1
Corporate cash holding: Do firms use
cash holding as an instrument to mitigate
future risks? A Brexit analysis
Name: Paul Koopman
Student number: 10642366
MSc Finance: Corporate Finance (track)
Master thesis, University of Amsterdam
Supervisor: Dr. V. Vladimirov
2
Statement of Originality
This document is written by Student Paul Koopman who declares to take full responsibility for the contents of this document.
I declare that the text and work presented in this document is 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.
3 Abstract
This study investigates the effect of the Brexit referendum announcement on corporate cash holding. Using the announcement as an exogenous event the precautionary motive of corporate cash holding is examined. A cross-sectional time-series model and a differences-in-differences model resulted that no evidence is found in favor of the precautionary motive. However, evidence is found that the cash holdings of UK firms have significantly declined after the Brexit announcement. In addition, the effects between financial constraint and not financial constraint, public and private firms and firms experiencing high agency frictions to low agency frictions are examined. From these tests, weak evidence is found that financial constraint firms, public firms, and firms without analyst coverage had greater declines in cash holding. No difference is found for the appearance of a credit rating. The robustness checks showed that UK firms have greater (but insignificant) decreases in cash holdings relatively to German firms. A possible reason for the decline in corporate cash holding for UK firms is that firms are reacting on the uncertain future by investing in their competitive and export position.
Keywords: Brexit, corporate cash holding, precautionary motive, financial constraint, differences-in-differences, cross-sectional time-series
4 Table of contents
I. Introduction ... 5
II. Theoretical framework and empirical hypotheses ... 7
A. Theoretical framework ... 7
B. Brexit – implications and scenarios ... 10
C. Hypothesis development ... 13
III. Methodology ... 14
A. Cross-sectional time-series model ... 14
B. Differences-in-Differences model ... 16
IV. Sample formation and descriptive statistics ... 18
A. Sample formation ... 18
B. Construction of variables ... 19
C. Descriptive statistics ... 20
V. Results ... 24
A. Results from cross-sectional time-series model ... 24
B. Results from differences-in-differences model ... 29
C. Robustness checks ... 34
VI. Conclusion ... 36
Reference list ... 38
5 I. Introduction
On the 22nd of February 2016, Prime Minister David Cameron announced a referendum in which the
residents of the UK could choose whether leaving the European Union or not. When the UK is leaving the EU, it is also called ‘Brexit’, referring to an exit of Great-Britain. On the 24th of June 2016, the
outcome of this referendum was announced: the UK residents voted, with 51.9 percent of the total votes, to leave the EU. The vote against staying in the EU was an unexpected event, both domestically as internationally, and has severe implications for the countries as well as for the firms operating in the UK. After the results were announced to the public, the market, credit agencies and central banks reacted on this event. The FTSE100 and 250 indexes both lost approximately 5.6 percent and 13.6 percent respectively in the four days after. The pound sterling depreciated to a 30-year low level and the UK lost its AAA-rating given by the credit ratings. In addition, the Bank of England (BoE) dropped its quoted interest rate from 0.5 percent to 0.25 percent to maintain a sustainable money market. From these reactions, it has shown that the market was surprised and did not incorporate the presumably negative effects in their assets.
The determinants and implications of corporate cash holding is a widely-researched topic in the past literature. From this literature, it is concluded that there are four motives to maintain higher cash holdings than needed: the transaction motive, precautionary motive, tax motive, and the agency motive1.
However, research on the motives of corporate cash holding in relation with an exogenous shock is limited. Since the Brexit increases the uncertainty about the future, it becomes riskier. This study will therefore focus on the precautionary motive in relation with the announcement of the Brexit referendum. By using this event it is possible to analyze if the precautionary motive holds.
The research question of this study reads if firms use cash holding as an instrument to mitigate future risks. The hypothesis is formulated as that firms in the UK have increased their cash holdings after the announcement took place. In addition, it is hypothesized that the increase in corporate cash holding should be greater for firms which are more financial constraint, firms which are private, and firms which have more agency problems. The prove or disprove of these hypotheses and research question will be obtained by employing two complementary approaches: a cross-sectional time-series and a differences-in-differences model. Annual sample data from UK firms over the period 2014 to 2016 and quarterly data over the period of Q1-2015 to Q4-2016 are used to check the consistency of the results.
First, a static and dynamic time-series model (Guney, Ozkan and Ozkan, 2003) is used. The results show a significant decline in cash holdings in 2016, relatively to the previous two years. This result is in contrast with the hypotheses. A possible explanation for the decline in cash holdings is that firms are
1 The corporate cash holding motives are widely discussed by Keynes (1936), Opler et al. (1999), Bates et al. (2009), Foley et al. (2007) and Jensen (1986), among others.
6
investing more in their domestic competition position. In addition, it is possible that firms are using (already accumulated) cash to invest in foreign projects to ensure incoming cash flows and to being able to pursue the business after the Brexit. As stated above, the pound sterling depreciated to a 30-year low level and could therefore have a large effect on the export of a firm.
The announcement of the Brexit referendum gives an opportunity to examine other determinants of cash holding. Therefore, this study is extended with tests which investigate if firms with different kind of characteristics are affected differently. The first test is analyzing the difference in cash holdings between private and listed firms. Brav (2009) found that privately held firms are more sensitive to cash flow shocks and should therefore rely more on precautionary cash holdings. In this study, weak evidence is found that public firms have decreased their cash holdings more than private firms. Secondly, two tests analyze the effect of information symmetry and the effect of the Brexit referendum on corporate cash holding. Firms with higher information asymmetry experience more trouble in raising new funds (Jensen, 1986). Dittmar et al. (2003) find that these frictions result in higher cash holdings. The first test focusses on firms which have been assigned a credit rating to firms which are not. A credit rating or more analyst coverage mitigate these agency problems. The results show that there is no difference between firms which have a credit rating relatively to firms without a credit rating. The second test shows that firms without analyst coverage have greater decreases in cash holdings than moderate or intensive analyst coverage.
Secondly, a differences-in-differences model is conducted to test the consistency of the results and to examine the effect on financial constraint firms. The treatment groups are separated in two variants, namely an industry risk variable and two financial constraint variables. The industry risk variable divides the treatment and control group by the standard deviation of the cash flows per three-digit SIC industry classification, while the financial constraint variables divide the sample by the Hadlock and Pierce (2010) financial constraint index (HP-index) and the variable dividend payout. The results show that firms in more riskier industries and firms which are more financial constraint hold significantly more cash. The coefficients of interest show that firms in riskier industries and financial constraint firms have decreased their cash holdings more than their counterparts. However, these results are not significant.
Two robustness checks are performed to critical assess the obtained results. Firstly, a comparison is made between the UK and Germany with regards to corporate cash holding after the announcement of the referendum. By using data from Germany, no evidence is found that the UK firms are behaving significant differently than German firms. The results show that UK firms have greater declines in their cash holdings than German firms. Secondly, the treatment and control group are separated by an additional financial constraint measure: The Whited and Wu (2006) index. In contrast to Hadlock and Pierce (2010), Whited and Wu use accounting variables to measure in which extent a firm is financial
7
constraint. The results are in line with the other financial constraint measures, and show that financial constraint firms did not increase their cash holdings more than less financial constraint firms.
