Role of International Knowledge Spillovers

The analysis presented here is carried out by means of the World Induced Technical Change Hybrid (WITCH) model, an energy-economy-climate model developed by the climate change group at FEEM. The model has been used extensively for the analysis of the economics of climate change policies.

WITCH is an economic model with a specific representation of the energy sector, thus belonging to the new class of fully integrated (hard link) hybrid models. It is a global model, divided into 12 macro-regions. For the present analysis the distinguishing features of the model are two. The first one is the representation of endogenous technical change in the energy sector. Advancements in carbon mitigation technologies are described by both diffusion and innovation processes. Learning by Doing and by Researching allow to devise the optimal investment strategies in technologies and R&D in response to given climate policies. Moreover, knowledge in a country does not depend solely on R&D investments in that country but it is partially affected by other countries R&D investments, via an international spillovers mechanism. The second relevant modelling feature is the game-theoretic set up. The model is able to produce two different solutions, one assuming countries fully cooperate on global externalities, the so called globally optimal solution. The second is a decentralized solution that is strategically optimal for each given region in response to all other regions choice, the definition of a Nash equilibrium. This modelling features allows us to account for externalities due to all global public goods (CO2 , international knowledge spillovers, exhaustible resources etc.), making possible to model free riding incentives. In terms of innovation market failures, the WITCH model captures the international flow of energy related knowledge via spillovers across national boundaries, see Bosetti et al. 2008. It also accounts for higher social returns of R&D by calibrating their investments marginal price of capital to four time the interest rate.

These modelling characteristics allow us to depict both policies aimed at internalizing the technological externality as well as those dealing with the climate externality, independently or in combination. This allows checking for potential mutual interactions of overlapping policies. However, it should be noted that important additional externalities such as appropriability and knowledge protection issues are not captured by the macro-economic structure of the model.





1. Even though costs are expressed as % of GDP, these costs do not represent GDP losses but rather loss of World surplus.

Basic Stabilization Policy Cases

The model results we present here relate to two basic stabilization policy scenarios that we measure against a business as usual (BaU) scenario, a 3.5 and a 4.5 W/m2 radiative forcing (RF) scenarios. As in Bosetti et al. (2009) we analyse the effect of different stabilization policies on energy investments and innovation, though we concentrate here on a different set of long term target. In the subsequent section we investigate how alternative scenarios on international spillover of energy knowledge might affect these results. The idea is to see how assumptions concerning spillovers in energy knowledge affect global and regional costs, hence how innovation policies can be effectively used to make climate policies more efficient and to improve the distributional effects of an international climate agreement.

Let us start by investigating basic implications of the stabilization scenarios. Figure 1 shows the cumulated GHG emissions implied by the two stabilization scenarios and the BaU case.

Figure 1: Cumulative emissions for three scenarios (in GtC)

Figures 2-4 report on fossil fuel carbon emission pathways consistent with the two policy scenarios for the whole world, OECD countries and non OECD countries, respectively, while the resulting effect on radiative forcing is reported in Figure 5. The price of carbon for these two scenarios is reported in Figure 6. As it can be noted, while the price for the 4.5 W/m2 RF scenario stays in the range of 200 $/ton CO2 eq, the more stringent scenario entails a radical shift in the energy sector paradigm, which means a price reaching the 1000 $/ton CO2 ceiling.

Figure 2: Fossil fuel emission pathways (GtC/Y) for three scenarios: whole World
Figure 3: Fossil fuel emission pathways (GtC/Y) for three scenarios: OECD countries
Figure 4: Fossil fuel emission pathways (GtC/Y) for three scenarios: non OECD Countries
Figure 5: Radiative forcing effect for three scenarios
Figure 6: Carbon price ($/tC) for three scenarios




Breakthrough in power generation technologies is modelled by introducing a Backstop technology, that can be better thought of as a compact representation of a portfolio of advanced technologies that can substitute nuclear power. In the model, energy consumption in the non-electric sector is based on traditional fuels (traditional biomass, oil, gas and coal) and biofuels. In order to account for food security concerns, overall penetration of biofuels is assumed to remain modest over the century. The consumption of oil can be substituted with a second carbon free Backstop technology, which could be thought as next generation biofuels or carbon free hydrogen. R&D investments facilitate the competitiveness of Backstop technologies in both the electric and non-electric sectors whose initial price is 10 times higher than conventional technologies.

Figures 7 and 8 show the technological changes required in the energy sector to cope with the two targets. Fossil fuel consumption is progressively replaced by carbon free technologies as renewables, IGCC plus CCS, nuclear, and a backstop carbon free technology 2 in both the electricity and non electricity sector. An important role is also played by energy intensity improvements. A detailed description of how different energy technologies responds to different climate policies can be found in Bosetti et al (2009).

