The internal- and external liquidity trade-off with
respect to future investments.
How did the global financial crisis disturb this relation?
Willem van Rijckevorsel Student Number: 6182631
Faculty of Economics and Business Corporate Finance
Abstract
This paper examines the companies’ trade-off between internal and external liquidity when considering future investments. This paper uses an extensive hand-collected dataset on revolving credit facilities for a wide array of companies during 2005 and 2008. Scrutinizing how businesses manage their needs when there are significant movements in the world of liquidity, including the global financial crisis. Carefully planned future investment decisions form the basis of corporate growth, but these investments can only be made when a company is sufficiently liquid. This paper provides detailed insights into how a firm’s liquidity position affects future investment decisions by employing several analytic models. These models evolve during the progression of this paper, subsequently displaying liquidity management alongside investment decisions. The outcomes of these models are consistent with existing empirical findings as well as their theoretical backgrounds.
Table of contents
Abstract ________________________________________________________ ii Table of contents _________________________________________________ iii 1. Introduction __________________________________________________ 1 2. Theoretical background ________________________________________3
2.1. Revolving credit Facilities _______________________________________ 3 2.2. Existing literature on liquidity management ________________________ 4 2.3. Literature inferences ____________________________________________ 6
3. Data ________________________________________________________7
3.1. Data collection process and generation of key variables _______________ 8 3.2. Additional data information ______________________________________ 9 3.3. Descriptive statistics___________________________________________ 10
4. Revolving credit facilities versus cash ____________________________ 13
4.1. Subsample analysis ____________________________________________ 13 4.2. Correlation Matrices ___________________________________________ 17 4.3. Credit line behaviour __________________________________________ 19 4.3.1. Probit Regression ____________________________________________ 20 4.3.2. Credit line size ______________________________________________ 23 4.3.3. Drawdowns ________________________________________________ 30 5. The effect on investment ______________________________________ 32
5.1. The investment model _________________________________________ 32
5.1.1. General statistics _____________________________________________ 33
5.2. Regressions __________________________________________________ 34
5.2.1. Credit lines to spending________________________________________ 35 5.2.2. Drawdowns ________________________________________________ 37 6. Conclusion __________________________________________________ 39 7. Bibliography_________________________________________________ 40
1. Introduction
Successful investments are the foundation of corporate growth, but these investments can only be made when a company is sufficiently liquid. Liquidity sources are fragmented, consequently enlarging the amount of potential techniques to which the financing aspect of investments can be approached. In determining the optimal way of funding operational costs and projects, firms have to find their ideal combination of engaging internal and external liquidity (Lins et al. 2010).
Internal liquidity mainly consists of a firm’s cash holdings and cash flows while financial intermediaries supply external liquidity; the latter exists of e.g. long-term debt and credit lines. When weighing the benefits, cash seems to be the most logical choice since it is low-cost and immediately available. Nevertheless, businesses frequently encounter difficulties managing their cash flows due to the variance in time gaps between capital needs and revenue realization. Therefore, revolving credit facilities also known as “credit lines” have become a widely used form of financing. This is one of the main characteristics of credit lines; henceforth they accommodate the credit demands of a firm alongside their fluctuations in cash flows.
There are many requirements for using a credit line e.g. a commitment fee, interest payments, financial covenants and their financial ratios. These financial constraints forces companies to trade-off internal and external liquidity. Especially during the global financial crisis, companies were forced to adjust their liquidity management strategies. This change led to a unique possibility. Examining how businesses manage their needs when there are significant movements in the world of liquidity. This brings me to the main research question: How do companies trade-off internal and external liquidity when considering future investments? Furthermore, how did the global financial crisis disturb this relation?
This paper uses a unique dataset, including an extensive hand-collected segment on credit lines for a large set of companies during 2005 and 2008. Additionally, the research analysis stands out compared to previous articles, since it not only includes publicly listed firms but also private companies required to submit 10-K’s.This paper mostly follows theories proposed by Campello et al. (2011) but differs in several aspects. Various features are taken into account such as the global financial crisis and firm characteristics. Also, this
paper includes a new angle that focuses on what in reality happened to investments during the global financial crisis.
The paper develops as follows. Section 2 clarifies the theoretical background where likewise a brief explanation on credit lines in general is given. Section 3 presents detailed information on the collection process of the datasets and provides summary statistics. Section 4 starts off with a univariate analysis followed by a discussion of the results from the multivariate analysis. Section 5 presents the research analysis with respect to investments and provides additional argumentation. Section 6 concludes the paper.
2. Theoretical background
The significance of internal and external liquidity for firms when making investment decisions continues to be heavily studied, supplying a large quantity of academic literature. Most of the preceding literature is of a theoretical nature, but recent empirical research focused on liquidity management has been conducted. This section starts by describing what revolving credit facilities are and for what reasons they are employed. Followed by an evaluation of existing literature focused on liquidity management. A selection of the most important articles examining the effect of financing on companies’ investments is also included. Overall, the intention of this section is to summarize the objectives of existing articles, and to show how their findings form a base of inspiration for the research in this paper.
2.1. Revolving credit Facilities
During 2005 and 2008 over 60%1 of all commercial and industrial loan commitments were made via revolving credit facilities. A revolving credit facility is a contract through which a financial institution gives a firm the option to borrow up to a predetermined amount over a fixed period. This contract provides additional flexibility to financing choices during the contract period, especially when considering capital expenditures that fluctuate heavily over time. Generally, revolving credit agreements are renewable at the expiry date unless certain customary covenants, defined below, are breached (Martin & Santomero, 1997).