This paper does contribute in three ways to the recent literature. First, the current literature does not provide an approach in which firms tend to hold less cash because of an event in which the future is more uncertain. Since the Brexit is regarded as an exogenous event in which the risk increases, this event is used to mitigate endogeneity problems encountered by the recent literature of corporate cash holding. Secondly, because the Brexit has recently occurred, limited literature is present about countries leaving the EU. This paper will shed light on the implications to firms which operate in the first country that leaves the EU in depth. Thirdly, this paper is the first one to investigate the effect of the Brexit on corporate cash holding of UK firms.
This paper will continue as follows: In Section II the related literature about corporate cash holding and the Brexit will be elaborated. In addition, the hypotheses will be developed. In Section III the sample will be formatted; the variables will be constructed and the descriptive statistics will be presented. Section IV outlines the methodology which is used to conduct the tests. In Section V the results from the analyses will be showed. Section VI concludes and discusses this study.
II. Theoretical framework and empirical hypotheses
Before analyzing the impact of the ‘Brexit’ on corporate cash holding, the recent literature is used to develop the hypotheses. It is necessary to understand the broad underlying theory of corporate cash holding. Therefore, Section A will focus on the motives behind corporate cash holding. Section B will elaborate the event of an European country leaving the EU and which implications it has on the firms operating in this country. At last, in Section C, the hypotheses will be developed using the literature explained earlier.
A. Theoretical framework
One of the most interesting debates in the corporate cash holding literature is the reason why firms excessively hold cash on their balances. In the perspective of perfect capital markets, in which capital can be transferred from one party to another without any costs, cash holding would have no value increasing function. Firms in need of financing will always be able to issue funds from outside investors without any frictions. This view, introduced by Modigliani and Miller (1958), is an utopic assumption. Firms must deal with certain market frictions. Transaction costs, agency costs, corporate taxes, and distress costs are a few examples of these frictions. These frictions affect the operating process of firms. For example, an increased variability in cash flows and investment making decisions.
Keynes (1936) is among the first to provide motives for cash holding. Keynes described that firms have an increased demand for cash due to cost limiting reasons. Two motives resulting from this insight
8
are the transaction motive and the precautionary motive. An additional motive, the tax motive, is argued by Foley et al. (2007). The last motive for corporate cash holding stems from Jensen’s (1986) agency theory. These motives are explained in detail below.
The first motive is the transaction motive. This motive relies on the fact that firms derive the optimal demand for cash when they do not have to convert an asset into cash for operating reasons. When firms must convert assets to raise cash, they will incur transaction costs (Keynes, 1936). In addition, raising funds externally by entering the external capital markets is assumed to be costly in other dimensions. These costs can have a negative impact on firms which are in need of liquid assets to finance certain investment opportunities. Even when firms already assessed the capital market or having credit lines available it can be costly to maintain low cash levels. When liquidity is most valuable for firms, banks which offer credit lines and investors are most likely reluctant to deliver funds (Opler et al, 1999). In addition, economies of scale is related to the transaction motive. Mulligan (1997) found, in a sample of 12,000 firms in the period of 1961 – 1992, that larger firms tend to hold less cash on their balance. Larger firms have an improved access to capital markets with respect to smaller firms. In addition, larger firms are more likely to be under supervision of credit agencies, analysts, and financial institutions (which relates partly to the agency motive which is explained below).
The second motive is the precautionary motive. Under this theory, firms are expected to maintain higher cash levels as a buffer to be able to protect themselves against future cash flow shocks. Consistent with this view, Opler et al. (1999) find that firms with riskier cash flows and firms with financial constraints hold more cash to reduce risk. Additionally, it suggests that when firms have better opportunities to invest, they should hold more cash because inability to invest is costlier for these firms (Bates et al., 2009). Recent empirical research on corporate cash holdings has found consistent support for this precautionary motive. Furthermore, corporate cash holding is more important for firms with external investors. Bates et al. (2009) found in a sample of public US firms additional support for precautionary cash holdings. They show that one of the main reasons for precautionary cash holding is cash flow risk. In addition, Brav (2009) used a sample of private UK firms to show the difference between privately and publicly held firms. There are large differences between the financial policies of public and private firms. Private firms have leverage ratios which are approximately 50 percent higher on average than public firms. Private firms raise approximately 10 percent of their financing by equity offerings while their public counterparts issue around 40 percent of their capital from equity offerings. Private firms rely on internally generated capital, and are therefore affected by dilution of this generated capital. Furthermore, firms which access capital markets less, will stockpile cash in good times and dilute cash holdings faster in worse times. Hence, the results show that private firms are more sensitive to cash flow shocks, which results in increased precautionary cash holding.
9
The precautionary motive is closely related to the extent in which firms are financial constraint. Financially constraint means that firms are not able to finance all projects in which it wants to participate. This means that firms will not invest in possible positive net present value projects, even when the project is value enhancing. Duchin et al. (2010) showed, by studying the effect of corporate investment during the recent financial crisis, that uncertainty in cash flows explains the excessive cash levels of firms. This uncertainty is especially important for firms with financial constraints, firms which are dependent of external finance, and firms with low cash reserves. Kaplan and Zingales (1997) investigated the effect of financing constraints on investment cash flow sensitivities for a sample of 49 low dividend firms. They found that firms with less financing constraints experience significantly higher cash flow sensitivity than firms with higher constraints. This means that cash flow sensitivity is not a good proxy to measure financing constraints. With the results obtained by Kaplan and Zingales, Lamont, Polk and Saá-Requejo (2001) created and index which is called the KZ-index. With this index, they found that financially constraints affect firm value. In addition, financial constraint firms experience lower returns, and that these returns are not more cyclical than unconstraint firms. However, Hardlock and Pierce (2010) cast doubt on the validity of this index. Hardlock and Pierce conducted an estimation with an ordered logit model to evaluate the KZ index and found that cash holdings have a positive and significant coefficient in the models predicting the constraints. This is consistent with the precautionary motive, and means that financial constraint firms should hold higher cash levels.
One way to decrease the sensitivity in expected cash flows (and therefore the amount of precautionary cash holding) is the use of financial derivatives. Financial derivatives provide the opportunity to ensure cash flows which are expected in the future by hedging the exposure. This exposure results from foreign exchange rates, change in commodity prices, and interest rate risks. With efficient risk management, firms can create a situation in which they receive consistent cash flows, and therefore be able to participate in all investment opportunities (Stulz, 1996). However, the extent to which firms create value with hedging is ambiguous. Not all firms engage in hedging, and not all firms engage in fully hedging their exposures. It is hard to measure the effect on cash flow sensitivity and this is therefore excluded from this analysis.
The third motive is the tax motive. Using a sample of US firms from 1982 to 2004, Foley et al. (2007) found that firms incur increased tax costs when repatriating earnings generate higher levels of cash. Additional tests comparing cash holdings of domestic affiliates with comparable foreign affiliates show that affiliates in lower tax countries have higher cash holdings and that affiliates organized as branches abroad, in which host country tax rates do not vary, hold lower cash levels. This finding suggests that the cash levels of multinational (and larger) firms are more likely to accumulate relatively to firms operating solely domestic.
10
The last motive, the agency motive, is based on the free cash flow theory, also known as the agency theory developed by Jensen (1986). Jensen showed that entrenched managers prefer to hold increased levels of cash, rather than payout the cash to shareholders, when firms encounter a decreasing amount of positive investment opportunities. Dittmar et al. (2003) estimated this effect by using a sample from 45 different countries, and found that firms in countries with an increased amount agency problems hold more cash, while controlling for the transaction and precautionary motives. These results provide support that corporate governance and information asymmetry is an important determinant in corporate cash holdings. Corporate governance is an extension of the agency motive.
