The behaviour on which revealed preference valuation methods are based almost inevitably results from thought processes involving bounded rationality.  What does that imply for such methods?

The insights of behavioural economics have been applied to many sub-fields of economics, and environmental economics is quite properly no exception.  If actual human behaviour is often found to differ from that of homo economicus due to bounded rationality, cognitive biases, and pro-social behaviour, then it is hardly to be expected that such differences will not be encountered in human decision-making relating to the environment.  Indeed, the complexity of issues such as climate change and sustainability may suggest that  environmental decision-making is a habitat in which homo economicus is especially unlikely to be found.

Academic surveys of what has been termed ‘behavioural-environmental economics’ include Shogren & Taylor (2008) (1) and Croson & Treich (2014) (2).  There has also been considerable interest by the UK authorities including publications by DEFRA (3) and the Forestry Commission (4).  All of these consider, among other topics, the application of behavioural economics to the valuation of non-market environmental goods.  I am struck, however, by the fact that in each case their focus is primarily on stated preference valuation methods such as contingent valuation.  This is not to deny the relevance of behavioural economics to stated preference methods: issues concerning the cognitive biases of survey respondents and their implications for the framing of questions are an obvious link.  Here, however, I will explore the possible relevance of behavioural economics to two revealed preference valuation methods, the travel cost and the hedonic pricing methods.

The travel cost method estimates the direct use value of recreational sites such as parks from the travel costs incurred by their visitors.  Travel cost is generally taken to be the sum of out-of-pocket expenses such as petrol and the opportunity cost of the time taken.

Decisions on whether to visit a park are unlikely to be made in the manner of homo economicus.  There are likely to be many alternative ways in which a potential visitor might spend their time: visiting other parks at different distances and with different characteristics; enjoying various other leisure activities or entertainments; doing housework, gardening, DIY  or even paid work.  Depending upon the location of the park and of alternative activities, a potential visitor might have a choice of modes of travel and routes.  They might regard some journeys as contributing positively to their whole experience, and others simply as sources of disutility.  Factors they might consider include the weather forecast, the likely appearance and condition of the park at the time of year, and what is known of air quality in the park and elsewhere. Then there are considerations about journeys, such as the likelihood of traffic jams, and the reliability of public transport.  And all that is before the additional complexity arising when a trip has more than one purpose, or when it is a compromise between the differing preferences of group members travelling together.

Note also that the decision whether to visit a park is in most cases not worth thinking about in too much detail.  The worst that can happen, unless something very unfortunate happens, is that a limited time will be spent somewhat less enjoyably or productively than it might have been.

Almost inevitably, the decision whether or not to visit the park will involve bounded rationality.  Even a person who is quite capable of acting as homo economicus for a big decision such as accepting a job offer will be unlikely to take that trouble in deciding how to spend a little of their leisure time.  What most people will do is apply heuristics (mental short cuts) in reaching their decision.  That does not mean making an instant, thoughtless decision.  What it will often mean is focusing on some but not all of the relevant factors.  For example, a person might have regard to the weather but not consider possible travel difficulties; and might consider only a very few possible alternative activities.  And even for the factors they do consider, their thinking is unlikely to go very deep. If they consider the cost of travel by car, for example, they may well roughly estimate the cost of fuel, but have no regard to vehicle wear and tear and the effect of mileage on depreciation (5).

The hedonic pricing method is used to estimate the value of an environmental characteristic of a locality from the prices of its homes relative to those in other localities.  It involves analysing the relation between home prices and the many factors which influence them so as to isolate the effect of the characteristic of interest.

Suppose the method is used to estimate the (negative) value of air pollution in a neighbourhood due to industrial emissions, using house price data from the neighbourhood and from other neighbourhoods in which air pollution is insignificant.  Ideally for the method, the neighbourhoods would be the same in all respects other than the air pollution, but in practice this will never be the case.  Both within and between neighbourhoods, homes will differ in number and size of rooms, quality, size of garden and availability of parking.  Accessibility of  shops, schools and places of work, and characteristics of such key facilities, will differ.  Home locations will also differ in other environmental characteristics such as traffic noise, views, proximity to parks, other forms of pollution, and risks of extreme weather events or earthquakes.