Conventionally the client is subject to two fees in connection with the credit line. The first one is called the commitment fee; a fee charged from lender to borrower for unused credit, in other words it is a continuous charge on the unused portion of the credit line. Secondly, a drawdown, a loan resulting from the use of a credit facility, charges interest usually at a fixed markup over some benchmark variable rate, e.g., prime or LIBOR (Martin & Santomero, 1997).
It is ordinary for revolving credit facilities to be accompanied by certain contractual restrictions better known as covenants. A covenant is a promise in an indenture that certain activities will or will not be carried out. Covenants are mostly put in place to protect borrowers against themselves from defaulting on their debt obligations. In finance, covenants most often relate to terms in financial contracting, such as loan documentation stating the limits at which the borrower
can further lend, called affirmative covenants. Typically negative covenants are likewise included in the contract, which limits a firm in their conduct of business. Negative covenants are most often constituted in terms of financial ratios, such as a maximum debt to total capitalization, or similar ratios. Covenants can involve everything from restrictions on dividend payments as limits on the disposal of assets. Once a covenant is broken, the lender normally has the right to accelerate the loan and demand full and immediate payment of the entire unpaid balance. A renegotiation of the revolving credit facility usually follows (Lins et al. 2010).
2.2. Existing literature on liquidity management
Sufi (2009) conducted one of the first empirical examinations that comprise the use of revolving credit facilities in public companies’ liquidity management. The intention of his research is to explore companies’ positions within cash and credit lines relative to liquidity management. Sufi (2009) stresses that previous literature mainly argues whether the driver for liquidity is either cash holdings or credit lines. He believes that preceding research neglects the presence of an interaction effect between the two; hence he makes this a key element in the research methodology of his article. Sufi (2009) uses a unique dataset consisting of two sets of variables collected directly from 10-K SEC filings on a random sample of 300 companies from 1996 through 2003. He finds evidence that firms who maintain higher cash flows have easier access to credit lines and therefore also depend more on them. In contrast with low cash flow firms, who have difficulties in obtaining credit lines and therefore depend more on the use of cash holdings.
The results of Sufi (2009) are applicable to a relatively stable time-period, presently more research exists on financially troubled times, especially on the global financial crisis. The paper of Ivashina & Scharfstein (2010) studies the effect of the banking panic on the supply of credit to the corporate sector. Although they do not observe direct drawdowns in their data, they state that during the fourth quarter of 2008 syndicated lending was 79% lower than at the peak2 of the credit boom, while at the same time banks experienced an increase in drawdowns on credit lines. One of the interesting aspects in their paper is that the increases in drawdowns are believed to be part of a bank-run. This suggests that, during times of financial distress liquidity choices are made from a different perspective. The study performed by Ivashina & Scharfstein (2010) suggests that additional
2) Ivashina & Scharfstein (2010) split the crisis into quarters based on the amounts of granted credit loans. They state that the drop in credit was particularly steep during 2008 and call this the peak of the credit boom. 2008 will be referred to as the crisis year.
drawdowns are related to the exposure of the lending bank. This might not seem important but could be a signal that instigates firms to increase their use of external equity.
The article of Lins et al. (2010) aims to disprove the notion that credit lines serve as a substitute to cash holdings. They lay emphasis on the fact that credit lines are accompanied by certain costs and clauses which separates it from cash holdings. They expand the framework Sufi (2009) put in place regarding the positive relationship between cash flows and the use of credit lines. Using data from surveying chief financial officers from 29 countries, Lins et al. (2010) deliver evidence on whether firms indeed consider external liquidity as a substitute to external liquidity. According to the outcomes of the survey, cash holdings are used as a buffer for cash flow difficulties and lines of credit are used for future investments. Furthermore, Lins et al. (2010) conclude that firms generally do not consider credit lines as a substitute to internal liquidity.
One of the most comprehensive papers on liquidity management is written by Campello et al. (2011). Their main goal is to empirically examine the way liquidity positions are managed. A dataset with a wide array of liquidity characteristics was assembled by surveying 800 chief financial officers during 2009. In contrast to other empirical studies they observe direct drawdown information as well as many other attributes concerning credit lines. E.g. their costs and the consequences of financial covenants. Their article provides evidence that higher cash flows do not necessarily lead to an increasing usage of credit lines; it rather shows that firms with low credit lines save cash for investments. They do mention, that there is a certain threshold in cash holdings at which this characteristic changes.
Preceding literature mainly considers the tradeoffs between internal and external liquidity independent of investments. However, the objective of this paper is to provide evidence between the management of liquidity positions alongside future investment plans. So far, previous research employed various methods to establish a relation between liquidity and investments. Almeida & Campello (2001) relate liquidity choices to the sensitivity of investment spending. In their research this is done by examining the relationship between financial constraints and investment-cash flow sensitivities. The paper of Almeida & Campello (2001) suggests a theoretical framework to link the effect of financial constraints on investment behavior. Boyle & Guthrie (2003) build on this framework of investment decisions relative to the use of liquidity. Within their model the level of cash is identified as one of the main drivers for investment decisions and has two consequences. First, a decrease in the level of cash increases the need for external financing, which shrinks the
possibilities of making new investments. Second, they believe that today’s level of internal liquidity is associated with the availability of future investments.
The article of Campello et al. (2011) provides the first quantitative research connecting firms’ liquidity positions to investments. According to Campello et al. (2011), investments are divided into three elements; fixed capital, technology, and employment. Estimates for these elements are provided by extensively surveying CFO’s on their future expectations. Subsequently, surrogate variables are created from these expectations. One of their main findings considering future investments suggests that firms with large cash holdings prefer not to use credit lines. Again, this refers to whether internal and external liquidity can be substituted. To perform a similar research with actual data, proxies for these investment sections are to be defined.