In addition to the four motives, the pecking-order theory is also related to corporate cash holding of firms (Myers and Majluf, 1984). The pecking-order theory plays a role in the decision-making process of firms, and states that firms prefer to finance positive investment opportunities with internally generated cash because it is the cheapest form of financing, and sends a good signal to the public. If the firm is unable to do so, they favor issuing debt, which is cheaper than equity. At last, firms issue equity to raise capital because adverse selection costs make equity the most expensive option. The pecking-order theory is another reason for increased levels of cash holding, and is line with the transaction motive.
B. Brexit – implications and scenarios
Since the establishment of the European Union in 1993, and the introduction of the monetary system in 19992, no country has ever left the union. On the 23th of January 2013 during his Bloomberg speech,
David Cameron promised a referendum to vote about the future of the UK: should the UK pursue staying together with the EU or leave. On the 24th of June 2016, the outcome of the referendum was announced.
51.9 percent of the voting residents chose for the ‘Brexit’, which means that 48.1 percent voted against3.
The referendum turn-out was measured at 71.8 percent, which means that more than 30 million people voted. This result was a surprising outcome, concluded by the market reaction, the expectation at forehand and the percentage of victory (51.9 percent).
There were several direct implications on the economy of the UK. Figure 1 shows the value of the British Pound relatively to the US dollar during 2016. A sharp spike is seen on the day of the announcement, 24th of June 2016. The pound depreciated over 9 percent relatively to the US dollar after
the announcement was made public. This spike gives an intuition in how surprised the investment market was after the result was announced.
2 The ‘monetary union’ is established in 1999, and fully operational in 2002. It is composed from 19 EU states, which use the euro as main currency.
3 England voted for leaving the EU by 53.4 percent to 46.6 percent. Wales voted also for leaving the EU: 52.5 percent against 47.5 percent. Scotland and Northern Ireland both wanted to stay, and voted respectively 38 percent to 62 percent and 44.2 percent to 55.8 percent.
11
[INSERT FIGURE 1 HERE]
Besides the effect on the exchange rate, the inflation has increased from the announcement date. As we can see in Figure 2, the CHIP (Consumer Prices Index including owner occupiers’ Housing costs) rose excessively comparing to 2015. Most of the recent inflation is due to increasing prices of food, tobacco, alcohol, clothing, footwear, miscellaneous goods, and services4. Prices are partly offset by
decrease in prices from transport and to a lesser extent motor fuels. The Centre for Economic Performance (part of the London School of Economics) announce in their forecasting Brexit paper (Breinlich et al, 2016), that most of the price increase would result from transport, alcohol, food and clothing. This seems partly in line in what the UK government found.
[INSERT FIGURE 2 HERE]
In addition to the inflation and exchange rates, the FTSE100 and FTSE250 are showed in Figure 3. On the vertical axis, the index level is shown, while the horizontal axis defines the date. The vertical line corresponds to the date one day before the Brexit announcement, 23 June 2016. As the figure shows, the Brexit announcement results in a sharp drop on the value of both indexes. However, after the announcement both the indexes recover and sustain at a higher price level than before the announcement.
[INSERT FIGURE 3 HERE]
Besides the direct implications of the Brexit, several researchers forecast the long-term consequences of the Brexit. In 2009, the procedures for leaving the EU were formally introduced by the Lisbon Treaty. When a country wishes to leave the EU, it should notify the other members of the EU and engage in a series of negotiations over a withdrawal agreement. When no agreement is reached, it is possible for the country to leave after two years. From Dhingra and Sampson (2016), several scenarios were elaborated which the UK could follow after exiting the EU: the Norwegian model, the Swiss model, re-joining the EFTA, and the WTO model. The Norwegian model refers to Norway, which is not part of the EU, but is part of the Single Market. Norway must comply with the Single Market rules implemented by the EU. However, being part of the EU does not oblige countries to be part of the monetary union, security policies, EUs justice and home affairs policies. The price which Norway must pay is by paying a fee and by contributing to several development funds and costs the EU makes. The second scenario is the Swiss model. Switzerland is neither part of the EU nor the Single Market. This scenario provides Switzerland the flexibility to choose between the different EU initiatives it wishes to participate. Some drawbacks are no guarantees of market access which results in more uncertainty. It is most likely that this model would result in increased economic segregation, and that the UK would no longer have any voting rights regarding the EU decision-making. Thirdly, re-joining in the European
4 For more information on how this prediction is determined consult, the UK government website: https://www.ons.gov.uk/economy/inflationandpriceindices/timeseries/l55o/mm23
12
Free Trade Association (EFTA) is mentioned. The EFTA is a free trade agreement which covers the trade of all export and import products, except for agricultural goods. Re-joining would guarantee trading tariff-free to the EU and ensure that the UK will not impose any tariffs when other EU countries import from the UK. Costs associated with the Brexit on the economy of the UK would come primarily from tariff barriers, which would be the economic price the UK must pay for joining the EFTA (Ottaviano et al., 2014). The last scenario, leaving the EU without any agreements, is the WTO5 scenario.
The UK would be eligible to set its own import-tariffs under the WTO rules. However, this will certainly reduce the access to the EU market for UK producers and service providers. When considering this overview of possible scenarios, no scenario seems perfect when leaving the EU. As it is still uncertain which type of Brexit the UK will follow (either ‘soft’, like the Norwegian or Swiss model, or ‘hard’ like the WTO scenario), the future seems uncertain and therefore riskier after the ‘victory’ of the Brexit-voters.
There is an ongoing debate about the costs and benefits for the UK when leaving the EU. Dhingra et al. (2016) are the first assessing the costs and benefits of the Brexit. With a general equilibrium trade model, which covers 31 sectors, they defined distinct scenarios which could happen after the Brexit. Furthermore, they measured the changes in welfare by real consumption. The findings showed that the welfare changes from -1.28 percent to -2.61 percent, depending if the Brexit occurs in respectively an optimistic or pessimistic scenario. In the pessimistic scenario (‘hard Brexit’) it is assumed that the UK will leave the single market and will trade under the regulations of the WTO. In the optimistic scenario (‘soft Brexit’) the UK will negotiate a deal (such as Norway, Iceland or Switzerland) and no import tariffs are set between the EU and UK. In addition to the equilibrium trade model, Dhingra et al (2016) used the reduced form approach and found even greater welfare losses. This model predicted losses of welfare between -6.3 percent and -9.5 percent. However, many factors are not included in both models such as immigration, increases in R&D, and changes in productivity. In addition to research performed after the referendum was announced, some predictions were made before this announcement. Bruno et al. (2016) conducted an econometric analysis of the foreign direct investment (FDI). Using FDI data from 34 countries from the period 1985 – 2013, Bruno et al. found that members of the EU have an increased amount of FDI inflow by 14 percent to 38 percent relatively to countries outside the EU. When the UK leaves the EU, they predict a decrease of FDI of approximately 22 percent.
