Check Your Spillover

The  ​ American economist William Nordhaus opened his Nobel Prize speech in 2018 with a slide of the painting El Coloso, traditionally attributed to Goya and completed sometime between 1808 and 1812. Like Goya’s better-known images of the Madrid uprising of 2 May 1808 and the bloody retribution that came after, the painting depicts the calamitous violence of the Peninsular War, which followed Napoleon’s invasion of Spain. But while Goya’s intentions are clear in El Dos and El Tres de Mayo, it is much less obvious what is going on in El Coloso. A giant man, shrouded in mist, looms over the hills. In the dark valley below, people and animals are caught in desperate flight. Only a single donkey remains still, unflustered. The giant is half-turned away from us and from the refugees, his fist raised, ready to fight. But his eyes appear to be closed. Who or what presents the threat isn’t visible to us.

Is the colossus protecting the people, or is it him they fear? Does he symbolise the French armies wreaking havoc across the Spanish hills or Spain standing up to the invaders? If the giant is Spain, which Spain is he: the Bourbons demanding restoration or the liberal republicanism that flared briefly at the time Goya was painting? Nordhaus, whose speech was about the contribution he has made to the economics of climate change, didn’t engage with any of this. For him, El Coloso was simply a metaphor for global warming, which ‘menaces our planet and looms over our future’. It stands outside and above humanity, ‘the colossus of all environmental externalities’, before which any single actor – person, nation, region – is powerless.

Nordhaus doesn’t claim to be an art historian, so perhaps he can be forgiven for not recognising that in the world of the colossus, there is no such thing as an ‘externality’, the term economists use to describe an unintended non-market outcome of market activity (Nordhaus often uses the word ‘spillover’ instead). Economic externalities can be negative or positive: a new initiative might generate pollution, for instance, but also knowledge that spills out into the wider world. The key feature of the externality, though, is that it is unpriced. The social costs (or benefits) of market activity are not accounted for in the production process, and the producer does not pay (or receive) prices that accurately reflect the costs or benefits of their activities. If you poison the groundwater fertilising your crops, but don’t pay the costs created for everyone else, then losses due to poisoned groundwater are not reflected in the price of the farmer’s produce. This misaligns market incentives, over-rewarding destructive behaviour.

The concept of externality is crucial to the often awkward attempts of economics to analyse human relationships with the non-human world. Pollution, biodiversity loss, soil degradation and ocean acidification are all defined as externalities – negative environmental outcomes of quite rational market behaviour. The solution, we are told, is straightforward: internalise the cost of the externality. If we get the prices right – by including the costs to society of poisoned groundwater, for example – the market will send the right signals, the incentives will align themselves and the externality will disappear.

This is the idea behind carbon pricing as a way to mitigate climate change. Regulators can use either taxes or markets in scarce emissions permits (‘cap and trade’) to create conditions in which the cost of emitting greenhouse gases reflects the actual costs it imposes on society. According to this way of thinking, the problem lies not with the market, or capitalism, or growth, and doesn’t necessitate radical change; it’s just a hiccup in the price mechanism. This, in a nutshell, is Nordhaus’s argument in The Spirit of Green. He also provides brief, superficial and cherry-picked overviews of ‘green’ ethics, political theory and politics, but the book engages with only a tiny sample of the ideas in these fields. It refers almost exclusively to the work of Nordhaus’s white male Ivy League colleagues and spends nearly as much time on plans to settle other planets as it does on political theory and politics combined. Ultimately, his point is simply that if we fix the price distortions caused by externalities, we can be as David to Goliath without really changing anything. Probably.

Credit where it’s due: Nordhaus was one of the first among his peers to name the climate colossus. It has taken more than forty years for the rest of the business community, and the elite economists and financial media that defend its interests, to do as much. Now even the global economy’s core institutions fret openly, and some of them seem more worried than Nordhaus. Until recently the International Monetary Fund had stuck closely to his non-alarmist rules of thumb: markets are the best and most efficient means of addressing climate change, and carbon pricing is the key (but it should be introduced and increased gradually, to minimise shocks to the economy). In September 2019, however, the IMF published a working paper that acknowledged the ‘growing agreement between economists and scientists that the tail risks are material and the risk of catastrophic and irreversible disaster is rising, implying potentially infinite costs of unmitigated climate change, including, in the extreme, human extinction’.

