The Economics of Climate Change: a Primer/Appendix

he economics literature contains hundreds of estimates of the costs or benefits (or both) of slowing, mitigating, or adapting to changes in the global climate resulting from human activities. Those estimates are derived mainly from a variety of computer models of economic activity that have been developed for other purposes and adapted to climate policy analysis. The variety of analytic approaches used makes it difficult even for modelers to interpret the differences among results from different studies. Researchers and policymakers are forced to integrate information from many sources and develop a synthesis based on a range of studies and approaches, each of which provides insight into some aspects of the problem while ignoring others.

Like their climate-related counterparts, modern models of the economy are composed of systems of mathematical equations that represent distinct but interacting processes in the real world and that are solved together to represent the simultaneous interaction of the parts within the whole. Even the most complex models of the physical climate or the economy inevitably lack detail. For example, many studies focus on the costs of mitigating climate change and ignore the potential benefits. Others focus only on costs in one sector or only on certain kinds of costs. Engineering studies evaluate the direct costs associated with adopting specific efficient technologies but tend to ignore larger-scale economic issues such as macroeconomic costs and impacts on international trade. Economic studies attempt to include a wider range of direct and indirect economic costs associated with emissions controls or with the effects of climate change, but they tend to use simple representations of technology. Some integrated assessment studies go further and try to incorporate both the costs of mitigation and the economic impacts of climate change. But to capture those very large-scale aspects of the problem, they rely on simplistic representations of many economic and environmental details.

Types of Models
Economic analyses yield a wide range of cost estimates for a given climate change mitigation policy, but most of the variation in results is due to identifiable differences in the approaches and assumptions that the studies use. Many analyses use one of several "top-down" approaches that represent the entire economy in an internally consistent way: they account for more or less all production and consumption; inputs of capital, labor, and energy; investment, taxes, and government spending; international trade; prices; interest and exchange rates; and so on. Top-down approaches allow researchers to account for many of the indirect economic effects of climate change policies that would primarily affect markets for energy; but they often ignore important details involving the gradual turnover of energy-using equipment, the choice of equipment, energy market barriers, and other factors. Nevertheless, they tend to produce fairly reasonable projections of overall energy use and thus emissions.

In contrast, "bottom-up" models draw on engineering cost studies to represent the details of specific energy-related technologies, but they tend to include much less detail about nonenergy sectors and other aspects of the economy. Unless constrained to do otherwise, bottom-up models always choose the most cost-effective technologies (from an engineering standpoint)—and therefore tend to produce rather unrealistic results.

Top-down modeling approaches, which as a class are sometimes referred to as macroeconomic models, generally fall into one of two groups. The traditional macroeconometric, or "macro," forecasting models that make up the first group are particularly useful in simulating the gradual adjustment of the economy to various kinds of shocks, such as changes in monetary and fiscal policy, higher energy prices, and exchange rate fluctuations. Macro models are particularly helpful in studying short-term (for example, five-year) responses and adjustments to economic shocks, but they do not represent specific markets in detail. Nor do they represent forward-looking expectations and behavior—an important element of economic activity, as discussed in the next section.

In contrast, computable general equilibrium (CGE) models, which form the second group, are useful in analyzing long-term responses to policies, over a decade or more. State-of-the-art CGE models incorporate forward-looking behavior, fairly detailed markets for specific factors and products, some types of gradual adjustment, aspects of long-run growth and technological progress, and detailed representations of the tax system and of international trade and finance. Some CGE models also include different groups and generations of households so that they can analyze the distributional impacts of climate change policies. Their disadvantage is that they do not capture gradual adjustments or elements of the business cycle very well. In particular, they have a hard time representing the gradual process through which industries and households replace equipment that is outmoded by policy shifts (as when consumers replaced cars that used leaded gas) and the gradual process through which a market economy adjusts to the economy-wide inflation that could result from significant increases in energy prices brought on by restricting emissions.

Many researchers combine several different models within a single modeling framework. For example, the Energy Information Administration’s National Energy Modeling System integrates a set of models of particular energy sectors, a national macroeconometric model, and an international econometric model. Other frameworks add models of the agriculture and forestry sectors to simulate flows of carbon dioxide and methane in those areas of the economy. Models that treat in detail the economy’s energy- or carbon-intensive subsectors tend to provide greater insight into those sectors’ responses to climate policies than do less complex approaches.

