Integrated Assessment - What Can It Tell Us?

James Edmonds
Battelle Pacific Northwest Laboratories

My charge is to look at integrated assessment and what can it tell us. Integrated assessment isn't a new area, though there has been a great deal of interest in it lately.

There are a variety of definitions of integrated assessment (IA) that have been put forward. The one that I find most appealing is: "Integrated assessment examines knowledge derived from diverse disciplinary research endeavors for the purpose of understanding the implication of their interactions and to gain insights not otherwise available." That is, IA creates interdisciplinary knowledge from multidisciplinary sources of information. The classic example is energy-economic models coupled to carbon cycle or atmospheric chemistry models. The field has of late gone much further and preceding speakers have pointed out that IA models have looked at sea level rise, climate change, economic and non-economic damages from climate change, and have ultimately been put together into a cost-benefit framework.

When IA modelers are successful, they engage in interdisciplinary research. That is, there is enough knowledge transferred from the relevant disciplinary areas to affect more than the simple butting together of models that were created for other purposes.

Very importantly, IA models contain feedbacks and interactions of sub-systems. If it were not for that fact, the problem would be a simple linear one. That is, it would be possible to take results from one disciplinary area and simply add them linearly to the results of another disciplinary area. But IA produces insights from the interaction of disciplinary areas that were not obvious to researchers working separately in the individual disciplinary areas.

As I indicated, IA is an field which has received a great deal of attention of late. In 1977, when Bill Nordhaus put together the first of the integrated assessment models, there was just one model. Last April, there were 14 models participating in the Energy Modeling Forum, inter-model comparison. There are a large number of models operating at the present.

Because there are a great number of computer models operating, there has been a sense that integrated assessment is nothing but modeling. This is not correct. Integrated assessment occurs any time knowledge is combined to gain some insight otherwise unavailable. IA therefore need not be embedded in a computer code. The Inter-governmental Panel on Climate Change (IPCC) is an integrated assessment process without a computer code.

In looking back over the last few years of work in the field of IA I have chosen a set of four results derived from IA research efforts that have had a major impact:

  1. Sulfur emissions matter,
  2. Timing matters,
  3. Hedging helps us cope with an uncertain world, and
  4. Climate change is a global not regional problem.
The first result is that sulfur emissions matter. This is one of the most intriguing results that has come out of IA research. It stems from the fact that the use of fossil energy not only generates emissions of carbon compounds (all of which are greenhouse related gasses), but also sulfur emissions, which also affect the Earth's radiative balance.

Unlike emissions of carbon and nitrogen compounds, whose presence in the atmosphere tends to warm the climate, sulfur has a very powerful cooling effect. It also adds complexity to the formulation of policy that was not there before. This is important and needs to be recognized.

The second conclusion that previous speakers have mentioned is that timing matters. To implement the Framework Convention on Climate Change (FCCC), which has as its objective the stabilization of concentrations of greenhouse gasses in the atmosphere in a cost effective manner requires the distribution of emissions over time.

The third result grows out of the research on uncertainty and value of information. That is, it pays to hedge against unlikely but high-consequence events. If decision makers are risk averse, the existence of a high-consequence event, even if highly unlikely, tends to skew the response toward more action sooner than compared with a policy formulated on the basis of the most likely outcome.

Finally, climate change emissions are a global, not a regional or national, problem. It matters a great deal whether or not emissions controls are formulated as a cooperative or a non-cooperative endeavor. There is a lot of money at stake in this decision.

I will elaborate on each of these points. I will begin with sulfur. The first time that I encountered the sulfur result was a couple of years ago in doing some work for IPCC Working Group II, in which we were examining advanced energy technologies, which were presumed to be available at costs which were highly competitive. We didn't grapple with whether or not those costs were available today or not; we simply assumed for the sake of the calculation that we could achieve these goals of getting very low cost technologies which do not emit carbon freely to the atmosphere.

