Equivalence Scales and Intra-household Distribution Hilde Bojer Department of Economics, University of Oslo Pobox 1095 Blindern 0317 Oslo Norway E-mail: hilde.bojer@econ.uio.no Abstract Equivalence scales are dened as the ratio between the cost functions of households of dierent size and composition. It is shown that the household cost function depends on the distribution of income or utility within the household. Therefore, neither cost functions nor equivalence scales are uniquely determined by the size and composition of households. In analysis of personal income distribution, income is usually dened as total household income per equivalent adult. The number of equivalent adults assigned to a household depends on the size and composition of the household. Household income is used because members of a household are assumed to pool their income. Pooling of income means that the consumption possibilities of each individual may be dierent from her individual income, larger or smaller as the case may be. There are two reasons why household income is not then simply divided by the number of persons. Firstly, economies of scale are assumed to be present in household consumption, for instance because of public goods in the household. Even if two cannot live as cheaply as one, they certainly do not need two kitchens. Secondly, there may be systematically 1 dierent preferences between dierent kinds of persons. Typically, children are assumed to dier from adults. Intuitively, equivalent adult scales or equivalence scales should give the answer to the following question: How big an income does a household consisting of several persons need in order to obtain the same (material) standard of living as a single person with a given income? A widely used equivalence scale, the OECD scale, assigns the number 1 to a single adult, 0.7 to the second and each following adult and 0.5 to each child. A household consisting of two adults and one child will thus consist of 2.2 equivalent adults, or be assigned the number 2.2 on the OECD equivalence scale. Formally, equivalence scales are dened as the ratio between the cost of living functions of households. The standard of living is represented by the utility function of the household. Utility is assumed to be a function of household composition as well as commodity quantities. The cost function is derived from the indirect utility function. For a given price vector p and a vector of demographic characteristics z it is written c(u; p; z). Let the reference household have characteristics zr. A schedule of equivalence scales1 is given by (See e.g. Blundell, Preston and Walker 1994) m(u; p; z; zr) = cc((u;u;p;p;zzr)) (1) The utility level u represents the utility level of the household as a whole, in accordance with the traditional approach in economics of treating the private household as a single entity without distinguishing between the dierent individual members of the household. In the literature on equivalence scales there is very little discussion of the relationship between the welfare of the individual and the welfare of the household as a whole. One exception is Nelson 1993, another Blackorby and Donaldson 1994, where internal household distribution is assumed to make all individual utility levels equal. Household utility is then set equal to this common utility level. There is a growing literature in economics analysing decision processes within the household, with particular attention being paid to the possibility that distribution within the household may well not secure equality. (Apps and Rees 1988, Bourgignon and Chiappori 1994, Haddad and Kanbur The term equivalent adult scales seems to imply that the reference household is a single adult. The term equivalence scales may be taken to be more general. 1 2 1990, Wooley and Marshall 1995) The purpose of this paper is to explore some consequences for the theory of equivalence scales of the individualistic approach, and of intra-household inequalities. Let us then take a closer look at the household as a collection of individuals. In the following, I shall simplify the analysis by assuming that all individuals have the same preferences. This assumption, while acceptable in a world of adults, means that children are not included in my analysis.2 I shall study a household consisting of two identical adults, while the reference household consists of one adult. Prices will be assumed to be constant, so that I can ignore the price vector in (1). To start with, I shall also assume that all consumption goods are individual: there are no public goods within the household and no economies of scale. The concept of a household is still meaningful: the two adults constitute a household because income is shared between them. One advantage of these strong assumptions is to create a quite clear intuition of what the value of our household ought to be on an equivalence scale: it should equal two. In the absence of public goods and economies of scale, two adults need exactly twice as much as one adult to obtain the same utility level. In this simple world, all individuals have the same individual indirect utility function, u = V (x), where x is total consumption expenditure (individual). The utility function is assumed to possess the usual properties, in particular it is strictly increasing and strictly concave in x. Inverting V gives the cost function c = C (u). From the properties of V (:), it follows that C (:) is strictly increasing and strictly convex in u. The cost of living function of the household is now a simple sum of the two identical cost of living functions. The two household members may, however have dierent levels of utility. Therefore, the cost function of the household will have as arguments the two individual utility levels. Superscripts h, 1 and 2 denote the household and the two individuals respectively. ch (u1; u2) = c(u1) + c(u2) (2) The question now is, what should we mean by the utility of the household as a whole? One denition of household utility might be the total utility uh = 2 As I argue in e. g. Bojer 1996, it is not enough to allow for dierent preferences in order to incorporate children in the analysis. There are great diculties in assigning preferences to children. 