Currency Trading

A subscriber asked whether the U.S. dollar is a safe haven from the U.S. stock market. One way to address the question is to repeat the tests used in “Best Safe Haven ETF?” on Invesco DB US Dollar Index Bullish Fund (UUP). Specifically, we look at:

1. Contemporaneous UUP return correlation with the S&P 500 Index during all market conditions at daily and monthly frequencies.
2. UUP performance during S&P 500 Index bear markets as defined by the index being below its 10-month simple moving average (SMA10) at the end of the prior month.
3. UUP performance during S&P 500 Index bear markets as defined by the index being -20%, -15% or -10% below its most recent peak at the end of the prior month.

Using daily and monthly dividend-adjusted closing prices for UUP since inception in March 2007 and contemporaneous daily and monthly levels of the S&P 500 Index since June 2006, all through April 2022, we find that: Keep Reading

A subscriber asked which exchange-traded fund (ETF) asset class proxies make the best safe havens for the U.S. stock market as proxied by the S&P 500 Index. To investigate, we test 15 ETFs/funds as potential safe havens:

Utilities Select Sector SPDR Fund (XLU)
iShares 20+ Year Treasury Bond (TLT)
iShares 7-10 Year Treasury Bond (IEF)
iShares 1-3 Year Treasury Bond (SHY)
iShares iBoxx $ Investment Grade Corporate Bond (LQD)
iShares Core US Aggregate Bond (AGG)
iShares TIPS Bond (TIP)
Vanguard Real Estate Index Fund (VNQ)
SPDR Gold Shares (GLD)
Invesco DB Commodity Index Tracking Fund (DBC)
United States Oil Fund, LP (USO)
iShares Silver Trust (SLV)
Invesco DB G10 Currency Harvest Fund (DBV)
SPDR Bloomberg Barclays 1-3 Month T-Bill (BIL)
Grayscale Bitcoin Trust (GBTC)

We consider three ways to find safe havens for the U.S. stock market based on daily or monthly returns:

1. Contemporaneous return correlation with the S&P 500 Index during all market conditions at daily and monthly frequencies.
2. Performance during S&P 500 Index bear markets as defined by the index being below its 10-month simple moving average (SMA10) at the end of the prior month.
3. Performance during S&P 500 Index bear markets as defined by the index being -20%, -15% or -10% below its most recent peak at the end of the prior month.

Using daily and monthly dividend-adjusted closing prices for the above 15 funds since their respective inceptions, and contemporaneous daily and monthly levels of the S&P 500 Index since 10 months before the earliest inception, all through April 2022, we find that: Keep Reading

Are there short-term patterns in bitcoin returns? In their April 2022 paper entitled “Seasonality, Trend-following, and Mean Reversion in Bitcoin”, Matus Padysak and Radovan Vojtko explore short-term bitcoin return behaviors. They look at:

• Daily patterns with respect to NYSE trading hours, defining intraday return as 10:00-16:00, overnight return as 16:00-10:00 and daily return as 1600-16:00 (all New York times).
• Next-day trend-following/reversal based on proximity to maximum or minimum price over the previous 10, 20, 30, 40 or 50 days.

Using hourly and daily bitcoin prices in U.S. dollars from the Gemini exchange during mid-September 2015 through early March 2022, they find that:

Keep Reading

Should investors consider allocations to bitcoin (BTC) in their investment portfolios? In their February 2022 paper entitled “Asset Allocation with Crypto: Application of Preferences for Positive Skewness”, Andrew Ang, Tom Morris and Raffaele Savi investigate how investors should think about allocations to BTC based on features of its historical return distribution. Specifically, they model optimal holdings of BTC, stocks and bonds for investors with either power utility or Cumulative Prospect Theory (CPT, loss aversion) preferences. Using monthly returns for BTC, the S&P 500 Index as a proxy for stocks and the Bloomberg Barclays Treasury Index as a proxy for bonds during July 2010 through December 2021, they find that: Keep Reading

What are the return behaviors of non-fungible tokens (NFT), which employ blockchain technology to convey ownership of unique digital or physical items? In their March 2022 paper entitled “The Economics of Non-Fungible Tokens”, Nicola Borri, Yukun Liu and Aleh Tsyvinski assemble a comprehensive dataset of NFT transactions (including digital art/media and objects related to virtual worlds) and create NFT overall market and sector indexes based on a repeat sales method. They then test:

• NFT market exposure to cryptocurrency market, size, value, momentum and attention factors.
• NFT market exposure to traditional equity, commodity and currency market factors.
• NFT market return predictability based on NFT market volatility, index-to-transaction valuation ratio, volume, momentum and attention factors.
• Individual NFT return predictability based on size and momentum/reversal.

