Measuring community health through community tokens

One of the key innovations that blockchain technology enables is the ability to easily create ventures that are fully owned (and operated) by their communities, through a community token. In most cases, ownership of the community token doesn’t just provide access to the economic upside of the venture but is also the primary mechanism to influence the most critical decisions: on-chain governance decisions assign a community member a certain number of votes that’s proportional to the number of community tokens that they hold.

Therefore, as I’ve argued before, understanding the patterns of community ownership and how it evolves over time is critical to maintaining the health of the community in the long run. Which brings us to today’s question - how?

As I initially suggested, a good starting point might be the Gini coefficient, a measure of wealth inequality in a nation or social group. The Gini coefficient is relatively easy to calculate, and relatively easy to understand - its values vary between 0 and 1, where 0 is the most equal distribution - all community members hold exactly the same number of community tokens; 1 is the most unequal distribution - a single community member holds all the tokens. Since it’s a normalized coefficient, it’s also easy to compare across different points in time and across different communities.

However, the Gini coefficient is not without its flaws, and Vitalik being Vitalik wrote a wonderful piece highlighting a key challenge with the Gini coefficient - its actionability.

He argued that it mashes together two separate community challenges:

  • Community suffering due to lack of resources - illustrated in Dystopia A where half of the community has no tokens at all and half of the community equally shares all tokens.
  • Community suffering due to concentration of power - illustrated in Dystopia B where one person holds half the tokens and the rest of the community equally shares the remaining tokens.

Both Dystopia A and Dystopia B will have the same Gini coefficient of 0.5…


Vitalik went deeper than this, reminding us that since a community token is only a portion of a person’s wealth, the number of tokens they own is not just a measure of the resources that are available to them, but also a measure of their interest in that particular community. Which certainly invites a deeper reflection on the moralistic implications of community token inequality, and perhaps an area for future research exploring how different funding sources (DAO treasuries, DEX, etc.) affect token ownership distribution. Measuring token ownership distribution is a dimension of community health, but looking at community health just through that lens is rather myopic.

Nonetheless, we want to overcome the bundling of dystopia A and dystopia B challenge and Vitalik proposed a few indices that do just that:

So I went ahead and did just that. I’ve built a Dune Analytics dashboard that calculates the various indices for a given ERC-20 community token. You can find it here.

Here’s how the different indices look like for the Bankless DAO community token (BANK):

BANK indices
BANK indices

And for the Braintrust community token (BTRST):

BTRST indices
BTRST indices

Along the way, I’ve come across some unique challenges in measuring these indices for community tokens. Some I was able to overcome. The rest are areas for further research:

  1. Dust: the exchangeable nature of different community tokens means that even people who’ve opted to leave their community and sell all their community tokens may still be left with some “dust”, a small fractional position that they can’t get rid of. If we count them in our analysis - it’ll likely skew the results. Overcoming this challenge is relatively straightforward by imposing a minimum number of community tokens that a wallet address must own in order to be counted in the analysis. Note, however, that this threshold may vary greatly from one community to the next, while holding 0.1 BANK is negligible, holding 0.1 of BTC isn’t. The dashboard above accounts for that and allows the user to set their own threshold for the token they are exploring.
  2. 1 wallet address != 1 person: this feature of Web3 can be broken down into two separate challenges. The first challenge is “one wallet, many people”: a smart contract or a multisig-treasury that is not directly linked to a community member. For community tokens, many of the top wallet addresses actually fall into this category so the generic but crude solution I’ve used is adding the ability to omit the N top wallets from the addresses. The unfiltered “Token current holders” table gives the user a rough sense of how many top addresses to omit. For a specific token, specific known treasury or smart contract addresses can be added to an explicit “black list” that’s ignored in the analysis. The second challenge is “many wallets, one person”. This one requires an explicit decision of the community to use some variant of a proof-of-humanity technology in order to fully overcome it.
  3. Community patterns: the community context in which those indices are used creates its own set of challenges. The population of a country tends to change rather slowly, a single-digit % a year if not less. Communities tend to be more volatile than that with double and triple-digit % change annually. In countries, new adult members (both graduating children and immigrants) join its ranks in various places across the social-economic spectrum, whereas in communities the vast majority of new joiners join with zero tokens. The combination of these two patterns means that several of the indices covered here are particularly sensitive to community growth and therefore the community growth rate needs to be taken into account in interpreting the results.
  4. Collusion: as Vitalik pointed out, these wallet-based indicators only tell part of the story. Even if we had full mapping of wallets to community members, it would still not capture the relationships between those community members and their ability to collude with each other. For example, a small cohort of whales deciding ahead of time to always vote the same way. The indices only give us a partial view that needs to be supplemented with other means. Probably a topic for a whole new post :)

About talentDAO: talentDAO is a community of organizational scientists, strategists, and researchers with a shared mission to unlock human potential in the decentralized, digital economy. We conduct scientific research that helps DAOs thrive while educating the public on the greater decency and agency offered from this decentralized future of work.

Subscribe to itamarg
Receive the latest updates directly to your inbox.
This entry has been permanently stored onchain and signed by its creator.