Does ESG divestment cut emissions?
- Robin Powell
- 11 minutes ago
- 10 min read
Sustainable funds move money away from the companies doing the most environmental damage and towards the ones doing the least. A revised working paper by two US finance academics puts numbers on a distinction hidden inside that idea, and finds that two versions of ESG divestment can move global emissions in opposite directions.
By the middle of July, 2026 had become the first year on record in which the UK reached 35°C in May, June and July, according to the Met Office. Weather like that is why a lot of people want their savings pointed somewhere useful. Whether the most popular way of doing that works is a narrower question, and a harder one.
The promise behind ESG divestment is simple enough to print on a factsheet. Buy the clean companies, avoid the dirty ones, and capital drains away from the polluters. Their cost of raising money rises, their expansion plans get harder to finance, and emissions fall. Much of the case for exclusion-based sustainable investing rests on that reasoning.
Samuel Hartzmark of Boston College and Kelly Shue of Yale School of Management have put numbers on a distinction that mostly goes unmentioned. In a July 2026 working paper issued by the National Bureau of Economic Research, they separate two versions of ESG divestment that tend to get treated as one strategy: avoiding a high-emitting industry altogether, and staying in that industry while holding the cleaner companies within it. On a factsheet, both look like money moving from dirty to clean. In the paper's calibrations, they don't do the same thing. One brings emissions down. The other, under the assumptions the authors regard as more relevant, pushes them up.
The two versions of the same strategy
Across-industry divestment is the version of ESG divestment most people picture. Capital leaves energy, materials and heavy manufacturing and moves into technology, finance and healthcare. Within-industry tilting is narrower. It stays in the high-emitting industry and holds the cleaner companies inside it: the better cement maker rather than the worse one.
Hartzmark and Shue model both in a stylised world containing one brown firm and one green firm, each with $35 billion in annual sales. The brown firm emits 1,309 tonnes of carbon for every million dollars of sales; the green firm emits five. A sustainable strategy opens up a one-percentage-point gap between what the two pay for capital.
Run the across-industry version using an industry-level output elasticity of −0.2 — the figure the authors say is the more relevant one for a strategy applied to a whole industry — and total emissions rise by 83,720 tonnes. Run the within-industry tilt with the same gap and they fall by 18,326 tonnes.
These are back-of-the-envelope calculations for two imaginary firms, not forecasts for the world economy, and the authors present them as such. They also say that characterising an optimal sustainable investing strategy is beyond the paper's scope.
The gap between the two versions comes down to how easily customers can switch. Within an industry they switch readily. If one cement maker's costs rise, buyers go to a competitor and the firm sells less. Across industries, switching is much harder. Cement and software aren't substitutes, so making cement more expensive to finance barely dents how much cement gets made. Output stays roughly where it was, and the offsetting benefit that divestment relies on never arrives.
The authors' conclusion is that 'tilting capital allocation within industries is more likely to reduce global emissions than reallocating capital across industries'. The prescription isn't new. Alex Edmans of London Business School, Doron Levit and Jan Schneemeier made a version of the argument in a 2022 paper on socially responsible divestment, which Hartzmark and Shue cite. What's new is the arithmetic, and the finding that the sign can flip.
Why higher capital costs can make brown firms browner
The mechanism runs through the discount rate. Make a company's capital more expensive and you raise the rate at which it discounts future cash flows, which makes money arriving soon worth more than money arriving later. Cleaner production usually means new equipment: a large bill now and a payback measured in years. Carrying on as before pays sooner. Expensive money tilts a manager towards the dirtier option.
That matters more at some companies than others, because brown and green firms are not in the same league. Comparing the top and bottom quintiles of US-listed companies on scope 1 and 2 emissions scaled by revenue, the average brown firm's emissions intensity is 257 times the average green firm's. A large percentage improvement at a green company is a rounding error next to a small one at a brown company.
Hartzmark and Shue call the relationship they are measuring the impact elasticity: the change in a firm's environmental impact caused by a change in its cost of capital. In their data it is asymmetric. A one percentage point rise in a brown firm's implied cost of capital is associated with 6.2 more tonnes of emissions per million dollars of sales, and specifications using the cost of debt and the weighted average cost of capital give four to five tonnes.
Green firms show no statistically significant response in either direction. An insurer or a software company has very little left to decarbonise, so cheaper capital buys almost no environmental improvement.
