Climate promises: does losing credibility hurt?
How does uncertainty about climate policy, arising from gaps between governments’ announced climate commitments and their actual credibility and follow-through, affect macroeconomic outcomes? This is the question Fulvia Marotta, Maria Sole Pagliari, and Jasper de Winter attempt to answer in this paper.
The authors build a novel, sentiment-weighted measure of climate policy uncertainty (CPU) for the Netherlands using the full archive of Financieele Dagblad (the leading Dutch financial newspaper), covering over 765,000 articles from 1985 to June 2025. An article is classified as CPU-related if it contains keywords from three simultaneous categories: climate-related terms (e.g., CO₂, renewable energy), policy-related terms (e.g., regulation, carbon tax), and uncertainty-related terms (e.g., risk, unpredictable).
Two key methodological innovations distinguish this index from earlier approaches:
- Normalization by climate policy articles (not total news volume), so the index reflects uncertainty within the climate policy debate rather than shifts in general media activity.
- Sentiment weighting using a Dutch-adapted Loughran–McDonald financial dictionary (1,345 terms), so the index captures whether media coverage frames CPU negatively or positively and not just how frequently it appears.
To quantify macroeconomic effects, the authors estimate a monthly Bayesian Vector Autoregression (BVAR) model (2006–2024) including the CPU index alongside business confidence, economic sentiment, industrial production, private investment, and the AEX stock index.
Findings
- CPU follows the legislative cycle: The index rises sharply when climate bills are submitted to Parliament and declines once they are formally ratified and published in the Government Gazette. This confirms that institutional processes, not just policy content, shape perceived uncertainty.
- The expectations channel drives transmission: CPU shocks reduce economic sentiment by 1.6% and industrial confidence by 1.1% within 3 months. Counterfactual simulations show that blocking the sentiment response attenuates the effect on industrial production by 60% and on stock prices by 30%, confirming that CPU works primarily through coordinated expectation formation rather than direct real-economy disruption.
- Real activity and financial markets contract: A one-standard-deviation CPU shock reduces private investment by 0.24%, industrial production by 0.55%, and the AEX stock index by 2.16% at peak.
- Asymmetric effects: Negative CPU shocks (pessimistic, concern-laden media framing) generate significantly larger and more persistent macroeconomic contractions than positive shocks. Positive shocks produce small, short-lived improvements. This asymmetry is most pronounced for confidence indicators and stock valuations.
- Supply-shock vs. demand-shock: Including consumer prices (HICP) in the model reveals that negative CPU shocks behave like negative supply shocks as output falls while prices rise temporarily, while positive CPU shocks resemble positive demand shocks, with both output and prices rising. This has direct implications for monetary policy: central banks should not respond uniformly to CPU shocks but tailor their reaction to the shock’s nature.
Commitment vs Credibility: Macroeconomic Effects of Climate Policy Uncertainty
Authors: Fulvia Marotta, Maria Sole Pagliari, Jasper de Winter
From: De Nederlandsche Bank
Climate policies: does unpredictability hurt?
This paper by Konstantinos Gavriilidis, Diego R. Känzig, Ramya Raghavan, and James H. Stock attempts to answer a similar question as the Marotta paper above, namely what happens to the economy when governments fail to provide clear, stable, and credible climate policy?
The main difference is that the Marotta et al. paper is motivated by the Netherlands’ specific predicament: a country with legally binding, ambitious climate targets but a persistent track record of implementation delays, political reversals, and credibility gaps. The Gavriilidis et al. paper is motivated by the extreme policy volatility of the U.S. political system, where climate policy has swung sharply between administrations: from Obama-era regulation, to Trump-era rollback, to Biden’s Inflation Reduction Act, to renewed rollback under the second Trump presidency.
Gavriilidis et al. construct a newspaper-based index of U.S. climate policy uncertainty using 7.87 million articles from four major outlets, namely the New York Times, Wall Street Journal, Washington Post, and Los Angeles Times, spanning the mid-1980s to 2025. An article is classified as covering CPU if it simultaneously contains terms from climate change, policy, and uncertainty dictionaries, where the dictionaries were built from a specialist climate policy news corpus and validated through both human audits and an LLM-based classifier (GPT-4o-mini), achieving a precision rate of around 80%.
The paper’s central methodological contribution is a narrative-based instrumental variable. The authors compile a comprehensive record of 146 major U.S. climate policy events, covering legislative actions, presidential decisions, regulatory rulings, and judicial proceedings, and measure the unexpected surge in media coverage around each event. They then purge this media response of any changes in policy stringency (whether the policy tightened or loosened regulation), so that the instrument captures pure uncertainty shifts rather than news about the direction of policy. This instrument is used to estimate dynamic causal effects in a VAR model, with local projections used as a robustness check.
The paper also conducts firm-level analysis using the full universe of listed U.S. companies, exploiting variation in firms’ climate change exposure, as measured from earnings call transcripts, to test whether more exposed firms respond more strongly to CPU shocks.
Findings
- A 50% increase in CPU which is roughly the scale of the swing observed over the 2016–2020 U.S. election cycle, leads to a 0.5% fall in GDP, a nearly 2% decline in private investment, and a 0.2 % rise in unemployment. At the same time, commodity prices rise by around 2.9% and headline consumer prices increase by 0.2%, meaning output falls while prices rise simultaneously. This opposing pattern of quantity and price responses is the signature of a supply shock, and the authors show this sharply distinguishes CPU shocks from standard economic policy uncertainty shocks, which propagate like demand shocks.
- Because CPU acts as a supply shock, it creates a direct dilemma for monetary policy: the central bank cannot simultaneously stabilize output and inflation.
- On emissions, the paper finds no aggregate green paradox: emissions do not rise in the short run but instead fall alongside the decline in economic activity, leaving emissions intensity largely unchanged. However, at the sectoral level, fossil-fuel-related industries do show a short-term increase in investment, consistent with a micro-level green paradox in which firms accelerate brown projects ahead of potential future regulation.
- At the firm level, CPU shocks lead to average declines in sales of around 1%, employment of 0.7%, investment of 2%, and R&D expenditure of 1.6%. These effects are significantly larger for firms with higher climate change exposure. R&D spending falls particularly sharply in fossil-fuel sectors, and total factor productivity declines persistently both at the aggregate and firm level, indicating that CPU exacerbates transition costs through misallocation.
The Macroeconomic Effects of Climate Policy Uncertainty
Authors: Konstantinos Gavriilidis, Diego R. Känzig, Ramya Raghavan, James H. Stock
From: University of Stirling, Northwestern University, Harvard University
