Chattoe and Gilbert: Understanding Consumption

Edmund Chattoe and Nigel Gilbert (2001) 'Understanding Consumption: What Interviews with Retired Households Can Reveal About Budgetary Decisions'
Sociological Research Online, vol. 6, no. 3, <http://www.socresonline.org.uk/6/3/chattoe.html>

To cite articles published in Sociological Research Online, please reference the above information and include paragraph numbers if necessary

Received: 4/7/2001 Accepted: 16/11/2001 Published: 30/11/2001

Abstract

This paper uses interview data from retired households to inform a discussion about economic models of consumption. It is divided into two parts. In the first part, the economic models are described. The paper then discusses several different types of reasons for finding them unhelpful in explaining consumption. The second part of the paper considers the role of 'middle range' theories in developing plausible models of household behaviour. Phenomena which the interviews suggest are important in explaining consumption, such as time allocation, the labour supply decision, the ubiquitous durability of goods and the structure of the household, are not typically supported by middle range theory in current models. Without the constraints of such theory, it is very hard to distinguish models providing genuine explanation from those that merely fit the data. The latter part of the paper also discusses aspects of a new middle range theory of consumption suggested by the interviews.

Keywords:

Budgeting; Consumption; Family Practices; Middle Range Theory; Rational Choice Theory; Retired Households

Introduction

1.1

In a previous paper (Chattoe and Gilbert 1999), we described and discussed the findings of a series of interviews with retired households about their management of money. The purpose of this paper is to extend that work by examining what these findings imply for our understanding of consumer behaviour. In particular, the behaviour of our interviewees can be contrasted sharply with the assumptions typically made in economics about "rational" financial planning. In order to criticise the economic theory effectively it is necessary to have a clear picture of it. For this reason, the first half of the paper provides a description of economic theories of consumption. The second part of the paper draws on the interview data to focus a number of specific problems with those theories, both in the particular context of retired households and more generally. (These problems can be divided into two groups: assumptions that seem empirically false and important dimensions of the social situation neglected by the theory.) Discussion of the data also reveals ways in which the theory of consumer behaviour can be re-orientated to make it more satisfactory from a sociological point of view without abandoning the idea of building testable models. The gap between the economic theories and the sociological findings of the interviews is also methodologically revealing. One underlying reason for the weaknesses of the economic approach is that it typically involves what has been criticised as "variable sociology" (Esser 1996), an attempt to explain social behaviour by reference to the correlation between aggregate variables. This unsatisfactory methodology can be contrasted with the kind of "middle range" (Merton 1968) or (more recently) "mechanism based" (Hedström and Swedburg 1998) development of theories in sociology. This approach makes explicit claims about the ways that individuals decide and act in response to the environment and other actors.

1.2

There is an interesting irony here. Many critics of variable sociology urge rational choice as a generally applicable model for social action. However, de facto "variable sociology" may nonetheless result from the limited applicability of rational choice theory to real behaviour. These potential difficulties are usefully discussed byGoldthorpe (1998).

Economic Theories of Consumption

2.1

In this section, we discuss economic theories of consumption, dealing first with traditional approaches and their shortcomings and then showing how some new theories that have been proposed also suffer from the same difficulties on closer examination.

Traditional Theories of Consumption

2.2

Standard economic theory provides two classes of model for understanding consumer behaviour. The first consists of models of intra-temporal choice described in detail by Deaton and Muellbauer (1980) and Blundell (1988). These consider the allocation of a consumer's fixed budget to goods within a single period, according to preference. In addition, it is usually assumed that preferences do not change systematically and that they satisfy axioms of rationality (Deaton and Muellbauer 1980, pp. 26-30). Such models would describe a household with a regular income and a predictable lifestyle, such that it was possible to allocate money on one payday to all the goods that the household was going to desire before the next. (By contrast, many retired households have a mixture of small income sources with different payment periods. In addition, the regularity of their budgeting may be further complicated by long term investments like maturing insurance policies.) Such models are tested by transforming the budget constraint and preference ordering into a set of demand functions.These relate the quantity of each good or group of goods purchased to the income of the decision-maker and the prices of goods. This type of model has proved problematic, and there is some evidence that both its assumptions (Deaton and Muellbauer 1980, pp. 78-82) and the axioms of rational choice on which it is based (Sippel 1994) may be inadequate.

2.3

In consequence, models of inter-temporal choice, forming the second class, currently dominate consumer theory (Blundell 1988, Deaton 1992). These consider a different question, how decision-makers allocate money to different periods over their lifetimes. If the inter-temporal and intra-temporal decisions can be separated effectively, these models are complementary. The decision-maker initially decides how much to spend in every time period and then what to spend it on in each particular period. Papers by Blundell (1988) and Blundell et al. (1994) are unusual in making this complementarity explicit. In practice, modeling the aggregate allocation of consumption over time has proved more theoretically satisfying and, at least predictively, more successful. For this reason, following current economic practice, most of the discussion in this paper is devoted to models of inter- temporal choice.

2.4

Inter-temporal choice models are invariably operationalised using the Life Cycle and Permanent Income Hypotheses, abbreviated to LCH and PIH throughout this paper (Deaton 1992, p. viii). Both sharpen the intuition that people may attempt to trade off the relative benefits of present and future consumption (by borrowing and saving), rather than spending all their money at each instant.

