Introducing CIRI: Crypto Interest Rate Index

Introduction

Fixed income markets allow borrowers to finance everything from day-to-day living expenses, college tuition, and real estate purchases, to government infrastructure projects, welfare programs and military expenditures.

Fixed income is the largest asset class in the world, rivaled only by real estate, and completely dwarfs the likes of commodities, currencies, and equities.

Crypto will likely follow the path of traditional finance, and fixed income markets will become a far more important component of the ecosystem in the years to come.

Before 2019, most crypto lending came from margin traders on exchanges like Bitfinex, Poloniex, or BitMEX who would borrow crypto to short sell or borrow fiat to leverage-long. Miners also used fixed income instruments to generate yield on the large amounts of crypto they sat on or to hedge volatility risks given their massive operating capital commitments.

This year, though, we’ve started to see the market really take off.

• Genesis Capital did $425 million in Q1 originations; their active loan book is $181 million.
• BlockFi now has $53 million in interest-bearing customer deposits.
• There is $85 million in DAI currently outstanding from MakerDAO’s CDPs.

Still, the markets are highly fragmented and inefficient. That’s due in part to their nascency. Another major issue is the lack of reliable benchmark interest rates.

In Search of “Crypto LIBOR”

Healthy fixed income markets require well-known reference interest rates. In the absence of such an index it is difficult for market participants to fairly price various fixed income instruments. Simply put, they don’t know what price everyone else is paying.

Without this information, it becomes difficult for parties to agree on a deal, and for the market to price credit risk appropriately. Interest rate indices would remove this informational friction, and ultimately help boost overall market liquidity.

In traditional finance LIBOR and the federal funds rate are two of the most important interest rate benchmarks, underpinning hundreds of trillions of dollars of fixed income instruments and their derivatives.

A change in the benchmark rate quickly permeates the rest of the market, as participants adjust their rates relative to the index. It even drives prices in all other non-debt asset classes as it can serve as a “risk-free rate” for invested capital. Below is an example of how FOMC announcement on 03/20/2019 impacted S&P500, U.S.10-year treasury, gold, JPYUSD, and Bitcoin.

In order for a thriving fixed income market to emerge in crypto, we need to identify the appropriate “risk-free” benchmark rate.

Parameters That Drive Interest Rates

The task of creating an interest rate benchmark index in crypto is nontrivial.

Often, an index is some sort of average between multiple instruments. For instance, the S&P 500 is a market-cap weighted average of ~500 large cap publicly traded companies in the US.

In crypto fixed income markets, there could be numerous platforms on which loans are originated and traded. On each platform, instruments differ across a variety of parameters like:

• Instrument borrowed
• Maturity
• Instrument collateralized
• Collateralization amount
• Counterparty risk
• Custody risk
• Synthetics

Which platforms and which instruments should we include in a reference rate index? How do we weigh them? How do we normalize loans with different parameters into one cohesive index?

To answer these questions, we must first understand each one of the parameters and how they apply to crypto markets.

Instrument borrowed

Supply and demand of loans clearly depends on what asset is being borrowed. Interest rates on Litecoin might be higher than that of Bitcoin because the market is less liquid, or because a relatively higher percentage of margin traders want to borrow Litecoin to short it (e.g. a supply shortage drives up borrow costs). As such, each asset deserves its own index or set of indices.

Maturity

This is what is commonly referred to as the “yield curve”.

Typically, the longer the terms of a loan, the higher the interest rate. A 30-year mortgage tends to have a higher interest rate than a 10-year mortgage. A 10-year treasury note tends to have a higher interest rate than on a 1-month treasury bill. That is because lenders require a higher rate of return when they don’t have access to their funds for a longer period of time. Despite these general trends, the shape of the yield curve can (and does) change continuously and in some circumstances long-term rates may even be lower than short-term rates.

Bitfinex operates a short-term loan book where the lender can customize the maturity. They also publishes what they call the “Flash Return Rate”, which is a volume-weighted average of loan rates. Unfortunately, this methodology mixes various maturities into one number.

Ideally, we should have reference interest rates across various maturities - 1 day, 7 days, 30 days, 3 months, 1 year, 5 years, 10 years, etc. For now, we’ll focus on shorter durations to capture the implied “risk free” rate of lending crypto.

Instrument collateralized

Lenders often require borrowers to deposit collateral to protect against default. The type of collateral affects the interest rate, depending on whether the lender can generate yield on the collateral.

