Cryptoasset Valuation: Considerations regarding “On Value, Velocity and Monetary Theory”

Note: This article is complementary to “Introducing Beta of Velocity.” The two pieces are best understood when read together.

Written by Johnny Antos and Justin Shin

Image Source

Blockchain Advisory Group (BAG) provides technical and principled strategic advisory services to high-quality crypto core teams and traditional private and public corporate management teams.

These are constructive thoughts on Alex Evans’ article “On Value, Velocity and Monetary Theory.” They are meant to be thought-provoking and challenging. The aim is that they allow us all to gain more precision with the way we discuss cryptoasset valuation.

There are some key points that readers should keep in mind when thinking through the VOLT model and evaluating Evans’ conclusions. A few of the points below are clarifications of selected statements, while others are suggested edits to be considered for the VOLT model.

The key takeaway is that simply making velocity endogenous and using Baumol-Tobin does not lead to a perfect “MV=PQ 2.0” model. There are many embedded assumptions in Baumol-Tobin that don’t apply to cryptoassets.

Moreover, the precise definition of transaction costs must be clarified. This is important because in the VOLT model, the shape of the transaction decline curve relative to the market adoption curve is the key determinant of average utility token price in each period. This fact has largely gone unappreciated in the crypto valuation community and the modeling of these curves merits careful consideration.

We hope that this piece enables the creation of new theories beyond the cryptoassets-as-money framework and allows for new valuation paradigms that incorporate the unique and innovative aspects of cryptoassets.

Table of Contents

1. Definition of Transaction Costs
2. Transaction Decline Curve vs. Market Adoption Curve
3. VOLT Model Clarifications
4. Criticisms of the Baumol-Tobin Model that VOLT Model Users Should Consider

Definition of Transaction Costs

According to Evans, C, transaction cost, is “a fully-loaded representation of friction in the VOLT economy”, including (but not limited to):

1. Network transaction fees
2. Exchange fees and spreads
3. Other illiquidity costs
4. Any instability (real or anticipated) motivating the holding of precautionary VOLT balances to safeguard smooth consumption
5. The extent to which any given transaction constitutes a taxable event in the jurisdiction in question
6. Time cost of waiting for confirmations
7. Implicit costs relating to asset custody and counterparty risk, inconvenience, and cognitive load

Note that originally in the Baumol-Tobin model, C was simply the fixed cost of converting between the Store of Value (SoV) asset and the Medium of Exchange (MoE) asset. It isn’t necessarily wrong to expand the definition of C in a cryptoasset context, but it is imperative to distinguish between:

Type I: Costs incurred in the process of acquiring the MoE (e.g., Cost of converting between SoV and MoE)

A. Exchange fees

B. Cognitive load cost (only when incurred at conversion between fiat and crypto)

Type II: Costs incurred when using the MoE as a MoE (e.g., Usage friction)

A. Network transaction fees

B. The extent to which any given transaction constitutes a taxable event in the jurisdiction in question

C. Time cost of waiting for confirmations

D. Implicit costs relating to actually using the MoE, including asset custody and counterparty risk, inconvenience, and cognitive load

Type I costs are the type of SoV-to-MoE transfer costs from the traditional Baumol-Tobin model, whereas Type II costs are incurred during the usage of the MoE. In the causal logic of Baumol-Tobin, it is Type I costs that drive how often agents go to the ATM. All else equal, when the transfer cost between the SoV asset and the MoE asset is high, on average agents hold higher MoE balances and go less frequently to the ATM.

Intuitively, high Type II costs actually have the opposite effect that high Type I costs have on average MoE asset holding: they decrease MoE holding.

This is because these costs aren’t incurred when acquiring the MoE, but rather are incurred during usage of the MoE itself. The best treatment would be to embed both Type I and Type II costs in a “real yield calculation formula” of the entire portfolio of asset holdings. Staying within Baumol-Tobin, the optimal MoE holding level is actually inversely proportional to Type II transaction costs. This is the opposite of how Evans considers Type II costs. Type II costs directly impact the usage of the given cryptoasset as a MoE in the first place, and therefore interplay between PQ and Type II costs should be considered. Type II costs should be reflected in adoption rate and therefore PQ, if staying within a Baumol-Tobin framework.