2.5

The Life Cycle Hypothesis (Modigliani and Brumberg 1954, 1979) asserts that decision-makers not only take account of current income in determining consumption but income over the whole lifetime. This results in saving when income is greater than that required for the currently preferred level of consumption, and dis-saving (or borrowing) when it is less. This model explains certain broad patterns of consumer behaviour. Students take out loans for education on the strength of future income. Most saving takes place in late middle age when disposable income is greatest. Most importantly for this study, retired households are supposed to dis-save so they can maintain something like their pre-retirement lifestyle. We broadly observed this behaviour in the interview sample, though with important qualifications which will be discussed later.

2.6

The Permanent Income Hypothesis (Friedman 1957) approaches the same question in a slightly different way. As Deaton (1992, p. 76) points out, under certain subsidiary assumptions the LCH and PIH are very similar. Friedman argues that decision-makers relate consumption in a particular period not to disposable income in that period but rather to some estimate of long run or permanent income. Friedman argues that the PIH can be modelled in terms of learning, and that the present value of permanent income depends on previous values of actual income.

2.7

Both hypotheses link current consumption to something other than current income, allowing both expectations of the future and experience of the past to make a difference. Plainly, the intuitions on which these hypotheses are based are plausible, but the way they are traditionally converted into models suggests difficulties of several different kinds.

Theoretical Difficulties

2.8

These are distinguished from implausible assumptions, which are discussed in the following sections. There is nothing theoretically problematic about assuming decision-makers know when they will die even if it is incorrect. However, there are theoretical difficulties in testing the assumptions of the LCH and PIH. Both hypotheses generate incomplete models. In particular, these models involve parameters whose values cannot be determined independently of the behaviour that they are supposed to explain.

2.9

In the case of the LCH, there is no theoretical or empirical explanation of the "preference" for any particular allocation of consumption over a lifetime (for example in terms of luxuries and necessities or family commitments). There is also no independent method of determining what weighting the decision-maker gives to income in different periods when determining the smoothed pattern of consumption. Any lifetime consumption profile can be justified within the LCH by suitable assumptions about preferences and the relative importance of income in different periods. For this reason, the criterion for a successful model tends to be its goodness of fit or predictive ability ex post rather than whether it provides an illuminating explanation of consumer behaviour. (For example, it would be far more convincing if these models were able to predict "anomalous" lifetime consumption patterns displayed by certain groups ex ante.A relevant case for retired households would be particular patterns of bequests and inter-generational transfers. To say that some households have a "preference" for bequests while others do not is not saying much. Those households that discussed bequests tended to do so in terms of their children's "needs" such as commitments to their children.)

2.10

The LCH and PIH are sometimes interpreted in terms of period to period stability or consumption smoothing rather than sophisticated planning and adaptation. This interpretation is consistent with more plausible bounded rationality models. This is because it does not require agents to have any overview of the life course. However, the problem of parameter identification remains, since we cannot assess the "expected" level of observed stability using the theory independently. For households on a subsistence income, even quite small variations in consumption might be unpleasant and vigorously avoided. One interviewee discussed the difficulty she had paying for the repair of a pair of broken glasses. For households with adequate income, however, one can easily envisage them giving up the majority of luxuries – and thus displaying quite "bumpy" consumption – to save for a holiday without any negative effects.

2.11

The same problem of identification arises in the PIH. In this case, there is no way of measuring permanent income independently of the claim that current consumption depends on income in previous periods. As long as any lagged terms in income are significant, it can be argued that the PIH is correct. However, the observed lags can be explained in at least two other ways. Firstly, they could be the aggregate result of individuals' learning strategies which might have very little to do with optimisation. (Most retired households appear to do little explicit calculation in their budgeting.) Secondly, they could be artefacts of serial correlation. This can arise when individuals make decisions at different times (Deaton 1992, p. 96-97). Serial correlation will be particularly severe when agents use very little data about the past. (Repeated decisions by agents at different times will look like much longer lags in aggregate.) In fact, retired households generally do appear to make repeated comparisons month-on-month and week-on week. Although there are econometric methods for detecting these problems, there is no behavioural insight within the PIH that will distinguish one set of plausible lags from another. In consequence, models are again evaluated on econometric rather than empirical or theoretical grounds.

Institutional Difficulties

2.12

Another important question is whether agents could behave in the ways suggested by the LCH and PIH even if they chose to. For example, markets for loans and insurance are imperfect, so that many lifetime consumption paths are inaccessible in practice. (This effect also differs across household types. Several interviewees reported that – as single or divorced women – it was extremely hard for them to get mortgages or private pensions as little as twenty years ago. This has clearly had an important effect on their present financial situations.) The unwillingness of institutional lenders to loan on the strength of future events suggests that the problem is not merely one of individual laziness or lack of foresight. Lenders and borrowers invariably negotiate loans for well-defined purposes, based only on current assets or income. Student loans are one exception, being provided on the strength of future income, but such schemes are almost universally set up and maintained only through government intervention, perhaps for ideological reasons. Other exceptions are provided by credit cards and general bank loans. However, both of these forms of credit are relatively short term, expensive and still based predominantly on current disposable income. In particular, they are unlikely to be available to those at or below subsistence levels when the LCH assumes they will be sought. Similar comments apply to the use of savings to generate particular income streams through investment or insurance. For most major life events (marriage, children, serious illness, divorce, long term unemployment), the available forms of insurance and investment will only cover part of the cost. The price of certain future income streams, such as a post-retirement income equal to the pre-retirement level, is prohibitively high. It is clear that decision-makers are heavily constrained in adjusting their chosen levels of consumption away from current income even assuming they wish to do so. (Several interviewees, having paid off their mortgages and gained their "independence", were very scathing about any form of credit as a money management strategy.) Generally, even those who did not disparage credit were not using it to smooth their incomes. One recent financial "innovation" bearing on the ability of retired households to "boost" consumption displays the classic emphasis on current capital rather than future income. Only one household – a single man without dependents – was considering making over his house to a company that would provide him with an extra income and the right to live in the house until he died.

