This paper reports on a study of the way that people make
decisions about the allocation of income between categories of
expenditure, such as food, rent and leisure. The most common
(economic) approach to formalising budgetary decision processes
is deductive: a theory is proposed and its implications are then
explored. However, it is also possible to develop formal models
inductively and this is the approach used here. Starting from
interviews with a sample of eight full time postgraduate
students, we have constructed a simulation of the budgetary
decision-making process which is presented in this paper. The
budgetary practices described by subjects were formalised as a
set of rules for each respondent. The budgetary behaviour
produced by these rules was then explored for different potential
incomes and expenditure levels. It was shown that the existence
of even quite limited foresight in agents enabled them to balance
their budgets even when potential outgoings considerably exceeded
income. The paper also discusses some implications of the
inductive approach to model building and the contrast between
economic theory and the interview data. It concludes by outlining
some directions for future research.
The paper reports on a study of the way that people make decisions about the allocation of income between categories of expenditure, such as food, rent and leisure. This is an issue which has been addressed by economists who have attempted to develop theories of consumption based on rational choice. However, these theories have not had much success in either predicting behaviour or explaining it. One reason is that they have difficulty in incorporating factors such as social influences, personal heuristics and conflicts between preferences.
The approach that has been adopted by economists is essentially deductive: a theory is proposed and its implications explored. An alternative strategy is to proceed inductively, starting from self-reports about budgetary decision-making and constructing generalisations from these. However, most people's self reports about the reasons for their decisions are likely to be uninformative because they allocate their income on the basis of habit and convention much of the time. This is less true of those who have recently had a significant change in their economic circumstances and who have therefore been forced to consider their decision processes, such as students starting on a postgraduate course, or people recently retired.
We have collected interview data from a sample of eight full time postgraduate students about their budgetary decision-making. They were asked to give details of major sources of income and outgoings using their own classifications. They explained, among other things, how they ensured that regular bills were met, how they would deal with unexpected expenses and which categories they regarded as fixed or negotiable.
The processes that they described are being formalised as a set of rules for each respondent. As a first step, the implications of these rules will be explored by applying them to varying incomes and prices and observing changes in expenditure patterns. As a second step, the rule sets will be compared to identify commonalities. As a third step, the aggregate relationship between income and consumption patterns over all the rule sets will be compared with nationally representative data from the Family Expenditure Survey. The rule sets will then be compared with those obtained from interviews with another group: the recently retired.
While the analysis to this point is static, that is, the rule sets describe decision making at the moment of the interview, the intention is to extend it to investigate how respondents adjust their allocation rules in the light of experience and through imitation of other's decision-making behaviour.
This paper is a work in progress. The simulation is still being
designed and constructed. In order to refine the questions to be
used in subsequent interviews, to provide practice in the
techniques of interviewing and to highlight important issues for
further investigation, it was decided to carry out a number of
preliminary interviews with students prior to interviews with the
recently retired. Students constitute another group with highly
constrained budgets who are forced to be relatively aware of
their financial decision processes. In a tentative way, these
interviews have supported the view that it is possible to build
simulations inductively. They have suggested both a metaphor for
describing consumption and a number of important and specific
patterns of behaviour that must be incorporated into a model of
consumer decision. This paper discusses these patterns of
behaviour, describes the simulation and some preliminary results,
outlines the direction of future work and considers the relations
between interview findings and economic consumer theory.
The eight students who were interviewed each spent between thirty minutes and an hour describing their budgeting behaviour and their attitudes to it. The interviews were tape-recorded and semi-structured, in that the interviewer worked through a list of prepared questions, but the subjects had considerable latitude to decide what constituted a relevant answer. After the interview, once they were in an appropriate frame of mind and thinking about their budgets, subjects were given a brief questionnaire to fill in describing their income and outgoings.
Generally, subjects were able to talk coherently about their budgetary practices and were seldom ambiguous or hard to understand. They were also frank, being prepared to talk about illegal activities and embarrassing financial misjudgments without prompting. However, there were definite limits to their awareness of the reasons for some actions and practices. Even so, they were often able to reflect on these limits and go some way towards explaining them when prompted. In some cases, the difficulty simply arose from asking people to become proficient at performing an unfamiliar task. Many subjects commented that it was the first time they had thought and talked in detail about some aspects of their budgetary behaviour.