The question arises now why the UK residents still voted in favor of the Brexit while these predictions were so negative. The proponents of the Brexit are concerned about the current state of the country in relation with the EU. Most arguments are concerning the politics, about the immigration, and they believe that the bureaucracy of the EU is costly and complex. Proponents of the Brexit also argue that the UK could benefit economically by removing the import-tariffs to lower the cost of imported
5 The WTO (World Trade Organization) is a forum for trade negotiations, including over 162 countries. The WTO sets a framework for trade between these countries regarding import tariffs and commitments.
13
goods (Dhingra et al., 2016). Altogether, it is concluded that the Brexit will result in an increase in uncertainty for everyone. The question is now how the firms which are operating in the UK will be affected by decisions made by the residents.
C. Hypothesis development
The basis for liquidity preferences comes from the theory of risk-avoiding behavior introduced by Tobin (1958). The changes in risk for investors are affected by market actions and are influenced by the monetary and fiscal policy of governments. Changes in interest rates and tax rates are both influencing the capital gains and losses, interest earnings and therefore the expected returns of investments. This is in line with the precautionary motive for firms, in which firms accumulate excessive cash holding to prevent future inability in financing its activities.
From the Brexit analysis in part B, Figure 1, Figure 2, and Figure 3, it is stated that the future of the UK is uncertain and that the market participants are experiencing this uncertainty also. Uncertainty means that firms do not know what to expect in the future and that it is hard to predict what the outcomes of such an event are. This implicates that for instance the incoming cash flows in the future are harder to predict. Therefore, this uncertainty does have implications for firms. According to the precautionary motive for cash holding, one of this implication is that firms should increase their cash levels to mitigate these increased amounts of risk. Uncertainty in cash flows can result in firms not having enough liquidity to maintain their operating business. Following this reasoning the first hypothesis is formulated as: Hypothesis 1: The announcement of UK leaving the EU is positively associated with corporate cash holding.
In this paper, the research is extended by performing additional tests while controlling for financial constraint firms. The second hypothesis follows the theory of Kaplan and Zingales (1997), Whited and Wu (2006), and Hadlock and Pierce (2010), which all construct an index to measure in which extent a firm is constraint and find that firms with increased financial constraints experience an increased variability in stock returns. This means that firms with increased financial constraints should increase their cash levels even more, relatively to firms which are less financial constraint. This is consistent with the view of the precautionary motive. The second hypothesis is therefore formulated as:
Hypothesis 2: The announcement of UK leaving the EU has a larger positively effect associated with corporate cash holding for firms experiencing financial constraints.
The main focus lies on the precautionary motive in relation with the Brexit. However, the Brexit gives the ideal opportunity to analyze other determinants of corporate cash holding. As stated above, privately held firms are more sensitive to cash flow shocks and therefore rely on precautionary cash holdings more often (Brav, 2009). In addition, the capital structure of private firms significantly differs
14
from their public counterparts. The third hypothesis will test if this difference between private and public firms is still consistent. From this line of reasoning and evidence, it is expected that private firms have higher cash holdings than public firms. The announcement of the Brexit referendum can have different effects on different type of companies. Therefore, the announcement of the Brexit referendum is used to determine if the effect on private and listed firms differs with respect to corporate cash holding. This results in the development of the third hypothesis:
Hypothesis 3: The announcement of UK leaving the EU has a larger positively effect associated with corporate cash holding for private firms.
In addition to the precautionary motive, the effect of the agency motive will be investigated with respect to the Brexit. As stated above, Dittmar et al. (2003) found that agency problems are important when analyzing firms’ corporate cash holding. Firms are holding more cash when access to new funds is easier. The fourth hypothesis will test the difference between firms with low agency frictions and high agency frictions. The uncertainty for firms about the future of the UK after the Brexit could have a different effect on firms with high agency problems relative to firms with low agency problems. This line of reasoning results in the fourth and last hypothesis:
Hypothesis 4: The announcement of UK leaving the EU has a larger positively effect associated with corporate cash holding for firms with greater agency frictions.
III. Methodology
Part III describes the methodology used to test the hypotheses developed above and to give an answer on the research question. The Brexit referendum announcement gives an ideal setting to test the hypotheses. In Section A, a static and dynamic cross-sectional time-series model with annual and quarterly data will be conducted. In Section B, a differences-in-differences model (Diff-in-diff) will be used to check if the results are consistent.
A. Cross-sectional time-series model
The data sample consists of data from annual and quarterly financial statements, respectively 10-K and 10-Q statements. To be able to test the hypotheses, the calendar year data is used instead of fiscal year data since fiscal year data could have different time ranges. The first approach is conducting a cross-sectional time-series model. By using annual financial statements over the period of 2014 to 2016 and quarterly data over the period of Q1-2015 to Q4-2016, hypotheses one, three and four are investigated. Unfortunately, the quarterly sample is smaller than expected on forehand. This is due the change in regulation of publishing interim accounting statements for firms in the UK. This rule6 is
6For more information consult the website of the Financial Reporting Council: https://www.frc.org.uk/Our-Work/Publications/Accounting-and-Reporting-Policy/FRS-104-Interim-Financial-Reporting.aspx
15
introduced in January 2015 and states that “FRS 104 does not require any entity to prepare an interim report”. Therefore, this has a negative result on our sample size.
This model will follow the methodology used by Guney, Ozkan and Ozkan (2003). In their analysis both a static as a dynamic model are used to confirm the validity of the results. The second model is dynamic because a first lag of the cash ratio is included to the regression equation. Including a first lag of the cash ratio will control for the persistence between the years since firms cannot instantaneously change their cash level follow a shock (Han and Qiu, 2007). The static model (Equation 1) will follow an adjusted form, however closely related to Guney, Ozkan and Ozkan’s (2003) model:
𝐶𝐴𝑆𝐻 𝑅𝐴𝑇𝐼𝑂 𝑖𝑡 = 𝛽 1𝑃𝑂𝑆𝑇𝐵𝑅𝐸𝑋𝐼𝑇𝑖𝑡+ 𝛽 2𝑀𝐴𝑅𝐾𝐸𝑇𝑇𝑂𝐵𝑂𝑂𝐾𝑖𝑡+ 𝛽 3𝑆𝐼𝑍𝐸𝑖𝑡+
𝛽 4𝐶𝐴𝑆𝐻𝐹𝐿𝑂𝑊𝑖𝑡+ 𝛽 5𝐿𝐸𝑉𝐸𝑅𝐴𝐺𝐸𝑖𝑡+ 𝛽 6𝑁𝑊𝐶𝑖𝑡+ 𝛽 7𝐵𝐴𝑁𝐾𝐷𝐸𝐵𝑇𝑖𝑡+ 𝛽 8𝑅&𝐷𝑖𝑡+ 𝛽 9𝐶𝐴𝑃𝐸𝑋𝑖𝑡+ 𝛽 10𝐴𝐶𝑄𝑈𝐼𝑆𝐼𝑇𝐼𝑂𝑁𝑆𝑖𝑡+ 𝛽 11𝑆𝐼𝐺𝑀𝐴𝑖𝑡 + 𝛽 12𝐷𝐼𝑉𝐼𝐷𝐸𝑁𝐷𝑖𝑡+ 𝛼𝑖𝑛𝑑𝑢𝑠𝑡𝑟𝑦+ 𝑢𝑖𝑡 (1) And where the dynamic model (Equation 2) also follows an adjusted form of Guney, Ozkan and Ozkan’s (2003) equation and is therefore defined as:
𝐶𝐴𝑆𝐻 𝑅𝐴𝑇𝐼𝑂 𝑖𝑡 = 𝛽 1𝑃𝑂𝑆𝑇𝐵𝑅𝐸𝑋𝐼𝑇𝑖𝑡+ 𝛽2 𝐶𝐴𝑆𝐻 𝑅𝐴𝑇𝐼𝑂𝑖𝑡−1+ 𝛽 3𝑀𝐴𝑅𝐾𝐸𝑇𝑇𝑂𝐵𝑂𝑂𝐾𝑖𝑡+ 𝛽 4𝑆𝐼𝑍𝐸𝑖𝑡 + 𝛽 5𝐶𝐴𝑆𝐻𝐹𝐿𝑂𝑊𝑖𝑡+ 𝛽 6𝐿𝐸𝑉𝐸𝑅𝐴𝐺𝐸𝑖𝑡+ 𝛽 7𝑁𝑊𝐶𝑖𝑡+ 𝛽 8𝐵𝐴𝑁𝐾𝐷𝐸𝐵𝑇𝑖𝑡 + 𝛽 9𝑅&𝐷𝑖𝑡+ 𝛽 10𝐶𝐴𝑃𝐸𝑋𝑖𝑡+ 𝛽 11𝐴𝐶𝑄𝑈𝐼𝑆𝐼𝑇𝐼𝑂𝑁𝑆𝑖𝑡+ 𝛽 12𝑆𝐼𝐺𝑀𝐴𝑖𝑡+ 𝛽 13𝐷𝐼𝑉𝐼𝐷𝐸𝑁𝐷𝑖𝑡+ 𝛼𝑖𝑛𝑑𝑢𝑠𝑡𝑟𝑦+ 𝑢𝑖𝑡 (2)
Cash ratio is defined as cash and equivalents divided by total assets. Variable 𝛼𝑖𝑛𝑑𝑢𝑠𝑡𝑟𝑦 controls for industry fixed effects. The remaining variables are explained in the data section. As stated above, the dynamic model includes a lagged value of the cash ratio to control for the inability of firms to instantaneously adjust their cash levels after a shock. All time-series regressions are performed with the cluster option. The clustered standard errors will account for within-cluster correlation and heteroskedasticity problems.
