Stock Investment Strategies
James O’Shaughnessy’s What Works on Wall Street looks at the most effective long-term investment strategies on Wall Street. It shows that selecting stocks using rational methods can beat the simple strategy of indexing to the S&P 500. Empirical evidence reveals that stocks selling at deep discounts to cash flow, sales, and earnings consistently beat the market in the long run. Few investors however are capable of rigorously sticking with such a strategy during market turmoil. The key to long-term success is an unwavering disciplined implementation of an investment strategy. That is why the S&P 500 consistently beats 70% of traditionally managed funds. While managers change their style over time and rotate between funds, the S&P 500 never varies from its systematic bet on large capitalization stocks.
Quantitative stock-picking models created from large datasets consistently make better predictions than human forecasters. That’s because models consistently apply the same criteria every time. Active managers on the other hand prefer gut reactions and colorful stories over impersonal statistics. Complicated explanations and ornate anecdotes are more convincing to humans than simple theories. We additionally overweight recent information to the more distant past. Despite the widespread dissemination of O’Shaughnessy’s now well-known investment strategies, opportunities to exploit them still remain profitable given humanity’s persistent irrationality.
Rules of the Game
Stocks will continue to be mispriced in the market as long as people let emotion cloud their judgment. This provides opportunities for those who make investment decisions based on methodical, scientific methods. The strategy should not be the product of data mining (i.e. finding relationships that are the result of chance). There should be a sound theoretical, economic, or common sense explanation for why it works. The best way to confirm strategies is to test them on different periods or in different markets. Bootstrapping can also be used, which involves randomly resampling selected sub-periods and retesting the strategy to confirm the relationships consistently exist.
O’Shaughnessy Asset Management relies on Standard & Poor’s Compustat (1963-2009) and CRSP (1926-2009) datasets when back-testing their strategies. The database lags quarterly data by three months and annual data by six months to avoid look-ahead bias. The universe of investable stocks is limited to those with market capitalizations in excess of $200M (adjusted for inflation). For back-testing purposes, portfolios are created 12 times per year, one in each month. Strategies are rebalanced annually. No trades are made throughout the year unless a stock goes bankrupt or is acquired.
(In real time, stocks are considered for removal if a company (1) fails to verify its numbers, (2) is charged with fraud, (3) restates numbers such that the company would not have qualified for purchase, (4) receives a takeover offer and the stock moves to within 95% of takeover offer price, (5) drops by 50% from date of purchase and is in the bottom 10% of all stocks for previous 12 months, or (6) cuts its dividend by 50% or more and is part of a dividend strategy.)
Back-tested portfolio strategies are compared on a risk-adjusted basis using the Sharpe ratio (using 5% as the risk-free rate). Only strategies that beat their benchmark on an absolute and risk-adjusted basis are considered for implementation. Downside risk is measured using the semi-standard deviation below zero or the Sortino ratio (using 10% as the required rate). Maximum declines are included so investors can decide if they’re capable of stomaching the volatility of any particular strategy. Bid/ask spreads and transaction costs are not included in back-tested results as each investor faces a different situation. As a general rule of thumb, bid/ask spreads on small-cap trades average 0.5% and large-cap trades average 0.15%. Many online brokers allow individuals to trade any number of shares for a flat $7 commission.
Ranking Stocks by Market Capitalization
Several portfolios are constructed looking only at stocks by market capitalization between 1926 and 2009. All Stocks are comprised of stocks with market capitalizations in excess of a deflated $200M. Large Stocks are those with a market capitalization greater than the dataset average (usually top 17% of entire dataset). Small Stocks have market capitalizations greater than a deflated $200M but less than the database average. Market Leaders are nonutility stocks with market capitalizations greater than the average (but from a dataset excluding stocks with market capitalizations less than $50M), greater than average shares outstanding, greater than average cash flows, and sales 50% greater than the average stock. As with all tests in this book, stocks are equally weighted and all dividends are reinvested.
Academic studies nearly unanimously find that small stocks do significantly better than large ones in the long run. This is confirmed here. However, since it is nearly impossible to purchase micro-cap stocks (market capitalizations less than deflated $25M) the advantage of smaller-cap stocks is not of the magnitude of other studies. One explanation for why small- and mid-cap stocks outperform large-cap stocks is that they are less efficiently priced. This is because they are not adequately covered by analysts given their sheer number. Small-cap performance is virtually indistinguishable from large-cap when risk is taken into account (see Sharpe ratio).
Market Leaders handily beats the S&P 500, a large-cap index. It also appears to perform relatively well during bear markets. One possible reason is that it includes foreign-domiciled companies in the form of American Depository Receipts (ADRs) rather than solely US-based companies. Another explanation is that it is equally-weighted rather than cap-weighted. We’ll see later in this book that investors should ultimately concentrate on stocks in the All Stocks group.
