Behavioural Economics and Environmental Valuation

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
  3. Department for Environment, Food & Rural Affairs (DEFRA) (2013) Behavioural Economics in DEFRA: Applying Theory to Policy
  4. Moseley D & Valatin G (2013) Insights from Behavioural Economics for Ecosystem Services Valuation and Sustainability Forestry Commission$FILE/FCRP022.pdf
  5. Re the effect of mileage on depreciation see this AA site:
  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
Posted in Pollution, Recreation | Tagged , , , , , , , , , | Leave a comment

Explaining Environmental Policy Failure

A proposed new approach to environmental policy fails to convince.

Why Environmental Policies Fail by Jan Laitos is a curious book (1). Its subject is certainly important, and it’s hard to disagree strongly with the bleak picture it presents of risks to the environmental conditions needed for humans to live safely  (pp 59-76).  It contains interesting discussions of particular situations and policy failures.  What makes it curious is the abstract framework within which its proposals are presented, based on what it describes as “fundamental laws of nature” (p 171).

The author is a professor of environmental and natural resources law, and it’s interesting  to see how he pigeonholes what economists regard as two key policies to address environmental issues: Pigovian taxes and marketable permits.  Grouped within the heading of economic policies, these form just one of ten items on his list of current environmental policy strategies (pp 139-140 & 145).  Among the others are some which are commonly discussed by economists as alternatives to market-based instruments, eg regulations (command and control), and adjustments to property rights.  Some others are new to me: legal rights for nature (pp 153-9) and human-nature linkages (pp 160-3).

Laitos’ comments on all these policy strategies are generally critical.  In the case of economic policies, his main criticism is that they are based on assumptions that human behaviour is deliberate, rational and utility-optimising. He refers to evidence that people often behave quite differently (pp 32-3, 145-8). But a crucial question, which he does not consider, is how much difference this makes.  Consider a case where emissions of pollutant are initially at a rate of 100 units per day.  Suppose that, applied to relevant data, conventional economic analysis predicts that a proposed tax would reduce the rate to 50 per day.  Given a more realistic view of behaviour, we would not expect the outcome of the tax to be exactly 50 per day.  But will it be, for example, 55 per day or 90 per day?  The former would suggest that the theory works fairly well in this situation (perhaps because departures from rationality largely offset each other in aggregate); the latter that it does not.

Environmental Policy Failure – an Orthodox Account

Laitos’ overall assessment, as the book’s title suggests, is that environmental policies to date have been unsuccessful.  His argument is essentially that severe environmental degradation has occurred despite the presence of many environmental policies (p 61).  I will not dwell on the objection that some policies, for example the Montreal Protocol on Substances that Deplete the Ozone Layer, have been fairly successful (2).  Of more importance here is that he shows little interest in distinguishing types of failure.  The following is I suggest a useful classification:

Type 1 Failure: Environmental policies which fail to achieve their objectives.

Type 2 Failure: Environmental policies which produce undesirable environmental side-effects.

Type 3 Failure: Environmental policies for which the objectives might be considered insufficiently ambitious.

All three types occur (some policies may exhibit failure of more than one type).  As examples I offer the following:

Example of Type 1 Failure:  The US Lead and Copper Rule, intended to minimise lead and copper levels in drinking water (3), which did not prevent lead contamination in Flint, Michigan (4).

Example of Type 2 Failure:  Germany closed eight of its seventeen nuclear reactors in 2011 following the Fukushima disaster in Japan (5).  This significantly reduced its exposure to the environmental and safety risks of nuclear power.  But to ensure adequate electricity supply, Germany has since increased production of brown coal which is an even worse source of air pollution and greenhouse gas emissions than ordinary coal (6).

Example of Type 3 Failure: Australia’s greenhouse gas emissions target for the Kyoto Protocol’s first commitment period was an increase of no more than 8% on its baseline level (7).

