Climate Change, Climate Policy, and Economic Growth
2020; University of Chicago Press; Volume: 34; Linguagem: Inglês
10.1086/707193
ISSN1537-2642
Autores Tópico(s)Market Dynamics and Volatility
ResumoPrevious article FreeClimate Change, Climate Policy, and Economic GrowthJames H. StockJames H. StockHarvard University and NBER Search for more articles by this author Harvard University and NBERPDFPDF PLUSFull Text Add to favoritesDownload CitationTrack CitationsPermissionsReprints Share onFacebookTwitterLinked InRedditEmailQR Code SectionsMoreThe topics of climate change and climate change policy encompass a complex mixture of the natural sciences, economics, and a mass of institutional, legal, and technical details. This complexity and multidisciplinary nature make it difficult for thoughtful citizens to reach their own conclusions on the topic and for potentially interested economists to know where to start.This essay aims to provide a point of entry for macroeconomists interested in climate change and climate change policy but with no special knowledge of the field. I therefore start at the beginning, with some basic background on climate change, presented through the eyes of an econometrician. I then turn to climate policy in the United States. That discussion points to a large number of researchable open questions that macroeconomists are particularly well suited to tackle.1Let me summarize my four main points. First, although a healthy dose of skepticism is always in order (as academics it is in our DNA), simple and transparent time series regression models familiar to macroeconomists provide independent verification of some key conclusions from climate science models and in particular confirm that essentially all the warming over the past 140 years is because of human activity, that is, is anthropogenic. Figure 1 shows time series data on annual global mean temperature since 1860, when reliable instrumental records start. As seen in the figure, the global mean temperature has increased by approximately 1 degree Celsius, compared with its 1870–90 average value. This increase in temperatures drives a wide range of changes in climate, including droughts, more hot days, and more intense rainfalls and storms, all of which vary regionally. Because climate science uses large, opaque calibrated models of the climate system, there is room for confusion among legitimately skeptical outsiders about just how much of the global warming observed since the industrial revolution results from human activity, that is, is anthropogenic. Standard time series regressions provide a simple, transparent, and (I argue) reliable alternative, at least for modeling the relation between emissions and temperature. According to a regression decomposition I present later, anthropogenic sources account for essentially all of the warming in figure 1. The main driver of that warming is anthropogenic emissions of carbon dioxide (CO2) from burning fossil fuels. The simple regression on which these estimates are based lacks nuance but the results accord with and, therefore, provide support for the more complex models used by climate scientists.Fig. 1. Global mean temperature deviated from its 1870–90 mean (Hadley Earth Observatory, HadCRUT4 series at https://crudata.uea.ac.uk/cru/data/temperature).View Large ImageDownload PowerPointSecond, policy will play a crucial role in decarbonizing the economy. As shown in figure 2, in the United States, energy-related CO2 emissions peaked in 2007 and then fell 12% by 2018. This fact has led some on the environmental left to argue that we have turned a corner and are on an inevitable path to decarbonization and some on the right to argue that the free market will lead to decarbonization so policy interventions are costly and superfluous. But this narrative, however appealing, is false. Instead, the decline in emissions since 2007 is mainly the consequence of the financial crisis recession and the fracking revolution, which made natural gas cheap enough that it has partially replaced a higher-carbon fossil fuel, coal, for generating electricity. In contrast to the rosy narrative, the most recent projections by the US Energy Information Administration (EIA) indicate that, under current policy, the United States will not be close to hitting its pledged 2025 emissions-reductions target under the now abandoned Paris climate accord.Fig. 2. US CO2 emissions from energy consumption, 1973–2018, with US Energy Information Administration projections (dashed), 2019–50 (US Energy Information Administration, Monthly Energy Review [June 2019] and 2019 Annual Energy Outlook, reference case).View Large ImageDownload PowerPointThird, looking beyond the short-term Paris target, the multitude of climate policies currently in place in the United States, from federal to state to local, fall far short of what is needed to achieve decarbonization on a timescale consistent with avoiding very severe damages from climate change. With some exceptions, existing