Frequently asked questions

General questions

How often will the NGFS scenarios be updated?

We are currently exploring options for ensuring regular updates.

Do the NGFS scenarios reflect the economic, energy and emission impacts of COVID-19?

The NGFS scenarios do include these effects. We updated policy and baseline GDP trajectories by using IMF projections from October 2021 for 2025 GDP values. SSP2 growth rates are assumed for the rest of the century meaning that the new GDP projections do not catch up to the original absolute SSP2 values. Scenarios also reflect the capacity build-up until 2020 and baseline energy demand has been updated to include the effects of COVID-19. The resulting impacts on emissions are endogenously computed by the models.

Where can I find the documentation of scenarios and models available?

The technical documentation of the scenarios and models used to generate them is available at this URL, including links to further information:

How can I integrate these data into my risk assessment analysis?

The NGFS has prepared a guide which is available here:

Which scenario is the most realistic and likely to occur?

Future developments are inherently uncertain and impossible to predict. For instance, nobody had anticipated the rapid and widespread impacts of the COVID-19 pandemic and subsequent socio-economic and policy responses.

Scenario analysis is one approach to tackle this uncertainty. On the one hand, the NGFS scenarios provide plausible future developments, because they are constructed with models designed to simulate the complex and non-linear dynamics of the energy, economy and climate systems. On the other hand, they account for various possible policy and technology assumptions. Therefore they allow a rich exploration of various plausible future developments and an understanding of the trade-offs between various policy and technology choices.

For instance, the Current Policies scenario (Hot house world) projects a frozen policy implementation scenario, so it will likely underestimate the change of transition brought about by future new legislation. The other Hot house world scenario, NDCs, projects current targets (mostly for the year 2030) to be achieved. Some states are currently not on track for meeting their targets, so this might be an optimistic assumption in some cases. Nonetheless, the current targets have mostly been formulated in 2015, and ongoing technological development in some cases has already overtaken projected targets, so some targets might also be strengthened. In fact, the Paris Agreement foresees a periodic strengthening of ambition to occur every 5 years. The 2022 scenarios account for NDC target published by the UNFCCC until March 2022.

What are the differences between the IEA scenarios and the NGFS scenarios?

There are a number of features that distinguish the NGFS from the IEA scenarios.

First, there are more NGFS scenarios. There are 6 NGFS scenarios produced by 3 IAMs that span a wider range of policy and technological dimensions, including exploration of disorderly scenarios. In contrast, there are 3 main IEA scenarios in the World Energy Outlook 2020: Current Policies scenario (“a baseline picture of how global energy markets would evolve if governments make no changes to their existing policies and measures”), Stated Policies scenario (“a detailed sense of the direction in which today’s policy ambitions would take the energy sector”) and Sustainable Development Scenario (“a way to meet sustainable energy goals in full, requiring rapid and widespread changes across all parts of the energy system”). However, the IEA scenarios can be roughly mapped to the NGFS scenarios. The Current Policies scenario correspond to each other. The IEA Stated Policies scenario corresponds to the NGFS NDCs scenario. The Sustainable Development scenario is roughly in line with the orderly Below 2 °C scenario, while the NGFS Net Zero 2050 scenario corresponds to the new IEA scenarios NZE published in May 2021.

Second, the NGFS offers 2 sets of scenarios that have been created using the REMIND-MAgPIE that integrates the macro-economic climate damages into the optimization procedure.

Third, the NGFS modelling framework includes 3 models whereas the IEA uses a single model (i.e. the World Energy model). This allows for exploring the uncertainty related to model structures and techno-economic (and potentially other) assumptions.

Fourth, while the temporal resolution of both the IEA and NGFS scenarios is in the order of 5 to 10 years, there are at least 25 regions represented in the World Energy model (up to 120 regions in the oil and gas module) and 12 to 32 in the IAMs used to generated the NGFS scenarios.

Fifth, these large-scale energy-economy models represent sectors that are relevant from an energy viewpoint, i.e. coal, oil and gas, power, buildings, industry and transportation. Each sector is represented with a different level of detail depending on the model.

Sixth, most IAMs cover land use (e.g. agriculture, forest) and its related emissions. For the recent IEA NZE scenario, the IEA linked their model to the GLOBIOM land use model. However, the data is not available yet for other scenarios.