The article of Baum et al. (2013) elaborates on the effect investment decisions have on internal liquidity. As believed by Baum et al. (2013) preceding papers use two proxies for future investment expenditures. They mention current expenditures together with expenditure expectations and Tobin’s Q. A theoretical framework employing Tobin’s Q is mentioned in the article of Bolton, et al. (2011). Campello et al (2011, p. 1948) refer to this framework as a suitable method to connect liquidity with investments. However, the mathematics of that technique exceed the scope of this paper. The objective of the research by Baum et al. (2013) is to empirically define changes in firms’ accumulations of cash reserves to future investment expenditures. Essentially, they examine a comparable relation to the one posed in this paper approaching it from the opposite perspective. Nonetheless, the research of Baum et al. (2013) applies interesting proxies for investments. In addition to Baum et al. (2013), the article of Opler et al. (1999) examines the determinant of corporate holdings of cash among publicly traded U.S. firms by using a “dated” sample (1971-1994). The outcomes of their research imply that firms' cash holdings increase capital expenditures as well as R&D expenses.
2.3. Literature inferences
This paper incorporates the theoretical aspects mentioned in previous research to actual observed investment expenditures during the global financial crisis. When putting all the reviewed articles together it becomes clear they display many similarities regarding the theory of liquidity management. Mostly all articles stress the importance of liquidity management, especially during times of general credit declines. A recurring notion in many of the reviewed articles seems to be
whether internal and external liquidity are interchangeable3. However, the articles do differ in their considered time periods and research methods. Especially, the absence of a consolidated model that captures the relationship between liquidity management and investments is interesting for the investment section of this paper. Accordingly, this paper tries to establish such a consolidated model by engaging a combination of several formerly mentioned theories with respect to investments.
In order to create such a model, this paper continues to build on existing frameworks. The main line of reasoning comes from the article of Campello et al. (2011), since it is the paper with the clearest organization. Campello et al. (2011) use a very clear construction method, adding new elements step by step. Consequently, this paper employs their established framework4. The interaction effect mentioned by Sufi (2009) will be taken into account when discussing the internal/external liquidity trade-off. Furthermore, the influence of the financial credit crisis, mentioned by Ivashina & Scharfstein (2010), is sure to be included by using a matching dataset.
As the subsequent sections will show, the employed research methodology is inspired by a combination of various concepts from preceding literature. Subsequently the outcomes of the analysis may well equal those of earlier authors. However, this paper aims to provide a solid empirical background to existing theories on liquidity management by using a wide-ranging dataset. Furthermore, the outcomes of this paper provide new insights on the management of liquidity whilst evaluating future investments.
3. Data
This section describes the data collection process and provides additional reasoning to some of its limitations. In order to perform a rigorous analysis, detailed information on both the definitions of internal and external liquidity is acquired for the years 2005 and 2008. The collection process of the data in this paper has led to a unique dataset, consisting of hand-collected data from the U.S. Securities and Exchange Commission, together with the necessary balance sheet information acquired through the Compustat database.
3) Henceforth, this notion will further be referred to as the substitute theorem and will form one of the main discussion elements of this paper.
4) Additional comments on the relation between this article and previous literature complement the outlining of the employed models found in section 4 and 5.
3.1. Data collection process and generation of key variables
Data on external liquidity is extracted from firms’ annual SEC-filings, in short: K or 10-KSB. A 10-K is an extensive summary report of a company's performance that must be submitted annually to the Securities and Exchange Commission, a 10-KSB refers to an annual report concerning small businesses. The SEC requires that companies with over $10 million in assets and a class of equity securities held by more than 500 owners must file annual and other periodic reports, regardless of whether the securities are publicly or privately traded. The process of manually collecting data from 10-K’s focused on whether companies hold credit lines. When a credit line is present, all accompanying information is included e.g. commitment fees, drawdown activities, interest rates, financial covenants, covenant violations, etc. Since there is no general way of identifying these aspects directly from an SEC filing, assumptions are made to generalize the possession of credit lines. Therefore, search criteria are based on the most common names5 related to revolving credit facilities, also used by Campello et al. (2011) and Sufi (2009).
Data on internal liquidity is collected by extracting the relevant variables of all manufacturing companies from the Compustat database. For some of the cash related variables lagged values are collected as well to use as instrumental variables. At that point, the internal liquidity data is merged with the data on revolving credit facilities. Firms are legally required to report their credit facilities, however sometimes it is hard to confirm whether a credit line is present or absent. Consequently, in order to still receive a balanced dataset this paper only considers companies of which information on whether a credit line (yes or no) exists. Therefore, all the companies where missing values occur are
left out of the sample. In the end, the final sample for the years 2005 and 2008 is made up of 2382 and 1932 firms, respectively.
The key financial variables are defined in the following manner. ‘Cash holdings’ consists of the
amount of cash and short-term equivalents relative to total assets. EBITDA relative to total assets is used as a proxy for a firm’s ‘Cash flow’. The data on revolving credit facilities among other things
comprise the following variables. ‘Line of credit’ is an indicator variable equal to 1 when a firm has a
line of credit in place. ‘Credit line to assets’ represents the size of the available credit line, defined as the
total loan commitment available relative to total assets. ‘Drawdowns’ covers funds drawn from a line
5) The most common names to search for credit lines in a 10-K follow from (Sufi, 2009) and are defined as: “revolving credit facilities”, “credit facility”, “revolving credit agreement”, “working capital line”, “credit lines”, “bank credit lines” and “lines of credit”. When a 10-K does not contain any of these keywords, the debt agreement is fully checked whether a credit line is indeed not in place.
of credit, measured by the amount companies have drawn from their total credit line in that particular year. ‘Covenant breach’ is an indicator variable that equals 1 when a firm has breached at
least one of the covenants accompanying a line of credit. For the investment part of the analysis additional data is included materialized6 by forward looking growth rates for a firm’s ‘Fixed capital’-, ‘Technology’-, and ‘Employment’ investments.
