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

A Valuation Case Study: the Great Barrier Reef

A valuation of the Great Barrier Reef illustrates many common issues in the economics of environmental valuation.

A recent report by Deloitte Access Economics (1) estimates the asset value of the Great Barrier Reef at A$56 billion (2).  This is the sum of:

  • Value to local recreational users and tourists from elsewhere in Australia: A$32 billion, estimated by the travel cost method (pp 39 & 40).
  • Value to Australians of the Reef’s existence, regardless of having visited it or intending to do so; A$24 billion, estimated by contingent valuation (p 34).

The main data source was a representative survey of over 1,000 Australians (p 30). The survey also included 500 residents of 10 other countries world-wide.

The report also identifies the following components of value for which it does not offer numerical estimates:

  • Value to tourists from outside Australia: not estimated due to data limitations (p 85).
  • Existence value to non-Australians: not estimated as difficult to allow for contextual cultural factors, language barriers and purchasing power differences in using survey responses from some countries (p 35).
  • The value of certain ecosystem services provided by the Reef such as maintenance of water quality and storm protection for the adjacent coast (pp 40 & 42): not estimated as difficult to separate from ecosystem services provided by neighbouring terrestrial and river ecosystems (p 84).
  • “Traditional owner value” to peoples who have lived in the vicinity of the Reef since before Australia was settled by Europeans, relating to cultural heritage, sacred sites and archaeological sites: not estimated because such unique sites lack the substitutability on which non-market valuation relies (pp 45-7).
  • Brand value, that is, the contribution of the Reef to “Brand Australia”. This is given a whole chapter of the report (pp 50-7) which is not easy to summarise.  One key point is how far the contribution is from the existence of the Reef and how far from perceptions of Australia’s performance as guardian of the Reef (p 56).

Taking the above together the report suggests, reasonably, that the full value of the Reef is much more than A$56 billion.  Some, though not all, of the points below tend to reinforce that conclusion.

There are other components of value which the report does not mention.  One is the possible medicinal use of plants and animals found in the Reef (3).  A negative component – illustrating the general point that natural capital is often associated with natural hazards  –  is its role as a shipping obstacle and hazard (4).

As is common with environmental valuation studies, therefore, the components of total economic value fall into three categories: those that can be measured with the available data; those that are in principle measurable but for which the necessary data is difficult to obtain; and those that are intrinsically unmeasurable (but still perhaps important).

As a contingent valuation sceptic (see this post), I would have been inclined to put the existence value of the Reef in the unmeasurable category.  A specific difficulty with the approach adopted is that respondents were asked how much they would be willing to pay weekly to “guarantee” that the Reef is “protected”, having been told that such a payment would be in the context that all Australians would have to pay (p 80). This is a strange question given that global climate change is a major threat to the Reef.  It seems quite possible that some respondents viewed the payments as to guarantee that the Australian government would do what it could to protect the Reef (eg from risks originating within Australia such as mining activity), while others viewed them as guaranteeing the preservation of the Reef, implying protection against all risks including climate change.

The remainder of this post considers the method used to estimate the value of the Reef to tourists from within Australia. The individual version of the travel cost method was used, given that individual data was available and showed good variation between individuals in visit numbers (p 78), so that it could be expected to give more precise results than the alternative zonal method (5).  The variation between individuals was not just due to luck: it was facilitated by the decision to ask respondents how many times they had visited the Reef in the last five years (not merely the last year).

With the individual method, the first step is to regress individual visit numbers against individual travel costs to the Reef and other variables that may influence visit numbers.  The report does not state the full regression model, but from the survey questions (pp 76-82) it appears that the other variables included age, gender and education.  Variables that seem not to have been included are income and travel costs to one or more substitute sites.  Although income is mentioned in discussion of the method (p 85), there is no indication that income data was collected.  This may have been because it was anticipated that such data would be difficult to obtain (US studies often collect income data but it may be that willingness to disclose one’s income varies between cultures).  The issue of substitute sites may have been ignored because of the complexity of allowing for many possible substitutes, and lack of consensus in the literature as to how this should be handled.  Whatever the reasons, omission of these variables can result in omitted variable bias leading to a biased estimate of the travel cost coefficient and hence of the site value, although the magnitude and direction of bias will depend on the circumstances (6).