The variable of interest in both equations is the POSTBREXIT dummy for all regressions. This binary variable is equal to one if the period is equal to 2016 and zero if otherwise. For quarterly data, POSTBREXIT is equal to one if the period is equal to Q3-2016 or Q4-2016 and zero if otherwise. A positive coefficient in this variable means an increase of the cash ratio relatively to the previous periods, and a negative coefficient a decrease of the cash ratio relatively to previous periods. A positive coefficient is therefore in line with the hypotheses.
To be able to give an answer to the additional hypotheses, the methodology described above will be used with several adjustments. The adjustments are implemented by dividing the sample in subsamples to prove or disprove the hypotheses. To provide evidence for hypothesis three, the sample will be divided between publicly and privately held companies. If a company is listed on any exchange in the world, it
16
will be regarded as a public company. As stated above, Brav (2009) showed that private firms hold higher amounts of cash levels than public firms.
Secondly, a subsample will be created by dividing the sample between firms with low agency frictions and high agency frictions. The determinants used to divide these sample are analyst coverage and the credit rating of a company. Information asymmetries decrease the easiness for firms to obtain external funds. External lenders discount securities and bonds because they know less about the firm and the firm’s management. Therefore, firms which are assigned credit ratings and have analysts following the doings of a company are believed to be more transparent (Jensen, 1986). The result of this should be maintaining decreased amount of cash holdings due to the easiness of obtaining funds (Opler et al., 1999). The sample is therefore divided by no analyst coverage (zero analyst following a specific firm), moderate coverage (1 to 5 analysts follow a firm), and intensive coverage (6 or more analyst following a firm). In addition, the sample is divided by firms with and without a credit rating. Firms with credit ratings are believed to be monitored by the external lenders (especially banks). This will also decrease possible information asymmetry problems.
B. Differences-in-Differences model
The second approach will be conducting a differences-in-differences (Diff-in-Diff) model with both the quarterly and annual data sample. Since the Brexit result of the referendum was announced on the 24th of June 2016, the sample period ranges from January 2015 to December 2016 for quarterly data and
from 2014 to 2016 for annual data sample. As stated above, the quarterly data sample is smaller than expected due to the new accounting rule implemented in 2015.
Policy analyses mostly rely on the Diff-in-Diff model. Diff-in-Diff models have become a popular way to analyze causal relationships after exogenous events have taken place (Bertrand, Duflo and Mullainathan, 2003). The Diff-in-Diff model compares differences in outcomes before and after the treatment and control groups are affected by this event. Thus, a Diff-in-Diff estimator observes outcomes for two different groups in two time periods. The first group, the so-called “treatment group”, is affected by the event in the second period. The second group, the “control group”, should not be affected at all in either periods. Therefore, to be able to use a Diff-in-Diff model, an exogenous event should have taken place, also called a natural experiment, which affects only one group. A natural experiment is an event where decisions of firms are effected for exogenous reasons. This model is designed to estimate the difference for the firms affected or not affected by this event. The event used in this model is the announcement of the outcome of the referendum on 24th of June 2016 which is believed to affect certain
firms more than others. To be able to conduct this Diff-in-Diff model a control and treatment group is needed. The Brexit does most likely not affect the control group, or significantly less than the treatment group.
17
The advantage of the Diff-in-Diff model comparing to other models is that a Diff-in-Diff model resolves possible endogeneity problems and biases when the event used is as good as random, conditional on the time and firms fixed effects. However, it is hard to address the true exogeneity or randomness of an event occurred. Therefore, from Section II it is concluded that the announcement of the UK leaving the EU is considered as a random, exogenous event. This conclusion is drawn from the market reactions, the number of voters against and advocating this decision, and the expectation from researchers and newspapers of UK leaving the EU.
The Diff-in-Diff model uses a normally distributed OLS regression with a dummy variable identifying whether a firm belongs in the treatment or control group. However, in certain models the treatment variable is a continuous variable (Acemoglu et al., 2004). This approach will also be used in this analysis. The Diff-in-Diff model follows the equation:
𝐶𝐴𝑆𝐻 𝑅𝐴𝑇𝐼𝑂𝑖𝑡 = 𝛽0+ 𝛽1 𝑇𝐼𝑀𝐸𝑖𝑡+ 𝛽 2𝑇𝑅𝐸𝐴𝑇𝑀𝐸𝑁𝑇𝑖𝑡+
𝛽3 (𝑇𝐼𝑀𝐸𝑖𝑡∗ 𝑇𝑅𝐸𝐴𝑇𝑀𝐸𝑁𝑇𝑖𝑡) + 𝛽 4𝑋𝑖𝑡+ 𝜇𝑖𝑡 (3)
Where TIMEit is one if its equal to a period after the announcement of 24th of June 2016. TREATMENTit is the treatment variable, and can be either a continuous or a dummy variable. Xit specifies all control variables which are used in the time-series regression also. The variable of interest is 𝛽3 (𝑇𝐼𝑀𝐸𝑖𝑡∗ 𝑇𝑅𝐸𝐴𝑇𝑀𝐸𝑁𝑇𝑖𝑡), which is also known as the differences-in-differences coefficient. A positive value of this coefficient means an increased amount of cash ratio for the treatment group relatively to the control group after the event period. A negative value implies vice versa. When the treatment group is based on a continuous variable, the variable will be interpreted different. It means that when the intensity of the treatment variable increases, the cash ratio will also increase.