2.13

Standard economic theory acknowledges these liquidity constraints but typically attempts to explain them in terms of asymmetric information (Deaton and Muellbauer 1980, p. 319). It is not that everyone necessarily finds predicting changes in future income problematic, but that lenders contingently find that they cannot get enough information about the characteristics of borrowers. If the difficulty is merely contingent, it is perhaps surprising that so few market mechanisms have evolved to circumvent it. Attempting to assess liquidity constraints using consumption and income data alone (without reference to the policies of banks and the decision processes of households) also exacerbates the problem of identification discussed in the last section. Such data provide no way of distinguishing (without the use of subsidiary theoretical assumptions) between households that would like credit and cannot get it and those which do not want it. (Both situations were discussed by the interview sample.) As we shall see shortly, other mechanisms of consumption smoothing may be more relevant to retired households than the financial markets emphasised by traditional consumer theory.

Empirical Difficulties

2.14

The institutional restrictions in the last section strongly suggest that uncertainty, rather than mere risk (Knight 1921) is paramount in real decision-making. In a risky situation, the decision-maker can still assign subjective probabilities and utilities to all relevant future events. In an uncertain situation, at least one of these assignments is impossible. In the case of major life events, both probabilities and utilities are very hard to assess. Actuarial estimates fail at the individual level and are complicated by private knowledge. Utilities are heavily dependent on circumstances so it will also be necessary to assign probabilities to other relevant events. Many of these probabilities will have to be conditional on future events. (For example, pregnancy may be viewed as pleasant when planned but unpleasant when unplanned.) In practice, retired households appear to take some account of future expenses but not of future income until it has actually materialised (Wilcox 1989). The result is a significant amount of precautionary saving when money is available, in addition to any saving for particular purposes such as holidays or house repairs. Households also use very short periods of history to determine their present actions. In weekly decisions, for example, comparisons will only be made with the previous week and the same applies to monthly and yearly decisions. (One open question is how these different "scales" of decision-making are reconciled.)

Econometric Difficulties

2.15

Some of the concerns raised in the previous sections would be less pressing if we were confident that econometric results about the quality of models were robust. If it were possible to compare models readily and to rank their stability, there would be far less of a role for theory, except as a matter of personal preference. However, many examples provided in Deaton (1992)suggest that, despite high quality econometric work, the results of studies are still heavily dependent on the data set used and the precise specifications of the model. Debates on the extent to which consumption is excessively sensitive (Flavin 1981) or smooth ( West 1988) are still wide open. Furthermore, in a recent series of papers, Attanasio and Weber (1993, 1994) have investigated the relationship between models based on aggregate data and those estimated from individual households in the Family Expenditure Survey. They find that several results suggested by the aggregate data are artefacts of econometric difficulties and restrictive aggregation assumptions.

2.16

Taken separately, these four kinds of difficulties might seem tolerable. Taken together, it is perhaps not surprising that the possibility of explanatory models based on the LCH and PIH is doubtful. In the next section, we consider the role of middle range theory, of which the LCH and PIH prove to be imperfect examples. We also discuss the need for middle range theories of other phenomena that are important in consumer decisions, such as family structure, labour supply and durable goods. We consider the implications of the interview data for new middle range theory, both in relation to the areas already mentioned and others. One important challenge is to build a theory that integrates different aspects of consumer decision in the ways suggested by the interview data. Such integration should also explain the distinctive "life world" of the retired households which was revealed by the interviews and will be described further.

New Theories of Consumption

2.17

Although the preceding criticisms apply to a significant fraction of current consumer theory, it is also important to show why two apparently "new" approaches to consumption do not in fact take us any further forward with explaining behaviour.

2.18

The first of these is the "household production function" approach, described by Becker (1981, pp. 14-37). Models of this kind explore such issues as the household "division of labour", marriage "contracts", "shirking" and so on. They do this by proposing a household production function, in which family members work together to maximise production of both normal market goods and non-market commodities like children and "health". This approach can be seen as suffering from three problems, clearly related to the discussion above. The first is that presenting social institutions in these terms fundamentally begs the question. It is certainly true that men and women engage in different household activities (though anthropological data suggests that divisions of labour can vary widely), but why they do so is far from clear. Specialisation certainly has some role to play (or at least it is often appealed to rhetorically in arguments about household chores), but so do "arbitrary" normative control and interaction effects. If a woman engages in "typically male" activities, she may be criticised or ostracised by both men and women, regardless of her aptitude. If a man is responsible for looking after small children, even if he is treated well at gatherings of other female carers, he may feel awkward simply because he is in a minority. The second problem is that the assumptions made in such models ought to require more justification than they do in the strictly economic sphere, but actually receive less. (It is extraordinary how little time Becker devotes to justifying his assumptions in the new context.) What objective criteria can we use to label "shirking" in households when ex hypothesi no specific labour contract exists? (Despite asserting that household activity involves optimisation, Becker says nothing about independently evaluating the "efficiency" of real family practices. Doing so in terms of adherence to rationality is self-evidently circular.) The third problem is that while particular models in this class can be falsified, the class of models as a whole cannot. There are always enough "preference parameters" and ill-defined forms of "family capital" to rescue things. What model results from the family production function would suggest to the theorist that they were actually concentrating on the wrong aspects of family behaviour? (Because this is certainly what the interview data suggests, as we shall see below.)