Several firm specific variables similar to the ones in Campello et al. (2011) are generated. The size of a firm is determined by the height of their sales/turnover, denoted by ‘Large’. Stock exchange
codes are used to identify on which stock exchange a firm is traded on, whether a firm is traded on a stock exchange is indicated by the variable ‘Public’. The credit rating for a firm is taken from
Standard & Poor’s and used to indicate if a firm is ‘Investment grade’ or not. ‘Profitable’ is an indicator
variable representing a positive net income. As performed before in the articles of Sufi (2009) and Campello et al. (2011) the companies can be differentiated in several ways. Firms are “small,” “private,” “bank-dependent,” “non-investment grade,” and are “unprofitable” if, respectively, their sales are less than $1 billion, they are privately held, they do not have a credit rating, their bonds are unrated or rated below investment grade (BBB–), and if losses are reported for the fiscal year.
3.2. Additional data information
Available data on revolving credit facilities is very limited; therefore this data had to be acquired manually. A small disclaimer has to be made concerning the manually collected data. The manually collected data is part of an agreement whereby the author of this paper collected supplementary data for a research project currently in progress, run by Dr. E. Giambona7. In exchange for supplying this data, Dr. E. Giambona granted an extensive amount of variables regarding credit facilities. However, there are some limitations to the acquired data. First, the sample is confined to all the available data of the pre-crisis year 2005 and the crisis year 2008. Second, the dataset only considers all manufacturing companies in the U.S. Regardless of these inescapable limitations the finalized sample holds unique information on a wide array of companies. Focusing on manufacturing companies might actually contribute to the investment aspects of the analysis. The two years are representative of the research scope for this paper and the sample size is still large enough for significant results. However, a larger time frame would actually be preferred.
6) See section 5.1 for a full explanation
Since a large spectrum of companies is included that differ widely in their size and conduct of business outliers may occur. Outliers can also be triggered when a combination of two or more variables have unexpected results. The existence of outliers can distort any inferences made from statistical tests based on means or standard errors. Therefore, every variable that is used in one of the models is individually checked for outliers by observing the data. When illogical values seem apparent, the relevant variable is winsorized at the upper, lower, or both 1% levels.
To see whether some specific covenants influence investments directly, additional detailed information on negative covenants is collected8. Included negative covenants are ‘Total leverage’, ‘Secured debt’, ‘Dividends’, ‘Share repurchases’, ‘Asset sales’, ‘Investment policy restrictions’ and ‘M&A restrictions’. These are indicator variables that equal 1 when a credit line is accompanied with the
specific covenant.
3.3. Descriptive statistics
Table 1 reports the descriptive statistics for every fundamental variable used in the analyses. This section is meant to provide a quantitative summary of the sample with respect to the observations that have been made. Clarifying the main image about the availability of revolving credit facilities to specific types of firms, outlining the use of credit lines during the crisis and beforehand. Furthermore, information on the forward looking growth rates for the investment variables are includes as well. This summary forms the starting point for the description of the sample as part of a more comprehensive statistical analysis.
The firm-specific variables in Table 1 clearly display the wide spectrum of companies included in the sample. During 2005, large firms make up 20% of the sample, 71% of the companies are traded on a stock exchange, around 21% of them are believed to be non-bank dependent, 9% of all the companies in the sample are rated as investment grade, and only 54% of all the companies had a positive net income at the end of the fiscal year. Relatively, there are only small upward changes when it comes to the composition of firms included in the sample during the crisis year (2008). The number of observations in the sample size for 2008 drops from 2382 to 1932, this reduction will not affect the quality of the inferences made in the analyses.
An evaluation of the liquidity variables in Table 1 immediately shows that revolving credit facilities are a common source of available funding. It is observed that 62% of the firms in the
sample have a line of credit available to them before the crisis; in 2008 a similar 64% is observed. Cash holdings to assets decrease slightly from 28% in 2005 to 26% during 2008. The descriptive statistics suggest the presence of the substitution effect between cash holdings and revolving credit facilities mentioned in the literature review. Firms with a line of credit at their disposal only hold a cash to total assets ratio of 16% in 2005 and 15% in 2008. Compared to firms without a credit line, who held a cash ratio of 47% in 2005 and 45% in 2008. Cash Flow is the only liquidity variable in table 1 displaying a negative mean; this can be explained directly from looking at the data. Firms in the lower percentiles of cash flows experienced some extremely negative results. Dealing with extreme outliers helped reducing this effect; however, still a considerable number of firms experienced very negative cash flows. On the other hand, median outcomes exhibit positive cash flows of around 8% for both years in the full sample. Firms holding a line of credit realized positive cash flows of around 11% of total assets. In contrast to firms without a line of credit who still experienced negative cash flows of around 20% in both years.
In both years, the sizes of credit lines roughly amounted to 13% of total assets. All of the firms holding credit lines approximately 19% of funds from their total credit. During 2008 this ratio increased to 23%, whether this was due to liquidity needs arising from financially difficult times cannot yet be determined. Forward looking growth ratios regarding investment spending were considerably high during 2005, opposite to times of financial distress where growth is seen to be absent. Overall, only minor differences are observed between 2005 and 2008, companies in the sample seem to have been financially troubled during both of these years. Similarities in these years will probably make it harder for the rest of the analysis to identify whether the global financial crisis disturbed any of the relations. The next section further elaborates on the underlying elements of the descriptive outcomes.
Table 1 Descriptive statistics
Descriptive statistics
Mean P25 P50 P75 Std. dev. Obs.