To report consistent results, two different treatment groups (with three different treatment variables) are created in which the treatment and control groups are defined and on which basis:
(1) Industry risk specified treatment group. Bates et al. (2009) found that firms in high-risk industries tend to hold more cash. Industries with an expected higher risk level could therefore be more affected than other industries. The extent in which firms are riskier is measured by the standard deviation per industry. The three digit SIC-code is used to determine the standard deviation in each industry classification. This standard deviation is used as a continuous treatment variable, which is higher (and more positive) for riskier industries.
(2) Financially constraint specified control group. This control group is separated between two different financial constraint measures: dividend payout and the Hadlock and Pierce (2010) financial constraint index (HP). The dividend payout divides the sample between dividend payout and firms which did not payout dividend in a certain period. Fazarri et al. (1988) studied the effect of financial constraints on
18
corporate investment policies of firms. They found that firms which are paying less dividend tend to be more financial constraint than their non-dividend paying counterparts. The reason behind this is that firms want to reduce the need of raising external funds in the future by keeping increased amounts of cash (instead of entitling it to the shareholders). The second financially constraint measure is the HP-index. Hadlock and Pierce (2010) found new evidence on the financial constraint measure literature. With an ordered logit model, they foound that only the variables leverage and cash flow consistently predict the extent in which a firm is financial constraint. However, they do not recommend using these variables due to their endogenous nature. Therefore, the HP index only uses the variables size (natural logarithm of total assets) and age (time since listing on any exchange by firm). The HP index is a formula developed by using regression output as evidence and is formulated as:
𝐻𝑃 𝑖𝑛𝑑𝑒𝑥 = (−0.737 ∗ 𝑆𝐼𝑍𝐸𝑖𝑡) + (0.043 ∗ 𝑆𝐼𝑍𝐸2𝑖𝑡) − (0.040 ∗ 𝐴𝐺𝐸𝑖𝑡)
Where 𝑆𝐼𝑍𝐸𝑖𝑡 is the natural logarithm of total book assets and 𝐴𝐺𝐸𝑖𝑡 is the number of years since listing on any exchange. 𝑆𝐼𝑍𝐸𝑖𝑡 is winsorized at the total of 4.5 billion dollar and 𝐴𝐺𝐸𝑖𝑡 is winsorized at 37 years. The variable age is gathered from CAPITALIQ and has the variable description of ‘date of first trading day’. These two variables have relatively exogenous characteristics and are therefore most reliable to use. This measure is used in favor to the KZ index constructed by Lamont, Polk and Saá-Requejo (2001). Hadlock and Pierce cast doubt on the KZ index7 because the findings show that three
out of five variables are either insignificant or show a different sign than expected and is therefore unreliable. In addition, the endogenous nature of these variables make the variables even more unreliable.
IV. Sample formation and descriptive statistics
Section A will explain how the data sample is formatted. In Section B, the dependent and independent variables are constructed and elaborated. Additionally, the need of including the variables in this analysis is explained. In Section C, the descriptive statistics are presented, elaborated, and compared with related studies.
A. Sample formation
An initial sample of firms is retrieved from the CAPITALIQ database, because the CAPITALIQ database provides quarterly data for UK firms. Data on market prices is gathered from DATASTREAM. First, all firms which are in primary operating in the United Kingdom are selected. To get to the final sample, additional filters are included. All firms which have not reported a financial statement after
7 The Kaplan-Zingales index is formulated as: 𝐾𝑍 = −1.001909 ∗ 𝐶𝑎𝑠ℎ 𝐹𝑙𝑜𝑤𝐶𝑎𝑠ℎ 𝐹𝑙𝑜𝑤
𝑃𝑃&𝐸𝑡−1 + 0.2826 ∗ 𝑇𝑜𝑏𝑖𝑛 ′𝑠 𝑄 + 3.1391 ∗ 𝐷𝑒𝑏𝑡 𝑇𝑜𝑡𝑎𝑙 𝐶𝑎𝑝𝑖𝑡𝑎𝑙 − 39.3678 ∗ 𝐷𝑖𝑣𝑖𝑑𝑒𝑛𝑑𝑠 𝑃𝑃&𝐸𝑡−1 - 1.3148 * 𝐶𝑎𝑠ℎ
𝑃𝑃&𝐸𝑡−1. Lamont, Polk and Saá-Requejo (2001) used the results of Kaplan-Zingales to construct this measure
19
January 2017 are excluded. This means for both annual as quarterly data that either a 10-K or 10-Q statement should be published after 2017. Secondly, financial firms (firms with Standard Industrial Classification (SIC) codes ranging from 6000 – 6999) are excluded from the sample, because financial firms use cash to meet capital requirements. Additionally, these firms are trading in marketable securities which are cash equivalents. Furthermore, firms operating in the utility industry (SIC codes ranging from 4900 – 4999) are excluded because utility firms are subject under regulatory supervision and therefore need to maintain certain cash levels. In addition, firms with no or non-positive sales, cash and equivalents or total assets are excluded. Firms with an incomplete amount of observations are also excluded.
B. Construction of variables
To construct the variables used in the empirical tests, the papers of Ozkan and Ozkan (2004) and Opler et al. (1999) are followed as a guideline. Opler et al. investigate the determinants and implications regarding corporate cash holdings in the US over the period of 1971 – 1994. Ozkan and Ozkan (2004) conducted an empirical analysis on a sample of UK firms over the period of 1995 – 1999. Many factors are influencing the cash holding of firms. Therefore, the following control variables are used to control for firm and financial characteristics. Additionally, to be able to differentiate the control and treatment group, financial constraint measures are created and have been explained in the methodology section.
The dependent variable which defines corporate cash holding is pronounced as the Cash ratio, and is defined as the total cash and equivalents divided by total assets. This variable is line with the dependent variable used by Bates et al. (2009) and Ozkan and Ozkan (2004). The first control variable is to measure the likelihood of a firm to have the possibility to engage in positive net present value projects: Market-to-Book ratio. The market-to-book ratio is constructed by dividing the market value of total assets by its book value of total assets. The intuition behind it is that the market value reflects the expected future growth options of a firm in the market price, and the book value does not. Therefore, it is expected that a higher market value, relative to the book value, is higher for firms with greater growth possibilities. The second control variable is Size, which is measured as the natural logarithm of total assets. The rationale of the variable size is that larger firms tend to hold relatively lower levels of cash because larger firms are less likely to be constrained or to experience financial distress (Ozkan and Ozkan, 2004). Cash Flow is the levered operating cash flow divided by the book value of total assets, where levered operating cash flow is net earnings after interest, dividends, and taxes. Leverage is defined as long-term debt plus short-term debt divided by the book value of assets, also known as the debt-to-assets ratio. Like the size variable, leverage is thought to influence the ability of firms to raise funds. Net working capital is measured by dividing the net working capital with total assets, where net working capital is current assets minus current liabilities. Firms can sell non-core assets in riskier periods to be able to increase the liquidity of the firm. Bank debt is measured as total bank debt divided by total debt.