2.19

It is also widely believed that the "behavioural" LCH (Shefrin and Thaler 1988) serves as a corrective to the problems of the more traditional models. Despite providing an effective critique of existing approaches ( ibid.p. 609), what Shefrin and Thaler actually do is replace one mysterious preference function with another. Specifically, they observe that people are less likely to spend out of money that they have labelled as "savings" rather than "income" or "windfall". This is clearly true, and is borne out both in the interviews with retired households and by other studies (Zelizer 1994), but what is important is the mechanism by which these labels are created and maintained. Plainly, saved money is spent without guilt on "appropriate" occasions. (There is a specific example of this issue discussed in Chattoe and Gilbert 1999, p. 94.) Similarly, a windfall may be saved in its entirety under certain circumstance. What determines such actions? The answer is the households budgetary "practices" or "mechanisms" which cause money to be placed in (or taken out of) certain categories. Without a theory of these practices, saying there are three "kinds" of money rather than one just introduces more free parameters into the model and worsens the identification problem.

2.20

To sum up, these models are new in what they try to explain but not new in the way they try to explain it. They still incorporate too many parameters like "preference" which lack independent measures. They make strong assumptions about individual rationality and motivation that are not plausible. They focus on fitting models to data (rather than identifying behavioural regularities empirically) resulting in models which are hard to identify or falsify and do not predict new data robustly.

The Role of Middle Range Theory in Explanatory Models

3.1

Despite concerns about the plausibility of the LCH and PIH, it is clear that similar behavioural hypotheses might play an important role in understanding consumption. By making falsifiable claims, they reduce the space of possible models to a manageable size. They also move us away from the purely econometric evaluation of models. (In fact, the first section of this paper suggests that the LCH and PIH are defective as middle range theories for both reasons. Firstly, because they do not make strong falsifiable claims and, secondly, because their incompleteness does not allow them to be evaluated except by econometrics.) Counter-intuitive behavioural implications generated by a hypothesis can be used to narrow the class of plausible models still further. (It is interesting that Becker's work generates very few such implications.) Such hypotheses can be seen as bridging the gap between very general methodological claims (for example, that decision-makers are individualistic optimisers) and very particular data about actual consumption and income. They are usefully referred to as middle range theories (Merton 1968, pp. 39-72). Model restrictions originating from middle range theory are also much easier to interpret. If a theory implies that all pensioners die in debt, it will be very easy to falsify a model based on that theory. On the other hand, models which are assessed primarily on econometric grounds are much harder to reject, since the evaluation depends on the intricacies of particular tests, the choice of data set and so on.

3.2

In this section of the paper, we discuss the possibility of using interview data to provide some middle range theory for four areas acknowledged to be important in econometric studies of consumer decision-making. These are habit and preference formation, the existence of durable goods, time use (including labour supply) and household size and composition. In most recent models (Attanasio and Weber 1993, Carroll 1994), variables for these factors are used with few theoretical restrictions to "pick up" unexplained variability in consumption. In consequence, the interpretation of these variables is problematic (Blundell et al. 1994, p. 58).

3.3

Our sample consisted of detailed interviews with 28 households containing at least one retired member. Interviewees were unpaid volunteers who responded to newspaper, magazine and newsletter advertisements. The interviews were semi- structured, following a script of open-ended questions about income and expenses, family circumstances, planning activities, past experience and aspirations. Further demographic details and summaries of budgeting strategies can be found in the appendix below. Despite the small sample, the interviews covered a wide range of demographic groups, income levels, lifestyles, accommodation types and locations. At the same time, they revealed several strong similarities in budgetary behaviour across the whole sample. This paper only discusses those parts of the interview data relevant to debates in consumer theory and the distinctive consumption situation of retired households. More detailed discussion of the interview data can be found in Chattoe and Gilbert (1999).

Time, Wellbeing and Labour Supply

3.4

One of the striking findings in the interviews is that the allocation of time to activities is far more important than the allocation of money to goods (Chattoe and Gilbert 1999, pp. 90-91). Decisions about expenditure and budgeting are thus taken to facilitate a preferred lifestyle (in terms of time use) rather than the other way round. One reason for this is that activities seem to be what provide pleasure. It is activity patterns rather than purchases which the retired households typically describe when they talk about the "good life". The role of many goods is thus to facilitate the pursuit of activities rather than typically to provide pleasure in their own right. (Of course, some goods combine these roles. Food may be merely nutritious and conducive to health, or it may be pleasant to eat, or both. Fortunately, this potential discrepancy becomes relevant only in very poor households.)

3.5

This observation has several practical implications. For all but the poorest households, or the most expensive goods, the time cost of price search makes it an unpopular practice. If the value of time is sufficiently great, it makes more sense to save it by combining a shopping trip at a more expensive supermarket and a visit to family in a nearby town. This shows a greater overall saving in "time plus money" than comparing prices in a number of shops. However, very poor households cannot afford to do anything more enjoyable so they have time to spare and often make a virtue of necessity.

