Testing the Theory: Is the Perpetual Swap funding rate predictive

Background: What is the Perpetual Swap

The perpetual swap is a synthetic margin trading instrument that was designed to appeal to crypto traders who were used to exclusively trading spot and had no history with the complex world of derivatives. They have, however, emerged as complex, nuanced financial instruments in their own right.

Perpetual swaps have no expiry and operationally work as a never-ending series of 8-hour futures contracts. At the end of every 8 hours, the contract holder is paid or charged an interest rate on the price difference between the price listed on the contract and the spot price at the time. A trader’s funding payment or payout is the size of the position times the funding rate.

A negative funding rate means short position holders pay long position holders, and vice versa for a positive funding rate. If at the end of the 8-hour period, the swap price was below the underlying, shorters have to pay to keep the position open. Perpetuals work this way to keep markets balanced, and it incentivizes traders to go against the popular trade. Savvy traders can trade out of their positions for a profit before they need to pay funding.

Exchanges differ in how they manage funding rates and position maintenance. Kraken, for example, requires that users pay to fund with interest every hour. Deribit takes things one step further and transfers funding every few seconds with a funding rate that is also calculated every few seconds. More frequent position maintenance means that while traders may have to pay interest more frequently it does reduce the likelihood of being auto-liquidated.

The Funding rate And Its Predictive Value Theory

While it may be tempting to use the funding rate as a forward indicator, generally, funding rates move along with the price trend. This is more likely because price impacts funding as opposed to the other way around.

Generally, upward price swings can cause outsized buying pressure in the perpetuals market as more traders want to be long than short. As a result, a premium emerges between the perpetual and spot markets, and the funding rate increases.

In October 2017, Bitmex discussed a potential strategy to test the potential predictive capabilities of the funding rate. Their basic hypothesis is that more extreme funding rate values lead to an increased likelihood of there being a positive or negative return in the next funding. The return is the difference in price between two funding periods.

Bitmex publishes its funding rate 8 hours ahead of time so traders know what it is going to be in the next interval. This is designed to be a grace period, to allow traders to exit positions before they are set to pay or receive funding. This system means we have something of a ‘natural experiment’ scenario. The Bitmex design which is meant to be a backdoor for traders who may be close to being underwater may have effects on other variables like price.

Periods of negative funding where shorts are paying longs may lead to some traders switching long in order to receive interest payments and this may lead to a price reversal. A likely condition for this to occur is a high funding rate, to incentivize a trader to swap out of a potentially profitable position to earn interest instead.

Setting up the test

This updated analysis deploys a similar strategy and makes some assurances that the data fits this type of analysis. For this analysis, Brave New Coin derivatives market data is used to source Bitmex funding rate data between the 2nd of November 2022 and the 22nd of February 2023. This data is used alongside Index Price data and market data.

Doing analysis in this period, we do observe a few sharp variations in funding rate, in general, however, it is low. The funding rate remains in a range between -0.05%-0.05% but in November 2022, it dropped to -0.2%. This is the largest funding rate requirement observed since November 2021. This means there are instances of extreme funding rate results in this dataset. This is important because extreme funding instances are more likely to impact price.

The closer to zero the funding rate is, the less likely it is to have predictive capabilities since traders are less likely to switch from long to short positions or vice versa if they are not going to receive higher funding payments.

The Z-score indicates the extremeness of a particular data point, or how many standard deviations away from the mean a particular result in a dataset is. There are extreme outlier data points in this data evidenced by Z-scores as high as 8 standard deviations from the mean. An outlier is considered to be +/- 3 standard deviations from the mean.

If there are extreme instances of funding rates in one direction or the other, it may be worth counter-trading the market because of the mean reversion theory. This is a theory in finance that says asset price volatility and historical returns eventually will revert to the long-run mean or average level of the entire dataset. Extreme values do not tend to last.

A positive return in the next period suggests a switch of traders from short to long. When there is a high negative funding rate, shorts are paying longs because shorting is the more popular trade.

So if a high negative funding rate is observed, it is likely that much of the market believes the price is set to go down.