policies interact in complex ways that lead to inefficiencies, are subject to industry capture, tend to be expensive as measured by cost per ton of CO2 avoided, and are small bore in the sense that their scope for emissions reductions is small. The large-scale, more efficient policies typically favored by economists, such as a carbon tax or its cousin, cap and trade, have dim prospects because they either have already been rejected politically (e.g., cap and trade), create significant political liabilities (e.g., a carbon tax), or have been weakened or reversed through the regulatory process (e.g., the Clean Power Plan [CPP], the Obama administration's plan for a cap-and-trade system within the power sector). Moreover, the absence of a price on carbon is but one of the externalities plaguing climate policy, and carbon pricing alone at politically plausible levels is unlikely to be particularly effective in reducing emissions from the oil and gas used in the transportation, commercial, and residential sectors.Fourth, the political constraints on and intrinsic limitations of Pigouvian carbon pricing mean that economists need to look elsewhere for efficient climate policies. I believe that the most important place that economists can add value to the climate policy discussion now is by focusing on policies that drive low-carbon technical innovation. This view is informed by positive political economy—what politicians seem willing to do, by empirical evidence and some key success stories about technology-pushing policies, and by a small but insightful literature on carbon prices, research and development (R&D) subsidies, and induced technical change. Ultimately, decarbonization will occur not by forcing consumers and businesses to choose expensive low-carbon technologies over inexpensive fossil fuels but by ensuring that those green alternatives are sufficiently low cost that they are largely chosen voluntarily. Consumers and firms will need to choose low-carbon energy not because it is the right thing to do, but because it is the economical thing to do, even if there is not a meaningful price on carbon. The transition to a low-carbon economy will require a low-cost alternative to fossil fuels. The key policy question is, How can we most efficiently promote the development of advanced low-carbon technologies? This difficult question is one that economists are well equipped to tackle.I. Some Climate Change EconometricsThe increase in global mean temperature in figure 1 happened in stages, initially rising starting around World War I, followed by a plateau in the 1950s through 1970s, then taking off in earnest around 1980. A natural question is, How much of this increase is anthropogenic? An oft-cited response is that 97% of climate scientists agree that global warming is mainly because of human activities (Cook et al. 2013). As part of the scientific community, we should trust in the peer review process and thus in the science underlying that consensus. That said, the models on which those conclusions are based are large, complex, and difficult for outsiders to evaluate. This complexity has opened the door to debate about the scientific consensus, which in turn raises the question of whether there are ways to estimate the extent to which this warming is anthropogenic that are simpler, transparent, and stay close to the data. Fortunately, the tools of time series econometrics provide such estimates.The starting point is the principle that Earth's temperature is proportional to the thermal energy flux hitting its surface. This includes energy from the sun and energy radiated from Earth that is absorbed by atmospheric gasses and reradiated back to Earth. This latter source is the greenhouse effect. These energy fluxes, called radiative forcings, are shown in figure 3: CO2, methane, trace gasses like hydrofluorocarbons, solar radiative forcing (the wiggles are sunspot cycles), and sulfur oxides, which have negative radiative forcings because they reflect sunlight back into space. All the gasses have natural components, but the changes in these radiative forcings over this period are almost entirely anthropogenic (sulfur oxides are also emitted from volcanic eruptions in addition to burning high-sulfur fossil fuels; however, their presence in the atmosphere is transitory). The dashed line is the sum of these radiative forcings.Fig. 3. Radiative forcings (see Montamat and Stock 2019 for original data sources).View Large ImageDownload PowerPointA very simple model of Earth's temperature is that it is proportional to the sum of the radiative forcings. With the additional assumption that total radiative forcings are an integrated process, this simple model implies that global mean temperature and radiative forcing are cointegrated (Kaufmann, Kauppi, and Stock 2006; Kaufmann et al. 2013); that is, there is a cointegrating relationship of the form Tt=α+θRFt+ut, where RFt is the sum of the radiative forcings in figure 3 and ut is integrated of a