According to Nordhaus, climate action will always involve trade-offs and we need to assess the anticipated costs and benefits rationally. His own assessment is that in light of those trade-offs we should settle for an average global temperature increase of 3.5°C above pre-industrial levels (in contrast to the Paris Agreement’s commitment to a limit of 1.5°C). But the IMF’s recent pronouncements reflect a more general change of pitch: cost-benefit analyses hardly seem appropriate when we are considering the possibility of human extinction. The less unlikely we consider catastrophic outcomes to be, the harder it is to argue that trade-offs are reasonable: any cost now is preferable to infinite costs later, let alone extinction.

Economists ‘set themselves too easy, too useless a task’, Keynes said, ‘if in tempestuous seasons they can only tell us that when the storm is long past the ocean is flat again’. But in the face of climate change, the long run – which remains the sacred temporality of economics – is a misleading guide not only to current affairs, but to the long run itself. There is no reason to expect the ocean to flatten again; we may be in for a permanent storm. The idea of eventual equilibrium belittles present crises by figuring them as mere bumps in the long arc of history, but even the promise that we might one day reach equilibrium now appears hollow.

Nordhaus has spent his career building a climate economics for the rose-tinted long run, developing and refining his version of the workhorse of climate economics, the ‘integrated assessment model’. The point of an IAM is to combine a climate model with a growth model – connected by a ‘damages function’ – to determine the ‘social cost of carbon’. The SCC is a measure of the discounted cost of future damages (we’ll come back to this) associated with carbon emissions in the present, meaning that, in theory at least, it is the optimal carbon tax: if we can force emitters to pay the cost of future damages now, we can tame Goliath.

In the eyes of Nordhaus’s many critics, part of the problem with climate economics is the disproportionate influence he and his IAM have exerted on subsequent efforts to design models and solutions for climate change. IAMs are the basis for pretty much all carbon abatement and taxation policy, yet they confine their economic analysis to a narrow range of futures in which the world remains the same while global temperatures slowly ratchet up. Many of their assumptions – consistent long-term growth rates, for example – are not guaranteed, or even that credible. The Dynamic Integrated Climate-Economy model, which Nordhaus and his team have been working on since the early 1990s, has become the benchmark IAM. Most conversations about climate-economy modelling begin with DICE, and anything that deviates from it must be justified. The 2006 Stern Report, for example, which uses an alternative IAM, expends a good deal of energy justifying its non-Nordhausian choices.

We know that Nordhaus’s cost-benefit models, for reasons technical and political, massively underestimate the scale and scope of the economic impacts of climate change. Because their conception of the future is a gradually warming version of today, with little impact on long-run growth, they conclude that global warming is best treated as a change in the cost structure associated with shifting temperatures: capitalism plus wind turbines. Even the loss of the Greenland icesheet is reluctantly monetised as just one more factor in the social cost of carbon for a set of ‘normal’ political-economic institutions that will somehow persist, no matter how high the seas or how parched the soils. This thinking is intrinsic to much mainstream climate economics. Whatever Nordhaus’s intentions, DICE has functioned as a technical expression of climate policy in the age of ‘new denialism’: we go through the motions of confronting the problem and by doing so justify the deferral of systemic overhaul.

One​ of the most significant differences between Nordhaus’s models and those, like the one used in the Stern Report, which point to the need for much higher carbon prices, is their respective discount rates. The discount rate is perhaps the most controversial term in climate economics. It is used to track the value we assign to the future relative to the present. A zero discount rate means the future is given the same weight as the present in our current choices; the higher the discount rate, the less weight is placed on the future, or on the future impact of current activity, in our decision-making.

At first glance, there’s something troubling about discounting. On what grounds can future states of the world be considered less important, less valuable, than the present? But in fact discounting is a crucial element in all investment decisions. One way to think about the discount rate is as the alter ego of the interest rate. If the discount rate is 4 per cent, then any investment that isn’t expected to provide an average annual return of 4 per cent leads to a ‘future’ which, when discounted, will return less than the current valuation of the ‘present’. In other words, at least in theory, the discount rate should equal the expected return on investment.