In summary, including both Type I and Type II costs together in a catch-all variable, C, demonstrates a fundamental misunderstanding of the definition and function of transaction costs in the Baumol-Tobin model.

Even overlooking these errors, we believe that it is inappropriate to characterize C as a catch-all to include 3) illiquidity costs and 4) consumption smoothing and behavior that protects MoE purchasing power (see the below final section “Criticisms of Baumol-Tobin model…”). It is strange to include illiquidity costs in C when the unit of analysis is a MoE asset where the sole purpose of holding the MoE asset is liquidity (to acquire the provisioned product/service in the decentralized crypto network).

Type I and Type II costs likely still move together over time. As usage friction decreases, the transfer friction of converting from fiat to utility token also decreases. However, as we have seen, a decrease in Type I costs causes an decrease in average MoE asset holding, whereas a decrease in Type II costs causes an increase in average MoE asset holding (decreasing usage frictions imply an increase in PQ, which increases average MoE asset holding, all else equal).

Transaction Decline Curve vs. Market Adoption Curve

Equipped with this more precise definition of which transaction costs are included in C and which ones instead should be reflected in PQ, we can analyze the drivers of average utility value in the VOLT model.

Evans summarizes the velocity thesis, writing:

“Tokens that are not store-of-value assets will generally suffer from high velocity at scale as users avoid holding the asset for meaningful periods of time, suppressing ultimate value.”

While the VOLT model is a useful way to incorporate dynamic velocity instead of making uninformed assumptions about changes in velocity over time, the crux of the model (and its “test” of the velocity hypothesis) relies on assumptions about how the market adoption curve looks relative to the transaction cost decline curve.

When incorporating Baumol-Tobin in a cryptoasset valuation context, the core network assumptions driving average utility token price are shifted to:

1. Initial market adoption (cryptoasset GDP, or “Annual Spending in VOLT”)
2. Initial transaction costs (Type I only)
3. The market adoption curve relative to the transaction cost decline curve over the projection period

In the base case of the VOLT model, the “Start of Hypergrowth” and “Start of Accelerated Decline” are set to align perfectly (both are set to 2023). This might appear to be an innocuous simplifying assumption, but it is worth critically thinking about whether a sharp decline in Type I transaction costs and rapid growth in user adoption could happen contemporaneously, or the extent to which one might precede the other.

The timing of these two phenomena is crucial, as it determines the year that the growth of velocity exceeds the growth of GDP (i.e., the year when average token utility price begins declining). As we showed above, in Baumol-Tobin, Type II costs should be implicitly reflected in PQ.

Is rapid user adoption (e.g., hypergrowth of PQ) a cause or a result of a sharp decline in Type I transfer friction? Is rapid user adoption a cause or a result of a sharp decline in Type II transaction costs? It’s worth exploring the causal logic to determine the relative timing of these effects.

In Baumol-Tobin, there is no explicit place for Type II costs and there’s no explicit place for the interactions between Type II costs and PQ. The best course of action would be to have a separate model (perhaps borrowing from diffusion theory) for the relationship between Type II costs and PQ.

Again, this is an important distinction because Evans (and others) discuss the C in Baumol-Tobin as if it is primarily composed of Type II transaction costs. As shown above, this is incorrect. Within Baumol-Tobin, what we really need to understand is the Type I costs.

Remember that Type I costs, C, should be composed only of:

• Exchange fees
• Cognitive friction (only when incurred at point of conversion between fiat and crypto)

Exchange spreads are not incurred by an individual when converting from fiat-to-crypto, so they should not be included. Remember that Type II costs should be reflected appropriately in PQ (see Appendix for further discussion on Type II costs).

