Who benefits from child benefit?

Blow, Laura ; Walker, Ian ; Zhu, Yu 等

I. INTRODUCTION

Over most of the developed world large financial transfers are made to parents by virtue of their parenthood. For example, in the United States the recently introduced Child Tax Credit (CTC), which goes to the vast majority of children, (1) costs almost $1 billion each week, or about 0.4% of GNP. The United Kingdom government spends on average about $25 (at present exchange rates) each week on each child in the form of a lump sum transfer called Child Benefit (CB) which goes to all children, and in addition, the United Kingdom has its own version of CTC which goes to almost all families with children--and together CB and CTC account for approximately 1% of GNP. (2) In the United Kingdom, CB and CTC represent a sizeable contribution to family net incomes, especially for the single parent households considered in this paper.

The typical rationale for these payments that are ostensibly earmarked for children is that they are good for our children. Such transfers are usually motivated by concern for the welfare of children and implicitly presume that there is some market failure that prevents parents from investing in the desired quality and/or quantity of children throughout their lives. This might arise, for example, through child quality being a household public good implying parental free-riding in quality investments, or through imperfections in fertility control, or credit market constraints that prevent households from smoothing the costs of children. A further motivation is that this is a form of intra-household redistribution so that, in households where resources are not pooled, a transfer payment via the mother may have different effects on spending than other forms of income. Moreover, particular concern arises for children in poor households, and the United States and the United Kingdom are distinctive in having child poverty rates that are considerably higher than that in most other developed countries. (3) Thus, there may be credit market constraints that prevent households, especially poor ones, from spreading the costs associated with children across their lifetimes that CB can help mitigate.

This paper is concerned with the impact on household spending patterns of exogenous changes in a lump sum cash transfer that is made to all parents. The United Kingdom is an excellent laboratory to address this issue because CB has been a simple lump sum universal transfer for a period of more than 20 years from 1980 up to the introduction of CTC, which is a means-tested supplementary transfer, in 2001. Since that reform in the late 1970's, the level of payments has varied dramatically over time.

Rather than consider intra-household distribution issues, our aim here is to try to complement existing research on the relationship between child outcomes and household income by trying to infer how CB is spent--in particular, we are interested in how CB affects spending on adult and child specific goods. That is, we investigate the impact of CB on household spending patterns with a view to estimating its impact on goods that are "assignable" to either children or adults. Thus, this paper takes a direct approach as to whether "money matters" by investigating the effect of variations in transfer households with children on household spending decisions. We provide separate estimates for lone mothers, where intra-household distributional issues do not arise, from mothers with partners present, where it may. Thus, for lone mothers our results indicate the pure effect of the fact that CB is "labeled," while for couples any such effect is confounded with any intra-household distributional effect. We are particularly concerned with spending on "child goods" and use spending on children's clothing to reflect this. In contrast, we also look at how transfers to parents affect spending on "adult goods" and use alcohol, tobacco, and adult clothing as examples of these. (4)

Our headline finding is that CB is spent differently from regular income--but, paradoxically, it is spent disproportionately on adult-assignable goods. This appears to be true for single parents as well as for couples and so is not attributable to just intra-household redistribution. On the face of it, this might be interpreted as implying that mothers have a callous disregard for the welfare of their children. We resolve this paradoxical finding when we disaggregate our variation in CB into anticipated variation and unanticipated variation in CB, which we are able to do by virtue of a peculiarity of the CB system in the United Kingdom. We are motivated to disaggregate variations in CB in this way because altruism toward one's children would, in a simple model at least, imply that mothers would insure their children against surprises in income, including CB. Thus, we would expect anticipated variation in CB to have an effect on spending on child-assignable goods that is the same as other anticipated variations in income. In contrast, unanticipated changes in CB should be spent on adult-assignable goods and not at all on child-assignable goods.

These theoretical propositions are broadly supported by our empirical analysis. Our cleanest results are for lone parents where there is no intra-household distribution issue. Here the unanticipated variation in CB, driven by policy changes, is spent disproportionately on adult-assignable goods. In contrast, we find that there is no significant overall difference in the way in which this is spent, relative to other income sources. Thus, it is parents who benefit from unanticipated variation in CB--a result that is consistent with the view that mothers are altruistic toward their children and so insure them against income shocks.

Although CB is universal (i.e., not means tested), it clearly contributes to a reduction in child poverty as measured in the United Kingdom. (5) In any event, it seems plausible that lessons that we learn here from this universal program applies to means-tested transfers that are explicitly aimed at relieving child poverty. (6) CB, in 2010, was worth 20 [pounds sterling] ($30) per week for the first child, (7) and 13.20 [pounds sterling] ($20) for subsequent ones, and this has recently been joined by the CTC which is a further program that provides a tax credit for children structured in such a way that its value only falls as income rises at a level of income that is far above the mean level of household income. (8) This credit was further superseded in April 2003 by the CTC worth 10.40 [pounds sterling] (around $15) per household per week, slightly more than the Children's Tax Credit, and where the means testing starts higher up the income distribution. The total of all child-related cash benefits amounts to 2% of GDP in 2010 in the United Kingdom, about half of which is accounted for by CB and CTC, compared to 1.5% in the late 1970's, despite the dramatic fall in fertility. Indeed, the recent reforms to the welfare system have been driven by the desire to ensure that absolute cash support for children is independent of parental circumstances such as unemployment, sickness, and disability. (9)

To anticipate our findings, we show that CB is spent differently from other sources of income--but, paradoxically, we find that it is spent disproportionately on adult-assignable goods, not on child-assignable goods. (10) We resolve this paradox by making a distinction between anticipated and unanticipated variation in CB. The plan of the paper is as follows: Section II outlines the existing literature on child outcomes and parental incomes which motivates our analysis and reviews the few existing papers that investigate spending patterns; Section III summarizes our data on CB variation and on household spending patterns; Section IV provides our empirical findings that relates the two; and Section V draws the conclusion.

II. LITERATURE

Economists take it for granted that giving additional income to individuals will improve their welfare. But understanding how important giving additional income to parents is likely to be for the well-being of their children is more complex. This is because children depend on the behaviors and decisions made by their parents to determine how much, and in what way, they will benefit from additional income into the household. Most straightforwardly, parental income could be important for child outcomes because parents could use additional income to buy goods and services that are good for their children and represent an investment in their children's future well-being. Such theories of parental investment in their children have been the focus of many economists' thinking about the role of parental income in determining children's outcomes (Becker and Tomes 1986).

Recent work on spending on child and adult clothing by Kooreman (2000) for the Netherlands suggests the fact that the money is labeled as child benefit motivates households to indeed spend it disproportionately on child goods essentially because of a "mental accounting" effect. (11) That paper exploits differential variation in CB by age of child for one-child households and finds that the estimated marginal propensity to spend on child clothing is higher for CB than for other income and so argues that this is evidence of a "labeling effect." However. identification relied on a single change in the rate for young children versus older children that was almost coincident with the change in the payment mechanism. Under this reform the recipient, in the overwhelming number of cases, ceased to be the head of household and became the mother. (12) The only statistically significant finding was for one-child married couples--for larger households and for single mothers there were no significant effects of CB.

Moreover, further work on Slovenia by Edmonds (2002) found no significant effects. However, this work exploited the dependence of Slovenian CB on household income and the number of children in the previous year and so requires that these have no direct effect on current expenditure patterns--something that seems unlikely because of serial correlation in incomes, habit persistence, and the fact that changes in the number of children in the household are likely to be anticipated. (13) As in the Netherlands, UK CB over the period 1980 to 2000 was a universal (not means-tested) program, where payments depended on the current number of dependent children, went to the mother, payments were not subject to taxation, and participation was effectively 100%. (14) Thus, the United Kingdom offers an interesting laboratory to study the effect of CB because we do not have to correct for program nonparticipation. Indeed, it was this absence of selectivity that allowed Lundberg, Pollak, and Wales (1997) to investigate the impact of the UK "wallet to the purse" reform in the late 1970's. The argument for such a reform was that mothers are better agents for their children than fathers. The authors show, in grouped data, that there is an increase in spending on child clothing relative to adult clothing and female adult clothing relative to male adult clothing following the reform which gave mothers control over this source of income. (15) These findings, that household members fail to pool their resources in making spending decisions, have been echoed in other studies (16) and suggest a rejection of the unitary model of household behavior. Here, we abstract from these considerations by only using data post 1979, by which time the wallet-to-purse reform had been fully implemented, (17) and using the couples samples separately from the lone parents sample. In the latter, there is no intra-household issue, whereas in the former our estimates are conditional on it. (18)

[FIGURE 1 OMITTED]

III. DATA AND IDENTIFICATION

Our analysis covers the 21 years from 1980 (when CB had finally entirely replaced the earlier system of Family Allowances whose main beneficiaries were fathers) to 2000 (after which tax credits for parents were introduced which would complicate our analysis because these credits were means tested and were subject to a potential take-up problem). Across this period there have been wide variations in real CB within years induced by differences in inflation across years, and large changes in the real value of CB between years driven by reforms whereby CB was reflated by more or less than the inflation rate from the previous uprating. For example, a large reform occurred in 1991 whereby CB entitlement of the first child rose by a considerable amount, and a further increase for the first child occurred in 1999. Figure 1 shows the two sources of variation in real CB for first and subsequent children and for lone parents and couples separately. (19) The sawtooth shape in the 1980's clearly shows the effects of inflation--something that is not obvious in later years when inflation was considerably lower. The real reductions over the period 1984 to 1990 shows the effect of not uprating in line with price inflation in the period when the Conservative government of the day had (implicitly) adopted a policy of targeting support on the very poorest households through real rises in the generosity of the in-work welfare program for parents (then called Family Credit) at the expense of CB. In 1991, a large real rise in CB for the first child of a couple was introduced--this distinction between first and subsequent children had always been a feature of CB for lone parents (lone parents received a supplement to CB known as OPB that created this wedge between first and subsequent children) but not for couples. In a controversial change in 1997, the new Labor government abolished the OPB and so effectively eliminated this distinction between couples and lone parents. (20) However, the adverse effect on (new) lone parents was soon ameliorated when the rate for all first children was subject to a large real increase.