2005 2008 2005 2008 2005 2008 2005 2008 2005 2008 2005 2008 Line of credit 62.17 64.44 0 0 1 1 1 1 48.51 47.88 2382 1932 Large 19.90 25.10 0 0 0 0 0 1 39.93 43.37 2382 1932 Public 70.57 79.35 0 1 1 1 1 1 45.58 40.49 2382 1932 Non bank-dependent 20.61 21.95 0 0 0 0 0 0 40.46 41.4 2382 1932 Investment grade 8.90 9.99 0 0 0 0 0 0 28.48 29.99 2382 1932 Profitable 54.11 47.31 0 0 1 0 1 1 49.84 49.94 2382 1932 Cash holdings 27.95 25.58 4.57 4.54 17.80 15.46 43.53 39.41 28.23 26.25 2382 1932 With line of credit 16.19 15.03 2.86 2.95 8.87 8.91 23.28 21.65 18.54 16.92 1481 1245 Without line of credit 47.28 44.70 19.62 19.27 44.92 42.45 73.86 69.30 30.70 29.21 901 687 Cash flow -22.25 -23.73 -18.02 -15.00 7.88 7.69 14.59 14.71 91.65 94.63 2382 1932
With line of credit 3.91 1.94 5.21 3.00 11.14 11.00 16.38 16.45 44.68 47.75 1481 1245 Without line of credit -65.25 -70.25 -66.72 -80.27 -21.49 -19.23 5.75 6.12 126.33 133.07 901 687 Credit Line to assets 12.42 13.72 0.00 0.00 7.33 8.08 18.58 20.43 17.41 18.87 2382 1932 Drawdowns 18.67 22.92 0.00 0.00 0.00 6.40 32.06 41.54 27.51 29.66 1470 1237 Fixed capital growth 59.24 -4.49 -26.31 -60.63 12.24 -33.33 60.92 -2.58 202.52 156.98 2039 1670 Technology growth 26.07 -3.02 -4.97 -24.22 10.24 -7.30 32.63 8.79 77.72 52.16 1623 1352 Employment growth -71.67 -7.93 -7.89 -22.62 6.81 -10.66 31.70 4.84 130.04 27.34 57 46
This table reports summary statistics for every fundamental variable used in the analyses. The data is collected from firms’ annual filings through the SEC website and through the Wharton Compustat database. The dataset comprises all the available data for the pre-crisis year 2005 and the crisis year 2008. Line of Credit is an indicator variable equal to 1 when a firm
has a line of credit in place and 0 otherwise. Large is an indicator variable equal to 1 when a firm’s turnover equals or exceeds $1 billion and 0 otherwise. Public is an indicator variable
equal to 1 when a firm is listed on a stock exchange and 0 otherwise. Non Bank-Dependent is an indicator variable equal to 1 when a firm has an S&P credit rating and 0 otherwise. Investment Grade is an indicator variable equal to 1 if its credit rating is BBB- or higher and 0 otherwise. Profitable is an indicator variable equal to 1 when a firm has a positive net income
and 0 otherwise. Cash Holdings are cash and short-term equivalents as a percentage of total assets. Cash flow is EBITDA relative to total assets. Credit line to assets is total loan commitment
available relative to total assets. Drawdowns are the total amount of drawn credit relative to the total loan commitment available. Fixed capital growth is represented by the forward-looking
growth rate of capital expenditures. Technology growth is represented by the forward-looking growth R&D expenditures. Employment growth is represented by the forward-looking growth rate
4. Revolving credit facilities versus cash
This section studies in what way firms control their on-going liquidity needs before and during the financial crisis. A wide array of firms is included in the dataset, accommodating the research to take on a firm-specific approach. The first results are presented by using a univariate analysis, consisting of mean difference tests and correlation matrices. After, a multivariate analysis comprising several regressions is used to further scrutinize the outcomes of the univariate analysis. The fundamental research methods used in this article are generally taken from the article of Campello et al (2011) but are also inspired by other reviewed articles9.
4.1. Subsample analysis
As already seen during the descriptive statistics section, every company included in the dataset has firm-specific characteristics. Table 2 groups all these characteristics together with several variables regarding liquidity, ultimately displaying their usage of credit lines and cash holdings in 2005 and 2008. To make the outcomes more comprehensible, all firms may be assigned to one of two classifications. The first classification, known as the “regular borrower”, covers all profitable, large, public, non-bank dependent and or investment grade companies. On the other side is the “regular borrower”, which includes all small, private, unprofitable, bank dependent and or non-investment grade firms (Campello et al. 2011).
Table 2 also reports mean difference tests concerning credit lines, cash holdings and cash flows. The percentage of firms having a line of credit is shown in column 1. When comparing the means, the largest difference resides in whether a firm is bank dependent or not, 96% of the non-bank dependent firms have credit lines, compared with only 56% of the non-bank-dependent firms. All of the mean comparison tests in this columns display very significant differences. It is immediately visible that nearly all the firms who are considered regular borrowers have a line of credit at their disposal. Conversely, the availability of credit facilities to non-regular borrowers seems to be only half of that. These results are expected, since banks obviously prefer to extend lines of credit to financially stable firms.
Regular borrowers may have easier access to lines of credit, but columns 2 and 3 show that non-regular borrowers exploit these facilities to a greater extent. In the first place this becomes
apparent from evaluating whether firms exercised their right to draw funds from their lines of credit (column 2). For example, in 2005 64% of private firms used their line of credit, relative to 45% of the public firms. This is the largest difference in means of column 2 but the net result is consistent for all non-regular borrowers. Not only is the number of firms that drew funds from their credit lines higher among the non-regular borrowers, but also the amount drawn is significantly higher in all cases. This outcome, supported by column 3, is very likely attributable to the financially troubled times firms were positioned in. The ratio of funds drawn increased for every type of firm, suggesting that troubled times called for additional liquidity.