Staying within Type I costs, intuitively, it doesn’t seem that the cognitive friction of going from fiat-to-crypto is particularly high. Many understand the general idea of being able to exchange fiat for crypto. In the early periods, the cost of acquiring education / knowledge on how to exchange crypto-to-fiat should perhaps be included in Type I costs, as it is technically incurred during conversion from crypto-to-fiat. On the other hand, it is more of an up-front, fixed cost. Every time you exchange between the SoV and MoE, you do not have to re-learn how to link a fiat bank account, click “buy”, etc. There may be some small time cost at every conversion of clicking through the website or app, although this will likely be automated in a similar way to how recurring fiat bank transfers can be set up today. So it seems that the main composition of C, perhaps after Year 3 when the cognitive load cost of transfer is significantly reduced, would be the explicit exchange fees.

To determine how exchange transfer fees will move over time, one needs to develop a view on how cryptoasset exchanges (and APIs / interoperability in general) will function at steady-state. Those enthusiastic about decentralized exchanges (DEX) might think that once kinks are worked out and DEX are ubiquitous, trusted, safe and provably secure, exchange fees between fiat and crypto will converge to zero. However, those who believe that regulatory bodies will never allow DEX to handle fiat (or tokenized representation of fiat) at scale could justify exchange fees remaining if the ecosystem is still driven by centralized entities controlling capital flows between fiat and crypto. Even competition between centralized exchange entities may not completely drive exchange fees towards zero if there is enough stickiness or path dependency in how users choose exchanges. The level of path dependency itself depends on the degree to which the fiat world and crypto world becomes interoperable. With perfect interoperability between all crypto exchanges and all fiat-to-crypto centralized exchanges, loyalty and trust in a preferred few exchanges might break down completely. Even with high-in-the-stack, consumer-facing utility tokens such as VOLT, it’s easy to imagine a world where an app or desktop client could scan and switch between exchanges to find the the lowest exchange fees in real time.

The underlying assumption of this discussion has been that the SoV asset is fiat-denominated, and not a crypto itself. If the SoV asset is bitcoin, then it again rests on whether you believe that there could exist a trusted (decentralized or centralized) exchange with zero exchange fees.

In addition to requiring a view on how the crypto exchange competitive landscape and market will develop, you also need to try to determine the timeline for when you expect these developments to happen. For instance, will exchange fees collapse completely and be ~0.001% by 2022, or will they remain ~0.25% for some period?

Where you stand in this discussion will determine how you model the transaction cost decline curve relative to the timing of mass user adoption and diffusion. This determines the timing of when average utility token price starts to decline.

In the VOLT model, Evans implicitly sets the effective Type I transfer fees at ~0.05% in 2018, which is likely a drastic underestimate given the current state of the exchange landscape. You can see this in row 28 of our updated VOLT model. If you bring the initial transaction transfer cost up to $800, it leads to a more realistic 0.30% effective fees as a percent of PQ paid in 2018. This is more realistic, as that breaks down to a 0.25% explicit exchange fee + ~0.05% exchange education cost. Notice how this assumption significantly affects velocity and average utility token price, as with higher exchange fees, a higher average MoE asset balance is held.

VOLT Model Clarifications

We agree with many of Evans’ conclusions throughout, but we have a few specific issues:

Evans writes:

“While we can deduce the strong correlation between velocity, VOLT GDP, and transaction costs just by looking at the formulas, the relation between GDP and utility value is what truly matters to our conclusion as per the velocity thesis.

(Alex Evans’ article “On Value, Velocity and Monetary Theory”)

Here we see the how endogenous velocity allows us to decouple utility value from GDP growth. In the INET model, the correlation between GDP growth and INET value is a perfect 1:1 and the correlation between velocity and GDP growth is zero (velocity is fixed). In our model, the correlation between GDP growth and utility value is 0.34. This is what is driving the result.” (emphasis our own)

The error is that basic correlation only measures the linear relationship between two variables. Below, we see that in the VOLT model, average utility value is directly related to the square root of PQ. Thus, focusing on a linear correlation coefficient between the magnitude of the variables fails to fully describe the relationship between GDP growth and average utility token price growth.

