Master Thesis
The Long-Run Impact of a Financial Crisis on Cash Holdings: A Study
Based on US Firms
Author: Huiwen Liu Supervisor: Dr. Vladimir Vladimirov MSc Finance – Quantitative Finance University of Amsterdam – Amsterdam Business School June, 2018
Statement of Originality
This document is written by Huiwen Liu who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Abstract
I investigate the long-run impact of the 2007–2008 financial crisis on cash hoardings using a sample of US firms. The evidence from my research indicates that the changes in companies’ characteristics are not the explanation for the growth in cash reserves after the financial crisis. The growth in cash hoardings is more likely caused by the changes in the function for cash hoardings. In other words, companies prefer to hoard excess cash out of their cash flows to secure liquidity buffer after the financial crisis.
Table of Contents
1. Introduction... 5
2. Literature review ... 7
2.1 Why does cash holding matter for firms?... 7
2.2 Motives for cash holding ... 9
2.3 Financial crisis and cash holdings ... 12
2.4 Determinants of cash holdings ... 13
2.5 Hypotheses development ... 17 3. Data ... 18 3.1 Data source ... 18 3.2 Cash ratio ... 19 3.3 Explanatory variables ... 19 4. Methodology ... 21
4.1 Index for identifying the financial constraints ... 21
4.2 Regression for cash holdings ... 23
5. Empirical Results ... 24
5.1 Descriptive statistics ... 24
5.2 Cash holdings, and financial ratios ... 26
5.3 Cash holdings and firm characteristics ... 28
5.4 Has the function of cash holdings changed after the 2007–2008 financial crisis? . 31 5.5 The increase in the cash hoardings and financial constraints ... 38
6. Robustness Checks ... 44
7. Conclusion ... 45
Appendix ... 47
The Long-Run Impact of a Financial Crisis on Cash Holdings: A Study Based on US Firms
1. Introduction
From the perspective of conventional understanding of corporate finance, hoarding cash can lead to agency problems. The management might spend funds on behalf of their interests, instead of maximizing shareholder wealth. Besides, a significant amount of cash reserves might send signals to outside investors that there is a decline in corporate growth opportunities. It seems that it is not ideal for companies to hoard excess cash. Even though there are costs in remaining cash, it does not slacken off the pace at which companies
accumulate cash piles. Between 2000 and 2016, the average cash ratio of US firms increased from 14.77% to 21.43%. After the 2007–2008 financial crisis, firms still maintain a high level of cash reserves. An extreme example is the cash balance of Apple Inc., over 200 billion dollars in 2017, which is more than the annual gross domestic product of Vietnam, according to data retrieved from the World Bank. Based on the results of the Liquidity Management Poll held in 2009, many financial executives claimed that they would alter their cash buffer to a higher level, and secure companies’ liquidity would be their top task. Meanwhile, the high cash balances of US firms also draw the attention of scholars. It seems that financial crisis has a short-run impact on corporate cash management. Empirical research shows that firms have to make a trade-off between cash reserves and investment opportunities in the short run; however, there have been a limited number of research of the long-run effect of financial crisis on cash hoardings. In this paper, I will assess the long-run influence of financial crisis on cash hoardings. In particular, after financial crisis, is there a permanent shift of demand for cash reserves, and how does financial crisis affect the determinants of cash holdings in the long run. I will try to answer these questions in this paper.
Once financial crisis happens, there is a contraction of capital supplies, and firms have to pay high prices to raise external funds (Campello, et al., 2018). Duchin, Ozbas and Sensoy (2010) indicate that business organizations choose to reduce investment activities in reaction to financial crisis over a short-run period. Campello, Graham and Harvey (2010) indicate that companies also spend their retained cash as a strategy to withstand financial crisis. For the long-run impact of financial crisis, Song and Lee (2012) examine the Asian financial crisis and indicate that Asian companies’ sensitivity to cash-flow uncertainty is affected by Asian financial crisis over a long-term period. However, there is no related research of how the US firms’ cash management reacts to a financial crisis. The 2007–2008 financial crisis provides an opportunity to examine the long-run influence of financial crisis on cash hoardings.
This paper extends the literature about the long-run effect of a financial crisis on firms’ cash management. My study shows that the US firms’ decisions on cash holding are affected differently. Unlike Asian firms, which are more sensitive to cash-flow uncertainty, US firms tended to pay attention to their cash flows after the 2007–2008 financial crisis. As a precautionary motive of cash hoardings, financially constrained companies would hoard more cash to hedge financial uncertainty (Han and Qiu, 2007); however, my research also shows evidence against this argument. Meanwhile, my study supports the research of Ang and Smedema (2011), which suggests that cash-rich companies highly intend to accumulate excess cash as a liquidity buffer to prepare for the next crisis.
To conduct this research, I collected a sample of 29,690 firm-year observations representing 5,685 US companies over the period from 2000 to 2016. First, I analyze the trend of cash holding and some financial ratios of firms by year. Second, I examine the hypothesis that changes in firms’ characteristics give rise to increased cash holdings. Third, I investigate the long-run relationship between the 2007–2008 financial crisis and cash
The rest of this paper proceeds as follows: Section 2 has five subsections, containing related literature and hypothesis development. Section 3 discusses the source of data and a gives a full description of relevant variables. Section 4 describes the methodology used in this paper. Section 5 reports empirical results. Section 6 conducts several regressions for
robustness checking, and Section 7 concludes.
2. Literature review
2.1 Why does cash holding matter for firms?
In standard valuation models, academics and firms regard cash as negative debt, because firms treat cash balance as a source from which they can pay back debt promptly. Firms subtract cash balance from the debit account to calculate net leverage. They believe that net leverage is the essence of measuring the value of shareholders’ wealth (Acharya, Almeida and Campello, 2007).
In contrast, recent studies on cash holding claim that the meaning of cash may be more than that. Faulkender and Wang (2018) do empirical research on how the external capital market evaluates cash balances that firms hold and how the value of cash balances differs across different dimensions. In general, the marginal effect of cash on shareholders is $0.94 across all firms. This means that the extra dollar added to cash balances does not incorporate the entire value of the extra cash into companies’ share prices; however, the marginal effect varies for cash-poor and cash-rich companies. The marginal effect of cash is statistically larger (more than $1) for constrained firms than for unconstrained firms. This research indicates that the external market recognizes the value of companies’ liquidity management, which reflects in stock prices. Pinkowitz and Williamson (2004) apply similar research, but they pay more attention to investment opportunity. Their results show that
investment opportunity is one of the critical elements to increase the marginal effect of cash holding. In general, their results show that increasing cash balance could add value to firms. Fresard (2018) shows that the cash reserve is not merely negative debt, by examining the relationship between cash holding and the competitiveness of firms. He uses an
instrumental variable approach to capture the general impact of cash hoardings on market-share growth. After that, he investigates how discrepancies in ex-ante cash balances bring about changes in market share after increased competition due to import tariff cuts, by applying a difference-in-difference methodology. His research shows that firms can obtain growth in future market share by hoarding a significant amount of cash. If there is an exogenic shock, the influence of cash holding on market share would be more considerable. He concludes that cash reserves improve the growth of the market. However, Morellec, Nikolov and Zucchi (2014) point out that there is a causal effect of market competition on cash management. Instead of focusing on exogenic effect, their research concentrates on firm characteristics. Morellec, Nikolov and Zucchi (2014) find that it is crucial for firms to hoard cash in a competitive product market, but this finding does not apply to firms that have no financing constraints. To sum, their results suggest that cash reserves play an essential strategic role.
Acharya, Davydenko and Strebulaev (2012) indicate that cash holding can be an indicator of credit risk. They show that there is a positive relationship between the amount of cash reserves and the credit spread. Moreover, the long-term default risk climbs with
increasing cash balances. They explain that firms choose to hoard cash to avoid default risks, as distinct from pursuing another option of investing more to obtain more future cash flows. The limitation of their model is that it is applicable when firms face high cash-flow
volatilities and expensive default costs. Based on the previous literature, we can conclude that cash reserves should not be regarded merely as negative debt, and cash holdings have
significant effects on increasing corporate value, enhancing the competitiveness of firms, and recognizing the signal of credit risks.