Finally, the NGFS IAMs have been linked to a macro-economic model (NiGEM) to extend the macro-economic information.

Has the NGFS nominated representative scenarios in the same way as in Phase I?

Because we had a larger number of scenarios in Phase I, we designated some of them as “representative” to help users select scenarios. From Phase II onwards, we decided to focus on the scenarios that are most relevant for central banks and supervisors and reduced the total number of scenarios. Given this focus, we no longer need to differentiate between representative and alternate scenarios.

Why is there no “too little, too late” scenario?

You will see that this category in the scenario space has not yet been filled with scenarios. This quadrant is less researched and requires the development of novel scenarios. This category of scenarios would require rather specific assumptions which is not well generalizable. The lack of existing academic literature on scenarios that fit to this category meant that other areas were prioritized so far.

To what extent can the NGFS scenarios be used by various non-financial sector companies to conduct their own scenario analysis?

In terms of transition risks, all sectors can use the scenarios to analyse to which extent their own current energy consumption and inputs (both energy and land-use products) might require transformation in different climate scenarios, and how existing decarbonisation/net-zero plans they might have align with different targets. The Integrated Assessment Models used for the NGFS scenarios provide a number of results for industry sub-sectors like cement, steel and chemicals.

Transition scenarios / IAMs

General questions

In which year are variables in the NGFS scenarios computed endogenously in the models (i.e. starting value)?

Models have different calibration years, mostly 2005 or 2015. All policy scenarios are however assumed to follow (closely) the trajectory of the Current Policies scenario until 2020 so that stringent mitigation is only happening in the first time step thereafter, i.e. 2025.

Do the models consider (irreversible) climate impacts and positive feedback loops where warming leads to more GHG emissions (e.g. methane emissions from melting of permafrost)?

Most transition scenarios do not incorporate climate impacts, neither in terms of GDP reductions due to climate damages, nor in terms of feedback effects. However, the 2022 scenarios include a full sets of scenarios for two levels of damage created using a special version of the REMIND model that integrate macro-economic climate damage and responds endogeneously to them. Tipping points and impacts from extreme weather events are not accounted for. The scenarios are included in the NGFS Scenario Explorer.

Are the scenario data enough to carry out financial risk analysis? Should they be supplemented with results from additional models (e.g. short-term economic/financial models)? If so, which models would be better suited (e.g. agent-based models, DSGE, structural macro-models)?

For many financial applications, supplementing the scenario data with additional data and modelling will be necessary. This is a nascent field both in terms of application and research, so no recommendation can be given yet. However, the Bank of England and the Banque de France/ACPR have already designed frameworks to complement the output of integrated assessment models and enhance the uptake and usability of NGFS scenarios.

Socio-economic assumptions

Do the NGFS scenarios rely on or use all the SSPs or just SSP2?

All NGFS scenarios are currently based on SSP2 socio-economic assumptions. The sensitivity of the scenarios to alternative socio-economic assumptions could be explored in future work, depending on needs.

Do models include a dietary change, e.g. meat consumption per capita developments?

The scenarios assume an evolution of dietary change in line with historic trends, so increasing meat consumption with rising affluence. Dedicated lifestyle changes as a mitigation option are not considered in these scenarios.

Will the NGFS provide scenarios that cover the potential for de-growth (i.e. non-classical models) and the depletion of fossil fuels?

The current set of scenarios does not consider such potentialities, but future extension might explore broader sets of socio-economic futures (e.g. SSP1 etc.).

Do the NGFS scenarios also consider the social dimension of climate change, such as social polarisation or climate migration?

The NGFS scenarios do not model those. Most IAMs are working on incorporating social dimensions gradually, but those efforts are at an initial stage.

Why are the NGFS scenarios produced with three different integrated assessment models?

This allows users to obtain a range of results across models, thus capturing modelling uncertainty to some degree, and allowing users to draw robust insights across models. Furthermore, different models have different properties which make them more suitable for different applications; for example, GCAM offers the highest regional granularity, MESSAGE and REMIND are giving optimal benchmark scenarios, GCAM and REMIND offer higher detail for the industry sector.

Will the input files for the integrated assessment models be made available so that the results can be reproduced and additional model outputs not included in the IIASA database can be accessed?