The reasons for these different types of failure are likely to be quite different.  Type 1 failures are often due to poor technical advice to policy-makers and/or ineffective administration and enforcement (including regulatory capture by rent-seeking groups).  Type 3 failures tend to result from political compromises between differing interest groups, or between short and long term interests.  Type 2 failures may occur for either of these reasons, or because a particular environmental problem happens to be especially salient when a policy is adopted.  We have, therefore, the outline of an orthodox answer to the question implicit in the book’s title.

Environmental Policy Failure – Laitos’ Account

Laitos, however, largely neglects these sorts of explanation. Instead, he presents a framework of thought, the key elements of which can be summarised as follows (8):

Nature as a Complex Adaptive System, not necessarily in equilibrium, characterised by co-evolution of living organisms and their abiotic surroundings, and often achieving resilience between stability and instability in the state sometimes termed ‘the edge of chaos’ (pp 95-100).  This leads to the idea that environmental policy should not aim to conserve or restore a state of nature, because there never was such a stable baseline state (p 100).

The Requirement of Symmetry as fundamental to the laws of nature, in physics and in biology (p 174).  The following three consequences of symmetry are of particular relevance to environmental policy.

The Law of Conservation according to which certain quantities cannot be created or destroyed, an example being the First Law of Thermodynamics  (p 177).  This leads to the conclusion that, because the natural environment is a complex adaptive system, environmental policy should be simple, and complex policies are likely to fail (p 179).

The Equivalence Principle according to which seemingly different concepts are actually the same, an example being the equivalence of gravity and acceleration in Einstein’s general theory of relativity (p 180).  This leads to a rejection of the ideas that humans are morally superior to their environment, and have the ability to manage and master it.  Instead, environmental policy should treat humans and their environment as equivalent parts of a social-ecological system (p 181).

The Unification Principle according to which apparently opposite concepts are actually the same (pp 181-2.  Again, there are examples from physics.  The conclusion drawn is that environmental policy should address not simply the environment but the whole social-ecological system (p 183).

According to Laitos, therefore, environmental policies do not work well for four main reasons: they are based on an inaccurate model of nature; they often presuppose human superiority; they are inconsistent with the law of symmetry; and sometimes they are based on a false model of how humans behave (p 184).

The most persuasive part of this is the view of nature as a complex adaptive system.  We know for example that there are complex interdepencies between the biotic and abiotic elements of the environment, via processes such as photosynthesis and carbon sequestration.  We also know that such natural processes have not always been in equilibrium.  The oxygen content of the atmosphere, 21% at present, has during the last billion years been as low as 3% and as high as 35% (9).  Even within the much more recent past, there have been dramatic changes such as the retreat of glaciers at the end of the last ice age about 12,000 years ago (10), and a period about 10,000 years ago when much of what is now the Sahara desert received sufficient rainfall to support a savanna-like environment (11).  Certainly, therefore, environmental policies should be developed with awareness of the complexity of natural processes and the possibility that the past may not be a good guide to the future.

On the other hand I am not persuaded by the ‘no natural baseline’ argument. Consider how it might apply to the issue of waste plastic polluting the oceans with adverse consequences for wildlife and indirectly (via food chains) for humans.  The argument would presumably be that the former plastic-free ocean was just one state which had persisted over a certain period and which, given the complexity of nature, might have come to an end without human intervention.  Hence to regard plastic-free oceans as a baseline to be restored if possible would be misguided.  Now that argument seems to me mistaken: it ignores both the implausibility of plastic being produced by natural processes, and the fact that, for thousands of years, fish caught from plastic-free oceans have been an important component of human diets.  What’s more, it illustrates a general shortcoming of Laitos’ approach.  His criticism of actual environmental policies contains many examples of particular policies in particular situations.  But in his presentation of his own policy ideas, I can find little indication that he has tested his abstract principles by considering what they would imply for specific situations.

So far as the requirement of symmetry is concerned, I will not comment on its application to theoretical physics, which is not my specialism.  But I have to say that I find the drawing out of consequences for environmental policy not just unconvincing but in places positively bizarre.  Nevertheless, let us consider the three conclusions on their merits, ignoring their provenance.