It is common wisdom on Wall Street that stocks with low price-to-earnings ratios are a bargain. High PE ratios are the consequence of strong earnings growth expectations by the market, which are generally thought to be unrealistic. O’Shaughnessy tests the results of investing in low PE stocks by assuming an investor buys the first 10% of stocks (decile 1) with the highest earnings-to-price ratios (i.e. the inverse of the PE ratio).
On a rolling 10-year basis, low PE stocks historically beat All Stocks 99% of the time. While a 10-year outlook is unconventional, it highlights the importance of a long-term perspective (the average investor’s time horizon is less than five years). It would be difficult, after all, for an investor to passively watch her portfolio plunge by 59% between 2007 and 2009. It would have been even more catastrophic however for that same investor to abandon her low PE strategy and invest in T-bill at the end of the decline, thereby missing out on the subsequent recovery. The key to success is finding the right underlying investment strategy and slavishly sticking with it through everything.
The data also confirms that buying high PE stocks is a foolhardy endeavor. Investing in the highest PE ratios (decline 10) yielded returns worse than U.S. T-bills. The results were even grimmer after adjusting for risk. The Sharpe ratio was 0.02.
EBITDA to Enterprise Value
Many investors believe EBITDA/EV is a better ratio to assess value than the PE alone since it is agnostic to a company’s financial leverage, capital expenditure, and tax rate. Enterprise value is calculated as the market value of common and preferred equity, plus debt and minority interest, less the market value of any associates and all cash and cash equivalents.
Empirical evidence suggests that, on average, high EBITDA/EV stocks outperform All Stocks. There were however five year periods where the EBITDA/EV strategy performed poorly relative to its benchmark. Sticking with such an investment strategy after it had underperformed would have been both extremely difficult and extremely wise. It is therefore important to make a long-term commitment and remember that the ups and downs of a strategy are part of the bargain.
Another advantage of stocks with the best EBITDA/EV is that they offer downside protection, thereby softening the blow of serious bear markets. The worst EBITDA/EV stocks, on the other hand, earn less than 30-day U.S. T-bills and tend to outperform All Stocks only in the middle of a stock market bubble.
Stocks with high EBITDA/EV earn the best absolute return over the 1963-2009 period relative to all other value ratios. However, ongoing research suggests that the ranking of value factors continuously changes over time. Enterprise value can also be used to create additional ratios that perform similarly well. For the All Stocks universe, free cash flow to enterprise value and sales to enterprise value had annual compounded returns of 16.10% and 15.79%, respectively.
Price-to-Cash Flow Ratios
Another measure for determining whether a stock is cheap or not is the price-to-cash flow ratio. Cash flow is defined as net income plus depreciation and other non-cash expenses. Some investors prefer using cash flow ratios to find bargain stocks because cash flow is traditionally more difficult to manipulate than earnings. Utility stocks, however, should be excluded since they frequently show up and would create a bias toward one industry if not adjusted out. Given Compustat’s ranking function, O’Shaughnessy actually looks at stocks with the highest cash flow-to-price ratios, the inverse of the price-to-cash flow ratio.
To reiterate a conclusion made in regards to other value ratios, it is always best to bet on factors that generate the best long-term results. This is especially true when those factors are underperforming and fall out of favor among investors.
Long-term data confirms that you should avoid high price-to-cash flow stocks. Actually avoiding your favorite stocks with high price-to-cash flow ratios in real time, however, may prove difficult. Although high price-to-cash flow stocks perform well in speculative markets, their long-term results are abysmal. Once again, the highest decile stocks generated returns less than T-bills.
An alternative to net cash flow is free cash flow, defined as net cash flow minus capital expenditures, dividends and preferred dividends. While the traditional net cash flow number performs marginally better, free cash flow will sometimes work better in multifactor models where the interaction effect with other factors comes into play.
In the previous edition of What Works on Wall Street, the price-to-sale ratio (PSR) was the single-best value factor. In the updated edition, EBITDA/EV was the winning factor. The fact that just a few years can change the results of findings highlights the challenge of relying on a single factor. More broadly inclusive composited value factors, on the other hand, are less subject to tumultuous changes over a short period of time.
Investors prefer low PSR stocks for the same reason they like low PE stocks. They believe they’re getting a bargain. Low PSR stocks from the All Stocks and Large Stocks universes experienced their worst declines during the bear market of 2007 through early 2009, losing 66% and 60% – respectively. Between 1964 and 2009, the lowest PSR stocks lost more than 20% of their value on seven occasions. Anyone pursuing a low PSR strategy should keep this in mind.