The integrated assessment model codes are openly available, so the modelling can be transparently reviewed. Fully replicating the NGFS scenarios would require the input data, part of which is not openly available due to licencing. Please contact us if you are missing specific variables; we can consider requests on a case-by-case basis.

Policy assumptions

Which policies are considered under the scenario “Current Policies”?

The Current Policies scenario includes a stylised representation of existing policies. Most of this work is based on the climate policy database established as part of the CD-Links projects (McCollum et al. 2018, Roelfsema et al. 2020). The scenarios reflect the status of the database from the end-of 2019. Please note, however, that in all models only a subset of policies reported in the database can be represented due to the granularity of the modelling, and some policies are also captured via proxies, e.g. carbon prices or overall final energy reductions to represent efficiency policies.

Spatial and temporal aspects

How granular are modelling results in terms of regions and sectors? Where can I find the list of countries included in each region?

Regional granularity differs between the participating models. The MESSAGEix-GLOBIOM model and REMIND-MAgPIE have 12 model regions, whereas the GCAM model has 32. The regional definitions are summarised in Table A1.1, Table A1.2 and Table A1.3 for the individual models and Table 8 for the aggregate regions.

Sectoral granularity also differs, but results are reported in a harmonized way for economic and energy sectors.

Is it possible to get a more granular breakdown of the model pathways, e.g. by country rather than by world regions?

Key economic, energy and emissions variables have been downscaled to about 100 major countries.

Why are data missing for several countries in the NGFS scenarios (e.g. no BRA, JPN, MEX, RUS data from MESSAGEix-GLOBIOM)?

These countries are only represented explicitly in the GCAM model (and partly in the REMIND-MAgPIE model). For these countries, you have to rely on the downscaled model data.

Emissions

How large are cumulative carbon emissions across scenarios?

Cumulative total emissions of CO₂ differ across models, as scenarios were harmonized to arrive at comparable warming levels which depends on all greenhouse gases. Carbon budgets can be calculated from the provided Emissions|CO2 variable in the database.

Why do global net CO₂ emissions in the Net Zero 2050 scenarios not cross the zero line exactly in 2050?

Achieving zero net CO₂ emissions by 2050 is often seen as synonymous with keeping warming below 1.5°C. Indeed, many governments and economic and financial actors have adopted a climate target of zero net CO₂ emissions by 2050 to meet the temperature goals of the Paris Agreement. However, in the three Net Zero 2050 scenarios, global net CO₂ emissions are not exactly at zero by 2050. For instance GCAM projects −300 MtCO₂ whereas REMIND-MAgPIE and MESSAGEix-GLOBIOM feature 2400 and 4300 MtCO₂ respectively.

Why?
The transition scenarios developed in this project are designed to keep warming below a certain threshold (e.g. 1.5°C, 2°C). All Net Zero 2050 scenarios limit warming to 1.5°C above pre-industrial levels. However, this does not necessarily mean that CO₂ emissions must reach exactly net zero by 2050. In fact, the IPCC SR1.5 report found that in “model pathways with no or limited overshoot of 1.5°C, global net anthropogenic CO₂ emissions […] reach net zero around 2050 (2045–2055 interquartile range)”.

Is it still OK to use these scenarios?
It is perfectly fine to use these scenarios as if their net CO₂ emissions reached zero in 2050. Indeed emissions in 2050 are broadly in the range of today’s emission uncertainty levels (i.e. +/− 0.5 GtCO₂ for fossil fuel and industry emissions and +/− 2.6 GtCO₂ for land use emissions, source Global Carbon Project). They could also be compensated by trade of carbon permits between countries. Importantly, the net zero targets announced by several countries (e.g. China, EU, Japan, USA) are specifically included in the Net Zero 2050 scenarios.

Economics

Is GDP computed endogenously or exogenously in the models?

GDP is an exogenous input assumption in the GCAM model, but a semi-endogenous output for REMIND-MAgPIE and MESSAGEix-GLOBIOM. The latter models take the SSP2 GDP trajectories (adjusted for COVID-19) for calibrating assumptions on exogenous productivity improvement rates in a no-policy reference scenario. GDP trajectories in other scenarios thus reflect the general equilibrium effects of constraints and distortions by policies (so changes in capital allocation and prices, but without taking potential damages from climate impacts into account).