Columns 4 and 5 of table 2 show the ratio of cash holdings- and credit lines to total assets respectively. Column 4 focuses on the difference in the usage of credit lines among the groups of firms where column 5 examines the size of cash holdings. Overall, cash holdings seem to play a more significant role when it comes to comparing balance sheets directly. The amounts of cash holdings are generally three times more abundant than credit lines. This relates to the notion made by Lins et al. (2010) and others, that credit lines cannot be seen as a substitute to holding cash reserves. Possibly, credit lines are an important form of liquidity but are merely used for different purposes.
The inclusion of the sixth and final column in table 2 follows from the research of Sufi (2009). He examines whether the height of firm’s cash holdings and credit lines are connected to their cash flows. This theory stems from the notion that cash flows are essential to meet loan covenants; therefore higher cash flows would ensure easier usage of credit facilities. Column 6 of table 2 only observes the relation between cash holdings and credit lines since that is where the rest of the paper also focuses on. Any correlations between cash holdings and cash flows are presented by the correlation matrices in the next section. Column 6 shows that the presence of cash flows is generally similar amongst the different kinds of borrowers. It does not prove any of the relations mentioned by Sufi (2009).
Table 2 is a simple overview to how different firms’ approached liquidity management across 2005 and 2008. Overall, large mean differences are mainly seen between the regular borrowers and the non-regular borrowers. Somehow liquidity management does not seem to change much during 2008. Perhaps the next sections where correlations between key variables and firm characteristics are examined more closely leads to more clarification.
Table 2 Subsample analysis
Line of credit
Diff 08-05 Active drawdowns Diff 08-05 Average size of drawdowns Diff 08-05
2005 2008 2005 2008 2005 2008 Small 0.55 0.55 0 0.51 0.55 0.03 0.22 0.26 0.04** Large 0.91 0.94 0.03 0.43 0.54 0.11*** 0.11 0.18 0.07*** Difference 0.36*** 0.39*** -0.08*** -0.01 -0.11*** -0.08*** Private 0.44 0.43 -0.01 0.64 0.71 0.07** 0.31 0.38 0.07*** Public 0.7 0.7 0 0.45 0.52 0.07** 0.16 0.21 0.05*** Difference 0.26*** 0.28*** -0.19*** -0.19*** -0.15*** -0.17** Bank Dependent 0.54 0.56 0.02 0.5 0.55 0.05** 0.22 0.26 0.04** Non-Bank Dependent 0.93 0.96 0.03 0.46 0.53 0.07** 0.11 0.17 0.06 Difference 0.39*** 0.4*** -0.04 -0.03 -0.11*** -0.09* Non-Investment Grade 0.59 0.61 0.02 0.5 0.57 0.06** 0.2 0.25 0.05*** Investment Grade 0.94 0.97 0.03 0.39 0.43 0.04 0.08 0.12 0.04* Difference 0.35*** 0.37*** -0.11*** -0.13*** -0.12*** -0.13*** Unprofitable 0.41 0.52 0.11*** 0.61 0.64 0.03 0.29 0.32 0.02 Profitable 0.8 0.79 -0.01 0.44 0.48 0.04 0.14 0.17 0.03** Difference 0.38*** 0.27*** -0.17*** -0.16*** -0.15*** -0.15***
This table groups all the internal liquidity variables by their firm specific characteristics for 2005 and 2008. Mean difference tests concerning credit lines, cash holdings and cash flows are reported. The data is collected from firms’ annual filings through the SEC website and through the Wharton Compustat database. Line of credit is an indicator variable equal to 1 when a
firm has a line of credit in place and 0 otherwise. Large is an indicator variable equal to 1 when a firm’s turnover equals or exceeds $1 billion and 0 otherwise. Public is an indicator variable
equal to 1 when a firm is listed on a stock exchange and 0 otherwise. Non bank-dependent is an indicator variable equal to 1 when a firm has an S&P credit rating and 0 otherwise. Investment grade is an indicator variable equal to 1 if its credit rating is BBB- or higher and 0 otherwise. Profitable is an indicator variable equal to 1 when a firm has a positive net income and 0
otherwise. Cash holdings are cash and short-term investments as a percentage of total assets. Cash flow is EBITDA relative to total assets. Credit line is total loan commitment available
relative to total assets. Drawdowns are total credit drawn relative to the total loan commitment available. Note: ***, **, and * indicate statistical significance at 1%, 5%, and 10% (two-tail)
Table 2 (continued)
Credit line to assets
Diff 08-05 Cash holdings to assets Diff 08-05 Cash flow to assets Diff 08-05
2005 2008 2005 2008 2005 2008 Small 0.12 0.13 0.01* 0.32 0.31 -0.01* 0.12 0.13 0.01* Large 0.14 0.15 0.01 0.12 0.11 -0.02* 0.14 0.15 0.01 Difference 0.02** 0.02** -0.20*** -0.20*** 0.02** 0.02** Private 0.13 0.15 0.02** 0.28 0.24 -0.04*** 0.13 0.15 0.02** Public 0.12 0.14 0.01 0.28 0.26 -0.02 0.12 0.14 0.01 Difference 0.01 -0.01 0.01 0.02 0.01 -0.01 Bank Dependent 0.12 0.13 0.01 0.33 0.3 -0.03** 0.12 0.13 0.01 Non-Bank Dependent 0.15 0.15 0 0.1 0.1 -0.01 0.15 0.15 0 Difference 0.03** 0.02* -0.22*** -0.2 0.03** 0.02* Non-Investment Grade 0.12 0.14 0.01 0.3 0.27 -0.02** 0.12 0.14 0.01 Investment Grade 0.15 0.15 0 0.1 0.09 -0.01 0.15 0.15 0 Difference 0.03** 0.01 -0.19*** -0.18*** 0.03** 0.01 Unprofitable 0.1 0.13 0.03** 0.38 0.32 -0.06*** 0.1 0.13 0.03** Profitable 0.15 0.15 0 0.2 0.19 -0.01 0.15 0.15 0 Difference 0.05*** 0.02* -0.18*** -0.13*** 0.05*** 0.02*
This table groups all the internal liquidity variables by their firm specific characteristics for 2005 and 2008. Mean difference tests concerning credit lines, cash holdings and cash flows are reported. The data is collected from firms’ annual filings through the SEC website and through the Wharton Compustat database. Line of credit is an indicator variable equal to 1 when a
firm has a line of credit in place and 0 otherwise. Large is an indicator variable equal to 1 when a firm’s turnover equals or exceeds $1 billion and 0 otherwise. Public is an indicator variable
equal to 1 when a firm is listed on a stock exchange and 0 otherwise. Non bank-dependent is an indicator variable equal to 1 when a firm has a S&P credit rating and 0 otherwise. Investment grade is an indicator variable equal to 1 if its credit rating is BBB- or higher and 0 otherwise. Profitable is an indicator variable equal to 1 when a firm has a positive net income and 0
