economic models, aggregated expectations

USE OF THE AGGREGATED EXPECTATIONS HYPOTHESIS IN THE CONTEXT OF ECONOMIC MODELS

The Aggregated Expectations Hypothesis offers a new way of examining likely future outcomes, based upon the most important contributor - expectations - to the individual decisions which aggregate to create these final macro-outcomes. It also offers the more powerful actors, especially governments, a new tool for influencing some of those future outcomes. One outcome is that (economic) models should be viewed primarily as communications devices. Their power depends upon how well they persuade the actors, on whom they are targeted, to bring their behaviour into line with that incorporated in the model.

Unlike the more specific rational expectations hypothesis, the ‘general hypothesis of aggregated expectations’ does not assume a rational actor. Nor does it require a limited number informed actors to directly influence the macro-economic outcomes. Instead, the hypothesis typically looks to the very much larger number of uninformed, though not necessarily irrational, participants throughout the whole population.

It represents a viable hypothesis which allows such managers to make meaningful predictions, even in an otherwise uncertain macro-environment, and hence to permit the possibility of successful intervention. It is based upon observations made, as part of the Open University’s ‘Millennium Project’, working with more than a thousand large organisations and government departments. The ‘Millennium Project[2]’, overall, attempts to improve the accuracy of long-range forecasting using a range of new research techniques (Mercer, 1995, 1996, 1997) including those derived from a combination of scenario-planning and focus-groups.

A number of existing hypotheses refer to ‘expectations’ of future outcomes. Perhaps the most important of these, and the most directly related to the new hypothesis, in the specific field of macro-economics, has been adopted by a number of economists as rational expectations. Although this also deals with aggregated expectations, it is a more specific hypothesis, which is normally described in terms of the (economic) actors in the 'market' (usually a financial market) immediately recognising, and discounting, the expected effects of any outside intervention in that market; so that - apart from a role for short-run stabilisation policies (Friedman, 1968) - any such intervention, by government in particular, is bound to be ineffective.

The ‘general hypothesis of aggregated expectations’ does not assume a rational actor. Moreover, where rational expectations is based upon the actions of the informed actors (who are usually quite limited in number) most directly influencing macro-economic outcomes, the new hypothesis typically looks to the very much larger number of uninformed, though not necessarily irrational, participants throughout the whole population.

In general, it is derived from management practice, which has taken a relatively pragmatic approach, looking to alternative forecasts, typically using scenario planning processes; which derive from ‘soft’ practical judgements about the future (usually on the basis of experience) rather than the ‘hard’ theoretical equations favoured by many economists. Scenario planning techniques have been most extensively developed by Shell Oil (Wack, 1991, Schwartz, 1991) with whom we worked - to develop the basic research techniques which now support this present work.

The future outcome of a (macro) issue - economic or political - will be largely determined by the expectations of those, in the population affected, whose aggregated individual decisions will shape that outcome.

The corollary is that changing the expectations of this population will change the (macro) outcome.

The outcome will be decided by many individual decisions[3], such that no one such single decision will unduly affect the overall aggregate decision.

The outcome will not be unduly constrained by any resource scarcity, or bottle-neck.

The separate individual decisions, and hence the overall aggregated outcome, will on average be swayed by the individual expectations[4] of this population, as well as specifically by individual needs and/or wants; and the latter can be assumed to be random in nature, cancelling out when aggregated[5].

In essence, practical use of the aggregated expectations hypothesis involves three stages:

1. observation/measurement of existing expectations - the starting point must be an understanding of what the population's existing expectations are; and, hence, what will happen if no intervention occurs (Mercer, 1997). Without knowing where you are, and in what direction you are currently going, it is impossible to steer a course to the destination you want. The new research techniques, now developed[6], allow this information - which is typically quantified - to be obtained.

