Tether vs. USD, is a Dollar a Dollar when it comes to trading?

Liquidity and Volatility Implications

Over the past couple of years, the crypto market has experienced a ‘love/hate’ relationship with stable coins. Though initially designed to increase market liquidity and facilitate settlements, questions are regularly raised regarding their actual safety. The pros and cons of using stablecoins are beyond the scope of this article. Instead, we decided to use market data at our disposal to analyze liquidity (in terms of trading volumes) and volatility implications from a trading perspective. We focused our attention on Tether vs. fiat USD in the case of Bitcoin.

Our dataset contains 365 days of daily OHLCV data over multiple exchanges covered by Kaiko.

Main takeaways

  1. Trading volume is much higher on Tether than on fiat USD and this holds true over time.
  2. BTC-USDT and BTC-USD closing prices (cross exchanges) closely mirror each other with very small tracking errors.
  3. BTC-USDT and BTC-USD daily returns display different volatility regimes over time.
  4. Intraday extremes between BTC-USDT and BTC-USD show higher dispersion than the usual close to close measure.

Aggregate Liquidity statistics

Table 1: aggregate statistics, Kaiko OHLCV data, 365-day window, April 1st, 2019 — March 31, 2020

A first look at aggregate numbers from our dataset draws immediate conclusions regarding trading volumes. Over our 365-day window, the total volume of Bitcoin traded vs. Tether was more than 7 times greater than the volume traded vs. fiat USD. More exchanges reported trades vs. Tether than vs. fiat USD (31 vs. 23) and the market was also less concentrated in the case of Tether: The top 3 exchanges for BTC-USDT had around 13% market share each while in the case of fiat, Coinbase itself reported 26.91% of all BTC/USD volume. This indicates that the Tether market is more fragmented, and hence potentially more competitive. One explanation for fiat-based trading being more concentrated is regulation and KYC requirements.

Exchange statistics

First, we analyzed the trading volume on exchanges for each asset pair (BTC/USD and BTC/USDT). Among all the exchanges considered, 5 exchanges exhibit a market share greater than 5% for the USD pair, whilst 10 exchanges do so for the Tether pair, as shown in Figures 1 and 2, respectively. This shows there is greater dominance of a handful of market places for the fiat market, relative to the Tether market.

BTC-USD pair

Figure 1: Exchanges traded volume, BTC-USD, 365-day window

Exchange selection

In order to take a representative snapshot of the aggregate market we decided to focus our volatility analysis on the most liquid exchanges displaying stable average traded volumes. The market share of the main exchanges (in terms of traded volume) is shown over our one year period in Figure 2. We noticed that all exchanges had a very stable market share except for TideBit displaying a few days with extremely high volumes.

For the rest of our analysis we decided to focus on Bitfinex, Coinbase, Gemini, Kraken, and Bitstamp as representative of the market.

Figure 2: Exchanges daily market share in terms of volume, BTC-USD, top 6 exchanges by market volume.

BTC-USDT pair

Figure 3: Exchanges traded volume, BTC-USDT, 365-day window

Exchange selection

Similarly, we looked at the daily market share of the top 6 exchanges in terms of total volume traded for Bitcoin-USDT in our dataset (Figure 4). We decided to remove BitForex for our volatility analysis since we did not have data for it until late August.

We focused on OkEX, Binance, HitBTC, Huobi, and BeQuant.

Figure 4: Exchanges market share (daily) in terms of volume, BTC-USDT, top 6 exchanges by market volume

Liquidity analysis

We then looked at the distribution of liquidity between fiat USD and Tether over our sample period (the timespan in Figure 4). Looking at the daily ratio of volume traded across our selected exchanges between BTC-USD and BTC-USDT (5 exchanges each) over a one year period we noticed different regimes with an average of 5.85x in favor of Tether. Broken down, this means that in our subset of top 5 exchanges USDT had a 5.85 times higher volume on average compared with USD.

The 25%-75% percentile band came out at [4.41x, 6.95x] while the minimum came out at 2.35. It is interesting to note that over our sample period, Bitcoin always had more than twice the volume vs. Tether than vs. fiat, even at its lowest ratio. The maximum ratio of USDT/USD trading volume was 13.95, indicating at one point, BTC/USDT daily trading volume was 13.95 times greater than BTC/USD.

Figure 4: Daily traded volume ratio BTC-USD / BTC-USDT, 365-day window

Volatility analysis

Beyond pure liquidity, when comparing Tether or fiat USD based trading, market participants will be interested in the pairs’ respective volatilities. In a perfect and efficient world and assuming that Tether is managed as described (with reserves constantly ensuring a 1 to 1 conversion ratio), traders should have no preference between using fiat or the stablecoin as the base trading currency.

In practice, there is an implied credit basis between the two pairs reflecting at any time the market’s assessment of each base currency’s liquidity and credit risk. Factors contributing to the basis are among others, perceived convertibility of Tether into fiat USD, limits, and capital controls to move fiat USD or KYC (Know Your Customer) related premiums.

If the market assumes that this basis is negligible, arbitrageurs should quickly replicate any movement on one of the pairs onto the other.

One way to look at price correlation at a day scale (we are not looking at market microstructure) is to study the pair’s rolling historical volatility. Computing rolling volatilities helps us quantify how much prices move away from their historical averages. If arbitrage strategies are efficient, we expect to see closely related historical volatilities between the two pairs.

Close to close volatility

The usual definition of historical volatility is from close to close. We computed rolling volatilities (30-day rolling window) of one-day log returns. We constructed our closing price as the simple average (equally weighted) of the different closing prices among exchanges (midnight UTC time).

We first charted the average closing price for BTC/USDT and BTC/USD.

