NB: authors are listed alphabetically.
NB: authors are listed alphabetically.
A neural network approach to the environmental Kuznets curve
Mikkel Bennedsen, Eric Hillebrand, Sebastian Jensen
Energy Economics, 2023, Volume 126, 106985
We investigate the relationship between per capita gross domestic product and per capita carbon dioxide emissions using national-level panel data for the period 1960-2018. We propose a novel semiparametric panel data methodology that combines country and time fixed effects with a nonparametric neural network regression component. Globally and for the regions OECD and Asia, we find evidence of an inverse U-shaped relationship, often referred to as an environmental Kuznets curve (EKC), in production-based emissions. For OECD, the EKC-shape disappears when using consumption-based emissions data, suggesting the EKC-shape observed for OECD is driven by emissions exports. For Asia, the EKC-shape becomes even more pronounced when using consumption-based emissions data and exhibits an earlier turning point.
Long-Lead Forecasting of El Nino Southern Oscillation Using Score-Driven Models with Many Predictors
Janneke van Brummelen, Sebastian Jensen, Siem Jan Koopman, Desislava Petrova
Conditionally accepted, International Journal of Forecasting, 2026
We consider long-lead forecasting of the El Niño Southern Oscillation climate phenomenon. This is widely seen as a challenging task from a methodological perspective. At the same time, El Niño events are of substantial social importance. We adopt the class of score-driven models to capture key dynamic features of the Niño3.4 index and also include a large set of previously designed predictors. We provide details of our forecasting methodology and assess the forecast accuracy for a range of different model specifications and estimation strategies. We conclude that a model with a good description of the dynamic features in the data is also able to provide accurate forecasts for almost all lead times, up to 18 months ahead. The large set of predictors is shown to support the provision of accurate long-range forecasts for specific lead times.
CO₂ Emissions Projections with Neural Networks
Mikkel Bennedsen, Eric Hillebrand, Sebastian Jensen
We propose a long-short term memory neural network model adapted to regional panels (RP-LSTM) for the relation of per-capita gross domestic product (GDP) and carbon dioxide (CO₂) emissions for a yearly panel of the countries considered in the Shared Socioeconomic Pathways (SSPs). We compare the RP-LSTM model to commonly used econometric panel data models for the environmental Kuznets curve in a pseudo out-of-sample forecast exercise and find that it has better predictive power. We use the model to generate future CO₂ emissions projections using GDP projections from the SSP scenarios. In a comparison with the corresponding emissions projections from integrated assessment models (IAMs), we find that the RP-LSTM model projects lower emissions compared to IAM baseline scenarios and higher emissions compared to IAM mitigation scenarios. This reflects the training of the RP-LSTM model on historical data that contain all dynamics in play in reality, including both technological progress and climate policy. The RP-LSTM model projections do not vary across the different SSP scenarios to the same degree as the IAMs do, but they align with the mitigation challenges mapped out in the different SSPs. We advocate the use of the model as a benchmark for IAMs that has a higher degree of empirical realism but a lower degree of scenario resolution.
Neural Networks for Nonlinear Regression with Serially Correlated Disturbances: Evidence from Cloud Cover
Sebastian Jensen, Siem Jan Koopman
We propose a new treatment of nonlinear regression with serially correlated disturbances that incorporates autoregressive moving average structures into feedforward neural networks. The resulting model provides an alternative to modeling temporal dependence using lagged variables. In simulations, the proposed method accurately recovers regression functions of varying complexity and the underlying error dynamics across a range of time-series lengths and signal-to-noise ratios. Finite-sample properties and out-of-sample predictive performances are shown to be robust to model misspecification induced by omitted lagged variables and incorrect specification of the error dynamics. Cloud cover is an important factor in climate projections. In an empirical study of cloud cover prediction for a grid of locations within and around the Mediterranean Sea, our proposed model yields more accurate predictions than existing methods, including long short-term memory networks. Serially correlated disturbances in place of lagged variables improve predictive accuracy across a range of land and ocean environments. Improvements over linear models with serially correlated disturbances are particularly pronounced in mountain areas, consistent with the presence of stronger nonlinear effects in cloud formation in such regions.
Mapping Instability in Sequential Nowcasting
Sebastian Jensen
Evidence on the Climate–Growth Relationship: A Panel‑Based Neural Network Approach
Mikkel Bennedsen, Eric Hillebrand, Sebastian Jensen, Marius Just
Nonlinear Long-Term ENSO Prediction
Janneke van Brummelen, Sebastian Jensen, Siem Jan Koopman, Desislava Petrova
The Conditional Akaike Information Criterion for Time-varying Parameter Models
Sebastian Jensen, Siem Jan Koopman, Jan van den Brakel
Use of Machine Learning in Climate Econometrics
Sebastian Jensen
This dissertation consists of three self-contained chapters on the use of machine learning in climate econometrics and is particularly concerned with how tools and ideas from the fields of econometrics and machine learning can be combined to shed new light on the relationship between macroeconomic activity and carbon dioxide (CO2) emissions.