otherwise. Cash holdings are cash and short-term investments as a percentage of total assets. Cash flow is EBITDA relative to total assets. Credit line is total loan commitment available
relative to total assets. Drawdowns are total credit drawn relative to the total loan commitment available. Note: ***, **, and * indicate statistical significance at 1%, 5%, and 10% (two-tail)
4.2. Correlation Matrices
The mean comparison analysis already hints to the presence of relationships between the variables. This subsection evaluates the degree of linear association between them, presented in the form of a correlation matrix. The variables will be treated in a completely symmetrical way. It will not convey the impression that one variable actually cause changes to another. However, it means to show that there is evidence for a linear relationship between any two variables and that those variations in the two are linked to a degree given by the correlation coefficient.
Table 3 displays the correlation matrices for the years 2005 and 2008 respectively. The first noticeable element is displayed by the firm-specific variables, which are all significantly correlated to each other. This confirms the existence of the regular borrowers group, as all the characteristics of that class are linearly associated with each other. E.g. large companies are most often non-bank-dependent as with a correlation of 0.674 and 0.708 in 2005 and 2008 respectively. Furthermore, investment-grade companies have a high chance being of a large size at a correlation of 0.605 in 2005 and 0.568 in 2008. Additionally all the coefficients of these variables are significantly correlated with having a line of credit, which coincides with the previous finding that regular borrowers generally probably have easier access to credit lines.
Access to lines of credit is not only defined by the kind of company but also by their balance sheet items. It is expected that cash holdings and cash flow are of importance for firms when lines of credit are considered. There seems to be a negative linear association between cash holdings and the line of credit indicator variable with -0.534 and -0.541 for 2005 and 2008 respectively. This outcome suggests that firms with higher cash holdings are less likely to have lines of credit on their balance sheets. In both years, cash holdings seem to negatively affect the amount of drawdowns as well with a correlation of -0.093 in 2005 and -0.103 in 2008, this could further add to the substitute theorem, because having cash makes it unnecessary to draw any funds. These findings are consistent with the article of Lins et al. (2010) and will be examined further in the multivariate analysis, since the presence of linear association by itself is insufficient evidence.
Table 3 Correlation matrix
2005 Line of credit Large Public Non bank-dependent Investment grade Profitable holdings Cash Cash flow Credit line to assets Line of credit 1 Large 0.298* 1 Public 0.243* 0.227* 1 Non bank-dependent 0.327* 0.674* 0.147* 1 Investment grade 0.207* 0.605* 0.199* 0.613* 1 Profitable 0.394* 0.318* 0.398* 0.246* 0.276* 1 Cash holdings -0.534* -0.274* 0.008 -0.317* -0.194* -0.319* 1 Cash flow 0.366* 0.203* 0.410* 0.201* 0.132* 0.427* -0.220* 1
Credit line to assets 0.557* 0.039 -0.004 0.080* 0.043* 0.132* -0.346* 0.088* 1 Drawdowns 0.018 -0.185* -0.227* -0.179* -0.154* -0.259* -0.093* -0.214* 0.207*
Table 3 (continued)
2008 Line of credit Large Public Non bank-dependent Investment grade Profitable holdings Cash Cash flow Credit line to assets Line of credit 1 Large 0.355* 1 Public 0.233* 0.216* 1 Non bank-dependent 0.350* 0.708* 0.156* 1 Investment grade 0.230* 0.568* 0.166* 0.628* 1 Profitable 0.279* 0.291* 0.276* 0.224* 0.276* 1 Cash holdings -0.541* -0.323* 0.032 -0.320* -0.206* -0.255* 1 Cash flow 0.365* 0.235* 0.429* 0.212* 0.144* 0.384* -0.206* 1
Credit line to assets 0.540* 0.042 -0.021 0.046* 0.019 0.050* -0.350* 0.089* 1 Drawdowns . -0.130* -0.201* -0.140* -0.162* -0.253* -0.103* -0.232* 0.245*
This table reports correlation coefficients between every fundamental variable used in this analysis for the years 2005 and 2008. Line of credit is an indicator variable equal to 1 when a firm
has a line of credit in place and 0 otherwise. Large is an indicator variable equal to 1 when a firm’s turnover equals or exceeds $1 billion and 0 otherwise. Public is an indicator variable
equal to 1 when a firm is listed on a stock exchange and 0 otherwise. Non bank-dependent is an indicator variable equal to 1 when a firm has a S&P credit rating and 0 otherwise. Investment grade is an indicator variable equal to 1 if its credit rating is BBB- or higher and 0 otherwise. Profitable is an indicator variable equal to 1 when a firm has a positive net income and 0
otherwise. Cash Holdings are cash and short-term investments as a percentage of total assets. Cash flow is EBITDA relative to total assets. Credit line to assets is total loan commitment
Before a bank provides a firm with a revolving credit facility, one of their main concerns will be whether a firm receives stable cash flows. Overall, these facilities concern vast amounts of debt that are accompanied by high interest payments. The ease at which a company can cover their interest expenses is pivotal to the height of the credit line and the strictness of covenants. Therefore it is relevant to consider the relation between the height of firms’ cash flows and the presence of credit lines. The outcomes of the correlation matrices show that a linear association to this conjecture may be consistent since correlations of 0.366 in 2005 and 0.365 in 2008 are observed. Furthermore, in 2005 cash flows might be negatively related to the drawdown ratio at a coefficient of -0.214 in 2005 and -0.232 in 2008. These correlations propose that the height of cash flows could be a determinant for a bank to grant a line of credit but on the other hand they reduce the need for these facilities.
