CASH HOLDINGS AND COUNTRY
CORPORATE GOVERNANCE
Master’s thesis IFM Supervised by: Peter Smid JEL-classifications: G32; G34; G38
Key words: Cash Holdings, Country Governance, Ownership structure
Abstract
This paper investigates the impact of country corporate governance on corporate cash holdings for relatively small firms in 59 countries from 2011 to 2016, using pooled OLS. We find no consistent evidence that the level of country governance affects corporate cash holdings. Our sample seems to be affected by a small number of countries and has endogeniety problems when we use 2SLS. The findings are not robust for different country samples or other definitions of cash holdings. We also find no impact of country creditor rights, concentrated ownership or shareholder rights on cash holdings.
JANUARY 11, 2019
UNIVERSITY OF GRONINGEN Jeroen Rozema
1
1. Introduction
Cash management should be high on the priority list for managers of relatively small firms. In our sample of 59 countries, cash holdings averaged 14.2% of total assets. Previous evidence suggest that companies of smaller size face more difficulties and costs to obtain finance and have more information asymmetry problems than companies of large size. Therefore, they keep higher cash levels. There are also cross-country differences that cause managers to keep different cash levels. Whether these differences lead to different cash holdings for smaller size is not widely investigated. Therefore, the goal of this paper is to investigate whether the corporate cash holdings of relatively small listed firms are influenced by the level of country governance.
Large listed firms face more agency problems than companies of smaller size. This difference is possibly driven by a separation of ownership and control. For smaller firms, ownership and control is often not separated. If a large shareholder is also the manager, it becomes easier to extract value from debt holders and minority shareholders. There are only a few studies that directly investigate the relationship between cash holdings and governance for smaller firms (see e.g. Belghitar and Khan, 2013; Al-Najjar 2015). These studies investigate single countries. This paper contributes to the literature by using a large cross-country dataset that investigates the relationship between corporate cash holdings and country level governance for listed firms up to $ 600 million in total sales and $ 500 million in total assets. If these firms are also affected by the level of country governance than this implies that managers of smaller companies consider the macro environment of countries into account before going there.
2 opacity, which leads to financial asymmetry problems. Furthermore, smaller firms often have coincided ownership and management, making issues between shareholders and managers possibly non-existent. It return, it likely makes issues associated with debt more severe. This might imply that strong creditor rights have a different impact on cash holdings of smaller firms. To explore whether this holds, this paper will investigate the association between country level governance and cash holdings for listed firms of relatively small size. The central question of this paper is: to what extent are corporate cash holdings of smaller listed firms affected by the level of country governance? Is there an association between shareholder rights and corporate cash holdings when ownership and control are concentrated for smaller firms? Are cash holdings of smaller firms lowest when both the level of country governance and creditor rights are strong?
We investigate the effects of country level governance on the level of corporate cash holdings for 59 countries over the 2011 – 2016 period using both pooled OLS and random effect regressions. We find some evidence that companies of small size are also affected by the level of country governance. However, our results seem to be driven by a small group of countries and likely suffers from endogeniety. We do find different results for shareholder rights. We argue that the new methodology of WB leads to different results for shareholder rights than using the Anti-director index.
This paper is organized as follows. The relevant literature is reviewed in section two and ends with the hypotheses. Section three lists the data and methodology. The initial results are presented in section four. Section five presents the robustness tests and possible other explanations for the main findings. The conclusion follows in section six.
1. Literature review
1.1 Determinants of corporate cash holdings
3 we refer to Opler et al. (1999). The focus of our literature review lies on the relationship between country governance and cash holdings and the possible impact of different company size (in terms of total assets or sales) and cash holdings.
A. Trade-off theory and corporate cash holdings
According to the TOT, firm’s cash holdings should be set at such a level that the marginal benefit of cash holdings equals its marginal costs. Trade-off theory implies that cash holdings can be explained by either precautionary motives or transaction cost motives. The transaction motive implies that a firm will hold more liquid assets, when either the costs of raising outside funds; and / or the risk of cash shortfalls, are higher. The second motive of TOT is the precautionary motive, which is based on the impact of information asymmetry on getting capital. When the costs of acquiring external capital are high, firms will likely hold more cash. It may also reduce the likelihood of entering into financial problems. Opler et al. (1999) find that the firms who have the greatest access to the capital market and are most creditworthy hold the least amount of cash. This firms are large in terms of size and have economies of scale. Fazzari and Peterson (1993) argue that small firms are more likely to be affected by financial constraints due to limited internal finance options. They also state that external finance is more costly for small firms, since there are a lot of fixed costs for outside financing. Therefore, it is likely that smaller firms find it more difficult to obtain outside financing, and will hold relatively higher levels of cash than large firms.
4 are higher. It can also reduce dividends to increase cash levels. Therefore, dividend is negatively related to cash holdings.
Small firms face more information asymmetry problems between equity holders and debt holders. This is another reason why smaller firms will likely hold relatively more cash. Dittmar et al. (2003) describe that research and development expenses (R&D) can be used to measure asymmetric information between equity holders and debt holders. R&D is typically used in companies that are less transparent and have relatively more intangible assets. In contrast to operating assets, relatively more intangible assets over total assets implies less borrowing power and hence, higher cash levels. It is also arguable that more information asymmetry implies less cash holdings. Garcia et al. (2008) argue that firms who have a relatively high amount of short-term bank debt relative to total debt oblige themselves to negotiate the refinancing of their funds regularly, which implies that they are monitored more strongly. In doing so, valuable information about the quality of the company is disclosed. This can both improve the conditions and availability of financing, since it leads to less credit rationing in the bank credit market. They describe that having this source of external finance reduces information asymmetry between borrowers and lenders. Therefore, a negative relationship is expected between short-term bank debt and corporate cash holdings.
B. Pecking order theory and corporate cash holdings
5 companies with low cash flows. The variables to test this view are equal to TOT, but the sign is the opposite. POT predicts a negative relationship between investments and cash levels (R&D and CAPEX). Garcia (2008) argues that POT is important for small firms. A positive effect of cash flow on cash holdings supports the idea that in the presence of information asymmetries firms prefer to finance themselves from internally generated resources, since smaller firms face more information asymmetries regarded with debt. Hence, POT might explain cash holdings for smaller firms.
C. Agency theory and corporate cash holdings
Managers and shareholders view the costs and benefits of holding cash differently. As described by Dittmar (2003), there are two types of costs for holding cash. Under the assumption that managers maximize the value for their shareholders, the only costs of cash is the cost-of-carry. Carry costs can be described as the difference between cash return and the opportunity interest that would have been paid to finance an additional unit of cash. If the assumption of wealth maximization is relaxed, it is possible to argue that the cost-of-carry increases, since it implies that managers have opportunities to spend the cash on activities that destroy value.
Both the TOT and POT are based on the assumption that executives always act in the best interest of shareholders, implying that there are no agency problems between them. However, as stated by Jensen (1986), listed firms often face significant agency costs and the costs increase with size. Managers who are not guided by shareholder maximization are more inclined to divert cash to themselves, as explained by studies such as Opler et al. (1999) and Ozkan and Ozkan (2004). Harford et al. (2008) find, for their sample of US firms, that weakly controlled managers prefer to spend cash quickly on acquisitions and capital expenditures over accumulating it.
6 In contrast, agency issues between shareholders and debtholders might be more severe for companies of small size. As explained by Cheung (2016), agency issues between these parties are caused by a conflict of interest associated with debt. When firms have higher leverage, it becomes both more difficult and expensive to obtain additional financing (Opler et al. 1999). As described by Myers (1977), this can lead to underinvestment problems, since raising more capital contributes relatively more to debtholders than for shareholders. Therefore, shareholders prefer not to invest, even though the firm has valuable projects. Garcia et al. (2008) argue in their study of small Spanish firms, that small firms with high growth opportunities have the most severe agency conflicts associated with debt. Furthermore, debt levels are likely to be negatively related to cash holdings as the costs involved in liquid assets increase with leverage. They also argue that asymmetry and agency problems can be reduced by establishing relationships with banks.
