standards data and methods to see whether it produces similar results.

This involved modifying the new budget standards so that their coverage and underlying assumptions more closely resembles that developed by Lovering. In the area of clothing and footwear, for example, the scope of the Lovering clothing budgets is far narrower than the budget standards in terms of the number of items included; the Lovering clothing budget for a 5-year-old girl contains just 12 separate items (some of them, like socks and underwear, in multiples), whereas the new clothing and footwear budget for a 6-year-old girl contains 49 different items. In light of these differences, the scope of the new clothing budget has been restricted so that it more closely resembles that developed and costed by Lovering.

The resulting estimates are shown in Tables 1 and 2, which compare the updated Lovering 'low income' and 'middle income' child cost estimates for the March Quarter of 1997 with the corresponding 'low cost' and 'modest but adequate' budget standards, respectively. The budget standards results were derived by adopting the same assumptions and methods that were used in Lovering's original study, although it should be noted that it was sometimes difficult to reproduce the Lovering methods precisely because of lack of information. Details of how the results in Tables 1 and 2 were produced will be provided in a forthcoming SPRC report.

Note:(a) Lov = Lovering estimates; BS = limited budget standards estimates; C = child (age in brackets); G = girl (age in brackets); B = boy (age in brackets).
Source:See text.

Note:(a) Lov = Lovering estimates; BS = limited budget standards estimates; C = child (age in brackets); G = girl (age in brackets); B = boy (age in brackets).
Source:See text.

Table 1 indicates that at the lower standard, the budget standards child cost estimates tend to be around $10 a week higher than the updated Lovering figures, except for the teenage boy for whom the budget standard estimates are slightly lower. Both methods produce remarkably similar food, energy and (to a lesser extent) clothing costs, with the higher budget standards figures mainly reflecting higher costs in three areas - household provisions, entertainments and toys. In all three of these areas, the new budget standards include a far greater range of items than it was possible for Lovering to include some two decades ago.

The comparisons in Table 2 reveal a similar pattern to those in Table 1, although now the budget standard estimates are lower for both the teenager and the 10-year-old boy. In both cases, the difference mainly reflects lower energy and clothing costs. As before, the food costs are very similar, although the budget standards again produces higher costs of household provisions, entertainments and toys.

Overall, Tables 1 and 2 indicate that the new budget standards estimates are capable of producing quite similar results to the Lovering estimates if they are adjusted to be more comparable in scope and method. Put differently, although they were constrained by lack of data, the Lovering estimates have withstood the test of time remarkably well and the decision to continue to update them was clearly a wise one. Their main limitation, however, continues to be their incomplete coverage, and now may be the appropriate time to cease updating and publishing the Lovering estimates given their restricted coverage and the time that has elapsed since they were first developed. This is not to deny the very important contribution they have made to our understanding of the cost of children and to the design of family policy over the last 15 years.

In time, the new budget standards may provide the basis to replace the Lovering costs of children estimates by a more up to date and comprehensive set of estimates. Because the budget standards study was conducted within a tight time frame, it was decided that the difference method was the only practical option for estimating the costs of children. The results are described and analysed at length in Chapter 14 of Saunders et al. (1998) and the main features are summarised by Chalmers (1998).

The estimates are presented in Table 3, in the form of both the total costs of different children in one-child families and the incremental costs of children as family size increases. As noted earlier, these estimates were costed at February 1997 prices and refer to households living in Sydney who are renting privately.

Note: At the low cost standard, the husband is assumed to be unemployed and the wife not in the labour force. Both are assumed to be employed full-time at the modest but adequate standard.
Source: Chalmers 1998.

Table 3 illustrates the points made earlier in that they show how the cost of a child varies not only according to the standard of living assumed (low cost or modest but adequate), but also with the circumstances of the child and those of other children in the household. The increase in costs going from low cost to modest but adequate is particularly high for families with children under school age because of the costs of formal child care that have to be incurred by parents at the modest but adequate standard (who are both assumed to be in full-time work).

The Table indicates that while the incremental costs of a 6-year-old girl are slightly higher than her total cost (compare line two in the upper panel with line two in the lower panel), the opposite occurs in the case of a 3-year-old girl (compare line one in the upper panel with line three in the lower panel). These differences mainly reflect the age-sex combinations of other children and what this implies for the number of bedrooms that are needed.

More results: The Lee method

Table 4 compares the breakdown of the budget standard difference method cost estimates with the Lee estimates. The two sets of estimates are not directly comparable because each refers to different families at a somewhat different standard of living. The Lee estimates relate to a one-child, one-income family with a gross income of $688.70 a week, while the modest but adequate budget standards assume both parents are in full-time work and vary (according to the age of the child) between $653 and $669 a week.

However, these figures are net of tax and thus would require a somewhat higher gross income, implying that the modest but adequate standard is somewhat above the implied Lee standard. (It is perhaps also worth noting that the budget standards tend to fall in the second quintile of expenditure for all households with one child - see Saunders et al. 1988, Figure 12.1 - implying that they are not high overall, at least when compared with the expenditures of all one-child families.)

Notes:(a) All figures have been rounded to the nearest dollar. The budget standards assume that the household is renting privately. ‘Other’ includes health and personal care expenses.

Although there are similarities between the total cost estimates shown in the final row of Table 4, there are several important differences in the basis of the two methods. The most important of these is that while the budget standards estimates purport to measure what it would cost to meet the needs of a child at a given standard of living, the Lee estimates are derived from survey data which record what households actually spend on their children. Even though the budget standards are modified in the light of actual spending patterns, while the Lee estimates attempt to unravel the impact of needs from resources available, the two sets of estimates still differ in important respects and this must be kept in mind.