The estimates come from S&P Trucost scope 1 and 2 emissions data scaled by revenue, covering US-listed companies from 2002 to 2020, with firms sorted each year into quintiles on the previous year's emissions intensity. Rather than isolate cost-of-capital shocks caused by sustainable investing specifically, the authors test five different sources of shock, from equity returns to the share of long-term debt falling due during the 2007 credit crisis. All point the same way.
The paper is an NBER working paper, number 35519, and has not been peer-reviewed. The sample also stops in 2020, so it says nothing about the years since.
What sustainable funds and indices actually hold
The same paper turns to what sustainable funds own. Hartzmark and Shue compare aggregate sustainable fund holdings against a value-weighted market portfolio. Industries with high emissions are significantly underweighted and low-emitting industries tend to be overweighted, which is what the strategy says on the tin.
The second half of that analysis is the part that doesn't follow. Repeating the exercise for the 20 per cent of firms with the lowest emissions intensity within each industry and year, the greenest fifth of companies inside high-emitting industries are underweighted too, relative to their market capitalisation. These are the firms doing best at precisely what sustainable investors say they want. The same pattern shows up in popular sustainability benchmark indices, not only in active fund holdings.
Part of the explanation the authors offer is definitional. Many products that describe themselves as sector-adjusted work with very broad sector definitions. MSCI's sustainable indexes adjust using 11 GICS sectors, and consumer staples — one of the 11 — contains both agriculture and drug retail, businesses no customer would treat as substitutes for one another. This is US holdings data, and like the rest of the paper it stops in 2020.
What happens when a company sells the problem
The same logic shows up one level down, in the real economy rather than in portfolios. Here it isn't investors selling shares but companies selling factories, under pressure from their own shareholders and regulators.
Ran Duchin of Boston College, Janet Gao of Georgetown University and Qiping Xu of the University of Illinois Urbana-Champaign matched 888 divestitures involving 1,105 plants between 2000 and 2020 against the US Environmental Protection Agency's Toxic Release Inventory. Their subject is toxic chemical release, not greenhouse gases. The study was published in the Journal of Finance. It asks what happens to a polluting plant after it changes hands.
No detectable improvement follows. Pollution at divested plants doesn't fall relative to comparable plants that stayed put, and the estimates are statistically indistinguishable from zero. The specifications can detect effects of around 2 to 3 per cent of the sample standard deviation, so this is not a null produced by a blunt instrument.
What changes is the seller. Environmental ratings rise by 160 per cent of the sample mean. Sellers become roughly 5 percentage points less likely to receive an EPA enforcement action. The share price rises on the announcement, and the dirtier the plant, the better the reaction: an interquartile increase in the pollution of the plant being sold is worth another 3 to 4 percentage points of announcement return, against a sample average of 2.5 percentage points.
The buyers tend to be firms facing less scrutiny, and often ones that already do business with the seller. Duchin, Gao and Xu describe the result as 'a cosmetic redrawing of the firm's boundaries'.
The case that ESG divestment does nothing at all
Everything above assumes that ESG divestment moves a company's cost of capital in the first place. There is peer-reviewed work saying it barely moves it at all.
Jonathan Berk of Stanford University and Jules van Binsbergen of Wharton published a model in the Journal of Financial Economics in 2025 putting the effect of ESG divestment on the cost of capital at 0.44 basis points, or less than half a hundredth of a percentage point. Their reasoning is that clean and dirty stocks are close substitutes, with a correlation of 0.93 between the two portfolios, and that investors who refuse to hold the dirty ones control around 2 per cent of wealth. When almost everyone will buy what you're selling, and it's nearly interchangeable with what they already own, the price barely has to move to clear the trade. 'A difference of half a basis point cannot meaningfully affect the capital budgeting decision', they write.
For the effect to reach a full percentage point, on their model, 84 per cent of investors would have to hold only clean stocks.
The force of the paper is its model rather than its data. The authors' empirical test of index inclusions and exclusions produced estimates that were not statistically different from zero, and they note there is close to a 50 per cent chance of estimating the wrong sign. That is an absence of a detectable effect, not a demonstrated one of zero.
So the two papers disagree about whether the divestment lever is attached to anything. They agree about what to reach for instead. Berk and van Binsbergen conclude that investors who want impact should buy and use their rights of control rather than sell, which is close to where Hartzmark and Shue end up.
What the evidence points to instead
Hartzmark and Shue are explicit that their critique targets one strategy rather than the field as a whole, and say so in the paper's closing pages. They name engagement, and transition funds that reward credible improvers, as approaches their argument does not reach.