The extent of the agency conflicts might also be mitigated by either internal or external governance measures. Good country governance is assumed to decrease corporate cash holdings by reducing agency costs (Seifert and Gonenc, 2016).
2.2
Empirical evidence
7 with low protection lack enforcement power to decrease excess cash levels held by executives. Therefore, it is arguable that shareholder protection is another measure to implicitly measure country governance. Guney et al. (2003) find a similar result with respect to shareholder protection and corporate cash holdings using a sample of large listed firms in Japan, Germany, United States and France. Furthermore, they argue that there is a negative relation between ownership and cash holdings. When insiders hold more shares, they have more enforcement power on managers to reduce cash holdings. Another cross-country study, performed by Kalcheva and Lins (2007), find some evidence that managers who are in control hold more cash and that this relation is stronger when country-level shareholder protection is weak. Although they did not study multiple countries, Harford et al. (2008) provide evidence of the contrary. They find, for their sample of US firms, a country with high shareholder protection, that companies with weak corporate governance structures have lower cash levels. Their study also divided their listed US sample into 5 size quantiles. They find that small firms have more inside ownership, less institutional ownership, less remuneration, lower board independence and smaller boards. They find that the smallest size quantile holds 8% more cash than the largest quantile.
8 state that, since the UK has high investor protection, shareholders might have sufficient legal protection to inhibit management from pursuing self-interests. Therefore, publicly traded firms of smaller size in the UK might hold large cash reserves and not necessarily incur severe agency costs. Saeed et al. (2014) provide support for this argument for a sample of Pakistan firms. Their study investigated the relationship between corporate cash holdings and political connections. They state that Pakistan has a low governance score, driven by a high level of political corruption. Their main finding is that political connected small firms face severe agency problems and therefore hold large cash reserves to pursue the political objectives. This might be driven by a lack of legal protection. Umrani et al. ( 2015) describe that in Malaysia, concentrated owners can more easily exploit minority shareholders in small firms since there country has a weak legal system. They state that more concentrated ownership results in less protection for minority shareholders. A study of Nicolov and White (2014) separated ownership into institutional, management and block holder ownership. They find significant relationships with respect to corporate cash holdings, but with different signs. They find that when block holder and institutional ownership is low, managers are monitored less, resulting in more private benefits and a higher level of cash. When mangers hold more shares, cash holdings decrease. Therefore, they argue that companies with low block and institutional ownership suffer more from agency issues. They find that agency costs matters less for companies of small size. This might be due to the coexistence of ownership and management within small firms. Above evidence suggests that the measurement of ownership matters a great deal.
9 country governance is strong. Within these circumstances, companies need less cash for future investments, because firms expect that funds will be available, both today and in the future. A similar finding is presented by Bae and Goyal (2009). They state that the enforcement of creditor rights is more important in explaining loan maturities, amounts and spreads than the creditor rights itself. They argue that a strong application of creditor laws leads to a higher recovery rate and therefore to more available funds. This suggest that creditor rights is positively related to corporate governance and therefore should also lead to lower cash levels for companies.
As explained by Garcia et al (2008), cash levels of small firms in Spain are lower when these firms have relationships with banks. This argument suggests that stronger creditor rights would also mean less cash holdings for relatively smaller firms when country governance is good. Seifert and Gonenc (2016) argue that the impact of strong credit rights is bigger for private credit agreements than for public bond markets, since private agreements are renegotiated after a firm violates the contract.
The evidence of a relationship between country governance and the level of cash holdings for listed firms of smaller size is limited. Most evidence is based on single country studies. This implies that there is limited evidence whether country governance has an impact on cash holdings of smaller size. Therefore, our paper uses a sample of listed firms up to $ 500 million in terms of total sales and $ 430 in terms of total assets and investigate whether country governance reduces the cash holdings of these firms.
1.2 Hypotheses
10 Second, we expect that cash holdings are lower in countries that have strong shareholder protection, with a different effect for companies with concentrated ownership. When concentrated ownership is high and shareholder rights is low, it becomes more easy to extract value from minority shareholders. There are also possibly less agency problems with managers and shareholders when there is concentrated ownership, since a shareholder can also be the manager. All in all, the findings seem to indicate that there might only be a relationship between shareholder rights and cash holdings when there is no concentrated ownership. Therefore, our second hypothesis is:
Hypothesis 2: The corporate cash holdings of relatively small listed firms in countries in countries with good shareholder rights are only lower than poor shareholder rights countries when corporate ownership is not concentrated.
Thirdly, based on insights gained so far, we all-in all expect that cash holdings are lower when both country governance and creditor rights are strong. Therefore, our third hypothesis is:
Hypothesis 3: Corporate cash holdings of listed large companies are lower when both country governance and creditor rights are strong
3.
Data and methodology
3.1
Model
3.1.1 Cash holdings basic model
11 parentheses [ ]. We use pooled cross country regressions with standard errors clustered at the company level. For the precise definitions of the variables we refer to table 1.
𝑪𝑯 = 𝜶𝒊𝒕+ 𝜷1𝑮𝑶𝑽𝑱𝒕+ 𝜷𝟐𝑺𝑹𝑱𝒕+ 𝜷𝟑𝑪𝑹𝑱𝒕 + 𝜷4𝑮𝑫𝑷𝒋𝒕 + 𝜷𝟓𝑪𝑪𝑴𝒋𝒕+ 𝜷6𝑪𝑺𝑀𝒋𝒕+ 𝜷7𝑳𝒔𝒊𝒛𝒆𝒊𝒕+ 𝜷8𝑳𝑬𝑽𝒊𝒕 + 𝜷𝟗𝑩𝑨𝑵𝑲𝑫𝒊𝒕+ 𝜷𝟗𝑫𝒅𝒊𝒕+ 𝜷𝟏𝟎𝑪𝑨𝑷𝑬𝑿𝒊𝒕 + 𝜷𝟏𝟏𝑾𝑪𝒊𝒕 + 𝜷𝟏𝟐𝑹𝑫𝒊𝒕+ 𝜷𝟏𝟑𝑺𝑮𝒊𝒕+ 𝜷𝟏𝟒𝑴𝑻𝑩𝒊𝒕 + 𝜷𝟏𝟓𝑶𝑾𝑵𝒊𝒕+ 𝜷𝟏𝟔𝑪𝑭𝒊𝒕+ 𝜷𝟏𝟕𝑺𝑻𝑫𝑬𝑽𝒊𝒕 +
𝜷18𝑶𝑾𝑵𝒊𝒕∗ 𝑺𝑹 + 𝑪𝑹 ∗ 𝑮𝑶𝑽𝑱𝒕 + ∑𝒄𝒀𝑬𝑨𝑹𝒋+ ∑𝑲𝒋+ ∑𝑪𝑶𝑼𝑵𝑻𝑹𝒀𝒋+ 𝜺𝒊
(1)
Within this model, the country level governance variables for country j at year t are, governance scores (𝑮𝑶𝑽𝑱𝒕) as defined by Kaufmann (2009), shareholder rights (𝑺𝑹𝑱𝒕) originally developed by La Porta et
al. (1999) and revised by Djankov et al. (2008), creditor rights (𝑪𝑹𝑱𝒕) originally developed by Djankov
et al. (2007), Gross Domestic Product growth (𝑮𝑫𝑷𝒋𝒕), Country private Credit Market over GDP
(𝑪𝑪𝑴𝒋𝒕) and Stock Market capitalization over GDP (𝑪𝑺𝑴𝒋𝒕), both from the World Bank (Beck and
Demiguc-Kunt 2009). Following Seifert and Gonenc (2016), we control for both the country debt market (CCM) and equity market (CSM) since on a conceptual level, a company’s cash holdings can be influenced by both the rights (in terms of creditors or shareholders) and the size (availability of financing) of these markets. A GDP dummy is incorporated to control for country GDP growth. Our firm level variables are denoted by company i and year t. Our dependent variable is cash holdings (𝑪𝑯) which is measured as total cash and short term investments over total assets per year end. The following controls are also measured per year end. We use book [market] leverage (𝑳𝑬𝑽𝒊𝒕), a bank
dummy (𝑩𝑨𝑵𝑲𝑫𝒊𝒕) denoting 1 when a company has bank debt and 0 otherwise, net working capital
minus cash over net [total] assets (𝑾𝑪), an ownership dummy (𝑶𝑾𝑵𝒊𝒕) denoting 1 when a single
shareholder holds at least 50% of the shares, sales growth (𝑺𝑮𝒊𝒕), monthly return volatility over total
market value [cash flow volatility over the last three years] (𝑺𝑻𝑫𝑬𝑽𝒊𝒕), and market to book ratio
𝑴𝑻𝑩𝒊𝒕. Our other control variables are flow variables measured over the years. These variables are
Cash Flow (𝑪𝑭𝒊𝒕), investments over total assets (𝑪𝑨𝑷𝑬𝑿𝒊𝒕), research and development expense over
12 otherwise. The 𝒄𝒀𝑬𝑨𝑹𝒋 variable indicates a set of year dummies, 𝑲𝒋 denotes a set of 10 industry
divisions based on SIC codes. We also use country dummies ∑𝑪𝑶𝑼𝑵𝑻𝑹𝒀𝒋 as an alternative for fixed
effects. We leave out the first group to avoid the dummy trap.