3.6

Furthermore, Deaton's (1992, p. viii) observation that it is reasonable to model consumption independent of the labour supply decision is almost certainly not correct at the household level. Even allowing for institutional constraints on hours of employment, finding work and adjusting the hours worked to achieve a better trade-off between income and time continue to be part of the decision- making repertoire in many retired households (Chattoe and Gilbert 1999, pp. 90-91). Furthermore, the constraints that do exist are steadily declining owing to part time working, job shares and temporary contracts. Reassessing consumption in terms of activities rather than purchases reminds us that it is hard to justify separating out one use of time (labour) from all other uses (leisure). This is particularly true in retired households where voluntary work so obviously serves social and emotional needs as well as financial ones.

3.7

Another important consequence of the activity- based approach is that decision-makers have two different aspects of their behaviour available to control. If there is a discrepancy between planned activities and the current budget, households can change the set of planned activities or the budget, depending on which is easier. Economising by changing activities (staying at home and watching television rather than going out) is far more important in practice (and effective) than making economies based on price search (Chattoe and Gilbert 1999, pp. 91-92). It is that adjustment to "time and durable rich – income poor" status which gives retirement its distinctive character. Retired households "use" the investments they have in household, social networks of friends and family, and (in some cases) skills to occupy their time as pleasantly as their incomes will allow. Contraction of income can often be "managed" to maintain lifestyle as when families come to visit rather than being visited or voluntary jobs can pay relevant expenses.

3.8

The observation that pleasure appears to inhere in activities (rather than purchases) has another important implication. The economic assumption of separability, that the contribution of alternative purchases to overall wellbeing is fixed and known is far less plausible for activities than it is for goods, particularly if the role of most goods is actually instrumental. (Strictly speaking, these assumptions are relaxed even in traditional theories. The existence of complements and the Diminishing Marginal Rate of Substitution must be justified by some connection between the wellbeing contributions of successive purchases.) It is easy to say whether a spanner helped to fix your bike, but rather harder to identify the features of a particularly pleasant day. If the well-being resulting from patterns of activity is not readily separable then we should anticipate the existence of lifestyles, socially reproduced patterns of activities known to provide an adequate level of wellbeing. (The evolution of lifestyles is modeled in Chattoe and Gilbert 1997.) It is this non-separability makes it much harder for an individual to develop a satisfactory lifestyle from scratch. (In addition, there are more subtle forms of pleasure than those suggested by the paradigm of eating and drinking. Perhaps these have an even greater degree of non-separability and since their causes are harder to deduce, they are more heavily contested.)

3.9

The importance of these observations for models of consumption is that constraints on time and compatible activities combine to produce considerable restrictions on the patterns of consumption that will be observed in real households. We would expect different decisions or even decision-making processes among those with different lifestyles: consumption of convenience foods and taxi rides being correlated with long hours of work and high income, for example. A final point is that the constraints on activities can often be investigated somewhat independently of data on consumption. For example time diary data (Anderson et al.1994) provides a separate source of information about what people do and (in some cases) how much they enjoy doing it.

The Importance of Durables, Luxuries and Necessities

3.10

The ubiquity of durable goods and the existence of necessities require two other important qualifications to the conventional LCH and PIH based models of consumption which are particularly relevant to retired households.

3.11

Durable goods can pose difficulties for simple models of consumption (Deaton 1992, pp. 10-12). Paradigm cases of durable goods such as cars and fridges are expensive, indivisible and provide flows of services that continue over time, contradicting the assumption that purchases and their effects can be allocated to separable decision-making periods. Many models of consumption attempt to solve this difficulty by excluding durable goods altogether (Attanasio et al.1995, Carroll 1994) although there are exceptions (Bertola and Caballero 1990). Unfortunately, it seems goods that are either durable or deferrable are the norm rather than the exception. (Deferrable goods do not provide a flow of services, but their use can be adjusted because they are stored without deterioration: washing up liquid for example.) About the only commodity that must be consumed immediately and cannot be stored is fresh food (Hayashi 1985). In consequence, stocks of other goods, not always treated as durable, are important in determining new consumption and activity choice. For example, in retired households, expenditure on new clothing and furniture can be cut dramatically by conserving existing stocks (Chattoe and Gilbert 1999, pp. 91-92). For poorer households, an important expenditure smoothing mechanism is a stock of tinned food. Likewise, many retirement activities focus on the home (a pre-existing durable), for example tools for gardening or D.I.Y., equipment for crafts, furniture and "decor" for social events. In addition, the durability of many small household goods introduces further possibilities such as second hand markets, sharing or loaning, gifts and recycling of old goods for different uses. (It is well known that charity shops are important consumption sites for retired households.) Durable goods thus provide important possibilities for consumption smoothing which augment those offered by financial assets.

3.12

Ironically, the ubiquity of durable goods also provides an explanation for a significant source of uncertainty about consumption. The paradigm for durable use is a gradual "melting away" of the flow of services provided by a particular durable (Deaton and Muellbauer 1980, pp. 345-358). In practice, many durable goods fail completely after a certain time. It may be wasteful or impossible to keep spares of all durable goods, so households will always have to face the possibility of a major expenditure at an uncertain time. Many retired households had ensured the adequacy of their budgets for everyday needs but were very worried about "the roof falling in", that is, having to deal with the unforeseen failure of expensive durable goods. Even fixing up the house and buying a new car before retirement, which was fairly common where it could be afforded, only offered imperfect protection against this problem, particularly if the household members lived significantly longer than they expected. (This study was designed to look at the transition from retirement but disproportionate extreme poverty among the very elderly is well known in the literature.)