Until 1999, and the Labour government's commitment to abolish child poverty, the real value of CB was lower than it had been when it was first introduced in 1978 and that remained the case for the first children of lone parents and for all subsequent children in 2001, and still remains to the present. The real value of CB for the first child of lone parents had fallen by more than 10%, whereas the value for all subsequent children had fallen by more than 15%. It is only with the recent introduction of the supplement to CB known as Child Tax Credit (CTC) that the real values of child-contingent financial support enjoyed by parents back in 1979 have been matched. Our analysis relies on the real variation in CB for given household types. That is, we make no attempt to exploit the variation in CB across household types at a point in time. We do this because we do not want to rely on functional form assumptions that restrict how different numbers of children affect household spending. Moreover, we do not want to make any assumptions about the nature of intra-household distribution of income so we present estimates separately for lone parent households and couples (which include repartnered divorcees). Finally, we also decompose our data into those on in-work welfare (WFTC) and out-of-work welfare (IS) and those not. CB interacts with the latter because CB counts as income for the purposes of computing IS payments and nominal CB rises are effectively taxed at 100%--although the situation is complicated by the fact that the child-related component of IS is also increased over time. We choose not to attempt to exploit this source of variation on the grounds that it may be too subtle for consumers to detect and the group affected is, in any case, quite small. Thus, most of our analysis will be conducted over households who are not on either in-work or out-of-work welfare.

Effectively, identification comes from two sources: the variation in inflation rates across years that ensures that we can identify anticipated effects independently of seasonality (effectively we assume that the seasonality in the data is orthogonal to inflation); and from the various reforms that ensure that there are discontinuities in anticipated CB (that cannot account for smooth changes in expenditure patterns).

We use Family Expenditure Survey (FES) data on household spending patterns, which contain detailed household (21) expenditure information, constructed from two consecutive weekly diary records supplemented with information about regular payments. The expenditure data is regarded as being quite accurate with the exception of alcohol and tobacco, (22) which are under-recorded relative to other sources of information. Moreover, there is considerable consistency over time. The data also records sources of income and their levels and periodicity, and the detailed characteristics of respondent households including the number and ages of children. (23) Table 1 shows the breakdown of the data by household type. Table 2 shows some summary statistics for households with exactly one child.

IV. ECONOMETRIC ANALYSIS

In our parametric work, we test for differential marginal propensities to consume out of CB compared to other income for different commodity groups. Unlike earlier research, we model the whole of household (nonhousing) spending--both child-assignable goods as well as those that are adult-assignable and those that are not assignable at all (food, and all other nonhousing expenditure (24)). Identification relies on the sizeable real variation in CB over time--at least part of which is discontinuous arising from reforms. Because we exploit only time series CB variation, we present estimates in the body of the paper based on samples of households that contain only one child. We assume that expenditure on good i by household h is given by [e.sub.ih] = [f.sub.i] ([X.sub.h], [CB.sub.h]) + [Z.sub.h] [[beta].sub.i] + [[epsilon].sub.ih] where [x.sub.h] is household h's other income (25) (defined as total expenditure minus CB), [Z.sub.h] is vector of exogenous characteristics such as age and age squared of the household head, dummy variables to control for having a child aged 0-4 and 5-10 (relative to 11-15), region to control for regional differences in spending, and a linear trend (26) and a vector of month dummies to capture seasonal variations in spending, and [[epsilon].sub.ih] captures the unobservable determinants of spending patterns. (27)

Because each of the expenditure equations contains the same explanatory variables we estimate the system using the usual Seemingly Unrelated Regression method to allow us to test cross equation restrictions. We impose adding up in the usual manner of omitting one arbitrary equation. We omit all other expenditure apart from the assignable ones (male, female, and child clothing, alcohol, and tobacco) and food so just six equations are reported.

In our parametric analysis discussed later, we further assume that [f.sub.i]([x.sub.h], [CB.sub.h]) is linear and additively separable. Linearity here is unlikely to be important--we are estimating a local approximation around the mean of total expenditure and the effect of CB is, itself, small variation around that mean. The specification follows earlier research by Kooreman (2000) and Edmonds (2002) who estimate simple specifications where expenditure on each good is assumed to be a linear function of CB and of total expenditure less CB. To ensure that our results are as robust as possible we select relatively homogenous samples to minimize the importance of Z. Our objective is to test whether [f.sub.i]([x.sub.h], [CB.sub.h]) is such that child benefit has the same effects on expenditures as total expenditure CB does--we refer below to this latter effect as the Engel curve slope. (28) We estimate separate systems for couples and lone parents. (29) We are particularly interested in this distinction for two reasons. Firstly, the single parents sample is immune from the problem that there may be an intra-household pooling issue which might cause CB, which is given to mothers, to have different effects from other sources of income because, in the case of lone mothers, all sources of income are at the disposal of the mother. Secondly, if underinvestment in child quality arises from each parent free-riding on the other, then this would be reflected in the behavior of couples and not in that of lone mothers.

A. Benchmark Results

The benchmark results are shown in Table 3, which provides estimates using the couples and lone parents data for those with one child aged under 16, (30) who are not on welfare. (31) The assignable goods equations and the food equation are presented (the residual spending equation is not presented and the estimates are independent of the excluded equation). The coefficients show the effect of 1 [pounds sterling] of CB and of other income on spending on each good. The key result here is that it is alcohol spending that changes when CB changes with a marginal propensity of 0.49 for couples (0.21 for single parents)--much larger than the marginal propensity to spend on alcohol from other income. For lone mothers we find that there is a significant effect (0.71) on adult women's clothing. In the case of couples the CB effect on alcohol (and for lone parents the effect of CB on mother's clothing) is more than ten times larger than the Engel curve slope. The [chi square] and p-statistics test for the restriction that marginal propensity to spend out of CB income is the same as that out of other income (defined as total expenditure minus CB). The restriction that the marginal propensities to spend out of CB and other income are the same is rejected for alcohol in the couples sample, and for women's clothing in the lone parent sample. The overall [chi square] and p values test the restrictions, across all goods, that the effects of CB and other expenditure are the same. We strongly reject this restriction for couples although the value for lone parents is not quite significant. (32)

B. Robustness of Benchmark Results

Infrequency of purchase is clearly an issue in short-survey datasets. This gives rise to a measurement error problem that would lead to biased estimates. Keen (1986) shows that this can be resolved by instrumenting total expenditure, and here we use total household income as an instrument. Moreover, alcohol is well known to be under-reported in survey data. Because alcohol is a component of total expenditure then this would normally give rise to the other income coefficient being biased toward zero. Under-reporting of spending on any good induces nonclassical measurement error in total expenditure and, because of adding up it seems likely that bias will affect all equations. There do not appear to be any analytical results of the effects of this sort of measurement issue in the literature and there are no strong a priori grounds for thinking the bias should be systematically in one direction. (33)

The results are reported in Table 4. In comparison with Table 3 there are some changes in magnitude but there is no change in the pattern or significance of results. In Table 5, we re-estimate using Tobit to allow for the zeroes in the expenditures. There is no change, relative to Table 3, for couples but for lone parents the result for women's clothing becomes insignificant while alcohol becomes larger and significant. Thus, it seems unlikely that our results are driven by measurement error. If anything, our instrumental variable and Tobit results strengthen our conclusion from Table 3.

The identification of the CB coefficients in Table 3 derives entirely from the time series variation. Although the real value of CB does not exhibit a time trend (and, in any event our modeling includes both a linear trend and a set of month controls), we first test for the robustness of the results in Table 3 by re-estimating over the 1980's data (1980-1989) separately from the 1990's (1990-2000) data. These results are presented in Table 6 for the 1980's and the 1990's separately. The results in Table 3 for the pooled data over the whole period are confirmed--with alcohol being the source of rejection for couples--men's clothing in the latter period, and women's clothing being the problem for lone mothers but only in the 1990's. In Table 7, we re-estimate for subsamples of mothers with different levels of education: left school at 16 (the minimum), at 17/18, or 19+. Our conclusion remains: couples reject through alcohol, whereas lone mothers reject through mother's clothing.

Table 8a and 8b divides the samples into the top, middle, and bottom thirds of the respective income (total expenditure) distributions because one might be concerned that Engel curves are nonlinear. Again the headline results are broadly confirmed: all but the bottom third of couples significantly reject because of alcohol; while the top third of the lone mothers reject because of women's clothing. Even for the bottom third the alcohol and women's clothing coefficients on CB are much larger than the respective other income coefficients, albeit not significant.

Table 9 replicates Table 3 but uses only the data for children under 11. We do this in case the benchmark results are contaminated by the possibility that parents may be wearing child clothing. (34) The strong results for couples remain although the precision of the lone mothers sample falls sufficiently that the effects become insignificant. Nevertheless, the sizes of the coefficients for lone mothers are comparable with Table 3.

C. Anticipated and Unanticipated Variation

Despite the weight of evidence here that suggests that variations in CB are reflected in adult-assignable, and not in spending on child-assignable, goods it would be inappropriate to conclude that the lack of equivalence between CB and other income implies that parents put less weight on the welfare of their children than on their own so that, at the margin, they favor expenditure on adult goods. Rather, an alternative explanation would be that parents may place so much weight on the welfare of their children that they fully insure them against income variations so that, at least unanticipated, variation in incomes does not affect spending on the children.

