Cornerstone Growth – Strategy Introduction and Out of Sample Results

Last week, I wrote about the cornerstone value strategy from James O’Shaughnessy’s first edition of What Works on Wall Street, published in 1996. This week I’ll introduce the cornerstone growth quantitative investing strategy and show out of sample backtesting results, net of taxes and fees. The strategy buys the top 50 stocks (by 12-month price performance), rebalanced annually, that have the following characteristics:

  • 5 consecutive years of earnings growth
  • Price-to-sales ratios of less than 1.5
  • Have a market cap larger than $150 M (in 1996 $, about $235 M today), which he calls “All Stocks”

As of October 23, 2017, 3645 of 6084 stocks in the Compustat database (~60%) are large enough to be considered.

Cornerstone Growth vs Large Stocks, Nominal Returns (12/31/51 - 12/31/94)
Cornerstone Growth vs Large Stocks, Nominal Returns (12/31/51 – 12/31/94)

Backtesting results were promising (and significantly better than the cornerstone value strategy), although these numbers do not reflect transaction costs, taxes, or slippage.

The out of sample backtest below includes 0.25% slippage for all transactions and are shown for various starting capitals, trade fees and tax brackets (with a 1-year rebalance period, we can sell (winning stocks) on day 366 and take advantage of lower long-term capital gains tax rates). For losing stocks, you typically want to sell them on day 364 to be able to use them as short-term capital losses, which can be used to offset any short-term capital gains you have, which are taxed at a higher rate (therefore, offsetting them is advantageous; this is one method of tax loss harvesting). Although this is a good idea for implementation of this strategy, tax loss harvesting was not utilized in the backtest below.

The backtest is also composited weekly, meaning each year’s return is the average return of 52 different portfolios. So if the return for the strategy in 1999 was 10%, that’s the average of 52 portfolios — 1/1/1999 to 1/1/2000, 1/8/1999 to 1/8/2000, 1/15/1999 to 1/15/2000, etc.

The benchmark this time is the SPDR S&P 500 ETF SPY (which includes dividends so the returns are total returns). Returns are shown as excess returns over buying and holding SPY.

Cornerstone Growth vs SPY, Excess Total Returns (01/09/99 - 10/23/17)
Cornerstone Growth vs SPY, Excess Total Returns (01/09/99 – 10/23/17)

With no transaction costs, the amount of starting capital doesn’t affect the excess return. If you’re paying $4.95 or $6.95 per trade, you would have needed to start with at least $10,000 to outperform buying and holding SPY. If you’re paying $19.95 a trade, you’d have needed $15,000 to start depending on your tax bracket.

The top excess return shown (for $0 trades and 10/15% tax bracket) resulted in an excess return of 413%, turning $15,000 into an astonishing ~$113,000 over ~18 years instead of ~$42,000 for buying and holding SPY.

Backtesting also indicates a slightly smaller drawdown (~50%) for the cornerstone value strategy than SPY (~55%) over this time period.

Cornerstone Growth vs SPY, Nominal Returns & Risk Statistics (01/02/99 - 10/23/17)
Cornerstone Growth vs SPY, Nominal Returns & Risk Statistics (01/02/99 – 10/23/17)

Based on the weekly composited portfolios, this strategy performs better during down markets (when SPY is decreasing) than up markets.

Cornerstone Growth vs SPY, Weekly Composited Portfolio Nominal Returns (01/02/99 - 10/23/17)
Cornerstone Growth vs SPY, Weekly Composited Portfolio Nominal Returns (01/02/99 – 10/23/17)

So the key takeaways:

  • Free trades are awesome (like with Robinhood).
  • For the most part (with the exception of low starting capital/high transaction fees/high tax bracket), cornerstone growth outperformed buying and holding SPY out of sample (1999 to 2017), and significantly outperformed the cornerstone value strategy.

As of 10/23/17, the top 5 stocks for the cornerstone growth strategy (highest 12-month price performance meeting the criteria from above) are LGIH, SKX, THO, CDW, & CBZ.

This strategy screen is available to my group on Portfolio123, where you can customize it or implement it as is. If you just want the stock picks, you can subscribe on Patreon below.

Cornerstone Value – Strategy Introduction and Out of Sample Results

In the first edition of What Works on Wall Street, published in 1996, James O’Shaughnessy outlines the framework for the cornerstone value quantitative investing strategy. The strategy buys the top 50 yielding stocks, rebalanced annually, among a universe of stocks he calls “market leaders.” (Although dividend yield is no longer a good value indicator for stocks).