Mention of substitute sites may seem surprising given the unique nature of the Great Barrier Reef.  But a substitute, in economics, does not have to be a perfect substitute.  One definition is that two goods are substitutes if their cross-elasticities of demand are positive.  Applied to two tourist sites, that would be the case if a higher travel cost to either were associated with a higher demand for visits to the other – a plausible scenario.

The decision effectively to ignore travel time (p 85) is surprising. Common practice is to include the value of time as well as expenditure on transport and accommodation within travel cost to the study site.  The report cites three Australian studies which applied a zero value to travel time, but this is rather selective even among Australian studies (7), and certainly not representative of the global literature including US and UK studies.  Admittedly there is no consensus on how the value of time should be determined.  But to assume a nil value is likely to result (other things being equal) in under-estimation of the value of the Reef.  A fairly simple and still conservative alternative would have been to value time at a suitably small fraction (say a quarter) of the average hourly wage, and to include the effects of different fractions in the sensitivity analysis.

The statistical methods used to estimate the trip-generating function and then derive the value (consumer surplus) per visit (A$662) are not described in full, but appear from the outline provided (pp 85-6) to have been appropriate.  As is common in individual travel cost studies, negative binomial regression was used because of overdispersion (p 86), which is a slightly misleading term for a feature (not a defect) often found in the distribution of individual visit numbers, namely that the variance is greater than the mean.

I would have liked to see some consideration of what potential visitors might have done instead if the Reef had not existed.  Many would probably have visited other tourist sites, increasing the value of those sites and offsetting to some degree the lost value associated with the Reef.  A case can be made that the most useful concept of value for a recreational site is not the gross site value but the contribution which the site makes to the total value of all such sites (8).

To obtain the total annual value to tourists, value per visit was multiplied by the annual number of visits to the Reef, which was estimated as 2.3 million (p 86).  This number was inferred from Tourism Research Australia (TRA) data on numbers of visits within Australia to regions adjacent to the Reef.  However, these numbers include visits to the regions but not to the Reef itself, and some crude round-number assumptions, claimed to be conservative, were made to eliminate such visits (p 86).  An alternative approach would have been to include in the survey a question designed to distinguish visits to the Reef itself from visits to the adjacent regions, and to use that data to estimate, for all visitors to those regions, the proportion who visit the Reef itself.

The final step was to capitalise the stream of annual values so as to obtain the asset value.  The result is heavily influenced by the choice of time horizon and discount rate, as is illustrated by the sensitivity analysis (p 88).  A time horizon of 33 years was adopted, one stated reason being the severe threats to the future health of the Reef (pp 87-8).  Within that time frame, annual consumer surpluses as calculated from the survey data were discounted at a rate of 3.7%, determined using the Ramsey formula (p 87):

Social discount rate  =  Rate of time preference + [Annual growth rate of consumption

                        x Minus the elasticity of marginal utility with respect to consumption]

Given assumptions of a very low rate of time preference (0.05%) and an elasticity of 1, the discount rate largely reflects the growth rate of consumption, which was assumed to equal the average GDP growth rate over the previous 30 years (which can be inferred to have been 3.65%).

While the particular figures used in the capitalisation could be challenged, there are some more fundamental issues here.  One could value the Reef on the assumption that it will degrade along a defined path, or on the basis of its remaining in its current condition.  The key question here is not which of these scenarios is more likely, but which basis of valuation will yield more useful information.  If what we are interested in is the value which is potentially at risk from degradation of the Reef, then it is valuation on  the ‘current condition’ basis which is more useful, and the ‘degradation’ argument for the 33-year time horizon does not apply.

Whether future economic growth is likely to be at a similar rate to that of the previous 30 years can be debated.  My instinct would be to calculate the central estimate of value on a zero-growth basis, and to consider the effect of positive rates within the sensitivity analysis.  If, however, a positive growth rate is assumed, then for consistency it needs to be considered that, since expenditure on tourism is discretionary, demand for tourism is likely to be highly income-elastic.  Subject to the condition of the Reef,  and to the adequacy of infrastructure to accommodate visitors, demand for visits might be expected to grow even faster than GDP.  The report, however, does not consider whether the number of visits to the Reef may change in future.  The implicit assumption is that annual visits and annual consumer surpluses will remain as estimated (9).