Some of the cost compositional differences shown in Table 4 are a direct reflection of the different assumptions used in the two studies. Child care costs (already referred to in discussing the results in Table 3) are a good example of this. Because budget standards assume that both parents are working full-time, the family has to incur considerable child care costs that are not present to anything like the same extent in the Lee estimates (because only one parent is assumed to be in work). This explains why the budget standards household goods cost component (the budget area where child care is recorded in the budget standards study) - particularly for the family with a 3-year-old girl - are so high.

Another striking difference is in the area of transport costs, where the Lee estimates are between $35 and $40 a week above the budget standards figures. It is not clear why the Lee transport costs appear so high, although the low cost budget standards figures for transport reflect the use of the difference method; families are assumed to own a car whether or not they have children and so the additional costs they incur when there is a child present depends only upon the extra distances travelled in order to meet the needs of the child (to go to school, or the child care centre, or to visit the doctor) , and these involve relatively low extra costs.

The Lee cost estimates for housing and energy seem very low and are well below the corresponding budget standards estimates, while the reverse is true in the case of recreational expenses. It is only in the areas of food and clothing (where the problems of shared consumption largely disappear) that the two sets of estimates are consistently close together. This reinforces the earlier conclusion that these pure individual cost items are capable of quite precise estimation, whatever method is used.

Summary

The comparisons presented in Tables 1-4 reveal that although there are many broad similarities between the child cost estimates that have been published by the Australian Institute of Family Studies and those produced from the new SPRC budget standards research, there are also a number of significant differences in the size and composition of some of the estimates.

It would appear on the face of it that there is a good case for replacing the Lovering estimates by the budget standards, which are both more up to date and more comprehensive in their coverage. Some of the differences between the Lee and budget standard costs shown in Table 4 are striking and, at the very least, suggest that more work is required to reconcile the observed differences.

It is important to emphasise that the budget standard estimates reported here have been derived from the difference method which is not the most ideal way of calculating the costs of children from a budget standard for the household. Research is currently underway within the SPRC which is applying the normative method to produce better estimates of the costs of children and a report is planned for release later this year that will summarise those results.

Concluding comments

There is no ideal way to derive the costs of children, nor will any estimate ever be free of criticism. To expect otherwise is to fail to understand the inherent difficulties of the task. This does not mean that we should throw up our hands in despair, but rather that in striving to improve on our understanding of the concept of 'the costs of children', and in trying to develop better ways to estimate them, we should acknowledge that there will always be imperfections in what can practically be achieved. This is no different from many other areas where research cannot provide a perfect answer and, like elsewhere, to strive for this ideal often helps to elucidate the issues by shedding light on why it is unattainable.

It seems likely that interest in the costs of children will continue to grow, both amongst the public generally and amongst those responsible for the development of policies that affect the wellbeing of families and children. To inform our understanding of the issues and to provide an improved basis on which to decide whether and how to intervene, we need to continue to update the available estimates of the costs of children and to search for improved methods.

The new budget standards represent a step along this path, and while they provide the basis to estimate better the costs of children, there are good reasons to explore also different methodologies and other approaches to the issue. It was noted earlier that a budget standard refers to a particular household living in a particular place at a particular point in time. These 'particulars' are important and there is already a need to improve upon some of them.

For example, in relation to household type, the SPRC research included only a limited range of children and there is a good case for extending the research to include very young children and youth. (Some European countries already estimate the costs of childbirth and pregnancy.) The estimates presented here also refer only to the costs of children in couple households. Budget standards have been estimated for a single woman as well as for female sole-parent households with one and two children. In principle, these can be used to derive estimates of the costs of children in sole-parent families, although using the single woman as the benchmark from which to estimate child costs is more problematic given the different behavioural patterns of single women and single mothers.

In addition to expanding the range of household types, there is also a strong case to re-price the budgets in areas other than Sydney, and preliminary work along these lines has already indicated that this can make a major difference to the budgets, as well as to the costs of children (Saunders 1998c). There is also a need to update the budgets so that they reflect changes in price (as a result of the introduction of a GST, for example), and how this should be done is currently under consideration. Over the longer term, new budgets will need to be developed in response to changes in community norms and patterns of behaviour.

Finally, it is important to emphasise that the budget standards approach is only capable of providing an indication of the direct budgetary costs of children - what it costs from the family purse (and wallet!) to maintain a child at a given material standard of living. There are equally important indirect costs of children associated with the foregone earnings of those who assume responsibility for caring for children within the home. Although the budget standards described here do incorporate the work-related costs of parents (including child care costs) they do not include estimates of the loss in earnings associated with caring for children within the home. Neither is any account taken of the many and varied benefits associated with child-rearing. These important factors would need to be incorporated into any comprehensive assessment of the value - as opposed to the cost - of children.