Simplicity can indeed be a virtue of environmental policy.  The five pence charge for single-use plastic bags introduced in England in 2015 is an example of a fairly simple policy (12).  It’s easy to think of ways in which the policy might with apparent justification have been made more complex, such as different charges for different bag sizes or exemptions for those on low incomes.  Nevertheless, the policy seems to have been widely accepted.  One likely reason is that the charge is too small to affect anyone’s vital interests.  Another is that it seems obvious that the administrative costs of the sort of complications listed above would greatly outweigh the benefits.

On the other hand, consider the case of international negotiations on reducing greenhouse gas emissions so as to mitigate climate change.  Here vital interests are at stake, along with different understandings of what the scientific evidence shows and different views of fairness in apportioning reductions.  A requirement that any agreement be simple, such as a uniform percentage reduction in emissions, would be fairly certain to ensure failure via a collapse in negotiations. Acceptance that an agreement will need to be complex keeps open the possibility of a partial success – an agreement that if implemented would lead to a reduction in emissions, albeit not as large a reduction as many consider necessary.

Laitos’ rejection of human superiority might suggest that he is an advocate of animal rights.  In fact, the concept of animal rights receives only one brief mention (p 156).  Nor does he ascribe value to nature generally, except in so far as the degradation of nature affects humans (p 38).  The superiority that Laitos rejects is the idea that humans have the ability to manage the environment (pp 29 & 181).  This highlights another weakness of his approach.  Perhaps as a consequence of his high level of abstraction, he tends to make statements that are too absolute and do not allow for matters of degree.  If he had just said that, when trying to manage their environment, humans often fail due to lack of ability, or make mistakes due to lack of knowledge or cognitive errors, then it would have been easy to agree.  But to deny outright  that humans have the ability to manage the environment is too strong.  Human actions do affect the environment, and humans do have the ability to choose among different actions that have different consequences.

If however one really did consider that humans cannot manage their environment, then the natural conclusion would be that any environmental policy is doomed to failure.  It is hard to see how such a belief could be a basis for choosing one policy rather than another.  Yet as we shall see, Laitos does advocate a particular type of environmental policy.

Laitos’ third conclusion – that environmental policy should address the whole social-ecological system – could be taken as just a way of saying that policy formulation should have regard to political feasibility (the social part of the system) as well as to its effect on the environment (the ecological part).  If so it would be entirely acceptable.  But what he actually means by this seems implicit in his rights-based approach to environmental policy, to which I now turn.

Laitos’ Proposed Environmental Policy: A Rights-Based Approach

Laitos proposes that environmental policies consistent with his framework would confer a certain type of right and impose a corresponding duty (p 36).  The right would be:

Positive, that is, a right entitling the holder to the help of others (as opposed to a negative right which merely requires others not to take certain actions) (pp 38-9 & 185-6).

Held by the social-ecological system, that is, by both humans and the environment. Thus it would be quite different from a specifically human right such as a right to clean water (p 185).

To environmental conditions in which humans can live safely (p 185). Thus although the right is held by humans and the environment, its purpose is solely to meet a human requirement, not to save or protect the natural environment (p 187).

Inclusive of the power to alter the right (pp 191).  Thus any legal right granted to the social-ecological system should be capable of being amended (p 192).

Laitos’ account of this right leaves unaddressed the key question of who would grant it.  The answer, at his abstract level, must be humans.  But that leads back to issues of political authority and international negotiation.  For he consistently identifies the right-holder as the social-ecological system. He is not talking about local ecosystems and their human inhabitants, on which one might envisage rights being conferred by local or national governments, but about one system consisting of the global environment and the whole of humanity.

That the right should be capable of being amended seems to amount to saying that there should be flexibility to adjust environmental policies in the light of circumstances (p 192).  As such it seems entirely acceptable.  But again it raises issues of authority which Laitos does not address.  How far should administrators have delegated authority to adjust policies (with risks of uncertainty for those affected and even outright abuse), and how far should changes be a matter reserved for law-makers?

So far as enforcement of the right is concerned, this is best considered after introducing the corresponding duty.  This duty is:

Imposed on humans only, not on the whole social-ecological system (p 185).