Notice that the combination of a very low rate of time preference and a zero rate of economic growth will yield, via the Ramsey formula, a very low discount rate.  In conjunction with a long time horizon this could imply an asset value many times higher than the report’s A$56 billion.

Notes and References

  1. Deloitte Access Economics (2017) At what price? The economic, social and icon value of the Great Barrier Reef   All page references are to this report.
  2. A$ = Australian dollars. A$56 billion is equivalent to GB£33 billion or US$43 billion.
  3. Bruckner A (2002) Life-Saving Products from Coral Reefs Issues in Science and Technology XVIII(3)
  4. See for example The Guardian (6/4/2010) The Great Barrier Reef scandal
  5. King D M. & Mazzotta M Ecosystem Valuation – Options for Applying the Travel Cost Method
  6. Re omission of substitute site variables see Caulkins P, Bishop R & Bouwes N (1985) Omitted Cross-Price Variable Biases in the Linear Travel Cost Model: Correcting Common Misperceptions Land Economics 61(2) pp 182-7
  7. Two Australian studies which applied positive values to travel time are: a) Lansdell N & Gangadharan (2003) Comparing Travel Cost Models and the Precision of their Consumer Surplus Estimates: Albert Park and Maroondah Reservoir Australian Economic Papers 42(4) p 403  b) Whitten S & Bennett J (2001) A travel cost study of duck shooting in the Upper South East of South Australia  Private and Social Values of Wetlands Research Reports No. 7
  8. This point is briefly noted in Bateman I (1993) Valuation of the environment, methods and techniques: revealed preference methods, in Turner R (ed) Sustainable Environmental Economics and Management: Principles and Practice Belhaven Press, London  pp 218-9
  9. This is shown by the fact that a stream of consumer surpluses of A$662 x 2.3 million = A$1.523 billion annually for 33 years, discounted at 3.7% per annum, approximates to the A$29 billion stated on p 39.
Posted in Environment (general) | Tagged , , , , , , , , | Leave a comment

Energy and Environment in China

Arthur Kroeber’s book on China’s economy includes an excellent section on energy but a rather selective account of its environmental issues.

Arthur Kroeber’s book China’s Economy in Oxford University Press’s What Everyone Needs to Know series (1) deserves a wide readership.  Admittedly it’s rather dry: those who like their reading on serious and important topics to be spiced with anecdotes or cultural references had better look elsewhere.  But for the general reader (in English) who wants to understand China’s development and possible future – to go beyond journalistic impressionism and simplistic or politically-motivated judgment – I doubt whether there is anything better.  This is a properly researched book at a well-chosen level, a serious piece of description and analysis which avoids over-technicality.  The chapter on ‘Changing the Growth Model’, for example, makes extensive use of key ratios such as the capital-output ratio while avoiding the complexities of, say, total factor productivity.  The tables and charts are helpful and not overdone. Also helpful is the division of each chapter into sections headed by questions.  Sensitive issues such as corruption and possible exchange rate manipulation are treated in a fair-minded and temperate manner.

A full review would be beyond the scope of this blog, but I offer here some comments on Chapter 8 entitled Energy and the Environment.

Starting with energy, I commend the book for using consistent, well-defined and sensible units (p 150).  Writings on energy often confuse matters by switching between different units, using vague units (“enough to power a million homes”) or, worst of all, failing to distinguish between units of energy and units of power (2).  By contrast, Kroeber sets out very clearly the main facts of China’s energy use. The total in 2014 was 22 billion barrels of oil equivalent, as compared with 17 for the US, 12 for the European Union and 95 for the whole world. Of China’s 22, 14 are from coal (which is about half of world coal consumption) and another 5 from other fossil fuels, underlining the importance of China in worldwide efforts to address climate change by limiting greenhouse gas emissions.

Kroeber also presents two important per unit measures of energy use.  As might be expected given China’s huge population, its per capita energy use is not especially high: much less than the US, less than the EU, and only slightly above the world average. More surprisingly, perhaps, China’s energy use per unit of output (GDP), also termed energy intensity, is more than twice that of the US and the EU, and almost twice the world average.  This is despite considerable improvements in energy intensity already achieved, eg a 19% improvement during 2005-2010 after the government had set energy efficiency targets for large firms in heavy industry (p 160).