2. decision on the possibility of intervention - the next, key, step is to decide whether any intervention - to change the outcomes - might be feasible. This aspect may now be supported by the research techniques recently developed (Mercer, 1997); which are based on qualitative results, but obtained on a scale sufficient to allow a higher degree of confidence than that which normally applies. This approach formalises the recent recognition that governments are not all powerful, but may only be successful in certain classes of action. This basic decision, whether or not to act, leads naturally to two types of subsequent activity. At one extreme, therefore, the expectations may be so firmly held, or the tools for changing them so weak, that it would be unrealistic to expect the outcomes to change. In this case the third step, where intervention simply is not possible, becomes:

3a actions to ameliorate the outcomes - actions are then taken not to change the outcomes, but to alleviate - or intensify (if positive) - the symptoms arising from the inevitable outcomes. Damage limitation may then offer the most practical route.

On the other hand, if it appears possible to modify the expectations, such that more beneficial outcomes will be achieved:

3b modification of expectations - the various tools, including those of marketing but also those of persuasive modelling, can then be deployed to steer expectations in the desired direction. Conviction marketing techniques (Mercer, 1996) are then likely to become the methods of choice.

The main use of aggregated expectations theory is to be seen in the field of macro-economic policy in general, and in particular in terms of its application to real economies. In the specific context of this paper, however, it also suggests - somewhat counter-intuitively in view of its emphasis on practical measurement in the field - that economic models, which have figured so prominently in the theory over recent years, still have an important role. But it sees them as fulfilling a very different purpose, as potentially powerful elements of the communication process, central to the process of persuading the actors involved, rather than as accurate representations of the actual economic processes. Some models, in this context, can offer a very potent means of influencing expectations; as is now recognised by those governments which use interest rates to communicate (signal) their own expectations of future changes in inflation. Paradoxically, one of the best examples was to be seen in the influence of the Chicago School of economists in general, and of Friedmann (1968) in particular; arising from their promotion of the simple model(s) behind monetarism. Their theories were eventually accepted by the majority of decision-makers; and built into their own expectations - in the terminology of the hypothesis - of future developments. Under those circumstances, monetarist theory was able to predict outcomes, for the relatively short time that these expectations were shared by the wider world, and enabled governments to influence those outcomes.

The nature of the model, as a communication device, becomes most apparent when we compare the success of this (monetarist) model with the equal success of that of the previous period - covering several decades - when the very different theories previously put forward by Keynes (1936) were just as well accepted. In their time, they as powerfully influenced expectations, with an even better track record in terms of predictability and of government use.

The change from Keynesianism to Monetarism also illustrated another feature predicted by the hypothesis of aggregated expectations:

When there is a general shift in expectations from a basis in one model to that in another - paralleling Kuhn's (1970) paradigm shift - relative stability (in terms of predictability, at least) is replaced by a period of uncertainty; as the proportion of the population supporting the two competing models progressively shifts from one to the other (and the strength of their individual belief in the new model grows, as that of the old one wanes).

This may, thus, be seen as one contributor to the uncertainty which has been a characteristic of the macro-environment in recent decades. It has been compounded by the need for 'monetarists' to regularly modify their earlier, over-simplistic, theories to allow for discrepancies in the observed outcomes. The shifting nature of this recent (‘monetarist’) theory has possibly been one contributor to the on-going dissonance. In the context of stable, and predictable, expectations it is even arguable that it is better for a government to be consistently wrong - as in retrospect the Thatcher governments were - than to be inconsistently right - as John Major's governments have sometimes been subsequently.

This (expectations) communications aspect of modelling imposes rather different requirements. Previously, models were deemed most academically worthwhile - and supposedly more effectively productive - if they were reducible to exact mathematical equations. Indeed, they were sometimes thought to be even more worthwhile if such equations were so complex ('mirroring the complexity of nature') that only the most erudite elite could understand them.

In the new context, of communicating expectations, such mathematical complexity should be seen to work against many models - not least those developed by econometricians. Instead, in line with the parallel requirements in the commercial sector for 'conviction marketing' (Mercer, 1996), the key parameters for a persuasive model - one which most effectively influences expectations - are:

simplicity and clarity - the concepts have to be easily grasped by the population at large. The very simple message of monetarism, that of the devaluation caused by governments 'printing money', was easy to understand at a superficial level - even if it did not attempt to fully describe the complex issues involved.

distinctive, rich identity - the ideas have to be clearly differentiated from competitive offerings, as was monetarism, but ideally should also tap into a much richer 'value system', as was monetarism supported by the panoply of the (Chicago School) free-market debate.