2.2 Motives for cash holding
Even though academics do not give an answer as to how much cash firms should have, it is crucial for firms to choose the level of cash balances they want, since cash is the basic element for corporate daily operations and liquidities. From existing literature, academics have determined four generally accepted motives for accumulating excess cash.
The transaction motive, primarily raised by Baumol (1952) and extended by Miller and Orr (1966), is that companies need liquid assets to pay transaction costs when they conduct their daily business. Transaction costs could arise when the firms raise external financing or asset sales. Moreover, in the transaction-demand model, Miller and Orr (1966) posit that cash holdings are applicable to an economy-of-scale hypothesis. Subsequently, economists have examined the transaction motive empirically. Using data from 1961 to 1992 collected from Compustat, Mulligan (1997) studies the observations of 12,000 US firms. By adopting a methodology of cross-sectional studies and ordinary least squares (OLS)
regression, he proves that cash holdings provide scale economies at both intra-industry level and cross-industry level.
Firms have a precautionary motive of accumulating cash in the event of unexpected shock, when they may face expensive costs in raising external funds. Opler, et al. (1999) show that corporates that have limited ability to get access of the external capital market prefer to hoard extra cash, and this is in agreement with the precautionary motive. Moreover, they find that cash holdings have significant relations with cash flows and the uncertainty of these cash flows. The precautionary motive also drives firms to hoard more cash if they encounter with profitable investment opportunities, and Opler, et al. (1999) prove this by adopting research and development (R&D) ratio as the proxy. Almeida, Campello and
Weisbach (2004) posit that financial constraints can affect the amount of cash holdings. They test their theory by using the precautionary-demand function of cash holdings and conclude that financially constrained companies reserve more cash from their cash flows compared with the cash reserved by financially unconstrained companies. Han and Qiu (2007) continue the research of Almeida, Campello and Weisbach (2004) to further investigate cash flows. They find that there is a statistically significant relationship between cash-flow volatility and the cash reserves for financially constrained firms; however, this relationship is not
significant for financially unconstrained companies. In their paper, they also indicate that changes in firm's characteristics can explain the increase in cash reserves without changing the demand function for cash holding. García-Teruel and Martínez-Solano (2008) study the determinants of cash holding by analyzing the data of small and medium firms in Spain. Their research indicates that the cash ratio is statistically correlated with the leverage ratio, cash-flow ratio, and short-run debt ratio in a positive way; meanwhile, there are negative relations between long-run debt ratio and cash reserves. From the previous literature review driven by the precautionary motive, we can conclude that academics keep adding new determinants to the demand function for cash holding. Moreover, the research subject has expanded from listed firms to small firms.
Some academics indicate that US multinationals tend to hoard more cash due to the tax motive (Foley, et al., 2007). They explain that multinationals may face costly tax payments if they repatriate cash reserves from their overseas subsidiaries. Instead of repatriating cash, the multinationals use their subsidiaries as a tax shelter, which
consequently results in a high level of cash holding. Gu (2017) examines the tax motive by applying a dynamic model with a sample of the US multinational and domestic firms. He shows that the cash differential between the multinational and domestic firms may reduce by 42% if there is no repatriation cost. Xing (2018) supports this by studying the Japanese
tax-system reform executed in 2009. After the reform, the Japanese tax tax-system switched from a global tax system to a territory-based system. Since the reform significantly decreases repatriation costs, Japanese multinationals reduce the level of cash holding relatively. In general, research shows that multinationals reserve relatively more cash than domestic firms, and this phenomenon may diminish if the tax costs of repatriation are reduced or waived.
Applying agency theory, Jensen (1986) argues that the interest conflicts between shareholders and management teams are irreconcilable, and the managers might withhold cash on behalf of their objectives, which recognizes the agency motive. He points out that the managers would increase the cash holding to invest low-profit projects later, instead of distributing the cash reserves to the shareholders. Opler, et al. (1999) also suggest that risk aversion could be another explanation for management holding a high level of cash reserves. It is noticeable that they find no evidence to support the agency motive. However, Harford (1999) shows opposite results, that companies with excess cash reserves prefer to involve in acquisition activities actively, and the value of these firms decreases significantly from the evidence of stock returns. Besides, inadequate protection for shareholder rights can
exacerbate the agency problem. In some countries, the legal protection of shareholder rights is not well-established. Dittmar, Mahrt-Smith and Servaes (2003) find that companies in these countries tend to retain excess cash holdings. In fact, these companies hold two times more cash reserves compared with companies in countries with well-established protection for shareholders. In this scenario, the explanatory power of the precautionary motive is less significant. Dittmar and Mahrt-Smith (2007) find supportive evidence for the agency motive by evaluating companies’ governance performance. The managers in poorly governed firms prefer to retain more cash reserves than the managers in well-governed firms, and it seems that the management in poorly governed firms could burn cash reserves speedily. The result
is that there could be a deterioration in the value of poorly governed companies due to poor cash management.
2.3 Financial crisis and cash holdings
In addition to the above literature, some scholars find that macroeconomic factors could affect the behavior of cash hoardings, such as financial crisis. Financial crisis can represent a systemic negative shock to the external capital supply for non-financial
companies. Furthermore, the firms have to handle more expensive costs for raising external finance due to the contraction of the capital supply (Campello, et al., 2018). Duchin, Ozbas, and Sensoy (2010) examine the short-term impacts of the 2007–2008 financial crisis on firms’ investment activities and cash-management policies using a difference-in-difference approach. Their investment activities decrease dramatically from the beginning of the financial crisis. The investment decline reaches its highest level for cash-poor companies. Campello, Graham and Harvey (2010) study this subject from a different angle, by directly surveying over 1,000 Chief Financial Officers in different continents. There are significant declines in investment activities systemically. For financially constrained firms, the situation is even worse; they have to cut more budget for technology and capital expenses. Meanwhile, financially constrained companies choose to burn through their cash reserves to support their operational activities. For firms with no access to the external market, they have to delay or cancel their investment plans. The drawback of this paper is that the research is survey-based, and the measurements in their research are not quantitative. In general, the research shows that the excess cash hoardings from precautionary motive are valuable when there is financial crisis.
The determinants of the cash holdings are also affected by the business cycle. Almeida, Campello and Weisbach (2004) show that there is an increase in cash-flow sensitivity during recessions. Some scholars follow their research to further examine the
effect of macroeconomic volatility on cash management. It seems that financial crisis causes the macroeconomic environment to become volatile, which further leads to a situation wherein managers implement inaccurate cash-management policies. Baum, et al. (2006) indicate that macroeconomic volatility hampers the ability of the managers to make good use of firm-specific information, and they would conduct similar policies to cash management.
Moreover, there is evidence showing that financial crisis has a long-run impact on companies’ cash reserves, because companies hoard excess cash in preparation for recessions that could happen in the future. Song and Lee (2012) study the long-run impact of the Asian financial crisis on cash reserves. They conclude that cash management underwent a structural change after the Asian financial crisis; the amount of cash reserves increased in the post-crisis period. This argument is in line with managers valuing the precautionary motive of cash management more after a negative shock. Ang and Smedema (2011) investigate how future crisis affects the management's cash-management choices. They show that cash-rich companies tend to hoard excess cash to secure their liquidity buffer in case that next crisis happens. They explain that financially unconstrained firms are cash-poor firms, and they are not able to prepare cash buffer. Generally, the results prove that the effects of the financial crisis are not just short term.