Suppose the simplest case where all goods are exclusive to either adults or children and the utility function of the altruistic parent is defined as [V.sub.a] (y - x) + [delta][V.sub.c] (x + b) where [delta] > 0 indicates altruism, y is the household income (assumed to be the adult's (a)), x is the transfer from parent to child (c), b is a transfer from the government to the child. Differentiating with respect to x shows that the equilibrium transfer to the child is such that [[lambda].sub.a] = [delta][[lambda].sub.a] (for an interior optimum where some positive transfer takes place), where the [lambda]'s are the respective marginal utilities of income. The optimal transfer, [x.sup.*], is such that it would be the same if the welfare transfer, b, had been made to the parent rather than the child. (35) In the case where b is uncertain it is useful to consider a simple benchmark case of [V.sub.a] and [V.sub.c] being CRRA functions of y - x and x + b, respectively. In this case, the 35. See Bergstrom (1989) for discussion of Becker's rotten kid theorem. optimality property allows us to solve for x in terms of b. As before, the optimal [x.sup.*] depends on the value of b, but the size of the effect of b on x now depends on the ratio of the degrees of relative risk aversions and the extent of altruism. Only if the parents are sufficiently risk averse with respect to the child's consumption, relative to their own consumption, and altruism is sufficiently large, will x vary inversely with b. In general, parents will not fully insure their children unless they themselves are risk neutral.

There is some qualitative evidence that suggests that parents (especially mothers) are likely to "go without" to protect spending on their children in the face of adverse shocks. (36) To investigate this issue we assume that households form static expectations of real CB. That is, we assume that households expect the government to reinstate the real level of CB to the value at the previous uprating date 1 year ago by an appropriate increase in the nominal CB. We assume that, between uprating dates, households form rational expectations about the price level and so anticipated real CB falls according to the actual inflation rate. That is, we assume that households assume that CB will be indexed in line with inflation because the last increase--and so have static expectations of policymakers. Thus, we decompose real child benefit according to the following formula:

[CB.sup.a.sub.ym] = [CB.sub.y - 12]/([P.sub.y-m]/[P.sub.y-12])

where [CB.sup.a.sub.ym] is the level of child benefit that would be anticipated in year y some m months after the uprating, [CB.sub.y-12] is the nominal value of CB at the last uprating and [P.sub.y-m]/[P.sub.y-12] is the inflation adjustment over the last m months since the uprating. This captures the variation in CB arising from the inflation that has occurred since the last uprating. The difference between actual CB and anticipated CB captures the change in CB that has occurred because of the nominal uprating that last occurred--which we assume is unpredictable and call unantici pated CB, [CB.sup.a.sub.ym]. We allow for there to be a differential effect of these two components by writing our Engel curves as

[e.sub.ih] = [[alpha].sub.i][CB.sup.a] + [[gamma].sub.i][CB.sup.u] + [[eta].sub.i][M.sup.h] + [Z.sub.h][[beta].sub.i] + [[epsilon].sub.ih]

where M is other expenditure which we assume is driven by long-run differences between households arising from skills differences. The results are reported in Table 10 in the case where we assume that expectations of inflation are formed rationally.

The anticipated CB effects are generally badly determined and therefore are not significantly different from the coefficients on other expenditure. This is reassuring: nominal CB shocks associated with the annual changes only have a temporary impact on spending on adult goods. Thereafter, the CB becomes part of permanent income and is spent like other permanent components of income. However, the unanticipated CB effects are consistent with our earlier results and with the interpretation that parents do insure their children against shocks so that unanticipated CB is spent disproportionately on adult goods. For couples, spending on alcohol out of unanticipated CB is significantly different from spending out of other income. For lone parents the same is true for both alcohol and women's clothing. The F- and p-statistics show that in the couples' sample the restriction that the marginal propensity to consume out of unanticipated CB is the same as that out of other income jointly for all equations is strongly rejected. However, the same restrictions cannot be rejected in the lone parent sample because of a smaller sample size and a lack of precision.

V. CONCLUSIONS

Our analysis finds that unanticipated variation in CB that is driven by policy induced changes in its real value is disproportionately spent on adult-assignable goods. The results for couples suggest that, at the margin, as much as a half of unanticipated changes in CB is spent on alcohol. The results for lone parents are less strong but nonetheless still apparent. These findings contrast with those of Kooreman (2000), which exploits variation in Dutch CB, and of Edmonds (2002), based on data from Slovenia. However, this earlier work made no distinction between anticipated and unanticipated variation in CB.

A weakness of this line of research is that it is unclear what inferences can be drawn from an equivalence (or lack of it) between CB and other income. One might be tempted to conclude that CB is treated differently because there is something different about it. For example, CB is usually given to the mother so that a lack of equivalence may suggest imperfect pooling of household incomes. However, our results are also true for lone parents where there is no intra-household distributional issue, so this cannot account for all of this lack of equivalence. It is true that the effect for lone parents is less pronounced, the alcohol coefficient for CB is around half the size as in the couples samples, and this is consistent with the idea that there is some free-tiding between partners which does not occur in single parent households. A second issue might be that real CB variation tracks the business cycle implying that our results are attributable to cyclical effects in spending. However, we find no such cyclical effects in the spending patterns of households without children and there is little reason to expect households with children to differ in this respect.

Finally, a simple but important innovation in this work has been to distinguish between anticipated and unanticipated variation in CB. We find that it is unanticipated CB variation that is reflected in adult-assignable good expenditure suggesting that parents are successful in providing at least some insurance for their children. This finding suggests a high degree of altruism on the part of parents. The implication is that CB may simply finance spending on children that would have otherwise occurred.

ABBREVIATIONS

CB: Child Benefit

CRRA: Constant Relative Risk Aversion

CTC: Child Tax Credit

DWP: Department of Work and Pensions

FES: Family Expenditure Survey

GDP: Gross Domestic Product

GNP: Gross National Product

JSA: Job Seeker's Allowance

IS: Income Support

OPB: One Parent Benefit

TANF: Temporary Assistance for Needy Families

WFTC: Working Families' Tax Credit

REFERENCES

Adam, S., and M. Brewer. Supporting Families: The Financial Costs and Benefits of Children since 1975. York, UK: Joseph Rowntree Foundation, Policy Press, 2004.

Becker, G. S., and N. Tomes. "Human Capital and the Rise and Fall of Families." Journal of Labor Economics, 4, 1986, S1-S39.

Bergstrom, T. C. "A Fresh Look at the Rotten Kid Theorem--And Other Household Mysteries." Journal of Political Economy, 97, 1989, 1138-59.

Bingley, P., and I. Walker. 1997, "There's No Such Thing as a Free Lunch: Evidence of Altruism from In-kind Transfers to Children." IFS Working Paper 97/07.

Blundell. R. W., and I. Walker. "A Lifecycle Consistent Model of Family Labour Supply." Review of Economic Studies, 53, 1986, 539-58.

Blundell, R. W., P. Pashardes, and G. Weber. "What Do We Learn About Consumer Demand Patterns from Micro Data?" American Economic Review, 83, 1983. 570-659.

Bourguinon, F, M. Browning, P.A. Chiappori, and V. Lechene. "Intra Household Allocation of Consumption: A Model and Some Evidence from French Data." Annales d'Economie et Statistiques, 29, 1993, 137-56.

Bradbury, B. "Consumption and Within-Household Income Distribution: Outcomes for an Australian Natural Experiment." Economic Studies, 50(3), 2005, 501-540.

Bradshaw, J., and N. Finch. "A Comparison of Child Benefit Packages in 22 Countries." Department of Work and Pensions Research Report 174, 2002. Accessed November 13, 2010. http://research.dwp.gov.uk/asd/ asd5/rrep174.pdf.

Brewer, M. "The New Tax Credits," IFS Briefing Note 35, 2003.

Burman, L. E., and L. Wheaton. "Who Gets the Child Tax Credit?" Washington, DC: Urban Institute, 2005. http://www.urban.org/url.cfm?ID=411232.

Edmonds, E. V. "Reconsidering the Labeling Effects of Child Benefits: Evidence from a Transitional Economy."

Economics Letters, 76, 2002, 303-09.

Farrell, C., and W. O'Connor. "Low-Income Families and Household Spending." Department of Work and Pensions, Research Report 192, 2003.

Greener, K., and R. Cracknell. "Child Benefit." House of Commons Research Paper 98/79, 1998.

Gregg, P., J. Waldfogel, and E. Washbrook. "That's the Way the Money Goes: Expenditure Patterns as Real Incomes Rise for the Poorest Families with Children," in A More Equal Society? edited by J. Hills and K. Stewart. York, UK: Joseph Rowntree Foundation, 2004.

Hotchkiss, J. L. "Do Husbands and Wives Pool Their Resources?" Journal of Human Resources, 40, 2005, 519-31.

Jacoby, H. "Is There an Intrahousehold "Flypaper Effect"? Evidence from a School Feeding Program." Economic Journal, 112, 2002, 196-221.

Keen, M. "Zero Expenditures and the Estimation of Engel Curves." Journal of Applied Econometrics, l, 1986, 277-86.

Kooreman, P. "The Labeling Effect of a Child Benefit System." American Economic Review, 90, 2000, 571-83.

Lundberg, S., R. A. Pollak, and T. J. Wales. "Do Husbands and Wives Pool Their Resources?" Journal of Human Resources, 32, 1997, 463-80.

Micklewright, J. "Child Poverty in English-Speaking Countries." IZA Working Paper 1113, 2004.

Middleton, S., K. Ashworth, and I. Braithwaite. Small Fortunes. York, UK: Joseph Rowntree Foundation, 1997.

Phipps, S. A., and P. S. Burton. "What's Mine is Yours? The Influence of Male and Female Incomes on Patterns of Household Expenditure." Economica, 65, 1998, 599-613.

Schluter, C., and J. Wahba. "Are Poor Parents Altruistic?: Evidence from Mexico." Journal of Population Economics, 23, 2008, 1153-74.

Tanner, S. "How Much do Consumers Spend? Comparing the FES and National Accounts," in How Reliable Is the Family Expenditure Survey: Trends in Incomes and Expenditures, edited by J. Banks and P. Johnson. London: IFS, 1998.

Thaler, R. "Savings, Fungability and Mental Accounts." Journal of Economic Perspectives, 4, 1990, 193-205.

UNICEF. Poverty Reduction Begins with Children. New York: UN, 2000.

