Inflation and Participation in Stake Based Token Protocols

Popular crypto token and blockchain based protocols use different models for the issuance of their tokens.

  • Some, such as Augur and Golem, issue a fixed amount of token initially, and there will never be any more tokens created.
  • Others, such as Bitcoin and ZCash, release token gradually over time via mining or other means, yet there is a fixed supply such that the final amount of token issued will never exceed a certain total.
  • And then there are protocols like Ethereum, where there is continuous issuance, with a temporarily fixed, yet fork-moveable inflation rate, such that the total amount of token is uncertain.

Depending upon the type of protocol, and what sort of behavior it is incentivizing, each of these models can make sense. In this post, I’ll describe an algorithmically adjusting token issuance model for stake based protocols, where high quality participation is vital to the quality and security of the network. The case for such a model in this type of network is as follows:

  1. Participants contribute value to the network.
  2. Greater participation rate yields greater security.
  3. Incentives must exist for participation.
  4. Targeting a participation rate, rather than a token issuance rate, aligns better with ensuring a high quality network.
  5. Participation targets may be easier to move via decentralized governance than token issuance rates.

The result of working through these assumptions is that a protocol can set a participation target, say 50% of all tokens staked and participating, and that inflation rate can adjust automatically to incentivize this target. If less than the target % of token are bonded, then the inflation rate can creep up and the protocol will issue more token each round. If the participation rate is above the target, the inflation rate can reduce itself and fewer token will be issued each round.

Inflation, and hence token issuance, adjusts through three rounds until the participation rate matches the target.

This leads to an unpredictable issuance of token over time. Let’s explore why this may be more likely to lead to a network that delivers its core value proposition as efficiently as intended.

Note — A quick background on proof of stake, bonding, and delegation: in a proof of stake, or delegated proof of stake protocol, users deposit their tokens for a period of time in exchange for having the right to do a job within the protocol, such as produce a block, earn a block reward, participate in governance, or elect someone else to play these roles. See this post on bonding and delegation in Livepeer for a short overview and elaboration on these terms.

Participants contribute value

In these scheme, we’ll define participants as users who bond their token. Since they’re bonding token and have to wait some period of time before being able to unbond and withdraw (possibly weeks or months), they are necessarily aligned with the long term success and usefulness of the network.

But the value they contribute must extend well beyond simply being aligned with the network. They must also be performing a vital function to the quality and security of the network. In proof of stake (PoS) protocols, they may be block producers or validators. It’s important that they maintain connectivity and produce or sign a block each round, lest the protocol be unable to move forward securely and efficiently. In the case of delegated proof of stake (DPoS) protocols, they may provide the valuable function of researching and electing 3rd parties to play either these roles or others. They delegate towards these parties, and risk losing some of their stake if the 3rd party doesn’t act in the best interest of the network.

More participation yields more security

In DPoS systems, there are often a limited number of special roles that are being delegated towards. The higher % of the total network that is participating in this election process, the more expensive it becomes to elect someone into one of these critical roles in order to mount an attack on the network.

To give a concrete example, in Cosmos there are a limited number of validators who collect transactions and coordinate to produce/sign blocks in the Tendermint consensus algorithm. Due to practical limitations of networking and coordination, there are a limited number of these roles (perhaps 100 initially). If someone can acquire enough stake to elect 1/3rd of these nodes, then the protocol potentially breaks down and needs to fork if they act dishonestly. Now try to examine the cost of acquiring 1/3rd of the total elected validator nodes relative to market cap of Atoms and the participation rate: if 10% of the total Atom tokens are bonded then an attacker only needs to acquire 3.3% of the total outstanding token to issue this attack, but if 90% of the token are bonded, then an attacker must pay up to acquire 30% of the token.

While this example simplifies the math a little bit, hopefully it helps illustrate that more bonded stake increases the cost of attack, and adds more security to the network. Remember that bonded stake is locked up long term, so therefore is more likely to prioritize long term usefulness of the network, rather than short term opportunity at the expense of the network.

You need incentives for participation

We’ve addressed how participation through staking provides security and adds value, but why would someone dedicate their resources — knowedge, participation, precious time — and lock up their token in a bonded state for a period of time in order to participate? Economic incentives of course.

Some networks entitle bonded participants to collect a portion of network fees that flow through the platform, others give governance rights over the future direction and utility, and others use inflation to incentivize participants.