2.4 Determinants of cash holdings
The determinants of cash holding, enlightened by the research of Opler, et al. (1999) adopted for this research are as follows:
Investment opportunities. For companies with more investment opportunities, if they
are in the situation of having a cash shortage and cannot access external funds on time, they have to bypass valuable investment projects. Meanwhile, it is costly for these companies to face financial distress. Once they are in bankruptcy, the value of their investment projects will no longer exist. Since these firms have to suffer the loss of bypassing good investment
opportunities and costly financial distress, they prioritize the role of cash retention. Empirical research also verifies that investment opportunity is one of the determinants for cash hoarding by precautionary motive (Opler et al., 1999).
Firm size. Empirical results suggest that corporate size is negatively related to cash
hoarding. Since large firms may get more public attention, they are more transparent and have less asymmetric information between the management and shareholder compared with small firms. Thus, large companies can easily access the external capital market, so they may hold less cash. On the other hand, small firms have higher costs for raising external funds and a higher probability of default. Therefore, small firms prefer to hoard more cash to secure their operations. So, firm size should adversely affect cash holding. Meanwhile, some academics believe that the relation between cash hoardings and firm size is positive, in accordance with the pecking-order theory. Typically, large firms may have successful operations in their business, and they may have more retained earnings compared to small firms, so holding the level of investment is consistent. This means that large firms may have more cash holdings on hand.
Cash flow. Based on the pecking-order theory, established by Myers and Majluf
(1984), companies apply a preference ranking for different financing sources. The implication of the theory is that internal funds are their first option when the firms are involved in investment activities. Hence, when the companies have a high volume of cash flow, they pay back their liabilities and accumulate retained earnings for future activities. It suggests that there is a positive relation between cash flow and cash hoardings. However, some scholars claim that corporates with high-volume cash flow have a low probability of being in financial distress or surrendering profitable investment opportunities, which leads to these firms lowering their cash reserves they hold. Hence, cash flows could have a negative influence on cash holdings.
Industrial cash-flow uncertainty. Idiosyncrasy risk can affect all corporates in specific
industries. The cash flow of the affected firms becomes volatile, and this increases the difficulties for the management to evaluate future cash flows accurately. The precautionary motive suggests that companies in industries with highly volatile cash flow accumulate more cash compared with the others (Bates, Kahle and Stulz, 2009). In general, there is a positive relationship between industrial cash-flow uncertainty and cash hoardings.
Intangible investment. Opler, et al. (1999) show evidence that R&D expense is a valid
proxy of growth opportunities and financial distress. However, this proxy can be improved and modified. Nowadays, the structure of corporate investment has changed. The amount of intangible investment has increased dramatically (Corrado and Hulten, 2010), and R&D is only a part of the intangible investment. With the increasing amount of intangible investment, it could exacerbate the information-asymmetry issue for highly intangible firms, because outside investors cannot appropriately determine the value created by the intangible investments. In other words, these companies are more likely to encounter higher financial distress costs if their cash balance dries out. Thus, it may be better to use
intangible-investment ratio as the proxy for financial distress and growth opportunities, and we expect that the relation between intangible investment ratio and cash holding is statistically positive.
Net working capital. Compared with other assets, it is cheaper to convert cash
substitutes into cash. When firms have a cash shortage, they can choose to sell cash
substitutes to avoid costly external financing. Thus, we assume that companies with a high level of cash substitutes might hold less cash.
Capital expenditure. Applying the pecking-order theory (Myers and Majluf, 1984),
we can conjecture that there is a negative relation between cash holding and capital
expenditure. If retained cash cannot fulfill the need for investment activities, companies have to maintain their investment activities by raising external debt or using cash reserves, which
leads to a decline in cash holdings. Besides, Riddick and Whited (2009) indicate that a productivity shock can boost investment activities. This means that companies would save less cash to implement more investment activities in short run. Meanwhile, Bates, et al. (2009) evaluate the cost of financial distress and business opportunities using capital
expenditure. In this case, capital expenditure could be positively correlated to cash holdings.
Financial leverage. Leverage is an indicator showing the ability to issue debt for
companies. A high level of leverage means that the firms can easily raise external funds when they have a cash shortage. Moreover, the pecking-order theory (Myers and Majluf, 1984) implies that high-leverage companies have lower cash reserves compared with low-leverage companies. This means that leverage is negatively correlated with cash holdings. However, high-leverage firms have more chances to encounter financial distress. In other words, the probability of facing financial distress increases with the level of leverage. To avoid this situation, high-leverage firms prefer to hold excess cash. Thus, the relationship between leverage and cash holdings could be either positive or negative.
Dividend payments. Regarding cash shortage, firms with dividend payments can
reduce or cut dividends to secure the cash supply. Thus, the firms paying dividends tend to hold lower cash reserves, and the evidence from Opler, et al. (1999) verifies this conjecture. However, Ozkan and Ozkan (2004) believe that the firms paying dividends prefer to maintain the consistency of their dividend policies. In other words, these companies would tend to hold excess cash reserves to support the consistency of dividend policies, which indicates that there is a positive relation between dividend payment and cash reserves.
Financial crisis. As discussed in the previous literature review, financial crisis has a
positive influence on cash reserves. Managers recognize that the value of cash hoardings is increased after the financial crisis. Song and Lee (2012) show evidence that this impact exist
over the long term. This suggests that financial crisis is positively correlated with cash reserves.
2.5 Hypotheses development
As discussed in the previous literature review, Han and Qiu (2007) suggest that the changes of companies’ characteristics can explain an increase in cash hoardings. Before we examine the research question, we need to exclude the conjecture that the changes of firm's characteristics explain an increase in cash holdings.
Hypothesis 1: The changes of firms’ characteristics cannot explain the growth in cash holdings.
If the capital market were perfect, there would be no presence of information asymmetry and the agency problem. The market information would be accessible for all participants, and no agency cost would exist. However, the real capital market is not perfect, and market participants might pay a high price for information asymmetry and the agency issue. Until now, the 2008 financial crisis has been regarded as an unpredicted event that still influences countries, companies, and households. Information asymmetry plays a vital role, and this proves itself in financial crisis. Before Lehman Brothers Holdings announced the news of seeking bankruptcy protection, which signaled the financial crisis, nobody realized the future shock they were about to experience, and no one prepared for it. Because of the uncertainty, information asymmetry now exists in the value of liquid assets and current valuation (Murphy, 2008). The management had arbitrary power in representing the agency problem. These two issues are responsible for the 2007–2008 financial crisis. The imperfect market and the financial crisis imply that companies modified their cash balances
accordingly. Moreover, empirical research from Song and Lee (2012) suggests that financial crisis has a long-run influence on Asian corporates’ cash hoardings.
Hypothesis 2: Financial crisis is positively correlated to cash holdings in the long term.
As discussed in the previous literature, there is evidence that financial constraint influences the amount of cash hoardings. It is better to identify the category of different firms by the index of financial constraint. Han and Qiu (2007) indicate that financially constrained firms are more likely to hoard excess cash. However, Ang and Smedema (2011) show that unconstrained firms prefer to prepare for a future financial crisis. For unconstrained firms, they are not able to do that. Thus, it is needed to examine the impact of the financial crisis on cash management for subgroups.
Hypothesis 3: Compared with constrained firms, financially unconstrained firms hold more cash reserves because of a financial crisis
3. Data
In this section, there are three parts. First, I will explain some basics of my sample. The second is about how to measure cash holding in different ratios. After that, there is a discussion about the explanatory variables.
3.1 Data source
To answer the research question, I will study the long-run effect of the 2008 financial crisis on the US firms’ cash reserves in the period from 2000 to 2016. The period from 2000 to 2006 is identified as the pre-crisis period; the period from 2009 to 2016 is the post-crisis period. The annual financial data of the US firms used in this paper are collected from Compustat. The sample set excludes utility firms with Standard Industrial Classification (SIC) codes from 4900 to 4999 and financial firms with SIC codes from 6900 to 6999, because these industries are under strict government regulation or have a minimal capital requirement. I exclude companies with incomplete variable data in the regression models and
restrict the sample to firms incorporated the US. Besides, to eliminate the effect of outliers, I winsorize all ratios at one percentile tails. Following the above process, there are 43,494 firm-year observations remaining.