For instance, a borrower may borrow Bitcoin from Genesis by collateralizing it with USD. Genesis may be able to park the USD in a money market instrument and to earn 2% per year. On the other hand, a Maker CDP owner borrows Dai by depositing Ether into the smart contract, but MakerDAO holders cannot generate any yield on the Ether (at least, not yet).

When lenders can generate yield on collateral, they are willing to offer lower interest rates.

Collateralization amount

Lenders may or may not require borrowers to fully collateralize their loans. In cases of default a lender can recover their principal by taking possession of or liquidating the collateral. If the loan is not fully collateralized, the lender will require a higher interest rate to compensate for default risk.

“Fully collateralized” does not mean 100% collateralized. It actually often means significantly more than 100% so that the lender is protected against violent price movements of the collateral.

For instance, margin books on exchanges such as Poloniex are virtually fully collateralized, as the borrower will get margin-called by the exchange when the market moves against them. Each Maker CDP is also fully collateralized at 150%. Below is the collateralization ratio of CDPs according to https://mkr.tools/.

On the other hand, when a retail lender opens a Bitcoin interest account with Blockfi, they effectively lend Bitcoin to Blockfi without the company posting collateral. The lender thereby takes on counterparty risk.

Counterparty risk

The credibility of the borrower plays an important role when loans aren’t fully collateralized. That’s why credit scores exist for individuals in most developed countries, and why ratings agencies exist in the fixed income markets. The higher the credit score or debt rating, the lower the interest rate the lender will likely require.

In traditional markets, standard credit ratings are provided by agencies such as S&P. These standard ratings allow the creation of bond indices based on credit levels. For instance, governments will be rated as AAA or AA+. A Fortune 50 bond may secure an A rating while riskier debt offers (B and C ratings), will have much higher yields.

In the world of crypto lending among centralized parties, there are no standardized credit scores or ratings (yet). Borrowers go by reputation. A reputable high-frequency trading shop like Tower Research Capital will likely borrow Bitcoin at a much lower rate than a retail borrower.

The lack of standardized credit scoring makes it hard to objectively normalize different levels of counterparty risk into one index, or to create a separate index for each level of counterparty risk.

Custody risk

The custody risk is something that is truly unique to crypto. Exchanges and other lending platforms have a track record of getting hacked or operating under fractional reserve. Lenders on these platforms theoretically require a higher interest rate to compensate for the risk of losing their funds.

Optionality

Fixed income instruments may come with different optionalities. For instance, the borrower may have the option, but not the obligation, to unilaterally end the loan before the maturity date and pay it off. The borrower benefits from such an option at the expense of the lender, and therefore tends to pay a higher interest rate.

The lender may also have such an option, in which case they will be willing to charge a lower interest rate. On Bitfinex and Poloniex, margin traders have the option to end the loan and repay the lender before the maturity date.

Synthetics

A synthetic financial instrument is one that is engineered to simulate another instrument.

One of the most liquid synthetic loan instruments in crypto is to short the CME Bitcoin futures and long an equal notional amount of spot Bitcoin. By entering these two positions, one effectively borrows Bitcoin by fully collateralizing the loan with USD with a maturity date equal to the futures expiration date and no optionality.

The annualized interest rate of such an instrument is:

A similar example is to simultaneously short the BitMEX perpetual and long an equal notional amount of spot Bitcoin. One effectively borrows Bitcoin by fully collateralizing the loan with USD, with a maturity time of 8 hours, no optionality, and an annualized interest rate of

Finally, one can short a call option and long a put option with the same strike price and maturity on Deribit, while simultaneously longing the spot.

The annualized interest rate on such an instrument is:

Note that with synthetic instruments, it’s entirely possible that interest rates are negative, for instance, when the future trades above the spot. Theoretically, this happens when the lender can generate significant yield on the collateral.

Index Methodology Overview

Let’s take a look at the largest lending markets in crypto right now. The following table describes each of the leading markets in terms of the parameters explained above.

There are five broad categories of markets. 1) Implied rates on derivative exchanges such as Bitmex. 2) Margin books operated by crypto-native exchanges such as Bitfinex. 3) Lending businesses such as BlockFi. 4) OTC desks such as Genesis. 5) DeFi applications.

Note that 1, 2, and 5 generally publish rates that everyone can pull. 3 and 4 are privates businesses that generally do not publish their rates today. If 3 and 4 were to be included in an interest rate index, a daily anonymous survey would need to be conducted as it is withLIBOR.