20
A higher portion of bank debt to total debt decreases the ability to raise additional bank debt or credit lines. R&D is a ratio of research and development expenses to sales. This variable is a measure for potential financial distress costs. Capital expenditures is measured by the amount of capital expenditures divided by total assets. Acquisitions is the amount of cash acquisitions divided by total assets. The variables R&D, Capital expenditures and Acquisitions control for the ability of a firm to pay with cash for non-operating projects. Firm sigma is the standard deviation of the cash flows generated in all periods. Dividend is a dummy variable which equals one in the periods the firm has a dividend payout, and zero otherwise. All continuous variables will be winsorized at a 1.0 percent level, except for Size while this variable is expressed in natural logarithm.
As the yearly data includes both public as private firms, a variable Listed is included which equals one if the firm in question is operating in private, and zero if otherwise. Brav (2009) showed that private firms hold different amounts of cash levels than public firms. This data will be used to test hypothesis three. Unfortunately, the quarterly data cannot be used while this dataset contains not enough information for private firms.
As stated above, Dittmar et al. (2003) found that corporate governance and information asymmetry are important determinants of firms’ corporate cash holdings. In a separate analysis, the difference in effect of agency problems before and after the Brexit are investigated. To give an answer on hypothesis four, the analyst coverage and credit rating data will be used to distinguish firms from being transparent or not. Jensen (1986) stated that analyst play a role in monitoring firms by generating information for the public and therefore taking care of firms being opaque. In addition, Harford and Uysal (2014) stated that firms with a credit rating are more transparent with respect to information.
C. Descriptive statistics
Table 1 shows the overview of the cash ratio in each corresponding period. Panel A shows the cash ratio for annual published data from 2014 to 2015, while Panel B shows the cash ratio for quarterly statements from Q1-2015 to Q4-2016. The dependent variable Cash ratio is defined as cash and equivalents divided by total assets. Cash ratio is winsorized at the 1.0 percent level. From Panel A it is shown that the cash ratio declines from 0.1615 to 0.1549 after the announcement of the Brexit referendum. This contrasts with Hypothesis 1 stated above. The announcement of the outcome of the referendum took place on 24th of June 2016. In the hypothesis, it is expected that the cash ratio should
increase after this announcement has taken place. The number of firms in this sample is 649, with a total of 1947 observations.
21
Panel A: Annual data
Variable Period Mean Standard 25th Median 75th N
Deviation Percentile Percentile
Cash ratio
2014 0.1686 0.1968 0.0436 0.0924 0.2102 649 2015 0.1615 0.1853 0.0432 0.0975 0.2047 649 2016 0.1549 0.1733 0.0443 0.0908 0.2014 649
Panel B: Quarterly data
Cash ratio Q1-2015 0.1396 0.1718 0.0370 0.0780 0.1702 520 Q2-2015 0.1391 0.1688 0.0385 0.0784 0.1710 520 Q3-2015 0.1436 0.1705 0.0393 0.0832 0.1741 520 Q4-2015 0.1435 0.1721 0.0399 0.0860 0.1624 520 Q1-2016 0.1342 0.1580 0.0411 0.0777 0.1499 520 Q2-2016 0.1328 0.1575 0.0404 0.0755 0.1498 520 Q3-2016 0.1370 0.1600 0.0390 0.0825 0.1638 520 Q4-2016 0.1373 0.1595 0.0398 0.0850 0.1624 520
Panel B presents the overview of the cash ratio from Q1-2015 to Q4-2016 in quarterly periods. The number of firms with quarterly data is 520 with a total of 4160 observations. From these statistics, it is shown that from Q1-2016 to Q4-2016 the mean cash ratio has increased from 0.1342 to 13.73, which means an increase of 2.31 percent. This increase is in line with the hypothesis formulated above, which presumes an increase of the cash ratio after the announcement. This should result in either an increase in 2016 relative to 2015 and 2014 or from Q3 and Q4 of 2016 relative to the quarters of 2015 and Q1 and Q2 of 2016. However, these results cannot be interpreted as sound as stated here because no control variables are used and therefore no causality can be derived from these tables. For this reason, the analysis is extended with an in-depth regression analysis.
Table 2 shows the descriptive statistics of the full sample of both quarterly as annual reported data. The construction of the variables is elaborated in section B. All continuous variables are winsorized at the 1.0 percent level. Panel A presents the descriptive statistics of annual data, while Panel B presents the descriptive of quarterly data. The dependent variable is Cash ratio, while all other variables are used as control variables. Exceptions are the variables Listed, Credit rating and Analyst coverage which are used to separate the sample in subsamples. This is conducted to test the hypotheses three and four. All continuous variables are winsorized at the 1.0 percent level.
Panel A describes the annual Cash ratio of 649 firms over the period of 2014 - 2016. Panel B describes the quarterly cash ratio for 520 firms over the period of Q1-2015 – Q4-2016. Cash ratio is defined as the total cash and equivalents divided by total assets and winsorized at the 1.0% level. Data is gathered from CAPITALIQ database.
22
From this table, it is shown that firms hold an average cash ratio of 0.162 during 2014 to 2016, with annual reported data. The quarterly data shows an average cash ratio of 0.138 for 2015 and 2016. A bank debt of over 1.00 can be explained by credit lines which are (over)used. From the annual data, it is shown that 94.5 percent of the firms are listed, 12.9 percent have a credit rating and that the average analyst coverage is 6.247 analysts. In further notice no odd observations are shown relatively to Ozkan and Ozkan (2004).
Opler et al. (1999) used a sample of US based public firms in their cash holding analysis. The dependent variable is cash and equivalents over assets minus cash and equivalents, which is different from the dependent variable used in this analysis. From the descriptive statistics it is shown that, with a total sample of 87,117 firms years from 1971 - 1994, that the mean and median cash ratio are 0.170 and 0.065 respectively. When comparing these numbers in percentages, it is found that Opler et al. had a mean which is approximately 4.7 percent lower with respect to the cash ratio measured from the annual sample used in this analysis. However, it is hard to compare it to each other. It can be concluded that it differs a lot, the cash ratio of the sample of Opler et al. seemed to be higher, since deducting cash and equivalents from the denominator will increase the cash ratio significantly.
While Opler et al. used a sample of US firms, Ozkan and Ozkan (2004) conducted an analysis on characteristics of UK firms over the period 1995 - 1999. The total sample size is gathered from 839 firms. The results should be more in line with the descriptive statistics used in this analysis. The dependent variable is cash and equivalents over total assets, which means that the cash ratio of Ozkan and Ozkan can be used to compare easily. Ozkan and Ozkan’s sample presented a mean and median of respectively 0.099 and 0.059. A comparison with the annual descriptive statistics presented below, show that the mean of Ozkan and Ozkan is approximately 39 percent lower. The median is approximately 58 percent lower in their analysis with respect to these descriptive statistics.