References

• Bradshaw, J. ed. (1993), Budget Standards for the United Kingdom, Avebury, Aldershot.
• Chalmers, J. (1998), 'Using budget standards to estimate the costs of children', Budget Standards Unit Newsletter No. 4, Social Policy Research Centre, University of New South Wales, p. 6.
• Espenshade, T. (1984), Investing in Children: New Estimates of Parental Expenditures, The Urban Institute Press, Washington DC.
• Lovering, K. (1984), Cost of Children in Australia, Working Paper 8, Institute of Family Studies, Melbourne.
• McDonald, P. (1990), 'The costs of children: a review of methods and results', Family Matters, no. 27, November, pp. 18-22.
• McDonald, P. (1997), 'Gender equity, social institutions and the future of fertility', Working Papers in Demography No 69. Research School of Social Sciences, Australian National University, Canberra.
• Oldfield, N. (1993), 'The cost of a child', in J. Bradshaw (ed.) Budget Standards for the United Kingdom, pp. 177-195.
• Rowntree, B. S. (1901), Poverty: A Study of Town Life, Macmillan, London.
• Saunders, P. (1998a), 'Household budgets and income distribution over the longer term: evidence for Australia', Discussion Paper No. 89, Social Policy Research Centre, University of New South Wales.
• Saunders, P. (1998b), 'Budget standards and the poverty line', Australian Economic Review, vol. 32, no. 1, pp. 43-61.
• Saunders, P. (1998c), 'Using Budget Standards to Assess the Well-Being of Families', Discussion Paper No. 93, Social Policy Research Centre, University of New South Wales.
• Saunders, P., Chalmers, J., McHugh, M., Murray, C., Bittman, M. and Bradbury, B. (1998), Development of Indicative Budget Standards for Australia, Policy Research Paper No. 74, Department of Social Security, Canberra.
• Watts, H.W. (1980), New American Budget Standards: Report of the Expert Committee on Family Budget Revisions, Special Report Series, Institute for Research on Poverty, University of Wisconsin, Madison.

Peter Saunders is the Director of the Social Policy Research Centre, University of New South Wales.

New estimates from the ABS household expenditure survey

by Rebecca Valenzuela

A sign on a Melbourne shop says 'I can imagine a life without children but it wouldn't have as much laughter!'. This aptly summarises how most people feel about children. Indeed, children add an interesting and unique dimension in one's life - and most people would not exchange that for anything in the world!

However, raising a child is not simply a labour of love - it takes a lot of time, a lot of energy and, without doubt, a lot of money. And while parents say that it is impossible to quantify the joys and satisfaction of having a child, they will also be the first to admit that having a realistic idea of the costs involved in child-rearing can help them to plan better.

There are other important reasons for knowing the money cost of a child, particularly in the economic and social policy areas. For example, money costs of children are directly relevant in setting levels of child-support payments following divorce. Cost of children must also be taken into account in assessments of the distribution of income, the progressivity and effectiveness of tax and social security systems, and the impact of government policies on living standards. In particular, cost estimates for children are used to determine whether or not the government's package of benefits and services to assist parents in the important task of raising children is adequate.

Measuring child costs

This article calculates estimates of the money costs of children based on actual expenditures incurred by families in the years 1984, 1988-89 and 1993-94. The costs of children are estimated by comparing the expenditure of families with children to those without children to determine the child's share of family expenditure. These money-cost estimates (also known as equivalence scales) are used to show how much income families with different numbers and ages of children would need to achieve comparable standards of living.

The procedure used here to estimate child costs falls under the 'large scale survey' (or 'demand based') approach to estimation. The label 'large scale survey' comes from the fact that such estimates are based on information gathered from asking thousands of householders how much they actually spend on specific commodity items and inferring child costs from this large set of quantitative information. This approach assumes that standard of living of a household is largely determined by the household's consumption or expenditure (or demand) behaviour and its demographic characteristics. One specifies the sources of expenditure differences - income and other household characteristics - and posits the form of their statistical linkages in 'demand equations'. The parameters of these demand equations are next estimated empirically using information from the household expenditure records, from which the costs of children are finally calculated.

The cost estimates in this article are based on the 1984, 1988-89 and 1993-94 Household Expenditure Surveys conducted by the Australian Bureau of Statistics (ABS). They provide answers to the question of 'What is the cost of a child?' as per the experience of Australian households. They indicate how much parents actually spend on their children, even though the amount spent might be considered inadequate or excessive by some other standards.

An alternative procedure for estimating costs of children is called the 'budget standards (or 'basket of goods') approach. This approach is characterised by the creation of an 'ideal' basket of goods and services for certain model families - ideal according to some expert opinion. Nutritionists and other medical experts are among those who are relied on to provide some minimum diet requirements for particular family types. Other experts who may be involved in the process are educators, psychologists and similar social scientists. From the prescribed basket of goods and services, the cost of a child is derived from that component of the budget that is attributable to children.

The basket of goods approach indicates how much parents ought to spend on their child if the child was to enjoy the fruits of the basket specified by the 'experts'. In this sense, it provides an 'ideal' or desirable costing. The estimates based on the basket of goods approach therefore provides answers to the question: 'What should be the cost of a child?'.

Estimates of child costs from these two alternative methods are not comparable in that each one is intended to measure two different costs. The basket of goods approach clearly involves some normative judgement and is ideal for computing the minimum cost requirements for keeping children at acceptable standards of living. On the other hand, the demand system approach indicates what is actually spent on children by households and families according to the expenditure behaviour of families. Because of their very nature, it would therefore not be surprising to see that estimates from these two approaches are not necessarily the same. Studies reveal that costs from the demand system approach tend to be lower on average compared with those estimates coming from the basket of goods approach.

As a result of their basic differences, the cost estimates emanating from these two different approaches are used for different purposes. Budget standard estimates, for instance, would be more appropriate for the estimation of poverty lines while analysis of income distributions and related welfare issues can be more properly performed using demand system estimates.

Spending patterns of families

Detailed expenditure information from the ABS household surveys reveal interesting spending patterns of Australian families. For the typical household composed of two adults and two children, about 50 per cent of the household budget is spent on Food, Housing, and Transport items. (See Figure 1. Details on the classification of expenditures are in the Appendix).

Figure 1. Distribution expenses of typical 2-adult 2-children family: 1993-94 HES: average weekly income $634.78

The next 25 per cent of this household's budget is devoted to items on Recreation and entertainment (including holidays and travel), and on Household goods. The remaining 25 per cent is distributed across the other commodity groups.

