---
title: effective altruists invented the impact certificate
canonical_url: https://ensurance.app/guide/effective-altruists-invented-the-impact-certificate
markdown_url: https://ensurance.app/guide/effective-altruists-invented-the-impact-certificate.md
subtitle: "it pays people back for good already done. a certificate for a place does a different job: it says where the money goes now"
category: nature-finance
---

# effective altruists invented the impact certificate

*it pays people back for good already done. a certificate for a place does a different job: it says where the money goes now*

Effective altruism has had a certificate of its own since 2014. It was run with real money, and the people who ran it published what they learned. Their experience is a test anyone can put to a certificate of any kind.

**What are impact certificates?** An impact certificate is a transferable claim on the credit for a piece of good work. Whoever does or backs the work holds it and can sell it later to a funder who judges that the work mattered. The sale repays the early backer, so a risky project can be financed before anyone knows how it turns out.

A river, an elk herd's winter range and the wetlands under a migration route keep working whether or not anyone issues a certificate on them. The tests in this series can be run on the money that reaches one of them at a named place. [**ensurance**](/?from=guide) is how that money gets an account, a public record and a route to the place. It is not the place. Its certificate is a cousin of the impact certificate and does a different job.

:::johnson
**one pays for the past. the other funds the present.**

An impact certificate lets a later funder repay whoever backed good work early. A certificate for a place names the account its money goes to today.

[see the certificates that name a place →](/specific?from=guide)
:::

![photo by Rose Li (@roselibre) on unsplash](https://images.unsplash.com/photo-1792557828531-595a6093015f?ixid=M3w4NjQwMjZ8MHwxfHNlYXJjaHw3fHxzbm93JTIwZ2Vlc2UlMjBmbG9jayUyMHdldGxhbmR8ZW58MHwwfHx8MTc5NTE0MjU4MXww&ixlib=rb-4.1.0&w=1200&fit=crop&auto=format)

## how impact certificates work

A funder has to choose projects before the results exist, and prediction is the hard part of the job. Judging a finished project is easier than forecasting an unfinished one.

An impact certificate moves the funder's decision to the point where it is easiest. Someone starts a project and issues a certificate for the impact it will have. An investor who believes in the project buys a share of that certificate, which pays for the work up front. If the project turns out well, a philanthropic funder buys the certificate afterward at a price that reflects what the result is worth to it. The investor is repaid, with a gain if the bet was a good one.

The thing sold is credit. Whoever holds the certificate is entitled to count the impact as theirs, and the seller gives that up.

## where the idea came from

Paul Christiano proposed certificates of impact in 2014. In 2015 he ran the idea as The Impact Purchase with Katja Grace. Manifund later ran impact-certificate rounds for ACX Grants and other programs through early 2024.

Manifund's retrospective on those rounds, in May 2024, says the team was "less excited by impact certs than before". Its review of 2023 reports that certificates were $45,000 of the $2.06 million it sent that year, about 2%.

Retroactive public goods funding is the onchain cousin, in which a funder pays for a public good after it has shown its use; Optimism ran it as Retro Funding and, as reported on its governance forum, paused it after its seventh season.

## what the idea gets right

It pays for results. A funder who buys after the fact pays for what happened, and the evaluation rests on evidence instead of a forecast.

It puts the risk with the people who chose it. An investor who backs a long shot carries the loss if it fails and is repaid if it works. That is a way to do hits-based giving while the funder pays only for the hits.

## where it stalled

Manifund's own account is that investor interest was hard to attract, and that it had not found a case where certificates led to better funding decisions. Two more difficulties show in the design, as we read it.

The first is the final buyer. Every price along the chain is a guess about what one funder will pay at the end. If that funder has not committed in advance, the investor carries two risks at once: that the project fails, and that it succeeds and nobody buys.

The final buyer also faces the question this series keeps asking. The work has already happened, so buying the credit changes nothing about it. What the purchase can change is the next investor's confidence, and that takes a buyer who returns year after year.

The second is resale. Each certificate is a different project, so each has few possible buyers and no price to look up.

None of this makes the idea wrong. It has been tried at small scale by people who reported their numbers.

## an impact certificate beside a certificate for a place

On 2026-11-20 there were 26 ensurance certificates. This post describes four of them: one place, the [Arno River](/arno-river.basin?from=guide) through Florence, and three purposes that cross many places, [wildlife corridors](/wildlife-corridor.syndicate?from=guide), [elk range](/elk.syndicate?from=guide) and [bird flyways](/flyway.syndicate?from=guide).

Read the five row labels as questions. They work on any instrument that calls itself a certificate, ours included.

| | impact certificate | a certificate for a place |
|---|---|---|
| What is sold | A claim on the credit for a piece of impact, usually one project | A unit of a certificate that names one account |
| Who is paid, and when | The people who did or backed the work. Investors pay first and a funder pays last, often after the work is done | The named account, when the certificate is minted. For the four certificates above, the mint price goes to that account |
| What the holder has | The transferable right to count the impact as theirs, and whatever a later funder will pay for it | An entry on the account's public funding record, and a pro-rata share of distributions the protocol is designed to pay. No distribution has been paid |
| Who the last buyer is | A philanthropic funder. If none appears, the investor is not repaid | None is needed for the money to reach the account. A holder who wants to sell still needs a buyer, and we do not claim one is waiting |
| What can be checked | The project's record, and the judgment of the funder who bought it | The address the certificate pays and every payment into it, on a public ledger. Nobody verifies outcomes at the place for the certificate yet |

A certificate for a place is not a claim on credit for impact. Holding one does not entitle anyone to say a river runs cleaner or a herd is larger because of them. It is not an offset or a credit. [Every credit is trying to become money](/guide/every-credit-is-trying-to-become-money?from=guide) covers how a credit, which is counted once and retired, differs from an instrument that is held.