3.13

Another important restriction on possible consumption patterns is imposed by relaxing the assumption that economic agents are "disembodied", that their preferences have no underlying justification or pattern. For economics, the paradigm of preference is whether to "sleep on your back or on your belly" (Sen 1982, p. 285). This is a choice between alternatives that only matter in terms of subjective satisfaction rather than having any "real" effects (for example on health). However, even for relatively wealthy retired households, a substantial proportion of income is spent on activities which cannot be deferred for long without real negative effects: food, heating, house maintenance, contact with family and friends. The theory of wellbeing is much too large and controversial a topic to be dealt with here, but it is important to understand that the simplifying assumption of "ungrounded" preferences has a cost for explanatory models of consumption. These concerns have also been addressed by Sen (1985), Lutz and Lux (1988) and Maslow (1954, 1968).

Accounting Procedures, Bounded Rationality and the Social Nature of the Household

3.14

We have already mentioned the fact that retired households can adjust both the budget and the set of preferred activities when there is a mismatch between them. The accounting procedures used by households reflect practical limitations on rationality, which determine whether it is easier to modify the activity plan or the budget.

3.15

The interviews showed that almost all households divide their expenditure into three parts that are managed differently (Chattoe and Gilbert 1999, pp. 88-89). Bills such as those for a mortgage and electricity can often be predicted exactly up to a year in advance and are often paid "automatically", not even being regarded as part of disposable income. Expenditure of the second type can be made more or less regular by habit (or other mechanisms of monitoring and control), so it is possible to budget approximately for it, based on past experience. An example of expenditure of this kind is provided by the weekly shopping trip, which always costs about the same, provided the same sorts of things are bought. It is true that price determines whether one brand of biscuits is bought rather than another or whether sardines are substituted for salmon. However, these decisions take place within the framework of assumptions about how much it is appropriate to spend on food each week. The final kind of expenditure involves items that are large, unexpected or unique and these will be financed out of money that is deliberately kept separate, for example, in a savings account. (Note that while these categories bear some relation to those proposed in the Behavioural LCH, they are defined by the nature of the expenditure and budgetary practice rather than some abstract "preference".)

3.16

Alterations to the budget and the set of preferred activities take place not by redesigning the whole plan, but by incremental alterations or reallocations. For example, if a shopping trip costs too much, the typical response is to recover the "shortfall" on the next flexible expenditure that arises, lest it be forgotten. This might involve cutting out some "little extra" like the cup of coffee in the supermarket café. This is another important source of observed behavioural stability. The same is true of alterations to the activity plan. Because of the difficulties of determining wellbeing, the complexity of budgeting and the multiplicity of constraints, the tendency is to minimise significant alterations in the set of activities. Where these cannot be avoided, because of the loss of income and the gain of free time associated with retirement, for example, many households expressed concern about the effects of the transition (Chattoe and Gilbert 1999, pp. 91, 93-94).

3.17

Another example of "budgetary practice" is provided by the general acceptance of direct debits as a costless way of smoothing monthly consumption on certain regular expenses. Poorer households generally try to regularise their activities far more than those with more resources do. This reduces the risk of having to deal with unexpected events. The difficulty of maintaining a good plan is such that decision-makers sometimes transfer allocations of money from one purpose to another, seemingly without re-evaluation. For example, money paid for school fees may be devoted to a pension once the children leave college, with the level of investment determined by what the school fees happened to be rather than any sort of calculation. Different sources of income may also be allocated to different accounts by a rule, so that state pension may pay the bills with any remainder going to luxuries, while the occupational pension is used to cover all other expenses. (Clearly, these rules must be sustainable but this does not mean they will be optimal. Different rules sets may also be more or less robust to novelty and unforeseen events.)

3.18

The use of budgeting rules of these kinds is also important in families as a way of reaching agreements and avoiding confusion about responsibility. Different frameworks of budgetary control are possible (Pahl 1989). For example, the husband may receive all the income and be responsible for the bills, while allocating money to his wife for housekeeping. Alternatively, each family member may keep their income separate while paying into an account for agreed purposes such as bills. (Social groups are also important because they negotiate shared activities such as meals, placing further constraints on the possible set of individual activity plans.)

3.19

The combination of budgetary rules of thumb, built around the monitoring of accounts and the existence of necessities and durable goods, has important consequences for the role of family size and composition on consumption. Larger households are likely to have a much higher proportion of essential expenditure for a given income. Small households will still tend to acquire the same set of "basic" durable goods (kettle, cooker and perhaps television) as those with many more members. In each case, the difficulties of maintaining desirable patterns of activities and a sustainable budget are likely to induce considerable stability in both budgeting and patterns of activities. (Some households we interviewed had been using the same accounting procedures to inform their decisions for over thirty years.) However, the "dimensions" along which behaviour is stable (such as activity plans and budgeting strategies) do not appear to be those emphasised by economic theory. By contrast, the things that are assumed to be stable over the life course – like preferences – actually change significantly for children, students, footloose young adults, parents of small children and retired households.

