How robust is the evidence on carbon pricing?

Carbon pricing is effective in reducing emissions while having limited negative economic effects. However, researchers and policymakers should be aware of several methodological issues that may reduce the reliability of the evidence on carbon pricing, writes Kasper Vrolijk.

Most economists agree that carbon emissions are most efficiently reduced by carbon pricing and as a result, close to 61 carbon pricing initiatives have been introduced or scheduled by governments around the world. Given that climate change is accelerating and materialising into large-scale environmental and economic damages, it is important to understand the effectiveness of such marked-based policies. Further, because carbon pricing may create substantial shocks throughout the economy, a comprehensive understanding of the economic effects of such mechanisms is necessary in order to design relevant economic policies and institutions.

Yet, there remain three challenges that limit our understanding of carbon pricing effects. First, a large part of the evidence is based on economic models that predict, rather than evaluate, the effect of carbon pricing policies. Second, of the empirical studies that evaluate carbon pricing, there are large differences in findings, depending on the empirical design and framing of carbon pricing. Lastly, the empirical methods that are frequently used, such as cross-sectional regressions, do not always enable causal relations to be inferred between carbon pricing and environmental and economic outcomes.

Fortunately, there has been a wave of new empirical research that provides useful evidence. This research utilises innovative quasi-experimental designs to examine the economic and environmental effects of carbon pricing. By exploiting changes in policy over time and within a specific geographical area, these studies can examine whether a causal relationship exists between carbon pricing, and emissions and economic outcomes. According to a review I recently conducted, this evidence suggests that carbon pricing is able to reduce carbon emissions without significantly affecting economic outcomes.

Carbon pricing effects

Most quasi-experimental studies consider either the emissions-related or the economic effects of carbon pricing. For example, a study on Sweden found that carbon taxation reduced emissions in the transport sector by 11 per cent on average per year between 1990 and 2005. In British Columbia, the carbon tax had no effect on aggregate emissions, but resulted in a reduction in transport emissions of 5 per cent. In Germany, taxation on electricity had no substantial effects on firm turnover, exports, value added, investment or employment.

Some studies have also examined the combined emissions and economic effects of carbon pricing, although the evidence is limited. In the United Kingdom, for example, a carbon tax on electricity consumption was found to reduce energy intensity by 18 per cent and electricity use by 23 per cent, while it did not negatively affect plant employment, revenue, productivity or exit.

Similar evidence is found on cap-and-trade policies. In California, emissions at plants that were included in a emissions trading policy fell on average by 20 per cent. In Germany, the EU emissions trading system (ETS) was found to contribute to a reduction in emissions of 20 per cent – although only during the second phase of the scheme – and it did not reduce firm employment, turnover or exports.

However, for cap-and-trade policy, the evidence is contradicting: in Norway the ETS did not have an effect on emissions, but boosted firms’ value-added and productivity, while in France it reduced emissions by 15 per cent and employment by 7 per cent, although in both cases these effects were only observed in Phase II.

Methodological shortcomings in the evidence

Despite this new evidence, some studies are characterised by methodological shortcomings, which makes the evidence unreliable.

There are three main issues. First, in many studies, parallel policies (such as renewable energy or subsidy programmes) may affect the carbon pricing effect that is measured, but not every study reports whether such policies existed at the time of study and how they may have influenced results.

Second, many studies use the difference-in-differences method, which compares firms affected by a carbon pricing policy with others that are not, but few studies give sufficient consideration to the problematic issue of auto-correlation. This arises when comparing changes in emissions-related or economic outcomes over time and these changes are correlated with one another, meaning that the researcher does not measure the true effect of carbon pricing.

Lastly, in some studies, results are affected by ‘treatment spillovers’, which means that a firm that was not covered by a carbon pricing policy may indirectly be affected by a firm that was. However, studies infrequently report whether these spillovers could have taken place or use available approaches to deal with this in the studies.

Lessons for policymakers

Quasi-experimental designs can provide useful insights about carbon pricing effects, but researchers and policymakers should be careful in how they use these studies when informing policy and should carefully consider their methodological limitations. 

Generally, policymakers should be aware of the benefits and limitations of quasi-experimental designs and how they may affect the validity of the results. Also, as with other studies that gather evidence from specific contexts in time or space, it is relevant to examine to what extent the localised evaluations of interventions can be generalised to scale.

The good news is that there is considerable scope for cross-learning across studies to address these methodological issues. Furthermore, a series of recent quasi-experimental studies provide innovative approaches in measuring carbon pricing effects, such as using prospective policies (rather than actual policies) to study effects and new methods that enable the study of carbon pricing effects even where there is limited variation in the data.

Kasper Vrolijk is a Research Fellow at the German Development Institute. His paper Quasi-Experimental Evidence on Carbon Pricing was published in March 2021.

The views in this commentary are those of the author and do not necessarily represent those of the Grantham Research Institute.