Evans’ table focuses on the correlation between the magnitude of the variables, but he includes discussion about the correlation between the growth of the variables. We propose a new table to reflect this:

It’s clear that there are stark differences between the two tables. The correlation between the change in GDP growth and change in average utility value is still almost perfect, as it is in Burniske’s INET model. This illustrates that correlation alone does not tell us much in the context of the velocity thesis. Neither the correlation between the magnitude of variables nor the correlation between the percent change in variables allow us to test the velocity thesis. To test the velocity thesis, one would need to develop and test hypotheses about the Beta of velocity (concept outlined here).

Moreover, we believe it’s inaccurate to state that “utility value is decoupled from GDP growth.” Note that in Evans:

where between some time period t=1 and t=2:

and

so substituting, we have:

Thus, we see that utility value is not decoupled: it is still directly related to the square root of GDP (PQ), albeit the relationship is not linear anymore as in the INET model.

This may just be a phrasing issue, because it is true that in the VOLT model that the magnitudes of GDP and utility value no longer have perfect correlation. This is better than setting V (velocity) as a fixed parameter because velocity changes along with the growth of the network, rate of adoption, and transaction costs. But we believe that it incorrect to characterize the relationship as “decoupled.”

Again, note that correlation is an insufficient metric to determine how changes in PQ affect changes in V (velocity), as outlined here. Evans’ thinks of MV=PQ using Baumol-Tobin, so it is really “average M” and “average utility value” that the VOLT model refers to. This terminology has been added in an updated VOLT model, found at the link below. The sensitivities table values are also corrected (the original VOLT model values were incorrect).

https://docs.google.com/spreadsheets/d/1mW5MJ33Yrxi2xSHsWVitAkmDUYN-eZVJYz_4TmzZzvY/edit?usp=sharing

Furthermore, Evans writes:

“If token value is equal to PQ/VM, we can agree that the problem is not “high” vs “low” velocity, as some writing the topic seems to suggest (though that certainly matters, all things being equal). The real question is how changes in velocity correlate with changes in PQ. Strong positive correlations approaching 1 effectively decouple token value from network transaction growth (note that while this is a drag on the upside, it is protective of value on the downside). If the two are uncorrelated, then token utility value grows (and declines) linearly with demand for the underlying utility (this is what happens in the INET model). Negative correlations act as a lever, generating outsized price swings relative to growth or decline in PQ.” (emphasis our own)

Again, as discussed here, correlation is an inappropriate measure of how one should think about the relationship between changes in PQ and changes in V. Below, we see that even if the correlation is ~1, the monetary base (and thus utility value) required may still shrink dramatically over time.

Because Evans refers to average M and average utility value in each period (and not their values at a specific moment in time), the Beta of Velocity concept doesn’t make sense to apply to the VOLT model, as beta focuses on the changes from one moment in time to another.

Though both have correlation of 1, monetary base decreases when V growth > PQ growth. This demonstrates that correlation doesn’t tell us much.

Evans also writes:

“In the INET model, velocity and PQ have no correlation, while price and PQ have a perfect 1.0 correlation coefficient.”

Note that in the INET model, the correlation between V and PQ is not 0, it is undefined, as V doesn’t change at all over the time period (it is a specified parameter).

Evans:

“The final point to note about our results is that the utility value of VOLT largely depends on factors outside of VOLT’s ecosystem. Namely, expected returns and transaction costs.” (emphasis our own)

Revisiting the derived VOLT formula for average utility value per token, we see that it depends on three variables:

• PQ (GDP)
• C (Type I transaction costs)
• R (the opportunity cost of capital or the rate of return on the external SoV asset)

It seems imprecise to say that the utility value largely depends on factors outside of VOLT’s ecosystem. Although core clearly teams don’t control R, the choices of core teams control aspects of PQ and C, to some extent. For example, by choosing an appropriate target market based on product functionality and using effective marketing strategies to gain higher market adoption, PQ is increased and the network is grown. Although the core team doesn’t control every aspect of the broader crypto exchange ecosystem’s transaction costs, C, by ensuring functionality to allow for users to seamlessly minimize Type I exchange fees of converting from fiat-to-crypto (or in general from an SoV crypto to a MoE crypto), some idiosyncratic exchange frictions can be reduced.