A person deciding whether to buy a particular home therefore has many factors to consider.  Now it is true that buying a home is a big decision which most people will consider quite carefully.  Even so, the complexity of the decision is such that the rationality people bring to bear will probably still be bounded.  One reason is the time and effort that would be needed to  obtain all relevant information.  Take schools, a key issue for many families.  An online search may quickly yield information about a school’s examination results.  But it may be much harder to assess whether good results reflect the school’s teaching or the standards of its pupil intake, or the likelihood that a particular child will be able to obtain a place at that school.  And that information may be needed for numerous schools near different homes under consideration.  Then there are home characteristics.  Perhaps the homes under consideration differ in their heating systems and extent of insulation.  A lot of information might be needed to assess the consequences for comfort and running costs.  As for the air pollution itself, it may be difficult to extract reliable information on pollution levels from competing claims by interested parties quoting different measurements and statistics, and to assess the associated health risks.

Another reason for complexity is that moving home is a decision with long-term consequences.  The health risks from the pollution may accumulate with exposure over many years.  Most fundamentally, because moving home is costly, both financially and in terms of disruption to normal life, people will be extremely reluctant to move on a ‘trial’ basis, that is, with the thought that if the home or its neighbourhood proves unsatisfactory they can simply move again.  Most people buying a new home will expect to live in it for a number of years, and therefore have an interest not only in what the neighbourhood is now like, but also in how it may change in future.  But prediction is difficult: past trends may not continue, and promised improvements may not come to fruition.  And even if the future could be perfectly foreseen, there is the issue of discounting to obtain the net present value of the stream of future benefits and costs from a proposed move.

Given all this complexity, it is hardly to be expected that home buyers will take account of every piece of information relevant to their choice.  Even for those whose inclination is to do as much research as possible, the discrete nature of the housing market and consequent risk that properties on offer will be snapped up by others tends to force buyers who know roughly what they want to make quick decisions.

In summary, therefore, the behaviour from which the travel cost and hedonic methods draws conclusions about environmental values will almost always result from thought processes involving bounded rationality.  That conclusion seems solidly based.  But what it implies for revealed preference valuation is much less clear. The remainder of this post is accordingly much more tentative.

Where rationality is bounded, various cognitive biases may come into play.  It is not difficult, in discussions of behavioural economics, to find lists of such biases.  Sometimes, however, the argument does not go beyond the broad generalisation that, where such biases are present, conclusions based on an assumption of rational behaviour will be unreliable.  That I submit is an oversimplification.

Consider for example confirmation bias.  Suppose a couple are considering buying a home in a neighbourhood which some claim to suffer from significant air pollution.  She, we suppose, really likes the home, and dismisses the air pollution as a minor issue.  He, on the other hand, is less keen on the home, and regards the air pollution as a serious problem.  Looking at it from a neutral standpoint, we may suspect that their thoughts have both been affected by  confirmation bias.  But as this example illustrates, confirmation bias can bias different people in different ways.   There is no obvious reason why the result of a valuation should be biased just because some of the underlying behaviour was influenced by confirmation bias.  It is plausible that the effects of such bias on individual behaviour will tend to cancel out when a valuation is based on large sample data.

Are there, then, types of cognitive bias which can be expected to affect the results of revealed preference valuations because they are likely to influence the behaviour of many individuals in the same way, so that their effects will not cancel out in large samples.  I offer the following suggestions:

Publicity Bias: People considering alternative activities may be unduly influenced by advertising and publicity.  Since the attractions of non-market goods tend to be less publicised than commercial goods (because there is less incentive for their owners or providers to do so), people may visit free parks less and go to paid entertainments more than they would if they chose their activities fully rationally.

Loss Aversion: Most parks are permanent facilities, whereas many alternative activities such as shows and concerts are only available for a limited period.  People may decide to go to such activities rather than visit a park because they would perceive not doing so as a loss in the form of a missed opportunity, whereas they know that they can visit a park at any time.  The same could apply where a more permanent activity offers a discount for a fixed period, an option not available for facilities that are free anyway.

Visibility Bias: People assessing air pollution may be unduly influenced by what they can or cannot see.  Consequently they may underestimate the seriousness of invisible pollution in the form of small particulates.

Name Bias: People considering visiting a park which they do not know may be unduly influenced by whether it has an attractive-sounding name.  Among London parks and open spaces, for example, they might be more inclined to visit St James’s Park, Kensington Gardens or Primrose Hill than Wanstead Flats, Figges Marsh or Wormwood Scrubs.  For home-buyers, this form of bias could be prompted by names of neighbourhoods.

