Terra Classic Proposal 5234
A recent proposal has posted by a community member Akujiro that has picked up quite a bit of traction. The proposal aims to lower the taxes from 1.2% to 0.2% and and take 10% of the collected seigniorage and add it to the community pool at the end of the epoch. Before explaining my decision, I would like to express my appreciation to community member Akujiro and commend them for their work. I believe this is what governance was meant to look like, where anyone from the community is engaged enough with the ecosystem, identify a policy that they believe would benefit the ecosystem and do something about it.
That being said, I believe there are essentially two issues being voted on here and so I would like to break down each part of the proposal separately and share some thoughts. I’ll first start with the 10% seigniorage being sent to the community pool.
Collect 10% Seigniorage
In the traditional definition of the word, seigniorage is the revenue for a government when the money it creates is worth more than it costs to produce. In the Terra treasury, this is described as the cost of acquiring external collateral such as fiat for the stablecoin.
seigniorage = new minted currency - cost of acquiring collateral
Indeed, what we had before the depeg, was the swapping of Luna into UST, which would burn Luna and mint UST. That burned Luna would be recorded by the seigniorage and then minted at the end of the epoch. Based upon the RewardWeight parameter, some of that minted seigniorage would be given to the oracle pool, and some would be given back to the community pool. With the introduction of columbus-5, the oracle pool was replaced with a permanent burn, so now the split of the seigniorage would be between permanent burns and the community pool.
When the market swap was disabled between Luna and UST, and the entire Lunc burn movement started, seigniorage took on a different meaning all together. Now, you can think of it as representing the amount of the total supply that has been burned in the last epoch. This is directly computed by subtracting the recorded old total supply with the new total supply and after the epoch (approximately 7 days).
The treasury parameter RewardWeight is a number between 0–1 that represents a percentage of how much of the seigniorage each epoch is redirected to the community pool versus being burned. A value of 1.0 indicates that 100% of the seigniorage should be burned after each epoch. But wait, hasn’t the seigniorage already been burned? It has. At the end of the epoch all of the burned tokens are immediately re-minted and then immediately burned again. This means that even though the reward weight might be set to 0.9 or 90% burned, it will seemingly function as a setting of 100% burns during the week, and then mint 10% of that burn back.
Another thing to realize is that you are voting to burn 90% of the total burn during the epoch. These burns are not just the ones as a result of the on-chain tax, but it also includes any manual burns sent to the burn wallet. As a result, if this passes, you should not be alarmed if you see a green “mint” candle on the Lunc burn chart. This is simply the mechanism by which seigniorage works.
That being said, I have written about how we need to raise funds for the chain. A community pool can be used for emergency funds, attracting dApps and projects back to the chain, and paying developers to improve the chain. I vote yes for the 0.9 parameter change.
Reduce the Tax Burn from 1.2% to 0.2%
Now to the more controversial change, the decrease of taxes from 1.2% to 0.2%. I would like to preface my explanation with the disclaimer that I am not an economist and cannot make any predictions about how this parameter change will affect on chain liquidity, traffic, or price. However, I am an expert on optimization (specifically machine learning optimization) and so I am drawing upon that knowledge to guide my decision.
Arguments to Decrease the Tax — I have heard quite a number of people expressing both positive and negative sentiment towards decreasing the tax. The positive group believes that decreasing the tax will bring back the on chain liquidity and that the current rates are prohibitively expensive. They also often times cite one or two examples of projects or dApps that have mentioned they are shutting down or not coming back to the chain because of the high tax. I personally know of at least one project that is doing multiple hops in their dApp and thus being taxed multiple times.
While I sympathize with these projects, it is hard for me to objectively assess the impact of this criticism as I believe there may be a component of confirmation bias where the tax reduction crowd may just be cherry picking these projects to make their point. I say this because I know of other projects that have migrated just fine and are building without hesitation on chain.
As to the argument that the volume on-chain will increase with the 0.2% tax reduction, I think this is a logical argument, but I am very skeptical that it will make that much of a difference. Furthermore, as new validators come back, I believe we will naturally see an uptick in on chain volume as validators attract new delegations. Uptick in volume will be confounded with the unlocking of new utility on the chain and thus may not be accurately measurable with the timing of this proposal.