For the purposes of climate economics, the discount rate is the value, in the present, of the expected benefits and costs of projects to mitigate the possible effects of global climate change. In public policy, investment decisions cannot be driven by returns. Public programmes and spending decisions are (we hope) not made by autonomous agents seeking profits, but by or for people with a variety of attitudes towards the future, differing degrees of uncertainty and unequal access to income, wealth and political power. Because dealing with climate change requires not only expensive investments in the future (renewable energy infrastructure, public transport etc), but also massive disinvestment in existing high-return activities (particularly fossil fuel extraction), discounting is central to the decisions that must be taken. But the degree of discounting is inevitably controversial. The farmer in her drought-stricken fields weighs the future impacts of fossil fuel use very differently from the Exxon shareholder, and their grandchildren are likely to experience the impacts of today’s choices very differently too.

One of the most common ways to incorporate discounting in climate-economy models is to use the so-called Ramsey equation, named for the mathematician Frank Ramsey.* Ramsey formulated the equation as a response to the question: ‘How much of its income should a nation save?’ In order to answer this, we need to think about the trade-off between the benefits of consumption today and those of saving and consuming later. The equation itself isn’t intimidating, but it contains important subtleties. It looks like this:

r = δ + ηg

It tells us that r, the annual discount rate used to weigh future benefits against present costs, is determined not only by the degree to which we prioritise the present, but also by the relative impacts our actions will have on a future that will be different from today. The world will change, so the rate must register the effects of present-day activity not as if they were being experienced now, but in the future.

In other words, what we might elsewhere think of as ‘the’ discount rate – represented by δ, the ‘pure rate of time preference’ – is only one term in the Ramsey equation. This is the easy part: if we value having something now more than we do waiting for it, our rate of time preference is positive: ‘impatience means δ > 0’ as the economist Partha Dasgupta puts it. The second term, ηg, is more complicated: η represents what economists call the ‘consumption elasticity of marginal utility’ and g is the annual rate of economic growth. The idea here is that the costs or benefits of our actions today won’t necessarily be equal to the costs or benefits they bestow in the future, and that the further out we go, the greater the difference will be. Perhaps if we limit economic growth to reduce emissions today, we will cause ourselves such pain and suffering that it will outweigh the gains our sacrifices make possible in the future. Maybe in the future our descendants will have carbon capture technologies and we needn’t have made any sacrifices after all. Maybe properly priced externalities will have generated much ‘greener’ economies, and our bequests won’t contribute to future wellbeing nearly as much as they cost us today.

If that were the case, then even if we were infinitely patient (δ = 0), there might still be reasons to discount the future impacts of present activities. Say that we think our economies will continue to grow over the coming decades (most models set aside the question of how the new wealth will be distributed). It follows that we can reasonably expect our grandchildren to be better off than we are, and that therefore adding to their baskets won’t contribute as much to their welfare as it does to ours today. If, however, we expect everyone to be impoverished in fifty years’ time, we should by the same logic take action now to provide for them. A great deal rides on expected growth and on who might benefit from it.

While discounting is standard practice, not all climate-economy models use it for the same purposes. Some use a normative standard, meaning that δ and η are determined on an ‘ethical’ basis. Others use descriptive values based on real-world proxies. Insofar as it is also clearly an ‘ethical’ decision to adopt a descriptive stance towards planetary crisis, the line between the two is blurry. But the basic distinction replicates the perennial opposition of Sollen and Sein, ought and is: some models specify values the modeller feels society should choose, others the ones society actually does choose. Stern, for example, chooses a ‘prescriptive’ near-zero value for δ (0.1), meaning that future welfare should be considered roughly equivalent to that of the present.

On straightforwardly humanist grounds, this makes sense. As Ramsey put it almost a century ago, to set δ > 0 ‘is ethically indefensible and arises merely from the weakness of the imagination’. Even a cursory look at the world around us, however, suggests that the reality is very far from δ = 0. Nordhaus insists on distinguishing between the ‘welfare discount rate’ (the wellbeing of future generations) and the ‘goods discount rate’ (the return on capital investment). Since the former is unobservable, he says, the latter is all we have to go on – it is ‘observed in markets’. This is Nordhaus’s principal defence of the higher discount rate in DICE: it is calibrated ‘realistically’ to reflect the way real-life actors behave. Societies will compare the returns from reducing greenhouse gas emissions, via tax increases and infrastructure investment for example, the same way they compare the returns on any long-term investment.

