8.2.1.4 Common Messages from Bottom-up Results
Clearly, the impact of policy scenarios has a large influence on abatement
costs. Certain studies propose a series of public measures (regulatory and economic)
that tap deep into the technical potential of low carbon and/or energy-efficient
technologies. In many cases, such policies show low or negative costs. A comparison
with least-cost approaches is difficult because these evaluate systematically
both the baseline and the policy scenario as optimized systems and do not incorporate
market or institutional imperfections in the current world. It would be of great
interest to conduct a more systematic comparison of the results obtained via
the various B-U approaches, so as to establish the true cause of the discrepancies
in reported costs. A timid step in this direction is illustrated in Loulou and
Kanudia (1999a).
This leads to a general discussion about the extent to which all these results
suffer from a lack of representation of transaction costs, which are usually
incurred in the process of switching technologies or fuels. This category of
transaction cost encompasses many implementation difficulties that are very
hard to capture numerically. The general conclusion from SAR (that costs computed
using the B-U approach are usually on the low side compared to costs computed
via econometric models, which assume a history-based behaviour of the economic
agents) is no longer generally applicable, since some B-U models take a more
behavioural approach. Models such as ISTUM, NEMS, PRIMES, or AIM implicitly
acknowledge at least some transaction costs via various mechanisms, with the
result that market share is not determined by visible (market-based) least-cost
alone. Least-cost modellers (using MARKAL, EFOM, MESSAGE, ETO) also attempt
to impose penetration bounds, or industry-specific discount rates, which approximately
represent the unknown transaction costs and other manifestations of resistance
to change exhibited by economic agents. In both cases these improvements result
in partially eschewing the “sin” of optimism and blur the division
between B-U and T-D models. While the former, indeed, tend to be less optimistic
when they account for real behaviours, it is symmetrically arguable that the
latter underestimate the possibility of altering these behaviours through judicious
policies or better information. All this area still remains underworked.
A common message is the attention that must to be paid to the marginal cost
curve. Despite the limitations and differences in results discussed above, B-U
analyses convey important information that lies beyond the scope of T-D models,
by computing both the total cost of policies and their marginal cost. Very often,
indeed, the marginal abatement cost of a given target is high, although the
average abatement cost is reasonably low, or even negative. This is because
the initial reductions of GHG emissions may have a very low (or negative) cost,
whereas additional reductions have, in general, a much higher marginal cost.
This fact is captured in the curve representing marginal abatement cost versus
reduction quantity, which starts with negative marginal costs, as illustrated
in Figure 8.1. The initial portion of the curve
(section A–B) exhibits negative cost options, which may add up to a significant
portion of the reductions targetted by a given GHG scenario. As the reduction
target increases (section B–C–D of the curve), the marginal cost becomes
positive, and also eventually the total mitigation cost if the reduction target
is large enough. But there is systematically a wedge between the marginal and
total costs of abatement, and this wedge is all the more important as the macroeconomic
impacts of climate policies are driven in large part by the marginal costs (because
the latter dictate the change in relative commodity prices). They are driven
only modestly by the total amount of abatement expenditures.
A crucial, albeit indirect, message, is the importance of innovation: indeed,
B-U models depend on a reasonable representation of emerging or future technologies.
When this representation is deficient, the models present a pessimistic view
of the costs of more drastic abatements in the long term. This issue is not
one of the modelling paradigm, but rather of feeding the models with good estimates
of technical progress. Some works are currently underway to make explicit the
drivers of technical change, such as learning-by-doing (LBD) or uncertainty.
These studies are discussed further in Section 8.4.
|