What About Emissions?

Study shows green industrial policy is not enough—but it can be the right place to start.

By Wei Peng, Princeton University

For years, climate policy debates have been organized around two broad families of instruments. Economists have emphasized policies that impose system-wide costs on emissions, with carbon pricing as the canonical example. Governments facing political constraints have often favored incentive-based policies — subsidies, tax credits, and targeted investment — that lower the cost of clean technologies and build supportive constituencies. The first family imposes visible costs; the second creates concentrated benefits. That asymmetry helps explain why incentive-based policies often come first.

The real policy question is how to combine and sequence these instruments over time on a credible pathway to deep decarbonization. Historical evidence alone cannot fully answer this question since few countries have yet achieved the scale of system-level transformation required for deep decarbonization. Forward-looking energy system modeling can help by exploring plausible future policy pathways and identifying key tradeoffs.

Below, we describe insights from a modeling exercise that compares stylized future policy pathways to achieve deep decarbonization. It uses subsidies on one side and an economy-wide carbon price on the other as analytical proxies for two distinct functions: lowering the cost of clean technologies through “carrots” and forcing system-wide adjustment across incumbent activities through “sticks”. Our goal is to explore what happens when policy begins with the first function and delays the second, while aiming to achieve the same long-term decarbonization targets.

The modeling results suggest that starting with industrial policy can work — but only if the carrot phase is durable, and only if the shift to system-wide constraints is not delayed too long. These insights hold across different levels of decarbonization ambition, whether measured by cumulative or annual emissions targets, and are especially pronounced under stringent goals such as net-zero by mid-century. We should also note, there are many places that use both “carrots” and “sticks” and a wide field of research about the interaction and impact of using those policies in combination. This study does not consider using those policy combinations and instead focuses on using them in sequence.

There are, of course, important caveats when translating insights from stylized modeling into real-world policy debates. Notably, the exercise here focuses on one important consideration of climate policy: emissions reduction and the cost of decarbonization. In the real world, green industrial policy and decarbonization strategies are also motivated by many broader objectives — energy security, supply chain resilience, regional industrial capacity, jobs, and innovation — that sit outside this exercise. The modeling results are therefore best understood as one input to the climate question; a more comprehensive evaluation of green industrial policy needs to consider a wider range of societal impacts and tradeoffs.

Industrial Policy Carrots Can Deliver Early Emissions Cuts, but Not Long-Term Deep Decarbonization

Subsidy-driven industrial policy can substantially reduce emissions in the near term. In our modeling, widespread clean-energy subsidies reduce US energy-system CO₂ emissions by 32% in 2030 relative to 2015, a finding consistent with other modeling studies. When designed well and applied consistently, this kind of policy can achieve meaningful near-term emissions reductions on its own.

But subsidies on their own do not get us to deep decarbonization by mid-century in our core scenarios. Their effects are strongest early, when they close cost gaps, accelerate deployment, and scale clean technologies. Over time, the marginal impact of additional subsidy spending declines: as technologies mature, each extra subsidy dollar does less to shift outcomes. And because subsidies make clean alternatives cheaper without making fossil alternatives more expensive, they do not directly compress activity in the incumbent system. In industrial strategy terms, this kind of policy is effective at scaling new industries but weak at displacing incumbent ones. Reaching deep decarbonization on a credible trajectory requires complementary policies that impose system-wide constraints on emissions.

A Carrots-First Strategy Can Work, but Only if the Shift to Sticks Does Not Come Too Late

Starting with carrots can deliver two climate policy benefits at once: greater political feasibility and meaningful near-term emissions reductions. That makes the timing of the transition to sticks a central design question: if governments begin with industrial policy, how long can they wait before imposing stronger sticks?

Our modeling suggests that some delay is manageable. For example, compared to a stick-only pathway that reaches 80% decarbonization by 2050, a carrot-first scenario requires a carbon price that is only 11% higher in 2050 to achieve the same cumulative emissions, as long as the transition to sticks occurs within 10 years. This increase in carbon price is even smaller under less stringent decarbonization targets. This increase is not trivial, but it suggests that a decade of industrial policy can be a reasonable tradeoff if it helps build political support and technological capacity.

But the costs rise sharply when the transition is delayed too long. In our slow-stick scenario, where governments rely on carrots for twenty years before introducing sticks, the carbon price required in 2050 can be as much as 40% higher than in the stick-only case. The delayed transition also leads to much greater reliance on carbon dioxide removal (CDR) technologies later in the century, pushing more of the decarbonization burden onto large-scale deployment of CDR, an uncertain and still-nascent set of technology options. In effect, postponing sticks does not remove the need for them; it makes the eventual transition steeper and more expensive.