This means that the discount rate must match the real return on capital. If, say, the return on thirty-year bonds is 5 per cent, then we should assume that the expected ‘return’ on efforts to mitigate climate-oriented damages thirty years from now must compete with that rate. If they don’t, it won’t happen. It’s as simple as that. It is ‘unrealistic’ to expect society’s assessment of the social cost of carbon to reflect a discount rate any lower than expected rates of return. Nordhaus doesn’t quite argue that this is how it should be, only that, in a market-based society, this is how it is. And if this is how it is, then the only way to stop investments that further destabilise the climate is to reduce returns with taxes or carbon markets.

The numbers used in the Ramsey equation can seem quite small. Descriptive estimates vary, but it’s not uncommon to set δ, for example, at 2 per cent, and η at around 1.5. If we also expect annual growth to be about 2 per cent, this produces a discount rate r of 5 per cent. Because discount rates compound annually, the effect of seemingly ‘low’ rates on modelled outcomes remains substantial, even over relatively short periods. For example, the 1.4 per cent average rate adopted in the Stern Report means that when assessing current investments, the welfare of people living fifty years from now should be valued at just under half that of people alive today. Nordhaus is considerably less cautious in his assessment, positing a discount rate averaging 4.25 per cent over the period to 2100, which weights the welfare of people fifty years from now at less than one-eighth of ours. That is terrifying. Yet it must be admitted that as a characterisation of current collective behaviour it is a lot more plausible than a normative rate like Stern’s. One might even say that in light of the scientific consensus on the scale and pace of action required, Nordhaus’s defence of baby-step gradualist climate policy inadvertently shows why a market-driven approach is destined to fail.

As calculated using the Ramsey formula, the discount rate is clearly an ethical, rather than ‘scientific’, element of the IAM. It also has an extraordinary influence on the model’s output. This means the technicalities, like the difference between pure rate of time preference (δ) and consumption elasticity of marginal utility (η), matter a great deal. In the calculation of the discount rate, the more the former declines, the more important the latter becomes. If we have any concerns about intergenerational equity in consumption, then as δ falls, larger elasticities are the only way to signal it in the discount rate, and the effect of these concerns will be multiplied by growth or its opposite. In Dasgupta’s words, if δ should be 0, as Ramsey tells us, then ‘the whole weight of the ethics regarding the distribution of consumption’ falls on the consumption elasticity of marginal utility. ‘That’s an awful lot of work for a single number to do adequately.’

Here is the key to the problem with Nordhaus’s ‘rational’ cost-benefit approach to the future of life on Earth. It isn’t a single number that does all the work, but we treat it as if it did because the remaining factor in the Ramsey equation, g, the rate of growth, is taken as a given. While our rates of time preference and judgments of the value of future consumption seem like ‘obviously’ ethical parameters, the rate of growth is treated as a technical input to which the normative-descriptive distinction doesn’t apply. That’s hogwash. It may be an empirical variable, describing a state of the world, but long-run growth is not determined empirically in the scientific sense: it represents modellers’ expectations regarding the future path of the economy. These expectations may be based partly on facts derived from past experience, but in the end g is an uncertain forecast, subject to misspecification, error, ignorance and surprise.

In an article from 2018 in the Proceedings of the National Academy of Sciences, Nordhaus analysed the uncertainty associated with modelling global growth decades in advance. He and his co-authors forecast growth between 2010 and 2100 using two different methods, a statistical approach using historical data to project forward, and a survey of the views of sixteen ‘experts’. (Is that a lot? Too few? Unclear.) Each expert was nominated by a panel of peers based on their ‘contributions to the economic growth literature, familiarity with empirical research on medium-run and long-run growth, and diversity in regional expertise’. All of them are male economists working at prominent institutions and almost all of them are white. All but two are based in the US, only one outside the Anglo-American world and only two at public universities. All of them work in the mainstream of the discipline; not one of them is a climate scientist, or any other kind of scientist for that matter.