Salience Bias: For people considering buying a home in a neighbourhood without significant air pollution, the presence or absence of such pollution may not be salient.  They may therefore make no allowance in their thinking for the benefits of clean air.

Present Bias: People may underestimate long-term risks relative to present benefits.  They might for example be attracted to a coastal home offering fine views and nearby beaches, but give little weight to evidence of gradually increasing risk from coastal erosion or rising sea-level, or to the likelihood that such physical risks will affect the future sale value of the home long before they result in actual damage.

Probability Neglect: People may ignore real but low risks such as an earthquake in the vicinity of a possible new home or a murder in a park.  They may also greatly overestimate such risks if similar events have featured prominently in recent news, even if the events are far away.  At any one time, either of these effects may predominate.

Taking these forms of bias together, it seems quite possible that the result of a revealed preference valuation could be significantly different from what it would have been if all the behaviour on which it was based had been fully rational (I pass over the pertinent but difficult issue of how the latter might be estimated).

This leads to a fundamental question:

Suppose a revealed preference method estimates the value of an environmental good (or bad), based on actual behaviour, as V, but would have estimated the value,  had all the relevant behaviour been fully rational, as V*.  Should we say that:

1. A revealed preference value estimate based on behaviour that is not fully rational but influenced in a consistent manner by cognitive biases must itself be biased. The true value is V*.  If V* cannot be estimated directly, then an attempt should at least be made to adjust V for the most obvious forms of bias relevant to the circumstances.   OR

2. Value as estimated by a revealed preference method is value to actual people. It isn’t relevant how those people decide on their behaviour.  Consumer sovereignty should apply in valuing non-market goods, just as (in a market economy) it applies to market goods.  The true value is V.

I don’t have a clear answer to this question.  But it is clearly important.  If the answer is 1, then any revealed preference valuation study needs to include a consideration of the effects of cognitive biases such as those listed above, and estimated adjustments for any significant effects.  If the answer is 2, however, we can ignore such cognitive biases even when we know them to be present, and simply conduct revealed preference studies in the way in which they have always been done.

What can be said is that the question is one illustration of a broader debate in welfare economics as to how welfare should be defined and measured in situations where there are indications that people’s behaviour may not be fully rational.  I use the word indications advisedly as there is a line of argument that, notwithstanding appearances to the contrary, all behaviour is fully rational or at least can plausibly be rationalised, albeit sometimes the implied preferences (utility functions) are unusual.  To me that seems unconvincing, for the reasons set out at length above.

Some have drawn a distinction between decision utility, the preferences that would be implied by an assumption of rationality, and true utility, the preferences that really would maximise a person’s well-being (6).  On that basis one can define alternative concepts of welfare or value.  It could then be argued that answers 1 and 2 above simply reflect those two different value concepts.  But such definition-making does not address what is clearly a substantive issue.  If for example a revealed preference value estimate is needed to determine the value at which a non-market environmental good should be included in a cost-benefit analysis of a proposed project, it will not do just to offer two numbers reflecting two different value concepts.  We need to know which is the appropriate one to use.

There is surely much more to be said on this topic?

Notes and References

1. Shogren J F & Taylor L O (2008) On Behavioral-Environmental Economics Review of Environmental Economics and Policy 2(1) pp 26-44
2. Croson R & Treich N (2014) Behavioral Environmental Economics: Promises and Challenges http://www2.toulouse.inra.fr/lerna/treich/BEE.pdf
3. Department for Environment, Food & Rural Affairs (DEFRA) (2013) Behavioural Economics in DEFRA: Applying Theory to Policy https://www.gov.uk/government/uploads/system/uploads/attachment_data/file/223835/pb13986-behavioural-economics-defra.pdf
4. Moseley D & Valatin G (2013) Insights from Behavioural Economics for Ecosystem Services Valuation and Sustainability Forestry Commission  https://www.forestry.gov.uk/PDF/FCRP022.pdf/$FILE/FCRP022.pdf
5. Re the effect of mileage on depreciation see this AA site: http://www.theaa.com/car-buying/depreciation
6. See for example Bernheim B D & Rangel A (2009) Beyond Revealed Preference: Choice-Theoretic Foundations for Behavioral Welfare Economics Quarterly Journal of Economics Feb 2009 p 52