There is evidence that engagement does something, with a condition attached. It comes from the research TEBI looked at in June on whether ESG funds change corporate behaviour.
Michelle Lowry of Drexel University, Pingle Wang and Kelsey Wei of the University of Texas at Dallas, writing in the Review of Financial Studies, compared actively managed US equity ESG funds that held much the same stocks but had different financial incentives to push the companies they owned. The measure of incentive is how much of a rise in a portfolio company's value would flow through to the manager's own fees. The sample runs from January 2013 to December 2020.
Funds with that incentive behave differently. They research environmental and social issues more, hold on through bad news rather than selling, and vote independently of proxy advisers. Analysts from those fund families are 54 per cent more likely to raise an environmental or social question on an earnings call, against an unconditional likelihood of 6.5 per cent. After an exogenous shock to fund flows in 2016, companies those funds overweighted recorded a statistically significant fall in Trucost onsite emissions releases, a narrower measure than carbon intensity. Companies overweighted by the other ESG funds recorded no significant change.
The catch is in how commitment was identified. It is a measure the researchers constructed from fee economics and validated against proprietary Morningstar and CDP data. It is not a label, a rating, or anything printed on a factsheet. From the outside, an investor cannot tell which fund is which.
What the UK rulebook already asks for
UK investors have had a partial answer to that since July 2024, when firms became able to use the FCA's sustainability labels. There are four: Sustainability Focus, Sustainability Improvers, Sustainability Impact and Sustainability Mixed Goals. The second is the one that maps onto the approach Hartzmark and Shue's calibrations leave open. Improvers exists for assets that are not sustainable now but can credibly become so.
What the FCA asks of an Improvers fund tracks the research closely. At least 70 per cent of the gross value of the product's assets must be invested in line with its sustainability objective, measured against a standard that is robust, evidence-based and, in the regulator's phrase, 'absolute (as opposed to relative)'. The fund must hold evidence that its assets have the potential to meet that standard over time. The FCA's own example of good practice is a defence and security company with a net-zero target validated by the Science Based Targets initiative, named accountability for hitting it, and annual disclosure of progress.
The fund must also have a clear escalation plan for assets that are not progressing sufficiently. The regulator's stated poor practice is continuing to engage with companies that are not improving, with no timeframe for a response and no next steps if none comes.
The absolute standard cuts against the simplest version of within-industry tilting. A fund cannot claim the label for holding the cleanest company in a dirty industry. Being better than the neighbours is not the test.
Whether funds meet any of this is another matter. The FCA reports that in its authorisations work it has not always been clear whether or how firms meet the labelling requirements, or whether disclosures accurately reflect what a fund invests in.
Still, the escalation plan is one of the few things in this entire argument that a fund has to write down and an investor can go and read. It points at a harder test than the one on the factsheet. Exclusion is easy to display. Improvement is harder to prove.
Resources
Berk, J. B., & van Binsbergen, J. H. (2025). The impact of impact investing. Journal of Financial Economics, 164, 103972.
Duchin, R., Gao, J., & Xu, Q. (2025). Sustainability or greenwashing: Evidence from the asset market for industrial pollution. The Journal of Finance, 80(2), 699–754.
Edmans, A., Levit, D., & Schneemeier, J. (2022). Socially responsible divestment (CEPR Discussion Paper No. 17262). Centre for Economic Policy Research.
Financial Conduct Authority. (2026). Sustainability Disclosure Requirements labels: Good and poor practice.
Hartzmark, S. M., & Shue, K. (2026). Counterproductive sustainable investing: The impact elasticity of brown and green firms (NBER Working Paper No. 35519). National Bureau of Economic Research.
Lowry, M., Wang, P., & Wei, K. D. (2026). Are all ESG funds created equal? Only some funds are committed. The Review of Financial Studies, 39(1), 79–113.
Checking what a fund is actually doing
For readers who have reached the end of this piece wondering whether the sustainable fund they hold does what its factsheet implies, TEBI's Find an Adviser directory is a place to start. Everyone listed has publicly committed to evidence-based investing, which includes being straight with clients about what the research supports and what it does not.
For readers who would rather work through it themselves first, How to Fund the Life You Want by Robin Powell and Jonathan Hollow sets out how to weigh costs, evidence and the claims funds make about themselves. Bloomsbury published the second edition, written for UK readers. Buy it on Amazon.