believability, especially of the champions - the model must ultimately be believed by the majority of the population. This typically occurs because the promoter is believable, as was Friedmann on behalf of monetarism - where his 'opponent', Keynes, was dead.

strength of opponents - indeed, the comparison with the old paradigm, and especially with its supporters, is a key element in the battle for hearts and minds. The Keynesians were handicapped not just by the fact that their leader was no longer there to defend his theories, but - even worse - by the fact that they had foolishly also espoused the Phillips J Curve as a core element of their expanded theory (even though Keynes himself had written his theories, and died, before Phillips had propounded his). They were, therefore, highly vulnerable when Friedmann proved this particular theory to be false!

match to 'consumer' needs - finally, the concept(s) have to resonate with the population on which they are targeted. The main target of monetarism was the business community, and the economists who aspired to mediate between it and government, both of whom found the notion of the free-market emotionally, and practically, satisfying at a time when it was increasingly claimed that there was too much government intervention.

In addition, if such a model is to survive the ravages of time, it needs to incorporate:

ambiguity - such that it can be progressively reinterpreted to meet changing circumstances. This has happened to an extent with monetarist theory - and rational expectations itself can be viewed as one such reinterpretation. Most notably, though, it has also happened to the longest lived models of all; religious texts - such as the bible - which have been interpreted very differently by different sects at different times in history.

The Aggregated Expectations Hypothesis offers a new way of examining likely future outcomes, based upon observations of the most important contributor - expectations - to the individual decisions which aggregate to create the final macro-outcomes. It offers the more powerful actors, especially governments, a new tool for influencing some of those future outcomes. On the other hand, it requires a different approach to the use of economic models - where these should now be judged in terms of their power to persuade economic actors to follow the patterns of behaviour incorporated in the models.

Mercer, D. 1996. A New Qualitative Research Technique for Exploring the Future, Marketing Education Group (MEG) Conference, Warwick (August, 1996)

Mercer, D. 1997. Determining Aggregated Expectations of the Future, Technological Forecasting and Social Change

Friedmann, M. 1968. The Role of Monetary Policy”, American Economic Review, 58, pp 23-41

Wack, Pierre. 1985. Scenarios: Uncharted Waters Ahead, Harvard Business Review, Sep/Oct 1985, pp 139-150

Wack, Pierre. 1985. Scenarios:Shooting the Rapids, Harvard Business Review, Nov/Dec 1985, pp 139-150

Mercer, D. 1997. A General Hypothesis of Aggregated Expectations, Technological Forecasting and Social Change

Shaw, G. K. 1987. Rational Expectations, Bulletin of Economic Research, 39(3), pp 187-209

Keynes, John Maynard. 1936. The General Theory of Employment, Interest and Money, Macmillan, London, 1936

Kuhn, Thomas S. 1970. The Structure of Scientific Revolutions (2nd Ed.), University of Chicago Press, Chicago

The hypothesis is best introduced in terms of a simple example, which parallels some of the earliest work on rational expectations but comes to a rather different conclusion:

Let us take the example of one individual who is asked to contract for the supply of a repeat purchase good at some time in the future. This example is chosen since it offers a more realistic example of price-setting than market-clearing by some form of auction (to an informed public) which is the starting point for much of neo-classical economics. Here, there is no direct balance of supply against demand, and bounded rationality is implicit in the process[7].

It is arguable that the analogue of a 'market price' enters into the equation in the shape of the expectations of the supplier, and of the purchaser, as to what will be - on the date of future supply - the generally prevailing price which they will be required to meet. As this is a repeat-purchased good, as indeed are most consumer goods, the price can be assumed to be based upon the currently applying price modified by an allowance for the expected inflation over the period in question.

Where P_f= Future Price, P_c = Current Price, I_e = Average Expected Inflation, and

The last term allows for the fact that the individual’s information is bounded, and their interpretation of it may not be as rational as many economists’ would wish. If, however, we aggregate a sufficient number of such individual decisions the resulting new aggregate price is:

As suggested earlier, it may then be reasonably assumed that the individual deviations are effectively random and will sum to zero - indeed that is, by definition, the case for the average price[4]. The resulting new ‘market price’ average (Pa) therefore will be:

It is, thus, possible to focus on the overall, average, expectations (of the total population) as to what inflation will be - and this becomes a self-fulfilling (prophetic) expectation.

