Full edition
Summer ESG Module · Postgraduate

Governance Factors in ESG Investing

The machine that turns what owners want into what a company actually does.
Jae Yung KimUniversity of Exeter29 June 2026
G
Scroll
Prologue

Who actually decides what a company does?

Not the managers. Not whoever owns the most shares. Not the law. Something subtler — a machine.
0

In May 2021, a hedge fund almost no one had heard of did something that, on paper, should have been impossible.

It was called Engine No. 1. It had existed for months. And it owned about one two-thousandth of one percent of ExxonMobil — one of the largest oil companies on earth.

0.02%
On a $10,000 company, that is two dollars. Round it off and it is nothing. A rounding-error shareholder.

On the strength of that vanishing stake, Engine No. 1 nominated four outside directors to Exxon's board, on one argument: the company was destroying its own future by ignoring the energy transition.

A brand-new fund holding 0.02% runs directors against management. How many of the twelve board seats does it win?
commit to a number  ·  none   one   three  ·  or is it hopeless from the start?
What happened

It won three of the twelve seats.

No new law. No boycott. No takeover. A rounding-error shareholder reshaped the board of an oil major with a single ordinary instrument — the shareholder vote. It did not need its own shares to win. It needed everyone else's, and the giant index funds that hold Exxon looked at the case and voted with the insurgent.

Governance is not a third silo beside E and S. It is the machinery that turns preferences into what the firm actually does.

That is the thesis. Environmental and social outcomes are not wished into being — they are produced by a mechanism. Understand the mechanism and you understand when ESG preferences become corporate action, and when they evaporate into press releases.

So we spend our time on the machine: how it is built, how power flows through it, where preferences are carried faithfully, and where they are gamed, rewired, or broken.

Part One

The agency problem

Every tool of governance exists to solve one problem — and the whole field becomes legible once you see it.
1

The people who own the firm are not the people who run it.

Berle & Means named it in 1932 — the separation of ownership and control: ownership dispersed across millions of small holders, control held by managers who own little. Jensen & Meckling priced the gap JFE 1976: whenever an agent acts for a principal, the relationship leaks agency costs — of three kinds.

Monitoring
what owners spend to watch — boards, auditors, analysts.
Bonding
what managers spend to promise — debt, pay put at risk.
Residual loss
the value still destroyed despite both. Never zero.
“Corporate governance deals with the ways suppliers of finance assure themselves of getting a return on their investment.”— Shleifer & Vishny, Journal of Finance 1997

Six mechanisms manage the gap.

1 Boards2 Voting3 Blockholders4 Market for control5 Disclosure6 Incentives

The thread through all six: every one runs on shareholders choosing to use it. The Exxon insurgents reached for two — voting and the market for control — and a mechanism only ever counts for as much as the other owners willing to back it.

PredictBoard diversity · Adams & Ferreira

Should a firm put more women on its board?

A board that is ninety percent male, and a proposal to add several women. The question divides people instantly — so be honest about your own answer first.

Does board gender diversity raise firm performance? And would a binding quota raise or lower the value of a typical firm?
raise   lower   no effect
EvidenceAdams & Ferreira, JFE 2009

Diversity reliably raises monitoring — but not value.

Behaviour → clear
attendance up, more seats on oversight committees, weak CEOs disciplined harder. Boards monitor more.
Value → depends on method
a positive raw correlation that flips negative as the statistics get careful — watch it below.
Figure 1The estimated effect of board gender diversity on firm value as the method gets more careful. A positive raw correlation (OLS) turns negative once firm fixed effects absorb selection, and strongly negative once diversity is instrumented for cause. The point is the change of sign, not the magnitude.

The cheerful correlation was selection — good firms were the ones adding women. Isolate the cause and the average effect is ~0 to negative, turning positive only where governance was weak to begin with. A blanket quota can destroy value in firms that did not lack oversight in the first place.

Verdict
Diversity reliably improves monitoring — but its average causal effect on value is about zero to negative, concentrated in poorly-governed firms.
“More is always better” is a slogan; the evidence replaces it with “it depends on the firm.” Diversity is at once a social goal and a governance-quality question.
PredictExecutive pay · Bertrand & Mullainathan

Are CEOs paid for luck?

