📈 economics

Global GHG Emission Dynamics, Regional Convergence, and Cross-Border Emission Spillovers: Evidence from a Hierarchical Dynamic Factor Model

Using a hierarchical Dynamic Factor Model on global EDGAR data from 1970 to 2024, this study reveals that greenhouse gas emissions are driven primarily by structural forces rather than climate policies, exhibit a persistent two-club divergence pattern across regions, and show evidence of significant cross-border spillovers from Europe to North America and Latin America, though causal attribution to carbon leakage remains statistically inconclusive.

Original authors: Aránzazu de Juan Fernández

Published 2026-07-15

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Aránzazu de Juan Fernández

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine the entire planet's atmosphere as a giant, chaotic dance floor where 225 different countries are trying to move in sync. For decades, we've hoped that by signing agreements like the Paris Agreement, everyone would start dancing to the same slow, low-carbon beat. But a new study by Aránzazu de Juan Fernández suggests the music is actually being controlled by a giant, invisible DJ that no single country can turn off.

The Invisible DJ and the Unstoppable Beat
The researchers used a super-sophisticated statistical tool called a "Hierarchical Dynamic Factor Model" to listen to the music of greenhouse gases (CO2, methane, and nitrous oxide) from 1970 all the way to 2024. They found that a single "global factor" is driving the dance for everyone. This invisible DJ explains between 41% and 68% of why emissions go up or down across the whole world.

Here's the twist: The authors looked closely at the timeline to see if major climate milestones—like the Kyoto Protocol in 1997, the Paris Agreement in 2015, or even the pandemic in 2020—changed the DJ's playlist. They found no discernible change. The beat kept rising steadily, driven by structural forces like population growth and industrialization that are too big for any single nation's "Nationally Determined Contribution" to stop. It's as if the DJ is playing a song that gets louder every year, and no matter how many countries promise to dance slower, the volume just keeps going up.

Two Different Dance Floors (The Clubs)
While the whole world is moving to the same loud beat, the dancers are splitting into two distinct groups, or "clubs," that are moving in opposite directions.

Club 1 (The Stuck Movers): This group includes the Former USSR, Latin America, and Sub-Saharan Africa. Their emissions are diverging from the first group, often staying high or rising.
Club 2 (The Fast Movers): This group includes Europe, Asia, the Middle East and North Africa (MENA), and Oceania. They are actually getting better at slowing down their emissions, forming a "declining club."
The Outlier: North America is so unique in its dance moves that it doesn't fit into either club.

The study suggests that while countries within these groups are starting to sync up with each other (converging), the gap between the groups is getting wider. It's like one group of dancers is learning a new, slower routine, while the other group is stuck in a fast-paced, old-school groove, and they aren't merging into one big circle.

The "Leakage" Mystery: Is the Energy Spilling Over?
One of the biggest fears in climate science is "carbon leakage"—the idea that if Europe turns off its lights, those lights just get turned on in North America or Latin America. The authors tested this using a method called "Local Projections."

They found that when Europe reduces its emissions, North America's emissions tend to go up by about 18%, and Latin America's go up by about 15%. This suggests a strong connection, like water leaking from a bucket into a neighboring one. However, the authors are very careful here: they cannot prove that this is definitely caused by companies moving factories to avoid rules. Their statistical "detective work" (using an instrument based on EU carbon prices) wasn't strong enough to pin the blame on a specific cause. The numbers suggest a leak, but the evidence is more of a "strong hunch" than a courtroom conviction.

The "What If" Experiments
To double-check their findings, the authors ran two side experiments:

The Connectedness Test: They measured how much countries "talk" to each other through their emissions. They found that 54% to 60% of the changes in one country's emissions are actually caused by shocks happening somewhere else. The world is incredibly interconnected; you can't change your own emissions without affecting your neighbors.
The Time-Travel Test (Synthetic Control): They asked, "What would have happened if the US hadn't left the Paris Agreement?" or "What if Brazil hadn't reversed its forest protection?"
- USA: When the US left the Paris Agreement in 2017, the study found no detectable change in its emissions compared to what would have happened otherwise. The authors suggest that local forces, like the shale gas boom and state-level rules, were doing all the heavy lifting, not the federal government's signature on a treaty.
- Brazil: When Brazil's government changed in 2019 and stopped protecting the Amazon, the study found a 30% increase in biogenic CO2 (emissions from nature, like trees burning). This was a "marginal" but significant finding, suggesting that specific policy changes in Latin America can have a huge, immediate impact on the planet.

The Bottom Line
The paper doesn't say climate action is useless. Instead, it suggests that the current system of individual country promises isn't strong enough to stop the global DJ from playing the same rising song. The world is splitting into groups that are converging internally but diverging from each other, and the massive interconnectedness of our energy markets means that a reduction in one place often triggers a rise in another. The authors suggest that to truly change the music, we might need a global approach that addresses these structural forces and cross-border leaks all at once, rather than hoping that individual countries can fix the whole dance floor by themselves.