Figure 7: Technological change in the energy sector

Energy R&D investments

In both stabilization scenarios total energy R&D levels have to increase sharply and ramp up to levels comparable to those registered in the 80's in order to trigger the backstop technology penetration (see Figure 9). In the more stringent scenario, not only the pick in energy R&D expenditure should be higher but it should also take place earlier, as the necessity to decarbonise the non electricity sector become more urgent in earlier periods.

Figure 8: Technological change in energy sector
Figure 9: Energy R&D investments

The cost to the world of the two policies is expressed as the difference in the Gross World Product in the policy case with respect to the BaU. The time profile of costs is shown in Figure 10.

Figure 10: Cost profile: whole World

A more compact cost measure is the discounted Gross World Product loss. Using a 5% discount rate, then policy costs are 0.37% and 1.55% in the 4.5 and 3.5 W/m2 radiative forcing scenarios, respectively. As we will look in detail to the costs of the two policies for different countries, we start by looking at the regional costs implied by the basic setting. Regional costs depend on three important forces: the stringency of the target, that is reflected in the price of permits; regional marginal abatement costs that are implicitly defined by the costs, penetration rates and availability of carbon free technologies; and the international scheme for the allocation of permits. For this analysis we assume permit are allocated on the basis of the Contraction & Convergence ( C&C ) rule, with full convergence in 2050. Starting for the first year of the agreement allocations are distributed proportionally to emissions in base year; allocations then converge to an equal per capita scheme by 2050. In Table 1 the regional policy costs measured as GDP losses discounted at 5% are presented. Main losers are TE, MENA and China, while Africa, India and SEASIA gain from the selling of emissions permits on the international market. Although a popular scheme, it is clear that a C&C based distribution of costs would not be considered fair by a large number of countries negotiating on a burden sharing rule.

Table 1: Regional policy costs measured as GDP losses discounted at 5%

Effects of Innovation and R&D International Spillovers

In the basic version of the model, energy intensity technical change is calibrated in order to replicate past trends in the energy improvements. In addition, two factor learning curves for breakthrough technologies have been calibrated using data on past evolution of other carbon free technologies. However, the degree of uncertainty surrounding these processes is very high, also because they will be affected by policies. In particular, the basic version of the model assumes that:

  • 1. the degree of spillovers of knowledge affecting breakthrough technologies depends on the absorption capacity of each country (approximated by that country's shareof world backstop R&D investments);
  • 2. learning by doing effect determines completely spillover, i.e. it is the world cumulative capacity installed that affects the backstop technology investment cost.

However, the economics of international R&D spillovers is not entirely settled, and empirical evidence is scarce, making it diffcult to model and quantify these effects.

Therefore any model-based analysis of international R&D spillovers and their implications for optimal international R&D policy should be interpreted with great care. Moreover, it is useful to perform extensive analyses to try to quantify the effect of policy designed to improve these spillover effects In the present section we want to see how different assumptions on innovation and the international transfer of this innovation might affect the cost of a climate policy and its distribution. To this aim we consider a set of three variations:

  • 1. «no breakthrough in technologies», in this scenario we assume that the R&D program set out to lower the cost of a carbon free alternative to fossil fuels both in the power sector and, more importantly, in the transport sector, fails to deliver the desired results;
  • 2. «increased spillovers through an international R&D fund», in this scenario we assume OECD countries contribute with 0.08% of their GDP to a fund that is used to directly finance dedicated breakthrough R&D in developing countries;
  • 3. «no international spillovers», in this scenario we switch off all transfers of technological change that are there in the basic version of the model in order to mimic a worse case scenario.

No breakthrough in technologies

The first variant case we consider shows what happens if a breakthrough in technology, both in the electric sector and in the non electric sector, does not take place, that is the dedicated R&D program fails to trigger the technological leap. CCS, nuclear, biomass and renewables are available and make up for the absence of the breakthrough technology. Such dramatic assumption has a very strong effect on the cost of stabilization policy, and in particular on the cost of stabilizing at 3.5 W/m2 radiative forcing, where a major breakthrough in the non electric sector appears to be crucial. Indeed the cost of the two stabilization policies would increase of 80% and 60% if no breakthrough in technologies has to take place, in the 3.5 and 4.5 W/m2 radiative forcing scenarios, respectively.

The regional effect exerted by the lack of a technological breakthrough depends on the effort allocation entailed by the agreement (C&C in the present analysis). The dramatic increase in marginal abatement costs, and thus in the carbon price, negatively affects permit buyers but positively affects permit sellers, as Africa, India and South East Asia. It should also be noted that discounted GDP losses for countries as China, Transition Economies and MENA rise well above 5%.