The rest of this section discusses a number of important themes
common to many of the interviews. The list is by no means
complete but highlights the behaviour that is most important and
interesting, either because it contrasts with the assumptions of
the more common rational choice models, or for the purposes of
designing the simulation. When compared to rational choice
models, subjects demonstrate a much less well developed awareness
of their budgetary priorities. However, by contrast, they have
a far richer repertoire of techniques for modifying consumption
over time. For example, subjects actively formulate plans and
make use of recording and other control systems that are
relatively robust to cognitive limitations and uncertainty. They
can also introspect these limitations, for example forgetting and
miscalculation, and to some extent work around them, although
they cannot simply eliminate them. The interviews also revealed
a number of important features of the environment through the
effects they had on the behaviour of subjects. These include the
substantial number of durable purchases involved in everyday
consumption, the importance of joint consumption with other
agents and the derivative nature of the benefits of consumption.
Generally, many of these themes can be characterised as
suggesting a complex environment in which consumption behaviour
must be both adaptive and based on heuristics. In turn this
suggests a need for computational simulation to model the
interaction of psychological, rational and social factors
explicitly within a common framework.
In the budget questionnaire, subjects were asked to prioritise the purchases they made, in the groups in which they commonly thought about them. These groups, for example "food", "clothing", "bills" and "recreation", are referred to here as expenditure groups. The priority was that of "importance": which purchases were regarded as most essential and which could be dispensed with in appropriate circumstances.
Subjects typically did not find it easy to provide an explicit
ordering of budget priorities, particularly at the level of
expenditure groups. The fundamental distinction made seems to be
between essential and unessential expenditure. Subjects who were
unable even to meet expenditure regarded as essential did not
attempt to economise further or to prioritise essential
purchases. Instead, they were prepared to accept increasing debt.
By contrast, subjects who were still spending on purchases they
regarded as unessential were typically taking steps to economise
further if outgoings exceeded income, rather than go into debt.
The view was frequently expressed that debt was justifiable to
meet essential commitments, but not otherwise. The same
distinction applied to parental assistance. Although the
definition of essential expenditure does include items that are
not strictly necessary to physical survival, when "social needs"
such as human interaction are considered, there seems to be
considerable agreement among the subjects as to what constitutes
a minimum standard of living. (In cases where this decision is
taken out of their hands, there is a large amount of evidence
that groups such as the unemployed suffer both physically and
mentally as a result of non satisfaction of social needs even
though they do not starve (Fagin and Little 1984, Smith 1987).
Thus, not only is there a shared perception of basic needs but
it appears to be empirically grounded.) Perhaps surprisingly,
rent, college fees and domestic bills, for example gas,
electricity and water, are of an even higher priority than food,
many subjects not considering rent as part of their disposable
income at all. Clothing and travel are lower priority items,
capable of being economised in extremis. All other items are
classified as unessential and it is here that differing tastes
for consumption can be observed. This is in accordance with
economic theory which defines a minimum survival set of goods and
considers decision making as relevant only to consumption outside
this set. Where these data diverge from economic theory is in
emphasising the fact that, for these students, unessential
consumption is often only a fairly small, and potentially
unimportant, proportion of total consumption and that the minimum
is defined not just physically, but socially.
Most subjects operated different budgetary practices over periods
as short as a term. During the term they might be budgeting very
precisely and living on savings, while they devoted themselves
to course work. During the vacation they might be spending quite
freely while working and living at home. Major decisions
concerning whether to change accommodation, cohabit and get or
quit a job have a considerable impact on budget decision making
as well as consumption patterns and objectives. Planning is also
observed in the medium term, with decisions to get a job before
travelling or working to save money for a return to college. It
even exists in the very short term with subjects reporting that
they relax quite stringent budget planning over Christmas or
during the course of social activities. These reports reflect the
adaptive and embedded nature of budgeting. They suggest that
rules of decision, as well as actual consumption, change to
reflect new objectives and that these patterns may in turn affect
the objectives that created them. The objectives also alter what
is feasible. A stringent budget can be maintained for a short
period to meet a fixed objective, but would not be sustainable
in the longer term, even to avoid debt.
There is an uneasy distinction between durable and non durable
goods in economic theory and it is generally agreed that durable
purchase is much harder to model (Deaton and Muellbauer 1980, pp.