Some variations across household types are worth noting. The budget share of Housing for single-parent families is considerably higher than for two-parent families, reaching as high as 23 per cent for single parents with one child (compared to 14 per cent for two-parent families). On the other hand, the budget share of Transport is only about 12 per cent for single-parent families compared to the 14 per cent share of Transport in two-parent families. The proportion of expenses devoted to Food is also slightly lower in single adult households, with or without children.

Figure 2. Income and expendature of different types of households: sample means, 1993-94 HES

Reported levels of total expenditures were, on average, consistently higher than reported total income but the differences are not statistically significant.¹ Further, two-adult households have higher weekly household incomes than one-adult households (Figure 2). Among households with two children, for instance, two-adult households average about $635 a week in income, compared to just $378 for one-adult households. Households with children have higher incomes than those without, in part because many of the households with no children are 'empty nests' with one or both senior adults retired. That difference in life course stage is one of the reasons that we need the elaborate model used here²: it systematically adjusts for these and other differences in family composition, to enable us to discover the 'pure' or separate effect of children on expenditure patterns.

Commodity-specific and general scales

Table 1 presents the estimates of commodity-specific scales. Commodity specific scales are numbers that show the change in the budget requirements due to the presence of a child or children in a household, per commodity, to maintain the same standard of living as a predefined reference household. Here, the childless couple is set as the reference household.

A scale value of 1.38 for Housing was estimated for a family of two parents and one child (Table 1). This means that a family of three needs a housing budget 38 per cent higher than that required by the two-adult childless household for them to maintain a comparable standard of living. The variances indicate that the estimated scales for two-parent families with two and three children, respectively, are not significantly different from case of the two-adult, one-child family, that is, the housing cost required for the second or third child in the family is highly likely to be negligible.

Source: Valenzuela (1999)

This might reflect couples' practice of buying (or extending to create) a family home big enough for several children around the time of the birth of their first child, so that little or no additional housing is required when the rest of the children arrive (since the overwhelming ideal in Australia is for more than one child in the family). Alternatively, it may be that the option of shared bedrooms may reduce subsequent-sibling housing costs for some families.

For single-adult households meanwhile, the Housing scale value of 82 per cent empirically verifies that it is more expensive to maintain a household on one's own due to the diseconomies of scale of living alone. If not for such diseconomies, the figure would be just 50 per cent.

Some interesting patterns are observed. For all family types, one child will increase food requirements by 22 per cent, two children by 44 per cent, and three children by 50-60 per cent. Because children are 'food intensive', these results indicate that there is not much to be gained in the way of economies of scale for food with the addition of more children in the family.

By contrast, besides the economies of scale in Housing noted above, economies of scales also seem to operate in the consumption of Fuel and power, Clothing and footwear, Household goods, and Transport items, for both single and two parent families. Actual family experiences verify that with the arrival of the first child, significant adjustments in the consumption of particular commodity items are incurred. These include such items as extra heating cost for child's room, purchase of an extra bed, and also a tendency to trade in a small car for a larger, roomier vehicle to accommodate a growing family. With the advent of subsequent children, however, the cost adjustments required on these types of goods would not be as substantial - as children can share rooms, toys, pass on clothes - therefore the scale values diminish with the addition of the second or third child.

Interestingly, the magnitude of the scales for Alcohol and tobacco declines as the number of children in the household increases (more obvious in estimated scales from 1988-89 data). By contrast, the scales for Medical care, Recreation and entertainment, Personal care, and Others, commodity groups exhibit no defined trend. It may be possible that the presence of children in the household tends to influence expenses away from 'adult goods' under which alcohol, tobacco and other miscellaneous goods are classified. Perhaps the aspiration of setting a good example for one's children influences these expenditures.

The general scales are presented in Table 2. These scales are used to estimate the cost of a child in the overall sense. In 1993-94, a first child required a once childless couple to increase their expenditure by about 18 per cent to maintain their pre-child standard of living (see Table 2). A second child will require a further increase in total expenditure by about 7 per cent, while a third child raises the requirements by a further 8 per cent. Consider, for example, a couple who spends about $325 per week in rent, grocery shopping, clothing expenses, petrol and heating costs, and eating out treats once a week. The arrival of the first child will entail shifts between types of commodities consumed, but overall, the scales imply that their pre-child standard of living can be maintained if they increase their budget expense by about $58 (that is, 18 per cent of $325.00). In the same way, having two children will require them to increase their expenditure by $81.25 (25 per cent of $325.00) compared to when they were childless. Having three children would require them to increase their expenditure by $107.25 (33 per cent of $325.00) each week, to maintain the standard of living they experienced when they were childless.

Source: Valenzuela (1999)

For single-adult households the cost requirements tend to be larger than those for the two-parent household, according to the scales in Table 2. Suppose that the income of this single adult was also $325 per week. Then having a first child will require her to raise her budget expense by about $72 dollars (or 22 per cent) to avoid any deterioration of her standard of living. Having two or three children would require an income gain of about $114 (or 35 per cent of total income), compared to her standard of living when she was childless.

What about changes over time? For the one-adult households with children, the estimated scales for Housing, Fuel and power, Food, Medical care, Transport, and Others have typically declined between 1984 and 1988-89 and then increased again between 1988-89 and 1993-94 (Table 1). In contrast, there have been some declines in the estimated scales for the two-adult households for these same commodities, although these declines are not statistically significant. An exception to this is the 1988 scales for Transport that registered significantly lower values than those for 1984. This may reflect some substitution effects in the later years, when family preferences began tending towards larger but cheaper-to-run family cars that started to flood the market towards the end of the 80s decade. The Transport scales did not change significantly between 1988-89 and 1993-94.