China’s high energy intensity calls for explanation, and Kroeber identifies three causes (pp 150-2 & 161).  One is the structure of its economy, with industry accounting for a high proportion of output, agriculture smaller proportionally than in many poorer countries, and services as yet smaller proportionally than in most developed countries.  Because industry is more energy-intensive than agriculture or services, and because a high proportion of China’s industry consists of especially energy-intensive heavy industry such as steel and cement manufacturing supporting its housing and infrastructure boom, its overall energy-intensity is high.  A second cause is China’s unusually high reliance on coal, which is a less efficient energy source than natural gas for generating electricity.  This is a consequence of the geographical accident that it has large reserves of coal but much less oil and gas.  The third cause relates to the efficiency with which China uses its energy sources. Here the picture is mixed.  Many of China’s coal-fired power stations have been built relatively recently to modern standards, and are somewhat more efficient than older power stations in the US.  Its fuel efficiency standards for vehicles compare reasonably well with those in developed countries. However, the energy efficiency of homes and offices is often poor, and many old, unprofitable and energy-intensive industrial plants have been kept open by local governments seeking to maintain employment and tax revenues.  Energy prices, though not especially low by international standards, are subject to controls which can reduce incentives to make energy-saving investments.

Kroeber does not attempt to quantify the overall effect of these causes, but it does seem plausible that together they go a long way towards explaining China’s high energy intensity.  A point he might have added is that the annual temperature range in much of China is such that homes need both heating in winter and air-conditioning in summer.

Like many countries, China has sought to diversify its energy sources in order to reduce its reliance on coal which is both a major source of local air pollution and a major contributor to global greenhouse gas emissions (pp 152-4).  It has become a major oil importer (although the IEA’s statistics do not seem to support Kroeber’s claim that in 2013 it became the world’s largest (3)).  It also imports natural gas.  Over the last decade, China has more than doubled its output of nuclear power and hydropower, and increased its output of electricity from renewables almost twentyfold.  However, the effect of all this in reducing coal’s share of China’s energy mix has been relatively small. In absolute terms coal consumption has continued to grow (from which it may be inferred that China’s overall energy use has also been growing).  Its coal consumption may now be close to peaking, although Kroeber advises caution on this point, both because of the short-term effect of macroeconomic fluctuations on energy demand, and because of possible under-reporting of output by smaller coal mines.

Kroeber states, correctly, that China produces over 90% of the coal it uses, and that its coal imports are only a modest proportion of its total use.  It might be added that in the context of world trade in coal, China is nevertheless a major player, and was the largest importer in 2014 (4).  Because its imports are the difference between two huge numbers (its demand and its domestic production), there is considerable potential for fluctuations in its imports to have a major impact on the pattern of trade in coal.

Because of its huge energy consumption and reliance on fossil fuels, China is the world’s largest emitter of greenhouse gases, accounting in 2012 for 24% of the global total.  Kroeber considers, but seems to me to do less than justice to, the fact that some of China’s emissions relate to the production of goods for export, and arguably should be attributed to the importing countries in any international apportionment of responsibility for climate change (pp 154-5).  I cannot see why he links the issue to that of multinational companies moving their production to China, as if exports produced by Chinese companies are irrelevant in this context.  He also states that most of China’s emissions relate to heavy industries supporting domestic construction and not to export industries.  It would have been useful to have quantified or given a source for this claim, and to have noted that domestic construction includes construction of factories producing goods for export and transport links to carry such goods.

Turning to other environmental issues, the book focuses mainly on the much-publicised issue of air pollution, treating other issues only indirectly via an environmental performance index.  Understandably perhaps given the broad scope of the book, it says little about soil and water other than noting their “extreme degradation” due to industrialisation (p 155).  A fuller treatment would have considered each of the following, and efforts to address them: soil erosion (5), soil pollution (6), reduced river flows (7), depletion of groundwater (8), and water pollution (9).  These are not minor or merely local issues.  Unless effectively addressed, they have the potential to constrain China’s food production and so increase its demand for food imports with impacts on world food prices (10); and the costs of addressing them are likely to divert significant resources from elsewhere in the economy.