3.2 Cash ratio
In previous studies, academics have identified different cash ratios to measure cash holdings. The first ratio is cash to book total assets, where cash is equal to cash and short-term investments (Bates, Kahle and Stulz, 2009). This is the most common ratio evaluate cash holdings. The second is the ratio of cash to total sales (Bates, Kahle and Stulz, 2009). Third, Opler, et al. (1999) use the ratio of cash to net assets. The net assets are the total book assets after subtracting cash and the equivalents. The drawback of this ratio is that it might generate extreme values if most of a firm's assets are in the form of cash. The fourth ratio is cash to net assets in the form of the natural logarithm, introduced by Foley, et al. (2007); this method could reduce the issue of extreme value (Bates, Kahle and Stulz, 2009).
3.3 Explanatory variables
Investment opportunities. Smith and Watts (1992) state that a valid proxy of the
investment opportunities is the market-to-book ratio (MTBR). Firms with either missing or negative book value of total assets are dropped from the sample set. The calculation of MTBR is:
"#$% = #()*+ *,,-), − (0(11(2 -345)6 + 895:- ∗ <ℎ*9-, (4),)*2>52?)
#()*+ *,,-)
Firm size (FZ). To measure firm size, I use firms' book value of total assets in the
form of the natural logarithm.
A591 ,5B- = ln ()()*+ *,,-),)
Cash flow. In this research, the proxy of cash flows is cash flow to assets ratio (CFR).
The process of calculating cash-flow ratio is:
0A% = EF-9*)52? 52:(1- G-H(9- >-F9-:5*)5(2 − I2)-9-,) − I2:(1- )*J − K5L5>-2>
Industrial cash-flow uncertainty. I use industry sigma (IS) to evaluate the risk of
industry cash flow. To measure IS, I first calculate firm-specific standard deviation (stdev) of cash flow to total assets ratio by using a rolling period of ten years. After that, I generate a two-digit SIC to classify firms into different industries. Then, I use the mean function to obtain average industry-specific standard deviation by the two-digit SIC code, which is the IS.
I< = *L-9*?- ,)>-L (H -*:ℎ 52>4,)96
Intangible investment. We calculate intangible investment as the sum of R&D
expense and 30% of selling, general, and administrative expenses (SG&A) (Peters and Taylor, 2017). Intangible-investment ratio (IIVR) equals that intangible investment divided by sales. The value of R&D and SG&A is zero if firms do not provide the information about actual R&D and SG&A. For missing or negative value of sales, I will drop it from the sample set. To compute IIVR, I use the following formula:
IIM% = %&K + 0.3 ∗ <R&S <*+-, (T-))
Net working capital. We can use net working capital ratio (NWCR), which is the ratio
of net working capital to sales, as the proxy and I calculate net working capital as working capital minus cash. The explicit formula of NWCR is:
TU0% =U(9V52? :*F5)*+ − :*,ℎ *2> ,ℎ(9)W)-91 52L-,)1-2)
#()*+ *,,-)
Capital expenditure. I use capital expenditure ratio (CapexR) as the proxy for capital
expenditure, which is calculated as that capital expenditure divided by the book value of assets.
0*F-J%*)5(= 0*F5)*+ -JF-2>5)49-#()*+ *,,-),
Financial leverage. Leverage ratio (LEVG) is used for this determinant. The
XYMR = +(2?_)-91 >-G) + >-G) 52 :499-2) +5*G5+5)6 )()*+ *,,-),
Dividend payment. I measure dividend payment using dividend dummy variable
(DIVD). If firms pay out dividends, the dummy variable equals 1; otherwise, it is 0.
Financial crisis. The measure of financial crisis is a dummy variable as well, called
post-crisis dummy (PCD). It equals 1 if the sample year belongs to the post-crisis period; otherwise, it is equal to 0.
Finally, the Compustat code table of above equation is in the Appendix.
Table 1
Explanatory Variables Summary Table
Explanatory Variable Factor Expected trend
Market to book ratio Investment opportunities Positive
Ln(Assets) Firms size Positive / Negative
Cash flow ratio Cash flow Positive / Negative
Industry ratio Industrial cash flow uncertainty Positive
Intangible investment ratio Financial distress / Investment opportunities Positive
Net working capital ratio Cash substitute Negative
Capital expenditure ratio Financial distress / Investment opportunities Positive / Negative
Leverage ratio Financial leverage Positive / Negative
Dividend dummy Dividend payment Positive / Negative
Post-crisis dummy Financial crisis Positive
4. Methodology
4.1 Index for identifying the financial constraints
In previous literature, scholars have identified several methods to measure financial constraint, but there is no consistent conclusion as to the optimal measure. In this research, I will adopt four kinds of measures to identify corporate financial constraint.
First, I use Kaplan-Zingales (KZ) index to measure the corporate financial constraint. Kaplan and Zingales (1997) conduct an in-depth study of 49 companies facing financial constraint, with a sample period from 1970 to 1984. By synthesizing qualitative and quantitative information, they divide the firms into five groups according to the level of financial constraint. They then use an ordered Logit regression to obtain the regression coefficients. Lamont, Polk and Saaá-Requejo (2001) adopt a wider sample of firms to build
the KZ index and further measure the level of financial constraint by using the coefficients. The strength of KZ index is that they take leverage and Tobin's Q. Accordingly, the higher KZ index firms have, the more severe are the financial constraint firms may face. In other words, firms with high KZ index belong to the fifth quintile, marked as financially
constrained firms (Chen and Wang, 2012). Firms belonging to other quintiles are financially unconstrained firms. The KZ index is derived as follows (Kaplan and Zingales, 1997):
‘[\ = −1.002 ∗abc_` + 3.139 ∗ XYMR + 0.283 ∗ f − 39.368 ∗hijabc− 1.315 ∗abc_c
where CF/LTA is the ratio of cash flow to lagged total assets; LEVG is the ratio of total debts to the book value of total assets; Q is the market value of total assets over the total asset book value; DIV/LTA is the ratio of cash dividends divided by lagged book value of total assets; and CA/LTA is the ratio of cash balances to lagged book total assets. However, Hadlock and Pierce (2010) argue that the explanatory power of the KZ index for measuring financial constraint is suspect. They explain that the dependent and independent variables in the KZ estimation adopt firm-specific information, which can lead to estimation bias. To handle extreme values, I winsorize the ratios at 1% tails to construct the KZ index.
Second, the Whited-Wu (WW) index is another method to measure financial constraint. Whited and Wu (2006) derive a Euler equation of investment by applying a generalized method of moments to construct a new index for financial constraint, known as the WW index. The WW index is a linear combination of six factors. The strength of this index is that the estimation of the WW index is built upon a large sample range, which could avoid issues such as sample selection and measurement error. Similar to the KZ index, firms with a high WW index have severe financial distress. As for KZ index, firms with the highest WW index are at quintile five, and they are considered financially constrained firms. The calculation of the WW index is (Whited and Wu, 2006):
UU = −0.0910A
#S− 0.062KIMK − 0.044ln (#S) − 0.035<R + 0.021
X#K
where CF/TA is cash flow to total assets ratio; DIVD is a dummy variable of cash dividends; ln(TA) represents the natural logarithm of total assets; SG represents the sales growth rate of firms; INSG represents industrial sales growth classified by three-digit SIC code; and
LTD/TA is the ratio of long-term debt divided by total assets.