23
Panel A: Annual data Quantiles
Variable Mean Standard Min 25th Median 75th Max N
Deviation Percentile Percentile
Cash ratio 0.162 0.185 0.001 0.043 0.093 0.206 0.873 1947
Market-to-book ratio 1.441 3.106 0.002 0.380 0.723 1.352 27.791 1623
Size 5.303 2.612 -4.075 3.434 5.051 7.190 14.304 1947
Cash flow -0.022 0.208 -1.043 -0.067 0.021 0.071 0.526 1884
Leverage 0.204 0.266 0.000 0.001 0.136 0.290 1.677 1938
Net working capital 0.022 0.190 -0.690 -0.064 0.006 0.104 0.680 1929
Bank debt 0.360 0.439 0.000 0.000 0.000 0.947 1.159 1947 R&D 0.336 1.880 0.000 0.000 0.000 0.035 16.311 1785 Capital expenditures 0.037 0.050 0.000 0.006 0.019 0.048 0.270 1947 Acquisitions -0.021 0.055 -0.317 -0.011 0.000 0.000 0.017 1947 Firm sigma 0.086 0.113 0.000 0.021 0.050 0.098 1.054 1944 Dividend payout 0.507 0.500 0.000 0.000 1.000 1.000 1.000 1947 Listed 0.945 0.229 0.000 1.000 1.000 1.000 1.000 1947 Credit rating 0.129 0.336 0.000 0.000 0.000 0.000 1.000 1947 Analyst coverage 6.247 8.399 0.000 1.000 2.000 8.000 42.000 1947
Panel B: Quarterly data
Cash ratio 0.138 0.165 0.001 0.039 0.080 0.162 0.851 4160
Market-to-book ratio 1.279 2.319 0.001 0.376 0.705 1.290 18.673 3560
Size 6.015 2.584 -2.273 4.333 6.165 7.769 14.229 4160
Cash flow 0.000 0.059 -0.284 -0.013 0.008 0.022 0.183 3948
Leverage 0.199 0.196 0.000 0.022 0.162 0.302 0.966 4160
Net working capital 0.019 0.170 -0.474 -0.064 0.007 0.097 0.630 4152
Bank debt 0.248 0.393 0.000 0.000 0.000 0.460 1.098 4160 R&D 0.091 0.452 0.000 0.000 0.000 0.005 3.970 4160 Capital expenditures 0.009 0.012 -0.000 0.002 0.005 0.011 0.069 4160 Acquisitions -0.006 0.020 -0.137 -0.001 0.000 0.000 0.002 4160 Firm sigma 0.056 0.210 0.002 0.015 0.024 0.045 3.731 4136 Dividend payout 0.544 0.498 0.000 0.000 1.000 1.000 1.000 4160 Listed 0.998 0.044 0.000 1.000 1.000 1.000 1.000 4160 Credit rating 0.173 0.378 0.000 0.000 0.000 0.000 1.000 4160 Analyst coverage 8.050 9.125 0.000 1.000 4.000 12.000 42.000 4160
Panel A describes annual data from 2014 to 2016. The number of firms is 649 with a corresponding number of periods of 1947. Panel B describes the quarterly data from Q1-2015 to Q4-2016 with a total of 520 firms and 4160 firm quarters. Cash ratio is defined as cash and equivalents divided by total assets. Market-to-book ratio is defined as the market value of assets divided by the book value of assets. Size is the natural logarithm of total assets. Cash flow is defined as the operating free cash flow divided by total assets. Leverage is defined as long-term debt plus short term debt divided by total assets. Net working capital is defined as net working capital divided by total assets. Bank debt is the ratio of bank debt to total debt. R&D is defined as research and development expenses divided by the total revenue. Capital expenditures and Acquisitions are defined as respectively capital expenditures and acquisitions divided by total assets. Firm sigma is the standard deviation (volatility) of cash flows in prior years/quarters. Dividend payout is a dummy variable, which equals one if a firm paid dividend in a specific quarter, and zero otherwise. Listed equals one if the firms are listed on any exchange in the world, and zero if the firm is privately held. Credit rating is one if a firm has been rated during this sample, and zero otherwise. Analyst coverage is the number of analysts which cover a firm. All continuous variables are winsorized at the 1.0% level. Data is gathered from CAPITALIQ and Datastream.
24 V. Results
Part V presents the results of this paper. Section A shows the results of the cross-sectional time-series regressions and elaborates how the variables of interest are interpreted. Section B presents the results from the differences-in-differences regressions with the various treatment variables. Section C will extend this analysis with two additional robustness checks to critical assess the results. The methodology will be extended with an additional financial constraint measure and will compare the cash ratio of the United Kingdom to Germany with relation to the outcome of the Brexit announcement.
A. Results from cross-sectional time-series model
In Table 3 and 4 the results from the first time-series model are presented. Table 3 presents the results of regressions with annual data, while Table 4 includes the quarterly data sample. As mentioned in the methodology section, two different approaches are used: a static model (regressions one to four) and a dynamic model (regressions five to eight). The dependent variable is Cash ratio, and used in all succeeding regressions performed during this paper. The dynamic model includes the first lag of the cash ratio to control for persistence between the years. Regressions two, four, six and eight include industry fixed effects. All regressions are clustered at firm level. In addition, to rule out unexpected impacts from outliers, all continuous variables are winsorized at the 1st and 99th percentile.
The coefficient of interest is the POSTBREXIT variable in either the annual and quarterly data sample. This variable takes the value of one if the period equals respectively 2016 or Q3 and Q4 in 2016, and zero if otherwise. By including this dummy variable, the effect of the announcement of the Brexit can be analyzed with ease. A positive coefficient means that the cash ratio has increased respectively to the periods before. Recall that Hypothesis 1 stated that the announcement of UK leaving the EU is positively associated with increased amount of cash holdings. Hypotheses 3 and 4 stated that this effect is greater for private firms and for firms with more information asymmetry frictions.
From Table 3 it is shown that the coefficient of POSTBREXIT is negative and significant at the 5 percent level in regressions one to four and seven to eight. Like stated earlier, the meaning of this observation is that the cash ratio declined from 2014 and 2015 to 2016. The decline in cash ratio varies statistically significant between 0.00725 and 0.0163. The mean cash ratio is 0.162, which results in a decline of cash ratio of approximately 4.5 percent and 10.0 percent over time. Since all regressions pronounces the same effect, it seems that the announcement of the Brexit referendum did not result in increased amount of cash holding. Therefore, these regressions present evidence contradicting the precautionary motive for cash holdings and Hypothesis 1 is rejected when following these results. However, additional tests should verify this conclusion. In Table 4, which includes quarterly data, it is also shown that the coefficient of interest is POSTBREXIT is negative in regressions one to four and varies between -0.00169 and -0.00557. However, the dynamic regressions five to eight show a positive
25
value between 0.000651 and 0.00300. No obvious result can be derived from these results since all regressions are insignificant.
Because the control variables are used from Opler et al. (1999) and Ozkan and Ozkan (2004), it is possible to compare their predicted signs with the results shown here. If these coefficients are in line with each other it can be concluded that the probability of data errors is low. Ozkan and Ozkan (2004) predicted that the variables Liquidity, Leverage, Bank debt, and Size should show a negative sign, which means that these variables have a negative impact on the cash ratio when they are increased and vice versa. The variables Cash flow, Market-to-book, and Variability are predicted to show a positive sign. The sign of Dividend payout is ambiguous. Note that the variables Liquidity and Variability are pronounced at Net working capital and Firm sigma in this analysis respectively. By comparing Table 3 and 4 it is shown that all these variables are in line with the predictions made by Ozkan and Ozkan. However, some variables are included which are covered by Opler et al. From Opler et al., the prediction is made that the R&D, Capital expenditures and Acquisitions variables should have a positive impact on the cash ratio. This is due to the reason that firms with higher cash levels invest more in assets and in taking over other companies. Table 3 and 4 shows that R&D and Acquisitions are indeed in line with this reasoning, while Capital expenditures is predicted in contrast.