Discussion and Conclusion

4.1

In this paper, it has not been possible to provide more than a brief outline of the interview data and their potential impact on consumer theory. Nevertheless, the shift in outlook suggested by the interviews is substantial. The first part of the paper considered the economic approach to middle range theory, exemplified by the LCH and PIH. In these models, apparently strong claims about individual rationality are subsequently weakened by the existence of parameters (like preferences) which are not determined by the theory and are hard to estimate or observe independently of the model as a whole. Such parameters permit flexibility, but push the models towards evaluation by purely econometric criteria (including prediction) rather than explanatory power or behavioural plausibility. In consequence, the form of models is strongly determined by currently available data and the limitations of mathematical modelling techniques. By contrast, the discussions in the second part of the paper suggest that there are considerable sources of stability in consumption behaviour, but that these rely on the very aspects of consumption which are typically excluded from consumer theory. The use of rules of thumb, the existence of durable goods and the importance (and difficulty) of decisions about time allocation and the pursuit of wellbeing all contribute to the stability of consumer behaviour. At the same time, many of these aspects of behaviour can be investigated independently if new kinds of data and data collection techniques are accepted. Some of these data have not yet been collected. Others kinds of data, like time diaries do exist but are not currently used as a basis for theory development. Interviews are indispensable for investigating the decision processes and practices used by real individuals while comparison of time diary data can give additional information about essential activities and the potential for variability in lifestyles.

4.2

A general finding from the interviews is that different aspects of consumption cannot usefully be split up and studied separately as is sometimes attempted in economics. Rules of thumb are used because the future is uncertain and the past requires interpretation. The future is uncertain because durable goods are ubiquitous and can fail without warning. The past is hard to interpret because our understanding of the world is incomplete. Some of the findings described here have already been incorporated into middle range theories of durable consumption or household budgeting. However, in addition to reservations about the type of theorising exemplified by the LCH and PIH, it would be fair to say that hardly any of these theories attempts to integrate more than one aspect of budgetary behaviour. Our first attempt to develop such an integrated model using simulation is described elsewhere (Chattoe and Gilbert 1997) though this work is ongoing.

4.3

This paper offers three contributions to the difficult task of refining economic models of consumption. Firstly, it considers some limitations of the traditional approach in the light of new data. Secondly, it describes the important parts of that data so they may form the basis for new models, either traditional or otherwise. Thirdly, it illustrates two ways in which economics makes the task of understanding consumption harder than it needs to be. The first is to exclude certain aspects of consumption (which are empirically important but theoretically problematic). The second is to limit the collection and use of data to that which may be treated econometrically. This reinforces the exclusion of certain important phenomena like rules of thumb and institutional constraints that cannot be deduced from behavioural data alone. If the task of consumer theory is to build genuinely explanatory models, it seems inappropriate to leave data sources and data collection techniques unused without very good reason.

Appendix 1: The Sample of Retired Households

The following table and key (reprinted with the permission of Cambridge University Press from Chattoe and Gilbert 1999, pp 100-101) provides background to the sample of retired households on which this discussion is based.

Table 1

Household

Householders (Column 1): Married/cohabiting (M) or single (S).

Others in Household (Column 2): None (-), young child or children in household (C), adult child or children in household (A), lodger (L).

Gender of Respondent (Column 3): Male (M), female (F), interview jointly with both male and female (J).

Status (Column 4): Respondent has recently retired (R), has been retired for up to 3 years (RR), has been retired for more than 3 years (RRR), is of pensionable age but self-employed/quasi- retired (E).

Employment (Column 5): None (N), voluntary/unpaid (V), part-time paid work (P), none but partner works full time (F).

Relative Income Level (Column 6): Low (L), medium (M) and high (H). 'Low income' is no holidays and/or no car, 'medium income' is car and/or UK holidays perhaps with an occasional foreign trip, and 'high income' is car and/or regular foreign holidays.

Sources of Income (Column 7): Occupational pension(s) (O), state pension (S), personal pension (P), investments or savings (I) and rent from lodger (L).

Accounts (Column 8): Pays bills directly from cash (M), has a current bank account (C), has a savings account (S), has investments (I), has account(s) hypothecated for specific purposes such as medical bills (H).

Main Budgetary Decision Maker (Column 9): Male (M), female (F), joint (J), mainly individual decisions and separate accounts, but some joint decisions (I).

Type(s) of Budgeting (Column 10): One-Off Financial Assessment at Retirement: Getting house into shape, sorting out lifestyle (O). Account Tracking: Watching the state of bank accounts rather loosely, checking the statements retrospectively (A). Project Management: Letting money pile up and dealing with items from a "wish list" when feasible (P). Regularising Activities: Buying the same food each week, reducing uncertainty (R). Calculation: Going beyond account tracking to make explicit allocations for particular uses or working out a quantitative financial plan (C).

Acknowledgements

: We are very grateful to all the households who kindly volunteered to take part in the interviews and for the comments of the anonymous SCROL referees whose input improved the final version of the paper. This research is part of Project L122-251-013 supported by the ESRC under their Economic Beliefs and Behaviour Programme.

References

ANDERSON, M., BECHHOFER, F., and GERSHUNY, J. (editors) (1994) The Social and Political Economy of the Household. Oxford: Oxford University Press.

ATTANASIO, O. and WEBER, G. (1993) 'Consumption Growth, the Interest Rate and Aggregation', Review of Economic Studies, Vol. 60, No. 3, pp. 631-649.