We measure size (𝑳𝒔𝒊𝒛𝒆𝒊𝒕) by taking the natural logarithm of total assets [sales], following Dittmar et
al. (2003). Since we want to investigate companies of smaller size, we require companies to have less than $500 million in terms of total assets and $600 million in terms of total sales. We acknowledge that this cutoff point is arbitrary, we could have used any number. However, this approach allows us to look more closely at smaller firms. These firms are mainly active in Asia.
A distinction is made between creditor rights (CR) and shareholder rights (SR), since we argue that this might have an different impact on corporate cash holdings of smaller firms. As explained by Seifert and Gonenc (2016), it can be argued that cash holdings are influenced by both the rights of financers and the size of the markets. Therefore, both markets are included as controls. We expect that firms of smaller size use relatively more debt financing compared to large companies, leading to a larger impact of creditor rights and credit markets. Seifert and Gonenc (2016) find that creditor rights did not fluctuate much over time in their research period of 1996-2003. If this is still the case for our research period, managers in strong creditor rights countries will have more confidence, both now and in the future, in their ability to obtain financing. This implies that less cash is required for future investments. In would also imply that we should find similar results.
13 access to external finance. Their study measured this relationship using short-term bank debt over total debt. Unfortunately, ORBIS seems to be unable to discriminate the maturity of bank debt: since the sum of total bank loans is provided. We come to this finding after comparing the bank debt according to ORBIS with data according to a number of different annual reports. We find that it is not consistent. Therefore, we will measure bank debt differently than Garcia et al. (2008), by applying a dummy variable denoting 1 when the company uses bank debt. We argue that having either long-term or short-term bank debt indicates better access to external finance and therefore reduces cash levels. Sales growth and MB are incorporated as controls for growth opportunities.
For our second and third hypothesis, most attention is given to the interaction between creditor rights and country governance (𝑪𝑹 ∗ 𝑮𝑶𝑽𝑱𝒕) and the interaction between ownership and shareholder rights
(𝑶𝑾𝑵𝒊𝒕∗ 𝑺𝑹). We argue that concentrated ownership might have an additional impact on the
relationship between cash holdings and shareholder rights, as concentrated owners can possibly attract more cash from minority shareholders when shareholder rights is low. With respect to country governance and creditor rights, Seifert and Gonenc (2016) showed that cash holdings are lowest when both creditor rights and country governance are strong. When firms have the ability to get financing in the future, they will hold less cash.
14 The variables in parentheses indicate that these variables are used for our alternative definition for cash holdings, which is measured as the natural logarithm of the ratio cash over sales. The data source in given in the column “Source” and the expected sign for our regressions is given in the last column.
Table 1
Definitions of variables Source Expected sign
Dependent variable
CH Cash holdings ratio, used in two alternative ways
(cash and short term investments) / book value of total assets ORBIS
Alternative measure: natural logarithm of cash and short term investments over total sales ORBIS
Net assets (Total assets - cash and short term investments) ORBIS
Independent variables 1. Country level variables
CCM Country private Credit Market (Beck and Demirguc-Kunt, 2009) WB
-Total credit by of deposit money banks and other financial institutions over GDP, deflated by CPI
GOV Weighted average of the six World Bank Governance indicators (WGI) ranging from -2,5 tot +2.5 WB -1. Voice and Accountability
2. Political Stability and Absence of Violence/Terrorism 3. Government Effectiveness
4. Regulatory Quality 5. Rule of Law
6. Control of Corruption
SR Shareholder rights WB
-Used the conflict of interest regulation index which is a weighted average score ranging from 0 to 10 based on: 1. Disclose index
2. Ease of shareholder suits index 3. extent of director liability index
(COML) Alternative measure: common law dummy denoting 1 if common law country and zero otherwise
CR Creditor rights (Djankov et al. 2007) WB -
GDP World Development Indicators, variable Economic expansion years
-(dummy variable denoting 1 for years with positve country GDP growth and 0 otherwise)
CSM Country annual Stock Market capitalization WB
-Total country market capitalization of listed shares to GDP, deflated by CPI (Worldbank) 2. firm level variables
Lsize Natural logarithm of book value of total assets ORBIS
(Lsales) When alternative measure of CH is used, size is measured as natural logarithm of total sales
OWN Dummy denoting 1 when company has 1 shareholder holding > 50% of total shares, otherwise 0 ORBIS
LEV Leverage ORBIS
+/-The ratio of the book value of (Short-term Debt + Long-term Debt)/ book value of total assets ORBIS (LEV2) When alternative measure is used, we measure leverage over market value of total assets
BANKD Total bank loans ORBIS +
Dummy variable taking the value of 1 if company has a bank loan, otherwise 0
Dd Dividend dummy ORBIS
-Dividend dummy denoting 1 when the company paid cash dividends and 0 otherwise
CF Cash flow. Measured as net income plus depreciation over total assets ORBIS STDEV Standard deviation of the market returns relative to the market value of total assets ORBIS +
Monthly return volatility. Calculated as (STDEV * sqrt 12) * (market value of equity / market value of total assets) (CFV) When alternative measure of CH is used, risk is measured as cash flow volatility over the last three years
MTB Market to Book Ratio ORBIS +
(Total debt + market value of equity) devided by total Assets
CAPEX Capital expenditures (WC08416) WS
+/-The ratio of funds used to acquire fixed assets other than those associated with acquisitions over bookvalue of assets
WC Working Capital
-(The ratio of current assets ‐ Cash – current liabilities) over book value of net assets ORBIS (WC2) When alternative measure of CH is used, WC is measured over book value of total assets
RD Research and development expense over total sales ORBIS +
15
3.2
Data sources
Our research investigates the effect of country level governance on cash holdings for companies of small size for the period 2011-2016. The firm level accounting data are collected from Bureau van Dijk (ORBIS) and CAPEX comes from the World Scope (WS) of Thomson and Reuters. We measure CAPEX using the WS database, as the CAPEX variable in ORBIS has a lot of missing data.
