This paper examines how firms' investments in pollution abatement are influenced by financial frictions and policy uncertainty. Our data analyses suggest that financially constrained firms are less likely to invest in pollution abatement and are more likely to release toxic pollutants. Such a pattern is intensified by policy uncertainty measured by close gubernatorial elections or uncertainty revealed in firms' earnings conference calls. We then develop a heterogeneous firm general equilibrium model, in which financially constrained firms face increased marginal costs of finance from pollution abatement. The marginal costs of finance are further amplified by policy uncertainty in environmental regulation, reducing firms' pollution prevention. The aggregate effect of environmental policy may therefore depend on the distribution of financial frictions and policy uncertainty.

This article considers the decision problem of a buyer who purchases emission allowances to compensate for carbon emissions. The buyer decides about the timing and quantity of allowances and must trade-off the overall cost of compliance with the risk of non-compliance. As with any commodity, the buyer faces a number of risks, such as market and credit risk, quantity risk, and regulatory risk. The work is motivated by a collaboration with a large chemical company. Carbon traders use the model to calibrate a purchasing strategy that minimizes cost and matches organizational risk preferences. A backtest on historical price data demonstrates that the proposed approach yields better outcomes inside the risk-reward spectrum than a practitioner benchmark. We formulate the problem as a risk-averse multistage stochastic programming problem that considers forward curve dynamics, transaction cost due to limited liquidity, as well as volumetric uncertainty. We develop a solution approach based on stochastic dual dynamic programming, and show how arbitrage-free scenario lattices can be trained on historical returns. The article contributes to the digitization of commodity procurement by enabling automation of decision processes. Model-driven decision-making thereby replaces manual, human-based analytics and risk management that is often error-prone and time consuming.

This paper proposes text-as-data methods relying on unsupervised machine learning algorithms applied to European Union (EU) law acts and newspapers. These are used to construct two monthly indices over a reference period 1997-2021: (i) First, a news- based index which underlies a conjunctural uncertainty about the international context in which the global energy and environment policy evolves (EnvPU). (ii) Second, a laws- based index which reflects structural changes of the European energy and environment regulations (EnvP). The main findings suggest both indices display, in some extent, a common evolutionary pattern around salient events in the history of the EU energy and environment policy. Moreover, EnvPU index appears to be more volatile and is driven in the short-run by EnvP index. Given the support of such a policy to carbon phase-out, we further examine, in what extent, each index relates to price uncertainty dynamics in fossil energy markets (oil, gas, and coal). As a result, we uncover that, increase in news-based EnvPU index has a positive impact on price uncertainty of all fossil energy markets, the effect being stronger and more significant for gas and coal markets. In contrast, while an exogenous shock in laws-based EnvP index has a negative effect on price uncertainty in oil and gas markets, it tends to increase the coal price uncertainty. Overall, EnvP index depicts a stabilizing effect on fossil energy prices.