Specifically, the insight about C suggests there’s a conflict of interest. Typically, core teams, investors and users are thought to be aligned with strong incentives to want the native utility token to appreciate (increase) in price relative to BTC/ETH/USD. However, users are certainly better off with a lower C (all else equal). Hence, all else equal, the effects are:

• A higher C implies higher utility token price from the formula
• A lower C implies lower utility token price from the formula

The conflict of interest is that the interests of consumers and core teams/investors are misaligned: all else constant, consumers prefer lower C so that they don’t have to hold as high of a MoE asset balance, while core teams/investors interested in a higher token price prefer higher C.

Unless PQ grows by a similar order of magnitude to that by which C shrinks over time, this seems to support Pfeffer’s thesis that consumers will reap most of the gains in decentralized ecosystems, with gains flowing to them in the form of consumer surplus (low C and thus minimal economic rent extraction in the form of exchange fees).

However, a more apt interpretation of this conflict of interest is that it highlights the inadequacies of using Baumol-Tobin in the context of cryptoassets-as-MoE. If the asset is solely used as a MoE, then as transaction costs decline, average holding of the MoE asset will necessarily decrease, all else equal. In other words, people go to the ATM more frequently when the conversion cost of each trip is lower. The underlying assumption driving this is that the MoE asset is held solely as a MoE, and not to generate a return (see next section). This is clearly not the case with the native tokens of even the most useful and functioning decentralized networks.

Clearly, when C and PQ are dynamic, the net effect on average token utility value is unclear (and possibly unobservable, although that’s an area of interest for further research). Without a solid fundamental understanding of how C and PQ move together over time, it’s not obvious which effect will dominate over any specific time frame.

Criticisms of the Baumol-Tobin Model that VOLT Model Users Should Consider

The Baumol-Tobin model, a development from the simplistic Keynesian theory of money demand, introduced the level of income and interest rate into money demand. The theory rests on several assumptions:

• Money’s primary utility is its use as a Medium of Exchange (MoE)
• Contrary to alternative investments (like bonds), money is a riskless, returnless asset with stable price levels over time

One can easily see that these assumptions break down in the context of cryptoassets. For instance, a MoE asset such as VOLT undoubtedly has a high degree of idiosyncratic risk (and perhaps market risk tied to relatively high volatility of cryptocurrency market) and there is an expectation of return.

In theory, what households consider when demanding two or more assets is their relative return profiles, including both direct and indirect yields. Money’s indirect yield is its convenience yield (liquidity premium). A MoE asset like VOLT bears this convenience yield, and as the Baumol-Tobin equation demonstrates, Type I costs (conversion costs) are directly proportional to optimal money holding. However, a MoE cryptoasset, unlike money, has more variables that affect its direct yield which are not captured by the Baumol-Tobin equation; for instance, a variety of Type II costs (crypto-specific usage friction) should necessarily lower the value of the crypto MoE because of effects on PQ.

Thus on the one hand, it is questionable whether the Baumol-Tobin framework of “money” demand is applicable to cryptoassets that provision decentralized goods or services, which have risk as well as expected return. On the other hand, the Baumol-Tobin formula of transaction demand for money, which is only a function of income, interest rate, and Type I transaction costs (conversion costs), does not seem robust enough to incorporate the direct and indirect yield of a MoE cryptoasset.

Is VOLT a MoE? Yes. Does VOLT fit into the context of “money” demand under Baumol-Tobin’s construct? Certainly not.

General criticisms of Baumol-Tobin model

While the Baumol-Tobin model is not meant to be a robust representation of realistic money demand, we believe that the model suffers from several theoretical shortcomings in the contexts of both fiat and cryptoassets. We challenge some of the simplifying assumptions and the model’s power to explain money (or MoE asset) demand.