Both utility and financial companies are excluded to avoid possible regulation issues and to improve the findings regarding leverage. All of the continuous firm variables are annually winsorized at 0.5th
and 99.5th percentile to reduce the impact of possible outliers. Data regarding all of the country
governance proxies come from the World Bank (WB), including creditor rights and shareholder rights. The main country governance variable GOV is measured as the weighted average score of six dimensions. Each dimension is ranked on a scale of minus 2.5 (bad governance) to plus 2.5 (good governance). The dimensions are explained by Kaufmann et al. (2009) and also used in the studies of Seifert and Gonenc (2016; 2018). The six dimensions are: “(1) voice and accountability, (2) political stability and absence of violence, (3) government effectiveness, (4) regulatory quality, (5) rule of law, and (6) control of corruption (Seifert and Gonenc, 2018; page 4). The WB incorporates annual country scores until 2016. Therefore, we do not include observations after 2016.
16 (Doing Business), we refer to Djankov directly. Creditor rights is an index on a scale of 1 to 12 based on a rating of the powers of lenders during bankruptcy. The pillars are as follows: “(1) whether there are restrictions, such as creditor consent, when a debtor files for reorganization; (2) whether secured creditors are able to seize their collateral after the petition for reorganization is approved, that is, whether there is no automatic stay or asset freeze imposed by the court; (3) whether secured creditors are paid first out of the proceeds of the liquidating bankrupt firm; and (4) whether an administrator, and not management, is responsible for running the business during reorganization” (Djankov et al. 2007; page 302).
17 prior to 2013. Therefore, we measure the shareholder rights as the weighted average of the disclosure, director liability and ease shareholder suits index, which have consistent data within our research period. Dittmar et al. (2003) showed that shareholder rights are stronger in common law countrie, possible because of better stock markets. Therefore, as a robustness check, we also use a common law dummy to measure shareholder rights.
Having a dataset of smaller firms reduces the possibility of having a lot of firm governance variables because of data availability issues. Nevertheless, we can investigate the link between ownership and cash holdings (𝑶𝑾𝑵𝒊𝒕) by using the Independence Indicator within ORBIS, which is a score between A
and D that reports the amount of direct ownership. When this indicator reports a score of D, it implies that the company has a single shareholder that holds more than 50% of direct ownership. We argue that when this is the case, there are likely no agency issues between shareholders and managers. Therefore, we measure ownership (𝑶𝑾𝑵𝒊𝒕) by using a dummy variable, denoting 1 when the
18
3.3 Representativeness sample
Our initial sample from ORBIS contained 59.702 firm-year observations for 9.873 individual firms over the period 2011-2016 for 106 countries. We used two years of cash flow data to calculate cash flow volatility, implying that we started with data in 2009. As in any research there are missing values. For example, for the variable bank loans, only 35.000 observations are available. Based on arguments by Garcia et al. (2008) bank loans might be of high importance for smaller firms. Therefore, we require that this data is available for our dataset. Another requirement is that country data is available, since we also required data from the WB. Due to the merge with the WB, we lost observations for 16 countries, or a total of 1098 firms. Main countries lost are Canada (483 firms) Cayman Islands (284), Bermuda (214) and New Zealand (38). The last requirement is that market data is available, since we investigate listed firms. We also dropped 19 countries with less than 3 firms to provide better estimates for our country effects regressions. After these adjustments, we end up with our final sample which has 5,373 individual firms in 59 countries with a total of 22,413 observations that have complete accounting and market data. A comparison between our initial dataset and final dataset is given in
19 of the final sample. Four of them are Asian countries (Taiwan, China, Japan and South Korea) countries and the other country is the United states. In terms of industries it appears to be more random. We lost most firms of the agriculture sector (73%) and the least amount in the Wholesale Trade sector (39%). The main differences are in appendix 2. Third, out of the 59 countries, 32 (55%) are developed and 27 are developing according to the International Monetary Fund's World Economic Outlook Database. However, the dataset is has more firms from developed countries, possibly due to better availability of market data. We report that 68% of the firms (3,796) are from developed countries and 32% (1,747) are from developing countries. Our initial sample contained 60% firms from developed countries. After these requirements and adjustments, we end up with a final sample of 25 individual observation from 5.373 different companies located in.
4. Results
4.1 Descriptive statistics
20 compare our SR variable with the original scores of the SR index of Djankov, which is an index of 0 to 5, we find some large differences. For example, Argentina has a SR of 8 out of 10 while the Djankov score is 2.5 out of 5. Opposite implies for Australia, that has an SR of 5 while the score of Djankov (4) would imply a score of 8. This either implies that the SR changes over time or that our SR variable is a
poor measure of shareholder rights.
Table 2
Cash holdings and country governance scores per country
Country Firms Mean Median SR
Common law SR Djankov CR CR Djankov Argentina 7 0.084 0.061 8.000 0.000 2.000 3.703 1.000 Australia 89 0.082 0.054 5.000 1.000 4.000 10.393 3.000 Austria 7 0.022 0.018 7.000 0.000 2.500 4.667 3.000 Bahrain 8 0.072 0.060 5.000 0.000 . 1.545 . Bangladesh 15 0.067 0.035 5.000 1.000 . 5.349 2.000 Belgium 18 0.096 0.051 5.000 0.000 3.000 4.455 2.000 Brazil 16 0.116 0.079 7.000 0.000 5.000 2.344 1.000 China 286 0.165 0.133 4.694 0.000 1.000 4.306 2.000 Croatia 11 0.042 0.026 8.000 0.000 2.500 5.565 3.000 Cyprus 7 0.093 0.030 6.000 1.000 . 7.571 . Denmark 27 0.065 0.038 6.000 0.000 4.000 8.339 3.000 Egypt 28 0.077 0.041 3.000 0.000 3.000 2.533 2.000 Estonia 9 0.114 0.075 8.000 0.000 . 7.000 . Finland 33 0.100 0.077 6.000 0.000 3.500 7.329 1.000 France 108 0.175 0.141 5.000 0.000 3.500 4.331 0.000 Germany 116 0.123 0.080 7.000 0.000 3.500 6.307 3.000 Ghana 6 0.068 0.063 6.000 1.000 5.000 7.429 1.000 Greece 56 0.071 0.030 7.000 0.000 2.000 3.436 1.000 Hong Kong 16 0.221 0.183 7.000 1.000 5.000 9.193 4.000 Hungary 3 0.050 0.027 6.000 0.000 2.000 8.875 1.000 India 377 0.054 0.026 8.692 1.000 5.000 6.550 2.000 Indonesia 126 0.070 0.045 5.000 0.000 4.000 4.662 2.000 Iran 5 0.064 0.035 3.000 0.000 . 3.125 2.000 Ireland 6 0.095 0.060 6.200 1.000 5.000 7.800 1.000 Israel 100 0.132 0.101 7.000 1.000 4.000 7.010 3.000 Italy 40 0.107 0.088 6.000 0.000 2.000 2.350 2.000 Japan 1,003 0.213 0.186 6.000 0.000 4.500 5.608 2.000 Jordan 31 0.031 0.013 2.000 0.000 1.000 0.673 1.000 Kenya 9 0.058 0.014 7.000 1.000 2.000 8.091 4.000 Kuwait 9 0.087 0.079 3.000 0.000 . 