Using blockchain-validated weekly data from major NFT exchanges during January 2018 through December 2021, encompassing about 1.3 million repeat sales, they find that: Keep Reading

Can machine learning usefully inform daily crypto-asset trading? In their August 2021 paper entitled “Boosting Cryptocurrency Return Prediction”, Ilias Filippou, David Rapach and Christoffer Thimsen apply the XGBoost algorithm to decision trees to generate next-day forecasts of bitcoin and Ethereum excess returns. They consider 39 potential predictor inputs, including: valuation ratios (network value-to-transactions, addresses-to-network value and fee-to-price); return volatilities over the past one, two or three months; deviations of prices from moving averages for four lookback intervals; cumulative excess returns (momentum) for three lookback intervals; and, sentiment indicators based on Google Trends searches, Reddit comments or Factiva articles. They further explore the relative importance of individual predictors. Their benchmark predictors are inception-to-date mean past returns. Using daily prices from CoinMetrics for bitcoin since since mid-July 2010 and for Ethereum since early August 2015, both through January 2021, along with the contemporaneous risk-free rate and data needed to calculate non-price input predictors, they find that:

Keep Reading

Does over $1 trillion in crypto-assets (having no fundamental value and paying no interest, but trading freely) now held by Americans represent new wealth? In his brief June 2021 paper entitled “Bubble Wealth”, Bradford Cornell addresses the apparent paradox between the view of economic theory (no new wealth) and the fact that those holding crypto-assets can use them for consumption (yes new wealth). He defines “bubble wealth” as the difference between wealth convertible to future consumption (real assets) and perceived wealth (not founded on real assets). Using a simple numerical example involving bubble wealth creation and destruction, he concludes that: Keep Reading

Are there plausible ways to forecast the price of bitcoin? In their March 2021 paper entitled “Bitcoin Price Forecast Using Quantitative Models”, Daniele Bernardi and Ruggero Bertelli examine fundamental bitcoin value from four perspectives:

1. Stock-to-Flow modeling that addresses evolving scarcity based on quantity of bitcoin already present in the world (Stock) and quantity of bitcoin extracted each year (Flow).
2. How bitcoin price bubbles relate to the halving rule (planned 50% reductions in the reward for successfully mining bitcoins).
3. Demand modeling based on the rate of bitcoin adoption.
4. Supply modeling based on costs and revenues of bitcoin production, emphasizing the role of hash rate (evolving system security).

Based on Bitcoin design and bitcoin price data spanning 2009 through 2020, they conclude that: Keep Reading

Speculator level of interest (attention) is plausibly key to bitcoin price behavior. Does the level of online searches for “bitcoin” as a proxy for attention usefully predict bitcoin return? To investigate, we examine interactions between monthly worldwide search intensity for “bitcoin” as measured by Google Trends to represent speculator attention and monthly bitcoin returns. Using monthly Google Trends data starting September 2014 (to coincide with inception of source price tracking) as retrieved on 9/6/2021 and end-of-month bitcoin prices during September 2014 through August 2021, we find that: Keep Reading

Does bitcoin (BTC) price predictably exhibit momentum or reversion? To investigate, we try three tests:

1. Calculate autocorrelations (serial correlations) between daily, weekly and monthly (4-week) BTC returns and BTC returns for the next five respective intervals (for example, correlation of daily return with returns the next five days). Positive and negative correlations suggest momentum and reversion, respectively.
2. Calculate correlations between next-week BTC return and current BTC price relative to its high or low over the last 13 weeks. A positive correlation for closeness to the recent high (low) suggests momentum (reversion).
3. Calculate average next-week BTC returns by ranked fifth (quintile) of BTC price relative to its high or low over the last 13 weeks.

Using daily, weekly and monthly (4-week) BTC closing prices during September 14, 2014 (the earliest available from the source) through August 31, 2021, we find that:

Keep Reading