Decile analysis indicates that decile 2 performs the same (or minimally better) than decile 1. Investors are therefore best off concentrating on the 20% of stocks with the lowest PSRs. Regardless of the stock market environment, high price-to-sale stocks rarely post positive returns.
Many investors believe that price-to-book value ratios are more important than PE ratios when looking for a bargain since earnings can easily be manipulated. Its popularity further stems from the fact that Ben Graham made it a central factor in his security analysis and Fama-French used it in their three-factor model.
Historical evidence suggests that stocks with low price-to-book value ratios work well over long periods of time, but there are long sub-periods where the opposite is true. In other words, performance is somewhat erratic depending on which sub-period is analyzed. This is because stocks with low price-to-book value ratios tend to be weak, risky companies in distress. They therefore tend to take a beating during periods of economic depression. This pattern highlights the importance of analyzing factors over long time horizons.
Based on decile analysis, investors are better off picking stocks ranked at the lowest 30% by price-to-book value rather than concentrating on the 10% of stocks that are cheapest. Stocks in deciles 2 and 3 primarily earned the advantage because decile 1 stocks suffered significantly in the 1930s and then again during the most recent crash of 2007-2009. Over the 84-year time period, the lowest price-to-book value stocks from the All Stocks universe lost 20% or more on 13 separate occasions.
Since one factor can underperform over a period of time, the implication is that investors should not rely on a single factor. Investors should instead strengthen the overall efficacy of their approach by combining multiple factors.
Investors like picking stocks that pay high dividends since dividends historically account for more than half a stock’s total return. The data suggest that investing in the decile of highest-yielding stocks does in fact outperform the All Stocks universe. However, the group also suffers from deeper declines, suggesting that stocks with the highest yields might exhibit additional risks. This implies that a high dividend yield may not be an ideal metric to use on its own when making an investment choice.
Decile analysis suggests that investors are better off selecting broadly from higher yielding stocks rather than focusing exclusively on stocks with the absolute highest yields. Further examination also indicates that capitalization matters. Companies from the Large Stocks universe tend to recover more quickly from their losses. This is because larger companies generally have stronger balance sheets and longer operating histories.
It is also worth paying attention to a payout ratio. Theory suggests that stocks with high payout ratios are poor investments because the companies are not reinvesting in their business. This is confirmed by the data. The results indicate that stocks with both the highest dividends yields and the highest payout ratios should be avoided.
Changes to a company’s dividend policy also affect performance. Companies that cut their dividend lost to their benchmarks, on average, by 3.6% in the following year. They underperformed by 5.1% when they suspended dividends altogether. Conversely, stocks that increased their dividends beat their benchmark by 4.5% the subsequent year. Stocks that started to pay a dividend for the firm time outperformed by 9.2%.
Buyback yields are not tracked by investors to the same degree as dividend yields. That’s a shame. They are an important factor to look at when determining a stock’s relative attractiveness. Share repurchases are increasingly substituted for dividends as the method for distributing earnings. The buyback yield is determined by comparing shares outstanding today with those outstanding one year earlier.
Companies that issue additional shares as opposed to buying them back have a negative yield. These stocks do not make good investments. They perform significantly worse than their benchmark.
The full decile analysis of the buyback yield is unusual because many companies neither issue nor repurchase stock over several months. This causes some of the middle deciles to be clumped together. As a result, the top eight deciles end up beating the All Stocks universe. Only deciles 9 and 10 dramatically underperform. Nonetheless, the significantly higher returns for decile 1 indicate that investors should focus only on stocks in the top 10% of buyback yields.
Shareholder yieed is the sum of a stocks’ dividend yield and buyback yield. It represents the proportion of total cash a company pays out to shareholders. While empirical evidence shows that the compound average annual return of total shareholder yield is less than the buyback yield itself, risk – as measured by the standard deviation of return – is actually lower. This translates to a higher Sharpe ratio.
In fact, the decile 1 portfolio lost less than 20% during the collapse of the tech bubble between 2000 and 2003. Other market declines are similarly muted compared to the dividend yield and buyback yield portfolios. The full decile analysis shows deciles descending uniformly in order, with the highest shareholder yield generating the best returns and the lowest earning the worst.
Empirical research has revealed that accounting variables can help investors make better stock selections. The following list shows several that offer predictive evidence of a stock’s future performance.
- Accruals-to-Price: The accruals-to-price ratio is a proxy for the quality of a company’s earnings. A high ratio might indicate that a company is accruing sales that are ultimately bogus, thereby increasing the likelihood of negative earnings surprises. Back-testing confirms that stocks with the lowest accruals-to-price ratios outperform the market.