1.361 3.000 Latvia 10 0.020 0.010 7.000 0.000 4.000 9.300 3.000 Lithuania 17 0.043 0.021 6.000 0.000 4.000 5.652 2.000 Macedonia 8 0.041 0.035 7.000 0.000 . 6.000 3.000 Malaysia 6 0.107 0.089 8.000 1.000 5.000 8.000 3.000 Malta 3 0.110 0.122 7.000 0.000 . 2.000 . Netherlands 10 0.087 0.037 6.000 0.000 2.500 3.182 3.000 Nigeria 8 0.091 0.090 5.000 1.000 4.000 6.600 4.000 Norway 19 0.107 0.078 7.000 0.000 3.500 5.328 2.000 Oman 17 0.071 0.036 4.000 0.000 . 2.320 0.000 Pakistan 64 0.044 0.012 8.000 1.000 4.000 2.571 1.000 Philippines 32 0.082 0.062 0.000 0.000 4.000 1.941 1.000 Poland 73 0.068 0.049 6.000 0.000 2.000 7.631 1.000 Portugal 7 0.041 0.033 4.000 0.000 2.500 2.342 1.000 Saudi Arabia 22 0.054 0.034 4.000 0.000 . 2.767 3.000 Singapore 117 0.153 0.120 7.000 1.000 5.000 8.578 3.000 South Africa 12 0.091 0.067 8.000 1.000 5.000 5.000 3.000 Korea, Rep. 370 0.082 0.060 7.000 0.000 4.500 5.333 3.000 Spain 19 0.084 0.071 7.320 0.000 5.000 4.860 2.000 Sri Lanka 64 0.054 0.027 7.000 1.000 4.000 3.195 2.000 Sweden 74 0.103 0.064 7.000 0.000 3.500 6.297 1.000 Switzerland 30 0.183 0.133 8.000 0.000 3.000 6.696 1.000 Taiwan 600 0.178 0.152 7.000 0.000 3.000 4.141 2.000 Thailand 191 0.079 0.044 4.000 1.000 4.000 3.942 2.000 Tunisia 6 0.057 0.047 5.000 0.000 3.000 4.158 0.000 Turkey 79 0.068 0.039 8.000 0.000 3.000 2.833 2.000 United Emirates 6 0.099 0.059 4.000 0.000 . 3.714 2.000 United Kingdom 171 0.123 0.082 7.000 1.000 5.000 7.123 4.000 United States 588 0.185 0.107 4.000 1.000 3.000 9.658 1.000 Vietnam 178 0.093 0.054 7.000 0.000 . 5.921 1.000 Total 5,373 0.142 0.100 6.138 0.308 3.783 5.803 1.987 0.602 1.440 1.248 -0.478 0.801 1.768 1.029 -0.299 -0.227 -0.152 0.225 0.757 0.836 -0.242 1.776 -0.296 0.863 0.982 -0.364 1.559 1.689 -1.105 1.771 0.148 -1.063 0.740 0.868 -0.023 0.416 1.113 0.499 1.326 -0.096 -0.634 -0.116 -0.281 -0.327 -1.055 1.467 0.682 1.510 0.056 0.288 1.478 0.619 1.757 -0.897 1.143 1.823 1.152 1.337 -0.017 -0.511 0.429 1.011 -0.313 1.584 1.501 -0.069 -0.851 CH Governance Index mean (GOV)
21 Table 3 gives the descriptive statistics for our descriptive statistics for our main variables. We use a similar setup as Seifert and Gonenc (2016) did for creditor rights with the difference that we focus on governance scores. We use three different samples: (1) all countries, (2) high governance countries, and (3) low governance countries. We define a low country governance score as a score below the mean of our sample, which is 0.80. The high and low countries are compared using statistics. Panel A provides statistics for country level variables and Panel B gives results for the firm level variables.
This table reports the mean, median, standard deviation, skewness and kurtosis of variables used in equation 1. Panel A lists our country level variables while panel B gives the firm level variables. Definitions of the variables are given in Table 1. The starts ***, **, and * indicate statistical significance at 1%, 5%, and 10% levels, respectively. TA is total assets and TS is total sales.
Table 3
Descriptive statistics
Mean Median Stdev Skew Kurt Mean Median Mean Median
Panel A: Country level variables
GOV 0.494 0.526 0.865 -0.145 1.870 1.366 1.416 -0.122 *** -0.132 *** SR 5.950 6.000 1.720 -0.971 4.145 6.265 6.000 5.728 *** 6.000 COML 0.304 0.000 0.461 0.851 1.724 0.289 0.000 0.315 *** 0.000 CR 5.294 5.000 2.644 0.115 2.252 6.367 6.500 4.536 *** 4.000 *** CCM 0.912 0.777 0.539 0.685 2.871 1.283 1.252 0.650 *** 0.598 *** CSM 0.706 0.393 1.440 5.893 41.212 1.069 0.488 0.449 *** 0.330 *** GDP 0.893 1.000 0.310 -2.539 7.446 0.898 1.000 0.889 1.000 SR Djankov 3.505 3.500 1.124 -0.335 2.259 3.623 3.500 3.404 * 4.000 *** CR Djankov 2.005 2.000 1.038 0.076 2.268 2.048 2.000 1.978 *** 2.000
Mean Median Stdev Skew Kurt Mean Median Mean Median
Panel B: firm level variables
22 We choose to split our data based on governance scores since hypothesis 1 and 3 are connected. We expect that cash holdings are the lowest when both country governance and creditor rights are strong. When we compare our data, we do find significant higher creditor rights scores in good country governance countries, both in terms of mean (+1.8) and median (+2.5), as given in Panel A of table 3. Solely for comparison purposes, we also report the original scores of Djankov (2007) for CR dated back in 2003. We find that CR are also significantly significant in terms of mean (2.05 vs 1.98), but not in terms of median. This either suggest that creditor rights change over time or that the new data of the WB is different compared to the study of Djankov, since the index is was on a scale of 0 to 4 in 2003. For SR, in terms of mean, good governance countries have 0.5 higher shareholder rights on a score of 0 to 10. The median of SR is not different between high and low governance countries. Both have a median of 6. Interestingly, the median of the anti-director index1 of Djankov (2008), which are also
solely given for comparison purposes, is actually 0.5 (out of 5) higher in low country governance countries. We also find that our dummy variable for shareholder rights (COML) is significantly higher in terms of mean in low country governance countries, this might imply that the new method of WB is different, since both the anti-director index and COML indicate different results than SR and the study of Dittmar et al (2003) showed that common law can we used as an alternative for SR. Also note that roughly 60% of our country observations have governance scores below the mean, this implies that we have more countries with low governance scores. In contrast, the results in panel B in table 3 indicate that 3,428 firms (66%) from 26 countries are from high governance countries. This implies that we have much more observations from countries with good governance. The cash mean over net assets is 19.9% percent higher (32.0%) compared to low governance (11.1%) countries. The median is also 10.2% higher. Similar findings are obtained when we examine the medians using the Wilcoxon rank-sum test. All variables except LEV and SG have significant mean differences. We also find that
23 low country governance countries have more twice as much concentrated ownership (22.5% vs 11.8%).
The correlation matrix is given in table A1.2 of Appendix 1. We find no significant correlation higher than 0.50 used in the same regression, other than between GOV and CCM, which is 0.595. We also find that CR is positively correlated with GOV (0.32), while SR is negatively correlated to GOV (-0.05).
4.2 Basic regression results
Our results for our firm-level regressions is reported in Table 4. We use cash to total assets as our dependent variable, following Seifert and Gonenc (2016). Model [1] of Table 4 contains a regression model using GOV variable with year and industry dummies, based on 10 SIC divisions. We find that GOV is highly significant with a positive sign, which is in the opposite direction of what we expect. It remains positively significant in model [4] after controls. A possible explanation is that OLS assumes independence of observations, while there might be interdependencies between countries. As argued by Dittmar et al. (2003), controlling for industries alone is likely not sufficient to fully capture dispersion in corporate cash levels. Therefore, Table 5 will also use country dummies as an alternative for using fixed effects and show that this is possibly the case.