Both methods produced an estimated median global annual growth rate of approximately 2 per cent between 2010 and 2100, which translates to an economy almost 500 per cent larger by the end of this century. (That the results were so similar isn’t surprising, given that the experts in the survey based their estimates on the same data from 1900-2010 that were used for the statistical exercise.) But as Nordhaus makes clear in the paper, if not in his book, projections of growth or any other future economic state beyond the short run are radically uncertain, and arguably more so today than they have been for centuries. Climate change means that when we try to look twenty or fifty years ahead, we have very little capacity to anticipate conditions even in what are, at least for now, the most affluent parts of the world, let alone in poorer regions already suffering climate-induced stress. Many have argued that the IAMs which rely on these forecasts are functionally incapable of taking catastrophic outcomes into account (a charge Nordhaus denies). But either way, a forecast of 2 per cent annual global consumption growth over the next decades – standard in DICE models – isn’t optimistic or pessimistic so much as baseless.

The most sophisticated​ climate science models describe a range of potential environmental conditions, shaped by a wide array of interconnected but poorly understood threshold effects and feedback loops – and that is just the non-human systems. Such models are rafts of mathematical complexity floating on seas of uncertainty, even about measures as fundamental as the relationship between emissions and global temperature change. State of the art climate-economy IAMs use these outputs as inputs and model economic impacts on the basis of the ‘damages function’, the predicted degree to which growth is affected by temperature change and attendant biogeophysical shifts (the rise in sea levels and so on). This is not a very precise exercise: damage function parameters are determined no more ‘objectively’ than growth forecasts. In the days when it seemed to many that climate change would mean gradual planetary warming, the idea of damages function guesstimates might have made some sense. But the truth of what scientists have been saying for a long time – that some of the most devastating effects of climate change will be more like flicking switches than turning dials and that the impact will continue to be extremely volatile and geographically uneven – is now widely recognised. Anyone who claims they can forecast rates of growth a hundred years from now, on a planet several degrees hotter than it is today, using temperature changes averaged across the entire planet and measures of climate sensitivity based on what we think we currently know, is fooling themselves.

Nordhaus attempts to make climate change compatible with ceaseless long-run growth by emphasising the global economy’s ‘carbon intensity’ instead of its carbon sensitivity. The Spirit of Green is most sanguine in its demonstration that a decreasing amount of CO2 is required to fuel a unit of global growth. I think his point is that since growth requires less carbon than it did in the past, we should be more hopeful. Maybe I am missing something, but it seems to me that continued emissions are continued emissions. Long-term studies of emissions pathways show that the problem, in the end, is the absolute volume of greenhouse gases in the atmosphere. We simply have to stop emitting them as soon as possible. There is no hope to be placed in the gradually declining carbon intensity of growth that nonetheless continues to add to the atmospheric pool. In fact, if growth accelerates, decreasing carbon intensity is quite compatible with increasing emissions. The problem, in the models and in our current reality, is the arrogant fidelity to growth.

In the discounting process, the politics of the so-called ethical parameters, δ and η, are explicit: both force us to make future welfare part of our present decision-making. Growth, however, is a hedge we can hide behind. The ethical problem here lies not in the calculation of incalculable uncertainties, but in the defence of an ‘expertise’ that can somehow navigate these uncertainties for us. The argument for the defence is that IAMs are better than nothing: like other economic models of complex adaptive systems, they are said to be helpful ‘tools’, useful ‘heuristics’ that allow us to examine fundamental dynamics. This is tantamount to saying that modelling badly or irresponsibly is better than not modelling at all. I’m not so sure.

Even if we accept this argument, it is implausible in the case of so fundamental a social variable as the discount rate. Nordhaus insists that ‘our primary ethical obligation is … to promote laws that correct the spillovers,’ which makes it sound as though correcting the spillovers were a technical problem. The result, according to DICE, is an estimated social cost of carbon per ton equivalent to the price of a moderately spendy bottle of pinot noir (currently $45 or so). Even if this SCC were put to work as a carbon tax – which, as Nordhaus repeatedly points out, is nowhere the case – it isn’t nearly high enough to incentivise behaviour that would meaningfully slow planetary warming (such as, for example, eliminating fossil fuel extraction). If our only hope right now is to internalise the colossal externality, that strategy was too little even before it was too late.

Listen to Geoff Mann discuss this piece with James Butler on the LRB Podcast.