[This was written prior to Dec 13 and is now outdated]
Author: Tim Daubenschuetz
Digital assets thrive on speculation. While our previous experiments at Kiwi News with Harberger taxes and partial common ownership showed promise, they missed a crucial element: enabling price discovery through "scalpers." Without premiums exchanging hands between buyers and sellers, the market lacked the speculative layer that drives early adoption.
Oh et al. [1] demonstrate that scalpers play an essential role in bootstrapping demand for digital assets. They act as market makers, providing liquidity and price discovery. This insight led us to redesign our onchain ad mechanism to explicitly encourage early speculation.
Pricing an ad with Harberger taxes has the benefit of the tax burden increasing with the ad's rising demand. This improves allocative efficiency for an asset compared to privately owned assets.
Unlimited private property titles, where the owner doesn't have to pay demand-based recurring fees, lead to inefficient allocations and give the owners monopoly power.
The most famous Georgist example of such a dynamic is exemplified by the above picture of a billboard reading "EVERYBODY WORKS BUT THE VACANT LOT," in which the private owner states to sell the lot only by the time it has doubled in value, hence demonstrating the allocative inefficiency of private inner city real estate.
As Georgists argue, an ideal inner city land policy then makes land so expensive to own over time, the cost incentivizes owners to perfectly utilize the land, which is why Georgists are known for promoting a continuous land-value tax where its value is assessed regularly, but the assessment only considers the "unimproved land value."
As crypto currency enthusiasts we understand the value of "network goods," goods which rise in value proportional to how well they propagate in the network. The above image exemplifies how this can be true for real estate too. As a land owner, having a dirty power plant as a neighbor is much worse than having nice cafes and gyms nearby. The absurdity of inner city real estate is that while the land appreciates from the networked improvements of the surrounding parcels, the owner itself only has limited capability of increasing the value of their own parcel. In that way, an inner city property owner isn't directly in control over their land's value, the network is.
This effect is what LTV is trying to dampen. With land being the source of all goods in the economy, and since it can create immense amounts of wealth without necessarily needing ongoing investment, taxing land's value would ensure that owners productively use the land or sell it to someone who can utilize it best.
But while true private ownership is unlimited over time, meaning that no one but the current owner has full and unlimited control over who else may own their property in the future. This causes inefficient allocation of unique and non-fungible goods and, in turn, leads to bad prices.
Demand-based recurring fees limit ownership by imposing a variable fee on holding a good. This fee is set such that, unless the owner makes good use of the property, it's not profitable for the owner to underutilize the good.
Land value tax, but also Harberger taxes, are types of demand-based recurring fees.
Compared to LVT, Harberger taxes decentralize the aspect of evaluating a property by requiring the owner to self-assess its value. Whereas the land valuation process in LVT is maliciously capturable through corrupting the centralized assessors, in the Harberger tax scheme, every owner self-assessing their property will also have to sell it at that self-assessed price immediately. In that way, prices in the Harberger tax model are self-correcting.
- Self-assess the value of your land too low, and therefore pay less taxes, and someone might buy it from you.
- Self-assess the value of your land too high, and hence make it less likely to be bought, and you're going to pay a lot of taxes.
Harberger taxes incentivize owners to self-assess the price of their property truthfully as to what its utility is as they're otherwise going to either lose access to the property or pay so much taxes that it makes holding the property unprofitable.
The table below compares LVT and Harberger taxes:
| Aspect | Land Value Tax | Harberger Tax |
|---|---|---|
| Valuation | Assessed by gov/market | Self-assessed |
| Sale | Owner keeps property | Must sell at declared value |
| Stability | Owner controls sales timing | Property can be bought any time |
| Gaming | Corruption | Some strategic pricing |
Taking a closer look at the above table, we can see that considering all aspects, Harberger taxes have the same benefit as LVT, but they also prevent the centralized value assessor from being capturable.
As for the stability of a Harberger tax priced property, since it may have to be sold instantly, it can have bespoke pros and cons depending on which asset types the policy is applied to. Especially the aspect that a sale can happen instantly and without the confirmation of the current owner creates fundamental problems in many asset markets, for example [2]:
- An ENS name that's instantly transferred to a highest bidder can make the old owner lose a lot of money.
- A restaurant which has to change its location "overnight" may lose many long-terms customers if their new location is too far away.
But there are types of assets where a high turn over rate is actually a positive externality. There is, for example, no harm done with replacing an online ad on a social media website several times an hour. In fact, users might even find that it enriches their experience. Which is why we see onchain ads, implemented as depreciating licenses, as a well-fitting model for pioneering a Harberger tax scheme on Ethereum.
In many ways, website real estate is similar to that of inner cities. There's only a limited amount of space and visitors' attention is scarce. The website's curators are responsible for increasing the ad space's value, not the ad publisher etc..
We're implementing a Harberger tax mechanism that follows a linear price decay function:
where:
-
$p(t)$ is the price at time$t$ -
$c$ is the collateral in ETH -
$t_0$ is the time of the last purchase -
$T$ is the period over which the collateral is fully taxed, e.g., 30 days in seconds (2,592,000)
The figure below shows the linear price depreciation, essentially a continuous dutch auction.
In the example above, the tax period is always 30 days, which means that the
entire collateral is taxed over that period. Notice what this does to the
steepness of price decay. The higher the initial price
In more mathematical terms: As the tax period
We call this concept Depreciating Licenses.
Now, when actually implementing the theory of LVT, Harberger taxes and depreciating licenses into Solidity there are all sorts of challenges which arise from the pseudonomous nature of online identity. We'll shed some light on the challenges and how we've overcome them in our implementation.