Market leaders are defined by the following:

  • They’re large (their market cap is greater than the universe average). These stocks are also called “Large Stocks”, which he used as his benchmark for this strategy.
  • They have more common shares outstanding than the universe average.
  • They have cash flows that exceed the universe average.
  • They have sales that are >=1.5x the universe average.
  • They are not a utility.

His universe was simply the Compustat database (the same used by Portfolio123). As of October 10, 2017, 441 of 6088 stocks in the database (~7%) are “market leaders.”

From the 1st Edition of What Works on Wall Street
Cornerstone Value vs Large Stocks, Nominal Returns (12/31/51 – 12/31/94)

Backtesting results were promising, although these numbers do not reflect transaction costs, taxes, or slippage.

The out of sample backtest below includes 0.25% slippage for all transactions and are shown for various starting capitals, trade fees and tax brackets (with a 1-year rebalance period, we can sell (winning stocks) on day 366 and take advantage of lower long-term capital gains tax rates). For losing stocks, you typically want to sell them on day 364 to be able to use them as short-term capital losses, which can be used to offset any short-term capital gains you have, which are taxed at a higher rate (therefore, offsetting them is advantageous).

The backtest is also composited weekly, meaning each year’s return is the average return of 52 different portfolios. So if the return for the strategy in 1999 was 10%, that’s the average of 52 portfolios — 1/1/1999 to 1/1/2000, 1/8/1999 to 1/8/2000, 1/15/1999 to 1/15/2000, etc.

The benchmark this time is the SPDR S&P 500 ETF SPY (which includes dividends so the returns are total returns). Returns are shown as excess returns over buying and holding SPY.

Each color represents a different starting capital
Cornerstone Value vs SPY, Excess Total Return (01/09/99 – 10/14/17)

With no transaction costs, the amount of starting capital doesn’t affect the excess return. If you’re paying $4.95 or $6.95 per trade, you would have needed to start with at least $10,000 to outperform buying and holding SPY. If you’re paying $19.95 a trade, you’d have needed $15,000 or $20,000 to start depending on your tax bracket.

The top excess return shown (for $0 trades and 10/15% tax bracket) resulted in an excess return of 236%, turning $15,000 into ~$72,000 over ~18 years instead of ~$42,000 for buying and holding SPY.

Backtesting also indicates a slightly larger drawdown (~63%) for the cornerstone value strategy than SPY (~55%) over this time period, which brings into question whether someone would be able to psychologically tolerate this strategy.

Cornerstone Value vs SPY, Nominal Returns & Risk Statistics (01/09/99 - 10/14/17)
Cornerstone Value vs SPY, Nominal Returns & Risk Statistics (01/09/99 – 10/14/17)

Based on the weekly composited portfolios, this strategy performs better during down markets (when SPY is decreasing) than up markets…

Cornerstone Value vs SPY, Weekly Composited Portfolio Nominal Returns (01/09/99 - 10/14/17)
Cornerstone Value vs SPY, Weekly Composited Portfolio Nominal Returns (01/09/99 – 10/14/17)

…so using a market timer to exit the market improves the drawdown but erases the alpha.

Cornerstone Value vs SPY, Nominal Returns & Risk Statistics w/ 50/200 day SMA Cross Market Timer (Exit) (01/09/99 - 10/14/17)
Cornerstone Value vs SPY, Nominal Returns & Risk Statistics w/ 50/200 day SMA Cross Market Timer (Exit) (01/09/99 – 10/14/17)

However, using the same market timer to add a low beta filter (if the market timer is signaled at a rebalance date, only repurchase stocks if their 3-year beta is in the bottom 10% of stocks in the universe) may preserve alpha while decreasing drawdown.

Cornerstone Value vs SPY, Nominal Returns & Risk Statistics w/ 50/200 day SMA Cross Market Timer (Lowest Decile Beta Filter) (01/09/99 - 10/14/17)
Cornerstone Value vs SPY, Nominal Returns & Risk Statistics w/ 50/200 day SMA Cross Market Timer (Lowest Decile Beta Filter) (01/09/99 – 10/14/17)

So the key takeaways:

  • Free trades are awesome (like with Robinhood).
  • For the most part (with the exception of low starting capital/high transaction fees/high tax bracket), cornerstone value outperformed buying and holding SPY out of sample (1999 to 2017).
  • This strategy can be implemented on its own or in conjunction with other strategies I’ve discussed, such as market timing.

As of 10/14/17, the top 5 stocks for the cornerstone value strategy (highest yielding among market leaders) are ETP, CTL, MBT, EPD, & ETE.

This strategy screen is available to my group on Portfolio123, where you can customize it or implement it as is. If you just want the stock picks, you can subscribe on Patreon below.