Each market in the above table has an outstanding loan notional amount of mid eight figures to low nine figures USD. From speaking to market participants, we estimate that aggregate outstanding loans is about one billion USD. For confidentiality reasons, we will not disclose the breakdown by each individual market.

First naive approach: Combinatorial explosion

In theory, we could create an index for each unique parameter set - for instance, an index for Bitcoin 3-month loans fully collateralized with USD, another index for Bitcoin 1-day loans 50% collateralized with call option for nearly risk-free counterparties, and so on.

There are a couple of issues with this approach. First, the large number of parameters would lead to a combinatorial explosion of indices. Which index should borrowers and lenders focus on? Second, and more importantly, the crypto fixed income market is still early. As you can see in the above table, the number of options is very limited. When there aren’t enough constituents or liquidity in an index, the latter can be unnecessarily volatile or manipulated.

Second naive approach: One index to include them all

The opposite extreme is to create one single index that is composed of all reasonably liquid instruments.

This is also problematic because, as we have seen, fixed income instruments aren’t homogeneous. Suppose we have one single index that encompasses loans with a 1-day maturity and loans with a 1-year maturity. How can a lender or a borrower use this index to price a loan with a 30-day maturity? The answer is it’s very hard.

As such, a useful index is one that includes instruments with similar, though non-identical, parameters.

Third approach: Striking a balance

We propose an index methodology that strikes a balance between the two extremes.

In particular, we propose a short-term risk-free rate index. Starting with Bitcoin.

This index includes only instruments with the following parameter set:

1. Either fully collateralized with fiat or crypto
2. No optionality
3. Either plain-vanilla or synthetic
4. Matures in less than 90 days

Why do we choose these parameters? Let’s look at each of the four parameters.

First, unless the borrower is someone with a giant balance sheet like the US government, there is always some degree of counterparty risk. As previously mentioned,

The lack of standardized credit scoring makes it hard to objectively normalize different levels of counterparty risk into one index, or to create a separate index for each level of counterparty risk.

The only type of loan whose counterparty risk can be objectively defined is those that are fully collateralized. We can call these loans “risk-free”, as the lender can take possession of the collateral in the event of a default.

With such a “risk-free” index, lenders can price instruments according to their own level of counterparty risk tolerance.

Second, the number of instruments with optionalities is limited. With a “no optionality” index, borrowers and lenders can price instruments with options according to their perceived value of the options.

Third, synthetic loans such as Bitmex futures should be considered as well, given their significant liquidity and the lack of plain-vanilla instruments in the ecosystem today.

Last, it’s difficult to create an index for each point along the yield curve. There is a major lack of options along the yield curve, especially for the longer terms (> 90 days). We could in theory use far-month derivatives, but they are currently very illiquid.

Mixing 2-day and 90-day instruments isn’t ideal, but we will justify this in the following section.

CIRI (Crypto Interest Rate Index)

Therefore, we have introduced CIRI: Crypto Interest Rate Index.

Given the requirements and constraints described above, the first iteration of CIRI will be a simple average between four rates:

1. Bitmex futures implied rate
2. CME futures implied rate
3. Bitfinex lending rate
4. Poloniex lending rate

The CIRI reference rate will be published daily, every weekday at 1pm UTC (9am EST).

Because we are focusing on the Bitcoin index, Maker will be excluded from the initial index. Additionally, Bitmex perpetual will be excluded Bitmex perpetual will not be included due to its unusual volatility. Historically, rates implied by the Bitmex perpetual was often negative or high double digit percent.

As with any index, the devil is in the details. Below is the detailed methodology for computing each of four initial constituents of CIRI to provide a better understanding of how the index works.

Bitmex futures implied rate

Every hour, we calculate a number called “basis”, as:

Where F is the latest execution price for the front-month future on Bitmex, S is the spot price that corresponds to the Bitmex Price Index. Note the subtle “division by 1.5” in this formula - we assume that the future has a margin requirement of 50%.

One week before the front-month future expires, we start using the next-month future. Future expiration can cause large volatilities.

Then, we take the 24-hour moving average (MA), using the last 24 outputs from the above formula.

Why is it important to use the MA as opposed to single snapshots? Because market volatilities can cause single snapshots to be very noisy.

Why choose 24 hours as the time horizon? Because too long a time horizon will cause the MA to be lagging; too short a time horizon will cause the MA to be noisy.

The basis moving average looks like the following for Bitmex June 2019 future.