According to the previous literature review, firm size and leverage are correlated with financial constraints, and scholars use firm size and leverage as proxies to identify financially constrained firms (Song and Lee, 2012). Thus, the third and fourth methods used in this research to measure financial constraint are firm size and leverage. As mentioned in Section 2.4, firm size equals the natural logarithm of a firm’s total assets. If the size of a firm belongs to the bottom 30% of sample firms, the firm is financially constrained. If the size of a firm belongs to the top 30% of sample firms, the firm is financially unconstrained. By using leverage as proxy, I divide and rank sample firms into ten groups. If the leverage ratio of a firm belongs to the top 30% of sample firms, the firm is high-leverage and financially constrained. If the leverage ratio of a firm belongs to the bottom 30% of sample firms, it is low-levered and financially unconstrained.
4.2 Regression for cash holdings
To test the long-term effect of financial crisis on cash hoarding, I insert the PCD variable into the demand function for cash holdings, and I adopt this variable from the work of Song and Lee (2012). The remaining explanatory variables are from the research of Opler, et al. (1999), which are more deeply elaborated in Section 2.4. The regression method adopted in this paper is panel OLS regression, and the model used for the research question is:
[Model A]
0%mn= o + pq,mn∗ "$% + ps,mn∗ A\ + pt,mn∗ 0A% + pu,mn∗ IIM% + pv,mn∗ TU0% + pw,mn∗ 0*F-J% + px,mn
∗ XYMR + py,mn∗ I< + pz,mn∗ KIMK + pq{,mn∗ 80K + |mn+ }mn
where 0%mnis cash ratio, which is the proxy for cash reserves; o stands for intercept; pq to
the data is of firm i in year t. Furthermore, I create interaction variables by using the PCD variable multiplied by different explanatory variables (excluding the PCD) of cash hoardings to examine how the financial crisis influences the determinants of the cash hoardings. The specific model is:
[Model B]
0%mn= o + pq,mn∗ "$% + ps,mn∗ A\ + pt,mn∗ 0A% + pu,mn∗ IIM% + pv,mn∗ TU0% + pw,mn∗ 0*F-J% + px,mn
∗ XYMR + py,mn∗ I< + pz,mn∗ KIMK + pic,mn(80K ∗ -JF+*2*)(96) + |mn+ }mn
where pic,mnare regression coefficients of the interaction variables; explanatory stands for independent variables of cash holdings excluding the PCD.
5. Empirical Results 5.1 Descriptive statistics
First, there is a description for regression variables, see Table 2. The table displays the number (N) of firm-year observations, mean, median, standard deviation (stdev), minimum (min) and maximum (max) value of each variable. There are four measures to estimate cash ratio, namely cash/asset ratio, cash/sale ratio, cash/net asset ratio, and the logarithm form of cash/net asset ratio. As seen from Table 2, corporate average cash holding is 18.18%
measured by cash/asset ratio, and the median of the cash holding is 10.24%, which is skewed. The average cash ratios measured by cash/sale ratio and cash/net asset ratio are higher than the cash/asset ratio, which is caused by the measure differentials of the denominators. There is not much difference in the median of the three cash ratios.
Table 2
Description of regression variables
This table presents descriptive statistics of variables used in the regression analysis for the sample of firm-year observations within a period from 2000 to 2016. It displays the number (N) of firm-year observations, mean, median, standard deviation (stdev), minimum (min) and maximum (max) for each variable. Cash/asset is cash to total asset ratio, where cash is cash, and short-term investment and asset is book value of total assets. Cash/sale is the ratio of cash to total sales. Cash/net asset is cash to net asset ratio, where the net asset is book total assets after subtracting cash and short-term investments. Ln(cash/net asset) is the cash to net asset ratio in the form of the natural logarithm. The numerator of market-to-book ratio equals that total book asset minus the sum of equity value and market capitalization, and the denominator is total assets’ book value. Firm size is the value of
book total assets in the form of the natural logarithm. Industry sigma is the average standard deviation of industrial cash flow ratio in a 10-year period, where firms are divided into different industries by 2-digit SIC codes. Cash flow/asset ratio is the ratio that cash flow divided by the total assets. Cash flow is calculated as operating income before depreciation minus the sum of interest, income tax, and dividend expense. Intangible Investment Ratio is the ratio of intangible investment to total sales, where intangible investment = R&D + 0.3*SG&A. NWC/Asset ratio is net working capital to total assets ratio, where net working capital equals working capital minus the value of cash and short-term investments. Capex/asset Ratio is the ratio that capital expenditure divided by the amount of total book assets. The leverage ratio is calculated as the sum of long-term debt and debt in current liability divide by the value of total book assets.
Variable N Mean Median Stdev Min Max
Cash/Asset 42,625 0.1818 0.1024 0.2013 0.0000 0.8930 Cash/Sale 40,886 0.2458 0.1026 0.3590 0.0000 2.1899 Cash/Net Asset 40,862 0.2535 0.1042 0.3674 0.0000 2.0563 Ln(Cash/Net Asset) 42,269 -2.2139 -2.1560 1.7184 -7.2629 2.0898 Market-to-Book Ratio 40,886 1.7684 1.4184 1.0488 0.5514 6.3211 Firm Size 41,755 5.6424 5.7108 2.2597 0.4929 11.3077 Industry Sigma 43,491 0.0612 0.0632 0.0157 0.0198 0.1000
Cash Flow/Asset Ratio 43,494 0.0175 0.0503 0.1332 -0.3882 0.3518
Intangible Investment Ratio 39,145 0.0610 0.0431 0.0894 0.0000 0.3600
NWC/Asset Ratio 35,143 0.0779 0.0576 0.1566 -0.2402 0.5409
Capex/Asset Ratio 42,845 0.0414 0.0262 0.0479 0.0000 0.3158
Leverage Ratio 42,625 0.2468 0.1900 0.2537 0.0000 1.5552
For most of the variables, the standard deviation is relatively small. It indicates that the distribution of the variables is concentrated, and the effect caused by outliers is under control. The numerator of the MTBR equals total book assets minus the sum of equity value and market capitalization, and the denominator is the book value of total assets. The mean and median of the MTBR are 1.7684 and 1.4184, respectively, and the standard deviation is 1.0488. Firm size is the value of total book assets in the form of the natural logarithm. The mean and median of the firm size are 5.6424 and 5.7108, respectively, which has a standard deviation of 2.2597. I use IS to capture the industrial cash-flow uncertainty. The IS is the average standard deviation of industrial cash flow in a ten-year period, and the detailed calculation is shown in Section 3. The mean and median of the IS are 6.12% and 6.32%, respectively, which are similar. Intangible investment contains the expense of R&D and SG&A, and IIVR is intangible investment divided by total sales. The mean IIVR is 6.10%, while the median is 4.31%. Another investment ratio is Capex/Asset Ratio, which is the ratio
of capital expenditure divided by total assets. The mean and median of the Capex/Asset Ratio are 4.14% and 2.62%, respectively, which has a standard deviation of 4.79%. NWC/Asset Ratio is the proxy of the cash substitute, which is the ratio of net working capital divided by the book value of total assets. The mean NWC/Asset Ratio is 7.79%, while the median is 5.76%. Leverage Ratio is the sum of the long-term debt and the debt in current liability divided by total book assets, and its mean and median are 24.68% and 24.68%, respectively. 5.2 Cash holdings, and financial ratios
This section presents the annual data of cash, leverage, and investment ratio with a period from 2000 to 2016. Cash ratio is the Cash/Asset Ratio, which is the sum of cash and marketable securities divided by the book value of the total assets. The second column of Table 2 presents the number of sample firms for each year. The third column shows that the mean cash ratio gradually increases from 18.45% to 21.45% between 2000 and 2007. In 2008, there a drop in the mean cash ratio, which proves that firms are choosing to spend their cash holdings to ensure business operation in the short run. From 2009 to 2012, the mean cash ratio has been steady at around 21%. After that, the mean cash ratio gradually grows in the following years, and reaches a level of 22.1% in 2016, peaking in 2014. The median cash holdings increase from 6.86% in 2000 to 10.27% in 2016. During this period, there is a drop between 2007 and 2008, which is in line with the trend of the mean cash ratio. Furthermore, I conduct the Chow test to examine whether there is a structural change of cash ratio after the 2007–2008 financial crisis. The p-value of the test results shows that it is significantly less than 0.01. This suggests that a time series of the post-crisis period cash ratio is not the m same as the cash ratio in the pre-crisis period.