Treatment of Expectations
One of the most complicated aspects of economics is that people decide what to do today in part on the basis of their expectations about the future. Modeling people’s expectations is crucial to forecasting, but there is no simple formula to describe how people form them. Economists typically model expectations in one of two almost polar ways. One method represents behavior as adaptive: in that representation, people do not have an explicit understanding of how the economy will evolve and simply extrapolate from past experience into the future. The other approach represents people’s behavior as forward-looking, which is also termed model-consistent or rational: in that representation, people correctly anticipate the future evolution of the economy unless the modeler engineers an explicit shock.

The assumption of adaptive behavior, which is generally used in macroeconometric models, yields forecasts in which people fail to anticipate known developments—for example, they will fail to prepare for a change in policy that is announced 10 years in advance. In contrast, the assumption of forward-looking behavior, which is used in a number of sectoral and general-equilibrium models, yields forecasts in which people perfectly anticipate all developments. Both assumptions are extreme, and they yield significantly different results.

When models with adaptive expectations are used to estimate the costs of a tax on emissions or a permit system, they tend to produce somewhat higher cost estimates than models with forward-looking expectations, all else being equal. That happens because people in forward-looking models have time to adapt to any policy that is announced in advance. Modelers who use adaptive expectations adjust for that limitation by gradually phasing in the policy, to simulate the anticipation of forward-looking individuals. A more difficult problem for models that use adaptive expectations is how to represent the gradual turnover of the capital stock in response to a policy shift. A few modelers combine adaptive and forward-looking assumptions, yielding results that in many ways are between the two extremes.

Technological Change
To model future economic growth and the effects of emissions restrictions, modelers have to guess what kinds of technologies are going to be available at various points in the future and simulate their effects under scenarios that include and exclude policies to reduce emissions. Forecasting the use of particular technologies for the near to medium term is relatively straightforward, for two reasons: much of the capital stock (especially energy-using capital) lasts a long time, and innovations usually take a fairly long time to make their way into the market. Nevertheless, analysts have often failed in the past to anticipate technological advances that seem fairly obvious in retrospect, such as the relatively rapid development of the Internet in the 1990s, or—an innovation that is closer to the issue of climate change—the adoption of natural-gas-powered combined-cycle electricity generation. As forecasters look forward over two or more decades, their ability to anticipate technological developments becomes increasingly weak.

Most models represent the pace and direction of technological change in a fairly simple way because the underlying forces of change are not well understood. The models typically assume that independent developments will gradually reduce the amount of capital, labor, and, in particular, energy required to produce goods and services. The process of reducing energy inputs per unit of output is often represented by a parameter called the autonomous energy efficiency improvement (AEEI) parameter. For given rates of growth of gross domestic product and energy prices, an assumption of a higher AEEI implies that energy efficiency will improve more quickly and emissions will grow less quickly. A lower AEEI implies the reverse. Such a model can be used to analyze how changes in energy prices might encourage more or less use of energy, relative to the autonomous trend. Somewhat more complex models extend that basic process by projecting a menu of technologies that are expected to be available in the future and then analyzing how changes in energy prices would encourage people to switch to more energy-efficient types of equipment.

However, as discussed in Chapter 4, technological developments also respond to shifts in relative prices. Virtually no model simulates the effect of energy prices on the autonomous trend or on the menu of available technologies. Although the size of that so-called inducement effect is controversial, cost estimates that ignore it probably overestimate the incremental cost of reducing emissions over the longer run—say, 20 years or more.

Integrated Assessment
A consistent analysis of both the costs and benefits of policies related to climate change requires a modeling framework with certain characteristics: it should cover national and international greenhouse gas emissions from many sectors of the economy; it should translate emissions of greenhouse gases into changes in the atmospheric and global climate; and it should evaluate the impacts of climate change on people and ecosystems. A number of so-called integrated assessment models are under development, as are simplified reduced-form models based on more complicated frameworks. To analyze distributional issues, some models separate the world into a number of regions or include several overlapping generations of households. A few of the models also incorporate a range of uncertainty in their choice of parameters or in their solution procedures (see Chapter 3).