Box 4-7: The Role of Prices in SRES Scenarios

The price of energy comprises many components:

• Costs to establish and maintain the production, conversion, transport, and distribution infrastructure of energy supply.
• Profit margins.
• A whole host of levies such as royalties and taxes raised at the points of energy production or use.
• Consumers' willingness to pay for quality and convenience of energy services.

Furthermore, given the importance of energy and the vast volumes traded, prices are influenced by a whole range of additional factors, from inevitable elements of speculation to geopolitical considerations, all of which can decouple energy price trends from any underlying physical balance between supply and demand. Taxes are especially significant. In a number of OECD countries, up to 80% of the consumer price of gasoline is taxes (OECD, 1998), and the differences between countries are enormous. In 1997, 27% of the price of gasoline in the USA was taxes, compared with 78% in France. Taxes vary substantially even between large oil producers (and exporters). In Mexico taxes are 13% of gasoline prices, but in Norway they are 75% (OECD, 1998).

Currently, no methodologies exist to project future energy prices taking all of above mentioned factors into account, nor were the SRES scenarios intended to make explicit assumptions on such factors such as future energy taxation. Price information enters long-term emission models either in the form of exogenous scenario assumptions, or it is derived internally in models based on simplified representations of price formation mechanisms usually based on (marginal) cost information.

The six models used for SRES range from detailed "bottom-up" models (e.g., AIM, IMAGE), through macro-economic (partial equilibrium) models (e.g., MARIA, MiniCAM), to hybrid approaches (successive iteration between the engineering model MESSAGE with a macro-economic model, or using the Worldscan model with IMAGE). Each has different representations of price formation mechanisms and their relationship to macro-economic or sectoral energy demand. These are summarized in Appendix IV. As a rule, "bottom-up" (optimization) models calculate only (average and marginal) costs endogenously. As a result of their sectoral perspective (energy, agriculture, etc.), these models cannot determine macro-economic feedbacks on other sectors or the entire economy and thus are unable to represent a consistent picture of price formation. Conversely, price formation is endogenized in "top-down" models; however, these rely on the stringent assumption that demand and supply must be in equilibrium and in addition provide little sectoral detail. Over recent years this simplified modeling dichotomy has progressively weakened because of further advances in methodology and the development of "hybrid" modeling approaches. To illustrate the methodologies deployed in the six SRES models, two (MARIA and MESSAGE) are discussed here, but (for space limitations) only in terms of one scenario (B2). (Table 4-9 gives additional details of an inter-scenario comparison of energy prices for the MiniCAM and ASF models. Owing to methodologic differences, a comparison of prices across scenarios is only possible within a consistent approach (i.e. be comparing scenarios quantified with the same model).)

The energy prices represented in MARIA (see also Mori, 2000) consist of energy production and energy utilization costs. Market prices are determined endogenously by model-calculated shadow prices (for further model details see Appendix IV and Mori and Takahashi, 1999). Among various parameters, the extraction costs of fossil fuel resources and the coefficients of utilization costs and their evolution over time are the most important determinants. For the MARIA runs, the resource estimates of Rogner (1997) were used as input. For the sake of simplicity, all fossil resource categories of Rogner (1997) were aggregated into two classes and a quadratic production function was used to interpolate the extraction costs of reserves and all other occurrences. For coal, long-term extraction costs range up to US$6.3 per GJ in 1990US$ prices, for gas up to US$25 per GJ, and for oil up to US$28 per GJ (see Appendix IV for further details). The energy cost coefficients (representing 16 different energy conversion technologies) are based on Manne and Richels (1992). For the B2 scenario quantification, the Manne and Richels (1992) estimates were largely retained. For instance, electricity generation costs range between 14 mills²⁶ /kWh for gas to 51 mills/kWh for coal. (For the other scenario quantifications these cost values were modified to conform to the different interpretations of a particular scenario storyline.) Together these assumptions determined long-run costs and shadow prices that were set equal to energy prices in the macro-economic production function of MARIA. The energy prices were combined with assumed (low) AEEI values and potential GDP growth rates (the latter from the B2 marker) to calculate the resultant aggregate energy demand in the model. The resultant primary energy demand was (with exception of the REF region) within 15% of the respective B2 marker quantification at the regional level and within 5% of global energy demand. As a result of different model structures, comparable price data for the MESSAGE model are only available for internationally traded primary energy forms (these are given in Table 4-8).

The bottom-up, systems engineering (optimization) model MESSAGE does not compute energy prices. Instead, the model is entirely based on cost information, but such costs are treated as dynamic. Their overall treatment follows the lines outlined above for the MARIA model, except that greater technology-specific detail is contained in the model. Altogether 19 different fossil resource grades are differentiated, based on the estimates of Rogner (1997). The resultant (levelized) extraction costs for the B2 marker are in the range US$1.1 to US$5.4 per GJ for coal, US$1.2 to US$5.3 per GJ for oil, and US$1.2 to US$5.7 per GJ for gas (range indicates costs variations between lowest and highest costs of the four SRES regions for 2020, 2050, and 2100 respectively, see Appendix IV). Technology-specific cost assumptions cannot be summarized here as MESSAGE contains literally hundreds of energy supply and end-use technologies. Examples of cost assumptions are given in Section 4.4.7 and more detail is reported in Riahi and Roehrl (2000). However, as in MARIA, MESSAGE also calculates shadow prices for internationally traded primary energy forms and therefore these two indicators can be compared (Table 4-8).

A. Costs include export and/or import infrastructure and transport.
B. Range between crude oil and light and heavy oil products.
C. Range between liquid natural gas and direct pipeline imports to North America, Europe, Japan, and North Africa.
D. Range between methanol, ethanol, and liquid hydrogen.

To achieve consistency between model-calculated energy cost dynamics and energy demand assumptions an iterative modeling procedure between MESSAGE and MACRO (a macro-economic production function model based on Manne and Richels, 1992) was used, on the basis of model-calculated shadow prices as indicators of future price dynamics. The methodology is described in more detail in Wene (1996). This approach requires time-intensive model iterations, but has the advantage that the impact of price increases can be separated from efficiency improvements through fuel substitution (e.g., a gas-fired cook stove energy end-use efficiency is up to 10 times higher than a traditional cook stove fired with fuelwood) as well as from "everything else," i.e., the AEEI in the traditional sense). Aggregated, the impact of (shadow) price increases in MESSAGE's B2 scenario accounts for 8% of global primary energy demand by 2020, 23% by 2050, and 30% by 2100. This impact is calculated as a reduction in energy demand compared to a hypothetical scenario with constant 1990 prices (and correspondingly higher energy demand). The impact of price increases on future energy demand in the B2 scenario is thus relatively small compared to that of other factors, although far from negligible. This also explains why the two B2 scenario quantifications by MARIA and MESSAGE have quite similar energy demand figures, even if international trade prices may differ. First, trade prices are only one component of the cost-price mechanism treated in the models (which also includes domestic energy production, conversion, and end-use costs). Second, models differ in their parametrizations of the "everything else" (AEEI) model parameters, for which a wide range of views on applicable ranges exists. Therefore it is one of the model calibration parameters frequently used to replicate existing scenarios or to standardize inter-model comparison projects such as EMF-14 (Weyant, 1995).

2000
2020
2050
2100