Strategic Allocation

Can machine learning perfect equity sector rotation? In the January 2023 version of their paper entitled “Deep Sector Rotation Swing Trading”, flagged by a subscriber, Joel Bock and Akhilesh Maewal present a sector rotation strategy guided by multiple-input, multiple output deep learning model. The strategy chooses weekly from among 11 U.S. sectors using exchange-traded fund (ETF) proxies. Specifically, each week during each year, they:

• Train the machine learning model on the last two years of weekly (Friday close) historical sector ETF prices and volumes and sometimes auxiliary economic data (10-year U.S. Treasury yield, USD currency index, crude oil proxy and stock market volatility) to predict next-week opening and closing prices for each ETF.
• Compare the predicted return estimate for each ETF to a dynamically updated threshold return to screen for potential buys.
• Apply additional filters to screen out potential buys with unusual past losses to accommodate investor loss aversion.
• At the next-week open, allocate available capital to surviving sector ETFs based on respective past win rate (profitable trade) and respective past sector trade momentum.
• Liquidate all positions just prior to the next-week close.

Their benchmark is buying and holding the S&P 500 Index with reinvested dividends. Using weekly inputs as described during January 2012 through December 2022, they find that:

Keep Reading

“Simplified Offensive, Defensive and Risk Mode Identification Momentum Strategy” describes a strategy that each month holds offensive (defensive) assets when average return on iShares TIPS Bond ETF (TIP) over the past 1, 3, 6 and 12 months is positive (negative). Is past return of TIP, which purely impounds investor expectations for U.S. inflation, a reliable indicator of future asset class returns? To investigate, we relate TIP returns to future returns for each of the following exchange-traded fund (ETF) asset class proxies:

• Equities:
  • SPDR S&P 500 (SPY)
  • iShares Russell 2000 Index (IWM)
  • iShares MSCI EAFE Index (EFA)
  • iShares MSCI Emerging Markets Index (EEM)
• Bonds:
  • iShares Barclays 20+ Year Treasury Bond (TLT)
  • iShares iBoxx $ Investment Grade Corporate Bond (LQD)
  • iShares JPMorgan Emerging Markets Bond Fund (EMB)
• Real assets:
  • Vanguard REIT ETF (VNQ)
  • SPDR Gold Shares (GLD)
  • Invesco DB Commodity Index Tracking (DBC)

We consider both linear correlation and non-linear ranking tests. We look at TIP returns over the past 1, 3, 6 and 12 months separately, and as an average of these past returns (average past TIP return). Using monthly dividend-adjusted returns for TIP and the above asset class proxies as available during December 2003 (limited by TIP) through February 2023, we find that: Keep Reading

Is portfolio downside risk better manageable by combining drawdown and recovery into a single “submergence” metric? In their February 2023 paper entitled “Submergence = Drawdown Plus Recovery”, Dane Rook, Dan Golosovker and Ashby Monk present submergence density as a new risk metric to help investors analyze asset/portfolio drawdown and recovery jointly. They define submergence (s) of an asset/portfolio as percentage of current value below its past highwater mark. They define submergence density (d) during a given sample period as the κ-root of the sum of measurement interval submergences each raised to the κ-power during the sample period, as follows:

When κ=1,  submergence density is arithmetic average drawdown. When κ=∞ (or a very high number), submergence density is maximum drawdown. They use κ = 5 in examples to represent typical investor drawdown sensitivity. They define submergence risk-adjusted return as excess return (relative to the risk-free rate) minus a fraction (Θ) of submergence density, using the range 0.2 to 0.5 as reasonable for Θ. They apply submergence to several market indexes and discuss portfolio diversification and rebalancing in the context of reducing submergence overlaps. Using monthly excess returns for U.S. stocks, corporate bonds and Treasuries (S&P 500 Index, ICE BoA Corporate Bond Index and Bloomberg Treasuries Index, respectively) during 1979 through 2022, they find that:

Keep Reading

Can investors achieve attractive asset class momentum strategy performance by applying mixed-lookback interval momentum to different risk-on (offensive) and risk-off (defensive) sets of exchange-traded funds (ETF), and to a separate risk mode identification ETF? In their February 2023 paper entitled “Dual and Canary Momentum with Rising Yields/Inflation: Hybrid Asset Allocation (HAA)”, Wouter Keller and Jan Willem Keuning present a simplification of the prior Bold Asset Allocation strategy. This Hybrid Asset Allocation strategy consists of the following baseline asset universes and rules, with a single asset momentum metric (equal-weighted average return over the past 1, 3, 6 and 12 months):

• When TIP momentum is positive (negative), use the offensive (defensive) mode.
• When in offensive mode, hold the equal-weighted four of SPY, IWM, VWO, VEA, VNQ, DBC, IEF and TLT with the strongest momentum, except replace any of the top four with non-positive momentum by the one of BIL and IEF with the strongest momentum for crash protection.
• When in defensive mode, hold the one of BIL and IEF with the strongest momentum.