345-7). In practice, it appears that very few goods do not have
some measure of durability. Many types of food are now available
in forms which are relatively long lasting so that only in a few
cases is it useful to equate purchase with consumption. Durable
goods pose a problem in modelling because their purchase can be
deferred, sometimes for quite long periods. This is precisely
because these goods are durable and because their capabilities
can be varied and combined in new ways. An old pair of trousers
can be patched and used for gardening to save wear on a newer
pair. A cracked jug may still function as a vase. In a similar
way, subjects report that they can economise by running down
various sorts of capital. They will empty their cupboards,
producing increasingly unpalatable meals, to survive until the
next grant cheque arrives. They are prepared to make use of
pre-existing parental housing capital, moving into the spare room
at home to save money, even when they are uneasy about accepting
a cash payment from their parents that would help with the rent.
Although subjects regard the ability to live at home as an
advantage, they also remark that it is a relatively "easy" thing
for parents to offer because the house is there already and
typically has room to spare - the opportunity cost to the parent
is less than the benefit to the student. In the longer term,
subjects report a steady decline in capital over their academic
careers particularly in the case of clothing, electrical goods
and transport. Large items such as cars are simply too expensive
to be replaced. Smaller capital purchases are repeatedly put off
as unessential in the light of current needs. Subjects thus have
to make decisions that relate not only to their immediate
commitments, but to the maintenance of various forms of capital.
Certain sorts of capital are accumulated because of the influence
of transactions costs, for example it is easier and cheaper to
buy food in quantity than an item at a time. (This can also
result in the gradual accumulation of foodstuffs as a form of
saving.) Other items are genuinely durable like clothing, while
still others are a form of insurance: it is useful to have an
umbrella in the house because you never know when you will need
it, but you know that you will.
The process of making budgetary calculations involves attempts to adjust the size of the budget (longer term planning) and to allocate resources within it. This allocation involves some measure of wellbeing. In the absence of a naturally defined "budget period", this usually involves the scheduling of an uncertain stream of sequential income and outgoings. This stream is "smoothed" by the existence of savings and capital, much as water flow is smoothed by the presence of a reservoir. If accurate information is available about previous expenditure and future needs, it may be possible to draw up an explicit budget, but there are a variety of types of uncertainty that make this difficult in practice. Instead, individuals develop processes that allow them to make decisions about individual purchases as they arise. One subject reported taking out eighty pounds in cash at the beginning of the week and spending nothing once that was gone. The sum was calculated separately on the basis of income and fixed outgoings. This is the simplest possible example of an "all or nothing" decision rule.
The agent's budgetary decision system constitutes a solution to
a control problem, subject to a number of constraints. These
include physical limitations (limited memory, will and
calculating power), organisational limitations (finite
overdrafts) and social and emotional limitations. (It would
typically not be acceptable, either to poor students, or to her
friends, to save money by agreeing to visit the pub but then
offering to buy no drinks.) As will be discussed in the following
section, agents have some awareness of these constraints, but by
its nature, this awareness must itself be imprecise. On the
positive side, the agent can make use of various capabilities,
such as calculation, projection and memory to decide whether to
make given purchases. Over time, the agent will learn or infer
various constraints and the effect that certain sorts of
consumption have on overall wellbeing. This learning involves
both introspection and social observation.
Most subjects remarked on situations where, despite being broadly aware of a behaviour they considered unhelpful, they were unable to control it directly, or to be aware of it while it was happening. As a result it was necessary to make use of "work arounds". One example is an unwillingness to carry large amounts of cash because it is spent faster. Subjects also remarked on the difficulty of balancing an account around a positive sum of money compared to balancing around zero. These are examples of psychological biases (Kahneman, Slovic and Tversky 1982) that may arise from experience that is relevant in other situations. They should not therefore be regarded as irrational, but perhaps non-rational or a-rational: despite awareness of their existence, it is not possible merely to eliminate them at will. An example of a bias that is more obviously helpful is provided by habituation. Once a behaviour has become habitual, and it can sometimes be made so by deliberate repetition, it can be carried out with negligible cognitive effort. Subjects who had kept detailed budgets for a long time found it very easy to do and were able to increase their efficiency incrementally. The process of habit formation is also discussed in Nelson and Winter (1982).