Affirmative, that is, a duty to provide something, not merely to refrain from harm (p 197).

To create public environmental goods and positive externalities (p 197).  In this case Laitos does give an example: a person planting a tree which will sequester carbon and so contribute to mitigation of climate change.

Laitos supports the affirmative nature of the duty with the argument that, as a matter of psychology, people respond better to being told what to do than what not do.  This seems problematic.  For one thing, the difference is often just a matter of polite wording, which is important in everyday intercourse (compare “Could you close the door” and “Could you not leave the door open”) but should not be a key consideration in framing laws.  For another, being told what to do can be far more restrictive of individual liberty than being told what not to do (compare requiring a car-owner to travel by public transport, or not to exceed speed limits).  Most fundamentally, the natural wording of some essential environmental policies is negative.  Consider safety policy for nuclear power stations. One might try to formulate a safety policy in terms of ‘preserving a radiation-free environment’, but the more obvious formulation is ‘avoiding radiation leakage’, and any idea that companies would respond better to the former than the latter is rather implausible.

The content of the duty is not specified in any detail. But let’s explore the tree-planting example. Suppose policy required every adult to plant at least one tree a year. That could help in mitigating climate change (though it is pertinent to ask how much else would be required to maintain the habitability of the planet).  But it could waste resources to little effect in places where the climate or soil is unsuitable for tree growth.  It could also do harm where tree-planting conflicts with other important land uses, or where the uptake of water by trees within a river basin would lead to reduced water flows with adverse downstream effects on agriculture or wetlands.  So a sensible policy focused on tree-planting would have to be more complicated, perhaps permitting people in places unsuitable for trees to enter into offsetting arrangements with those in more suitable places.

One way to give content to the duty while addressing such issues would be a policy telling each person in what way they should contribute to public environmental goods and positive externalities, but with the contributions tailored to people’s local circumstances: tree-planting in region A; rainwater harvesting in B; fish restocking in C; and so on.  At the other extreme, another would be to allow people discretion as to the nature of their contributions, but to require each person to submit an annual statement of their contribution for review by the authorities.  Intermediate arrangements would probably be more workable, for example, a list of acceptable contributions from which people could choose any one, or a system of points for contributions with an annual points target for each person.

What emerges from this is that giving practical form to a duty to create public goods and positive externalities would not meet Laitos’ requirement that policy should be simple.  Any apparent simplicity is entirely due to Laitos’ abstract presentation of his ideas.  Putting it into practice would raise the same sorts of issues as existing environmental policies, such as whether policy-makers have adequate technical advice, conflicts between different environmental objectives, and political feasibility.

Then there is the issue of enforcement.  Laitos states that environmental policy needs a “stick” instead of “carrots”, contrasting the “carrots” of subsidies with the “stick” of his positive duty (p 200).  This is an odd comparison for two reasons.  Firstly, subsidies are only one of many instruments used by existing environmental policies, and as a means of incentivising firms to reduce pollution are generally considered by economists to be inferior to taxes (because subsidies can also encourage new firms to enter polluting industries).  Secondly, a duty is not itself a “stick”. A “stick” would be a penalty that an authority could impose for failing to fulfil a duty.  One might have expected here a discussion of the process for imposition: how evidence might be gathered; whether the process would be administrative or judicial; types of penalty; and so on.  As throughout the positive parts of his book, however, Laitos seems uninterested in such details.

In summary, therefore, although Laitos is right to highlight the failure of many existing environmental policies, the alternative he offers owes whatever plausibility it may have almost entirely to its abstract formulation.  As soon as one attempts to flesh it out with practical detail, it becomes apparent that it would encounter much the same complexities and difficulties as existing policies.

