Articles & Media

In Part 1, I revisited some of the key findings from the influential study by Lara Cushing and Rachel Morello-Frosch et al. (Carbon trading, co-pollutants, and environmental equity: Evidence from California’s cap-and-trade program (2011-2015))  In this Part 2, I provide some extended details on the assessments presented in Part I.

Assessing the key findings of the original study

In evaluating research findings, it is important to consider:

  • What conclusions or interpretations can be reliably drawn from the data, including cause and effect and predictive power;
  • The state of the system as it was studied, and changes that may limit applying conclusions broadly and beyond the time frame of the study;
  • How these conclusions relate to other findings of similar systems and more recent data;

Effect on populations

Understanding the impacts on people is a fundamental, and perhaps the single most important, outcome to understand.  The key findings from the original study concentrate on average facility outcomes and the differences in median characteristics of census block groups (BGs), with little said about net or average exposure weighted by the population in proximity to these facilities

The only data presented on net population impacts shows a net of nearly 2.3 million people experiencing GHG decreasing over the timeframe examined (as reproduced in the table below).  The authors do not present this level of net population detail by population characteristic or for other pollution types.

Figure 1: Clip of original study Table 3 concerning full population characteristics sorted by GHG trends of nearby facilities

While very limited conclusions about net population exposure to air pollution can be drawn from the data presented in the original study, the findings are suggestive that all population characteristics were more likely to experience decreases in GHG pollution exposure from facilities in years 2013-2015 (see Part 1, Figure 3).  The data are also suggestive that white people had a higher likelihood than non-white people of experiencing decreases.  While the net effects or differences between different population groups cannot be determined with statistical significance using the available data, the results do call into question whether any net benefits or areas of increased harm have been distributed equitably.

In terms of demographic exposure, the data is unambiguous on two facts for the time-period considered:

  1. Communities that experienced increases in local GHG emissions were more likely to have higher share of POC than those who experience a decrease in local GHG emissions (part 1, Figure 1), and;
  2. Communities near facilities, whether emissions increased or decreased, were disproportionately POC or otherwise disadvantaged (part 1, Figure 4).[1]

Without more data on both average population and changes in co-pollutants, it is not possible to determine additional population-level, statistically-robust conclusions beyond this.  Therefore, it is essential that any update of the prior study include this perspective across all years for which data is available and for each pollutant measured.

Facility performance

The paper reports that 52% of facilities considered reported emission increases between 2011-12 and 2013-15.  The applicability of this observation is particularly influenced by boundary years, including:

  • Remarkably low emissions in the starting year, 2011, a year heavily impacted by the great recession, and a notable decline in emissions starting in 2016 (part 1, Figure 5)
  • The largest and one of the fastest emissions growing sectors in the study, power sector emissions were skewed by low availability of zero-carbon resources, including below average hydropower availability and the unexpected shutdown in 2012 of the San Onofre Nuclear Power Plant. Power plants made up the largest share of facilities (83 out of 322) examined and showed the second highest average increase in emissions for the time-period examined (Figure 2).

Figure 2: Original study figure 2 showing average and range of emissions outcomes across all facilities and for each sector, with highlighting of electricity generation sector.

  • No indication of statistical significance, which is normally included in papers when statistical significance was found. Statistical significance would be expected to include the influence of the choice of years and the potential outlier data from any of those years.
  • No indication that these findings are predictive of future results or inherent to the program. The authors stated in the original study that:

Our results indicate that, thus far, California’s cap-and-trade program has not yielded improvements in environmental equity with respect to health-damaging co-pollutant emissions.  This could change, however, as the cap on GHG emissions is gradually lowered in the future. 

The authors further note that lower GHG emissions in 2015 could suggest “that greater reductions may be achieved going forward as the cap is lowered further”, but caution that the impact of banking of excess allowances and the use of offset credits could undermine future reductions.  Another factor, as the authors note, is revenue-use:

ongoing investments of a significant portion of California’s cap-and-trade revenue in disadvantaged communities as mandated by law [18] to mitigate climate change could also potentially incentivize deeper local GHG and co-pollutant reductions in the future.

Key findings in the context of previous studies of emissions trading programs

The findings of the original study, examining the first three years of a longer-term program, are a valuable contribution to the growing body of research and should be viewed in that context.  The authors reviewed studies of other emissions trading programs, observing “little evidence that they produced socially inequitable outcomes,”  based upon:

studies of the US Acid Rain Program to reduce sulfur dioxide emissions from coal-fired power plants and of Southern California’s Regional Clean Air Incentives Market (RECLAIM) program to reduce NOx and SOx emissions from large facilities such as power plants, refineries, and manufacturing facilities found no evidence that the locations of emissions or purchases of allowances were disparate with respect to the racial/ethnic makeup or income of surrounding neighborhoods [34-36]. One exception is an analysis that incorporated dispersion modelling of emissions and found that high-income neighborhoods benefited more from RECLAIM than did low-income neighborhoods and that, conditional on income, African American individuals benefited more, and Hispanic individuals benefited less, than white individuals.

This suggests that emissions trading programs are not inherently socially inequitable, but that the design and implementation details are very important.

Updated and additional data

  • In the years following the original study’s timeframe, evidence has grown suggesting substantial improvements in program performance. This includes the aggregate additive impact of investments made with cap auction revenue, and more recent data from 2016-18 showing a greater than 10% reduction in covered facility GHG emissions from 2013-15.  This high-level data illustrates that an update to the original study is very important to understand program impacts, and that understanding the various driving causes of these changes is challenging but essential.

Figure 3: California covered facility greenhouse gas emissions, with box highlighting recent years not covered by the original Cushing and Morello-Frosch et al. study

  • A more recent paper looking at 5 additional years of data (2016-2017 during the program and 2008-2010 prior to the program) found “novel causal evidence that California’s GHG C&T program has reduced the EJ gap in NOx, SOx, PM2.5, and PM10 following its 2013 introduction.” After climbing prior to 2013, the EJ Gap dropped by 21-30% during the first five years of the program.[2] These trends depend on accounting for pollution transport by winds.


[1] For example, the median and median range of the population located within 2.5 miles of at least one regulated facility is 71% POC (44% to 91% “interquartile range” or IQR) compared to 53% (30% to 80%) beyond that range (see original study Table 1).  Within the population located near facilities, the median demographic was 68% (42%-90%) for GHG decreasing facilities and 79% (50%-75%) POC for GHG increasing facilities.

[2] Hernandez D & Meng KC, Do environmental markets cause environmental injustice? Evidence from California’s carbon market emissions data 2012-2017 (2020)