They reform the portfolio monthly, assuming constant 0.1% 1-way trading frictions. Using modeled monthly total returns prior to ETF inception and actual monthly total returns after inception for each specified ETF during December 1970 through December 2022, they find that: Keep Reading

A subscriber requested review of the Golden Butterfly (GB) portfolio, which assigns equal weights to the total stock market, small-capitalization value stocks, long-term government bonds, short-term government bonds and gold. To investigate, we use the following exchange-traded funds (ETF) as asset class proxies, respectively:

• Vanguard Total Stock Market Index Fund (VTI)
• iShares S&P Small-Cap 600 Value Fund (IJS)
• iShares Barclays 20+ Year Treasury Bond (TLT)
• iShares 1-3 Year Treasury Bond (SHY)
• SPDR Gold Shares (GLD)

We consider either monthly or annual rebalancings to equal weight, ignoring associated trading frictions. Using monthly dividend-adjusted prices for the five ETFs during November 2004 (limited by GLD) through December 2022, we find that: Keep Reading

Referring to “Asset Class Momentum Faster During Bear Markets?”, a subscriber asked about performance of a modification of the equal-weighted top three (EW Top 3) version of the “Simple Asset Class ETF Momentum Strategy” (SACEMS) which uses the baseline momentum ranking (lookback) interval when the S&P 500 Index is above its 10-month simple moving average (SMA10) and a shorter lookback interval when the index is below its SMA10. To investigate, we look at average monthly return, standard deviation of monthly returns, monthly reward/risk (average divided by standard deviation), compound annual growth rate (CAGR) and maximum drawdown (MaxDD) as key performance statistics. Using monthly SACEMS returns for different lookback intervals since July 2006 and monthly levels of the S&P 500 Index since September 2005, all through December 2022, we find that: Keep Reading

How would currently popular asset allocation strategies have performed back to the end of 1925? In their January 2023 paper entitled “A Century of Asset Allocation Crash Risk”, Mikhail Samonov and Nonna Sorokina test the following seven widely used asset allocation strategies back to 1926 using a combination of actual and modeled returns for many asset sub-classes and factors:

1. U.S. 60/40 – 60% U.S. large caps and 40% U.S. aggregate bonds
2. Global 60/40 – 60% global stocks and 40% global aggregate bonds
3. Diversified Multi-Asset – 60% to equity asset classes (15% U.S. large caps, 5% U.S. small caps, 5% U.S. growth, 5% U.S. value, 10% U.S. REITs, 10% international developed markets, 10% emerging markets); 34% to fixed income (8% 10-year U.S. Treasuries, 8% U.S. municipal bonds, 8% U.S. investment grade corporate bonds, 5% international bonds, 5% emerging market bonds); and, 6% to commodities.
4. Risk Parity – 33% of portfolio risk to each of U.S. large caps, 10-year U.S. Treasuries and commodities, leveraged to 11.4% target volatility (the retrospective volatility of U.S. 60/40).
5. Endowment – per the 2020 National Association of College and University Business Officers report, 13% U.S. public equities, 13% non-U.S. public equities, 7.3% global equities, 13.5% private equity, 9.3% venture capital, 20% hedge funds, 12.3% fixed income and 11.1% real estate (with further breakdowns to sub-classes within these classes).
6. Factor Investment – 70% U.S. 60/40 and 30% equally to 15 factor premiums.
7. Dynamic Asset Allocation – either U.S. large cap stocks or U.S. aggregate bonds based on higher trailing 11-month returns, leveraged to 11.4% targeted volatility (retrospective volatility of U.S. 60/40) using 5-year rolling volatility (as in “Momentum in a Mean-variance Optimization Framework”).