Figures 11 and 12: Primary Energy Consumption (3.5. W/m2 RF): with or without Backstop Technology
Table 2: Comparison of regional policy costs measured as GDP losses discounted at 5%

Increased spillovers through an international R&D fund

As an illustration consider a global fund which is financed through a share of OECD region's GDP, and provides a subsidy to each non OECD region allocated on an equal per capita basis that adds to their own backstop R&D expenditures. This additionality constraint is imposed because otherwise the optimal reaction of each region to the subsidy would be to cut their own R&D expenditures, i.e. the R&D spending spurred by the subsidy would fully crowd out other domestic R&D investments. The size of the fund is chosen to replicate the investments in energy related R&D that took place during the 80's and it is a tenth of the millennium goal for ODA.

www.unmillenniumproject.org/reports/costs_benefits.htm

The role of additional funding to R&D (on top of R&D investments stimulated by the carbon policy) or of enhanced international spillovers of knowledge is much more limited. In particular, when the target is stringent (as in the 3.5 W/m2 RF scenario) then the carbon signal is high enough to internalize completely the technology externality and stimulates very large investments in R&D both in developed and in developing countries. In particular, the presence of an international market for emission permits and the deriving benefits from selling permits fosters the development of carbon-free technologies throughout the globe. Therefore, policies enhancing knowledge spillovers or increasing investments in R&D (e.g. through OECD countries financing additional R&D investments in developing countries) has little effect. Indeed, the gain in policy costs implied by designing an international fund financing additional energy R&D investments in developing countries or improving their capacity to absorb knowledge, has a negligible effect on stabilization costs. Larger room for improvement is there when we consider a milder target (as the 4.5 W/m2 RF scenario), then an international R&D fund can bring stabilization policy costs down of 5% to 10%, as the much lower carbon signal fails to stimulate the socially optimum level of energy R&D.

No international spillovers

Let us now assume no spillovers take place in any of the carbon free backstop technologies. Let us look for example to the effect on the Non Electric sector backstop price in the 3.5 W/m2 RF scenario. What is immediately visible when comparing this with the basic version case is that the differences in the regional costs become wide and that mostly developing regions will be lagging behind. The spread in regional backstop cost is even more visible when considering a 4.5 W/m2 RF scenario. To make up for the absence of spillovers from other countries investments in R&D and actual deployment of the technology, all countries start to invest earlier in backstop R&D and early investments are undertaken also in non OECD countries as well as shown in figures 13 - 16. The effect on policy costs is limited (a 5% and 2% increase in costs for the 3.5 W/m2 RF and 4.5 W/m2 RF scenarios, respectively). This is mainly due to the fact that OECD countries, which need the technology in the shorter term, rely little on international spillovers in both cases, and mainly benefit from intra regional standing on shoulders and learning by doing effects, triggered by the carbon price. On the other hand, several non OECD countries benefit from larger carbon prices because they are carbon permit sellers.

Figures 13 and 14: Price of non-electric sector backstop technology with and without spillovers. 3.5 W/m2 RF scenario
Figures 15 and 16: Price of non-electric sector backstop technology with and without spillovers. 4.5 W/m2 RF scenario
Figure 17: R&D investment cost with and without spillovers. 4.5 W/m2 RF scenario

Policy Insights

The major insights from this sensitivity analysis hinge on the role of technological breakthroughs and on the role of international knowledge spillovers.

The major factor affecting policy costs, independently on the stringency of the target, is whether a major breakthrough in the non electricity sector takes place; indeed, if R&D is successful in bringing down the costs of a carbon free alternative to oil in the transport sector, then policy costs are decreased crucially. Conversely, costs might increase by 60 up to 80% if R&D programs fail in bringing down costs of alternative technologies in the non electricity sector.

On the other hand, the role of additional funding to R&D (on top of R&D investments stimulated by the carbon policy) or of enhanced international spillovers of knowledge is much more limited. In particular, when the target is stringent (as in the 3.5 RF scenario) then the carbon signal is high enough to internalize completely the technology externality. Therefore, policies enhancing knowledge spillovers or increasing investments in R&D (e.g. through OECD countries financing additional R&D investments in developing countries) has little effect. Indeed, the gain in policy costs implied by designing an international fund financing additional energy R&D investments in developing countries or improving their capacity to absorb knowledge, has a negligible effect on stabilization costs. Some more room for improvement is there when we consider a milder target (as the 4.5 RF scenario), then an international R&D fund can bring stabilization policy costs down of 5%, as the much lower carbon signal fails to stimulate the socially optimum level of energy R&D.

As we have seen, the basic model version assumes substantial technological transfers among countries, hence in addition to enhanced spillovers scenarios we have also seen what would be the effect of switching off completely spillovers, both in learning by doing and in learning by researching. What we found is that costs would not be substantially affected, although a different distribution of the R&D effort, both in time and through regions would be required. However, countries that would mostly be negatively affected by the absence of spillovers would anyway develop the backstop technology later in time and are allocated most of emission rights.