Inflation is interesting, because how it gets distributed determines what the network prioritizes long term. If newly minted token were distributed evenly across all token holders, it would have no effect. If distributed only to the few who play special roles, like miners, then it has the potential to concentrate wealth and power in the network. In a stake based protocol however, it can be distributed widely to every participating user in proportion to their stake. This has the effect of growing network ownership amongst those who participate, and shrinking it amongst those who do not participate.

We’ve already determined that participation adds value to the network, and inflation can be used as the incentive that leads to participation. Long term alignment, value creation, and increased ownership for those who are providing such, are the results.

Target a participation rate to ensure a high quality network, rather than an inflation rate

Taking the above points, it may seem like a network may want as near 100% participation through bonding as possible. It is also worth noting that while more participation adds security, there are also good uses for some token to remain unbonded. The ability to transfer the token, the ability to use it for the main utility of the network, and the ability to use it within ecosystem applications are all reasons not to stake. In Ethereum for example, when Casper goes live, people will still need liquid ETH in order to pay gas fees on the network, in order to deposit it for use within ENS or Maker, and to preserve the ability to cash some out to pay for goods and services in their local currency.

This line of thinking would lead one to the conclusion that it may be more important to target a specific participation rate to ensure a high quality and secure network, than it is to target a specific token issuance schedule with a fixed supply.

Since inflation can be used to incentivize participation, if participation is too low, inflation can gradually increase. This will lead more people to participate, as they see increasing benefits and opportunity to increase their network ownership. Or it will lead those participating to grow their network ownership at a faster rate, increasing their own incentives to add value to the network through providing more of the resource the network considers valuable (CPU, bandwidth, connectivity, time, expertise, etc).

If participation is higher than it needs to be, then the inflation can decrease, causing less of a penalty in terms of decreased network ownership to those who choose to use the token for other means. It can spur creation of additional ecosystem components that make alternate uses available for the token and therefore increase total utility for all.

And this inflation rate movement can happen algorithmically, without any human or governance intervention. In fact, it’s likely better that way. If the main lever to pull to increase or decrease participation (and hence network quality) was only moveable via governance, then this could paralyze a network or render it operating at sub-optimal efficiency, for quite awhile or permanently. Using an adjustment algorithm to move the incentive for participation is going to very quickly cause users to either change their behavior, or shift network ownership towards those who are acting in the best interest of the network.

Participation targets may be easier to move via decentralized governance than inflation rate

By now you’ve seen the above suggestions that inflation rate should be moved automatically by the protocol to incentivize a certain participation target. But what if you don’t set the right participation target? Isn’t this just shifting moving one parameter via governance, to another parameter via governance? Isn’t it very hard to enact a hard fork, or pass a parameter update vote, or get miners to signal for updates?

Well yes the parameter may need to be moved, and yes it will need to happen via governance, and yes each of these mechanisms brings its own challenges. But it may be possible that it is easier for stakeholders to mentally grasp the concrete effects of moving participation targets, than it is the effects of moving the inflation rate.

For example, imagine a protocol doesn’t appear to have a lot of participation, and a group is trying to lobby for a hard fork to move annual inflation rate for a token from 6.8% to 7.3%. On the nay side, it is easy for users to see that they’ll be further diluted if they don’t take more action. And at best on the yay side, they may believe that participation will actually increase…but who really knows by how much?

Now imagine in the same protocol, the participation target is set at 50%, and a group is trying to make the case that we’ve seen about 3% dishonest nodes acting in important roles, and there’s plenty of token liquidity for alternative uses, so boosting the participation rate to 60% will reduce the % of dishonest actors hurting network efficiency. Even if this is just as arbitrary as the inflation rate proposal in the previous scenario, it feels more human graspable and tied to the positive impact. There’s no mention of inflation, dilution, token issuance schedule. There’s just a long term, value aligned, argument to be made for the suggested change.

The End Result

A major shift under these circumstances is that it’s hard to calculate the value of a token if the eventual supply isn’t fixed or even mildly deterministic. It is possible however, to calculate the value of a network based upon the utility and future expected usage. And isn’t that what people should be doing anyway?

In stake based systems, such as described above, it’s important to note that the actors who add value by participating, playing the roles, and staking will always maintain or grow their ownership percentage in the network. It’s not about the value of any one token, but the value that a certain percent ownership in the network represents. That becomes protectable and increasable by adding value and security. And this is the alignment of incentives protocols should be looking for.

Building live streaming on the blockchain at Livepeer. Previously Founder, VP Eng at Wildcard and Hyperpublic (acquired by Groupon).