Notes and References

  1. Laitos J, with Okulski J (2017) Why Environmental Policies Fail  Cambridge University Press.  All page references above are to this book.
  2. Wikipedia: Montreal Protocol – Effect
  3. US Environmental Protection Agency: Lead and Copper Rule – A Quick Reference Guide
  4. Wikipedia: Flint Water Crisis
  5. Wikipedia: Nuclear Power in Germany – Closures and Phase-Out
  6. Der Spiegel (22/11/2017) Can Germany Break its Lignite Habit?
  7. UN Framework Convention on Climate Change – Kyoto Protocol – Targets for the first Commitment Period
  8. My policy in this blog when setting out the views of others is normally to paraphrase and rarely to quote. This is partly because a paraphrase highlighting key points can be more concise than a quotation, and partly for copyright reasons. I hope that my paraphrases are fair and, as ever, would appreciate advice of any misrepresentation.
  9. Wikipedia: Atmosphere of Earth – Evolution of Earth’s Atmosphere
  10. Wikipedia: Last Glacial Period
  11. Carey B, Live Science (20/7/2006) Sahara Desert was Once Lush and Populated
  12. The government regulations on the 5p charge are at
Posted in Environment (general) | Tagged , , , , , , | Leave a comment

Improving on GDP

The winning entries for the Indigo Prize highlight both the importance of improving economic statistics and some of the challenges of doing so.

The launch of the Indigo Prize by LetterOne is a welcome initiative which recognises both the limitations of conventional economic statistics and the need for a little razzmatazz to bring the issue to the attention of a wider public (encouragingly there have been reports in several UK newspapers).  Entrants were asked to submit a 5,000 word article presenting a design for a new economic measure for global economies and explaining how it should be used to improve the measurement of GDP.  The new measure was to take full account of social and economic factors and the impact of creativity, entrepreneurship and digital skills.

The remit made no mention of environmental or natural resource issues.  A clue as to why may perhaps be found in the contribution on the Indigo website by Mikhail Fridman, a co-founder of LetterOne, in which he rejects the view that there is an imminent problem of depleting resources.  Fortunately, both entrants and judges appear to have taken a broad view of their task.

Given the current state of knowledge, no entry could have been expected to offer a definitive solution to the problem of economic measurement.  Nor is it the sort of problem that lends itself to such a solution: what is best will depend on economic circumstances and (so far as high level presentation is concerned) on the capacity of policy-makers and the public to interpret economic indicators.  Nevertheless, both of the joint winning entries are insightful and thought-provoking (and contain much more than I shall mention here).  That by Haskel et al begins by identifying what it regards as two key strengths of GDP, namely, that it avoids double-counting of intermediate outputs, and that it adds outputs of different goods in an objective way using market prices as weights. Recognising however the limitations of GDP, it proceeds to identify various ways in which GDP could be “fixed” and “extended” to address those limitations while preserving its strengths. It outlines how it might be possible to define and estimate a single measure of well-being that would be a substantial improvement on conventional GDP.

The entry by Coyle and Mitra-Kahn proposes, for the long term, an approach which differs from that of Haskel et al and of conventional GDP in two key ways.  One is that it focuses on assets rather than annual flows.  If such an approach were adopted, people would no doubt consider annual increases or reductions in assets as well as their absolute magnitude. But that itself is quite different from focusing on annual GDP, much of which (eg most consumption) does not contribute to the maintenance and growth of assets.  The second difference is that it rejects the idea that economic debate should focus on a single key number, proposing instead a dashboard presenting separate numbers for broad asset groups.

Of these differences it is the first that is the more radical.  It has never been the case that public debate on economic matters focused solely on GDP.  A fairly conventional ‘dashboard’, though perhaps not presented as such, would include measures of output, employment, inflation, poverty and perhaps a few more variables.  The interesting question here is how far it is sensible or meaningful to aggregate items using weights, with the implication that categories that cannot sensibly be aggregated should be treated as separate indicators within a dashboard.  But this is really two questions, one presentational and one technical.  For presentation to the general public, and as a high-level framework for public debate, a dashboard of six indicators (the number proposed by Coyle and Mitra-Kahn) seems about right, in the sense that it would inform but not overload with detail the lay person with no special interest in economic policy.  The technical question is how many indicators we need to include relevant items that cannot meaningfully be aggregated using weights, and the answer to that question is probably many more than six (within natural capital, for example, consider whether it is meaningful to aggregate petroleum, fresh water, forests and fisheries).