Table 3
Cash, Leverage and Investment Ratio from 2000 to 2016
Table 3 includes the mean and median of cash, leverage and investment ratios by year. Cash ratio is the Cash/Asset Ratio, which is the sum of cash and marketable securities divided by the book value of the total assets. Leverage Ratio is calculated as the sum of long-term debt and debt in current liability divide by total book assets. Investment ratio is the ratio of capital expenditure expense to total assets.
Cash Ratio Leverage Ratio Investment ratio
Year N Mean Median Mean Median Mean Median
2000 3,564 0.1845 0.0686 0.2697 0.2235 0.0545 0.0381 2001 3,417 0.1850 0.0774 0.2666 0.2164 0.0468 0.0320 2002 3,187 0.1879 0.0843 0.2557 0.2000 0.0398 0.0265 2003 3,055 0.2007 0.0934 0.2419 0.1910 0.0378 0.0242 2004 2,934 0.2123 0.1043 0.2264 0.1656 0.0401 0.0258 2005 2,845 0.2139 0.1043 0.2235 0.1587 0.0416 0.0271 2006 2,692 0.2161 0.1023 0.2236 0.1668 0.0439 0.0274 2007 2,593 0.2145 0.0987 0.2343 0.1730 0.0438 0.0263 2008 2,414 0.1984 0.0935 0.2491 0.1919 0.0447 0.0280 2009 2,507 0.2111 0.1139 0.2308 0.1649 0.0324 0.0187 2010 2,249 0.2161 0.1197 0.2203 0.1567 0.0356 0.0221 2011 2,132 0.2136 0.1100 0.2342 0.1748 0.0406 0.0248 2012 2,101 0.2132 0.1087 0.2407 0.1940 0.0422 0.0251 2013 2,067 0.2241 0.1111 0.2420 0.1846 0.0390 0.0236 2014 2,011 0.2242 0.1061 0.2577 0.2039 0.0384 0.0242 2015 1,960 0.2215 0.1004 0.2931 0.2393 0.0385 0.0235 2016 1,766 0.2210 0.1027 0.2961 0.2503 0.0345 0.0219 Whole period 43,494 0.1903 0.0856 0.2467 0.1909 0.0414 0.0262
As with the increase in cash holdings, it may affect the measurement of investment activity and leverage. Column 5 and Column 6 of Table 3 report the average and median leverage ratios, respectively. Leverage ratio is calculated as the sum of long-term debt and debt in current liability divided by total book assets. The mean leverage increase is from 26.97% in 2000 to 29.61% in 2016. Meanwhile, the average leverage ratio keeps at a level of around 22% to 23% during the period from 2004 to 2010. Until 2014, the mean leverage ratio is 25.77%, which is lower than the mean leverage ratio in 2000. In 2015, the average leverage ratio increases to 29.31%, nearly 4%. The median leverage ratio rises from 22.35% in 2000 to 25.03% in 2016, and its growth pattern is similar to the pattern of the mean leverage ratio. Column 7 of Table 3 reports the mean of the investment ratio. The investment ratio is the ratio of capital expenditure to total assets. During the period between 2000 and 2016, the mean investment ratio declines from 5.45% to 3.45%. From 2009 to 2016, the mean investment ratio keeps at a level of around 3%, except for 2011 and 2012. This pattern
verifies the theory that investment activities may decline with the increase of the cash
holdings. The next column displays the median investment ratio of the sample companies by year. The median investment ratio falls from 3.81% in 2000 to 2.19 in 2016, which decreases to its lowest value at 1.87% in 2009.
5.3 Cash holdings and firm characteristics
The previous literature suggests that there is a chance that companies hoard more cash after financial crisis, caused by the changes in their characteristics. Table 4 reports the
average firm characteristics for the pre-crisis (2000–2006) period and the post-crisis (2009– 2016) period. I conduct t-tests to investigate whether the mean differences between the pre-crisis period and the post-pre-crisis period significantly different from zero, and the p-value of the t-tests is in Column 3 of Table 4. Compared with the variables of the pre-crisis period, the post-crisis variables significantly change at a significance level of 5%. The mean MTBR slightly declines from 1.785 to 1.784. The mean firm size increases from 5.323 to 6.03 after the financial crisis. A plausible explanation is that some small companies may not survive the financial crisis period. The industrial cash-flow risk (IS) increases from 5.7% to 6.3%. The average cash-flow ratio drops slightly from 2% to 1.4%, and the decline of cash-flow ratio suggests that the growth of the cash ratio cannot be explained by the increasing amount of internal cash-flow. The average ratio of the capital expenditure decreases from 4.4% to 3.8%, which is in line with the findings presented in Table 3. There is also a decline in IIVR, which drops from 7.8% to 5.9%. The average NWC/asset ratio declines from 8.4% to 7.1%, which means that the cash substitute of the firms decreases in the post-crisis period. The average ratio of leverage slightly increases from 24.6% to 25%, which is consistent with the findings presented in Table 3. The previous literature shows that the declines in MTBR and IIVR are negatively related to cash reserves, while the decrease in the ratio of net working capital and the increased IS may have a positive impact on the cash hoardings. Thus, due to the
significant post-crisis period changes in the firm characteristics, the effects caused by the firm characteristics on the cash reserves are mixed and may be offset by each other.
Table 4
Firm Characteristics Difference Test
Table 4 reports the mean firm characteristics for the pre-crisis (2000-2006) period and the post-crisis (2009-2016) period. I conduct t-tests to investigate whether the mean differences between the pre-crisis period and the post-crisis period are significantly different from zero, and the p-values of the t-tests present in column 3 of Table 4.
1 2 3
Pre-crisis Post-crisis p-value of
VARIABLES (2000-2006) (2009-2016) difference
Market-to-Book Ratio 1.785 1.784 0.033
Firm Size 5.323 6.030 0.000
Industry Sigma 0.057 0.063 0.005
Cash Flow/Asset Ratio 0.020 0.014 0.033
Intangible Investment Ratio 0.078 0.059 0.021
NWC/Asset Ratio 0.084 0.071 0.003
Capex/Asset Ratio 0.044 0.038 0.000
Leverage 0.246 0.250 0.000
I further estimate the post-period function of cash holding to eliminate the possibility that the changed firms' characteristics are responsible for the growth in cash hoardings by adopting the regression introduced by Fama-MacBeth (FM) (1973). The coefficients in FM regression are the average coefficients, which are from cross-section regression for each year.
Table 5 presents average predicted cash/asset ratios of pre-existing samples with a period from 2009 to 2016. The average differences are the values of the average cash/asset ratios minus the average predicted cash/asset ratios. The sample firms are divided into five groups, and the whole sample contains all sample firms. The rest of the groups are high KZ index, low KZ index, high-leverage, and low-leverage companies, where high KZ index and high-leverage firms are financially constrained. The FM regression measures the predicted cash/asset ratio with the assumption that there is no change of the cash hoarding function and firm characteristics. Based on this assumption, the estimated FM regression using the pre-crisis data should be able to predict the cash/asset ratio for the post-pre-crisis period. The description in Table 5 reports the estimated pre-crisis FM regression, and the coefficients of the pre-crisis FM regression are all significantly different from zero at the 1% significance
level. If the assumption is valid, the differences should not be significantly different from 0. However, the results presented in Table 5 display the opposite story. For the sample group of all firms, the average predicted cash/asset ratios range from 15.1% to 17.6%, and the average differences are significantly greater than 0 at a significance level of 1%. This means that the pre-crisis FM regression cannot estimate the post-crisis cash holding accurately. Moreover, unlike the findings in the research of Han and Qiu (2007), the changes of firm characteristics cannot adequately explain the increases in cash reserves after the financial crisis. Moreover, I move into the average predicted cash/asset ratio of financially constrained and unconstrained firms, using KZ index and leverage ratio as the proxy of financial constraints. The difference between the average cash/asset ratio and the average actual of the rest sample groups shows the same result as the whole sample group, for which the average differences are positively significant from zero at the 1% significance level.