Even in the case of the Special Hypothesis some individual inputs can distort the overall aggregate. In particular, governments - at the national level - can intervene to influence this process[10]. Indeed, despite economists’ rational expectations, this is arguably the one area where they have learned to intervene in order to manipulate expectations (as the hypothesis allows); most notably, in recent years, by using interest rates as a signal.

It may be possible, however, to adopt at least some of the concepts now employed in marketing practice to benefit from the Aggregated Expectations Hypothesis. Most notably, it is possible to adopt the marketing research philosophy that, if you cannot realistically model individuals' behaviour from first principles, you can 'ask' them what their expectations are - on any given issue. In particular, the critical ‘dimensions’ of the key issues (as seen by the population as a whole) can be derived, and subsequently quantified, by the new research techniques we have developed (Mercer, 1995, 1996, 1997) and, by use of repeat surveys, changes in these can be tracked over time.

The evidence is that the assumptions underlying the hypothesis are met in many circumstances:

It is only in the past half century that this assumption could said to be generally true - but, for the majority of issues across the spectrum, real resource constraints, as opposed to perceived constraints (which affect expectations), can no longer be seen as the major factor determining overall outcomes. This is especially true where developed economies are rapidly moving from a predominance of physical goods to one of intangible services which make relatively fewer demands on resources.

In addition, few issues are now decided by just a few individuals. Over recent years it has become apparent that even national governments cannot control many of the events within their boundaries. Thus, true democracy is gradually coming about not because the individual is the focus of politicians' activities, but because in almost all fields the important trends are now set by the aggregate of individual actions - from brand shares through to inflationary pressures.

Even so, the one factor which sways the overall 'votes' is not individuals’ wishes but their expectations. As we have seen, it can be assumed that, to a degree, individual wishes cancel each other out. More important, individuals - who are more capable and better informed than many politicians would allow for - recognise that the world will not necessarily change to give them just what they want. They do, however, assume that it will change, in line with their expectations. It is for this reason that their expectations are such a powerful predictor of the future.

Gerrard, Bill. 1994. Beyond Rational Expectations: A Constructive Interpretation of Keynes’s Analysis of Behaviour Under Uncertainty, The Economic Journal, 4 (March), pp 327-337

[1] Open University Business School, Walton Hall, Milton Keynes, MK7 6AA, United Kingdom
Telephone: 044(0)1908 655878 Fax: 044(0)1908 655898 Email: d.s.mercer@open.ac.uk

[2] Now also the UK focus of the American Council of the United Nations University (UNU) Millennium Project.

[3] Made by the whole population involved, not just by the key economic actors involved in the rational expectations hypothesis.

[4] Usually observed by practical marketing research techniques, not predicted by theoretical modes.

[5] This is similar to the forecast errors typically allowed for in rational expectations work, which are also essentially random with a mean value of zero (Shaw, 1987).

[6] The qualitative techniques - a combination of focus groups with scenario forecasting - were developed as part of a programme of research lasting more than half a decade; and tested against 17 groups containing managers from 140 organisations. The quantitative work - using semantic differentials to map the importance of the 162 dimensions discovered at the qualitative stage - has been piloted by a survey with respondents from more than 150 organisations (Mercer, 1997).

[7] Hence, one key reason for the divergence from classic rational expectations, even if wage or price stickiness is allowed for.

[8] This compares, for instance, with the rather more complex form of even the ‘simple’ representation of the rational expectations hypothesis given by Gerrard (1994)

_t-1 X^e_t=E(X_t|Ω_t-1)

where _t-1 X^e_tis the rational expectation formed in period t-1 of the variable, X, in period t and E(X_t|Ω_t-1) is the mathematical expectation of X_t conditional on the information set Ω_t-1 available in period t-1 (where the whole is characterised by unbiasedness and orthogonality)

[9] This is a key feature of the generalised hypothesis, as it is for the rational expectations hypothesis, hence the emphasis on aggregation

[10] In contrast with rational expectations theory, which typically only allows for short-run stabilisation policies to succeed (Friedmann, 1968)