This is why sequencing is more than a theoretical concern. A subsidies-first approach can work, but only if policymakers treat the incentive phase as the first stage of a longer strategy and actively prepare for the introduction of system-wide constraints. The point is not that governments must impose a carbon price immediately or fail. It is that delay shifts the adjustment costs into the future, where they are harder to absorb both economically and politically.

Carrots Build Green Industries, but They Do Not Automatically Weaken Fossil Incumbents

Can early subsidies make a later transition easier? The model speaks to one part of this question: the technology-cost and deployment dynamic. Subsidies accelerate the scale-up of renewable electricity, electric vehicles, and other clean technologies, and drive cost declines through learning-by-doing — especially under optimistic assumptions about how quickly investors and consumers respond. Lower clean-technology costs do reduce the eventual cost of the adjustment that system-wide constraints have to drive.

Based on the projected scale of technology investments and deployment, our modeling also shows that subsidies do not substantially weaken fossil incumbents. Compared with a stick-only scenario where system-wide constraints are doing the work, subsidy-driven pathways maintain high gas and oil use in many parts of the system: they make clean alternatives more attractive but do not reduce overall energy demand or force a phaseout of fossil activities. Put differently, subsidies change relative cost competitiveness at the margin; they do not force system-wide substitution.

This is a useful caution against an optimistic theory of change for green industrial strategy: that scaling clean industries will, on its own, transform the political economy of climate action. Falling clean-technology costs and growing clean-energy sectors do change the landscape—they reduce some sources of resistance and create new constituencies. But the modeling suggests these effects do not automatically displace incumbent activities, and the political economy still has to be built deliberately. If system-wide constraints are delayed for too long, their eventual size of adjustments may intensify opposition rather than reduce it.

The IRA Is Best Understood as an Incomplete Sequence

To bring the modeling closer to recent policy experience, consider the US case. The Inflation Reduction Act and the subsequent rollbacks are sometimes read as evidence that industrial policy has failed. The modeling here points to a different reading: the underlying issue is not the use of industrial policy, but that it was not embedded in a durable, multi-stage strategy.

In the model, a carrots-first strategy works only when two conditions hold: the subsidy phase is durable enough to support investment and deployment, and it is followed by a credible shift to sticks. The weakest policy pathway in our analysis combines inconsistent carrots with a delayed shift to sticks. In the near term, inconsistency already reduces the impact of industrial policy: intermittent carrots lower US energy-system CO₂ emissions by 24% in 2030 relative to 2015, compared with 32% under consistent carrots. That weaker start creates a larger long-term problem. Because fewer emissions are reduced early on, more decarbonization must happen later. If policymakers still aim for the same cumulative emissions outcome, the required carbon price in 2050 is 67% higher than in the stick-only pathway. But that kind of late adjustment may not be politically realistic. A more plausible response is weaker ambition. In essence, inconsistent carrots either make the later transition much more abrupt or lead policymakers to accept a weaker long-term decarbonization outcome.

That is the real warning from the IRA era. Industrial policy is not broken. But introducing it is only the beginning. Without durability in the incentive phase and a credible path to broader system-wide constraints later, governments bear the fiscal and political costs of industrial policy while capturing only a fraction of its long-term decarbonization potential. Policy durability, in other words, is not just a political concern — it directly shapes investment trajectories and emissions outcomes.

What the Next Round of Industrial Policy Should Do Differently

For policymakers designing the next phase of green industrial strategy, the first lesson is to stop treating incentive-based policy and system-wide constraints on emissions as rival answers to the same problem. They do different jobs, at different stages of the transition, and a durable decarbonization strategy will likely need both.

Incentive-based policy mobilizes investment, accelerates deployment, and builds constituencies around emerging technologies. System-wide constraints push the whole system — especially incumbent fossil-fuel activities — toward deeper structural change. The value of sequencing is that it combines the political advantages of the first with the structural force of the second. They are complements that operate at different stages of the transition.

The modeling points to four practical lessons. First, build the sequencing logic in from the outset: industrial policy should be the opening phase of a broader transition, not a permanent substitute for stronger measures later on. Second, strengthen policy durability—stop-start incentives weaken investor confidence, slow deployment, and erode emissions outcomes. Third, do not assume that scaling clean industries will automatically shift the politics of incumbent activities; without constraints on fossil expansion, industrial policy risks scaling clean supply alongside persistent fossil demand. Fourth, build the political coalition for the next phase deliberately, by making the benefits of the incentive phase visible early.

As noted at the outset, the analysis here focuses on emissions and the cost of decarbonization. It does not include the broader rationales often associated with green industrial policy—job creation, supply-chain resilience, energy security, and regional development. Those rationales matter on their own terms, and they may also broaden the political support that makes the next phase of decarbonization possible. Whether they do so is ultimately an empirical and political question rather than a modeling one. What this analysis can say is narrower but useful: the success of a subsidies-first strategy depends on whether it creates the conditions—economic and political—for the harder transition that follows.