Table 5
Predicted Cash Holdings and the Mean Differences in the Post-crisis Period
Table 5 presents the average predicted ratio of cash holdings (cash/asset ratio) of pre-existing sample companies with the period from 2009 to 2016. The difference is the value that the mean actual cash ratio minus the mean predicted cash ratio. There are 5 sample groups presented in Table 5, which are the whole sample, high KZ index, low KZ index, high-leverage, and low-leverage firms. High KZ index and high-leverage firms represent financially constrained firms; low KZ index and low-leverage firms represent financially unconstrained firms. I adopt Fama-MacBeth regression (1973). The coefficients in FM regression is the average coefficients which are from cross-sectional regression for each year. The estimation of the pre-crisis FM presents as follows:
Cash/Asset Ratio = 0.1967 + 0.0432 Market-to-Book Ratio – 0.0093 Firm Size – 0.2569 Cash flow/Asset ratio + 0.9517 Industry Sigma + 0.1970 Intangible Investment Ratio – 0.2513 NWC/Asset Ratio – 0.4524
Capex/Asset Ratio – 0.2769 Leverage – 0.0244 Dividend Dummy. The value in parentheses reports the p-value of t-test which investigates whether the average differences are significantly different from 0.
Whole Sample High KZ Index Firms Low KZ Index Firms High-leverage Firms Low-leverage Firms
Year Predicted Predicted Actual - Predicted Predicted Actual - Predicted Predicted Actual - Predicted Predicted Actual - Predicted Predicted Actual -
2009 0.170 0.024 0.072 0.071 0.183 0.023 0.076 0.040 0.255 0.060 (0.000) (0.000) (0.002) (0.000) (0.000) 2010 0.176 0.016 0.083 0.056 0.186 0.015 0.087 0.029 0.255 0.047 (0.004) (0.001) (0.030) (0.001) (0.000) 2011 0.165 0.022 0.066 0.082 0.177 0.018 0.073 0.046 0.251 0.054 (0.000) (0.000) (0.005) (0.000) (0.001) 2012 0.154 0.024 0.085 0.066 0.164 0.020 0.074 0.039 0.237 0.065 (0.001) (0.000) (0.010) (0.000) (0.000) 2013 0.166 0.028 0.102 0.061 0.175 0.027 0.086 0.046 0.253 0.060
Table 5 – Continued
Whole Sample High KZ Index Firms Low KZ Index Firms High-leverage Firms Low-leverage Firms
Year Predicted Predicted Actual - Predicted Predicted Actual - Predicted Predicted Actual - Predicted Predicted Actual - Predicted Predicted Actual -
(0.003) (0.000) (0.000) (0.000) (0.000) 2014 0.168 0.023 0.100 0.056 0.180 0.021 0.084 0.038 0.268 0.057 (0.000) (0.002) (0.005) (0.000) (0.000) 2015 0.151 0.028 0.081 0.063 0.165 0.025 0.080 0.029 0.247 0.066 (0.004) (0.004) (0.001) (0.001) (0.000) 2016 0.152 0.024 0.090 0.059 0.164 0.020 0.087 0.025 0.246 0.064 (0.000) (0.001) (0.000) (0.000) (0.000)
Moreover, the average predicted cash/asset ratios of low KZ index firms, ranging from 16.4% to 18.6%, are more significant than the average predicted cash/asset ratios of high KZ index firms, ranging from 6.6% to 10%. The sample groups using leverage ratio to identify financial constraints have a similar pattern. In other words, the FM regression predicts that financially unconstrained companies may hold more cash reserves compared with the financially constrained companies. The findings of Song and Lee (2012) do not make this kind of prediction. From the results presented in Table 4 and Table 5, the post-crisis firm characteristics are significantly different from the pre-post-crisis firm characteristics. However, the changes of firm characteristics are not responsible for the increase in cash reserves during the post-crisis period. Thus, I do not reject Hypothesis 1, that the changes of firm characteristics cannot explain the growth in cash holdings.
5.4 Has the function of cash holdings changed after the 2007–2008 financial crisis? In the previous discussion, I consider the possibility that changes of firm
characteristics are responsible for the increase in cash holdings. To further investigate the increased post-crisis cash reserves in the US firms, I extend the research by adding new variables to the equation of cash holdings. I follow the model of Opler, et al. (1999) to
estimate the regression for cash holdings. The type of the samples is panel data, and the panel OLS regression is applied in this paper. In Table 6, it reports the results of FM (1973)
regression and the panel OLS regression with firm fixed effect. Besides, there is an estimated regression for the changes in variables instead of using the levels of the variables as a
dependent variable. Based on the work of Bates, Kahle and Stulz (2009), there are four cash ratios used in the panel OLS regression as the proxy of cash holdings, which are cash/asset ratio, cash/sale ratio, cash/net asset ratio, and ln(cash/net asset). The estimates in Table 6 use the sample of all companies within a period from 2000 to 2016, and the sample set contains 29,690 firm-year observations. To examine whether cash holding is affected after the financial crisis, I insert the PCD variable in the panel OLS regression, Model A. Using this variable allows shifts of the intercept.
The dependent variable used in Model 1 of Table 6 is cash/asset ratio, and the estimated regression is the panel OLS regression. According to the precautionary motive, companies with more investment opportunities tend to hoard more cash reserves, because these firms suffer considerably from bypassing valuable investment opportunities. Following Opler, et al. (1999), I use MTBR as the proxy of investment opportunities, and I expect that MTBR has a positive relationship with cash reserves. The coefficient of the MTBR is 0.0273, and its p-value is less than 1%. This means that cash/asset ratio will increase by 2.73 basis point if MTBR increases by 1%. This result is consistent with the relationship between the investment opportunities and cash holdings. For firm size and cash flow, they can be either positively or negatively correlated with cash holding. From the estimated results of Model 1, they have negative coefficients, namely -0.0109 and -0.0522. The coefficient of firm size suggests that firms prefer to decrease their retained cash when the firm size is getting larger. The coefficient of cash-flow/asset ratio suggests that firms with high-volume cash flows would lower the number of cash reserves they hold. The results of firm size and cash flow show an opposing viewpoint to the pecking-order theory because the pecking-order theory suggests that firms with large size or high-volume cash flows tend to hold more cash. I expect
Table 6
Estimated Regressions on Cash Holdings for Whole Sample
Table 6 reports the regression results on cash holdings for used a sample of all firms. In Model 1, 2, 4, and 6, cash/asset ratio is the dependent variable. For the rest of models, ln(cash/net asset) is the dependent variable. Lag dcash is the difference between cash/asset ratio and lagged cash/asset ratio. Model 4 is a Fama-MacBeth regression. In Model 5 and 8, there are regressions with the fixed effect at the firm level. In parentheses, the values are robust standard errors for each coefficient. Note: ***, **, and * represent statistical significance level at the 1%, 5%, and 10%.