So what do the winning entries say about environmental and resource issues?  As part of ‘fixing’ GDP, Haskel et al propose treating enhancement of the natural environment (eg planting a forest) as investment (an addition to GDP) and degradation (eg destruction of a coral reef) as disinvestment (a deduction).  Presumably they would treat depletion of minerals in a similar manner.  This approach requires valuation of the enhancement and degradation, which is challenging both in principle and in practice.  Taking the coral reef example, an issue of principle is whether the loss of value attributed to destruction of a given quantity of reef should be constant, or depend on how much coral remains elsewhere in a country’s territorial waters, or elsewhere in the world (on the basis that scarcity of a good  increases the marginal value per unit of what is left).  A practical issue is the collection and assessment of data from below the ocean surface.

A more fundamental problem is how the measure obtained by adjusting GDP is to be interpreted. Conventional GDP is, at least, a fairly good measure of economic activity (though it could be improved even in that respect).  The measure that would be obtained after the above adjustments would be such that a year-on-year reduction could reflect either a fall in economic activity or a reduction in net enhancement of the natural environment.  Most people, I imagine, would want to know which, rather than focusing on a measure which combines both.

As part of ‘extending’ GDP, Haskel et al propose that it should count future as well as current consumption (a line of thought which may suggest a concern with resource depletion, although it is introduced in the context of the balance between consumption and investment).  Their reference to Weitzman (1) indicates that what they have in mind here is that suitable adjustments can convert GDP to a concept of net national product (NNP) which measures what Weitzman calls “the stationary equivalent of future consumption”.  In my view this is not a workable or especially useful proposal.  It isn’t workable because the adjustments needed to obtain Weitzman’s NNP depend upon shadow prices of capital goods, and these shadow prices are obtained by solving a long term dynamic optimisation problem requiring assumptions about production technology far into the future (2).  It isn’t very useful because, although a future-oriented NNP concept may sound as if it has something to do with sustainability, “the stationary equivalent of future consumption” is not the same as sustainable consumption.  The relation between the two concepts is quite complex, as explained in this post, and if (implausibly) we had the data needed to calculate the former, we would be able to calculate sustainable consumption directly.

Within the dashboard approach of Coyle and Mitra-Kahn, natural capital is one of the six groups of assets.  Their definition of natural capital as “the renewable resources provided by nature” (p 6) should presumably have read “renewable and non-renewable”, since their chart showing natural capital declining in England Wales (p 12) includes minerals, oil and gas.  They state that measurement of natural capital is currently very incomplete and based on market prices.  The former is correct, but the latter seems to refer to official statistics, and overlooks the considerable academic literature on non-market valuation techniques including the hedonic pricing method (for local environmental quality) and the travel cost method (for recreational sites).

I would agree however that, in the context of a high-level dashboard of six items, it is appropriate that one of the six should relate to the environment including natural resources, and should be a capital measure.  A slight disadvantage of such a measure is that it will tend to ignore the short-term (though often recurrent) effects of non-cumulative pollutants.  Greatly outweighing that is the fact that most of the more important environmental and natural resource issues (eg depletion of minerals, deforestation, over-fishing, greenhouse gas emissions) involve long-term effects that a suitable capital measure should reflect.

Coyle and Mitra-Kahn note that valuing their asset groups, including natural capital, is a major challenge.  I wonder however if it is necessary or appropriate to value each group.  If each item in a dashboard is quantified in monetary terms, that could be taken as an invitation to add the values together and focus on the total rather than the separate items.  Having distinct indicators that cannot be aggregated would seem more consistent with a dashboard approach (think of a car dashboard including speed, fuel level and oil pressure, each in separate units that could not be aggregated, even if bizarrely someone wished to do so).  In the case of natural capital, there is also what might be termed the ‘baseline problem’.  For example, should the UK place a positive value on the fact that it does not suffer from serious earthquakes?  Or should the valuation baseline be defined to include ‘absence of serious earthquakes’ so that the value of natural capital for Japan, say, would include a negative element for earthquake risk?   There is a case therefore for including natural capital in terms of some non-monetary measure based on an assessment of its adequacy to sustain current living standards into the future, although so far as minerals are concerned that leads back to the issue of assumptions about future production technology.  So although the idea may have merit, working out the detail would be just as challenging as trying to value the whole of natural capital.