ATTANASIO, O. and WEBER, G. (1994) 'The UK Consumption Boom of the Late 1980's: Aggregate Implications of Microeconomic Evidence', The Economic Journal, Vol. 104, No. 427, pp. 1269-1302.

ATTANASIO, O., BANKS, J., MEGHIR C., and WEBER, G. (1995) 'Humps and Bumps in Lifetime Consumption', Working Paper Series Number 5350, National Bureau of Economic Research.

BECKER, G. S. (1981) A Treatise on the Family.Cambridge, MA: Harvard University Press.

BERTOLA, G. and CABALLERO, R. (1990) 'Kinked Adjustment Costs and Aggregate Dynamics', in O. Blanchard and S. Fischer (editors) NBER Macroeconomics Annual 1990. Cambridge, MA: The M. I. T. Press.

BLUNDELL, R. (1988) 'Consumer Behaviour: Theory and Empirical Evidence - A Survey', The Economic Journal, Vol. 98, No. 389, pp. 16-65.

BLUNDELL, R., BROWNING M., and MEGHIR, C. (1994) 'Consumer Demand and the Life-Cycle Allocation of Household Expenditures', Review of Economic Studies, Vol. 61, No. 1, pp. 57-80.

CARROLL, C. (1994) 'How Does Future Income Affect Current Consumption', The Quarterly Journal of Economics, Vol. 109, No. 1, pp. 111-147.

CHATTOE, E. and GILBERT, N. (1997) 'A Simulation of Adaptation Mechanisms in Budgetary Decision Making', in R. Conte, R. Hegselmann and P. Terna (editors) Simulating Social Phenomena. Berlin: Springer-Verlag.

CHATTOE, E. and GILBERT, N. (1999) 'Talking About Budgets: Time and Uncertainty in Household Decision Making', Sociology, Vol. 33, No. 1, pp. 85-103.

DEATON, A. (1992) Understanding Consumption. Oxford: Clarendon Press.

DEATON, A. and MUELLBAUER, J. (1980) The Economics of Consumer Behavior. Cambridge: Cambridge University Press.

ESSER, H. (1996) 'What is Wrong with "Variable Sociology"', European Sociological Review,Vol. 12, No. 2, pp. 159-166.

FLAVIN, M. (1981) 'The Adjustment of Consumption to Changing Expectations About Future Income', Journal of Political Economy,Vol. 89, No. 5, pp. 974- 1009.

FRIEDMAN, M. (1957) A Theory of the Consumption Function. Princeton, NJ: Princeton University Press.

GOLDTHORPE, J. H. (1998) 'Rational Action Theory for Sociology', British Journal of Sociology, Vol. 49, No. 2, pp. 167-192.

HAYASHI, F. (1985) 'The Permanent Income Hypothesis and Consumption Durability: Analysis Based on Japanese Panel Data', Quarterly Journal of Economics, Vol. 100, No. 4, pp. 1083-1113.

HEDSTRÖM, P. and SWEDBURG, R. (editors) (1998) Social Mechanisms: An Analytical Approach to Social Theory. Cambridge: Cambridge University Press.

KNIGHT, F. (1921) Risk, Uncertainty, and Profit. Boston, MA: Houghton Mifflin.

LUTZ, M. and LUX, K. (1988) Humanistic Economics: The New Challenge.New York, NY: The Bootstrap Press.

MASLOW, A. (1954) Motivation and Personality.New York, NY: Harper and Row.

MASLOW, A. (1968) Toward a Psychology of Being.New York, NY: Van Nostrand Reinhold.

MERTON, R. (1968) Social Theory and Social Structure,enlarged edition. New York, NY: The Free Press.

MODIGLIANI, F. and BRUMBERG, R. (1954) 'Utility Analysis and the Consumption Function: An Interpretation of Cross Section Data', in K. Kurihara (editor) Post-Keynesian Economics. New Brunswick, NJ: Rutgers University Press.

MODIGLIANI, F. and BRUMBERG, R. (1979) 'Utility Analysis and the Consumption Function: An Attempt at Integration', in A. Abel (editor) The Collected Papers of Franco Modigliani, volume 2. Cambridge, MA: Cambridge University Press.

PAHL, J. (1989) Money and Marriage. London: Macmillan.

SEN, A. (1982) 'The Impossibility of a Paretian Liberal', in Choice, Welfare and Measurement. Oxford: Basil Blackwell.

SEN, A. (1985) Commodities and Capabilities. Amsterdam: North- Holland.

SHEFRIN, H. M. and THALER, R. H. (1988) 'The Behavioral Life- Cycle Hypothesis', Economic Inquiry, Vol. 26, No. 4, pp. 609-643.

SIPPEL, R. (1994) 'An Experiment on the Pure Theory of Consumer's Behaviour', Projektbereich B, Discussion Paper Number B-274, Rheinische Friedrich-Wilhelms- Universität, Bonn.

WEST, K. (1988) 'The Insensitivity of Consumption to News about Income', Journal of Monetary Economics, Vol. 21, No. 1, pp. 17-33.

WILCOX, D. (1989) 'Social Security Benefits, Consumption Expenditure, and the Life Cycle Hypothesis', Journal of Political Economy,Vol. 97, No. 2, pp. 288-304.

ZELIZER, V. A. (1994) The Social Meaning of Money: Pin Money, Paychecks, Poor Relief, and Other Currencies. New York, NY: Basic Books.