Ward-Batts, J. "Out of the Wallet and Into the Purse: Modeling Family Expenditures to Test Income Pooling." Journal of Human Resources, 43, 325-51.

(1.) See Burman and Wheaton (2005).

(2.) See Bradshaw and Finch (2002) for details of 22 countries.

(3.) For international comparisons, see Micklewright (2004) and UNICEF (2000).

(4.) Our analysis is one of a complete demand system where we impose the adding up condition. Thus there is an excluded category of expenditure whose coefficients are implicit.

(5.) Child poverty is measured by counting the number of children in households whose incomes fall below 60% of the median of the overall household income distribution so that it is a relative measure. Thus, a reform that increases the incomes of households with children through CB and leaves other households unaffected must decrease child poverty even though it is not a means-tested transfer. The UK finance minister, then Gordon Brown, described child poverty as "a scar on the soul of Britain" in a speech at the 1999 SureStart Conference. He went on to promise that increases in CB under the Labour government were part of "immediate and direct action" to provide "cash help to lift children out of poverty".

(6.) Income Support (IS) and Job Seekers' Allowance (JSA), the out-of-work welfare programs (mainly for poor lone parents, the disabled, and the unemployed), have also benefited from increasingly generous additions for dependent children, as has Working Families' Tax Credit (WFTC), the main in-work welfare program.

(7.) Prior to 2007 there was a higher rate of CB paid to lone parents, called One Parent Benefit (OPB). This was fixed in nominal terms in 2000 and inflation between 2000 and 2007 closed the differential.

(8.) WFTC and Children's Tax Credit has recently been replaced by Working Tax Credit (WTC) and CTC but they broadly retain their earlier structure (Brewer 2003). In contrast to the extensive cash support for children in the United Kingdom and the relative unimportance of means-testing, the United States, until recently, relied heavily on in-kind transfers such as food stamps, targeted nutrition schemes such as the school breakfast program, the health care cover provided by MedicAid, and Temporary Assistance for Needy Families (TANF) which typically provides extensive childcare support but rather little explicit cash. Indeed, the cash that is provided is time limited.

(9.) See Adam and Brewer (2004) for a review of the development of all UK child-related benefits including CB.

(10.) Remember that our result is for lone mothers so has nothing to do with intra-household distributional issues as in Lundberg, Pollak, and Wales (1997).

(11.) See Thaler (1990) for why this phenomenon might exist and why it leads to differences in marginal propensities to consume out of different forms of income.

(12.) Thus. the paper places some weight on the presumption that this 'wallet-to-purse'" transfer had an equal effect on spending patterns across households with different aged children. Since maternal market labor supply may be affected by the intra-housebold transfer this seems unlikely.

(13.) Jacoby (2002) investigates in-kind (food) transfers targeted on children and finds no evidence of a "flypaper" effect of such transfers increasing the calorific intake of the children. Bingley and Walker (1997) consider the effects of giving food and milk to children on household spending patterns--we find significant effects on household milk spending. Schluter and Wahba (2008) examine the effects on household spending patterns of the Mexican Progressa experiment whereby schooling subsidies were randomly assigned. They show significant effects of the subsidy on child clothing expenditure which they interpret as altruistic behavior. However. the subsidy is conditional on attending schools and it seems likely that this conditionality affects how the money is spent--for example, attending school may itself have an impact on clothing needs.

(14.) Private correspondence with DWP confirms that this also applies to the supplement to CB that is paid to lone parents--OPB.

(15.) However, analyses of the microdata by Ward-Baits (20001 and Hotchkiss (2005) cast doubt on the original conclusions. Limited studies exist elsewhere: for example, Bradbury (2004) analyzes a similar natural experiment in Australia and finds no effect of the redistribution. Here we abstract from this issue entirely by concentrating on the spending patterns of single parent households. We intend to revisit the intra-household distributional issue that Lundberg et al. investigated because there is significant variation in the level of CB paid to mothers in couples that could be very informative.

(16.) See Phipps and Burton (1998) and Bourguignon et al. (1993), for example. However, these studies simply examine whether spending patterns are affected by the individual composition of household income without regard to the potential endogeneity of that composition.

(17.) Our data record who receives the CB in the household: the proportion of two-parent households where the mother is the recipient is 99.1%.

(18.) More recently Gregg. Waldfogel, and Washbrook (2004) have described how patterns of spending have changed between 1996/97 and 2000/01, for low income households relative to other households as their relative disposable incomes varied (for a variety of reasons, not just CB). They find that spending on alcohol and tobacco for low income households with children relative to those with higher income has fallen, and that spending on toys, games, and clothing and footwear has risen. However, their analysis takes no account of changing composition of the low income group relative to the rest--which will have been dramatic because of the large change in in-work welfare entitlements that occurred in 1999, the introduction of the National Minimum Wage in 1999, and the unfolding New Deals, especially for lone parents, all of which will have contributed to a reduction in worklessness among this low income group of parents. Moreover, there will have been cyclical effects that have more pronounced effects on the bottom of the distribution than the rest.

(19.) See Greener and Cracknell (1998) for the historical background and development of CB in the United Kingdom.

(20.) Lone parents who were already in receipt of OPB prior to 1997 were allowed to retain it.

(21.) Spending data at the individual level is not available in the public use files. However, since 1995 the data has separately recorded the expenditure of all children aged 7-15.

(22.) See Tanner (1998) for an analysis of the reliability of FES expenditure data. The deficiency in the alcohol and tobacco categories is thought to be largely associated with differential response rates of smokers and drinkers and not because of under-recording by respondents. We find no evidence that under-recording is correlated with the real variation in CB.

(23.) We drop all households where the youngest child is 16 and over because the FES treats the clothing of children aged 16 and over as adult clothing. We also exclude multiple benefit unit households so that our sample consists of "nuclear" families only.

(24.) This latter is the excluded category. Homogeneity of demands would allow us to recover the parameters of this excluded category form the parameters estimated assuming that adding-up holds. The estimates are guaranteed to be independent of which commodity forms the excluded category.

(25.) We use total expenditure (minus CB) as our explanatory variable rather than income. This is to ensure consistency with an intertemporally separable lifecycle maximizing model (Blundell and Walker, 1986). Results using total (net of tax and welfare) income (minus CB) are essentially the same and are available on request.

(26.) We included a cubic spline in month of survey to capture smooth changes in tastes but were able to reject this in favor of a simple linear trend.

(27.) Estimates which include relative prices are available on request. We do not control for relative prices here because when we tested for the time series correlation between CB and monthly relative prices we found an insignificant partial correlation of only 0.088. Including relative prices does not affect our estimates in any way apart from slightly increasing their precision.

(28.) We experimented with nonlinear Engel curves. For example, we found that when we entered CB and other expenditure quadratically the marginal effects, evaluated at the means, were essentially unchanged. In any event we do go on to provide estimates for subsets of the data broken down by income and find that our main results carry over to each subset of the income distribution.

(29.) We refrain from using childless households in the analysis because they are uninformative about the question at hand. They clearly cannot be used to estimate the effect of CB. Moreover, while they can be used to estimate the Engel curve slope for adult goods, they cannot he used to estimate the Engel curve slopes for child-assignable goods since there is no such expenditure in childless households. Although they can be used to estimate the Engel curve slopes for adult-assignable goods, such estimates would not be comparable with the estimates for household with children since "adding up" (i.e., homogeneity of degree zero) is imposed across a smaller set of goods.

(30.) Results restricted to children under 11 are almost identical.

(31.) We investigated the sensitivity to including welfare recipients in the samples. For welfare recipients CB counts as income when computing other welfare payments to households. Thus, we do not expect any effect of CB in such households and this is, indeed, what we do find.

(32.) Clearly part of the variation in real CB arises because of differential inflation rates across years. There is a possibility that the differential effect on spending patterns is due to business cycle effects that are correlated with inflation and not adequately controlled in the model by the inclusion of total expenditure. If the variation in the expenditures of households with children was being affected by the business cycle rather than by real CB variation then we would expect the same to be true of households without children. We investigated this by looking at the correlation matrix between expenditures and inflation for both singles and couples without children. We found no correlation. Thus. we feel that our results are not contaminated by omitted business cycle effects.

(33.) This instrument is commonly used in demand system estimation (Blundell, Pashardes, and Weber, 1993). The absence of any analytical results of the effects of this sort of measurement issue in the literature prompted us to simulate some data with varying degrees of under-recording and our consequent estimates (not shown here but available from the authors upon request) suggest that the basic findings still hold, even with substantial degrees of under-reporting (e.g., with up to half of households under-reporting true alcohol expenditure by 50% on average).

(34.) Although there is a sales tax distinction between adult and child clothing that is defined by sizes, the FES clothing data is self-reported as child or adult.

(36.) Two recent examples of such work are Middleton, Ashworth, and Braithwaite (1997) and Farreli and O'Connor (2003). However. the datasets used in these studies are small and formal hypothesis tests are not conducted. Indeed, such qualitative research makes no attempt to distinguish between anticipated and unanticipated variation in income in any very formal way. Thus. the work here complements that qualitative research.

Online Early publication December 15, 2010

LAURA BLOW, IAN WALKER and YU ZHU *

* This research was funded by Her Majesty's Treasury's Evidence Based Policy Fund with the support of the Department for Education and Skills, Department of Work and Pensions, Department of Culture, Media and Sports, and the Inland Revenue. The work was completed while Walker was a visiting fellow at Princeton University, funded by the Leverhulme Trust. Material from the Family Expenditure Survey is Crown Copyright and has been made available by the Office of National Statistics through the Economic and Social Research Council's Data Archive at the University of Essex. The data is used with the permission of Her Majesty's Stationery Office but can be made available to other researchers under conditions imposed by the Archive. We are grateful for comments from an anonymous referee, Mike Brewer at the Institute for Fiscal Studies, Mike Bielby, Ilona Blue, and Mario Pisani of the Inland Revenue. Andrew Oswald at Warwick, and seminar participants at the Department for Education and Skills, the Department of Work and Pensions, the University of Warwick, the University of Kent, the Cardiff Business School, and Queen Mary and Westfield College of the University of London. The opinions expressed in this paper are those of the authors.