Testing whether Funding rate has an effect on price changes

For the next sections, T0 will refer to the initial time period when the funding rate is published, and T1 as the period when it is implemented. As mentioned earlier, Bitmex publishes its funding rate 8 hours ahead of time, so traders know what it is going to be in the next interval. Essentially, this is designed to be a grace period, to allow traders to exit positions before they are set to pay or receive funding.

The scatter chart above has the funding rate at T0 on the x-axis and the spot price return over the following eight hours on the y-axis(Price at T1-T0).

On the scatter we see some evidence that if the funding rate is extreme, there is some effect on the premium in the direction we’d expect. In the instances when funding rate was the lowest at -0.25%, we observe a high positive return, as we may expect a switch from short-to-long worked in these instances – perhaps because of mean reversion theory.

There is a possibility that both the following could be true. One, that traders observed the large negative funding rate and thought that BTC was over shorted or two, that they switched long knowing that they’d receive funding in the next period.

For the most part, however, this link appears to be loose. The connection between funding rate and returns in the next period is mixed.

Factors During the Testing period

Interestingly there was a lot of shorting in the market in November 2022 because of the FTX debacle. November 7th-10th was the peak of the FTX crisis. During this period, price and market sentiment for crypto was tanking. The funding rate had to reach such high negative levels to attempt to balance out the market’s bearishness.

If during this period investors had more confidence that switching long would eventually be how the market would shift, then there may be more instances of high negative funding leading to high positive premiums in November. It is likely that sentiment was so dire that it did not make sense to switch until the market chaos settled. Shorters for the most part stayed short. Mean reversion did occur but not immediately after a high negative funding rate rounds.

This is the funding rate during the period when it was extremely low in November 2022.

The Funding rate may work as a signal to suggest the general sentiment of a market or it could impact price directly. Periods of negative funding where shorts are paying longs may lead to some traders switching long in order to receive interest payments and this may lead to a price reversal. A likely condition for this to occur is a high funding rate, to incentivize traders to swap out of a potentially profitable position to earn interest instead.

However, even if there is a large switch of short positions to long positions on Bitmex, if the action on the spot market is larger then the price may not be impacted significantly. The Bitmex index price is composed of BTC/USD price feeds across Coinbase, Gemini, Kraken, Bitstamp, Bitfinex, and Binance.US. If traders on these spot markets are outselling traders switching long on the perpetual swap market to get interest payments, then the price will still slide and there won’t be a positive return.

It is perhaps the case that during this period, the intensity of the external bearish headwinds superseded the effect of potential switching in perpetual positions to realize potential gains from funding payments.

Conclusions

• Testing whether funding rate publication has the power to predict price changes in future periods yields interesting, insightful results.
• While it does not appear that there is a strong relationship between funding rate and price, in the most extreme cases when funding payments are set to be high, there are market reversals.
• This suggests that there are occasions when a funding rate payment is high enough to compel traders to go against momentum, or that they are backing a mean reversion narrative.
• In the last four months, while funding rates have been primarily close to zero, black swan market events led to extreme funding rate instances.
• It appears that even when funding rates were extremely negative in November 2022 and it was evident that longs were set to receive a high-interest rate payment in the next funding cycle (or that mean reversion may have been possible), in most cases, this was not enough to switch markets.
• The bearishness of the FTX collapse appears to have taken over markets and left traders bearish for a period, irrespective of potential interest payments.

Note:
The above is just a sample of the potential analysis that can be conducted with Brave New Coin’s derivative market data. More detailed analysis, and spreadsheets used for the analysis are available upon request.

The Brave New Coin Derivative data product offers a single source to access standardised market data across the leading derivative exchanges.

Other Regression Test Results

The table below shows the results of a regression where y = Log(Price at T1/ Price at T0) and the variable is the funding rate published in T0. The R-squared (Coefficient of determination) is extremely low suggesting that there is not much predictive power in the funding rate.

Using another variation on ‘premium’, in this case simply the index price in T1 minus the Index price in T0, The R-squared is higher but still not large.