Notes and References

  1. Weitzman M L (1976) On the Welfare Significance of National Product in a Dynamic Economy  The Quarterly Journal of Economics  90(1) pp 156-162
  2. Weitzman, as above: the “investment prices” he introduces on p 158, which vary over time, would have to be obtained as shadow prices in the solution of the dynamic optimisation problem set out on p 159. The relevance of future production technology is implicit in Equation (6) on p 158 which features the marginal product of each capital good.
Posted in Environment (general), Macroeconomic Theory | Tagged , , , , , , , | Leave a comment

A Difficulty in Assessing Sustainability

Long-term forecasting of the aggregate production function is essential for assessing sustainability, but very difficult.

Is our current living standard sustainable?  Or does depletion of non-renewable natural resources such as fossil fuels and metal ores mean that it must eventually decline?  And if the latter, what lower standard of living, if any, would be sustainable?  A well-founded sustainability criterion, enabling us to give clear and convincing answers to those questions, could make an important contribution to public debate on economic policy.  In particular, it could help counter the view that growth in GDP is the key indicator of economic performance.

Qualitatively, it is fairly clear how it might be possible to sustain consumption while non-renewable resources are progressively depleted.  We start from two reasonable assumptions:  that production technology will permit a considerable degree of substitution between inputs, and that there must always be at least some input of non-renewable resources. Inputs of non-renewable resources must therefore be ever-decreasing so that reserves are never totally exhausted (in mathematical terms, the stock of non-renewable resources must be asymptotic to zero).

To maintain output and consumption in those circumstances, there must be increasing inputs of reproducible capital (which includes both capital equipment and infrastructure and renewable natural resources).  Furthermore, it will not be sufficient simply to maintain output: output must grow to provide not only for consumption but also for the necessary investment in reproducible capital and for offsetting depreciation of the increasing stock of that capital.

Whether this scenario is feasible will depend on a number of variables: the initial stocks of capital and resources, resource extraction costs, production technology, population and labour inputs, current consumption and (if we are considering the economy of one country rather than of the whole world) opportunities for trade. There are difficulties of measurement and/or forecasting associated with several of these variables. This post will argue that, in particular, the forecasting of production technology presents a major challenge in assessing sustainability.

That might seem a surprising claim.  After all, economists have put much effort into the estimation of aggregate production functions and into economic forecasting using models in which a production function is one component (1).  The problem is that, to assess long-term sustainability, we need to estimate the relation of output to inputs for future periods in which the ratio of reproducible capital to input of non-renewable resources will be vastly greater than at present.  And that is very difficult.

To illustrate the point, I shall consider the assessment of sustainability within a model having the following simplifying assumptions:

  1. Output of a single good which can be either consumed or invested as reproducible capital.
  2. A Cobb-Douglas production function with technical progress Y = AKαRβ(1.01t) , where K is man-made capital, R is use of a non-renewable resource S, extracted at nil cost, t is time in years from the present, and A, α and β are parameters.
  3. Constant population and labour (this is why, although production requires labour, labour does not appear as a variable in the production function).
  4. Depreciation of reproducible capital at 5% per annum.
  5. Initial stocks: 100 units of K and 100 units of S (the respective units need not be the same).

The assumption of technical progress at a constant 1% per annum, implicit in 2 above, is made to facilitate a focus on the rest of the production function and keep the mathematics of the model reasonably simple.  The actual rate of future technical progress is difficult to forecast, but that is a reinforcement, not a criticism, of the argument presented here.

The question I shall explore is whether, within this model, annual consumption of 100 units (the same units as those of K) is sustainable.  This will depend on the parameters of the production function.  Consider the following functions:

F1:  Y = 15K0.3R0.2(1.01t)

F2: Y = 24K0.3R0.1(1.01t)

This pair of functions has the following property: if K approximates to 100 and K/R to one, then they yield very similar values of Y (because 15 x (1000.1) ≈ 24).  This is illustrated, for the case t = 0, by Chart 1 below.