Overall, the direction of the change in the size of the scales is the same for the two adult households with one child and those with two children, while the change in the scale values for those with three children families was in the other direction. Interestingly, the only consistent (direction of) change for all household types was observed for Alcohol and tobacco. For this commodity, scales in 1984 did not show any discernible pattern associated with the addition of children in the family. In the later data sets, however, the scales indicate a steady decrease in the consumption of these 'adult goods' with every additional child for both single-parent and two-parent families. It would seem from this that the presence of children has increasingly had a deterring effect on the consumption of Alcohol and tobacco items.

Changes over time in the estimates of the general scales can be seen from Table 2. For two-adult household, the 1993-94 scales are smaller than both the 1984 and 1988-89 scales, although the scale estimates from the later year data sets are closer to each other. It appears that for the 1984 Household Expenditure Survey, there is a strong evidence of economies of scale in the second child but adding the third child increases these households' needs considerably. However, this effect is diminished for the later data sets. The 1993-94 scales, in particular, suggest equal cost requirements for the second and third children.

The economic model which formed the basis of the estimation of these scales allowed for the possibility that consumption behaviour of families varies systematically by the income level of the household. As the figures show, however, the estimated cost ratios for families of different compositions appeared stable across the various income levels. This suggests that the overall cost requirements (in per cent), associated with the presence of children, is the same whether the household is rich or poor. However, these percentage rates imply significant differences in dollar value terms, which may further carry important implications for families in transition.

For example, we compare the pre- and post-divorce standard of living of two families - one, a couple with two children having an income of $1250 a week, and the other, a couple with two children having an income of $375 a week. If these couples divorce and one partner becomes a one-adult-and-two-children family, the high income single parent will need to have $920 to maintain his/her pre-divorce standard of living, along with the two children. The lower income single parent, meanwhile, will need $276 to do the same. This suggests that the child support requirements to assist single-parent families cannot be generalised into standard dollar figures for the average case. If such dollar amounts are fixed, a significant decline in the standard of living for many recently divorced single parents is bound to happen. In general, though, higher income households where both spouses are highly educated and in secure employment situations would have better coping mechanisms than their lower income counterparts.

Cost of children estimates

Table 3 presents estimates of the money costs of children as per the experience of Australian households. The values in the table are in nominal dollar terms for the years the surveys were conducted. These cost estimates were derived from the cost ratios in Table 2 and were calculated based on specific gross family incomes. The table indicates the expected costs for low, medium and high income households and for certain numbers of children in the family. Disaggregation of the costs by the age of children or by specific commodity groups are not given here, although such estimates can be provided by the author upon request.

Conclusion

This paper presents estimates of the cost of raising children in Australia based on actual expenditures incurred by families as revealed in the 1984, 1988-89 and 1993-94 Household Expenditure Survey.

In general, the results show that to be able to maintain the living standards of the household before the arrival of children, the family budget would have to be increased by about 18 per cent for the first child for two-parent families, and by about 22 per cent for single-parent families. The additional budget requirements for the second and third children will still be positive but not as much as that of the first child. Again, there is more adjustment required by single-parent households than two-parent families.

The commodity specific scales show that the budget requirements vary across the various commodity groups. For the first child, there is a significantly higher budget requirement (38 per cent) to meet Housing needs of the child while in terms of Food, the budget needs to be adjusted by only 20 per cent. The results show that single-parent households need more assistance to meet Housing adjustment needs than two-parent families. After the first child, large gains in economies of scale were observed for Housing, Fuel and power, Household goods, and Transport items. Also, the findings suggest that the presence of children induces adult family members to consume less of the adult goods' such as Alcohol and tobacco.

One important finding relates to the implication of the estimated scales for households in different income groups as they go through certain transition phases. For instance, to maintain the pre-divorce standard of living of many single parents and their children, it will be necessary to recognise significant differences in needs according to people's pre-divorce income levels. If such differences are not properly accounted for, a significant decline in the standard of living for many just-turned single parents is bound to happen.

References

• Australian Bureau of Statistics (1992), Information Paper 1988-89 Household Expenditure Survey of Australia (July 1992 Update), Canberra.
• Australian Bureau of Statistics (1996), Detailed expenditure items: 1993-94 household expenditure survey, Australia, Catalogue No. 6535.0, Canberra.
• Kakwani, N. (1980), Income Inequality and Poverty: Methods of Estimation and Policy Applications, World Bank, Oxford University Press.
• Lluch, C. (1973), 'The extended linear expenditure system', European Economic Review, vol. 4, pp. 21-23.
• Valenzuela, M.R. (1997), 'Alternative approaches to the estimation of household equivalence scales: an Australian application', Ph.D dissertation, University of New England.
• Valenzuela, M.R. (1999), Costs of Children and Living Standards in Australian Households, Working Paper No. 8/99, Melbourne Institute of Applied Economic and Social Research, The University of Melbourne.

About author

Dr Rebecca Valenzuela is a research Fellow at the Melbourne Institute of Applied Economic and Social Research of the University of Melbourne where she is conducting research on such issues as consumer demand, costs of children, poverty and income distribution.

Detailed definition of commodity classification used

1. Housing

rent, mortgage, property rates, house and contents insurance as well as housing repairs and maintenance.

2. Fuel & power

electricity, gas and other fuels (excluding car petrol).

3. Food

bakery products, flour and other cereals, meat and fish, dairy products, fruits and vegetables, miscellaneous food (jam, jellies, coffee, tea), non alcoholic beverages, meals out and take away food.

4. Alcohol & tobacco

cigarettes and all types of alcoholic beverages.

5. Clothing & footwear

clothing and footwear for men, women and children, clothing accessories (e.g. ties, gloves, hankerchiefs) as well as clothing and footwear services (e.g. drycleaning and shoe repairs).