When we measure CR without controls in model [3], we find a positive and insignificant sign, which is the opposite what we expect. Model (9) measures shows that our CR becomes negatively significant after controls. Since there is a sign switch and since we did not control for shareholder rights and country effects, we do not interpret this finding.
24 significance, our result signals that a SR increase from 0 to 92 is associated with a decrease in cash
holdings of 1% (0.011 * 9) compared to the mean CH of 14.2%. This is a relative associated decrease of about 7% (0.010 / 0.142). Our result is 11% less compared to Dittmar et al. (2003) who used a similar model. Our finding suggests that smaller firms are also affected by the level of shareholder rights. When we interact SR with OWN, we do not find a significant result, which suggest that concentrated ownership does not alter this association. However, we note that OWN is not significant in all shareholder rights models. This would imply that concentrated ownership by itself does not alter cash holdings in this regression.
When we measure shareholder rights with our dummy variable for shareholder rights, COML, (model 7) we also find a negative relationship without interaction. The economic implications are stronger for COML than SR, similar to Dittmar et al. (2003). Common law countries hold 3.1% less cash in this regression, a decrease of 21.8% relative to the mean (0.031 / 0.142). In contrast to SR, when we interact COML with OWN in model [8], we find a positive effect. This result suggests that firms with concentrated ownership in common law countries are associated with 1.5% more cash relative to the mean compared to firms without concentrated ownership. One of the interpretations of this result is that concentrated owners in common law countries divert cash to themselves to spend these funds on private benefits, which have a negative impact on shareholder wealth. However, since minority shareholder rights in common law countries have better rights than countries with civil law, this is likely not the case. A more benign explanation is that there are possibly country specific differences for which we did not control yet.
We believe that there is heteroscedasticity present in our model. If this is the case, then that means that the error term is correlated with at least one of our explanatory variables, implying we have to adjust the model. In fact, a Breusch-Pagan tests rejects the null hypothesis that the error terms are independent. Therefore, we run a Housman test which rejects the null hypothesis that random effects
25 are consistent, implying that fixed effects is the appropriate model. However, if we use fixed effects, this would cancel out OWN and COML in our analysis because these variables do not vary over time. Therefore, we use country dummies as an alternative to capture heterogeneity between countries and then use random effects. For the sake of comparison, Table 5 also list a OLS model with year and company fixed effects [1]. We find a similar result regarding GOV in model [2].
This table reports our pooled cross-country regression. The dependent variable is cash and short-term investments over total assets. The definitions of all variables are given in Table 1. All of the regressions use both year and Industry dummies (based on SIC code divisions). Standard errors corrected for heteroscedasticity by using White’s correction and reported in brackets. ***, ** and * denote significance at 1%, 5% and 10% respectively. The sample period is from 2011 to 2016. The year 2011, the industry group agriculture and country Argentina are left out to prevent the dummy trap.
Model [1] of Table 5 uses company and year fixed effects and finds a significant negative association between the level of country governance and corporate cash holdings. Note that OWN drops out since this variable does not fluctuate over time. As an alternative, in model [2], we use country and industry dummies to also capture heterogeneity within countries and find a similar result. Regarding economic impact, we find that an increase of one standard deviation in GOV (0.865) is associated with a decrease in CH of 0.0329 (0.038 * 0.865) which implies a decrease of 23.1% of the mean value of CH (0.142). Our OLS regression in model [1] finds a similar result (coefficient of 0.036). Model [3] shows that SR remains significant when we use country dummies compared to Table 4. In terms of economic impact, we find
Table 4
Country-level governance and cash holdings using pooled OLS
[1] [2] [3] [4] [5] [6] [7] [8] [9]
Panel A: Country Governance
26 a much stronger effect. An increase in SR of 0 to 9 is associated with a decrease in CH of -64.8% in this case ((0.0102 * 9) / 0.142). In model [4], on the 10% percent level, we find that the interaction of OWN and SR is positively significant, which is different than model [6] of Table 4 where we did not find a significant result. We find that firms with concentrated ownership hold 3.1% less CH relative to the mean value (decrease of -21.8%). The interaction effect implies that if SR increases with one standard deviation (1.72), cash holdings of concentrated firms are associated with an increase of 0.0067 (0.0039 * 1.72) which represents an increase of 4.7% of the main value of CH. This finding is less strong when we consider the fact that it is only significant at the 10% level. In unreported results, we find no significant results when we use COML. This is likely due to the fact that common law captures an effect between country group while our model now allows for individual country effects.
Our results regarding creditor rights are presented in models [5] until [7] of Table 5. CR is positively significant in all models, which is in the opposite direction of what we expected. For hypothesis 3 we argued that cash holdings are lowest when both GOV and CR are strong. Our result in model [7] indicates the opposite3. Model implies that CH is lowest when country governance is good and creditor
rights is weak. Our result indicates that for each additional point in creditor rights, cash holdings increase with increase with 0.00085 (0.00085 * 11), meaning that cash holdings increase with 0.62% relative to the mean (0.142) in when country governance is good. This result could imply that creditors are not fully rational, since, as described by Seifert and Gonenc (2016), creditors would face lower recovery rates, higher costs, and more time claiming collateral in countries with low creditor rights. A more benign explanation lies in questioning our data, on which we come back in our robustness tests. Regarding our control variables, most variables have the appropriate sign, with a view exceptions. CCM, our variable that represents the ratio of private credit by deposit money banks and other institutions relative to GDP is negatively significant, suggesting that the credit market leads to a lower
3 In unreported results, we find a similar finding when we use the original scores of Djankov of 2003 with a
27 need for precautionary cash holdings. Our BANKD variable is also negatively significant which suggests that firms with bank debt have more confidence that they can easily get additional financing. We also find that more leverage implies lower cash levels, similar to Dittmar et al. (2003). Our cash flow (CF) variable is always significant with a positive sign, which could imply smaller firms with high cash flows have less confidence that they can easily get additional financing. Our RD variable indicates that firms with more RD likely have more information asymmetry and therefore hold larger cash levels. We find that companies with more working capital (WC) hold less cash, which would imply that WC is a substitute for CH for smaller firms. Our dividend dummy has the opposite sign, which suggests that dividend paying firms hold more cash. This is likely not the case, since paying dividend reduces cash levels.
This table reports our pooled cross-country regression using fixed effects (only model 1) and country dummies. Model [1] uses company and year fixed effects. All other models have year, industry and country dummies and are regressed using random effects. The definitions of all variables are given in Table 1. Standard errors corrected for heteroscedasticity by using White’s correction and reported in brackets. The stars ***, ** and * denote statistical significance at 1%, 5% and 10% respectively. The sample period is from 2011 to 2016. The year 2011, industry group agriculture and the country Argentina are left out to prevent the dummy trap.
Table 5
Pooled regression with industry, country and year dummies.
[1] [2] [3] [4] [5] [6] [7]
Panel A: Country Governance
28
5. Robustness checks
5.1 Alternative measure for cash holdings
We repeat the analysis in Table 6 for our country dummy regression by using an alternative measure for cash holdings used in the literature. For this, we scale cash and short term investments by total sales and take the natural logarithm, similar to Dittmar et al. (2003) and Harford et al. (2008). They argue that cash can also be seen as necessary investments for the working capital of firms. We take the log of this ratio to make it more normally distributed. We also use measure working capital over total assets, similar to Harford et al. (2008). In addition, we measure size as the natural logarithm of sales, cash flow volatility as an alternative for STDEV and market value for leverage.