The same situation arises in the introspection of wellbeing and
wants or needs. These are often relatively imprecise. Agents
learn to understand their desires over time by a process of
recollection and generalisation. Small children have to learn to
associate bodily sensations like nausea with their subsequent
effects. As a result, they become able to warn their parents
accordingly, and eventually take appropriate action on their own
(Parker, Mitchell and Boccia 1994). However, due to the
complexity of the determinants of wellbeing, this is a difficult
process involving a substantial amount of trial and error as well
as directed action. Sometimes it is not only unclear whether a
given purchase will result in satisfaction, but even what sort
of purchases should be considered to satisfy some unfocussed
desire.
There are relatively few mentions of joint consumption in the interviews, probably due to the social situation of most students. When joint consumption involves partners or families, it is presented as relatively unproblematic. This may be a result of habituation, as such relationships are typically of long standing. The main difficulties appear to involve casual friendships and pragmatic relationships where social and personal needs conflict. Socialising with acquaintances who are employed can result in overspending because of a lack of shared priorities and an unwillingness to complain. Subjects living in shared houses often described a tension between the communal atmosphere that they desired and an individualistic atmosphere that was far more common, shared bills displaying the classic structure of the Prisoner's Dilemma with respect to economising. Students were deliberately chosen for the research, as the possibility that agents need to model the beliefs and desires of others in making their own decisions would add considerable complexity to the simulation, though conceptually such models are not incompatible with the computational approach used here (Baron-Cohen, Tager- Flusberg and Cohen 1993, Whitten 1991). (There would also be a need to model the communication processes resulting in joint patterns of activity.) Nevertheless, despite this simplification, it should not be concluded that social consumption is unimportant, just that at this stage, it is too complicated for us to model.
A different, but related aspect of social consumption is the
existence of lifestyles, an important source of collective
knowledge about viable consumption patterns. The analysis of
lifestyles constitutes a major part of sociological analysis of
consumption (Appadurai 1986, Featherstone 1990, Lunt and
Livingstone 1992, Shields 1992, Tomlinson 1990) thus providing
an important link between economic and sociological analysis.
Rather than modelling other agents, it is possible to use them
as (more-or-less) approximate sources of data about what
consumption is conducive to wellbeing. We can observe, more or
less, what other people are doing, and how they feel about it.
(Indeed, much casual conversation is precisely about topics like
these.) Lifestyles that provide a substantial feeling of
wellbeing are highly constrained and therefore scarce. Others can
provide knowledge not only about the best options, but also, by
implication, the range of alternatives that are available. This
in turn allows the development of target levels of consumption
suitable for satisficing behaviour (Simon 1955, Radner 1975a).
One can also actively "pursue" a lifestyle, making it part of
one's planning, on the grounds that if some consumption in a
lifestyle is found to be agreeable, the rest may also be.
A number of subjects remarked on the trade-off between time and money, not only in labour decisions, but also in travel and food purchase. Goods that "save time" are often relatively expensive, because they involve the purchase of time belonging to other people, in the form of labour. (Ready cook meals are an obvious example, though their prices are also affected by such factors as economies of scale and the division of labour.) In the light of the discussion of planning above, it is clear that time is an important part of the wellbeing function in that both consumption and production require it. It is another form of capital that depletes rapidly, and can only be transferred indirectly and imperfectly.
More generally, the relationship between time and money raises the important issue of the derivative relationship between purchase, consumption and wellbeing. Not only can purchase not be equated with consumption due to the durable nature of most purchases, but the process of consumption, when it occurs, is typically an indirect result of the choice of activities which appear to be the primary objects of human desire. For everyday consumption at least, people desire or need goods for the capabilities or effects they provide, rather than as things in themselves. Because what is important about goods is the functions they are able to fulfil, objectively or subjectively, goods themselves cannot be used as the basic units of analysis. This is revealed in the way that subjects talk about necessities and trade-offs. A meal in a restaurant meets both the social need for interaction and the physical need for food. A cheaper way to meet the same needs might be to take a picnic to the park. The restaurant meal is thus "unnecessary" because the same needs can be addressed in other, cheaper, ways. (It is an empirical matter whether the picnic in the park actually does satisfy the same needs.) Although these "attributes" of goods, the needs or wants which they can satisfy, are not strictly objective, it is still easier to identify properties like substitutability with the attributes than with the goods alone. This point will be addressed in more detail in Section 5.4.
Continue to Part II
Return to Menu