3 u1+u2. However, this denition is clearly not useful for comparing households of dierent sizes. A better denition is average utility: u = 1=2(u1 + u2). None of these denitions gives cost as a well dened function of household utility. To obtain cost as a function of average utility, the internal distribution needs to be known. A simple and attractive rule would be to postulate that the household acts so as to maximise average utility for a given total income. It is well known that, when preferences are equal, maximum average utility is obtained when the two utility levels are equal. If u = u1 = u2, (2) becomes: ch(u) = c(u) + c(u) = 2c(u) (3) The equivalence scales value of the two-person household then simply becomes m(u; 2; 1) = 2c(u)=c(u) = 2, as it should. It follows from duality theory that the cost function (3) represents the minimum expenditure required to obtain a specied average utility level. The rule of equal utilities within the household postulated by Blackorby and Donaldson (1994) is therefore in some ways the natural assumption to make; it makes the denition of the household cost function similar to the denition of individual cost functions, it results in the intuitively right value of the equivalence scales, and it makes for a straightforward interpretation of household income per equivalent adult. But the assumption of equal utilities is a questionable one, and certainly not the only possibility. If utility levels are not equal, the cost function (2) no longer minimises the expenditure necessary to obtain the average utility of the household. Hence, we may be certain that, if u1 = u2, then c(u1) + c(u2) > ch(u) (4) Moreover, the cost-function depends on distribution of utility as well as average utility in the household. Let the distribution of utility be characterised by a parameter such that = 1 when utilities are equal, while 0 < < 1 when they are not. In general, therefore, the equivalent scales corresponding to a given average household utility must be written 1 + c(u2) m(u; ; 2; 1) = c(u )c( (5) u) It follows from (4) that m(u; 2; ; 1) > 2 whenever = 1. In other words, if internal distribution within the household results in unequal utilities, a 6 6 4 two person household needs more than twice the income of a single person household to achieve a given level of (average) utility. Now, household utility as such may well not be the important concern, either in distributional analysis or in distributional policies. What concerns us may well be the welfare of one particular household member. Let us make the reasonable assumption that number one has the higher utility level, but does not consume the whole household income. Since the preferences are identical, this implies c(u1) > c(u2) > 0 and therefore c(u1)=c(u2) > 1. If the utility of number one is what concerns us, the equivalence scales should be dened by 1 2 2 m(u1; 2; 1) = c(u c)(+u1c)(u ) = 1 + cc((uu1)) (6) In this case, 1 < m(u1; 2; 1) < 2: If what concerns us is the utility of number two, the denition is: 1 2 1 m(u2; 2; 1) = c(u c)(+u2c)(u ) = 1 + cc((uu2)) (7) In this case, we see that m(u2; 2; 1) > 2: It costs more to ensure the well-being of number two than that of number one. The unequal internal distribution implies that in order for one ECU to reach number two, more than one ECU must reach number one, and more than two ECUS accrue to the household as a whole. On the other hand, for one ECU to reach number one, number two receives less than one ECU and the household needs less than two ECUS. This is also the reason why equality of utilities is necessary for the household to have an equivalence scale value of exactly 2. Now, let us relax the assumption that preferences are identical. If the household distributes income to ensure maximum average utility, marginal utilities but not utility levels will be equal. If, on the other hand, income is distributed to secure equal utilities, the ensuing household cost function will not minimise the expenditure corresponding to the average utility level in the household. The researcher working with equivalence scales will either have to choose between one of two equally attractive assumptions when interpreting the scales, or fall back on the assumption of equal preferences. In the last case, equivalence scales handle the eect of household size and composition only in a limited sense: they only measure economies of scale. 5 Actual behaviour of households will probably neither maximise average utility nor secure equal utilities. Moreover, if internal distribution varies from household to household, cost functions will vary from household to household. It is, for instance, realistic to assume that internal distribution between adults will depend on individual earning power, which may well vary even between individuals who otherwise have the same preferences. Such variation raises problems for empirical estimation of equivalence scales. References Blackorby, Charles and David Donaldson (1994) 'Measuring the cost of children: a theoretical framework' in Richard Blundell, Ian Preston and Ian Walker (Eds) The Measurement of Household Welfare 51{69 Cambridge University Press. Blundell, Richard, Ian Preston and Ian Walker (1994) 'An introduction to applied welfare analysis' in Richard Blundell, Ian Preston and Ian Walker (Eds) The Measurement of Household Welfare 1{50. Cambridge University Press. Bourgignon, Francois and Pierre{Andre Chiappori (1994) 'The collective approach to household behaviour' in Richard Blundell, Ian Preston and Ian Walker (Eds) The Measurement of Household Welfare Cambridge University Press. 70{85. Haddad, Lawrence and Ravi Kanbur (1990) 'How Serious is the Neglect of Intra{household Inequality?', The Economic Journal, 100: 866{ 881. Nelson, Julie (1993) Household Equivalence Scales: Theory versus Policy?', Journal of Labor Economics, 11:471{492 Woolley, Frances R. and Judith Marshall (1995) 'Measuring Inequality within the Household', Review of Income and Wealth, 40, 415{432. 6