This table reports our pooled cross-country regressions. All models have year, industry and country dummies and are regressed using random effects. Dependent variable is the natural logarithm of total sales. The definitions of all variables are given in Table 1. Standard errors corrected for heteroscedasticity by using White’s correction and reported in brackets. The stars ***, ** and * denote statistical significance at 1%, 5% and 10% respectively. The sample period is from 2011 to 2016. The year 2011, industry group agriculture and country Argentina are left out to prevent the dummy trap.
Table 6
Country-level governance and cash holdings using country dummies and alternative measure of cash holdings
[1] [2] [3] [4] [5] [6]
Panel A: Country Governance
29 Our results change dramatically. Hypotheses two and three are never significant. We only find that governance is significant at the 10 percent level in model [1]. Since we find only support for our first
hypothesis, and inconsistent support for our other two hypotheses, we focus further on the association between country governance and cash holdings.
5.2
Equally weighted regression results.
As shown in Table 2, a view countries have a large weight in our analysis. Therefore, for our first hypotheses, we follow the approach of Seifert and Gonenc (2016) and perform an analysis where each country has a single observation per year. Regarding firm variables, we use the mean of each variable per year, reducing our observations to the same amount reported in panel A of table 3. We only perform the analysis for creditor rights, shareholder rights and country governance and present the results in Table 7 . Model [1] until [3] report the results for CH and model [4] until [6] for our alternative
CH definition. We find no significant results.
This table reports our pooled country regression where each country has a single observation per year. All models have year and country dummies and are regressed using pooled OLS. Dependent variable is cash holdings over assets in Panel A and the natural logarithm of total sales in panel B. The definitions of all variables are given in Table 1. Standard errors corrected for heteroscedasticity by using White’s correction and reported in brackets. The stars ***, ** and * denote statistical significance at 1%, 5% and 10% respectively. The sample period is from 2011 to 2016.
Table 7
Country-level regressions
[1] [2] [3] [4] [5] [6]
Panel A: Cash holdings over total assets Panel B: Cash holdings over total sales
30
5.3 Alternative sample compositions
The lack of significance in Table 7 suggests that our result is driven by a few countries, since we have a lot of countries with little observations and vice versa, our results are likely driven by a small group of countries. Therefore, as an robustness check, we follow Dittmar et al. (2003) and Seifert and Gonenc (2018) and take out the countries that have a big impact on our analysis for both measures of cash holdings. Model [1] and [4] leaves out Japan, that takes a weight of 18% in our regressions. We find that GOV is no longer significant in our alternative measure of cash holdings for model [4] but remains significant for our primary cash holdings variable. Model [2] and [5] also leave out the US and Taiwan, implying that we take out 40% of all observations. We find a similar result, that GOV is only significant for panel A. Model [3] and [6] use the opposite approach and only use countries with more than 100 individual firms. We find that GOV is significant for both cash holding measures. These findings indicate that our findings regarding hypothesis one, that good governance leads to lower cash holdings are (at least) partly caused by a few countries.
This table reports our results using less countries. All models have year, industry and country dummies and are regressed using random effects. Dependent variable is the ratio of cash and short term investments over total assets for panel A and the natural logarithm of total sales for panel B. The definitions of all variables are given in Table 1. Standard errors corrected for heteroscedasticity by using White’s correction and reported in brackets. The stars ***, ** and * denote statistical significance at 1%, 5% and 10% respectively. The sample period is from 2011 to 2016. The year 2 011, industry group agriculture and country Argentina are left out to prevent the dummy trap.
Table 8
Alternative sample compositions in terms of countries
[1] Excl. Japan [2]Excl Jap, Taiw, US [3] countries > 100 firms [4] Excl. Japan [5]Excl Jap, Taiw, US [6] countries > 100 firms
Panel A: CH over total assets Panel B: CH over total sales
31
5.4 Jointly determination of cash levels and country governance
It is likely that a number of our variables are jointly determined. If this is the case, then we need to change our measurements. Harford et al. (2008) describe that cash holdings and governance is likely to be jointly determined and that there are little instruments to test this endogeniety. The main reason is that, as described by Seifert and Gonenc (2016), an instrument should be correlated with the replaced variable but not with the error term. Seifert and Gonenc (2016) describe for their study of creditor rights and cash holdings that there is (at least) potentially endogeniety between leverage and cash holdings. Therefore, they used an instrumental approach for leverage using two staged least squares regressions (2SLS). They used the approach suggested by Rajan and Zingales (1996) and used the annual mean leverage of US firms per industry as the instrument for non-US firms.
32 which means that leverage is not exogenous. Put differently, our result is biased and we can not take any conclusions from it. Possibly we used too many controls.
This table reports uses pooled OLS in models 1 and 3 and uses 2SLS in the other models. All models have year and country dummies. PREDLEV is the leverage that is used in the first stage regression. Residual is the predicted residual from the first stage regression of model [1]. Dependent variable is cash holdings over assets in Panel A and the natural logarithm of total sales in panel B. The definitions of all other variables are given in Table 1. Standard errors corrected for heteroscedasticity by using White’s correction and reported in brackets. The stars ***, ** and * denote statistical significance at 1%, 5% and 10% respectively. The sample period is from 2011 to 2016.
Table 9
Country-level governance and cash holdings using less controls and 2SLS
[1] [2] 2SLS [3] [4] 2SLS [5]
Panel A: Cash over assets Panel B: Cash over sales
33
Conclusion
This paper tried to explain differences in corporate cash holdings for smaller firms by investigating the role of country governance, shareholder rights and creditor rights. We hypothesize that cash holdings are lower when country governance is strong. In addition, we hypothesize that cash holdings are even lower when both creditor rights and creditor rights are strong, following the approach of Seifert and Gonenc (2016). We find no evidence for the latter, but our analysis likely suffers from endogeniety problems, as outlined in Table 9. Therefore, our estimates are probably biased and we cannot establish causality. This is an important caveat of this study.
We also hypothesized that cash holdings of relatively small listed firms in countries in countries with good shareholder rights are only lower than poor shareholder rights countries when corporate ownership is not concentrated, but find no consistent evidence. All of our hypotheses are not significant in our country level regressions, implying that our findings are driven by countries with an high amount of observations.
34
References
Al-Najjar, B. 2013. The financial determinants of corporate cash holdings: Evidence from some emerging markets. International business review, 22(1), 77-88.
Beck, Thorsten, Asli Demirgüç-Kunt, and Patrick Honohan. Access to financial services: Measurement, impact, and policies. The World Bank Research Observer 24(1) 2009: 119-145.
Belghitar, Y., and Khan, J. 2013. Governance mechanisms, investment opportunity set and SMEs cash holdings. Small Business Economics, 40(1), 59-72.
Berger, A. N., and Udell, G. F. 1995. Relationship lending and lines of credit in small firm finance. Journal of business, 351-381.
Chakra and Kaddoura. 2015. Doing Business 2015. Protecting minority investors, 1-7
Dittmar, A., Mahrt-Smith, J., and Servaes, H. 2003. International corporate governance and corporate cash holdings. Journal of Financial and Quantitative analysis, 38(1), 111-133.
Djankov, S., La Porta, R., Lopez-de-Silanes, F., and Shleifer, A. 2008. The law and economics of self-dealing. Journal of financial economics, 88(3), 430-465.
Djankov, S., McLiesh, C., and Shleifer, A. 2007. Private credit in 129 countries. Journal of financial Economics, 84(2), 299-329.
Drobetz, W., Grüninger, M. C., and Hirschvogl, S. 2010. Information asymmetry and the value of cash. Journal of Banking & Finance, 34(9), 2168-2184.
García‐Teruel, P. J., and Martínez‐Solano, P. 2008. On the determinants of SME cash holdings: Evidence from Spain. Journal of Business Finance & Accounting, 35(1‐2), 127-149.