lower order than RFt and θ is the cointegrating coefficient.2Figure 4 overlays the global temperature series in figure 1 with the predicted value of temperature, θ^RFt. The estimate of θ used in figure 4 (0.489, standard error [SE]=0.041) is the benchmark estimate from Kaufmann et al. (2006, table 2, col. 2), which was estimated using data from 1860–1994, the full data set available at the time. The in-sample fit of the dynamic ordinary least square estimate (through the vertical line in 1994) captures the overall pre-1994 trend, although there are short-run fluctuations in temperature around this trend that are not captured by this long-run relationship.Fig. 4. Temperature and fitted values based on radiative forcings. Estimation 1860–1994. Shading is 67% confidence interval conditional on radiative forcing. Predicted value uses dynamic ordinary least square cointegrating vector from Kaufmann et al. (2006, table II[2]). Temperature (dashed line) is deviated from its 1870–90 mean. The solid line is the predicted value from the benchmark cointegrating regression in Kaufmann et al. (2006) (T^t=const+0.489RFt), which they estimated using data from 1860 to 1994. The vertical line demarks the in- and out-of-sample time periods for that estimate. The shading around the predicted value post-1994 is a one SE band for the predicted value using their reported SE of θ^. The dotted line is the contribution of natural variation in solar radiation to temperature, estimated using the Kaufmann et al. (2006) regression.View Large ImageDownload PowerPointBecause this model was fit using data through 1994, there is a clean out-of-sample test of this very simple model. The test is nontrivial: temperatures increased since 1994, but irregularly, with a famous decade-long "hiatus" starting in 1998. How did this simple model do?It turns out that it did quite well. As discussed in more detail in Kaufmann et al. (2011), the model provides a parsimonious explanation of the hiatus as due in part to a lull in solar activity and to new dirty coal-fired power plants coming online in China, which produced sulfur oxides and a cooling effect.3This simple model provides a standard regression decomposition of the post-1880 warming into a natural component, an anthropogenic component, and a residual. One way to do this is to consider the counterfactual in which all the gasses simply equaled their averages in the late nineteenth century. The dotted line in figure 4 is the predicted natural component arising from variation in solar flux. Initially, nearly all the variation in the predicted value of temperature was from variation in solar radiation. But starting around 1920, greenhouse warming started to kick in. During the 1950s through the 1970s, the warming effect of CO2 and methane was largely offset by sulfur oxides emitted from coal power plants. As those emissions were cleaned up to mitigate local pollution and acid rain, CO2 took over as the main driver and warming accelerated.According to this very simple model, of the 0.81 degree Celsius of warming from the 1870–90 average through the 2006–15 average, 0.84 degree (SE=0.07) is due to greenhouse gasses, 0.01 degree (SE=0.004) is due to an increase in solar intensity, and −0.04 degree is an unexplained residual.4 Thus, according to this decomposition, essentially all of the observed warming is anthropogenic in origin, up to a residual of approximately 5%.The full decomposition based on this simple regression model is given in Table 1. As this decomposition shows, the key driver is CO2, and its impact on warming would have been greater had it not been for the additional, and unhealthy, increase in SOx pollutants produced by burning high-sulfur fossil fuels, especially high-sulfur coal.Table 1. Decomposition of the Change in Global Mean Temperature from 1870–90 Average to 2006–15 Average Change or Predicted Change (°C)Standard ErrorGreenhouse gasses: CO2.96.08 Methane.24.02 Trace gasses.13.01 SOx−.49.04Subtotal, gasses:.84.07 Solar.01.004Subtotal, predicted:.85.07 Actual.81 Residual−.04 Note. Predicted values and standard errors are based on the cointegrating regression used for the predicted values in figure 3 and described in Section I.View Table ImageThe virtue of this model is its transparency and its good performance in a 2-decade, true out-of-sample test. But the model is an extreme simplification of highly complex climate processes and is silent about the wide variation in climate change effects stemming from this temperature increase. Those effects are extensively documented in the climate science literature.5 Many are also amenable to validation using econometrics.6 To me, the numerical alignment of the estimates from this very simple model with the climate models justifies confidence in the climate science models.II. What Is the Progress to Date on Reducing Carbon Emissions?As I mentioned, a popular narrative is that the downturn in US CO2 emissions since 2007 demonstrates that we have turned a corner and are