Blow: Senior Research Economist, Institute for Fiscal Studies, London, UK. Phone +44-20 7291 4800, Fax +44-20 7323 4780, E-mail l.blow@ifs.org.uk

Walker: Professor, Lancaster University Management School, Lancaster, UK. E-mail ian.walker@lancaster.ac.uk

Zhu: Senior Lecturer, School of Economics, University of Kent, Canterbury, Kent, UK. Phone +44-1227 827438, Fax +44-1227 827850, E-mail yz5@kent.ac.uk

doi: 10.1111/j.1465-7295.2010.00348.x

TABLE 1
Summary Statistics: Household Types (Numbers and Proportions)

 1 Child

 Married Lone All

Not on Welfare 8,575 744 9,319
 0.87# 0.25# 0.73#

On Out-of-Work Welfare 948 1,836 2,784
 0.10# 0.63# 0.22#

On In-Work Welfare 288 340 628
 0.03# 0.12# 0.05#

Total 9,811 2,920 12,731

 2 Children

 Married Lone All

Not on Welfare 12,967 570 13,537
 0.88# 0.25# 0.80#

On Out-of-Work Welfare 1,255 1,453 2,708
 0.09# 0.65# 0.16#

On In-Work Welfare 441 216 657
 0.03# 0.10# 0.04#

Total 14,663 2,239 16,902

 3+ Children

 Married Lone All Total

Not on Welfare 4,502 165 4,667 27,523
 0.76# 0.16# 0.67# 0.75#

On Out-of-Work Welfare 1,000 783 1,783 7,275
 0.17# 0.76# 0.26# 0.20#

On In-Work Welfare 422 81 503 1,788
 0.07# 0.08# 0.07# 0.05#

Total 5,924 1,029 6,953 36,586

Note: Figures in italics are column proportions.

Note: Figures indicated with # are column proportions.

TABLE 2
Summary Statistics: Expenditure Patterns for Households with 1
Child, Weekly Amounts ([pounds sterling]) and Standard Deviations

 Couples

 Not on On Out-of-Work
 Welfare Welfare

Child Clothing Expenditure 7.72 4.51
 (12.49) (8.90)
 Positive exp 61.94 52.95
 Expenditure|exp > 0 12.46 8.52
 (13.89) (10.75)

Women's Clothing Expenditure 10.24 4.25
 (19.08) (10.09)
 % Positive exp 61.17 41.98
 Expenditure|exp > 0 16.75 10.12
 (22.06) (13.536)

Men's Clothing Expenditure 6.64 3.08
 (18.34) (8.88)
 % Positive exp 36.00 24.68
 Expenditure|exp > 0 18.45 12.5
 (26.77) (14.24)

Food Expenditure 68.06 46.51
 (27.83) (18.69)
 % Positive exp 99.97 100.00
 Expenditure|exp > 0 68.08 46.51
 (27.81) (18.69)

Alcohol Expenditure 14.63 9.03
 (19.02) (14.05)
 % Positive exp 83.78 65.4
 Expenditure|exp > 0 17.47 13.81
 (19.55) (18.69)

Tobacco Expenditure 7.15 12.06
 (10.86) (11.8)
 % Positive exp 46.33 73.52
 Expenditure|exp > 0 15.43 16.41
 (11.26) (10.87)

Child Benefit Expenditure 11.38 11.44
 (1.91) (1.60)
 % Positive exp 100.00 100.00
 Expenditure|exp > 0 11.38 11.44
 (1.91) (1.60)

All Other Expenditure 298.30 157.03
 Expenditure (181.7) (96.07)
 % Positive 100.00 100.00
 Expenditure|exp > 0 298.30 157.03
 (181.7) (96.07)

Household Income Expenditure 395.59 213.77
 (239.5) (133.20)
 % Positive exp 99.93 100.00
 Expenditure|exp > 0 395.92 213.77
 (239.32) (133.2)
 No Obs 8,575 948

 Couples

 On In-Work
 Welfare Total

Child Clothing Expenditure 5.23 7.34
 (8.49) (12.13)
 Positive exp 56.25 60.90
 Expenditure|exp > 0 9.29 12.05
 (9.51) (13.60)

Women's Clothing Expenditure 4.59 9.50
 (8.62) (18.28)
 % Positive exp 45.83 58.86
 Expenditure|exp > 0 10.02 16.14
 (10.40) (21.46)

Men's Clothing Expenditure 3.38 6.20
 (8.93) (17.47)
 % Positive exp 29.86 34.73
 Expenditure|exp > 0 11.31 17.86
 (12.36) (25.90)

Food Expenditure 52.70 65.52
 (22.51) (27.77)
 % Positive exp 100.00 99.97
 Expenditure|exp > 0 52.70 65.84
 (22.51) (27.75)

Alcohol Expenditure 9.51 13.94
 (14.28) (18.56)
 % Positive exp 71.88 81.65
 Expenditure|exp > 0 13.23 17.07
 (15.32) (19.20)

Tobacco Expenditure 10.98 7.74
 (12.23) (11.11)
 % Positive exp 64.24 49.49
 Expenditure|exp > 0 17.10 15.64
 (11.33) (11.21)

Child Benefit Expenditure 12.16 11.41
 (1.93) (1.89)
 % Positive exp 100.00 100.00
 Expenditure|exp > 0 12.16 11.41
 (1.93) (1.89)

All Other Expenditure 188.71 281.43
 Expenditure (106.0) (179.1)
 % Positive 100.00 100.00
 Expenditure|exp > 0 188.71 281.43
 (106.0) (179.1)

Household Income Expenditure 245.47 373.62
 (123.9) (235.9)
 % Positive exp 100.00 99.94
 Expenditure|exp > 0 245.47 373.89
 (123.9) (235.81)
 No Obs 288 9,811

 Lone Parents

 Not on On Out-of-Work
 Welfare Welfare

Child Clothing Expenditure 9.16 4.81
 (15.78) (8.35)
 Positive exp 55.91 53.65
 Expenditure|exp > 0 16.38 8.97
 (18.09) (9.63)

Women's Clothing Expenditure 11.65 4.01
 (23.75) (8.80)
 % Positive exp 58.60 42.76
 Expenditure|exp > 0 19.88 8.39
 (28.28) (11.44)

Men's Clothing Expenditure 1.85 0.59
 (8.91) (4.75)
 % Positive exp 11.69 5.99
 Expenditure|exp > 0 15.78 9.90
 (21.51) (16.91)

Food Expenditure 46.93 30.34
 (20.95) (13.96)
 % Positive exp 100.00 100.00
 Expenditure|exp > 0 46.93 30.34
 (20.95) (13.96)

Alcohol Expenditure 6.54 2.61
 (9.96) (5.14)
 % Positive exp 66.94 43.74
 Expenditure|exp > 0 9.78 5.96
 (10.80) (6.36)

Tobacco Expenditure 4.61 6.67
 (7.42) (7.61)
 % Positive exp 37.77 60.29
 Expenditure|exp > 0 12.22 11.07
 (7.27) (6.89)

Child Benefit Expenditure 15.91 14.93
 (3.35) (3.55)
 % Positive exp 100.00 100.00
 Expenditure|exp > 0 15.91 14.93
 (3.35) (3.55)

All Other Expenditure 205.34 90.52
 Expenditure (158.3) (54.47)
 % Positive 100.00 100.00
 Expenditure|exp > 0 205.34 90.52
 (158.3) (54.47)

Household Income Expenditure 254.97 117.09
 (208.9) (55.65)
 % Positive exp 99.87 100.00
 Expenditure|exp > 0 255.32 117.09
 (208.92) (55.65)
 No Obs 744 1,836

 Lone Parents

 On In-Work
 Welfare Total

Child Clothing Expenditure 6.60 6.13
 (10.47) (11.11)
 Positive exp 55.59 54.45
 Expenditure|exp > 0 11.88 11.25
 (11.60) (13.00)

Women's Clothing Expenditure 7.34 6.35
 (15.48) (15.20)
 % Positive exp 54.12 48.12
 Expenditure|exp > 0 13.57 13.19
 (18.95) (19.74)

Men's Clothing Expenditure 0.99 (1.96
 (3.71) (6.02)
 % Positive exp 11.47 8.08
 Expenditure|exp > 0 8.62 11.86
 (7.43) (17.89)

Food Expenditure 38.53 35.52
 (16.39) (17.79)
 % Positive exp 99.71 99.87
 Expenditure|exp > 0 38.64 35.53
 (16.28) (17.78)

Alcohol Expenditure 4.80 3.86
 (7.83) (7.21)
 % Positive exp 59.12 51.44
 Expenditure|exp > 0 8.12 7.51
 (8.76) (8.57)

Tobacco Expenditure 6.15 6.09
 (7.84) (7.64)
 % Positive exp 52.35 53.63
 Expenditure|exp > 0 11.75 11.35
 (7.18) (7.00)

Child Benefit Expenditure 17.12 15.44
 (2.51) (3.47)
 % Positive exp 100.00 100.00
 Expenditure|exp > 0 17.12 15.44
 (2.51) (3.47)

All Other Expenditure 136.47 125.12
 Expenditure (72.59) (106.1)
 % Positive 100.00 100.00
 Expenditure|exp > 0 136.47 125.12
 (72.59) (106.1)

Household Income Expenditure 178.79 159.41
 (55.71) (130.1)
 % Positive exp 100.00 99.97
 Expenditure|exp > 0 178.79 159.46
 (55.71) (130.08)
 No Obs 340 2,920

TABLE 3
Estimated Effects of 1 [pounds sterling] of CB and 1 [pounds
sterling] of Other Income on Spending on Each Good: Parents with
One Child Not on Welfare, 1980-2000

Explanatory Variables Child Women's Men's
 Clothing Clothing Clothing

Couples, N = 8.575

 CB 0.014 0.213 0.196
 (0.2) (1.9) (1.8)

