Probability distribution forecasting of carbon allowance prices: A hybrid model considering multiple influencing factors

Abstract

This study presents a hybrid model for forecasting the conditional probability distribution of carbon allowance prices. We primarily use the singular spectrum analysis method to process the non-stationary signals, and the non-crossing composite quantile regression neural network algorithm to achieve accurate, robust, and realistic quantile forecasts. We also include multiple influencing factors to enhance the forecasting performance. Empirical applications to the China and Europe carbon markets show that the proposed model significantly outperforms other benchmark models in terms of point and density forecasting accuracy. In addition, distribution forecasts can lead to economic gains using a simple switching trading strategy. Our hybrid framework is also useful for risk measurement and management.