Figure 5: Market average closing price, daily, BTC-USDT and BTC-USD, 365-day window

We then calculated the tracking error, which is the percent difference between the two closing prices over time. The tracking error between BTC-USDT and BTC-USD remained very small over our dataset, reaching the 2% mark once and staying within the — 0.50% / +0.50% band most of the time (Figure 6).

Figure 6: Tracking error %, exchange selection average closing price BTC-USDT vs. BTC-USD

We then charted the normal annualized volatility of the two pairs. This strong replication is also visible when looking at the rolling volatilities.

Figure 7: Daily rolling (30 days) close to close volatility (log returns), BTC-USDT and BTC-USD, 365-day window

And finally, the ratio between the two volatilities.

Figure 8: Daily rolling (30 days) close to close volatility (log returns), BTC-USDT and BTC-USD, 365-day window

Normal volatilities for both pairs remained close to parity (Figure 8) though we noticed a few regime changes. Early on in our dataset (until about July 2019, Regime 1), BTC-USDT closing prices were slightly less volatile (between 92% and 98%) than their BTC-USD counterparts. From then on, we had a few months of near-parity (Regime 2) before the ratio dropped again for a short period of time in early 2020 (Regime 3).

These results could be explained by factors independent of the market structure such as exchange selection (the most liquid exchanges are not the same for both pairs) or mean price computation bias (all exchanges have the same weight in our computation so one remedy would be to consider a volume-weighted average for the closing price) or time difference (peak liquidity times will differ on exchanges with different location), among other reasons.

Nevertheless, when looking at the statistics of the normal volatilities ratio distribution, we find clear signs of a strong correlation between the two pairs:

Table 3: Daily ratio of BTC-USDT and BTC-USD rolling (30 days) close to close volatility (log returns) distribution

We plotted the rolling correlation (log returns, 30-day window as well) between our two trading pairs (BTC-USD and BTC-USD), to confirm the strong correlation between them even during the first Regime identified above.

Figure 9: Daily rolling (30 days) close to close correlation (log returns), BTC-USDT and BTC-USD, 365-day window

Daily High-Low volatility

The close to close volatility studied in the previous section is the usual definition of volatility usually considered by developed market practitioners. Still, it is important to remember that the measure is highly dependent on factors such as window size and sampling frequency and does not capture extreme values well.

One way to improve the measurement of tracking errors is to integrate the daily High and Low values provided in our dataset. This allows us to grasp the behavior of intraday volatility and not only close to close.

We define our own measure of intraday “high-low” returns and run the same rolling calculations as above to see if the three volatility regimes we identified using the first calculation are maintained. By taking the ratio of the new “high-low” volatility calculations, we can again try and identify specific volatility regimes, and compare them with those of the first volatility calculation.

We defined a daily high-low return as:

As done for our previous volatility analysis, we compared rolling 30-day volatilities of High-Low returns for our two trading pairs.

Figure 10: Daily ratio of BTC-USDT and BTC-USD rolling (30 days) close to close volatility (log returns)
Figure 11: Daily rolling (30 days) High-Low close to close volatility (log returns), BTC-USDT and BTC-USD, 365-day window

Over the one-year sample, the ratio of High-Low BTC-USDT to BTC-USD rolling volatilities follows the same pattern (and regimes) as the close-to-close volatilities computed previously, although with greater dispersion. The mean of the ratio’s distribution is close to parity (0.98) but the standard deviation is almost four times greater (0.07 vs. 0.02 earlier) and extreme values much further away (min and max of 0.76 and 1.14 respectively vs. 0.91 and 1.02 earlier).

Overall, this shows that intraday moves (in terms of extreme high-low values) can be quite different between the two pairs. Here again different factors may explain these differences (exchange selection, and timing issues for example), and getting a clear cut answer is not straightforward.

Table 4: Daily ratio of BTC-USDT and BTC-USD rolling (30 days) High-Low close-to-close volatility (log returns) distribution statistics

Conclusion

When looking at Bitcoin — USD and Bitcoin — USDT trading pairs, our liquidity analysis showed much higher volumes (factor above 5 on average) in favor of the stablecoin, as one would expect given the nature of the asset. Moreover, daily volume analysis shows that trading involving the stablecoin was at least 2.0x more liquid than fiat at all times, a sign of robustness.

When looking at close-to-close price tracking errors, we noticed that the two trading pairs mirror each other very closely (2% max error with less than 0.50% error most of the time). However, we were able to identify a few different volatility regimes in our dataset.

Finally, when looking at intraday price moves (in terms of high-low returns), we found out that our exchanges subset displayed a higher dispersion of rolling volatilities between the two pairs.

We can think of a few different reasons to explain the higher intraday volatility discrepancy (as per our measure) between the two pairs. The exchange selection is one: the main exchanges are different in both cases with different geographical locations (hence different peak liquidity times) and infrastructures. We may also consider that players between the two pairs are different (ratios of investors to speculators and market makers). Liquidity itself (if measured in terms of different trading volume) could also explain bigger intraday swings, especially for BTC-USD compared to BTC-USDT. Finally, we may consider more structural differences related to varying regulatory regimes between fiat and stablecoin based trading.

Overall, our analysis showed a very strong correlation between BTC-USD and BTC-USDT across our exchange sample. Over our one year dataset, it looked clear that traders did not show a significant preference for fiat USD vs. Tether. Still, a few differences and regimes appeared. Finally, when looking at historical volatilities, it is important to remember that we may define different measures that can capture varying patterns.


Tether vs. USD, is a Dollar a Dollar when it comes to trading? was originally published in Kaiko Data on Medium, where people are continuing the conversation by highlighting and responding to this story.

—Source link—

What do you think?

Making Maker: April 2020

Gnosis Launches Gnosis Safe Apps: Multi-sig DeFi Transactions