 Other Expenditure 0.017 0.039 0.028
 (22.8) (34.7) (24.9)

 [chi square] (CB=Other exp) 0.00 2.44 2.30

 p .97 .12 .13

 Overall [chi square] (6) = 26.87
 p = .0002#

Lone Parents, N = 744

 CB 0.154 0.706# 0.074
 (0.9) (2.9) (0.8)

 Other Expenditure 0.025 0.064 0.007
 (6.6) (12.4) (3.1)

 [chi square] (CB=Other exp) 0.54 7.11# 0.47

 p .46 .01# .49

 Overall [chi square] (6) = 11.81
 p = .0664

Explanatory Variables
 Food Alcohol Tobacco

Couples, N = 8.575

 CB -0.188 0.491# -0.005
 (1.3) (4.3) (0.1)

 Other Expenditure 0.075 0.033 0.000
 (51.8) (28.9) (0.6)

 [chi square] (CB=Other exp) 3.32 16.47# 0.01

 p .07 .00# .94

 Overall [chi square] (6) = 26.87
 p = .0002#

Lone Parents, N = 744

 CB -0.096 0.212 0.009
 (0.5) (2.1) (0.1)

 Other Expenditure 0.067 0.019 0.001
 (16.0) (8.6) (0.4)

 [chi square] (CB=Other exp) 0.68 3.51 0.01

 p .41 .06 .92

 Overall [chi square] (6) = 11.81
 p = .0664

Notes: Figures in parentheses are absolute t-values. Other
explanatory variables are: a linear trend; month, region, and
dummy variables for whether the child was aged 0-4, 5-10
(relative to I1-15); a quadratic in age of household head; and a
lone father dummy in the lone parent sample. The F statistic in
each equation is a test of whether the coefficient on CB and on
other income is equal, and the overall F statistic is a test that
all of the CB coefficients equal the corresponding other income
coefficients. Bold values indicate statistical significance at 5%
level.

Note: Values with # indicates statistical significance at 5% level.

TABLE 4
Instrumental Variable Estimates of Engel Curves: One Child Not on
Welfare, 1980-2000

Explanatory Variables Child Women's Men's
 Clothing Clothing Clothing

Couples, N = 8,560

 CB -0.011 0.152 0.158
 (0.1) (1.3) (1.4)

 Other Expenditure 0.006 0.014 0.008
 (10.6) (15.9) (8.5)

 [[chi square].sub.CB 0.05 1.38 1.72
 = Other exp)]

 p .83 .24 .19

 Overall [chi square] (6) = 25.72
 p = .0003#

Lone Parents, N = 738

 CB 0.169 0.712# 0.076
 (0.9) (2.7) (0.8)

 Other Expenditure 0.006 0.023 0.002
 (2.1) (5.6) (1.5)

 [[chi square].sub.CB 0.81 6.95# 0.54
 = Other exp)]

 p .37 .01# .46

 Overall [chi square] (6) = 12.52
 p = .0513

Explanatory Variables Food Alcohol Tobacco

Couples, N = 8,560

 CB -0.309# 0.434# 0.000
 (2.0) (3.7) (0.0)

 Other Expenditure 0.035 0.015 -0.003
 (29.0) (16.3) (6.2)

 [[chi square].sub.CB 4.74# 12.99# 0.00
 = Other exp)]

 p .03# .00# .96

 Overall [chi square] (6) = 25.72
 p = .0003#

Lone Parents, N = 738

 CB -0.062 0.244# 0.013
 (0.3) (2.3) (0.2)

 Other Expenditure 0.028 0.008 -0.001
 (7.9) (4.8) (0.7)

 [[chi square].sub.CB 0.16 4.94# 0.03
 = Other exp)]

 p .69 .03# .86

 Overall [chi square] (6) = 12.52
 p = .0513

Notes: Figures in parentheses are absolute t-values. Other
explanatory variables are: a linear trend; month, region, and
dummy variables for whether the child was aged 0-4, 5-10
(relative to 11-15); a quadratic in age of household head; and a
lone father dummy in the lone parent sample. Households with
negative other incomes are excluded. Bold values indicate
statistical significance at 5% level.

Note: Values with # indicates statistical significance at 5% level.

TABLE 5
Tobit Estimates of Engel Curves: One Child Not on Welfare, 1980-2000

Explanatory Variables Child Women's Men's
 Clothing Clothing Clothing
Couples, N = 8.575

 CB 0.014 0.275 0.174
 (0.1) (1.6) (0.7)

 Other Expenditure 0.024 0.053 0.057
 (21.2) (32.1) (22.4)

 [F.sub.(CB-Other exp)] 0.01 1.70 0.19

 p .93 .19 .66

Lone Parents. N = 744

 CB 0.152 0.673 0.678
 (0.5) (1.8) (1.0)

 Other Expenditure 0.039 0.085 0.031
 (6.4) (10.9) (2.3)

 [F.sub.(CB-Other exp)] 0.15 2.37 0.93

 p .70 .12 .33

Explanatory Variables
 Food Alcohol Tobacco
Couples, N = 8.575

 CB -0.188 0.522# 0.020
 (1.3) (4.0)# (0.1)

 Other Expenditure 0.075 0.038 -0.004
 (51.8) (29.1) (2.5)

 [F.sub.(CB-Other exp)] 3.31 13.89# 0.03

 p .07 .00# .86

Lone Parents. N = 744

 CB -0.094 0.421# 0.141
 (0.5) (2.8) (0.7)

 Other Expenditure 0.067 0.025 0.000
 (16.0) (8.1) (0.0

 [F.sub.(CB-Other exp)] 0.67 6.98# 0.45

 p .41 .01# .50

Notes: Figures in parentheses are absolute t-values. Other
explanatory variables are: a linear trend: month, region, and
dummy variables for whether the child was aged 0-4, 5-10
(relative to 11-15); a quadratic in age of household head: and a
lone father dummy in the lone parent sample. Bold values indicate
statistical significance at 5% level.

Note: Values with # indicates statistical significance at 5%
level.

TABLE 6
Engel Curves: Parents with One Child Not on Welfare: 1980-1989
and 1990-2000

 Child Women's Men's
Explanatory Variables Clothing Clothing Clothing

Couples, N = 4,554 (1980-1989)

 CB 0.019 -0.127 -0.311
 (0.1) (0.6) (1.6)

 Other Expenditure 0.017 0.045 0.033
 (16.0 (26.1) (20.1)

 [[chi square].sub.(CB=Other exp)] 0.00 0.67 2.96

 p .99 .41 .09

 Overall [chi square](6) = 19.95
 p = .0028#

Lone Parents, N = 325 (1980-1989)

 CB 0.223 0.305 0.198
 (0.8) (0.8) (1.0

 Other Expenditure 0.030 0.058 0.014
 (4.2) (6.2) (2.9)

 [[chi square].sub.(CB=Other exp)] 0.49 0.46 0.95

 p .48 .50 .33

 Overall [chi square](6) = 2.28
 p = .89

Couples, N = 4,021 (1990-2000)

 CB -0.045 0.265 0.456#
 (0.4) (1.8) (3.0)#

 Other Expenditure 0.017 0.036 0.024
 (16.0 (23.6) (15.5)

 [[chi square].sub.(CB=Other exp)] 0.34 2.41 7.92#

 p .56 .12 .00#

 Overall [chi square](6) = 23.25
 p = 0.0007#

Lone Parents, N = 419 (1990-2000)

 CB 0.166 1.043# -0.038
 (0.7) (3.2)# (0.4)

 Other Expenditure 0.022 0.066 0.003
 (5.1) (10.8) (1.9)

 [[chi square].sub.(CB=Other exp)] 0.39 8.78# 0.19

 p .53 .00# .67

 Overall F, p [chi square](6) = 16.52
 p = .0112#

Explanatory Variables Food Alcohol Tobacco

Couples, N = 4,554 (1980-1989)

 CB -0.682# 0.607# -0.003
 (2.7)# (2.6)# (0.0

 Other Expenditure 0.076 0.045 0.002
 (35.6) (23.0) (1.8)

 [[chi square].sub.(CB=Other exp)] 8.67 5.72# 0.00

 p .00# .02# .97

 Overall [chi square](6) = 19.95
 p = .0028#

Lone Parents, N = 325 (1980-1989)

 CB 0.080 0.094 -0.014
 (0.3) (0.7) (0.1)

 Other Expenditure 0.069 0.015 0.003
 (9.8) (4.1) (0.9)

 [[chi square].sub.(CB=Other exp)] 0.00 0.33 0.02

 p .97 .57 .89

 Overall [chi square](6) = 2.28
 p = .89

Couples, N = 4,021 (1990-2000)

 CB 0.241 0.509# -0.054
 (1.2) (3.9)# (0.6)

 Other Expenditure 0.075 0.025 -0.002
 (36.6) (19.0) (1.9)

 [[chi square].sub.(CB=Other exp)] 0.68 13.43# 0.33

 p .41 .00# .56

 Overall [chi square](6) = 23.25
 p = 0.0007#

Lone Parents, N = 419 (1990-2000)

 CB -0.333 0.337 0.045
 (1.2) (2.2) (0.4)

 Other Expenditure 0.066 0.020 -0.001
 (12.4) (7.3) (0.3)

 [[chi square].sub.(CB=Other exp)] 1.91 4.37# 0.15

 p .17 .04# .69

 Overall F, p [chi square](6) = 16.52
 p = .0112#

Notes: Figures in parentheses are absolute t-values. Other
explanatory variables are: a linear trend; month, region, and
dummy variables for whether the child was aged 0-4, 5-10
(relative to 11-15): a quadratic in age of household head: and a
lone father dummy in the lone parent sample. Bold values indicate
statistical significance at 5% level.

Note: Values with # indicates statistical significance at 5% level.