Model A Model B
1 2 3 4 5 6 7 8
Model OLS Changes OLS FM FE OLS OLS FE
VARIABLE Cash/Asset Cash/Asset Ln (Cash/Net Asset) Cash/Asset Ln (Cash/Net Asset) Cash/Asset Ln (Cash/Net Asset) Ln (Cash/Net Asset)
Intercept 0.202*** 0.0350*** -2.296*** 0.197*** 0.211*** -2.153*** -1.627*** (0.0101) (0.00123) (0.0684) (0.0066) (0.0104) (0.0907) (0.176) Lag dcash -0.0949*** (0.00952) Lag cash -0.205*** (0.00690) Market-to-Book Ratio 0.0273*** 0.0116*** 0.258*** 0.0432*** 0.191*** 0.0279*** 0.268*** 0.196*** (0.0013) (0.00139) (0.0083) (0.0020) (0.0094) (0.0015) (0.0122) (0.0139) Size -0.0109*** -0.00779** -0.0840*** -0.00933*** -0.125*** -0.0117*** -0.102*** -0.156*** (0.0011) (0.00330) (0.0073) (0.0006) (0.0139) (0.0012) (0.0098) (0.0252)
Cash Flow/Asset Ratio -0.0522*** 0.0380*** -0.303*** -0.257*** 0.153* -0.0626*** -0.310*** -0.0438
(0.0122) (0.0112) (0.0772) (0.0189) (0.0878) (0.0146) (0.106) (0.145)
Industry Sigma 0.878*** 0.139 9.466*** 0.952*** 6.965*** 0.814*** 9.161*** 6.444***
Table 6 – Continued
Model A Model B
1 2 3 4 5 6 7 8
Model OLS Changes OLS FM FE OLS OLS FE
VARIABLE Cash/Asset Cash/Asset Ln (Cash/Net
Asset) Cash/Asset Ln (Cash/Net Asset) Cash/Asset Ln (Cash/Net Asset) Ln (Cash/Net Asset)
Intangible Investment Ratio 0.142*** -0.0456 1.893*** 0.197*** 0.452** 0.158*** 1.983*** 0.528
(0.0238) (0.0392) (0.135) (0.0181) (0.214) (0.0257) (0.183) (0.325) NWC/Asset Ratio -0.245*** -0.256*** -2.053*** -0.251*** -1.915*** -0.258*** -2.260*** -2.149*** (0.0108) (0.0124) (0.0656) (0.0094) (0.0786) (0.0113) (0.102) (0.135) Capex/Asset Ratio -0.329*** -0.229*** -3.053*** -0.452*** -2.547*** -0.290*** -2.703*** -2.309*** (0.0205) (0.0191) (0.166) (0.0247) (0.187) (0.0231) (0.256) (0.292) Leverage -0.200*** -0.111*** -2.145*** -0.277*** -1.803*** -0.205*** -2.248*** -1.864*** (0.0075) (0.00858) (0.0420) (0.0070) (0.0488) (0.0083) (0.0851) (0.100) Dividend Dummy -0.00360 -0.00893*** -0.0444** -0.0244*** 0.0431* -0.00479 -0.0496 0.0361 (0.0027) (0.00132) (0.0201) (0.00171) (0.0222) (0.0032) (0.0344) (0.0393) Post-crisis Dummy 0.0137*** 0.0130*** 0.271*** 0.286*** (0.0020) (0.0011) (0.0125) (0.0134) dummy * Market-to-Book -0.00212 -0.0323* -0.0313 (0.0022) (0.0180) (0.0195) dummy * Size 0.00115 0.0411*** 0.0471*** (0.0009) (0.0092) (0.0098)
dummy * Cash Flow 0.0762*** 0.488*** 0.498**
(0.0216) (0.178) (0.194)
dummy * Industry Sigma 0.104 -0.127 -0.476
(0.105) (0.965) -1.019
Table 6 – continued
Model A Model B
1 2 3 4 5 6 7 8
Model OLS Changes OLS FM FE OLS OLS FE
VARIABLE Cash/Asset Cash/Asset Ln (Cash/Net
Asset) Cash/Asset Ln (Cash/Net Asset) Cash/Asset Ln (Cash/Net Asset) Ln (Cash/Net Asset) (0.0262) (0.201) (0.217) dummy * NWC 0.0406*** 0.589*** 0.668*** (0.0114) (0.130) (0.136) dummy * Capex -0.125*** -1.132*** -0.835** (0.0337) (0.381) (0.410) dummy * Leverage 0.0151* 0.261** 0.174 (0.0090) (0.103) (0.108) dummy * Dividend 0.0026 0.0112 0.0156 (0.0038) (0.0408) (0.0428) Observations 29,69 29,69 29,432 11,490 29,432 29,69 29,432 29,432 Adjusted !" 0.4116 0.1759 0.4088 0.368 0.7211 0.4103 0.4117 0.7232
that industrial cash-flow uncertainty and intangible investment are positively correlated with cash/asset ratio, and the estimated coefficients are consistent with the results from the previous literature. Net working capital is regarded as the cash substitute, and its coefficient is -0.245, which is as negative as I expect. The previous literature suggests that capital expenditure, leverage, and dividend payment can be either positively or negatively correlated with cash holdings. The coefficients of these variables are -0.329, -0.200, and -0.0036;
however, the coefficient (-0.0036) of the dividend payment is not significant. The coefficients of capital expenditure and leverage prove the viewpoint of the pecking-order theory that these variables have negative impacts on cash holding. The estimated coefficients in Model 1 are all significant at the 1% level, excluding the coefficient of the dividend dummy.
In Model 2 of Table 6, there is an estimated regression for the changes in variables instead of using the levels of the variables as dependent variables. This estimation can eliminate the effect of fixed imperceptible firm characteristics on cash hoardings. I include lagged level of cash and lag dcash as new independent variables, where lag dcash is the difference between cash/asset ratio and lagged cash/asset ratio. Due to these new variables, this approach allows partial adjustment for cash/asset ratio to the equivalent level. The coefficients of Model 2 are slightly different in both sign and significance level, compared with the results of Model 1. The coefficients of Model 2 are significantly different from 0 in a significance level of 1%, excluding the coefficient of IS and intangible investment. Model 3 runs the same regression as Model 1, using ln(cash/net asset) as the dependent variable. In Model 3, the sign of the dependent variables’ coefficient is consistent with the results of Model, and the p-values of variables are all significantly less than 1%.
To examine the effect of the financial crisis on cash holdings, I insert the PCD variable in Model A. If a firm-year observation is in the post-crisis period between 2009 and 2016, the PCD variable is equal to 1; otherwise it is 0. From the previous literature (Almeida,
Campello and Weisbach, 2004; Ang and Smedema, 2011; Song and Lee, 2012; Campello, et al., 2018), financial crisis is a negative shock to the external capital market, and firms tend to hoard more cash to avoid cash shortages due to the contraction of the capital supply. Thus, the relationship between financial crisis and cash hoarding should be positive. In Model 1, the estimated coefficient of the PCD is 0.0137, which is significant at the 1% significance level. It means that cash/asset ratio would increase by 1.37% after the financial crisis. The coefficients of the PCD in Model 2 and Model 3 are 0.013 and 0.271, respectively, with a p-value less than 1%. Because the sample is panel data, I run Model A with firm fixed effect, Model 5, and the coefficient of the dummy variable is 0.286, which is significant at 1% significance level. Model 4 in Table 6 presents an FM regression, but the PCD is not significant in this regression. For the rest of the coefficients in Model 4, there is no
significant difference compared with the results in Model 1. This indicates that the results of these two regressions are consistent. In general, these findings indicate that the demand for cash reserves has a significant upward shift after the 2007–2008 financial crisis, and Hypothesis 2, that financial crisis is positively correlated to cash holdings in a long-run period, is valid.
To further investigate how financial crisis influences the determinants of the cash holdings, I insert interaction variables in the regression of cash holdings, Model B in Table 6. Interaction variables are the variables which use the PCD variable multiply by firm
characteristics. Model 6 and Model 7 use cash/asset ratio and ln(cash/net asset) as dependent variables, and Model 8 re-estimates Model 7 with firm fixed effect. In Model 6, the
coefficients of the interaction variables with cash flow, intangible investment, capital
expenditure, and leverage are statistically significant. These results indicate that the relation between firm characteristics and cash hoardings has changed during the post-crisis period. It is worth noting that the coefficient of cash flow has a significant change from -0.0626 in the