As in other countries, air pollution in China includes both gases – notably sulphur dioxide – and particulates of various sizes.  Over 50% is attributable to burning of coal, 15-20% to vehicle emissions, and the remainder to other sources (pp 159 & 161).  Although Kroeber seems to suggest that the problem is most serious in Beijing and other northern cities (pp 155 & 161), he does not offer a systematic description of the geographical pattern of air pollution.  If, as seems likely, the air is cleaner in much of the countryside, then that surely needs to be taken into account in any assessment of rural-urban inequality (a topic discussed by Kroeber elsewhere in the book (pp 30-5))?

China has made some progress in addressing air pollution in that emissions of sulphur dioxide have been reduced, although concentrations of small particulates have continued to rise (p 161).  What progress there has been seems to have been achieved largely via improvements in energy efficiency and some diversification away from coal as described above.

Kroeber rightly rejects the idea that China’s environmental problems are “uniquely attributable” to its growth model or political system, pointing out that Japan, the UK and the US all experienced severe air pollution in the mid-twentieth century (p 156).  He argues however that its problems are particularly severe for a country at its stage of development.  As evidence for this he presents a version of the Environmental Kuznets Curve (a formulation of the tendency for countries to give a higher priority to environmental issues as they become richer), plotting scores on Yale University’s Environmental Performance Index (EPI) (11) against gross national income for 30 of the world’s most important countries (p 157).  This shows a fairly clear relation between EPI and income, albeit with, as is to be expected, some spread of points about the line of best fit.  China’s EPI  score is some 14% less than might be predicted from its income level.

Kroeber suggests that this can be explained in terms of China’s political system, its ‘East Asian’ approach to development, with an unusually high premium on maximising economic growth, and its aspiration to be a superpower (pp 157-8).  This seems questionable.  A possible alternative explanation starts from the fact that environmental improvement is usually a gradual process. This is for various reasons: some pollutants have a finite life over which they gradually degrade; fish stocks take time to recover from a pollution incident; newly planted trees take many years to mature; and so on.  When a country initiates the sort of environmental improvements typical of its income level, therefore, it is likely to take some years for the full benefit to be realised.  If the country’s economy has grown rapidly, as China’s has, then this time lag may result in a lower EPI score than that of another country which has a similar income level but has reached that level more gradually.  If for example countries A and B have similar income levels but A’s economy has grown annually at 8%  and B’s at 1%, then an average time lag of about 2 years would be sufficient to give A a score 14% below B.

Looking to the future, addressing air pollution is now a stated priority of the Chinese government (p 159).  The main policy instruments likely to be used are stricter environment laws and stricter enforcement.  Other approaches used in western countries, such as emissions trading schemes and class-action lawsuits against polluting companies, Kroeber suggests, are unlikely to be successful in the Chinese context (p 158).  On the other hand, the fact that a high proportion of emissions are from a small number of heavy industries may make the problem easier to address, especially, it might be added, as some of the companies in those industries are state-owned (p 100).  At any rate, Kroeber is optimistic that the next few years will see accelerated progress against air pollution.

Notes and References

  1. Kroeber, A R (2016) China’s Economy: What Everyone Needs to Know Oxford University Press.  Page references in the text are to this book.
  2. Difference Between Energy and Power
  3. International Energy Agency Key World Energy Statistics 2015  p 11
  4. International Energy Agency, as 3 above, p 15
  5. Xinhuanet (15/3/2017) Central China Province to Spend 2 Billion Yuan on Erosion Control
  6. Xinhuanet (18/1/2017) China Sets Up Lifelong Accountability System to Control Soil Pollution
  7. Earth Observatory Yellow River Delta
  8. Qiu J (13/7/2010) China Faces Up to Groundwater Crisis Nature News
  9. Xinhuanet (22/4/2014) China’s Underground Water Quality Worsens: Report
  10. OECD-FAO (2013) Agricultural Outlook 2013-2022 Chapter 2 Feeding China: Prospects and Challenges in the Next Decade  See especially Risks and Uncertainties pp 83-7
  11. Yale University Environmental Performance Index
Posted in Energy, Environment (general) | Tagged , , , , , , , , | Leave a comment