First of all, a basic implementation of depreciating licenses, using the
The figure below visualizes that process:
sequenceDiagram
participant Alice
participant Ad
participant Bob
participant Treasury
Note over Alice,Treasury: Day 1: Initial Purchase
Alice->>Ad: Buys for 1 ETH
Note right of Ad: Collateral: 1 ETH
Note over Alice,Treasury: After 15 days (50% depreciated), <br />worth 0.5 ETH now.
rect rgb(240, 240, 240)
Bob->>Ad: Buys for 1 ETH
Ad->>Alice: Gets back 0.5 ETH
Ad->>Treasury: Receives 0.5 ETH in taxes
end
What we can see here is that Alice initially buys the Ad for 1 ETH and that her collateral depreciates over 15 days by 50% (at this point, the ad is worth 0.5 ETH). Bob then buys the ad for 1 ETH, and, in the process of overbidding Alice sending her leftover collateral back and the tax revenue (0.5 ETH) to the treasury.
Note how Bob's interactions are highlighted with a background in grey. That is to show that Bob buying the ad for 1 ETH, and sending Alice's remaining collateral back, and sending the taxes to the treasury is just done in one transaction, all triggered by Bob overbidding Alice.
One challenge, however, with designing the system like this is that it's not passing along the higher sales price of Bob (1 ETH) to Alice. What economists call a buyer's premium (0.5 ETH) isn't passed along. Instead, Alice gets sent back her 0.5 ETH of leftover collateral although the ad is now worth 1 ETH. In the above implementation, this weakens participants' motivation to disover and hold onto a Harberger-tax priced property. That problem specifically arises as most digital property is priced through private ownership, giving scalpers both the possibility of using a bought asset's utility and speculative value.
Private property price discovery works well as utility value and speculative value successfully generate a flywheel to incentivize trading. And while, indeed, the utility value of the onchain ad may also be a motivation for ad-publishing scalpers to trade the ad, in practice, we've observed that its utility value alone is too weak for generating a price discovery flywheel, especially when, to exploit price differentials, private ownership is a widely understood substitution model.
Surprisingly, adding a buyer's premium into the above model isn't straight forward. Onchain Harberger tax properties require a buffer of Ether that's being depreciated. Adding extra logic to deprecate collateral at, for example twice the speed may add unnecessary complexity.
So if a premium is paid for acquiring the ad, the new collateral in the ad contract must reflect the level of premium paid during the owner replacement. Yet, during the overbidding, the excess Ether must also be sent to the seller. Which shows that, surprisingly, a premium has to be paid twice. Once as a to-be-depreciated buffer, remaining in the ad contract, and once as a one-off payment to the seller. This is arguably a bit more involved than, for example, selling a used car on a secondary market, where the buyer simply passes a premium to the seller.
sequenceDiagram
participant Alice
participant Ad
participant Bob
participant Treasury
Note over Alice,Treasury: Day 1: Initial Purchase
Alice->>Ad: Buys for 1 ETH
Note right of Ad: Collateral: 1 ETH
Note over Alice,Treasury: After 15 days (50% depreciated), <br />worth 0.5 ETH now.
rect rgb(240, 240, 240)
Bob->>Ad: Buys for 1 ETH
Note right of Ad: Collateral: 0.75 ETH
Ad->>Alice: Gets back 0.5 ETH + 0.25 ETH (premium)
Ad->>Treasury: Receives 0.5 ETH in taxes
end
In the above figure, Bob still buys the ad for 1 Ether, however, now the premium is:
where:
-
$c_{new}$ is the new collateral amount -
$p(t)$ is the last price before Bob's bid.
Meaning that Bob is actually re-pricing the ad to 0.75 Ether while forwarding 0.25 Ether as a one-off payment to Alice.
Beyond the curiosity of having the split the excess Ether in half there's also an issue with how a contract best passes on this premium from seller to buyer. As perfect financial anonymity is possible today, we have no choice but to design primitives with it in mind. The figure below shows one example we've encountered where Bob manages to artificially suppress the premium paid to Alice by creating two accounts, where Bob's first account buys the ad for a small premium, and then ramps up the price through his second account, effectively sending the premium from his second, to his first account.
sequenceDiagram
participant Alice
participant Ad
participant Bob_A1 as Bob 👨
participant Bob_A2 as Bob 🥷
Alice->>Ad: Buys ad worth 1 ETH
Note over Alice,Bob_A2: Alice owns ad for 6 months. <br />It depreciates to 0.5 ETH
Bob_A1->>Ad: Buys for 0.5 ETH + 0.000001
Ad->>Alice: Gets only 0.000001 ETH markup
Note right of Ad: Price artificially suppressed
Bob_A2->>Ad: 🎭 Secretly buys the ad for 1 ETH with anon account
Ad->>Bob_A1: Gets 0.5 ETH markup
- Oh, Sebeom and Rosen, Samuel and Zhang, Anthony Lee, Digital Veblen Goods (December 5, 2023). Available at SSRN: https://ssrn.com/abstract=4042901 or http://dx.doi.org/10.2139/ssrn.4042901
- Buterin, Vitalik. "The Endgame of Ethereum Name Service (ENS) Squatting." vitalik.eth.limo, September 9, 2022, https://vitalik.eth.limo/general/2022/09/09/ens.html (accessed October 28, 2024).
- Price to Demand chart: https://excalidraw.com/#json=YwxqR0Cp1i0a9_Foo-u7H,bPuMWrGwnU7TP10XqG5nZA
- p(t) on Desmos: https://www.desmos.com/calculator/dpev9ha8vf