First, as Pollock notes, the Baumol-Tobin model (derived from the theory of inventory holdings by a firm) takes a purely cost-minimizing approach to economic entities. That is, it assumes that rational agents will choose to minimize transaction costs (instead of focusing on maximizing utility, revenue, or profit). However, individuals make decisions based on a combination of objective functions — staying within a rational agent formulation, they likely consider how to achieve a target utility level while minimizing expenditure.

Furthermore, one questionable conclusion of the Baumol-Tobin model is that optimal money holding is proportional to the square root of income level. While it is perhaps true that wealthier individuals have higher demand for a medium-of-exchange asset (e.g., fiat or VOLT), it is questionable whether it will follow the square-root rule of inventory theory of demand (i.e., Baumol-Tobin predicts that when household income doubles, they will hold sqrt(2) = ~41% more money). In reality, households are diverse and heterogeneous agents, who choose an optimal money balance based on their consumption preference, budget constraint, risk profile, and perhaps perceived convenience yield. There are other societal effects not captured by the simple direct relationship between income and money demand that is implied by Baumol-Tobin. For example, socioeconomic status is likely a lurking variable on multiple factors that determine a household’s optimal money holding.

Baumol-Tobin also assumes that individuals are risk averse, and that money is a riskless asset. One of the reasons why money demand is inversely proportional to the interest rate is because when interest rates are high, forgone interest income increases, making holding money (which earns no return) less attractive. However, another reason for the inverse proportionality is that changes in bond prices (resulting from changes in the interest rate) make bonds (the SoV asset) inherently riskier than a MoE asset. Thus, individuals will balance allocation between money and bonds based on their personal risk preferences. So Baumol-Tobin implicitly assumes that, on average, individuals are risk-averse with homogeneous expectations of future interest rates. In addition, the model assumes that individuals on average have homogeneous sensitivity to changes in the interest rate (interest-elasticity of the demand for money).

Perhaps the most troublesome of the Baumol-Tobin assumptions is the failure to consider changes in the value of money itself (i.e., purchasing power and return on the MoE asset):

• The Baumol-Tobin theory has an implicit assumption of stable price levels of money, and assumes money to be a riskless asset
• The framework does not include the expected gain/loss of purchasing power caused by fluctuations in value or price levels of the MoE asset over time (e.g., In fiat, inflation rate)
• As a result, it insufficiently takes into account the relative costs and benefits of the MoE asset and the SoV asset, which are the key determinants of a household’s allocation between the MoE and SoV.

Alongside the rate of return of the alternative asset, a rational agent will take into account more factors:

1. Price level of the MoE asset in question
2. Price level of the SoV asset, and
3. Resulting real yield (capital yield above nominal yield) in dollar terms

The relative price levels of the two assets (MoE asset and SoV asset), which are exogenously determined, are not necessarily captured in the interest rate R (although bond interest rate and inflation generally have inverse relationships). For bonds, the real return (which, theoretically, should be compared with return of money) is determined not only by the stream of interest income (driven by R), but also by the price increase/decrease of the bond at a future date (dependent on changes in R). For fiat, its average price level is calculated by prices paid by consumers for a market basket of consumer goods and services. All else equal, an increase in price level of an MoE asset will decrease its purchasing power.

The Baumol-Tobin framework fails to consider these effects. This is primarily because the model constitutes, in the words of Pollock, “nothing but an explanation of the transactional premise for demanding cash.” It assumes that households hold money only to make purchases, and that they predominantly make decisions based on the transaction costs. In reality, agents allocate their wealth to a portfolio of assets based on each asset’s risk profile, return, relative valuation, and transaction costs, among other factors. The return of a MoE asset not only includes the convenience yield engendered by holding cash or cash-like instruments, but also incorporates the expected changes in price level.

Blockchain Advisory Group (BAG) provides technical and principled strategic advisory services to high-quality crypto core teams and traditional private and public corporate management teams across the organizational lifecycle, from initial concept to liquidity events.