6. Household furnishings & equipment

furniture and floor coverings, blankets and rugs, household linen and furnishings, household appliances, glassware, tableware, household utensils and cleaning agents; gardening services, housekeeping, childcare, and the repair and maintenance of household durables.

7. Medical & health care

accident and health insurance premiums, practitioner's fees, prescriptions, medicines, pharmaceutical products, hospital and other health charges.

8. Transport

motor vehicles, petrol and fuels, vehicle registration and insurance, vehicle servicing and repairs, driver's licenses, driving lessons, subscriptions to motor organisations, vehicle hire, as well as public transport fees.

9. Recreation & entertainment

television and other audiovisual equipment, books, newspapers and other printed materials, recreational equipment (cameras, musical instrument, toys), gambling, entertainment, recreational services, pets, holiday.

10. Personal care

toiletries, cosmetics, hair dressing and beauty services. 11. Others - miscellaneous goods (watches, jewellery, stationery), interest payments on selected credit services, education fees, life insurance and other miscellaneous services.

by Peter McDonald

Peter McDonald, AIFS Deputy Director (Research), examines the difficulties of estimating the direct costs of children, arguing that the cost of a child is not an objective fact but varies according to tastes and preferences, and the amount of money parents have to spend on their children.

The Institute receives many inquiries about the costs of children. The information is sought by parents, courts and government and non-government agencies dealing with issues such as child maintenance, fostering and, even, failed vasectomies. The expectation of most of those who inquire is that the cost of a child is a fairly straight-forward matter and, therefore, should be easily obtained. This is far from being the case; estimating the cost of a child is a complex and highly imprecise exercise. Therefore, estimates of the cost of children should be used with caution.

Conventionally, the costs of children are divided into two types: direct and indirect. Direct costs are the additional costs that a household has because of the presence of children. Indirect costs, on the other hand, refer to the loss of income that a household experiences because one or both parents spends time out of paid employment or takes a lower-paying job in order to look after the children. This paper deals only with the direct costs of a child; indirect costs have been addressed for the Institute by Beggs and Chapman (1988).

Measures of the Costs of Children

There are essentially three ways in which the direct costs of children have been estimated: the opinion survey approach; the budget or basket-of-goods approach; and the expenditure survey/ equivalent standard of living approach.

Opinion surveys

With the opinion survey approach, the researcher simply asks a representative sample of families how much it costs to keep their children. The data so obtained would then be analysed according to the ages of children and the number in a family

This method has been little used and it is easy to see why this is so. If you have children, can you give an estimate of how much they cost? Do not forget that because you have children, you may have needed a bigger car or a second car and you use cars more often and therefore have them serviced more often. Do not forget that because you have children, you may have needed a bigger house and its maintenance costs will be greater. Do not forget that children mean that your fuel and telephone bills are higher, and remember all those health and education costs. And did you buy the video for yourself or for the children, and what about the freezer and the larger refrigerator and washing machine? And do your furniture and household goods need to be replaced more often?

This is a sample of some of the prompts that the researcher may give to you on the doorstep. Little wonder that the opinion survey approach is regarded as being unlikely to produce reliable estimates.

Basket-of-goods approach

With the basket-of-goods or budget approach, the researcher specifies a standard 'basket' of goods that a child of a given age (and sex?) would need. This can be done at various levels - for example, the bare minimum, a comfortable level, and even an indulged level.

The method has two main difficulties: what should be included in the basket, and how much do the items in the basket cost? Returning to those earlier questions: what proportion of a car do we attribute to a child, what proportion of the furniture and household goods, what proportion of the fuel and telephone bills, how much of the petrol and car repair costs? The costs of children are not simply a matter of food, clothing and bus fares. And even then, whose standards of food and clothing do we use? And how do we take account of the fact that the costs of travel to school may vary greatly between children in different localities.

Despite these problems, the basketof-goods approach is likely to be far better than top-of-the-head guesstimates on the doorstep. The most detailed study using the basket-of-goods approach ever conducted in Australia was done at the Australian Institute of Family Studies by Kerry Lovering (1984). Lovering overcame the problems of those basket items which are difficult to calculate by simply leaving them out of the basket. Her basket does not include the child-related costs of housing, transport, school fees and uniforms, child care, medical and dental expenses. It does include food, clothing, fuel, household provisions, costs of schooling other than fees and uniforms, gifts, pocket money and entertainment. Even for the items included, the development of standards provided great difficulty. For example, comparison with other research to be discussed below would suggest that Lovering's standard for clothing was on the low side.

Limitations aside, the Lovering study has provided the best estimates available in Australia of the costs of children over the past seven years. The Institute updates the Lovering estimates each quarter on the basis of changes in the consumer price index.

The Lovering estimates are always provided to users with the accompanying footnotes which point out which costs are and are not included in the estimates. It is remarkable, however, how often the footnotes are ignored and the Lovering estimates are taken as the full costs of a child. The Lovering estimates are not the cost of a child but only part of the costs of a child. No attempt has yet been made in Australia to estimate the full costs of a child using the basket-of-goods approach.

Expenditure survey approach

The expenditure survey/equivalent standard of living approach (Espenshade, 1984) attempts to obtain estimates of the total costs of children including all those costs left out of the Lovering estimates. Basically, the method compares the household expenditure of a couple with children with that of a couple of the same age without children who have an equivalent standard of living. The difference in the expenditures of the two households represents the cost of the children.

Comparison between couples of the same age is specified so as to reduce the difference in tastes or needs that old couples would have compared to young couples. However, couples' tastes and needs also vary by characteristics other than age, such as education and occupation. It is desirable that these characteristics are also taken into account, but expenditure survey samples are sometimes of insufficient size to permit such detailed analysis.