Arguments to Keep the 1.2% Tax — Now to the tax keepers, the main argument that I hear is that we do not have enough data, and it is too soon to change course. To the argument of we do not have enough data, I completely agree, but to the argument of it is too soon to change course, I disagree. For this, I would like to reference two frameworks in mathematical optimization. The first is a reinforcement learning optimization algorithm called epsilon greedy.
Let’s say you go to the casino and are presented with 10 different slot machines. What you know is that each one of these slot machines have different reward schedules. In the very beginning, you have no idea what to expect in terms of rewards, you pick one at random. This is called the exploration phase. You sit in front of this machine for a bit and pull the lever and see what happens. Now the question is, how long do you stay at the same machine before moving to the next one? 5 pulls? 10 pulls? 20 pulls? In the epsilon greedy algorithm, in the early stages, the best optimization strategy is to explore as early as possible. You only have one data point and need more information in order to maximize future reward, thus, changing machines is important. I believe that is similar to where we are now, with the 1.2%. We do not have enough data, so I believe it is important to continue exploring.
Now, some of you may call foul, the scenario above is not completely synonymous as we do have information about the different slot machines. And in fact, smarter people than me have convincing economic arguments about the improvements to chain volume and project adoption for a lower tax. This leads me to a second optimization algorithm that is widely used to optimize an objective function called gradient descent.
Imagine you are on a rocky mountain and your goal is to get to the bottom of this mountain. It is really foggy or cloudy so you can see a few feet around you, so not that far. And for one last stretch of the imagination, imagine you can take really big steps, like giant mile long steps if you wanted too.
Now the process of gradient descent is to look around you and find which step will take you in the right direction. You see the right direction to go down, but now the question is how big of a step do I take? Do I take little steps, only where I can see? Do I take medium steps, half of my step stride? Or do I take a big step down the direction of steepest descent? In optimization theory, at the very beginning you should take big steps…. You do not know how far down this mountain goes and if you keep taking small steps, it will take you forever to get down the mountain. With gradient descent, your first step e.g. (learning step size) in the beginning is typically the largest step and is gradually reduced as you continue to traverse the mountains landscape. This address the question in my head about the big jump from 1.2% to 0.2%. Most people seem to agree that lowering taxes is the right way to go down the hill, but some are hesitant to take such a big step and some are worried it would drastically reduce the burn. Personally, this big step is consistent with the models of optimization theory in my decision process. I think early exploration with large deltas in tax jumps could help us find the best reward state of the parameters of the blockchain faster and more efficiently.
Limited Data Analysis
So to recap, I am not convinced by some of the arguments to decrease the tax, nor am convinced that it is too soon to change the tax rate, and I do believe a large step is justifiable. And let me emphasize that the reason I believe these change parameters are ok is NOT because I understand what will happen, but rather the exact opposite. I am admitting that I do not know and thus I believe exploration at this stage is more important than exploitation. We are in uncharted territory which indicates that we should explore rather than exploit, and we should take larger steps rather than smaller ones.
That being said, there is some hard data that we can analyze at this point which ultimately was my deciding factor. If we look at the on-chain volume over the past 23 days (since the start of the 1.2% tax), we have the following trend.
The blue line represents a simple linear regression analysis of the transactional volume using a mean squared error, essentially a line of best fit. A positive slope would indicate an uptrend, and a negative slope indicates a down trend. Given all of the data so far, we have a negative trend of on-chain volume. Maybe short term trends show something different?
Indeed, if we use less and less data closer to today, you can see that the slope of the linear regression line does flip positive, indicating that there were three days where it seemed like the on-chain volume was trending back up. However, that again flipped as we used less days and flipped back negative. Thus, the argument that we need to wait and volume will pick up does not hold weight with my limited data analysis. Furthermore, recent chain analysis by StrathCole indicates that the reduction in volume can be explained largely by CEX transactions and thus waiting for volume to pick up is also unlikely.
Summary
I was skeptical of arguments on both sides and I did my own research. I encourage everyone to do the same and not just jump on a particular bandwagon when it comes to these decisions. Furthermore, I am not advocating for any one side, but rather laying out the reasons behind my own personal decision. After weighing all the data, reflecting on what I know and recognizing what I don’t know, I voted yes to proposition 5234.