Appendix

Discussion on rapid user adoption (increase in PQ) and Type II cost decline:

In recent cases of technological adoption, a rapid decline in Type II transaction costs preceded the tipping point of mass adoption. For example, Uber initially had very low transaction friction (there was even a subsidy due to free rides from referrals) as many people were already comfortable using and trusting their smartphones to facilitate transactions. Even in earlier days, before Uber experienced mass adoption, transaction friction was fairly low (although higher than later on, when mass adoption reduced wait times as a result of more drivers providing network services). To overcome status quo bias, usage friction needs to be extremely low for mass adoption and diffusion of a new technology.

“Lazy User Theory,” the idea that users select the product/solution that requires the least effort on their part to fulfill their needs, supports this line of thinking. If humans are risk averse and use mental accounting heuristics, it makes sense that perceived potential losses loom larger than perceived potential gains. This would mean that the adoption cost (the perceived loss) would need to be low for adoption to occur.

Alternatively, it’s possible that if the perceived benefit is large, rapid user adoption could happen first, and then usage costs decline. What comes to mind are products where the perceived benefit is so large that it exceeds the high frictional costs incurred during usage because of the innovative new possibilities that the technology enables.

For example, the general public perhaps had high perceived transaction costs of acquiring and subsequently learning to use smartphones (and replacing many “old ways” of performing certain tasks). Yet only four years after the release of the iPhone, ~35% of American adults had smartphones. Many agree that Apple did a great job of marketing how simple/intuitive it was to use apps (to lower perceived friction), but there were still initial skeptics.

Wang et al. (2008) conclude that “When considering the possible adoption of a new product or service, cost considerations are important only if the adoption is desirable (i.e., high level of benefit), but benefit considerations remains important whether cost is high or low.” This supports the concept that the perceived benefit is the primary consideration in a potential consumer’s mind. Even if consumers are indeed lazy, the high perceived benefit can overcome laziness and perceived friction.

Which case applies specifically to utility token crypto ecosystems? Is low transaction friction a necessary prerequisite that drives adoption, or is transaction friction not as relevant to mass adoption because the perceived benefit is so high?

This is a bit of a false dichotomy because there’s a high degree of positive feedback (or circular causality) between mass adoption and low usage frictional costs. Though it depends on the developer incentives in the particular cryptoasset ecosystem, there are some intriguing crypto-specific nuances.

Network effects, driven by incentive-based token design (which typically encourages user / community advocacy), play a major role in the reduction of transaction costs, as well as in the perceived benefit. In decentralized systems, increased adoption results in each participant of the ecosystem enjoying a higher surplus, as per Cong et al. (e.g., A larger community will provide more trade counterparties, cheaper goods and services, etc.).

Low usage friction (e.g., better UI/UX, lower volatility for payment tokens, etc.) for specific cryptoassets and the ecosystem might only be achieved when there is a perception that mass adoption is occurring or imminent.

Cryptoasset price volatility, one form of usage friction, will likely not decrease to match the volatility of fiat currencies until cryptoasset capital inflows/outflows are stable. In turn, capital inflows/outflows likely will not stabilize until the entire sector reaches maturity, implying that this may be a rigid frictional cost for MoE cryptoasset ecosystems

Without the UI/UX of fiat-to-crypto exchanges (i.e., crypto “on and off ramps”) provided by centralized entities (e.g., Coinbase in the USA), how can crypto core teams reduce friction in the truly decentralized networks of the present and future?

What incentive is there for the creation of an interoperable, extremely thin UI/UX protocol if that protocol doesn’t extract any economic rent?

Will centralized entities continue to dominate ultimate mass market UI/UX for the general crypto sector, perhaps due to a “nimbleness” advantage of centralized entities over decentralized entities?

Empirically, it seems that few core teams are focusing much time on UI/UX (in part because many are low-level protocols, but even bitcoin still has significant perceived usage friction). Core teams and improvement protocols may only shift to focusing on UI/UX after they observe rapid usage growth in network usage and increased community participation.

Perhaps the perceived benefit of censorship-resistant, decentralized value transfer (or value provisioning, depending on the utility token) is so high that many are willing to incur high transaction friction to capture the benefit. Thus, frictional costs only start rapidly declining after mass user adoption, over time.