Considering that the models are assuming perfect foresight and optimising behaviour, do you think that in the “real world”, carbon prices have to be higher, or GDP effects would be stronger?

Not necessarily, as the models have simplifying assumptions on the functioning of economies that go both ways. Assuming perfect foresight and optimising behaviour indeed leads to a downward bias for carbon prices. On the other hand, assuming economies to be in a full equilibrium without idle capital or work-force, and with given fixed demands for energy services leads to an upward bias for carbon prices. Carbon prices are very sensitive to assumptions about the evolution of demand for food and energy (which in the real world will to some extent be sensitive to climate policies) and to technological development.

Are model prices real or nominal prices?

All model prices are real prices, so the indices are indices for real prices.

How are the particular carbon prices linked to a given carbon budget? Is it related to what the price has to be in order for alternative energy sources to be viable?

Yes, that’s a viable interpretation. The temporal trajectory of carbon prices is iteratively adjusted up or downwards until the defined temperature target/ carbon budget is fulfilled.

When carbon price is assumed to increase under a scenario, that should act as a (tax) wedge between demand and supply; the path for the oil price is reflective of the (gross) price from the consumer perspective or the (net) one from the producers’ angle?

The oil supply is modeled via supply-cost curves, and without explicit representation of the exploration infrastructure. This is a simplifying assumption for the modeling, but does not suggest that oil prices will indeed rise in such scenarios. A more detailed analysis of the oil sector is therefore a useful complement to the use of NGFS scenarios (using the scenario demand information).

Oil prices are reflective of the gross price paid by consumers. The price at the second energy level (i.e. Price|Secondary Energy|Oil) includes the costs of extraction, transport and transformation as well as the carbon tax. The price at the final energy level (i.e. Price|Final Energy|Transportation|Liquids) includes additional transport and distribution costs and other taxes and subsidies.

Even in the 1.5 °C scenarios, oil and gas prices seem to remain relatively high and in some cases increase over the scenario. Could you explain the intuition behind this?

In MESSAGEix-GLOBIOM and REMIND-MAgPIE, primary energy oil and gas prices reflect the dynamics of long-term oil and gas markets in equilibrium. This representation is, of course, different from reality but historically this has served the purposes of our research on the economics of climate change mitigation. More concretely, in a no to little climate action scenario (e.g. Current Policies, NDCs), oil and gas prices increase over time because demand for these fuels keeps rising while marginal production costs increase. Oil and gas prices are lower in 1.5°C and 2°C scenarios compared to Current Policies, but still, increase over time. In 1.5°C and 2°C scenarios, the demand for these fuels does not drop immediately because it is cost-effective to continue to use them for a while (e.g. EVs are more expensive and there is no policy push for them in those scenarios, natural gas can be cheaper than renewables in some regions and when used with CCS can act as a bridge fuel). Oil and gas are also used in the petrochemical industry. These dynamics explain why prices do not drop steeply. Both oil+gas demand and prices would drop more steeply if the policy assumptions do not only foresee a comprehensive carbon price (see also next answer), but also explicit policies to regulate oil and gas use (efficiency standards) or push their competitors (EV subsidies, direct support for industry electrification, change in energy taxation).

Why are oil prices not the same across regions in REMIND?

Trade costs for importing/exporting oil can lead to slight differences in oil prices across regions.

Are annual investments only related to the investments needed for changing the energy system, or does it also include what is needed to uphold/maintain demand from current sources? Like new upstream oil/gas projects.

Annual investments for fossil extraction include the required investments to maintain the production observed in the respective scenario. The investments are estimated using historical investment intensities of extraction, so might be an overestimate in scenarios with declining demand.

What is the correct interpretation of relatively smooth economic growth in face of such dramatic changes in the industry after 2030?

Several aspects of Integrated Assessment Models can explain these results. First they include a simplified representation of the economy. In particular, they assume economic equilibrium and only account for a few economic frictions. Second, these models are designed to find ways to sustain economic growth. When a climate constraint is added (i.e. keeping warming below 2 °C), IAMs switch to low-carbon energy technologies.

There is a significant difference in the GDP impacts from transition risk between the IAM forecasts and the NiGEM forecast. How can we interpret these differences?

