The third reason for concern relates to the overall (i.e., worldwide or aggregate) economic and ecological implications of climate change. Numerous studies have addressed aggregate impacts, particularly in the context of integrated assessment.
Estimating the aggregate impact of climate change is an intricate task that requires careful professional judgment and skills. Aggregate analysis is based on the same tools as most distributional analysis and uses regional data as inputs. Consequently, it shares with distributional analysis the methodological difficulties and shortcomings discussed more fully in Section 19.4:
In addition, analysts have to grapple with some issues that are generic to aggregate analysis. The most important issue is spatial and temporal comparison of impacts. Aggregating impacts requires an understanding of (or assumptions about) the relative importance of impacts in different sectors, in different regions, and at different times. Developing this understanding implicitly involves value judgments. The task is simplified if impacts can be expressed in a common metric, but even then aggregation is not possible without value judgments. The value judgments that underlie regional aggregation are discussed and made explicit in Azar and Sterner (1996), Fankhauser et al. (1997, 1998), and Azar (1999). Aggregation across time and the issue of discounting are discussed in more detail in TAR WGIII Chapter 7. Aggregate impact estimates can be very sensitive to the aggregation method and the choice of numeraire (see Chapter 1).
All of these factors make aggregate analysis difficult to carry out and reduce
our overall confidence in aggregate results. Nevertheless, aggregate studies
provide important and policy-relevant information.
Most impact studies assess the consequences of climate change at a particular concentration level or a particular point in time, thereby providing a static "snapshot" of an evolving, dynamic process. The SAR suggested that the aggregate impact of 2xCO2expressed in monetary termsmight be equivalent to 1.5-2.0% of world GDP. Estimated damages are slightly lower (relative to GDP) in developed countries but significantly higher in developing countriesparticularly in small island states and other highly vulnerable countries, where impacts could be catastrophic (Pearce et al., 1996). The SAR was careful, however, to point out the low quality of these numbers and the many shortcomings of the underlying studies.
Since publication of the SAR, our understanding of aggregate impacts has improved, but it remains limited. Some sectors and impacts have received more analytical attention than others and as a result are better understood. Agricultural and coastal impacts in particular are now well studied (see Boxes 19-2 and 19-3). Knowledge about the health impacts of climate change also is growing (see Box 19-4). Several attempts have been made to identify other nonmarket impacts, such as changes in aquatic and terrestrial ecological systems and ecosystem services, but a clear and consistent quantification has not yet emerged.
Table 19-4 contains a summary of results from aggregate studies that use money as their numeraire. The numerical results as such remain speculative, but they can provide insights on signs, orders of magnitude, and patterns of vulnerability. Results are difficult to compare because different studies assume different climate scenarios, make different assumptions about adaptation, use different regional disaggregation, and include different impacts. The estimates by Nordhaus and Boyer (2000), for example, are more negative than others because they factor in the possibility of catastrophic impact. The estimates by Mendelsohn et al. (2000), on the other hand, are driven by optimistic assumptions about adaptive capacity and baseline development trends, which result in mostly beneficial impacts.
Standard deviations rarely are reported, but they are likely to be several times larger than the "best guess." They are larger for developing countries, where results generally are derived through extrapolation rather than direct estimation. This is illustrated by the standard deviations estimated by Tol (2001b), also reproduced in Table 19-4. These estimates probably still underestimate the true uncertaintyfor example, because they exclude omitted impacts and severe climate change scenarios. Note that the aggregation can mask large standard deviations in estimates of damages to individual sectors (Rothman, 2000).
An alternative indicator of climate change impact (excluding ecosystems) is the number of people affected. Few studies directly calculate this figure, but it is possible to compare the population of regions experiencing negative impacts with that of positively affected regions. Such calculations suggest that a majority of people may be negatively affected already at average global warming of 1-2°C. This may be true even if the net aggregate monetary impact is positive because developed economies, many of which could have positive impacts, contribute the majority of global production but account for a smaller fraction of world population. The quality of estimates of affected population is still poor, however. They are essentially "back-of-the envelope" extensions of monetary models, and the qualifications outlined in that context also apply here. In addition, they do not consider the distribution of positive and negative effects within countries.
On the whole, our confidence in the numerical results of aggregate studies remains low. Nevertheless, a few generic patterns and trends are emerging in which we have more confidence:
Overall, the current generation of aggregate estimates may understate the true cost of climate change because they tend to ignore extreme weather events, underestimate the compounding effect of multiple stresses, and ignore the costs of transition and learning. However, studies also may have overlooked positive impacts of climate change. Our current understanding of (future) adaptive capacity, particularly in developing countries, is too limited, and the treatment of adaptation in current studies is too varied, to allow a firm conclusion about the direction of the estimation bias.
Other reports in this collection