Finally, we annualize the above rate

Out of all the implied rates, Bitmex futures implied rate has historically been the most stable. That’s why it’s important to include it in the short-term interest rate index even if its implied duration can be 3 months. Even then, on average its duration is 1.5 months.

CME futures implied rate

The CME futures implied rate is calculated in the same way, with one caveat.

Similarly to most traditional markets, CME Bitcoin futures don’t trade on holidays and weekends. More specifically, they start trading on Sunday 6pm ET and stop trading on Friday 5pm ET.

When it’s time publish CIRI at 1pm UTC (or 9am ET), there are only 17 hourly data points since the start of the trading week. In order to compute the 24-hour MA, we take last 7 hourly data points from the previous Friday.

Bitfinex lending rate

The Bitfinex lending rate is what they call “Flash Return Rate” (FRR), a number that is publicly available on their website for each top cryptoasset.

According to Bitfinex website, the FRR is based on the average of all fixed-rate positions, of all terms, weighted by their amount.

The FRR is a daily rate, so we annualize it as follows:

Poloniex lending rate

Unlike Bitfinex, Poloniex does not publish an aggregate rate. However, they do publish the lending order book.

Every hour, we take a snapshot of the midpoint between the best bid and the best ask. Then we compute the 24-hour MA.

Similarly to Bitfinex, Poloniex’s rate is daily. So it needs to be annualized.

Future Iterations

Early feedback for CIRI has been overwhelmingly positive, but we recognize that it is a work in progress and we foresee steady improvements in the future.

Primarily, the issues preventing a more robust index relate to the lack of data available for certain markets.

Private lending businesses

Private lending businesses are responsible for a significant market share. Intuitively, they should be a major component of the CIRI.

One way to do that, as previously mentioned, is to conduct an anonymous daily poll, similarly to LIBOR. However, even though banking is a heavily regulated industry, LIBOR still has been manipulated in the past by members who participated in the poll, making us skeptical of how well this would work in largely unregulated crypto markets

This is an extremely challenging problem. One way to alleviate this problem is to publish each category of CIRI side by side - exchange margin rates, exchange implied rates, and private lending rates. If private lending rates diverge significantly from exchange margin rates and exchange implied rates, people should be skeptical.

Another way is to include a large number of participants around the world and to trim outliers, so that each one of them cannot influence the rate significantly. In this case, if they wanted to influence the rate, they would have to collude together with each other.

A third way is to include a similar number of net borrowers and net lenders. Normally, net borrowers are incentivized to report a lower rate, while net lenders are incentivized to report a higher rate.

Continuous rates

As previously mentioned, in the first iteration we publish CIRI once per day at 1pm UTC. As exchange margin and implied rates change continuously, we may increase the frequency at which CIRI is calculated published. Smoothing techniques such as moving averages will become more important at increased frequencies. Without smoothing we may end up with noisy data.

Asset coverage

As it stands today lending markets outside of Bitcoin remain small, leading to a lack of data allowing us to calculate rates beyond those related to Bitcoin.

For instance, CME does not have a future contract for Ethereum. Bitmex supports futures trading for several assets, but the lack of liquidity makes it unclear that they are suitable for an index.

Genesis reports that nearly 70% of the total notional amount of their loans are for Bitcoin, while other large caps like Ethereum are in the low single digit percent.

DeFi

DeFi lending platforms such as Maker, Compound, and Dharma suffer a similar problem. While they are growing, they are still very small in the grand scheme of things.

Their advantage relative to private lending rates, or even exchange rates, is that they are fully transparent and very hard to lie about by one central entity.

Derivative exchanges

Options and swap exchanges such as Deribit are gaining momentum. As they become more significant liquidity-wise, the implied rates on these exchanges can become a major component of CIRI.

Expand the yield curve

Currently, most crypto loans are short-term, from 1 day to 1 year. By contrast, treasury bonds can often have a maturity of 30 years.

This is not unexpected, as crypto is a nascent market. It’s hard for people to plan for 30 years ahead.

However, as the industry matures and proliferates with more data points, CIRI will cover the entire yield curve. Below is a hypothetical CIRI yield curve.

CIRI is live on https://messari.io/. If you are a private lending business and interested in contributing, or have any suggestions on the methodology, please reach out at [email protected]

Let us know what you loved about the report, what may be missing, or share any other feedback by filling out this short form. All responses are subject to our Privacy Policy and Terms of Service.