on a path toward decarbonization. According to this narrative, we are reducing emissions because of energy efficiency improvements, the expansion of wind and solar for electricity generation, and an increasing cultural awareness of the importance of conserving energy and going green. This narrative is popular among environmentalists, who say that decarbonization will be cheap; conservatives, who say that market forces are resulting in decarbonization already; and green investors, who proclaim a bright future for their low-carbon investments.I wish that this rosy narrative were true, but it is not. Macroeconomists will not find it surprising that the big drop in emissions occurred in 2009, when energy demand plummeted as the economy tanked. Since then, the fracking revolution has resulted in low natural gas prices, which has led to replacing coal generation with natural gas generation.7 Because burning coal emits more CO2 than burning natural gas per kilowatt-hour of electricity generated, switching from coal to natural gas reduces CO2 emissions.Because the 2009 recession and the advent of fracking were one-time events, they do not constitute a change in the trend, just a shift in the level of emissions. Indeed, in 2018, US energy-related CO2 emissions increased by 2.9%. The US EIA projects coal use for electricity to be roughly flat from 2020 to 2050.8 As shown in figure 2, emissions are projected to plateau at current levels, as energy efficiency improvements and renewables just offset growing energy demand. Indeed, the silver lining of the substitution of natural gas for coal resulting from fracking hides a cloud, which is the substantial investment in natural gas pipelines and generating facilities that could lock in future emissions else risk the political and economic disruption of stranded natural gas assets.This projection leads to the question: If CO2 emissions remain at their current rate, what is their short-run effect on temperature? In recent work with Giselle Montamat, we use a natural experiment instrumental variables approach to estimate the short-run temperature effect of emission without adopting any particular model of long-run persistence. We estimate that 10 years of emissions at the current rate would increase temperature over those 10 years by 0.13 degree Celsius (Montamat and Stock 2019). This might not seem to be by much, but it is more than one-eighth the total warming to date and amounts to 1 degree Fahrenheit over 3 decades. Moreover, this is just the impact effect, and the cumulative effect would be even larger as the pulse works through Earth's system.In short, climate change is anthropogenic and it is happening now on a human timescale. The planet is already experiencing temperature records and increasingly damaging hurricanes and typhoons, wildfires, droughts, and heat waves. Additionally, sea levels have been and will be rising because of thermal expansion of water and melting of glaciers and ice sheets. Under a business-as-usual scenario, the mean sea level is projected to rise by between 55 and 95 centimeters by the end of this century.9 These consequences of human emissions of greenhouse gasses are not a "new normal." Rather, they will become more severe as temperatures rise.The future consequences of climate change remain uncertain. For example, the amount by which sea levels rise depends in part on the extent to which glaciers and ice sheets melt. In climate science, events such as the melting of the West Antarctic Ice Sheet or, much worse, the melting of the Greenland Ice Sheet, are referred to as abrupt irreversible events. Those events are not expected to happen in this century, although they could be triggered irreversibly in the first half of this century. They could add multiple meters to sea level rise. Similarly, there is considerable uncertainty about the pace of extinctions that are being and will be induced by climate change. The severity of these and other aspects of climate change depends on whether cumulative emissions get high enough to trigger such transformations.10 That, in turn, depends on climate policy decisions made by our generation, arguably within the next decade or two.III. US Climate Policies: Historical Evidence on Efficiency and EffectivenessThis brings us to a discussion of climate policies, where I focus on the United States. First, however, I digress briefly on the externalities these policies aim to address and on current estimates of the value of one of these, the carbon externality.A. Digression on ExternalitiesThere are two main market failures that climate policy aims to address: the carbon price externality and the R&D externality. In some instances, network externalities are also important, such as the chicken-and-egg problem of electric vehicles and charging stations.The climate externality that has received the most attention by economists is the carbon price externality. The starting point estimate for assigning a value to this externality is the social cost