The final interesting outcomes are found in the sizes of the credit lines. The sizes of the credit lines are an important proxy for external liquidity, and are compared to cash holdings throughout this paper. When looking at correlation coefficients with respect to the ratio of credit lines to total assets, it looks that a firms’ cash flows and cash holdings are important determinants. Cash flows were already mentioned when considering the presence of a credit line but might also be a positive influence on their size with a correlation of around 0.088 in both years. On the other hand, cash holdings could be a negative influence on the heights of credit lines with a negative relation of -0.346 in 2005 and -0.350 in 2008. The next sections perform actual regression analyses and provide supplementary material for further discussion of the suggested relations in this section.
4.3. Credit line behaviour
The univariate analysis is the foundation on which further research is conducted. The previous section defined the variables of greatest importance and also made some suggestions to possible prevailing relations and in what direction they move. Each of the tables in the following sections exists out of a stepwise constructed model. The tables are discussed column by column and every column contains the years 2005 and 2008. The regression in the first column of each table will comprise the dependent variable, one explanatory variable and the firm specific control variables. The regressions in the subsequent columns further specify the model by adding more explanatory variables. This is done to inspect the effect of including an additional variable and to verify whether it improves the model. To control for multicollinearity, the control variable ‘Profitable’ is left out of
performed with clustered standard errors to control for heteroscedasticity. This is more efficient than including indicator variables for each industry, because regressions are clustered by specific industry types within the manufacturing industry10.
The regression analyses presented in the following sections further scrutinizes the theories that presented themselves in the univariate analysis. The regression analysis will mostly replicate the models used in Campello et al. (2011) and Sufi (2009). However, the models used in this paper are augmented in certain aspects and outcomes contrast those of Campello et al. (2011) and Sufi 2009 by using different periods and using actual observed data. First a probit analysis is performed to see what drives the likelihood for a firm to actually have a credit line to its disposal. Second OLS regressions are performed in separate stages on the determinants of the sizes of credit lines. The third and final subsection reviews the ratio of drawdowns by the method of OLS regressions.
4.3.1. Probit Regression
Reviewing the descriptive statistics (table 1) of the full sample, there is a 0.62 (2005) or 0.64 (2008) probability that a firm actually has a line of credit to its disposal. This section further examines all the firms from the sample whether or not they have a line of credit. The objective here is to determine how firm characteristics, cash flow and cash holdings add to the probability of having a revolving credit facility. When evaluating the effect variables have on a certain probability, a non-linear model is considered. Therefore, to perform the analysis in this section a probit regression is employed. A full specification of the model can be expressed by the following equation:
𝐿𝑖𝑛𝑒 𝑜𝑓 𝑐𝑟𝑒𝑑𝑖𝑡 {0,1} = 𝑐 + 𝛽1𝐶𝑎𝑠ℎ 𝑓𝑙𝑜𝑤𝑖+ 𝛽2𝐶𝑎𝑠ℎ ℎ𝑜𝑙𝑑𝑖𝑛𝑔𝑠𝑖+ 𝛽3(𝐶𝑎𝑠ℎ 𝑓𝑙𝑜𝑤 × 𝐶𝑎𝑠ℎ ℎ𝑜𝑙𝑑𝑖𝑛𝑔𝑠)𝑖+ 𝛾𝑿𝒊+ 𝜀𝒊 The limited dependent variable is ‘Line of credit {0, 1}’, regressed on ‘Cash flow’, ‘Cash holdings’ and
their interaction term. Furthermore, the variable X exists of all the firm specific control variables. The constant in the regression equation is denoted by 𝑐 and 𝜀 is the error term.
The outcomes of the probit regression analysis are presented by table 4. However, interpretations of the results from a probit regression require a different approach than with a standard OLS regression. The probit model assumes, the higher the outcome of the probit value, the
10) Standard errors are clustered by the sub-classifications within the manufacturing industry by using their SIC-codes. Since there are quite a lot of sub-industries this is more efficient than using the fixed effects model.
higher the probability that a firm has a line of credit. Each coefficient in table 4 represents the marginal impact of a unit change in an explanatory variable, ceteris paribus, to the probability of having a line of credit. The center of attention in this analysis lies on column three, since it is the most complete and significant version of the model. First the mean probability of the probit model for each year is calculated by using the following formula:
𝑃𝑅(𝐿𝑖𝑛𝑒 𝑜𝑓 𝑐𝑟𝑒𝑑𝑖𝑡 = 1)𝑖 = 1+𝑒−(𝛽0� +𝛽1� ×𝐶𝐹����𