C.C. Koopmans, D.W. te Velde r Energy Economics 23 2001 57]75 69
ICARUS with a 25 discount rate, we may conclude that the part explained by NEMO corresponds to an additional discount rate of up to 10.
We have outlined how we obtained trends and elasticities from the bottom-up information in ICARUS. Table 4 at the end of this paper contains trends and
elasticities for 19 separate sectors, thereby distinguishing between fuel and electric- ity use, and the type of investment. For a detailed discussion of the parameters see
Ž .
Koopmans et al. 1999 . In this paper we will present some general findings and discuss the implicit discount rate in NEMO.
Ž .
Table 1 presents aggregated results over sectors from NEMO. The results show Ž
. a long-term autonomous energy efficiency trend of approximately 0.8 per year
and a long-term price elasticity of energy efficiency of approximately 0.3. The ‘raw’ Ž
ICARUS data suggests an energy efficiency improvement of 1.5]2 per year at a .
15 discount rate , which is caused by the fact that ICARUS describes a hypothet- ical situation of full penetration of available and cost-effective energy saving
techniques in 2015. This hypothetical situation will not occur in reality because actual energy efficiency lags behind technological possibilities.
Ž .
Table 1 also compares NEMOs parameters to results from top-down historical studies. One should however take caution in comparing the studies, for two
reasons. Firstly, the historical studies, using sectoral data on energy use, exclude inter-sectoral effects but include intra-sectoral effects: if energy-intensive sub-sec-
tors grow faster than energy-extensive sub-sectors, it leads to a decline in energy intensity of the sector as a whole. This effect is not accounted for by ICARUS or
NEMO: these focus on efficiency. Secondly, this study investigates energy saving techniques at the energy demand side only, while the study by the Ministry of
Ž .
Economic Affairs 1996 includes energy efficiency improvements at both the supply and demand sides. Table 1 shows that NEMOs elasticities are compatible
with results from other studies. NEMO’s trends, however, are lower. This might be caused by the differences described above. It might also indicate that new tech-
nology for energy efficiency improvements will become available more slowly in the future than it did in the past.
5. Historical simulation
NEMO is not especially suited to simulate energy use in the past. Most of NEMOs parameters were estimated using technological possibilities for the period
1990]2010r2015.
6
The parameters we obtained from this do not necessarily reflect technologies available before 1990. In fact, we would need a bottom-up database
for, say, the period 1970]1990 to estimate parameters describing this period. As such a database is not available, we analyze the effects on energy efficiency of the
6
Strictly speaking, this would mean that NEMOs parameters are not valid for the period 2015]2020, and we could not give results after 2015. However, if NEMOs parameters are different in 2015]2020,
this would not have a large impact on the total results for 1995]2020. Therefore, we do present estimates for 2020 in section 6.
C.C. Koopmans, D.W. te Velde r Energy Economics 23 2001 57]75 70
energy price increases of 1973r1974 and 1979r1980, and the price fall of 1985r1986, using NEMOs ‘1990]2010’ parameters.
Apart from energy prices, the volume of total fixed investments also strongly influences NEMOs outcomes. This captures business cycle effects. Historically,
investments are negatively related to oil prices. The combination of real energy prices, investment levels and government policies determines NEMOs outcomes
with respect to energy efficiency.
A first round of simulations confirmed the results from the previous section that NEMOs aggregate energy efficiency trend parameter is lower than in studies based
on historical data. That is, predicted energy efficiency improvements on the basis of NEMO were lower than observed improvements over the period 1970]1990.
Ž .
This might be caused partly by intra-sectoral changes see above , but as we used very disaggregated data in measuring energy efficiency improvement, we expect
this effect to be much smaller than the difference we found. Another explanation would be that the pace at which new technologies become available is substantially
Ž .
lower between 1990 and 2010 ICARUSrNEMO than between 1973 and 1995 Ž
. historical data . To test this hypothesis, we did a second round of simulations in
which we increased all trend parameters by 50. The outcomes of the second simulation are presented in Table 2. Energy
efficiency increases strongly between 1973 and 1985, mainly as the result of the real energy price increases of 1974 and 1979r1980. The oil price shocks increase
the total yearly efficiency improvement between 1973 and 1985 to 2.1 per year. We can compare this to historical data: between 1975 and 1985, the efficiency
improvement was 2.0 per year. After the price fall of 1985r1986, the predicted rate of efficiency improvement falls back to 0.6 per year; the observed rate was
0.8 per year. The difference may be caused by government policies aimed at
Ž energy efficiency. New policies which were introduced in the 1980s regulations for
Table 2 Energy efficiency, Netherlands, 1974]1995; NEMO simulation and historical data
1974]1985 1986]1995
a
Ž .
NEMO simulation Efficiency improvement per year
Households Fuel
3.3 1.4
Electricity 3.3
2.5 Industry total
Fuel 1.8
0.1 Electricity
1.5 0.9
Transport Fuel
1.3 0.6
Other Fuel
2.3 0.8
Electricity 2.4
0.9 Total final energy use
2.1 0.6
b
Historical data Total final energy use
2 0.8
a
Ž .
With trend parameters increased by 50 see text .
b
Ž .
CPB monitoring of energy data from Netherlands Statistics CBS .
C.C. Koopmans, D.W. te Velde r Energy Economics 23 2001 57]75 71
. new-built houses, subsidies for energy-saving investments are not included in these
predictions; they may have increased observed energy efficiency improvements. On the whole, the second set of simulations with increased trends appear to
describe energy efficiency reasonably well over the period 1970]1990. This suggests that the pace at which new technologies become available after 1990 is lower than
in the 1970s and 1980s. Therefore, it may not be adequate to extrapolate historical trends of energy efficiency improvements into the future.
6. Policy analysis