De-Risking the Cash Signal from Corporate Balance Sheets

Abstract: Cash to market value (CMV) is a traditional value factor that is designed to capture the amount of cash reserves a company has relative to its market cap. Historically, it was assumed that a higher cash reserve is a reflection of positive cash flows, indicates stability or at least a baseline level of liquidity, and that a higher proportion of CMV would indicate a discounted buying opportunity. However, research by the Quantitative Equity team at Glenmede Investment Management (GIM) has indicated that high proportions of cash or current assets to market value can actually be a proxy for higher volatility. Companies can hold additional cash for various reasons, but many are associated with higher risk (e.g., lower profitability, smaller cap or riskier, more cyclical business models). Rather than correct for these individual metrics, we found that a simple volatility adjustment for sector-neutral cash/market cap delivered better results, enhancing returns and lowering tracking errors.

Is Cash a Return Signal?

There are conflicting viewpoints on the return signal of cash reserves in current academic and practitioner literature. Michael Jensen’s 1986 free cash flow theory asserts that managers with access to large free cash flows are more likely to engage in negative net presence value (NPV) projects.¹ In a more direct 2004 study on the effect of cash upon returns, Robin M. Greenwood finds aggregate corporate cash holdings to be a negative predictor for future market returns.² Greenwood asserts that this is due to the manager’s motivation to time equity and debt markets.

On the other hand, there is empirical evidence from W.H. Mikkelson and M.M. Partch 2003’s study that found companies that hold large amounts of cash have superior operating performance when controlling for size and industry.³ Likewise, research articles from sell-side analysts suggest using cash holdings relative to a company’s market cap as a measure of safety. In 2014, Deutsche Bank’s quantitative research team pointed out that technology companies were better positioned than in the internet bubble in part because of a higher cushion of CMV.⁴ Shepherd 2007 shows that financial slack, as measured by CMV, not only improved operating performance amongst small firms, but it was not appropriately priced in the market, which led to more upside.⁵

Our internal research has shown CMV and current assets to market value to be strong return predictors over time. Since 1982, the top quintile of companies for CMV have provided an annualized excess return of 2.18% compared to the Russell 1000 benchmark, with the results declining monotonically (Exhibit 1). The excess return is even stronger at 4.01% when comparing high CMV stocks to the Russell 2000 benchmark, which is consistent with Shepherd’s findings.

At GIM, we have been using CMV as a factor in our models since Q3 2014. During this time, the top quintile of this factor has produced an excess return of 0.63% compared to the Russell 1000 benchmark and 8.03% when compared to the Russell 2000 benchmark. The standard deviations of the top quintiles are 6.26% higher compared to the Russell 1000 benchmark and 10.9% higher compared to the Russell 2000 benchmark, respectively. We have also been using Current Assets to Market Value as a stock selection factor since 2011 with similar results.

Our rationale for these excess returns is similar to the findings above – namely, that companies can operate more effectively with financial slack – but also that companies trading at a discount to their cash holdings may be overlooked by the market. Indeed, CMV, using a sector neutral basis, has about a 30% correlation to traditional value factors such as price to book as of November 30, 2022.

Cash to Market Value –Riskier than Other Value Factors

Regarding risk, our results somewhat contradict the academic literature. We find significantly higher volatility for high CMV firms than low CMV firms; from Exhibit 1, the top quintile has a standard deviation of 25.73%, while the bottom quintile has a standard deviation of only 17.46%.

There is some explanation from the comparison of cash specifically to market capitalization, a relationship between cheaper valuations and higher volatility is common to value factors. In other words, value factors tend to be more volatile on the long side and less volatile on the short side. Exhibit 2 shows correlation between traditional value factors and rolling three-year volatility. The moderately positive correlation between volatility and traditional value factors (such as price to book, price to sales, and free cash flow yield) is well known, and so is the notable exception, wherein dividend yield is negatively correlated with risk.

Despite these general relationships between value and volatility, the strong correlation of CMV seems to be an outlier. Over the last 40 years, there is an average correlation of about 0.75 between CMV as a factor versus volatility. Exhibit 3 highlights the fact that this correlation is not a recent phenomenon but has been true over multiple market cycles, with the relationship generally weakening during recessions or midcycle slowdowns. Presumably this relationship becomes clouded with other financial risks in the equity markets during such time periods. Our goal in this paper is to dissect this risk relationship and examine whether we can adjust the CMV factor to create a more stable return signal without compromising the historical excess return.

Cash Metrics Need to Be Sector Neutral

Before we address the risk metrics directly, we need to state that sector neutralization is necessary. Due to their business operating models, financials and technology companies tend to hold more cash than other sectors. Together, they make up about 61% of the top quintile of CMV within the Russell 1000 Index (Exhibit 4). If our goal is to address cash differences between companies with similar operating models and different levels of cash, rather than contemplating the merits of being in one line of business versus another, then we need to normalize for these differences. We have chosen to rank stocks on a sector-neutral basis, but many academic papers have instead just removed financials from the sample.

The Multifaceted Nature of Risk in Cash to Market Value

Companies hold cash for a myriad of reasons. Based on our analysis, most of these reasons are indicators of risk or liquidity needs within the businesses themselves. In other words, companies that hold more cash tend to do so because they have riskier business lines, are less profitable with volatile cash flows, or they are smaller and have more difficulty accessing the capital markets. Exhibit 5 shows some of the traits of high- versus low-cash companies within the Russell 1000 Index.

The statistics above show that not only are companies with higher cash levels smaller on average, but they also have less consistent earnings, worse interest coverage, less reinvestment, lower profitability, and higher market beta than low-cash firms. Each of these biases is intuitive:

• A lower market cap bias points to the fact that it can be more difficult for smaller companies to receive external financing, so they keep high levels of cash on hand.
• A lower reinvestment rate of cash-rich companies shows that these companies are not putting profits back into their operations but rather holding cash on their balance sheets.
• A lower ability to pay off interest expenses suggests an elevated level of debt relative to market cap.

While each of these tendencies are themselves a proxy for increased risk, we find that they are distributed differently cross-sectionally and over time for high- and low-cash firms. Some metrics, such as market cap and volatility (Exhibits 6 and 7), show uniform trends over time and across the sample. Other metrics such as return on equity (ROE, Exhibit 8) show cyclical patterns. In periods of stress, such as the continuation of the tech bubble burst in 2001, the Global Financial Crisis (GFC) of 2009, and especially during the COVID onslaught period, the gap is large. In more normal periods, these differences close almost completely. Moreover, metrics such as leverage (Exhibit 9) show results that are not monotonic over the cross section: very high-cash firms are likely to have very high leverage, but low-cash levels are not necessarily associated with lower leverage.

While we will consider each of the risk factors in Exhibit 5 as an adjustment for CMV, we would naturally favor the most stable relationships, namely volatility and market capitalization. Additionally, there are theoretical reasons to favor these factors; for instance, Campbell (2005) points out that when predicting failure over longer horizons, market capitalization and equity volatility become the most significant.⁶

Volatility Adjustment as a Solution

As discussed previously, CMV, unadjusted, is a higher volatility value factor with many characteristics simultaneously adding to the risk profile of this factor. Exhibit 10 shows panel results from the top quintile of CMV, along with adjustments for reinvestment rate, leverage, beta, and rolling volatility.

On a risk-adjusted basis, the reinvestment rate adjustment shows the most improvement in Sharpe and information ratios. However, when looking to implement in a multifactor stock selection model, it is important to note that this improvement for the top quintile does not translate to the rest of the sample. The Sharpe ratio of the CMV factor adjusted for reinvestment rate on a long-short basis is weaker than the original factor alone. This is due to inconsistencies in the cross-sectional as those shown in Exhibit 9 for interest expense. Indeed, the long-short Sharpe ratio for leverage-based factors also appears to suffer. Moreover, while these adjustments improve risk-adjusted performance for the top quintile of the factor, they do so at the expense of reducing average returns.

Meanwhile, the use of a rolling volatility adjustment to CMV produces much more consistent results. It produces significant improvements in long-only Sharpe ratio, long-short information ratio (+30%), positive frequency, average return, and relative and absolute standard deviation. Exhibit 11 shows the time-series back test for CMV with volatility adjustment, indicating fewer significant negative dips than the unadjusted backtest in Exhibit 1, particularly around recessions, post internet bubble, and during the GFC.

Additionally, when looking at the subindustry level data of the unadjusted versus the volatility adjusted CMV, we found that volatility adjustments had a small but noticeable effect in the subindustry composition, as the largest decrease in weight was in biotechnology toward more diversified, less risky health care industries (see appendix for reference). This further illustrates our rationale that such an adjustment to this factor reduces its exposure from traditionally volatile businesses.

Additionally, when looking at the subindustry level data of the unadjusted versus the volatility adjusted CMV, we found that volatility adjustments had a small but noticeable effect in the subindustry composition, as the largest decrease in weight was in biotechnology toward more diversified, less risky health care industries (see appendix for reference). This further illustrates our rationale that such an adjustment to this factor reduces its exposure from traditionally volatile businesses.

Conclusion

While academic literature has often hypothesized that companies with more cash should be safer with lower returns, our empirical analysis indicates that small and large cap stocks with high CMVs have had higher returns and risk for over 40 years. Our work suggests that combining sector neutralization and a volatility adjustment can mitigate the risks without reducing returns. We have found that rolling volatility produces the most consistent improvement, but due to the myriad of risk factors that may be present in companies with high levels of cash, other adjustments may be appropriate if investors seek additional risk management.

Appendix