The NiGEM scenarios demonstrate that the estimated impact on GDP depends heavily on the fiscal policy assumptions. In the NiGEM Net Zero scenario, 50% of carbon revenue is assumed to be channelled back into the economy in the form of government investment. This raises potential output, offsetting more of the transition and physical costs compared to the IAM scenarios.

What assumptions are made for the role of monetary policy, including any impact on inflation expectations?

For the NiGEM scenarios, central banks target a weighted average of inflation deviations from target and output deviations from target. There is a trade-off, as inflation is generally rising and output falling in the scenarios, so the interest rate response differs depending on the scenario and also on the individual county and their sensitivity to both inflation and output. It is assumed that 50% of the carbon price is passed directly to consumer prices.

GDP losses from physical risks are estimated somewhere between -5% and -15% by 2100, which may not necessarily be considered significant. What would be the best argument to push policy makers that climate change should be taken seriously?

GDP loss estimates are highly uncertain and typically do not include acute risks and indirect damages from societal repercussions, such as conflict. They also often do not account for tipping points in the Earth system. This is the case for the estimates from Kalkuhl and Wenz (2020) used in the scenario release. Uncertainty is only covered for the climate response, not for the choice of damage estimate or covering a broader range of damages.

What do the models assume about the pass-through of carbon cost, e.g., do the demand curves for a given product assume the end user will have to pay for all (or part) of the carbon cost?

The IAMs do not represent different users explicitly, or model explicit products, only energy services (e.g. mobility by mode for transport, floor space for buildings, and certain materials for industry). The demand for energy services is assumed to be fully elastic, so energy price increases lead to reduced demand and/or technology change. The prices of these energy services therefore fully reflect the carbon cost.

Within NIGEM is there any consideration of additional investment needed by firms and consumers in an economic transition (replacing machinery, retrofitting buildings...), and the opportunity costs for consumption from this extra depreciation?

In the Net Zero 2050 and Below 2°C scenarios, we have assumed that carbon taxes generate fiscal revenues which are then partially reinvested into the economy (50% of revenues is channelled into government investment). This crowds in further private investment, resulting in a moderate investment boom. We have not made additional assumptions regarding the depreciation rate or opportunity costs, but there may be some crowding out of private consumption compared to a scenario where carbon revenue is channelled into personal sector transfers or tax relief. In other scenarios, investment is determined endogenously within NiGEM without additional assumptions.

To what extent do the NGFS scenarios incorporate the impact of the pandemic on countries' capacities to finance the transition to a low-carbon emissions economy? Certain countries that are key to reduce emissions were heavily hit.

In order to account for the COVID-19 pandemic and its impact on economic systems and growth, the near-term GDP and final energy demand trajectories have been adjusted to incorporate the impact of the COVID-19 shock. In some cases, this has a permanent impact on the level of potential output. The short term (up to 2023) base files used by NiGEM include the COVID affected numbers from the NIESR forecast of 2021, while the IAM GDP projections include the IMF COVID adjusted GDP and final energy data. The baseline assumptions for the IAM’s and NiGEM are matched by 2025. The baseline values determine the potential for carbon revenue to be raised and channelled into investment, so scenarios are consistent with a post-COVID world.

Energy system

To what extent the models cover the limitations in the roll-out of (new) technologies? E.g. availability of lithium & other minerals for car/home batteries, solar PV?

The models do not endogenously track flows and stocks of minerals and raw materials, but assumptions on overall potentials are informed by studies (including from the life-cycle assessment, LCA community) that have analysed these requirements. Temporary price spikes (similar to the 2008 polysilicon price spike) of these materials are thus not included, but their impact might also be of minor importance on the 5-year time step perspective here.

Are efficiency improvements of solar PV, wave energy, etc. considered or is the roll-out of renewable energy based on current efficiencies?

Continued improvement of technologies in terms of decreasing capital costs is included, which is the biggest effect of efficiency improvements.

Are there any new energy generation technologies that are not included in the current versions of the models but are to be considered for updates?

Yes, the models only take proven technologies into account, so do not consider technologies like fusion reactors, small modular reactors or electric planes. As soon as these technologies are successfully demonstrated, they will be included in the models.

What variables does the “buildings” sector entail? Are there any particular real estate variables in the NGFS scenarios?