They gather commercially and academically available asset return series. They extend series for 23 sub-asset classes and 15 long-short factor portfolios back to the end of 1925 based on similarity arguments and known-versus-extended return/volatility/correlation comparisons. They update all portfolio allocations monthly. They ignore rebalancing frictions, shorting costs, leverage costs, other costs of complex portfolio maintenance and administrative/management fees. Testing the seven strategies on actual and modeled data during December 1925 through December 2020, they find that:

Keep Reading

A subscriber asked about the performance of a strategy that each month allocates funds to pairs of exchange-traded fund (ETF) asset class proxies according to the term spread, as measured by the difference in yields between the 10-Year constant maturity U.S. Treasury note and the 3-Month U.S. Treasury bill (T-bill). Specifically:

• When the term spread is less than 0%, hold 50% Invesco DB Commodity Index Tracking Fund (DBC) and 50% SPDR Gold Shares (GLD), rebalanced monthly.
• When the term spread is 0% to 2%, hold 50% SPDR S&P 500 ETF Trust (SPY) and 50% iShares MSCI EAFE (EFA), rebalanced monthly.
• When the term spread is greater than 2%, hold 50% iShares 20+ Year Treasury Bond (TLT) and 50% Vanguard Real Estate Index Fund (VNQ), rebalanced monthly.

Also, how does the performance of this strategy (Term Spread Strategy) compare to that of a portfolio that each month allocates 50% to Simple Asset Class ETF Value Strategy (SACEVS) Best Value and 50% to Simple Asset Class ETF Momentum Strategy (SACEMS) equal-weighted (EW) Top 2. We begin the test at the end of June 2006, limited by SACEMS inputs. We ignore monthly rebalancing frictions for both strategies. Using monthly dividend-adjusted prices for the specified ETFs starting June 2006 and monthly gross returns for 50-50 SACEVS Best Value and SACEMS EW Top 2 starting July 2006, all through November 2022, we find that: Keep Reading

A subscriber requested evaluation of a strategy that seeks to exploit U.S stock market reversion after dips by temporarily applying margin. Specifically, the strategy:

• At all times holds the U.S. stock market.
• When the stock market closes down more than 7% from its high over the past year, augments stock market holdings by applying 50% margin.
• Closes each margin position after two months.

To investigate, we assume:

• The S&P 500 Index represents the U.S. stock market for calculating drawdown over the past year (252 trading days).
• SPDR S&P 500 (SPY) represents the market from a portfolio perspective.
• We start a margin augmentation at the same daily close as the drawdown signal by slightly anticipating the drawdown at the close.
• 50% margin is set at the opening of each augmentation and there is no rebalancing to maintain 50% margin during the two months (42 trading days) it is open.
• If S&P 500 Index drawdown over the past year is still greater than 7% after ending a margin augmentation, we start a new margin augmentation at the next close.
• Baseline margin interest is U.S. Treasury bill (T-bill) yield plus 1%, debited daily.
• Baseline one-way trading frictions for starting and ending margin augmentations are 0.1% of margin account value.
• There are no tax implications of trading.

We use buying and holding SPY without margin augmentation as a benchmark. Using daily levels of the S&P 500 Index, daily dividend-adjusted SPY prices and daily T-bill yields from the end of January 1993 (limited by SPY) through November 2022, we find that: Keep Reading

In response to a prior analysis (updated here), a subscriber asked whether adding a simple moving average (SMA) filter to “Simple Asset Class ETF Momentum Strategy” (SACEMS) assets, either before or after ranking them based on past returns, improves strategy performance. SACEMS each month picks winners from among the a set of eight asset class exchange-traded fund (ETF) proxies plus cash based on past returns over a specified interval. Since many technical traders use a 10-month SMA (SMA10), we test effectiveness of requiring that each asset pass an SMA10 filter as follows:

1. Baseline – SACEMS as presented at “Momentum Strategy” (no SMA10 filter).
2. Apply an SMA10 filter after asset ranking (SACEMS R-F) – Run Baseline SACEMS and then apply SMA10 filters to dividend-adjusted prices of winners. If a winner is above (below) its SMA10, hold the winner (Cash).
3. Apply an SMA10 filter before asset ranking (SACEMS F-R) – If a SACEMS asset is above (below) its SMA10, apply SACEMS ranking rules to it (exclude it from ranking). If there are not enough ranked assets to populate multi-position SACEMS portfolios, put the positions in Cash.

We focus on compound annual growth rates (CAGR), annual Sharpe ratios and maximum drawdowns (MaxDD) of SACEMS Top 1, equally weighted (EW) Top 2 and EW Top 3 portfolios. To calculate Sharpe ratios, we use average monthly 3-month U.S. Treasury bill (T-bill) yield during a year as the risk-free rate for that year. Using monthly dividend-adjusted closing prices for the asset class proxies and the (T-bill) yield for Cash over the period February 2006 through November 2022, we find that: Keep Reading