We gradually introduced them into a scenario whose reference case was similar to the IPCC's case, IS92a, a scenario which is right down the middle of the range of research results from a vast literature. We also examined their implications when the reference case was similar to IPCC IS92c, which is toward the low end of emissions under business-as-usual, and IPCC case IS92e, a case toward the high end of the range of business-as-usual scenarios found in the literature. The introduced of advanced energy technologies reduces fossil fuel carbon emissions dramatically. In fact they reduce emissions to the point where the concentration of atmospheric CO2 stabilizes below 500 parts per million without further policy intervention.

But surprisingly, at very low emissions the global mean temperature was higher for the first 50 years than in the reference case. This was highly surprising. When we went back and unraveled this unexpected result, we found that as the emissions of carbon were being reduced, so too were the emissions of sulfur. We were cleaning up an acid rain problem at the same time as we were taking care of the greenhouse problem, but we had made the near-term greenhouse problem worse.

That result was very surprising. IPCC Working Group I found the sulfur effect was so powerful that it explains why climate models without sulfur in them show about twice as much warming as observed in the historical record. Sulfur emissions explain the apparent inconsistency in the history of climate modeling. But in an IA model sulfur implies that very aggressive emissions reductions in the near term lead to higher temperatures rather than lower temperatures. Note however, that past the year 2050 the effects of sulfur are dominated by the direct effects of carbon dioxide and other greenhouse gasses, and that reductions in emissions lead to lower temperatures than in the reference case.

The second noteworthy IA modeling result concerns the timing of emissions reductions. Timing matters in implementing the FCCC, which has as its goal the stabilization of greenhouse gas emissions at concentrations that would prevent dangerous anthropogenic interference with the climate.

Having said that, the FCCC was put together by teams of international negotiators who crafted language that everybody could live with. When the goal was established, it was a goal with which everyone could agree, but the FCCC never said what constitutes dangerous anthropogenic interference with the climate.

I will focus today on two concentration ceilings: 450 ppmv and 550 ppmv. They are interesting because they require policy intervention immediately to achieve. But they are not so low as to require negative emissions. And they are not so high as to require no near-term policy intervention at all.

The identification of an emissions trajectory which simultaneously minimizes economic cost and leaves an atmospheric concentration ceiling is a familiar problem to economists. Researchers in the area of climate change have examined variants of the problem for many years. Nordhaus (1979) was the first to examine the problem which has more recently been considered by Richels and Edmonds (1994), Kosobud et al. (1994), Manne and Richels (1995), and Wigley, et al. (1995). Several reasons argue for a particular pattern of emissions if the economic burden to society is to be minimized. The generic path for an emissions trajectory which stabilizes emissions has three phases:

  1. Increasing emissions,
  2. Emissions Peak, and
  3. Declining emissions.
The extent of each of these phases depends on height of the ceiling, the cost of emissions reductions and the nature of the cost function, and on the carbon cycle.

The rationale for the three phase optimal trajectory can be seen by considering a ceiling of 500 ppmv. Begin by noting that present fossil fuel emissions rates of 6 PgC/yr imply a CO2 concentration in year 2100 of 500 ppmv. Since constant emissions imply concentrations in the years beyond 2100 in excess of 500 ppmv, then observance of a ceiling implies that year 2100 fossil fuel emissions must fall below 6 PgC/yr. If this is the case, then emissions in some year prior to 2100 must be in excess of 6 PgC/yr. The economic and carbon cycle considerations listed below argue that any extra emissions be taken near the present:

  1. The Carbon Cycle Dividend: For any concentration ceiling, cumulative emissions tend to be greater the earlier in the period of analysis emissions are released. Thus comparing two emissions trajectories, both of which satisfy the concentration ceiling, the profile with larger near-term emissions will also have larger cumulative emissions. We refer to this phenomenon as the "carbon cycle dividend." The carbon cycle dividend can be substantial. For the 550 ppmv computation the difference between cumulative emissions for the stable emissions and atmospheric stabilization cases between 1990 and 2100 is 290 PgC (48 years of fossil fuel carbon emissions at 1990 rates). The "carbon cycle dividend" is so large that annual emissions need not return to 1990 levels until the year 2040 to achieve the 450 ppmv ceiling and the year 2100 to satisfy the 550 ppmv ceiling. This is despite the fact that steady-state 1990 emissions exceed a concentration of 550 ppmv early in the twenty- second century. The carbon cycle dividend operates because the sooner emissions enter the atmosphere, the longer natural removal mechanisms have to operate and the more that can be removed by a fixed date.