An oil company's share price doubles — not from strategy, but because the world price of oil doubled. The CEO did nothing.

Does his pay rise with the windfall anyway — and if so, how much, relative to pay for performance he actually produced?
Figure 2How strongly CEO pay responds to a dollar of genuine performance (“skill”) versus a dollar of pure luck — an oil-price windfall. On the market measure the two are essentially equal (skill 0.38, luck 0.35).
Verdict
CEOs are paid for luck about as much as for skill — and a real monitor strips out exactly the luck.
When a large shareholder sits on the board, pay-for-luck falls by roughly 23–33% while pay for genuine performance is untouched. Pay is partly a symptom of weak governance, not only a cure for it — which is why shareholders later demanded a vote on it.
Part Two

Shareholders & voting: the heart of the machine

Ownership has quietly re-concentrated — not into families, but into a handful of giant institutions. Who holds the votes has become a central fact.
2

The owners with the most votes are precisely the owners who cannot walk away.

The Big Three — BlackRock, Vanguard, State Street — now cast a huge share of U.S. votes, and they are mostly index funds. An index fund holds the market by construction, so it cannot sell its disapproval. Its only lever is voice, not exit.

The worry
tiny fees, gains shared with every rival → why monitor? They should be lazy owners.
The evidence
index-heavy firms get more independent directors, fewer takeover defences — “passive investors, not passive owners.” Appel et al., JFE 2016

One detail to hold for later — eligibility to vote is fixed on a single record date, but shares trade every day. The set of people allowed to vote can differ from the set who own the firm a week later. It looks like a technicality. In Part Four it becomes the whole story.

PredictHedge-fund activism · Brav et al.

When an activist arrives, is value built or stripped?

A fund quietly takes a six-percent stake in an underperforming company, files the disclosure that announces it, and demands change.

Over the following year, does the stock rise as lazy management is disciplined, fall as raiders strip it for parts, or pop and then revert as the hype fades?
Figure 3Cumulative abnormal return around an activist's 13D filing. The stock rises about 7% around the announcement and then holds — it does not give the gain back, the signature of genuine value creation rather than overreaction.
Verdict
On average activism creates value, and the gain sticks: about +7%, no reversal, with real operating and governance gains.
The short-termist asset-stripper is not supported in the broad data — though distributional fights, where gains come partly at the expense of workers or creditors, remain a fair and separate concern.
PredictProxy advisors · Malenko & Shen

How much does ISS really move a vote?

A dispersed owner cannot research the thousands of items it votes on each year, so it buys advice. Two firms dominate, and ISS alone covers some forty thousand meetings. The worry: unelected kingmakers. But you cannot just compare votes ISS opposed with votes it supported — it opposes the troubled proposals to begin with.

ISS flips a say-on-pay recommendation from for to against. By how many points does shareholder support fall — essentially zero, or huge?
Figure 4Support against firm performance, around the cutoff that triggers an ISS recommendation. Two firms either side of the line are near-identical except that one gets a negative recommendation; the jump at the threshold, scaled up, implies a causal effect of roughly 25 points.
Verdict
A negative ISS recommendation causally removes about a quarter of the vote — real power, and bounded at the same time.
Because roughly 80% of funds vote their own custom policy that ISS merely helps implement, “ISS decides everything” is an overstatement. The influence is genuine and limited.
Part Three

Where preferences become action — or don't

E and S reach a company through two channels: engagement behind closed doors, and the public ballot. In both, the quality of the machine decides whether a preference becomes action or a press release.
3
PredictStrategic voting · Michaely et al.

Do “ESG” funds actually vote their values?

A fund that markets itself as an environmental champion, and a climate proposal coming to a vote.

When its vote would genuinely decide the outcome — a close, pivotal vote — does it support the proposal more often than usual, the same, or less?
Figure 5An ESG fund's extra support for environmental and social proposals, relative to its peers. Large (about +22 points) when the vote is not pivotal — a costless signal — and collapsing by roughly 60% when the fund's vote would actually decide the result and impose the cost.
Verdict
ESG funds back proposals when it is costless, and quietly withdraw support precisely when their vote would be decisive.
Average support therefore overstates true commitment — the headline number is padded with non-pivotal yes-votes. Greenwashing shows up not only in brochures but in voting behaviour, which is why behaviour must be read strategically, not at face value.
PredictEngagement · Dimson, Karakaş & Li

Does quiet engagement pay?