In this series' terms, minting one from a taxable account or a wallet is a **funded position**: the money goes to the named account and the holder is on the record. A grant from a donor-advised fund cannot buy the donor a certificate; [what you actually hold after the grant](/guide/what-you-actually-hold-after-the-grant?from=guide) has that rule, and [a gift, a position or an investment](/guide/a-gift-a-position-or-an-investment?from=guide) sorts the pockets.

## why the difference matters

An impact certificate answers a funder's problem, which is how to pay for good work without having to predict it. A living place has an earlier problem. Without an account kept for it, money meant for a river lands in someone else's budget. A certificate for a place routes money to that account now, while the river is still doing its work, and the account stays open for whoever comes next.

Each instrument lacks what the other has. The impact certificate carries an evaluation, because a funder looked at a result and paid for it. A certificate for a place carries none. It shows where the money went and when. In the terms of [would it have happened anyway?](/guide/would-it-have-happened-anyway?from=guide), that record lets a stranger check the capital layer. It is not evidence of impact.

The certificate for a place does not solve the final-buyer problem. It never poses it, because the account is paid at the first sale, and in exchange nobody is repaid for past work. The holder's risk sits elsewhere: the distributions are a design, and the condition of a place is reported by its stewards without third-party verification.

The two could meet. A funder who wants to pay for results could pay into a place's account after an independent check shows the place held or improved. That is a design, set out in [holds or improves](/guide/holds-or-improves-outcome-payments-into-the-place?from=guide), and no reading has moved money yet.

## where this stands

The protocol is pre-seed. On 2026-11-20 it had 2,095 accounts, 26 certificates, 196 coins and 12 funded release schedules, and the live numbers are at `/api/protocol/stats`. The certificates sit on a public ledger, used here as a record anyone can read.

## read next

The five rows in the table need nothing from us. The certificates are listed at [specific ensurance](/specific?from=guide), each with the address it pays. The next post is [the moral circle has no bank account](/guide/the-moral-circle-has-no-bank-account?from=guide).

## frequently asked questions

### what are impact certificates?

Impact certificates are transferable claims on the credit for a piece of good work. A funder buys one, often after the fact, so that the people who backed the work early can be repaid.

### do impact certificates work?

They have run at small scale. Manifund sent $45,000 of $2.06 million through certificates in 2023, and wrote in 2024 that investor interest was hard to attract.

### is an ensurance certificate an impact certificate?

No. An impact certificate sells credit for work already done so early backers can be repaid, while an ensurance certificate names the account its money goes to now and carries no claim on credit for any outcome.

## sources

[EA Forum, certificates of impact](https://forum.effectivealtruism.org/topics/certificates-of-impact) — the definition; proposed by Paul Christiano in 2014; run as The Impact Purchase with Katja Grace in 2015. Read 2026-11-20.

[Manifund, "Q1 retro: learnings from impact certs", EA Forum, 2024-05-01](https://forum.effectivealtruism.org/posts/onFtsHPEpFNXxq9CN) — impact-certificate rounds for ACX Grants and others through early 2024; the team "less excited by impact certs than before"; investor interest hard to attract. Read 2026-11-20.

[Manifund, "2023 in review", 2024-01-18](https://www.lesswrong.com/posts/cxjRjs5BWw5pBHZiA/manifund-2023-in-review) — $2.06 million sent to projects in 2023, $45,000 of it to impact certificates. Read 2026-11-20.

[Holden Karnofsky, "Hits-based giving", 2016](https://coefficientgiving.org/research/hits-based-giving/) — accepting a high chance of failure for a small chance of a very large result. Read 2026-11-20.

[Optimism governance forum](https://gov.optimism.io) — Retro Funding through its seventh season, as reported there. Read from a search summary on 2026-11-20, not from the full page.

[26 U.S.C. §4967](https://www.law.cornell.edu/uscode/text/26/4967) — the excise tax on a donor-advised fund grant that gives the donor more than an incidental benefit. Read 2026-11-20.

[ensurance protocol stats](https://ensurance.app/api/protocol/stats) — 2,095 accounts, 26 certificates, 196 coins, 12 funded release schedules. Read 2026-11-20.

[ensurance certificates](https://ensurance.app/api/specific/tokens) — the 26 certificates, what each names and the address each pays. Read 2026-11-20.

Account pages read 2026-11-20 (each returned a live page): [arno-river.basin](https://ensurance.app/arno-river.basin), [wildlife-corridor.syndicate](https://ensurance.app/wildlife-corridor.syndicate), [elk.syndicate](https://ensurance.app/elk.syndicate), [flyway.syndicate](https://ensurance.app/flyway.syndicate).

## the series

1. [the test most impact investing fails](/guide/the-test-most-impact-investing-fails?from=guide)
2. [importance, tractability, neglectedness: scoring a living place](/guide/importance-tractability-neglectedness-for-a-living-place?from=guide)
3. [would it have happened anyway?](/guide/would-it-have-happened-anyway?from=guide)
4. [give now, give later, or fund something that keeps receiving](/guide/give-now-or-give-later?from=guide)
5. [effective altruists invented the impact certificate](/guide/effective-altruists-invented-the-impact-certificate?from=guide)
6. [the moral circle has no bank account](/guide/the-moral-circle-has-no-bank-account?from=guide)
7. [the risk factor nobody funds](/guide/the-risk-factor-nobody-funds?from=guide)
8. [a gift, a position or an investment](/guide/a-gift-a-position-or-an-investment?from=guide)

The giving side of the same question: [the gift isn't finished when the deduction posts](/guide/the-gift-isnt-finished-when-the-deduction-posts?from=guide).