The key to this approach, however, is the specification of what constitutes an equivalent standard of living. After investigation of several possible measures, Espenshade (1984) concluded that the best of the measures examined was the proportion of total current consumption that was spent on food consumed at home. That is, two households of differing composition will have the same standard of living if they spend the same proportion of their total household expenditure on food at home. Thus the cost of children is the difference in expenditures between a couple with children and a couple without children who both spend the same proportion of their consumption on food at home. The accuracy of the method is thus highly dependent upon the assumption that these two families have the same standard of living - a relatively loose assumption.

This approach has been applied in Australia by Donald Lee of Deakin University working under contract to the Institute. Dr Lee (1989) used the 1984 ABS Household Expenditure Survey to make his estimates. His methodology allows for the costs of children to be estimated according to the numbers and ages of the children in a family and the income levels of both parents. The inset presented here shows Lee's results updated to the June Quarter 1990 for a one-child, single-income family with an income of $578.01 per week (male average weekly earnings for this quarter were $554-60 per week). The Lee estimates from 1984 have been updated to 1990 using the change in average weekly earnings between May 1984 and May 1990 rather than the change in the consumer price index. This is done because the Lee estimates are derived for a family with a given income level: in contrast, the Lovering estimates should reflect the changing costs of a basket of goods and hence are updated using changes in prices rather than changes in earnings.

Inset: Costs of Children in Australia

Constraint of Family Income

Discussion of the relationship of costs of children to the income of parents gives rise to a fundamental difficulty related to estimates of the costs of children: the money that families spend on their children is in most cases very likely to be determined by how much money they have. So when we observe that a family has spent a certain amount on their children, this may not be determined so much by some objective measure of how much children cost but rather by the amount that the family has available to spend on the children. The lower the family income, the more likely that money spent on children is constrained by the income level. Furthermore, families on the same income level will have less to spend per child the more children that they have.

This indicates an important difference between results obtained by using the basket-of-goods method and the equivalence survey method. The basket-of-goods method indicates how much parents would spend on their children if the child was to enjoy the fruits of the basket specified by the researcher. In this sense, it provides an 'ideal' or desirable costing- In contrast, the expenditure survey method indicates how much parents actually spend on their children, even though the amount spent might be considered inadequate or excessive by the objective standards of the basket-of-goods method.

A difficulty with the basket-of-goods approach is that because it deals with desirable expenditure on a child, it becomes difficult conceptually to consider what happens when the number of children in a family increases while the parents' income remains the same. That is, the basket-of-goods approach has difficulty accommodating the constraint on expenditure of the level of family income.

The basket-of-goods approach also has a difficulty when it comes to assessing the effect on expenditure on children of family type. That is, what are the relative costs of children in one-parent and two-parent families? This is a problem also for the expenditure survey approach because, in practice, the number of one-parent families in expenditure surveys is too small for valid analysis.

With these reservations in mind, it is useful to consider some of the relativities regarding costs of children that emerge from the study conducted by Donald Lee using the expenditure survey approach.

Impact of Child's Age

The Lee study provides comprehensive costs of children across all areas of expenditure based on the diaries of expenditure kept by Australian families who participated in the 1984 Household Expenditure Survey. For reasons as yet unknown, the cost estimates for children aged 14 years and over obtained using this database appear to be too low and so the relativities to be discussed here will refer only to children up to their fourteenth birthday. Results showing the relative costs by age group are set out in Table 1. In the table, the average annual cost of children up to their 14th birthday is taken at the standard cost (1.00), while the costs at specific ages are shown relative to this.

*Data refer to two-income families where the wife earns $400 per week. Lower levels of income refer to one-income families. Source: Derived from results of Lee (1989).

The data suggest that while costs of children rise with the age of the child (except for the higher costs at ages 0 and 1), the rises are not dramatic. Furthermore, there is almost no variation in the pattern of costs by age of child as income rises. Thus, in broad terms, it could be said that the Lee data support the flat allowances for children under the age of 13 years in the social security system. The non-recognition of age of child in the Child Support Scheme formula is also supported by these data, at least for the pre- teenage years.

Effect of Family Income

What impact does the family's level of income have on the costs of children? The Child Support Scheme formulae apply a constant percentage of the noncustodial parent's income up to a limit of two- and-a half times average weekly earnings. This would imply that child costs increase proportionally as income increases. Family allowances provide flat rates of benefit as income rises, with the effect that these benefits would meet a higher proportion of the costs of children in lower income families. This can be justified by the principle of vertical equity (Brownlee and McDonald 1989).

The Lee data (Table 2) show that the dollar costs of children are in fact relatively 'flat' compared to rises in family income. Children in poor families cost proportionally more, and children in rich families proportionally less than those in middle income families. For example, children in families whose incomes are only 63 per cent of the average (AWE in 1984 was approximately $400 per week) cost 78-84 per cent of the costs of children in average families. On the other hand, families with two and a half times the average income only spend 53-56 per cent more than the average on their children.

The pattern of costs of children by income does not vary according to the numbers of children.

(a) Costs used in this table are the total costs of a child from birth to the 14th birthday.
*Data refer to two-income families where the wife earns $400 per week.
Lower levels of income refer to one-income families.
Source: Derived from Lee (1989).

Effect of Numbers of Children

The Lee study also allows us to examine whether the numbers of children in a family have any bearing on the cost of children. Table 3 shows that two children cost about 55 per cent more than one child, while three children cost about twice the cost of one child. We cannot say, however, the extent to which this is due to a notion that children come cheaper in numbers or simply that, at a fixed income level, parents are constrained to spend less per child as the number of their children increases. (An analysis later in the paper examines which particular expenditure items contribute to the reduced costs as numbers of children increase.)

