Sensitivity Analysis

Different assumptions for demand elasticities can have a large impact on simulated prices and traded volumes if the reference equilibrium deviates from the simulated equilibrium.

We will illustrate this effect in figure 3.4. Let S be the linear marginal cost curve, D and D’ the linear demand curve for different elasticity values, and E^ref the reference equilibrium. The reference equilibrium is determined by the reference quantity q^ref and the reference price p^ref. The capacity limit is denoted as q^max. The graph on the left depicts a situation in which the reference equilibrium coincides with a simulated competitive equilibrium. In this case, different elasticity values have little or no effect on prices and trade volume, as just the slope of the demand function changes, but not the intersection of demand with the supply curve.

q

Figure 3.4: Comparison of competitive market equilibria under different elasticity assump-tions

The graph on the right outlines a situation in which the real market equilibrium and the simulated competitive equilibrium in 2008 do not coincide. In contrast to the first setup, the reference equilibrium is not on the marginal cost curve, and different price elasticity values have a large effect and imply different model-based equilibria E’ and E”. In such a situation, a competitive model cannot reproduce the reference equilibrium.

Hence, either the capacity limit q^maxwas temporarily shifted to the left due to short-term bottlenecks or prices were strategically raised over marginal costs.

Consequently, we test our results presented in sections 3.6.1 and 3.6.2 for robustness by performing a sensitivity analysis regarding price elasticities of demand of importing

regions. We test for the following values, which broadly fall in the range presented in table 3.5: −0.1, −0.3, −0.5 and −0.6. Elasticities are assumed not to differ between the importing regions in each simulation run. Figure 3.5 presents a model and real market prices for different elasticity runs for the year 2006.

0 30 60 90

Actual eta = -0.6 eta = -0.5 eta = -0.3 Actual eta = -0.6 eta = -0.5 eta = -0.3 Actual eta = -0.6 eta = -0.5 eta = -0.3 Actual eta = -0.6 eta = -0.5 eta = -0.3 Actual eta = -0.6 eta = -0.5 eta = -0.3

Europe Japan South Korea Taiwan China 0 30 60 90

Actual eta = -0.6 eta = -0.5 eta = -0.3 eta = -0.1 Actual eta = -0.6 eta = -0.5 eta = -0.3 eta = -0.1 Actual eta = -0.6 eta = -0.5 eta = -0.3 eta = -0.1 Actual eta = -0.6 eta = -0.5 eta = -0.3 eta = -0.1 Actual eta = -0.6 eta = -0.5 eta = -0.3 eta = -0.1

Europe Japan South Korea Taiwan China

Figure 3.5: Prices in USD/t for different elasticity values (eta) in the year 2006 for oligopoly with fringe (left) and perfect competition (right)

Source: own calculations.

Clearly, model prices are more robust with regard to different elasticity assumptions in the competitive scenario, indicating that reference and simulated equilibrium are close to each other. Prices in the non-competitive scenario are significantly above actual 2006 price levels for all tested elasticity values. Also, capacity utilisation remains stable at 95% to 96% (see table 3.8) for different elasticities. These are indicators for competitive mechanisms to have applied in the real market during this year.

Figure 3.6 presents corresponding simulation results for the year 2008. In this case, the picture is less straightforward. In the competitive scenario, prices differ widely with regard to different elasticity values, thus indicating that reference and simulated equilib-rium differ significantly. Also, simulated prices remain lower than actual prices for any elasticity tested. Although the supply capacity limit is reached in the competitive sce-nario for higher elasticities (−0.5 and −0.6), model prices are still below actual prices⁵³. The simulated competitive market size is larger than the historic market volume (refer-ence quantities).

53Actually, traded quantity (reference quantity) is lower than the competitive market volume in the simulation runs. Hence, demand would have to be infinitely price elastic (horizontal demand curve) to match competitive model prices with real market prices at the capacity limit (vertical part of the supply curve).

0

Europe Japan South Korea Taiwan China 0

Europe Japan South Korea Taiwan China

Figure 3.6: Prices in USD/t for different elasticity values (eta) in the year 2008 for oligopoly with fringe (left) and perfect competition (right)

Source: own calculations. Note: Results for eta = -0.1 were omitted in the left graph for the sake of clarity. In this case, prices are well above 280 USD/t in most demand regions.

Both the oligopolistic and the competitive scenarios show relatively robust prices for higher elasticities (−0.5 and −0.6). However, oligopolistic prices fit actual prices better.

In the case of an elasticity value of −0.5, model prices are close to real market prices in major importing regions (except for China), and total model market volume is close to the real market outcome. However, trade flows are again distorted under the oligopoly market structure and therefore we still reject the oligopolistic scenario as an explanation for real market outcomes. See appendix B for a comparison of simulated trade flows with actual trade flows in this case.

Table 3.8: Capacity utilisation for different values of elasticity, in %

Oligopoly with fringe Perfect competition Capacity utilisation 2006 2008 2006 2008

eta = -0.6 85,9% 91,6% 96,0% 100,0%

eta = -0.5 85,3% 89,5% 95,9% 100,0%

eta = -0.3 83,9% 84,9% 95,5% 99,6%

eta = -0.1 82,2% 79,6% 95,0% 93,2%

Source: own calculations.

Furthermore, coal demand elasticity of more than −0.3 might seem to be unrealistic for major Asian importing countries. South Korean and Japanese coal-fired power plants experienced a major cost advantage compared to natural gas-fired power plants during most of 2008, which was also significantly higher than the cost advantage of coal in Europe (figure 3.7). As outlined in section 3.4.3, a detailed bottom-up analysis yielded a coal demand elasticity of −0.43 for the European power system in 2008. Due to the even higher advantage of coal in power generation in Asian import regions, it is therefore rather unlikely that coal demand elasticity for Asian importers has been as high as in Europe. Even after taking into account ramping capabilities and long-term fuel contracts, such high-cost advantages would probably lead to lower elasticity figures of around −0.3, the value we initially assumed in our base scenario for 2008. The cost

advantage of coal in Asia in power generation stemmed mostly from the very high oil-indexed LNG prices in 2008. LNG imports in these countries comprise virtually the total gas supply in Asian countries.

30 40 50 60 70 80 USD / MWh

0 10 20 30

1Q2008 2Q2008 3Q2008 4Q2008

Korea Japan Germany

Figure 3.7: Cost advantage of coal vs. natural gas in power generation

Source: IEA (2010a,b), EEX (2011). Note: We assume average coal power plant efficiencies of 41%

(Japan), 40% (Germany), and 36% (Korea) (IEA, 2010b). Assumptions for average gas power plant efficiencies are 47% (Germany, Korea, Japan). Carbon intensities are 0,335 tCO2/MWhth for coal and 0,201 tCO2/MWhthfor natural gas (Nagl et al., 2011).