TABLE 7
Engel Curves and Maternal Education: 1980-2000

 Child Women's Men's
 Clothing Clothing Clothing

Mother Left School at 16: Couples, N = 5.271

 CB 0.017 0.459# 0.590#
 (0.2) (3.4) (4.6)

 Other Expenditure 0.020 0.040 0.029
 (18.2) (27.8) (21.1)

 [[chi square].sub.(CB=Other exp)] 0.00 9.65# 18.75#

 p .98 .00# .00

 Overall [chi square](6) = 32.15
 p = .0000#

Mother Left School at 17/18: Couples, N = 1.980

 CB -0.146 -0.017 -0.129
 (1.2) (0.1) (0.7)

 Other Expenditure 0.016 0.043 0.026
 (10.4) (16.8) (11.7)

 [[chi square].sub.(CB=Other exp)] 1.74 0.09 0.79

 p .19 .77 .37

 Overall [chi square](6) = 17.24
 p = .0084#

Mother Left School at 19+; Couples, N = 1,324

 CB 0.543# -0.042 -0.377
 (2.0) (0.l) (0.7)

 Other Expenditure 0.016 0.035 0.029
 (8.9) (12.2) (8.5)

 [[chi square].sub.(CB=Other exp)] 3.66 0.03 0.57

 p .06 .86 .45

 Overall [chi square](6) = 9.92
 p = .1279

Mother Left School at 16: Lone Parents, N = 366

 CB 0.109 0.618# -0.092
 (0.6) (2.4) (1.2)

 Other Expenditure 0.054 0.095 0.012
 (8.9) (11.5) (5.2)

 [[chi square].sub.(CB=Other exp)] 0.08 4.03# 1.99

 p .78 .04# .16

 Overall [chi square](6) = 8.33
 p = 0.2151

Mother Left School at 17/18; Lone Parents, N = 154

 CB 0.318 1.185# 0.073
 (0.5) (2.1) (0.4)

 Other Expenditure 0.033 0.031 0.002
 (3.3) (3.3) (0.6)

 [[chi square].sub.(CB=Other exp)] 0.22 4.29# 0.17

 p .65 .04# .68

 Overall [chi square](6) = 8.03
 p = 0.2357

Mother Left School at 19+; Lone Parents, N = 224

 CB 0.051 1.111 0.497
 (0.2) (2.0) (1.9)

 Other Expenditure 0.009 0.061 0.007
 (1.7) (6.3) (1.5)

 [[chi square].sub.(CB=Other exp)] 0.02 3.67 3.59

 P .89 .06 .06

 Overall [chi square](6) = 8.82
 p = 0.1841

 Food Alcohol Tobacco

Mother Left School at 16: Couples, N = 5.271

 CB -0.086 0.378# 0.018
 (0.5) (2.8) (0.2)

 Other Expenditure 0.078 0.037 0.002
 (39.6) (26.1) (1.9)

 [[chi square].sub.(CB=Other exp)] 0.78 6.43# 0.02

 p .38 .01# .88

 Overall [chi square](6) = 32.15
 p = .0000#

Mother Left School at 17/18: Couples, N = 1.980

 CB -0.246 0.579# -0.013
 (1.0) (3.5) (0.1)

 Other Expenditure 0.074 0.030 0.003
 (23.9) (14.6) (2.0)

 [[chi square].sub.(CB=Other exp)] 1.72 11.27# 0.02

 p .19 .00# .88

 Overall [chi square](6) = 17.24
 p = .0084#

Mother Left School at 19+; Couples, N = 1,324

 CB -0.710 0.892 -0.014
 (1.3) (1.6) (0.1)

 Other Expenditure 0.068 0.031 0.001
 (19.0) (8.7) (1.0)

 [[chi square].sub.(CB=Other exp)] 1.92 2.40 0.01

 p .17 .12 .91

 Overall [chi square](6) = 9.92
 p = .1279

Mother Left School at 16: Lone Parents, N = 366

 CB -0.081 0.135 0.019
 (0.3) (1.2) (0.2)

 Other Expenditure 0.086 0.022 0.000
 (11.2) (6.0) (0.0)

 [[chi square].sub.(CB=Other exp)] 0.47 0.93 0.02

 p .49 .33 .88

 Overall [chi square](6) = 8.33
 p = 0.2151

Mother Left School at 17/18; Lone Parents, N = 154

 CB -0.252 0.367 0.182
 (0.6) (1.5) (1.1)

 Other Expenditure 0.077 0.009 0.003
 (10.7) (2.2) (1.2)

 [[chi square].sub.(CB=Other exp)] 0.59 2.19 1.23

 p .44 .14 .27

 Overall [chi square](6) = 8.03
 p = 0.2357

Mother Left School at 19+; Lone Parents, N = 224

 CB -0.224 0.171 -0.011
 (0.5) (0.7) (0.1)

 Other Expenditure 0.052 0.023 0.002
 (7.0) (5.6) (0.8)

 [[chi square].sub.(CB=Other exp)] 0.44 0.41 0.01

 P .51 .52 .93

 Overall [chi square](6) = 8.82
 p = 0.1841

Notes: Figures in parentheses are absolute t-values. Other
explanatory variables are: a linear trend; month, region, and
dummies for whether the child was aged 0-4, 5-10; and a quadratic
in age of household head. Bold values indicate statistical
significance at 5% level.

Note: Values with # indicates statistical significance at 5% level.

TABLE 8
Engel Curves and Household Income

(a) Couples with One Child Not on Welfare, 1980-2000

 Child Clothing Women's Clothing

Couples in Bottom Third of Income Distribution N = 2,859, Mean
Income = 215.81 [pounds sterling]/wk

 CB 0.060 0.275
 (0.4) (1.3)

 Other Expenditure 0.019 0.043
 (12.9) (20.0)

 [[chi square].sub.(CB = Other 0.08 1.18
 exp)]

 p .78 .28

 Overall [chi square](6) = 7.95
 p = .2415

Couples in Middle Third of Income Distribution N = 2,858, Mean
Income = 349.59 [pounds sterling]/wk

 CB 0.094 -0.017
 (0.8) (0.1)

 Other Expenditure 0.018 0.036
 (11.4) (17.1)

 [[chi square].sub.(CB = Other 0.42 0.11
 exp)]

 p .52 .74

 Overall [chi square](6) = 7.13
 P = .3085

Couples in Top Third of Income Distribution N = 2,858, Mean
Income = 621.44 [pounds sterling]/wk

 CB -0.086 0.343
 (0.6) (1.6)

 Other Expenditure 0.016 0.038
 (11.5) (18.1)

 [[chi square].sub.(CB = Other 0.57 2.13
 exp)]

 p .45 .14

 Overall [chi square](6) = 19.99
 p = .0028#

(b) Lone Parents with One Child Not on Welfare, 1980-2000

 Child Clothing Women's Clothing

Lone Parents in Bottom Third of Income Distribution N = 248,
Mean = 123.28 [pounds sterling]/wk

 CB -0.269 0.462
 (1.0) (1.8)

 Other Expenditure 0.060 0.074
 (5.1) (6.5)

 [[chi square].sub.(CB = Other 1.49 2.17
 exp)]

 p .22 .14

 Overall [chi square](6) = 7.8
 p = .2531

Lone Parents in Middle Third of Income Distribution N = 248,
Mean = 224.79 [pounds sterling]/wk

 CB 0.138 0.091
 (0.5) (0.3)

 Other Expenditure 0.057 0.047
 (6.7) (4.3)

 [[chi square].sub.(CB = Other 0.09 0.02
 exp)]

 p .77 .90

 Overall [chi square](6) 2.89
 p = 0.8221

Lone Parents in Top Third of Income Distribution N = 248,
Mean = 416.85 [pounds sterling]/wk

 CB 0.595 1.560#
 (1.7) (2.6)

 Other Expenditure 0.012 0.065
 (2.2) (7.1)

 [[chi square].sub.(CB = Other 2.83 6.41#
 exp)]

 p .09 .01#

 Overall [chi square](6) = 14.35
 14.35 p = .0260#

(a) Couples with One Child Not on Welfare, 1980-2000

 Men's Clothing Food

Couples in Bottom Third of Income Distribution N = 2,859, Mean
Income = 215.81 [pounds sterling]/wk

 CB 0.072 -0.571
 (0.4) (1.8)

 Other Expenditure 0.028 0.085
 (13.8) (27.0)

 [[chi square].sub.(CB = Other 0.05 4.43
 exp)]

 p .83 .04

 Overall [chi square](6) = 7.95
 p = .2415

Couples in Middle Third of Income Distribution N = 2,858, Mean
Income = 349.59 [pounds sterling]/wk

 CB -0.132 -0.016
 (0.8) (0.1)

 Other Expenditure 0.030 0.067
 (15.0) (22.9)

 [[chi square].sub.(CB = Other 1.09 0.14
 exp)]

 p .30 .71

 Overall [chi square](6) = 7.13
 p = .3085

Couples in Top Third of Income Distribution N = 2,858, Mean
Income = 621.44 [pounds sterling]/wk

 CB 0.508# -0.184
 (2.4) (0.7)

 Other Expenditure 0.027 0.065
 (12.6) (26.0)

 [[chi square].sub.(CB = Other 5.14# 1.01
 exp)]

 p .02# .31

 Overall [chi square](6) = 19.99
 p = .0028#

(b) Lone Parents with One Child Not on Welfare, 1980-2000

 Men's Clothing Food

Lone Parents in Bottom Third of Income Distribution N = 248,
Mean = 123.28 [pounds sterling]/wk

 CB 0.058 -0.101
 (0.5) (0.4)

 Other Expenditure 0.026 0.107
 (4.7) (9.9)

 [[chi square].sub.(CB = Other 0.06 0.72
 exp)]

 p .80 .40

 Overall [chi square](6) = 7.8
 p = .2531

Lone Parents in Middle Third of Income Distribution N = 248,
Mean = 224.79 [pounds sterling]/wk

 CB -0.081 0.101
 (0.8) (0.3)

 Other Expenditure 0.001 0.072
 (0.4) (7.2)