The private channel: a large investor pressing management behind closed doors, with no proxy fight.

Does it pay off in the stock always, only when it succeeds, or essentially never? And how often does engagement actually succeed?
Figure 6Abnormal return in the year after an engagement. The payoff is concentrated entirely in the successful cases (~+7%); failed engagements move the stock about zero. Because success is roughly one in five, the average effect is small.
Verdict
Engagement pays — but only when it succeeds: about +7% when it works, essentially zero when it does not, success roughly one time in five.
Active ownership creates value, but it is hard, selective work, not a free lunch. And that +7% is a statement about shareholder value, which is not the same as social welfare.
Part Four

The frontier: shareholder democracy

We have treated the electorate of a vote as fixed. The research frontier dissolves that — and with it, a good deal of our intuition about what a vote even means.
4

People trade before they vote.

Shares change hands right up to the record date, so the set of owners at the moment of the vote is not given from outside — it is chosen in the market. The electorate is endogenous Levit, Malenko & Maug, JF 2024. And it runs in a loop:

expectations→ valuations→ who owns the shares→ the vote→ validates the belief

So outcomes can be self-fulfilling, and a high price need not mean happy owners.

PredictPost-meeting trading · Li, Maug & Schwartz-Ziv

If shares sort by belief around a vote, then once a contentious meeting resolves, the losers should no longer wish to hold a company run against their preferences.

After a contentious meeting, who trades more — the winners or the losers? And in which direction do the losers go?
Figure 7Abnormal trading volume around a shareholder meeting. Volume peaks near 20% above normal and stays elevated for weeks, as funds on the losing side sell out and the shareholder base re-sorts toward agreement — even when prices barely move.
Verdict
After a contested vote the losers sell and the winners stay — the electorate re-sorts, and the base becomes more homogeneous.
Ownership at the moment of the vote is endogenous, and this post-meeting selling is the empirical fingerprint of exactly that mechanism. Once the electorate is endogenous, even “what is a vote worth?” becomes a formal, unsettled question.
Part Five

A live question: dual-class sunsets

We end not with a settled result but an open one — where the machinery of the lecture meets a fast-spreading structure, and the answer is genuinely unknown.
5

When Snap went public it sold shares carrying no votes at all. Founders at Google, Meta and Snap keep voting control while holding only a small slice of the cash flows. The dual-class share forces apart two things that usually travel together — and separating them is the key.

Would you pay more or less for a no-vote share? And, separately: is a founder's voting control good or bad for the value of the firm?
Figure 8Firm value against insiders' stake, split into its two components. More insider cash-flow ownership raises value, then flattens (incentive alignment); more insider voting control lowers it (entrenchment). The gap between them — the wedge — is where the danger lives.

Ownership aligns; control insulates — and the structure is now in nearly 30% of U.S. IPOs.

Value peaks around a 25-point wedge, then falls: a little founder control helps, too much destroys. So the question is not “ban or allow,” but how to make control expire — a sunset. Here the lecturer's own work begins, on a real puzzle:

Combined vote
the founder's super-votes count → he wins by definition. The sunset barely binds.
Majority-of-the-minority
looks like the fix — but he can buy out the dissenters, and bought-out shares leave the protected minority.
Figure 9A sketch from the model of the “sunset contest.” Whether the minority is genuinely protected (foreclosed from capture) or the outcome is up for grabs (contested) depends on how much ownership is locked beyond the founder's reach. Time-based sunsets do not foreclose capture; an ownership-band sunset can.
Can a rule meant to protect the minority end up hollowing it?
The open question is not whether sunsets are a good idea — almost everyone agrees they are — but which design actually disciplines a founder rather than merely appearing to. That is genuinely unsettled, it is where the machinery of this lecture points, and it is the question I put back to you: what do you think?