Suppose we had to decide whether the production function is F1 or F2 on the basis of input and output data for recent periods in which K/R is in the vicinity of one.  Given the inevitable random variation in output due to other variables, this would be extremely challenging. Statistical tests would at best point to one function as slightly more likely than the other.

At the very much higher K/R ratios that sustainability requires, however, the difference between these functions becomes very significant, as shown in Chart 2 below (note the log scale on the horizontal axis).

Because the same inputs yield more output with F2 than with F1 whenever K/R exceeds one, we expect that sustainable consumption will be greater given F2. To find out how much difference this makes I set up the model in discrete form as a spreadsheet with one row per year for 5,000 years.  Key features of the spreadsheet are as below:

  • Consumption in each year equals that for year 0.
  • Growth of capital equals output less consumption less depreciation.
  • Marginal product of capital is calculated directly using the standard formula (for a Cobb-Douglas production function) MPK = αY/K.
  • Marginal product of the resource is calculated directly (MPR = βY/R) for year 0, but subsequently using the Hotelling rule according to which the rate of growth of the marginal product of the resource equals the marginal product of capital (2).
  • Extraction and use of the resource after year 0 is calculated backwards from its marginal product and output (to avoid circularity, the output figure used is that for the previous year).
  • Resource stock after year 0 is that for the previous year less extraction and use for the previous year.
  • Consumption and extraction / use of the resource for year 0 are trial values. A pair of trial values is considered feasible if it generates time paths for the variables in which, during years 0 to 5,000, the resource is not exhausted and output is never less than consumption.

The maximum trial value of consumption consistent with feasibility as defined above is an approximation to maximum sustainable consumption (it’s only an approximation because of the discrete spreadsheet approach and because of the 5,000 year time horizon).  It was found that:

  • With production function F1, constant consumption of 100 units is unsustainable as maximum sustainable consumption is approximately 79.7 units.
  • With production function F2, constant consumption of 100 units is sustainable as maximum sustainable consumption is approximately 117.6 units.

So the assessment of whether consumption of 100 units is sustainable rests on the weak foundation of whether F1 or F2 in Chart 1 above provide a better statistical fit to current and recent data.

A possible objection is that the implicit sustainability criterion, that is, the feasibility of consumption at 100 units per annum forever, or for 5,000 years, is too demanding.  Suppose instead, therefore, that we set the time horizon at just 100 years.  Surprisingly, perhaps, this makes very little difference to the conclusion.  For F1, maximum sustainable consumption increases from 79.7 to 80.0 units.  For F2, it increases by such a tiny amount that on rounding to one decimal place, maximum sustainable consumption is still 117.6 units.

Addendum 1 November 2017

A further possible objection is that if the initial stock of the resource is 100 units, then it is not very plausible to assume that resource use in the preceding years used to estimate the production function was in the region of 100 units per annum.  To address this, I considered the effects of assuming initial stocks of the resource of 1,000 units and 10,000 units (in each case with initial capital of 100 units and over 5,000 years).  With 1,000 units, maximum sustainable consumption was 138 units with F1 and 151 units with F2.  With 10,000 units, maximum sustainable consumption was found to be higher with F1, 239 units, against 194 units with F2.  The important point, however, is that for each of these variations, maximum sustainable consumption depends significantly upon the production function.

The spreadsheet used to derive the above charts and results may be downloaded here:

Constant Consumption with Different Production Functions Adam Bailey

Notes and References

  1. One example is the NIESR’s NiGEM model: see
  2. A derivation of this form of the Hotelling rule may be found in Perman R, Ma Y, McGilvray J & Common M (3rd edn 2003) Natural Resource and Environmental Economics Pearson Addison Wesley  pp 660-1, the rule being equation 19.42j.
Posted in Macroeconomic Theory | Tagged , , , , , , | Leave a comment