Technical Summary: Global GHG Emission Dynamics, Regional Convergence, and Cross-Border Emission Spillovers

Problem and Motivation
Nearly a decade after the entry into force of the Paris Agreement, a fundamental empirical question remains unresolved: are global emission trajectories converging toward a common low-carbon path, or does the aggregate global trend mask deep, persistent divergences? Furthermore, if specific regions are reducing emissions, is this reduction offset by increases elsewhere via carbon leakage? Existing literature often relies on computable general equilibrium (CGE) models or firm-level studies that may miss macro-level structural dynamics. This paper addresses these questions by analyzing the most comprehensive monthly greenhouse gas dataset available, covering 225 countries, four distinct gases (CO2 fossil, CO2 biogenic, CH4, N2O), and a 55-year period (1970–2024).

Methodology
The study employs a unified hierarchical framework consisting of three primary analytical stages:

Hierarchical Dynamic Factor Model (DFM):
The core empirical strategy decomposes emission variation into three orthogonal layers: a global factor common to all countries and gases, continental factors specific to eight world regions, and idiosyncratic country-level residuals. The model is estimated using quasi-maximum likelihood (Doz et al., 2012) on standardized monthly data. This decomposition allows the authors to separate the aggregate global trend (driven by population and industrialization) from genuine regional divergences in decarbonization efforts.
Convergence Analysis:
Using the DFM-purged continental factors, the paper applies three complementary convergence tests:
- $\sigma$ -convergence: Testing for a monotonic reduction in the cross-regional standard deviation of factors.
- $\beta$ -convergence: Estimating a two-way fixed-effects panel model on country-level log-emissions to determine if higher-emission countries grow more slowly.
- Phillips–Sul Log-t Club Test: An iterative algorithm applied to the continental factors to identify "convergence clubs"—groups of regions sharing a common long-run trajectory—while explicitly removing the confounding influence of the global trend.
Spillover and Policy Evaluation:
- Local Projections (LP): Reduced-form impulse response functions are estimated using purged idiosyncratic regional shocks to measure cross-border emission spillovers (carbon leakage) over a 36-month horizon. A Bartik-style instrumental variable (IV) strategy using EU ETS prices is attempted for causal identification but is noted as underpowered.
- Diebold–Yilmaz Connectedness: A generalized forecast error variance decomposition (FEVD) is applied to annual log-differences to quantify the total interconnectedness of global emissions.
- Synthetic Control Method (SCM): Used to evaluate the causal impact of specific policy events (US withdrawal from Paris in 2017 and Bolsonaro's deforestation reversal in 2019) by constructing counterfactual trajectories for treated countries.

Key Results

Global Factor Dominance: A single global factor explains between 41% (CO2 biogenic) and 68% (N2O) of cross-country emission variance. Crucially, this factor shows no discernible inflection at major climate policy milestones (Kyoto, Copenhagen, Paris) or the 2020 pandemic shock, suggesting the aggregate trajectory is shaped primarily by structural forces beyond the reach of current nationally determined contributions (NDCs).
Club Convergence Structure: While individual countries exhibit accelerating convergence (half-lives shortening from ~5–6 years in the 1970s to ~2–3 years in the Paris era), the regional level reveals a persistent two-club structure for fossil CO2.
- Club 1 (Declining): Europe, MENA, Oceania, and Asia.
- Club 2 (Stagnating/Diverging): Former USSR, Latin America, and Sub-Saharan Africa.
- Outliers: North America is unassignable to either club for fossil CO2 due to its unique policy volatility and energy mix; the Former USSR is a non-converging outlier for the GHG aggregate due to the structural break of the 1991 collapse.
Cross-Border Spillovers: Reduced-form Local Projections find positive spillovers from Europe to North America (+18%) and Latin America (+15%), magnitudes consistent with carbon leakage via energy markets and consumption channels. However, causal attribution is not established; the Bartik-IV strategy is underpowered (FKP = 7.86), and all Anderson–Rubin confidence sets contain zero.
Connectedness and Policy Effects:
- The Diebold–Yilmaz Total Spillover Index is 54–60%, confirming high global interconnectedness.
- Synthetic control analysis finds no detectable domestic emission effect from the US Paris withdrawal (p = 0.75), suggesting US trajectories were driven by market forces and sub-national commitments.
- Conversely, the Bolsonaro administration's deforestation reversal is associated with a marginally significant (+30%, p = 0.06) increase in Brazilian biogenic CO2, corroborating that Latin American emissions respond to idiosyncratic policy shocks.

Significance and Claims
The paper claims to provide the first unified, high-frequency analysis of global emission dynamics at this level of disaggregation. Its primary contributions are:

Empirical Reframing: It documents that the global aggregate trajectory has not been durably altered by major policy milestones, implying that structural forces currently dominate NDCs.
Structural Heterogeneity: It reveals that convergence is occurring within specific regional clubs rather than globally, suggesting the NDC framework may be effective for deepening convergence among already-decarbonizing regions but ineffective at bridging the gap between diverging clubs.
Spillover Evidence: It provides reduced-form evidence of cross-border spillovers consistent with leakage hypotheses, while simultaneously highlighting the difficulty of establishing causal identification at the regional aggregate level.

The authors conclude that the decentralized nature of the Paris Agreement may be better suited to generating within-club convergence among countries with institutional capacity than to shifting the global trajectory or neutralizing cross-border spillovers. They caution that while country-level effects exist, they have not yet been sufficient to generate a detectable break in the global aggregate.

More like this