“Do firms park capital? Evidence from the U.S. manufacturing sector”

This study uses the “cost of carry” (CoC) measure to identify the motive for corporate cash holdings. Based on the historical, moving-average holdings of currency and liquid assets, the measure represents the net opportunity cost of corporate demand for money. This study finds that large manufacturing firms in the U.S. park their capital in short-term assets appealing to the agency motive for cash holdings. Because dividend-paying firms can choose to distribute their capital to equity shareholders when their investment opportunities are unfavorable, these firms might show a non-positive association between capital expenditure and the CoC measure, championing the transactions motive. Still, dividend-paying large firms exhibit an overall positive correlation, suggesting that they park their capital on the agency motive. A detailed literature review and discussions are followed.


INTRODUCTION
Agency theory suggests that firms are prone to make inefficient longterm investments. This research uses Azar et al.'s (2016) "cost of carry" (CoC) measure -defined as the spread of the risk-free rate over the return on the corporate investment portfolio of short-term financial instruments -to study the relationship between fixed investments and corporate cash holdings. The CoC measure is, thus, the net cost of holding cash and cash equivalents or just "cash". By documenting the time-series and cross-sectional variations of non-interest-bearing and liquid assets held by U.S. firms, Azar et al. (2016) conclude that once the cost of carry effect is accounted for, the current U.S. cash holdings do not appear abnormal. This study find evidence of capital "parking" by large firms in the U.S. manufacturing sector banking and report their inefficient cash holdings.
Applying the CoC measure to study the relationship between fixed investments and the cash holdings of firms has three noticeable advantages. First, the CoC measure, as a historical proxy for agency costs, can detect either temporary inefficiencies or chronic liabilities. Second, the measure contains information about the two distinct (transactions and agency) motives for cash holdings. Third, the CoC measure specifically incorporates the transactions motive to estimate the opportunity cost of holding cash. Because it is a historical measure, interpreting the CoC measure at the firm level should reveal how the historical proportion of different cash holdings influences present corporate financial policies.
There is an extensive literature analyzing the agency motive and inefficient cash holdings of large firms. Jensen (1986) illustrates how en-trenched managers would rather retain cash than increase payouts to shareholders when the firms have poor investment opportunities. Stulz (1990) argues that firms with large cash holdings may invest more than they should. In line with Stulz (1990), Shin and Kim (2002) report that large firms make inefficient corporate investments, in terms of Tobin's Q (1980), in the fourth quarters, controlling for changes in cash holdings. Dittmar et al. (2003) and Pinkowitz et al. (2006) also provide evidence of inefficiency caused by the agency motive for cash holdings.
At the firm level, the CoC measure is technically derived as the share of currency out of the total invested value in liquid assets, scaled by the (risk-free) 3-month Treasury Bill rate. For those firms with a high accumulation of liquid assets in the past (low in terms of the CoC measure), say, for a decade, fixed investments might either increase by "cashing out" the previously held short-term financial instruments or remain stagnant because of uncertain project prospects -leading to an overall non-positive correlation between capital expenditures (CAPEX) and the CoC measure. Corporate demand for money in this case is, thus, based on the transactions motive. However, for the companies that wish to "park" their capital in liquid, short-term assets, even with their large holdings of cash (again low in terms of the CoC measure), concurrent fixed asset investments will decrease -showing an overall positive association with the CoC measure. This scenario is backed by the agency motive for corporate demand for cash.
This study focuses on the large U.S. firms in the manufacturing sector for their sizable fixed asset investment and this feature would render the implication of this research that relates fixed investments to the cost of corporate cash holding more reliable. This study finds that large manufacturing firms in the U.S. park their capital in short-term assets appealing to the agency motive for cash holdings. Because dividend-paying firms can choose to distribute their capital to equity shareholders when their investment opportunities are unfavorable, these firms might show a non-positive association between CAPEX and the CoC measure, championing the transactions motive. Still, dividend-paying large firms exhibit an overall positive correlation, suggesting that they park their capital on the agency motive.
The remainder of this study is organized as follows: section 1 surveys the literature on the corporate motives for cash holdings. The data used and methodologies employed herein are discussed in section 2. Section 3 provides the main results and, lastly, final section concludes this research.

Inefficient cash holding and fixed investments
The agency cost of free cash flow was introduced by Jensen (1986), who illustrates how managers may be motivated to force a firm to grow to more than its optimal level. Such inefficiency arises from managers' being aware of the firm's capital in their control.  Stulz (1990) shows that management tends to over-and underinvest at times of low and high cash flow, respectively, resulting in a positive relationship between investment and cash flow. For quarterly investment analysis, Shin and Kim (2002) report an overinvestment pattern of large firms in the last-quarter fixed investments.
However, firms' historical pattern of agency costs has not been examined. It is important for whether agency cost is a temporary inefficiency that can be improved or a chronic complication that cannot be adjusted. Furthermore, the behavior pattern of large firms parking capital in short-term investment or in cash holdings has not yet been reported.