of carbon (SCC), which is the monetized net present value of the damages from emitting a marginal ton of CO2. The final estimate of the SCC released under the Obama administration is approximately $50 per ton for emissions in 2020 (US Government Interagency Working Group on the Social Cost of Greenhouse Gasses 2016). (To get a sense of orders of magnitudes, a short ton of subbituminous coal from a federal mining lease in the Powder River Basin currently sells for approximately $12; when burned, it emits 1.7 metric tons of CO2, which has approximately $84 of climate damages evaluated at an SCC of $50. The climate damages from burning a gallon of gasoline are approximately $0.45, also evaluated at an SCC of $50.) There is widespread recognition that the scientific basis for this $50 estimate of the SCC needs to be solidified. To this end, Resources for the Future is coordinating a major research project involving energy-climate labs at Chicago and Berkeley, along with academics from other universities, which (among other things) is implementing suggestions made by the National Academy of Sciences (2017) for improving the estimate of the SCC. Because this work is still in progress, for this paper I use the provisional $50 per ton estimate for the SCC.I now return to the discussion of US climate policies.11 These policies fall into four categories: regulation, narrowly targeted policies, carbon pricing, and technology-pushing policies.B. Sectoral Regulation Based on the Clean Air ActThe Clean Air Act is the legal authority used for the two most ambitious regulatory attempts to date to reduce greenhouse gas emissions, the CPP that applied to the power sector and the Corporate Average Fuel Economy (CAFE) standards that applied to automobile emissions (and thus mileage). With careful attention to detail, regulations under the Clean Air Act can be efficient and effective. For example, the CPP developed by the Obama administration used Clean Air Act authority to construct a mass-based cap-and-trade system for the power sector that is broadly considered to be workable and cost-effective. Estimates are that the CPP would have achieved substantial emissions reductions with an average cost around $11 per ton CO2, which is well below the SCC benchmark.12 Initial estimates suggest that the CPP would have led to significant emissions reductions and would have been a meaningful step toward decarbonizing the power sector. The CPP was, however, stayed by the Supreme Court and subsequently was replaced by the Trump administration with an alternative, the Affordable Clean Energy plan. Under that plan, there are strict limits on the measures that states can require, and states have the ability to waive or reduce the emissions reduction measures specified in the federal plan. As a result, the Affordable Clean Energy plan is projected to have negligible effects on emissions.13Regulatory approaches, whether under the Clean Air Act or more generally, have multiple drawbacks. Although some regulations can be efficient (the CPP being a prime example), many are not, in the sense that they result in emissions reductions that are costly per ton compared with the SCC. For example, there are many papers in environmental economics highlighting inefficiencies in the CAFE standards on automobile emissions.14 Estimates of emission reduction costs from that program range from $50 to more than $300 per ton. In addition, under existing legislative authority, regulatory approaches are limited in scope and are at best a partial solution to the climate problem. Moreover, regulations can be changed, and indeed the climate policy of the Trump administration largely consists of reversing Obama-era climate regulations. Finally, recent changes at the Supreme Court increase the odds that expansive interpretations of Clean Air Act authority to regulate greenhouse gasses will not be upheld. It is important to study the history of these regulatory approaches to inform policy design, and there are circumstances in which narrowly proscribed regulation might be the most efficient way to regulate emissions (e.g., command-and-control regulation of methane emissions in oil and gas drilling). That said, because of its limitations, I expect that regulation under the Clean Air Act is unlikely to play a major role in reducing emissions going forward.C. Narrowly Targeted PoliciesThe second category of climate policies is what I will call narrowly targeted. Examples include home weatherization programs, mandates to use biodiesel and corn ethanol in our fuel supply, and state-level renewable portfolio standards (RPSs). The costs of these policies vary widely. In a few cases, such as blending corn ethanol to comprise 10% of retail gasoline (the dominant blend in the United States), costs per ton are low or even negative. In many cases, however, the costs are high. For example, replacing petroleum diesel with biodiesel has a cost per ton of between $150 and $420, depending on the feedstock and how