  2. Discounting: Any economic activity such as emissions reductions requires the allocation of scarce resources. A positive marginal product of capital means that resources set aside today can accommodate a larger burden, if they are not needed until some future date. Put another way, the further in the future a given economic burden lies, the smaller the resources that must be set aside today to facilitate the activity. There is great leverage to be gained by postponing the burden of emissions reductions. This is not the same as procrastinating. There is still a present day resource equivalent burden, but the overall resource requirements are lessened. Furthermore, it does not mean doing nothing. Emissions must eventually peak and decline and therefore the nature of new investments must change immediately, because each piece of new capital installation must spend an increasingly long period operating in a reduced emissions environment.

  3. Technology Development: Emissions reduction technologies are not static. As the IPCC Working Group II notes, multiple technologies exist in various stages of development and deployment which have the capacity to reduce emissions in the future at costs lower than those required today. Since cumulative emissions reductions are largely fixed (subject to the qualifications of the carbon cycle mentioned above) then costs are lowered by undertaking emissions reductions most strenuously when they are cheapest -- in the future -- since a tonne of emissions reduction counts as much or more if it is undertaken in the future as it does if taken now.

  4. Capital Utilization: Both physical and human capital are created with a set of expectations about the environment in which they will be used. These expectations shape the character of these investments. Historical investments have been made with the expectation of little or no carbon emission penalty. There is a cost to utilizing capital in an environment for which it was not configured. As a general rule, the greater the departure from the anticipated operating environment, the larger the penalty sustained. For atmospheric stabilization ceilings which do not require immediate emissions reductions, there is a benefit to announcing future emissions controls, but implementing them over time. See point 2 above for further discussion.
Richels and Edmonds (1994) have shown that the cost of efficiently stabilizing the atmosphere at 500 ppmv may be only half the cost of efficiently stabilizing emissions.

The third observation that comes out of integrated assessment is that in the face of an unlikely but highly undesirable event an appropriate risk averting policy response is to hedge. In terms of the FCCC, hedging tells us to set a lower greenhouse gas concentration ceiling than we would otherwise. So these results are a different cut through the climate problem than the timing result.

The hedging result says that if you don't love risk, i.e. you are risk averse, that even if disaster is unlikely, but still possible, recognizing the potential for that disaster would lead you to take more action sooner than you would if you just took a statistical approach to the problem.

The literature that has looked at this is relatively small so far. But what it tends to indicate is that given the current probabilities of extreme consequences, that the appropriate financial surcharge is approximately $5 per tonne of carbon emissions.

The fourth result emerging out of the integrated assessment literature is one that joint implementation of emissions reductions are cheaper than uncoordinated policies which achieve the same emissions reductions. Climate change emissions are a global, not a regional or national problem.

Minimizing the cost of any emissions reduction goal requires joint implementation. That is we seek mechanisms that will induce all parties to reduce emissions up to the point where the cost of the last tonne of emissions reduction is the same for every one. If the marginal cost of emissions reductions are unequal, there are gains to be had by trading. If it costs me a hundred dollars to get rid of a tonne and it costs you one dollar to get rid of a tonne, there is a wide range of potential benefit to both of us for me to pay you to take out a tonne rather than me. I can pay you up to 99 dollars and be ahead. You can take anything above a dollar and be ahead. So there is tremendous room for benefit to trade, and we saw this result discussed in the cost-benefit analysis paper. Any time there is a difference in the marginal cost of emissions reductions, there is potential for everybody to be better off and nobody to be worse off by emissions trading.

The magnitude of these gains can be seen in the example of the AOSIS proposal. If OECD nations execute emissions reductions responsibilities individually, the total global cost is about a little bit more than two and a half trillion 1990 US dollars, discounted at five percent per year from 2050 to the present. If these emissions reductions were taken globally on a least cost basis, costs would be cut by two-thirds.

I am open to questions.

Go To Discussion - Session III