35 Jensen, M. C. 1986. Agency costs of free cash flow, corporate finance, and takeovers. The American economic review, 76(2), 323-329.
La Porta, R., Lopez‐de‐Silanes, F., and Shleifer, A. 1999. Corporate ownership around the world. The journal of finance, 54(2), 471-517.
Kaufmann, D., Kraay, A., and Mastruzzi, M., 2009. Governance matters VIII: Aggregate and individual governance indicators, 1996-2008. World Bank Policy Research Working Paper 4978.
Opler, T., Pinkowitz, L., Stulz, R., and Williamson, R. 1999. The determinants and implications of corporate cash holdings. Journal of financial economics, 52(1), 3-46.
Ozkan, A. and Ozkan, N., 2004. Corporate cash holdings: An empirical investigation of UK companies. Journal of Banking & Finance, 28(9), 2103-2134.
Myers, S. 2001. Capital Structure. Journal of Economic Perspectives 15(2): 81-102.
Nikolov, B., and Whited, T. M. 2014. Agency conflicts and cash: Estimates from a dynamic model. The Journal of Finance, 69(5), 1883-1921.
Rajan, R. G., & Zingales, L. 1996. Financial dependence and growth (No. w5758). National bureau of economic research.
Seifert, B., and Gonenc, H. 2018. The effects of country and firm-level governance on cash management. Journal of International Financial Markets, Institutions and Money, 52, 1-16.
Seifert, B., and Gonenc, H. 2016. Creditor rights, country governance, and corporate cash holdings. Journal of International Financial Management & Accounting, 27(1), 65-90
36
Appendix 1: Sample representativeness and correlation matrix
Table 1A shows the extent of representativeness for the sample.
Table 1A: Representativeness sample in terms of industries, countries and companies
Country Initial comps Final % lost Market cap 2016 Final sample 16 lost Industry Initial comps Final % lost
Argentina 27 7 -74% 1,673,544 905,218 -46% Agriculture 712 187 74%
Australia 582 89 -85% 33,283,718 7,823,962 -76% Construction 282 155 45%
Austria 11 7 -36% 609,126 132,700 -78% Manufacturing 5,182 2,900 44%
Bahrain 12 8 -33% 1,572,289 558,507 -64% Mining 933 541 42%
Bangladesh 28 15 -46% 3,834,257 903,121 -76% Public administration 144 83 42%
Belgium 32 18 -44% 6,711,253 494,995 -93% Retail trade 620 327 47%
Brazil 18 16 -11% 2,132,273 1,923,856 -10% Services 777 457 41%
China 641 286 -55% 396,857,820 103,610,956 -74% Transport & Gas 1,166 694 40%
Croatia 17 11 -35% 1,372,158 930,010 -32% Wholesale Trade 48 29 40%
Cyprus 10 7 -30% 279,897 34,262 -88% Total 9,864 5,373 46%
Denmark 43 27 -37% 8,162,305 2,709,917 -67%
Egypt 77 28 -64% 2,072,580 499,613 -76%
Estonia 11 9 -18% 1,382,429 967,214 -30%
Finland 45 33 -27% 5,020,858 3,982,627 -21% Country Initial comps Final % lost
France 198 108 -45% 13,994,616 5,897,050 -58% Morocco 29 2 93%
Germany 192 116 -40% 19,292,164 9,054,533 -53% Zimbabwe 19 2 89%
Ghana 11 6 -45% 633,067 203,837 -68% Tanzania 5 2 60%
Greece 98 56 -43% 2,316,098 1,278,251 -45% Mauritius 4 2 50%
Hong Kong 35 16 -54% 2,470,548 424,716 -83% Romania 4 2 50%
Hungary 6 3 -50% 252,032 55,149 -78% Iceland 3 2 33% India 588 377 -36% 118,610,624 44,434,703 -63% Mexico 3 2 33% Indonesia 180 126 -30% 23,198,906 13,327,103 -43% Serbia 3 2 33% Iran 18 6 -67% 1,427,056 908,052 -36% Luxembourg 6 1 83% Ireland 13 5 -62% 2,221,124 57,505 -97% Bulgaria 5 1 80% Israel 164 100 -39% 13,881,916 6,850,868 -51% Qatar 4 1 75%
Italy 50 40 -20% 3,675,058 2,380,146 -35% Russian Federation 4 1 75%
Japan 1,132 1,003 -11% 110,622,546 98,441,615 -11% Botswana 3 1 67%
Jordan 56 31 -45% 1,953,760 462,284 -76% Lebanon 2 1 50%
Kenya 14 9 -36% 1,329,078 49,070 -96% Panama 2 1 50%
Kuwait 31 9 -71% 3,907,370 582,457 -85% Czech Republic 1 1 0%
Latvia 11 10 -9% 288,337 277,114 -4% Slovenia 1 1 0%
Lithuania 17 17 0% 1,535,177 749,764 -51% Canada 483 0 100%
Macedonia 10 8 -20% 434,405 187,835 -57% Cayman Islands 284 0 100%
Malaysia 453 6 -99% 29,411,730 0 -100% Bermuda 214 0 100%
Malta 8 3 -63% 768,615 582,100 -24% New Zealand 38 0 100%
Netherlands 20 10 -50% 2,425,134 1,122,628 -54% Côte d'Ivoire 21 0 100%
Nigeria 15 8 -47% 1,231,874 47,005 -96% Palestinian Territories 13 0 100%
Norway 34 19 -44% 4,524,981 2,845,293 -37% Jamaica 3 0 100% Oman 34 17 -50% 3,528,171 963,785 -73% Syria 3 0 100% Pakistan 206 64 -69% 20,963,802 1,285,382 -94% Algeria 2 0 100% Philippines 77 32 -58% 7,301,498 1,794,950 -75% Namibia 2 0 100% Poland 98 73 -26% 6,655,907 3,172,106 -52% Uganda 2 0 100% Portugal 9 7 -22% 209,186 209,186 0% Zambia 2 0 100%
Republic of Korea 465 370 -20% 49,043,615 31,153,378 -36% Anguilla 1 0 100%
Saudi Arabia 37 22 -41% 10,298,263 1,946,754 -81% Bahamas 1 0 100%
Singapore 288 117 -59% 15,360,086 3,690,373 -76% Cambodia 1 0 100%
South Africa 59 12 -80% 4,929,328 871,606 -82% Fiji 1 0 100%
Spain 19 19 0% 2,712,612 2,495,362 -8% French Guiana 1 0 100%
Sri Lanka 123 64 -48% 6,535,532 1,134,349 -83% Gibraltar 1 0 100%
Sweden 140 74 -47% 16,860,094 7,484,342 -56% Kazakhstan 1 0 100%
Switzerland 45 30 -33% 14,638,481 4,584,149 -69% Malawi 1 0 100%
Taiwan 936 600 -36% 97,740,442 50,240,688 -49% Marshall Islands 1 0 100%
Thailand 295 191 -35% 31,272,830 14,635,342 -53% Nepal 1 0 100%
Tunisia 15 6 -60% 598,650 169,767 -72% Papua New Guinea 1 0 100%
Turkey 113 79 -30% 10,720,454 5,143,682 -52% Senegal 1 0 100%
United Arab Emirates 14 6 -57% 2,060,988 223,405 -89% Slovakia 1 0 100%
United Kingdom 387 171 -56% 41,764,074 19,142,545 -54%
United States 969 588 -39% 139,837,744 58,192,284 -58%
Vietnam 251 178 -29% 7,736,728 3,976,629 -49%
Industry comparison
Countries lost due to missing or poor data
37 Table 1.B shows the correlation matrix. We do not find a high correlation between variables that are used in the same regression. We Note that STDEV and CFV, Lsize and Lsales, LEV and LEV2, and WC and WC2 are never used in the same regression.