Descriptive statistics
In mean and median tests of cash proportion confirm the difference in transaction motives between large and small firms. Large firms in general have lower CoC compared to the full sample. As shown in Table 3, although the mean is not significant (0.0243 versus 0.0283), the median is significantly different (0.0187 versus 0.0235). The skewness measure shows that the distribution of CoC of large firms is more positively skewed, implying that most large firms have lower CoC than other firms. The median of short-term investment yields near zero value suggests that most firms in the sample do not hold or report short-term investments. It is possible that missing short-term investments caused small firms to have higher CoC close to the interest rate. This could be problematic if the full sample is used or size terciles are compared, but the focus of this research is on the agency motives for cash holding by large firms. Although the median for the large firms is also small (0.0012), the non-zero value indicates that large firms did report their short-term investment better than other firms did.  The statistics on the dependent variable and the control variables are generally consistent with prior studies. Investment is higher for large firms than for other tercile firms. This can be evident, since the cash flow proxy (CF) is higher and change in cash holding (∆Cash) is lower for large firms. The mean of Tobin's Q of large firms is lower than for the full sample, but the median is similar.
In Table 4, Pearson correlation estimates show a high correlation between the carry measure and the interest rate proxy (TB3MS). The correlation reduces for the large firms subsample, but it is still large and statistically significant enough to be used in the same regression model. Therefore, to investigate the difference between the carry measure and the interest rate, separate regression models are used for comparison of their results. Change in cash holding accounts for the firms' investment alternatives. If firms use cash holding as an alternative to capital expenditure, then cash holding will be higher when capital expenditure is lower and vice versa, yielding negative coefficient estimates. The change measure is different from the CoC measure in the sense that it is a proxy for the past one-year change in cash holding, whereas CoC is a historical measure.

Methodology
Firm and year fixed effects are included, considering the fact that the CoC measure is applied to capture the historical tendencies of individual firms. If only the cross-sectional variation is observed, then the fixed effects may not be necessary. Current focus, however, is to capture the time-series variation within an individual firm, as well as the cross-sectional variation. In addition, capital expenditure is currently calculated as a ratio by normalizing it with the book value of assets.  Note: Table 4 shows the Pearson correlation coefficients between the independent and dependent variables. The items are the estimates, p-values and observation numbers, respectively. ence fixed investments, because it can either provide firms with alternative investments or raise the cost of capital. Interest rate is positively correlated with firms' fixed investments in the descriptive statistics, but interest rates incorporate the aggregate economy's growth options. Controlling for the firm's specific growth opportunities, the relationship should be negative. If the effect of the carry measure on investment is dominated by the effect from the interest rate, then the signs of the coefficients of the carry measure and the interest rate will be identical.

MAIN RESULTS
In Table 6, the dividend dummy (equal to 1 if the firm paid a cash dividend or had a stock repurchase program in the observed year) is added and an interaction term with the CoC and the dividend dummy. The dummy and the interaction term are included to observe whether the historical cash-holding level affected fixed investments differently for firms that had a dividend payment schedule in the same year. Dividend-paying firms have the choice to pay dividends instead of parking capital. If such firms faced historically low investment opportunities, the firms can decide to pay dividends and reduce their cash holding. As a result, the tendency to park capital should decrease, leading to a negative coefficient for the interaction term.
Results show that large firms, whether they have a dividend payment schedule or not, do not decide to maintain optimal cash holding. Although the regression yields a negative coefficient for the interaction term, it is statistically insignificant. In contrast, the statistical significance of the CoC measure increases, implying that the historical level of cash holding negatively affects fixed investment. The result suggests that the agency motives for cash holding do not disappear after considering for possible different behaviors of dividend-paying firms. Including the dividend dummy has a minimal effect on the cash-flow proxy, implying that agency cost does not diminish for dividend-paying firms. The coefficient estimate for the dividend dummy yielded a positive value, indicating that firms with dividend policies make more fixed investments. The result is consistent with the agency theory literature, since managers have no incentive to temporarily increase divi- dends and instead invest more. As a whole, the result can be interpreted as meaning that the agency cost of large firms does not disappear, although much of the literature on finance identifies it as an inefficiency.
Strangely, for the full sample, the firms did lower their cash balance if they were not dividend pay-ing, but considering that the full sample does not control for the transaction motive for cash holding, there is a chance that the transaction motive influenced the result. Middle-size firms were not reported, because they showed statistical inconsistency when the dividend dummy was included. Without the dividend dummy, they appeared to behave like the big firms.

CONCLUSION
This study finds empirical evidence of inefficient behavior of capital parking in the short-term investments of large U.S.-based manufacturing firms, and the effect persists after considering the dividend-payment policies of firms. To identify the inefficiency, the CoC measure is augmented to Tobin's Q investment model. The carry measure has the merit of being a historical average, containing information on the ratio between the motives for cash holding, and calculating the exact opportunity cost of carrying cash as currency. By controlling for size, the large firm subsample is expected to be controlled for the transaction motives for cash, leaving the intermediate liquidity needs of cash holding, as well as the explanatory power of the carry measure on overall cash holding, represented by the cash-to-asset ratio, as proposed by Azar et al. (2016).
The historical examination of the agency cost of firms shows that firms facing such costs may not be able to overcome such problems. If agency costs were temporary irregularities that firms can solve when they are identified, the results would not have been observed. The positive high statistical significance of the CoC measure on firms' fixed investments shows that agency cost is still prevalent in firms and may need further investigation of the firm's fixed investments, the distinct cash holding motives for manager-driven firms, or suboptimal solutions for firms that cannot overcome the inefficiencies.
There are limitations to this study. First, although it empirically identifies inefficient cash holdings, it lacks suggestions on how the firms should behave in order to become efficient. There may be a suboptimal fixed investment and cash-holding level that firms with managerial costs cannot avoid. Second, the suggested model is exposed to misspecification errors that could lead to biased estimators caused by omitted variables or wrong error structures. Last, no robustness tests are provided. Although the first and the second problem are yet to be solved, simple OLS regression and Fama-Macbeth regression with a Newey-West standard-error-adjusted model lagged by 2 or 3 show consistent results for the large-firm subsamples.
This study contributes to the agency cost of cash holding and investment by investigating the relationship between the historical measure of cash holding and investment to identify a persistent anomaly. Dividend and repurchase programs do not mitigate the effect, further supporting the fact that agency cost is not dissipated by dividends and dividend-related payout policies. The result is independent from the effects of interest rates proxied by three-month T-bill rates.