 [[chi square].sub.(CB = Other 0.69 0.01
 exp)]

 p .41 .93

 Overall [chi square](6) 2.89
 p = 0.8221

Lone Parents in Top Third of Income Distribution N = 248,
Mean = 416.85 [pounds sterling]/wk

 CB 0.401 -0.607
 (1.7) (1.4)

 Other Expenditure 0.003 0.052
 (0.7) (7.6)

 [[chi square].sub.(CB = Other 2.74 2.22
 exp)]

 p .10 .14

 Overall [chi square](6) = 14.35
 14.35 p = .0260#

(a) Couples with One Child Not on Welfare, 1980-2000

 Alcohol Tobacco

Couples in Bottom Third of Income Distribution N = 2,859, Mean
Income = 215.81 [pounds sterling]/wk

 CB 0.315 0.068
 (1.4) (0.4)

 Other Expenditure 0.040 0.005
 (17.8) (2.8)

 [[chi square].sub.(CB = Other 1.54 0.12
 exp)]

 p .21 .73

 Overall [chi square](6) = 7.95
 p = .2415

Couples in Middle Third of Income Distribution N = 2,858, Mean
Income = 349.59 [pounds sterling]/wk

 CB 0.348# 0.035
 (2.3) (0.3)

 Other Expenditure 0.023 0.004
 (11.3) (2.7)

 [[chi square].sub.(CB = Other 4.60# 0.06
 exp)]

 p .03# .80

 Overall [chi square](6) = 7.13
 p = .3085

Couples in Top Third of Income Distribution N = 2,858, Mean
Income = 621.44 [pounds sterling]/wk

 CB 0.701# -0.036
 (3.3) (0.4)

 Other Expenditure 0.031 0.001
 (14.2) (0.8)

 [[chi square].sub.(CB = Other 9.87# 0.16
 exp)]

 p .00# .69

 Overall [chi square](6) = 19.99
 p = .0028#

(b) Lone Parents with One Child Not on Welfare, 1980-2000

 Alcohol Tobacco

Lone Parents in Bottom Third of Income Distribution N = 248,
Mean = 123.28 [pounds sterling]/wk

 CB 0.166 0.089
 (1.5) (0.7)

 Other Expenditure 0.015 0.001
 (3.0) (0.1)

 [[chi square].sub.(CB = Other 1.86 0.46
 exp)]

 p .17 .50

 Overall [chi square](6) = 7.8
 p = .2531

Lone Parents in Middle Third of Income Distribution N = 248,
Mean = 224.79 [pounds sterling]/wk

 CB 0.225 -0.040
 (1.3) (0.3)

 Other Expenditure 0.020 0.003
 (3.6) (0.5)

 [[chi square].sub.(CB = Other 1.34 0.07
 exp)]

 p .25 .79

 Overall [chi square](6) 2.89
 p = 0.8221

Lone Parents in Top Third of Income Distribution N = 248,
Mean = 416.85 [pounds sterling]/wk

 CB 0.196 -0.017
 (0.9) (0.1)

 Other Expenditure 0.015 0.002
 (4.2) (0.0)

 [[chi square].sub.(CB = Other 0.63 0.02
 exp)]

 p .43 .90

 Overall [chi square](6) = 14.35
 14.35 p = .0260#

Notes: Figures in parentheses are absolute t-values. Other
explanatory variables are: a linear trend; month, region, and
dummy variables for whether the child was aged 0-4, 5-10
(relative to 11-15); and a quadratic in age of household head.
Bold values indicate statistical significance at 5% level.

Note: Values with # indicates statistical significance at 5% level.

TABLE 9
Engel Curves and Household Income: Child Aged Up to 10 Only

Explanatory Variables Child Clothing Women's Clothing

Couples, N = 6.564

 CB -0.101 0.268#
 (1.4) (2.2)

 Other Expenditure 0.016 0.039
 (21.3) (29.8)

 [[chi square].sub.(CB = 2.71 3.52
 Other exp)

 p .10 .06

 Overall [chi square](6) = 22.92
 p = .0008#
Lone Parents. N = 404

 CB -0.093 0.592
 (0.5) (1.8)

 Other Expenditure 0.037 0.072
 (8.4) (9.8)

 [[chi square].sub.(CB = 0.46 2.60
 Other exp)

 p .50 .11

 Overall [chi square](6) = 5.01
 p = .5420

Explanatory Variables Men's Clothing Food

Couples, N = 6.564

 CB 0.267# 0.001
 (2.1) (0.0)

 Other Expenditure 0.031 0.077
 (22.9) (45.2)

 [[chi square].sub.(CB = 3.54 0.23
 Other exp)

 p .06 .63

 Overall [chi square](6) = 22.92
 p = .0008#
Lone Parents. N = 404

 CB 0.135 0.150
 (0.8) (0.6)

 Other Expenditure 0.014 0.076
 (3.7) (12.6)

 [[chi square].sub.(CB = 0.52 0.08
 Other exp)

 p .47 .78

 Overall [chi square](6) = 5.01
 p = .5420

Explanatory Variables Alcohol Tobacco

Couples, N = 6.564

 CB 0.486# 0.021
 (3.8) (0.3)

 Other Expenditure 0.033 0.000
 (23.8) (0.5)

 [[chi square].sub.(CB = 12.79# 0.09
 Other exp)

 p .00# .77

 Overall [chi square](6) = 22.92
 p = .0008#
Lone Parents. N = 404

 CB 0.134 0.095
 (1.0) (0.9)

 Other Expenditure 0.020 0.004
 (6.3) (1.6)

 [[chi square].sub.(CB = 0.69 0.74
 Other exp)

 p .41 .39

 Overall [chi square](6) = 5.01
 p = .5420

Notes: Figures in parentheses are absolute t-values. Other
explanatory variables are: a linear trend; month, region, and
dummy variables for whether the child was aged 0-4, 5-10
(relative to 11-15); a quadratic in age of household head; and a
lone father dummy in the lone parent sample. The F statistic in
each equation is a test of whether the coefficient on CB and on
other income is equal, and the overall F statistic is a test that
all of the CB coefficients equal the corresponding other income
coefficients. Bold values indicate statistical significance at 5%
level.

Note: Values with # indicates statistical significance at 5% level.

TABLE 10
Anticipated Versus Unanticipated CB Effects: Rational Expectations

 Child Clothing Women's Clothing

Couples, N = 8,575

 Anticipated CB -0.233 0.403
 (1.3) (1.5)

 Unanticipated CB 0.066 0.173
 (0.8) (1.4)

 Other Expenditure 0.017 0.039
 (22.8) (34.8)

 [[chi square].sub.(antCB 1.83 1.78
 = Other exp])

 p .18 .18

 Overall [chi square](6) = 23.84
 p = 0.0006#

 [[chi square].sub.anant(CB 0.34 1.19
 = Other exp])

 p .56 .27

 Overall [chi square](6) = 20.06
 p = 0.0027#

 Lone Parents, N = 744

 Anticipated CB 0.085 0.775
 (0.1) (0.7)

 Unanticipated CB 0.156 0.704#
 (0.9) (2.9)#

 Other Expenditure 0.025 0.064
 (2.4) (12.4)

 [[chi square].sub.(antCB 0.01 0.47
 = Other exp])

 p .94 .49

 Overall [chi square](6) = 1.28
 p = 0.97

 [[chi square].sub.anant(CB 0.55 6.93#
 = Other exp])

 p .46 .01#

 Overall [chi square](6) = 11.44
 = 0.0758

 Men's Clothing Food

Couples, N = 8,575

 Anticipated CB 0.141 -1.368#
 (0.5) (3.9)

 Unanticipated CB 0.208 0.059
 (1.7) (0.4)

 Other Expenditure 0.028 0.075
 (24.9) (51.7)

 [[chi square].sub.(antCB 0.17 16.66#
 = Other exp])

 p .68 .00

 Overall [chi square](6) = 23.84
 p = 0.0006#

 [[chi square].sub.anant(CB 2.16 0.01
 = Other exp])

 p .14 .92

 Overall [chi square](6) = 20.06
 p = 0.0027#

 Lone Parents, N = 744

 Anticipated CB -0.079 -0.326
 (0.2) (0.4)

 Unanticipated CB 0.079 -0.089
 (0.8) (0.4)

 Other Expenditure 0.007 0.067
 (3.1) (16.0)

 [[chi square].sub.(antCB 0.04 0.21
 = Other exp])

 p .84 .65

 Overall [chi square](6) = 1.28
 p = 0.97

 [[chi square].sub.anant(CB 0.53 0.61
 = Other exp])

 p .46 .44

 Overall [chi square](6) = 11.44
 p = 0.0758

 Alcohol Tobacco

Couples, N = 8,575

 Anticipated CB 0.330 0.174
 (1.2) (1.1)

 Unanticipated CB 0.524# -0.043
 (4.2)# (0.6)

 Other Expenditure 0.033 0.000
 (28.8) (0.6)

 [[chi square].sub.(antCB 1.16 1.11
 = Other exp])

 p .28 .29

 Overall [chi square](6) = 23.84
 p = 0.0006#

 [[chi square].sub.anant(CB 15.57# 0.32
 = Other exp])

 p .00# .57

 Overall [chi square](6) = 20.06
 p = 0.0027#

 Lone Parents, N = 744

 Anticipated CB 0.323 0.143
 (0.7) (0.4)

 Unanticipated CB 0.208# 0.005
 (2.0)# (0.1)

 Other Expenditure 0.019 0.001
 (8.6) (0.4)

 [[chi square].sub.(antCB 0.46 0.16
 = Other exp])

 p .50 .69

 Overall [chi square](6) = 1.28
 p = 0.97

 [[chi square].sub.anant(CB 3.31# 0.00
 = Other exp])

 p .07# .96

 Overall [chi square](6) = 11.44
 p = 0.0758

Notes: Other expenditure is defined as total expenditure minus
CB. Figures in parentheses are absolute t-values. The lone
parent's equations include a dummy variable for lone father. Bold
values indicate statistical significance at 5% level.

Note: Values with # indicates statistical significance at 5% level.