(a) Costs used in this table are the total costs of all children from birth to their 14th birthdays.
*Data refer to two-income families where the wife earns $400 per week.
Lower levels of income refer to one income families.
Source: Derived from Lee (1989).

Types of Expenditure: One-child, Average Income Families

The actual expenditure on the child in one-child families with incomes a little above average weekly earnings is shown by type of expenditure in the inset table relating to the Lee estimates.

Transport is the largest of the expenditure types. This does not refer simply to expenditure on public transport but also includes the additional expenses that one-child couple families have compared to couples with no children in purchasing and running private vehicles. Food is second ranked closely followed by recreation. Besides holidays and outside entertainment, recreation includes such items as toys, television, sound equipment, home computers, records and cassettes, books, newspapers, sporting and camping equipment and pet food. This helps to explain the apparently high level of recreation expenses related to children.

When these results are compared with those obtained by Lovering (1984), it is encouraging that both methods produce similar results in respect of the costs of food for a child in a one-child family. For example, Lovering found food costs for children in May 1983 ranged from $13 per week for a 5 year- old child to $27 for an 11 year-old in middle-income families. As at June 1984, the Lee study obtained results of $16 per week for a 2-4 year-old child to $25.46 for an 11-13 year-old child.

On the other hand, the costs of clothing were much lower in Lovering's standard basket than the expenditures on clothing indicated by the Lee study. For example, costs of clothing for a child in a middle-income family in May 1983 ranged from $4.55 per week for a 2 year-old to $ 5.25 per week for an 11 year-old. The Lee estimates showed a range for June 1984 of $8.69 per week for a 2-4 year- old to $13.44 per week for an 11-13 year-old. While the Lovering costs exclude school uniforms, this is unlikely to account for this difference.

Overall, as indicated by the two tables shown in the inset, for middle income families, the Lee estimates of total expenditure on the one child in a one-child family are about two and a half to three times the part-costs of one child according to Lovering's standard basket.

Types of Expenditure According to Family Income Level

Do the relative expenditures on types of costs related to children vary as the family income level changes? This issue is addressed by the data in Table 4.

(a) Costs used in this table are the total costs of all children from birth to their 14th birthdays
*Data refer to two-income families where the wife earns $400 per week. Lower levels of income refer to one-income families
Source: Derived from Lee (1989)

Not unexpectedly, expenditure on food is relatively inelastic to changes in income. At the other end of things, household goods and services and, particularly, 'other' expenditure (probably reflecting private school fees) were the most elastic expenditure types. At the extremes, families with incomes of $1000 per week compared to families on $250 per week spent 40 per cent more on food for the child, but 111 per cent more on household goods and 299 per cent more on 'other' costs. Wealthier families thus are in a position to spend on the more expensive, discretionary items of child expenditure and they do so.

Table 5 presents these data in a different way by placing the seven types of expenditure in rank order for each income level.

(a) Costs used in this table are the total costs of all children from birth to their 14th birthdays.
*Data refer to two-income families where the wife earns $400 per week. Lower levels of income refer to one-income families.
Source: Derived from Lee (1989)

At every income level, costs of transport retain the first ranked position in child costs. Food, on the other hand, drops from second to fifth ranking as income increases. 'Other' costs move from the lowest ranking for low income families to the second ranking for high income families. It should be noted that these estimates do not include costs from age 14 years onwards. The effect of education costs, particularly private school fees, is likely to be even more significant at these older ages.

Types of Expenditure According to Number of Children in Family

As observed above, costs of children decline in relative terms as the number of children in the family increases. Table 6 shows in which types of expenditure these savings are made.

All types of expenditure produce some reductions per child as the number of children increases, but the basic expenditures (food and clothing) are clearly the items upon which the reductions are least. Contrary to popular images, 'hand-me-down' clothing contributes very little to the relative savings in clothing expenditure for families with more than one child. Mirroring the conclusions in the last section dealing with income, Table 6 shows that the largest reductions occur in the more discretionary types of expenditure, household goods, recreation and 'other' expenditure. Thus increases in the number of children in a family have similar effects on expenditure patterns as declines in family income.

(a) Costs used in this table are the total costs of all chidren from birth to their 14th birthdays
Source: Derived from Lee (1989).

Conclusion

The analysis in this paper shows that the cost of a child is not an objective fact but varies according to tastes and preferences and according to the amount of money that parents have to spend on their children. In particular, the data indicate that families spend what they can on their children irrespective of any 'objective' costing that may be placed on a child by a researcher using an approach like the basket-of-goods approach. On the other hand, the basket-of-goods approach has the advantage that it Is easily understood, and users of the data are able to directly assess whether they regard the contents of the basket as appropriate or not. The expenditure survey approach, in contrast, is dependent upon the single assumption, the validity of which is difficult to assess, that households which spend the same proportion of their consumption on food at home have the same standard of living

The Institute is continuing Its research efforts in this field and is cooperating with other bodies which also have an interest in research on the costs of children.

References

• Beggs, J. and Chapman, B. (1988), 'The forgone earnings from childrearing in Australia', Centre for Economic Policy Research, Australian National University, Canberra.
• Brownlee, H. and McDonald, P. (1989), 'Proposals for the treatment of families in the tax system: an assessment of equity and efficiency', Paper presented at the Third Australian Family Research Conference, Ballarat, 26-29 November, Australian Institute of Family Studies, Melbourne
• Espenshade, T. (1984), Investing in Children: New of Parental Expenditures, The Urban Institute Press, Washington.
• Lee, D. (1989), A program for calculating the direct costs of children based on the 1984 ABS Household Expenditure Survey, floppy disk, Australian Institute of Family Studies, Melbourne.
• Lovering, K. (1984), Cost of children in Australia, Working Paper No.8, Australian Institute of Family Studies, Melbourne.