the incidence of the biodiesel tax credit is treated. Moreover, many of these policies interact in ways that increase costs but do not materially reduce emissions. For example, some states both have a RPS and participate in a regional cap-and-trade program for the power sector, such as the Regional Greenhouse Gas Initiative in the Northeast. Because electricity is provided on a multistate grid and cap-and-trade allowances are tradable across states, mandating clean energy in one state increases the number of allowances, reducing their cost and allowing more carbon emissions in other states in the regional program, a phenomenon that environmental economists refer to as "leakage."Within this catch-all group, one set of policies—namely, RPSs—does have the possibility of being impactful and cost-effective. Concerning impact, 29 states have renewable energy standards and some states, including California and New York, have announced midcentury goals of generating electricity that emits no greenhouse gasses. In theory, RPSs could become much more effective and efficient if all or nearly all states were to adopt them and if interstate trading of RPS allowances were introduced. With the important caveat that RPSs do not cover nuclear or other nonrenewable zero-carbon sources, a nationally tradable RPS system would approximate a national clean energy standard. This system would be less efficient than having a uniform price on carbon for the power sector, but it could come close (Goulder and Hafstead 2016, 2018), at least for the initial tranche of reductions. A noteworthy political economy feature of a nationally tradable RPS allowance market is that it would facilitate decarbonization in participating states with low RPS targets, more than achieving their targets with the cost underwritten by states with ambitious targets.With the exception of RPSs, this family of narrowly targeted policies tends to be small bore and in this sense is at best complementary in a broader package of solutions.D. Pricing CarbonThe third set of policies are carbon pricing policies. Although efforts to adopt a cap-and-trade program in the United States with the Waxman-Markey bill of 2009 failed, other countries and some states have adopted cap-and-trade systems or a carbon tax or fee on at least some sectors.The cost of a carbon tax depends on how the revenue is recycled. Here, I focus on the case in which it is returned by lump-sum rebates, as proposed by the Climate Leadership Council. In a recent book, Goulder and Hafstead (2018) use a multisector computable general equilibrium model to estimate the effect of carbon taxes with this and other revenue recycling schemes, along with other economy-wide climate policies. For a $20 per ton tax that increases by 4% per year and lump-sum recycling, they estimate that the level of gross domestic product (GDP) would be reduced by 1% over 30 years, amounting to an average reduction of GDP growth of just three basis points per year.It is also possible to look at actual macro outcomes for countries that have adopted a carbon tax. Preliminary empirical results for European countries, some of which have adopted carbon taxes, suggest small and statistically insignificant macroeconomic effects of a carbon price on growth (Metcalf and Stock, forthcoming; Metcalf 2019). These preliminary findings are consistent with the small GDP effect predicted by Goulder and Hafstead (2018).Goulder and Hafstead (2018) estimate that US emissions would be reduced by about one-third by 2050 if a $20 per ton tax were implemented. This finding aligns with estimates by the US EIA (2014, side case GHG25) and others (e.g., Larsen et al. 2018). These estimates underscore a key point: a carbon tax alone, at least at levels that are potentially politically viable, is insufficient to decarbonize the economy. An economist might retort that this statement is a non sequitur: if the carbon tax is set at the Pigouvian amount to equal the externality, then marginal cost equals marginal benefit and that is the optimal path and we should not adopt decarbonization as a goal or standard. But that reaction assumes that we can estimate the marginal benefit with some precision, it ignores the fact that other externalities are involved, and it fails to grapple with the deep uncertainty and potentially very negative outcomes arising from climate change.15It is important to understand that the emissions reduction from a carbon tax is nonlinear in the tax rate. A relatively small tax, say $20 to $30, essentially decarbonizes the power sector. But a tax of $20 per ton corresponds to $0.18 per gallon of gasoline. The demand reduction effects of this increase in driving costs are negligible: using the gasoline demand elasticity of −0.37 from Coglianese et al. (2017) and $3.50 per gallon gasoline, a $20 per ton carbon tax would decrease gasoline demand by only 2%. As inexpensive electric vehicles become increasingly available, the gasoline price elasticity could inc
Referência(s)