When the IAMs provide information at the sectoral level, it will always be focused on the energy-use aspects of that sector. So, final energy demand, energy efficiency, energy prices, as these vary between the sectors.

The buildings sector represents all residential and commercial buildings. One needs to keep in mind that our models are primarily designed to represent the dynamics of energy systems. From the energy demand side, it is useful to decompose energy demand by sectors because each one behaves differently. At the most aggregate level, IAMs divide energy demand into buildings, industry and transportation.

The typical variables available for the buildings sector are energy consumption, energy prices (e.g. electricity, natural gas, oil, coal, biomass). We can also provide estimates of investments in energy efficiency.

The real estate sector is not represented as such in IAMs but the buildings sector could be used as a proxy.

Why is nuclear investment lower in Net Zero 2050 than in Current Policies?

This is an intertemporal and international effect; given that competing technologies (solar, wind and storage) experience higher learning in Net Zero and are needed strongly in the long-term anyway, they also become more competitive than nuclear in some countries, while in the current scenarios policies nuclear is more competitive.

Climate and Energy Policies

How are current policies introduced in the model? Is it always through carbon price?

No, policies are represented in different ways, both via carbon prices but also via sectoral constraints and incentives.

Do the NDC scenarios represent conditional or unconditional NDCs?

They represent conditional NDCs, so provide an optimistic assumption on NDC pathways (all targets, even conditional ones are met in 2030).

Are net zero emissions targets included in the NDC scenario?

They are included in the Net Zero 2050 and Delayed Transition scenarios. For a complete list of the targets, see Table 1.4 in the Technical Documentation.

Why is the NDCs scenario developed with REMIND-MAgPIE so different than that developed with MESSAGEix-GLOBIOM?

The long-term developments beyond 2030 are not clearly defined by the scenario narrative. This explicitly specifies 2030 targets (from countries NDCs), but only defines “comparable ambition levels” for climate policy stringency after 2030, so as to not anticipate overly optimistic (2°C compatible developments after 2030) or pessimistic (full reversion to no-policy baseline) policy stringency after 2030.

Technical aspects

Is optimisation in the models performed at the global or regional level?

This differs across models.

GCAM is a dynamic recursive market equilibrium model and not an optimizing model. That is, no attempt is made to optimize the aggregate economic system over time or at any moment in time. While GCAM’s economic agents are assumed to individually optimize either expected utility or expected profits or minimize expected costs, economic activity is set in a probabilistic context. GCAM’s economic agents undertake investments without knowledge of future outcomes and thus can, and often do, hold unfulfilled expectations about the future.

In REMIND-MAgPIE, optimisation happens at the regional level, and a Nash algorithm ensures market clearance of trade at the global level.

Macro-economic impacts from climate physical risks

Are all chronic physical effects captured through damage functions and effects on GDP?

No. The aggregate empirical damage function captures productivity effects, e.g. from impacts on labour productivity or agricultural productivity, related to temperature shocks. It does not capture the effects from extreme events, sea-level rise, non-market impacts or indirect effects e.g. through conflict.

How are the parameters of the macro-economic damage functions calibrated?

They are based on an empirical analysis of the effects of historic climate on gross regional product with data for more than 1500 regions in 77 countries (Kalkuhl & Wenz 2020). It applies an annual panel model assessing effects on productivity levels and growth for temperature and precipitation. The parameters used are those from the preferred specification of the authors, based on the robustness of the regression, which is specification 5 in Table 4 of the paper, using the results for temperature change and the interaction term including one lag.

In the Phase I scenarios, physical risk damages were projected using three different damage functions, while only one damage function was used for Phase II. What was the rationale behind this decision?

A practical reason for this shift was that additional damage functions would have required running the scenarios with integrated physical damages and NiGEM for each of the damage functions, thus multiplying the number of model outputs by three. For Phase II, we prioritised a streamlined set of scenarios, thus favouring using of a single damage function. The range of damages spanned with this estimation is still considerable, given the uncertainty in temperature response.

Do the scenarios capture the interaction between transition and physical risks?

For Phase III, we produced two additional scenario variants, run with the REMIND integrated assessment model, that include internalized physical damages, so that the transition trajectory is reflecting the social cost of carbon. This variant of the scenarios can be selected separately in the IIASA NGFS data explorer. Also see section 3.2.3 in the technical documentation for more details: