Interest Rates, Cash and Short-Term Investments Bektemir Ysmailov * * Doctoral Student at the College of Business, University of Nebraska-Lincoln, 730 N. 14th Street, Lincoln, NE 68588; phone: 402-472-3450. E-mail: bysmailov2@unl.edu. For helpful comments, suggestions, and discussions, I would like to thank David Denis, Kathleen Farrell, Geoffrey Friesen, John Geppert, Emre Unlu, and Liying Wang.
Interest Rates, Cash and Short-term Investments Abstract I document that two components of corporate liquid assets, cash and short-term investments, have followed strikingly different paths as a percent of assets since the 1970s. These patterns contrast with a monotonic increase in their sum often referred to as the corporate cash puzzle. To explain the difference in trends for cash and short-term investments, I revisit and find support for the overlooked predictions of the transactions model of liquidity management. High interest rates are associated with high short-term investments and low cash holdings while low interest rates are associated with low short-term investments and high cash holdings. Moreover, I find that the sum of cash and short-term investments is unrelated to interest rates directly. This result suggests that the traditional definition of cash (as the sum of cash and short-term investments) masks an important relation predicted by the seminal transactions model. JEL classification: G30, G31, E41 Keywords: Corporate cash, Interest rates, Transactions model, Short-term investments, Liquid assets
1. Introduction Corporate liquid assets have more than doubled as a percent of assets over the last 40 years in the U.S. Researchers have proposed numerous explanations for this trend. 1 However, the evolution of its two components, cash and short-term investments (STI), has received limited attention. This lack of attention is because the corporate cash literature has traditionally referred to total liquid assets as cash and thereby imposed an implicit restriction on its two components to have identical determinants. In this paper, I document that cash and short-term investments have followed strikingly different paths as a percent of assets since 1974 and utilize the overlooked predictions of the transactions demand for cash model about interest rates to explain these trends. The outstanding evidence regarding the relation between interest rates and corporate cash holdings is both scarce and inconclusive. In one of the most important modern papers on corporate cash, Bates, Kahle, and Stulz (2009) conclude that T-Bill yields are uncorrelated with cash holdings. On the other hand, Azar, Kagy, and Schmalz (2016) find that T-Bill yields are negatively related to cash among firms with at least ten years of data. Both Bates, Kahle, and Stulz (2009) and Azar, Kagy, and Schmalz (2016) define cash as the sum of cash and short-term investments, to which I refer as total liquid assets. Further, Gao, Whited, and Zhang (2017) document a hump-shaped relation between interest rates and cash (where cash is defined as in this paper, i.e. one of the two components of total liquid assets): it is positive when interest rates are relatively low and negative when interest rates are relatively high. This paper aims to advance the understanding of the link between interest rates and corporate liquidity management. Figure 1 shows the evolution of corporate liquid assets. The sum of cash and short-term investments has increased steadily - a well-known trend that has been extensively studied. A surprising pattern emerges when looking at the evolution of cash and short-term investments separately. From 1974 to 1986, corporate short-term investments have more than doubled as a percent of assets while the cash to assets ratio stayed largely constant. This was followed by an abrupt reversal: in 1988, cash became the largest component of liquid assets and, in 1989, the share 1 Among the proposed explanations are: a shift in the composition of the US public firms towards research and development (R&D) intensive firms and higher R&D spending ( Bates, Kahle, and Stulz, 2009; Begenau and Palazzo, 2017; Graham and Leary, 2016); increasing cash flow volatility (Bates, Kahle, and Stulz, 2009); a rise of a multinational firm whose earnings are trapped abroad (Foley, Hartzell, Titman, and Twite, 2007); increasing refinancing risk (Harford, Klasa and Maxwell, 2014); a rise in the intangible capital (Falato, Kadyrzhanova, and Sim, 2013); and changes in the cost of carrying liquid assets (Azar, Kagy, and Schmalz, 2016; Curtis, Garin, and Mehkari, 2017).
of cash was more than two times the share of short-term investments. Subsequently, the growth of the cash to assets ratio has outpaced that of the short-term investments ratio. Thus, Figure 1 illustrates two distinct trends for cash and short-term investments as a percent of assets which contrast with a monotonic increase in their sum often referred to as the corporate cash puzzle. The offsetting relation between cash and short-term investments on Figure 1 is suggested by the transactions model of liquidity management that was originally developed by Baumol (1952) and was later refined in Tobin (1956) and Miller and Orr (1966). According to this model, interest rates represent an opportunity cost of holding idle cash and it is higher for larger firms. Firms can minimize this opportunity cost by parking cash in an interest-bearing asset until it is needed. Therefore, the model predicts that interest rates and size are negatively related to cash but are positively related to investments in an interest-bearing asset. The existing empirical tests of the transactions model of liquidity management define cash as the sum of cash and short-term investments but do not specify a proxy for an interestbearing asset (see, e.g., Opler et al., 1999; Bates, Kahle, and Stulz, 2009; Graham and Leary, 2016). I define cash as the Cash and Cash Equivalents balance sheet account, a Compustat data item CH, which allows me to proxy for an interest-bearing asset using the Short-Term Investments account, a Compustat data item IVST, and to test the predicted positive relations between that asset and interest rates and size. Short-term investments satisfy two criteria outlined in Tobin (1956): they are not a medium of payment and they earn a yield. Importantly, a seminal transactions model allows cash and short-term investments to have common determinants with opposite signs: size and interest rates. Consistent with the transactions model, I find that size is negatively related to cash and is positively related to short-term investments. Although the corporate cash literature has previously reported a negative relation between cash and firm size, this is the first paper to document a positive relation between an interest-bearing asset and size. 2 I also find that cash flow is negatively related to cash but is positively related to short-term investments. One interpretation is that, all else equal, high cash flow firms have higher cash balances and, thus, have higher opportunity cost of holding 2 A number of papers that define cash as the sum of cash and short-term investments find that it is negatively related to size. For example, Opler et al., 1999; Bates, Kahle, and Stulz, 2009; Graham and Leary, 2016.
cash. To minimize this opportunity cost, firms park cash in short-term investments until it is needed. More importantly, my results indicate that interest rates (proxied by the three-month T-Bill rate) are significantly negatively related corporate cash and are significantly positively related to corporate short-term investments. My estimates suggest that a 14% decrease in the T-Bill rate from 1981 to 2014 explains approximately half of the actual increase in the cash ratio during that time. Given that the transactions model implies an offsetting relation between cash and short-term investments, the corresponding decrease in short-term investments is also accommodated by my estimates (see Figure 5). Additionally, I find that interest rates are unrelated to total liquid assets, consistent with Bates, Kahle, and Stulz (2009). Thus, changes in interest rates explain a large portion of the evolution of corporate liquid assets indirectly, by explaining changes in its two components. One alternative driver of the patterns observed in Figure 1 for cash and short-term investments is the change in the composition of the U.S. public firms. Several papers attribute the increase in corporate liquid assets since 1980 to the entrance of firms with high levels liquid assets (e.g., hi-tech, R&D intensive firms) rather than to within-firm changes in liquid assets demand (Bates, Kahle, and Stulz, 2009; Begenau and Palazzo, 2017; Graham and Leary, 2016). It is possible that the same hi-tech, R&D intensive entrants that pushed the liquid assets ratio higher have also caused the switch in 1986 from short-term investments to cash as the dominant form of liquid assets and the subsequent run-up in cash, to the extent that these new entrants hold most of their liquid assets in cash rather than in short-term investments. To rule out this alternative explanation, I utilize the findings of Graham and Leary (2016) who report that the effect of new entrants on the corporate liquid assets ratio is mitigated among the New York Stock Exchange (NYSE) listed firms. I find that the evolution of cash and shortterm investments as a percent of assets in the NYSE subsample closely resembles that of the whole sample (see Figure 2). Furthermore, the predicted relations between cash and short-term investments and interest rates are confirmed in the regression analysis, and are more pronounced among firms in the highest quintile of cash flow volatility, which are predicted to have higher sensitivity to interest rates by the transactions model (Miller and Orr, 1966). These results suggest that the changes in the composition of corporate liquid assets are not driven by the new entrants.
I subject my findings to a number of robustness tests. First, I control for the effect of inflation by using a deflated T-Bill rate as a proxy for interest rates. Second, I control for the real GDP growth, productivity and market volatility which have been previously found to influence corporate liquid assets (Graham and Leary, 2016). Third, I re-run the main regressions over a different sample period from 1980 to 2006 that was used in the seminal Bates, Kahle, and Stulz (2009) study. Fourth, I use the corporate bond yield as a proxy for interest rates to address the changes in the composition of liquid assets towards more risky and less liquid financial securities (Dunchin et al., 2017; Cardella, Fairhurst, and Klasa, 2016). My results are robust to all of these tests. The corporate cash puzzle refers to the increase in total liquid assets ratio since 1980, originally documented by Bates, Kahle, and Stulz (2009). How do the findings of this paper relate to the existing explanations of this phenomenon? My findings are in line with explanations based on the shift in the cross-section of firms contributing to the increase in total liquid assets. On the other hand, the results point to interest rates as the primary driver of the changes in the composition of corporate liquid assets and the run-up in cash since the late 1980s. I conclude by pointing out that what explains the increase in total liquid assets does not necessarily explain the increase in cash, and vice versa. In a closely related study, Azar, Kagy, and Schmalz (2016) link changes in interest rates to changes in the level of total liquid assets. They argue that a decrease in interest rates since the 1980 and the dwindling share of cash and checking deposits held by the non-financial sector in the aggregate economy (based the Fed Flow of Funds data) made it less costly to maintain liquid assets portfolios and, therefore, resulted in the increase of total liquid assets as a percent of assets. The evidence in this paper suggests that the negative relation between interest rates and total liquid assets is not consistent across different specifications and subsamples and is potentially spurious in the main sample because of new entrants that drove an increase in total liquid assets holdings. Gao, Whited, and Zhang (2017) document a hump-shaped relation between interest rates and cash: it is positive when interest rates are relatively low and negative when interest rates are relatively high. They conclude that cash should have decreased as a percent of assets since 1991 because of lower borrowing costs. My approach is different from Gao, Whited, and Zhang (2017) in that I study the behavior of the two key components of corporate liquid assets collectively
instead of in isolation. This approach helps to put the trade-off firms face due to changes interest rates in a broader perspective. Specifically, the analysis in Section 3.3 demonstrates the power of interest rates to explain both the increase in cash and the corresponding decrease in short-term investments. The remainder of the paper is organized as follows. Section 2 documents the evolution of cash and short-term investments ratios. Section 3 reviews the transactions demand for cash model and discusses my empirical approach. Section 4 presents univariate and regression results. Section 5 provides several robustness checks. Section 6 concludes. 1. The Evolution of Cash Holdings and Short-Term Investments I construct my sample by merging CRSP and Compustat databases from the Wharton Research Database (WRDS) for the period from 1974 to 2014. I exclude utilities (SIC codes 4900-4999) and financial firms (SIC codes 6000-6999), firms not incorporated in the United States, and firms with non-positive values of total assets and sales revenue. Missing explanatory variables reduce the main sample to 127,927 firm-year observations. I define cash as the Cash and Cash Equivalents balance sheet account, a Compustat data item CH, and short-term investments as the Short-Term Investments balance sheet account, a Compustat data item IVST, and their sum as total liquid assets, a Compustat data item CHE. According to the U.S. GAAP, Cash and Cash Equivalents include short-term, highly liquid investments that are readily convertible to known amounts of cash and that are so near their maturity that they present insignificant risk of changes in value because of changes in interest rates. Short-term Investments balance sheet account consists of marketable securities intended to be sold within one year (or the normal operating cycle if longer) and may include trading securities, available-for-sale securities, or held-to-maturity securities (if maturing within one year), as applicable. Figure 1 illustrates how cash, short-term investments, and total liquid assets evolved as a percent of total book assets for the whole sample. Total liquid assets grew steadily from 7.9% in 1974 to 22.5% of assets in 2014, a pattern consistent with prior literature. Distinguishing between cash and short-term investments reveals surprising trends. Until late 1980s, short-term investments
were growing as a proportion of assets while cash ratio stayed constant. The share of short-term investments was almost twice as high as the share of cash in 1986, 9.2% to 5.0%. Subsequently, cash became a dominant form of liquid assets, growing to 17.3% of assets while short-term investments declined to 5.2% in 2014. Thus, there was a lot of variation in the composition of corporate liquid assets between cash and short-term investments since the 1970s while total liquid assets absorbed these changes and increased on a steady basis. There are several motives for firms to hold cash. Precautionary motive says that firms hold cash to insure themselves from adverse cash flow shocks in the presence of costly external financing. On the other end of the spectrum is the speculative motive that states that firms hold cash buffers in order to take advantage of attractive investment opportunities when, again, external financing is costly. There is also a tax motive to hold cash that stems from the repatriation taxes U.S. firms face on their foreign earnings. The agency motive posits that entrenched managers prefer to hoard cash rather than payout and, as a result, build up large cash reserves. Finally, there is a transactions motive to hold cash under which firms balance the opportunity costs of holding cash with the costs of converting interest-bearing assets into cash to arrive at optimal cash balances. The next section and the rest of the paper focus on the transactions motive for holding cash for one main reason. The transactions model delivers two key predictions (discussed below) that directly imply changes in the composition of corporate liquid assets over time. However, I will control for precautionary, speculative, agency costs and tax motives to hold cash in all of the multivariate tests in line with prior literature. 2. Theoretical and Empirical Framework In this section, I will review the transactions model of liquidity management that was originally developed by Baumol (1952) and was refined in Tobin (1956) and Miller and Orr (1966). More specifically, I will point to the overlooked predictions of this model regarding the relation between cash and short-term investments, on the one hand, and interest rates and size, on the other. The following discussion refers to Tobin (1956).
3.1 Transactions Demand for Cash As a starting point, consider a hypothetical scenario: firm s receipts (cash inflows) and expenditures (cash outflows) are perfectly synchronized. In this case, firm s cash balance to satisfy its everyday transactions is zero. 3 In reality, cash inflows and outflows are not perfectly synchronized, for example, a firm might receive a payment for its products today but has to pay employee salaries in a week. This creates a need to hold transactions cash balances. Then, the question becomes: Why not hold transactions balances in assets with higher yields than cash, shifting into cash only at the time an outlay must be made?... The advantage of this procedure is of course the yield. The disadvantage is the cost, pecuniary and non-pecuniary, of such frequent and small transactions between cash and other assets (Tobin, 1956). An interestbearing asset that Tobin considers in his model is a risk-free bond, which is different from cash in two respects: (i) it is not a medium of payment and (ii) it earns an interest rate. The two main empirical predictions of this model are: The optimal share of bonds in a transaction balance varies directly, and the share of cash inversely, with the rate of interest because interest rates represent an opportunity cost of holding cash. In other words, interest rates are positively related to investments in an interest-bearing asset but are negatively related to cash holdings. Small transactors (firms) do not find it worthwhile even to consider holding transaction balances in assets other than cash; but large transactors may be quite sensitive to interest rate. Thus, size is negatively related to cash but is positively related to an interest-bearing asset. This relation is a result of (i) higher opportunity cost in dollar amount of holding cash for large firms and (ii) lower transactions costs (such as brokerage fees) for large firms of converting bonds into cash and vice versa due to economies of scale. The existing tests of this model in corporate finance define cash as the sum of cash and short-term investments and focus on the predicted negative relation between cash and size. Opler et al. (1999) and Bates, Kahle, and Stulz (2009) confirm the negative relation between firm size and total liquid assets and interpret it as consistent with the economies of scale. The conclusions regarding the negative relation between total liquid assets and interest rates are not uniform. Bates, 3 Notice that I am not considering precautionary, speculative, agency costs and any other motives to hold cash at this point but will control for these factors in multivariate tests.
Kahle, and Stulz (2009) use T-Bill rate as a proxy for interest rates and find that it is unrelated to total liquid assets. On the contrary, Azar, Kagy, and Schmalz (2016) find a negative relation between the T-Bill rate and total liquid assets. Since the above-mentioned studies do not define a proxy for an interest-bearing asset, they do not test the predicted positive relation between that asset and size. The same is true for the positive relation between interest rates and an interest-bearing asset. In the next subsection, I will address these shortcomings by defining new proxies for cash and an interest-bearing asset. 2.2 Empirical Approach I proxy for cash using the Cash and Cash Equivalents balance sheet account, Compustat data item CH, and for an interest-bearing asset using the Short-term Investments balance sheet account, Compustat data item IVST. Short-term investments satisfy two conditions for an interestbearing asset outlined in Tobin (1956): they are not a medium of payment and they earn a yield. In fact, Tobin used a risk-free bond to proxy for an interest-bearing asset, which at the time was a primary component of short-term investments (Jacobs, 1960). Note that if cash earns a yield, the predictions of the model would still hold as long as the expected yield on short-term investments is higher. Using the new proxies, the empirical predictions of the transactions model become as follows: (i) interest rates are negatively related to cash but are positively related to short-term investments; and (ii) firm size is negatively related to cash but is positively related to short-term investments. Thus, despite the traditional restriction on cash and short-term investments to have identical determinants, the seminal transactions model predicts that they have common determinants with opposite signs. Equipped with this framework, I turn to empirical tests in the next section.
3. Determinants of Cash and Short-Term Investments 3.1 Univariate Evidence In this subsection, I will provide univariate evidence regarding the predicted relations between cash and short-term investments and interest rates. I plot a 3-month T-Bill rate, used as a proxy for interest rates, as well as cash and short-term investments ratios in Figure 2. Since nominal interest rates and inflation are highly correlated, I also consider whether real interest rates better represent the opportunity cost of holding cash. I proxy for the real interest rates using a difference between a 3-month T-Bill rate and inflation. Although the fit is not perfect, the trendlines are largely consistent with the model s predictions. During the high interest rate environment of the late 70s and early 80s, short-term investments holdings were consistently higher than cash holdings as a proportion of assets. Subsequently, when interest rates declined and remained at a relatively low level, cash became the dominant form of corporate liquid assets. 3.2 Empirical Design Next, I turn to multivariate analyses of the observed relations between interest rates and liquid assets. My main empirical model is the following OLS regression as in Bates, Kahle, and Stulz (2009): YY iiii = αα + ββ IIIIIIIIIIIIIIIIIIIIIIII tt + γγ XX iiii + ww ss + εε iiii where YY iiii is the ratio of cash, short-term investments, or their sum to assets; IIIIIIIIIIIIIIIIIIIIIIII tt is a 3- month T-Bill rate; ww ss captures industry fixed effects; and XX iiii is a set of control variables including the natural log of total assets in 2009 dollars (a size proxy), cash flow volatility, market to book ratio, cash flow, net working capital, capital expenditures, leverage, R&D expense, dividend dummy, acquisitions, repatriation tax cost, and Initial Public Offering (IPO) dummy. Variable definitions are provided in the Appendix. The sample consists of Compustat/CRSP firms from 1974 to 2014. I exclude utilities (SIC codes 4900-4999) and financial firms (SIC codes 6000-6999), firms not incorporated in the United States, and firms with non-positive values of total assets and sales revenue. Missing values reduce
the sample to 127,927 firm-year observations for 13,403 unique firms. Summary statistics for the main variables are provided in Table 1 and are consistent with prior literature. 3.3 Results Table 2 reports the main results. First, consistent with the transactions model, T-Bill is negatively related to cash (Column 2) but is positively related to short-term investments (Column 3). Second, size is negatively related to cash (Column 2) but is positively related to short-term investments (Column 3), also consistent with the transactions model. To compare these findings with prior literature, I use the sum of cash and short-term investments as the dependent variable in Column 1. In line with Bates, Kahle, and Stulz (2009), I find that the T-Bill rate is not related to the sum of cash and short-term investments. The remaining three columns of Table 2 incorporate industry fixed effects. The estimates are largely unchanged except that the T-Bill rate in Column 4 with the ratio of the sum of cash and short-term investments as the dependent variable becomes significant at the 10% level. The coefficients on the remaining control variables are in accordance with what the existing literature reports. Consistent with the precautionary savings and speculative motives, coefficients on cash flow volatility, market-to-book ratio, and research and development expense are positive and significant. Interestingly, cash flow is positively related to short-term investments but is negatively related to cash. This result is consistent with the transactions model: all else equal, high cash flow firms have larger liquid assets portfolios resulting in the higher opportunity cost of holding idle cash. Having established that cash and short-term investments have common determinants with opposite signs, I examine whether interest rates explain changes in the level and composition of corporate liquid assets. In Table 3, the dependent variable is the natural logarithm of the cash ratio. My estimates indicate that in response to a 14% decrease in nominal T-Bill rate from 1981 to 2014, cash ratio should increase by exp(-0.14*-7.351) 1 = 180%. This explains roughly half (180%/386%) of the actual increase in the cash ratio during that time. Next, I calculate the predicted response of the cash ratio to changes in interest rates, holding everything else constant, using the estimated coefficients from Column 1, Table 3. The predicted
series is normalized so that its average value is equal to the average value for the actual series. Further, since the transactions model implies that an increase in cash is offset with a corresponding decrease in short-term investments, the predicted changes in the short-term investments ratio is simply the opposite of the predicted changes in the cash ratio. The resulting patterns are presented in Panels A and B of Figure 3. The overall trends of the predicted ratios match those of the actual. The effect of the T-Bill rate alone seems to do a good job of capturing the changes in the composition of corporate liquid assets over time although the fit is not perfect. 3.4 Subsample Analysis In this subsection, I test the transactions model in the subsample of NYSE listed firms. There is a widely accepted view that the run-up in corporate liquid assets since the 1980s was primarily driven by the changes in the cross-section of firms rather than the time-series changes in firm-level liquid assets demand (Bates, Kahle, and Stulz, 2009; Begenau and Palazzo, 2017; Graham and Leary, 2016). It is argued that the entrance of firms with high ex-ante demand for liquid assets (e.g., R&D intensive firms) to public markets pushed the average ratio of corporate liquid assets to total assets higher. The exact same phenomenon might have caused the trends observed in Figure 1. To address this possibility, I utilize the findings of Graham and Leary (2016) that the effect of new entrants on the average corporate liquid assets ratio is mitigated among firms listed on the New York Stock Exchange (NYSE) due to the lower exposure of this subsample to R&D intensive firms. The evolution of the composition of corporate liquid assets for NYSE listed firms is illustrated in Figure 4. Consistent with Graham and Leary (2016), I find that total corporate liquid assets do not increase monotonically in this subsample (e.g., total liquid assets ratio was equal to 11.6% in 1983 and 11.9% in 2011). On the other hand, the changes in the composition of corporate liquid assets in the NYSE subsample closely resemble those of the whole sample. The ratio of short-term investments to assets was higher than the ratio of cash to assets until late 1980s. In 1983, the short-term investments ratio reached a peak of 9.0% while the cash ratio was more than three times lower at 2.6%. Compare that with 2014 when the average cash holdings were 83.9% (10.9% divided by 13.0%) of total liquid assets. These findings for the NYSE subsample suggest
that the firm-level demand for the two types of corporate liquid assets (cash and short-term investments) has changed over time. Additionally, I examine the subsample of new entrants in each year in Figure 5. These firms have higher ratios of liquid assets to total assets, also consistent with Graham and Leary (2016). Their entrance to the U.S. public markets is believed to be the driving force behind the increase in average total liquid assets ratio in the whole sample, as mentioned earlier. Notice that the subsample of new entrants also was subject to substantial changes in the composition of liquid assets especially during the 1980s. The regression results for the NYSE subsample are presented in Table 4. Cash to assets ratio is negatively related to T-Bill rate (Column 2) while short-term investments to assets ratio is positively related to T-Bill rate (Column 3). Total liquid assets are unrelated to T-Bill rate (Column 1) suggesting that the two effects offset one another. Next, I break the subsample of NYSE firms further into quintiles based on size (Columns 4 and 5) and cash flow volatility (Columns 6 and 7). Transactions model predicts that large firms have higher sensitivity to interest rates due to the economies of scale while firms with more uncertain liquidity needs (i.e., high cash flow volatility) have higher sensitivity to interest rates due to their larger transactions cash balances. Table 4 shows that firms in the highest quintile of cash flow volatility are more than two times (-0.601/-0.261 = 2.3) more sensitive to interest rates than those in lowest quintile. Highest and lowest size quintiles of firms do not show much difference in their sensitivity to interest rates with coefficients of - 0.411 and -0.425, respectively. Overall, the results in this section are consistent with the transactions demand for cash model: size and interest rates are negatively related to cash but are positively related to short-term investments. Changes in interest rates explain the composition and the level of corporate liquid assets over time indirectly, by explaining the evolution of its two components. In contrast to the increase in total liquid assets, these trends were primarily within-firm rather than cross-sectional as evidenced by the subsample analysis of the NYSE listed firms.
4. Robustness In this section, I report the results of several additional tests that address the choice of proxies, the effect of other macroeconomic variables and the sample selection. I end this section by discussing the remaining limitations of this study. First, throughout this paper I assumed that the yield on cash is zero and the yield on shortterm investments is the three-month T-Bill yield, which follows directly from the classic transactions model of liquidity management. Both Cardella, Fairhurst, and Klasa (2016) and Dunchin et al. (2017) report that firms increasingly hold both safe interest-earning instruments and less liquid, more risky securities such as corporate and municipal bonds in their liquid assets portfolios. These findings call for an alternative proxy to the three-month risk-free T-Bill rate as the yield on short-term investments. Thus, I repeat the main analyses of this paper using the AAArated corporate bond yield as a proxy for interest rates. Cardella, Fairhurst, and Klasa (2016) point out that although the nature of firms short-term investments is riskier in the most recent period, the actual probability of default and interest rate risk are still low because firms prefer only the highest credit quality securities and plan to hold them until maturity. Panel A of Table 5 shows that interest rates are still negatively related to cash and are positively related to short-term investments. Second, interest rates are highly correlated with inflation. I repeat the main analyses using the deflated interest rates defined as the difference between the 3-month T-Bill rate and inflation. 4 The results are not affected (Panel B, Table 5). Next, I add controls for several macroeconomic variables such as the real GDP growth, market volatility, and productivity. The predicted relations between interest rates and cash and short-term investments still hold (Panel C, Table 5). Fourth, the results hold in a different sample period, from 1980 to 2006, that excludes the financial crisis and was used in the seminal Bates, Kahle, and Stulz (2009) study (Panel D, Table 5). The remaining concern is that there is likely switching between interest-bearing assets and actual dollar bills within Cash and Cash Equivalents balance sheet account that I am not capturing. The ideal setting for analyzing the trade-off between cash holdings and interest-bearing assets would be to have detailed data on the composition of Cash and Cash Equivalents. These 4 Curtis, Garin, and Mehkari (2017) study the effect of inflation on corporate cash holdings in detail.
data were not available until the Statement of Financial Accounting Standards (SFAS) No. 157 was implemented in 2009, which requires firms to report the value of all financial assets on their balance. In addition, Azar, Kagy, and Schmalz (2016) discuss how changes in regulation (e.g., the elimination of interest rate ceilings) and developments in technology in the late 1980s contributed to firms holding more interest-bearing assets in their liquid assets portfolios. Money market mutual funds and time and savings deposits offer the benefits of both high liquidity and interest-earning ability. Therefore, to the extent that firms switched towards these financial assets while reporting them under Cash and Cash Equivalents, the documented run-up in cash in Figure 1 could be a direct implication of this fact, having nothing to do with interest rates. Overall, the evidence in this section shows that the main results of this paper are robust to using a different proxy for interest rates, additional controls for macroeconomic factors, and a different sample period. However, the effect of changes in the composition of Cash and Cash Equivalents towards interest-yielding assets and in the composition of Short-Term Investments towards riskier and less liquid securities introduces measurement error that is most pronounced in the second half of my sample period. Nevertheless, my proxies represent an improvement to the traditional definition of cash as the sum of cash and short-term investments and allow for a long horizon study of firm-level liquidity management practices. 5. Conclusion I document the evolution of two components of corporate liquid assets, cash and short-term investments. They follow strikingly different paths as a percent of assets since 1970s. I then revisit and find support for the overlooked predictions of the transactions demand for cash framework that interest rates are positively related to short-term investments but are negatively related to corporate cash. Changes in interest rates explain a large portion of changes in cash and short-term investment ratios but are unrelated to their sum. As the U.S. economy abandons the unprecedented low interest rate environment, firms may start switching the composition of their corporate liquid assets towards interest-yielding financial securities. However, the measurement error discussed in the previous section will likely mute the
shift towards short-term investments. An interesting research question is whether interest rates affect not only the cash versus interest-bearing assets trade-off but also safe versus risky and liquid versus illiquid financial securities trade-off in corporate liquid assets portfolios. In other words, did the low interest rate environment induce firms to invest in riskier and less liquid financial securities as documented in Dunchin et al. (2017)? Further study of factors determining changes in the composition of corporate liquid assets is warranted. Appendix Variable Definitions Compustat data items are in parentheses. Acquisitions: ratio of acquisitions [#129] to total book assets [#6]. Capex/Assets: ratio of capital expenditures [#128] to total book assets [#6]. Cash/assets: ratio of cash [#162] to total book assets [#6]. STI/assets: ratio of short-term investments [#193] to total book assets [#6]. (Cash + STI)/Assets: ratio of cash and short-term investments [#1] to total book assets [#6]. Cash flow/assets: ratio of operating income before depreciation [#13], after interest [#15], dividends [#21], and taxes [#16] to total book assets [#6]. Dividend dummy: indicator variable equal to 1 if a firm paid a common dividend in a given year (i.e., #21 is positive). Cash flow volatility: volatility of cash flow to assets within the two-digit SIC group of a firm. As in Bates, Kahle, and Stulz (2009), for a given year and two-digit SIC group, I calculate the standard deviation of cash flow / assets over the previous 10 years for each firm within that group. A firm must have at least three observed cash flow / assets over the previous 10 years to be counted. Industry sigma for a two-digit SIC group is the average of the standard deviations of cash flow / assets across all firms in the group. Leverage: ratio of the sum of long-term debt [#9] and debt in current liabilities [#34] to total book assets [#6]. Market to book: ratio of the market value of the firm to total book asset value [#6]. Market value is computed as book value of assets [#6] plus market value of equity (equal to the
stock price at fiscal year close [#199] times the number of common shares outstanding [#25]) less book value of common equity [#60]. NWC/assets: ratio of net working capital, net of cash and short-term investments [#179- #1], to total book assets [#6]. R&D/sales: ratio of R&D expenditures [#46] to sales [#12]. When missing from Compustat, R&D is set equal to 0. Real size: ratio of total book assets [#6] to the U.S. GDP deflator in the corresponding year (equal to 100 in 2009) multiplied by 100. The U.S. GDP deflator is obtained from FRED. Assets: book value of total assets [#6]. Repatriation tax cost: the product of a firm s foreign earnings [#273] times the statutory U.S. tax rate of 35% minus the firm s foreign tax credit [#64], scaled by total assets [#6] and multiplied by 100, if this quantity is positive, and zero otherwise. IPO dummy: indicator variable equal to 1 if less than 4 years have passed since the first year for which a stock price [#24] is observed. T-Bill: average secondary market rate on 3-month treasury bills from FRED. Inflation: percent change (December to December) in the CPI for all urban consumers: all items from FRED. Corporate bond rate: average Moody s seasoned AAA bond yield from FRED. Real GDP Growth: annual percent change in real US Gross Domestic Product (chained 2009 $) from FRED. Productivity: Nominal GDP scaled by total nonresidential fixed assets over the prior year from the US Bureau of Economic Analysis. Market Volatility: One-year lag of the standard deviation of the daily value-weighted market returns from CRSP. To limit the effect of outliers, I winsorize the data as follows: leverage is between zero and one; R&D/sales, acquisitions/assets, cash flow volatility, NWC/asset, cash flow/assets, capital expenditures/assets, market-to-book ratio are winsorized at the 1% level.
References Azar, J., Kagy, J. F., and Schmalz, M. C., 2016. Can Changes in the Cost of Carry Explain the Dynamics of Corporate Cash Holdings? Review of Financial Studies 29, 2194-2240. Bates, T. W., Kahle, K. M., and Stulz, R. M., 2009. Why do US firms hold so much more cash than they used to? The Journal of Finance 64, 1985-2021. Baumol, W. J., 1952. The transactions demand for cash: An inventory theoretic approach. The Quarterly Journal of Economics 66, 545-556. Begenau, J. and Palazzo, B., 2017. Firm Selection and Corporate Cash Holdings. Working Paper. Cardella, L., Fairhurst, D. J., and Klasa, S., 2016. What Determines the Composition of a Firm's Total Cash Reserves? Working Paper. Curtis, C.C., Garín, J. and Mehkari, M.S., 2017. Inflation and the evolution of firm-level liquid assets. Journal of Banking & Finance 81, 24-35. Dunchin, R., Gilbert, T., Harford, J. and Hrdlicka, C., 2017. Precautionary savings with risky assets: When cash is not cash. The Journal of Finance 72(2), 793-852. Falato, A., Kadyrzhanova, D., and Sim, J., 2013. Rising intangible capital, shrinking debt capacity, and the US corporate savings glut. Working Paper. Foley, C. F., Hartzell, J. C., Titman, S., and Twite, G., 2007. Why do firms hold so much cash? A tax-based explanation. Journal of Financial Economics 86, 579-607. Graham, J., and Leary, M., 2016. The Evolution of Corporate Cash. Working paper. Gao, X., Whited, T.M. and Zhang, N., 2017. The interest sensitivity of corporate cash. Working paper. Harford, J., Klasa, S., and Maxwell, W. F., 2014. Refinancing risk and cash holdings. The Journal of Finance 69, 975-1012. Jacobs, D. P., 1960. The Marketable Security Portfolios of Non Financial Corporations, Investment Practices and Trends. The Journal of Finance 15, 341-352.
Miller, M.H. and Orr, D., 1966. A Model of the Demand for Money by Firms. The Quarterly Journal of Economics 80, 413-435. Opler, T., Pinkowitz, L., Stulz, R. and Williamson, R., 1999. The determinants and implications of corporate cash holdings. Journal of Financial Economics, 52: 3-46. Tobin, J., 1956. The interest-elasticity of transactions demand for cash. The Review of Economics and Statistics 38, 241-247.
25% 20% 15% 10% 5% 0% (Cash + STI)/Assets Cash/Assets STI/Assets Figure 1 Average liquid asset holdings The sample includes all Compustat firm-year observations from 1974 to 2014 with positive values for book value of assets and sales revenue and nonmissing observations for explanatory variables. Financial firms, utilities, firms not incorporated in the United States are excluded from the sample.
(a) 18% 16% 14% 12% 10% 8% 6% 4% 2% 0% 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014 Cash/Assets STI/Assets T-Bill (b) 20% 15% 10% 5% 0% -5% 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014 Cash/Assets STI/Assets T-bill - Inflation Figure 2 Average liquid asset holdings and interest rates The sample includes all Compustat firm-year observations from 1974 to 2014 with positive values for book value of assets and sales revenue and nonmissing observations for explanatory variables. Financial firms, utilities, firms not incorporated in the United States are excluded from the sample.
(A) Cash/Assets 18% 16% 14% 12% 10% 8% 6% 4% 2% 0% Actual Predicted 14% 12% 10% 8% 6% 4% 2% 0% (B) STI/Assets 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 2011 2013 Actual Predicted Figure 3 Predicted versus actual values of the corporate cash and short-term investments ratios I calculate the predicted response of corporate cash holdings to variation in interest rates using the estimated effect from Specification 1 in Table 3, holding all other factors constant. I normalize the predicted series so that its average is equal to the average value for the actual series.
14% 12% 10% 8% 6% 4% 2% 0% (Cash + STI)/Assets Cash/Assets STI/Assets Figure 4 Average liquid asset holdings for firms listed on the NYSE The sample includes all Compustat firm-year observations from 1974 to 2014 with positive values for book value of assets and sales revenue and nonmissing observations for explanatory variables. Financial firms, utilities, firms not incorporated in the United States and firms not listed on the NYSE are excluded from the sample.
50% 45% 40% 35% 30% 25% 20% 15% 10% 5% 0% (Cash + STI) / Assets Cash / Assets STI / Assets Figure 5 Average liquid asset holdings for new entrants The sample includes all Compustat firm-year observations from 1974 to 2014 with positive values for book value of assets and sales revenue and nonmissing observations for explanatory variables. Financial firms, utilities, and firms not incorporated in the United States are excluded from the sample. New firms are defined as those entering the sample for the first time in each year t.
Table 1 Summary Statistics Variable Mean Std. Min Max N (Cash + STI)/Assets 0.16 0.20 0.00 1.00 127927 Cash/Assets 0.11 0.15 0.00 1.00 127927 STI/Assets 0.06 0.13 0.00 1.00 127927 Real Size 1475 7486 0.00 324392 127927 R&D/Sales 0.17 0.80 0.00 6.63 127927 Market to book 1.97 1.89 0.55 12.83 127927 Acquisitions 0.02 0.05 0.00 0.33 127927 CF volatility 0.14 0.12 0.01 0.56 127927 Capex/Assets 0.07 0.07 0.00 0.40 127927 NWC/Assets 0.10 0.25-1.05 0.58 127927 Cash Flow/Assets 0.00 0.25-1.46 0.27 127927 Dividend dummy 0.34 0.47 0.00 1.00 127927 Leverage 0.25 0.23 0.00 1.00 127927 The sample includes all Compustat firm-year observations from 1974 to 2014 with positive values for book value of assets and sales revenue. Financial firms (SIC codes 6000-6999), utilities (SIC codes 4900-4999) and firms not incorporated in the United States are excluded from the sample. The resulting panel contains 166,531 observations for 14,586 unique firms. Missing explanatory variables reduce the panel to 127,927 observations for 13,403 unique firms for the OLS regressions. Variable definitions are provided in the Appendix.
Table 2 Regressions of firm-level liquid assets demand Variables (1) (2) (3) (4) (5) (6) (Cash + (Cash + STI)/Assets Cash/Assets STI/Assets STI)/Assets Cash/Assets STI/Assets T-Bill -0.015-0.562*** 0.546*** -0.079* -0.574*** 0.495*** (0.053) (0.072) (0.064) (0.047) (0.073) (0.060) CF Volatility 0.210*** 0.151*** 0.059*** 0.112*** 0.129*** -0.017 (0.018) (0.014) (0.013) (0.016) (0.018) (0.012) R&D / Sales 0.067*** 0.028*** 0.039*** 0.064*** 0.028*** 0.036*** (0.002) (0.002) (0.003) (0.002) (0.001) (0.002) Repatriation tax cost 0.000 0.000-0.000* 0.000 0.000-0.000 (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Market to book 0.015*** 0.012*** 0.004*** 0.014*** 0.011*** 0.003*** (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) ln(real Size) -0.002** -0.004*** 0.002*** -0.001** -0.004*** 0.003*** (0.001) (0.001) (0.001) (0.001) (0.000) (0.001) Cash Flow / Assets 0.026*** -0.013** 0.039*** 0.026*** -0.013*** 0.039*** (0.007) (0.006) (0.004) (0.006) (0.005) (0.002) NWC / Assets -0.178*** -0.090*** -0.088*** -0.185*** -0.089*** -0.096*** (0.007) (0.006) (0.006) (0.006) (0.005) (0.005) Capex / Assets -0.388*** -0.206*** -0.182*** -0.350*** -0.186*** -0.164*** (0.021) (0.016) (0.010) (0.016) (0.014) (0.007) Leverage -0.347*** -0.194*** -0.153*** -0.344*** -0.191*** -0.153*** (0.008) (0.009) (0.008) (0.006) (0.008) (0.008) Dividend Dummy -0.033*** -0.023*** -0.010*** -0.031*** -0.021*** -0.010*** (0.004) (0.002) (0.003) (0.003) (0.001) (0.002) Acquisitions -0.270*** -0.142*** -0.128*** -0.268*** -0.143*** -0.125***
(0.017) (0.011) (0.009) (0.016) (0.011) (0.008) IPO dummy 0.041*** 0.021*** 0.020*** 0.041*** 0.020*** 0.021*** (0.004) (0.003) (0.003) (0.004) (0.003) (0.002) Constant 0.237*** 0.182*** 0.055*** 0.250*** 0.184*** 0.067*** (0.008) (0.008) (0.007) (0.005) (0.007) (0.006) Industry FE? No No No Yes Yes Yes Observations 127,927 127,927 127,927 127,927 127,927 127,927 R-squared 0.452 0.342 0.190 0.463 0.348 0.202 The sample includes all Compustat firm-year observations from 1974 to 2014 with positive values for book value of assets and sales revenue. Financial firms (SIC codes 6000-6999), utilities (SIC codes 4900-4999) and firms not incorporated in the United States are excluded from the sample. Missing explanatory variables reduce the panel to 127,927 observations for 13,403 unique firms for the OLS regressions. Standard errors allow for clustering by firm and by year in Columns 1-3 and by year only in Columns 4-6. Robust standard errors are in parentheses. Variable definitions are provided in the Appendix. Note: *** p<0.01, ** p<0.05, * p<0.1.
Table 3 Regressions of firm-level cash demand in logs 1 2 Variables ln(cash/assets) ln(cash/assets) T-Bill -7.351*** -7.483*** (0.899) (0.889) Firm characteristics? Yes Yes Industry FE? No Yes Observations 126,639 126,639 R-squared 0.266 0.281 The sample includes all Compustat firm-year observations from 1974 to 2014 with positive values for book value of assets and sales revenue. Financial firms (SIC codes 6000-6999), utilities (SIC codes 4900-4999) and firms not incorporated in the United States are excluded from the sample. Missing explanatory variables reduce the panel to 127,927 observations for 13,403 unique firms for the OLS regressions. Standard errors allow for clustering by firm and by year in Column 1 and by year only in Column 2. Robust standard errors are in parentheses. Variable definitions are provided in the Appendix. Note: *** p<0.01, ** p<0.05, * p<0.1.