- Assets-to-Equity: The assets-to-equity ratio measures a company’s use of leverage. Firms with low ratios are financing assets through cash flow whereas high ratios indicate a firm is taking on substantial debt to finance operations. Evidence suggests that stocks with the greatest leverage should be avoided. However, the decile analysis shows that some leverage is actually beneficial. Decile 7 was the strongest performer (11.56%).
- Asset Turnover: Asset turnover is defined as a company’s sales divided by its average total assets. It measures how efficiently a firm manages its assets. Stocks with the highest asset turnover ratios beat the market while stocks with the lowest ratios underperform it. The performance across deciles however is not very consistent. This means lower asset turnover ratios do not necessarily need to be avoided at all costs.
- Cash Flow-to-Debt: The cash flow to debt ratio (or coverage ratio) measures a company’s ability to satisfy its debt obligations. While the market severely punishes companies with the lowest ratios, it also rewards those that are moderately aggressive. Decile 6 was the strongest performer during the period of study (13.30%). A devastating bear market is likely to follow when decile 10 performs better than the market for any length of time.
- Debt-to-Equity: The debt-to-equity ratio provides another measure of a company’s financial leverage. Empirical evidence shows that it’s not much use on its own in helping investors in selecting the best stocks.
- External Financing: Academic research suggests that companies taking on a great deal of external financing (i.e. debt and issuance of shares) will face weak future stock returns. This is measured by cash flow from financing divided by average assets. Over the 38 years of study, the 10% of companies with the highest external financing actually lost money.
- Change in Debt: Companies with the highest percentage change in debt do substantially worse than the market. However, the most conservative companies (those in decile 10) are outperformed by those that are fairly aggressive with debt (deciles 4-8).
- Depreciation–to-Capital Expense: Companies that write down their physical assets faster than the pace at which they replenish them are considered conservative. More aggressive companies delay writing down their equipment, which can eventually produce negative earnings. Empirical evidence suggests that only deciles 9 and 10 really suffer. Deciles 2, 3 and 4 perform the best. The weaker performance of decile 1 might indicate those companies are backing off on investment, thereby hurting future growth prospects.
- Change in NOA: Net operating assets (or NOA) are a company’s operating assets less operating liabilities. Substantial increases in NOA may suggest accounting profits are outpacing cash profits. Historical research confirms that companies with the largest change in NOA do significantly worse than the market. Deciles 5-9 are the best performing.
- TATA: Total accruals to total assets (or TATA) sever as a proxy for income quality. A company’s earnings are considered more sound the lower the TATA ratio. Stocks with the lowest ratios do significantly better than the market. Stocks with the highest ratios are consistent underperformers.
- TAAA: The total accruals to average assets (or TAAA) ratio provides an even stronger signal of which companies to void. Quarterly data is used to determine the average assets.
- Composited Accounting Ratio: Research shows that combining several accounting variables is superior to relying on individual ones when looking for stocks with higher earnings quality. The combined group of factors includes (1) TATA, (2) Change in NOA, (3) TAAA, and (4) Depreciation to Capital Expense. The composited group generates better, more consistent overall results.
Combining Value Factors into a Single Composite Factor
Academic studies indicate that using different factors simultaneously provides better returns than single factors alone. This is especially true since the single best-performing factor continues to change over time. O’Shaughnessy combines the following factors into a single master composite: (1) Price-to-Book, (2) Price-to-Earnings, (3) Price-to-Sales, (4) EBITDA-to-Enterprise Value, and (5) Price-to-Cash Flow.
In order to determine the decile rankings, each factor (for each stock) is assigned a percentile ranking on a scale of 1 to 100. For example, a stock that has a PE ratio in the lowest (highest) 1% of the universe receives a rank of 100 (1). A similar convention is followed for each of the factors. (Rankings are reversed in instances where higher is better). If a value is missing for a factor, a rank of 50 is assigned. After all the factors are ranked, stocks are assigned an overall cumulative ranking by adding up the five factor rankings.
Stocks in decile 1, which comprise the best overall scores, earned an average annual compound return better than any of the single value factors studied. They also provided better downside protection relative to the All Stocks universe. In a separate value composite that also included shareholder yield, compounded returns improved 12 basis points (17.30%) and the standard deviation declined by 99 basis points (17.10%), translating to a Sharpe ratio of 0.72.
If the portfolio is narrowed down to just 25 or 50 stocks the performance is slightly better, but volatility is consequently greater. Sharpe ratios for the top 25- and 50-stock portfolios are 0.59 and 0.63, respectively. For investors looking to short the bottom 25 or 50 stocks, additional strategies such as moving averages and stop-loss orders should be incorporated since said stocks can perform exceedingly well during irrational market bubbles.
Additional commentary can be found on the O’Shaughnessy Asset Management (OSAM) website (here).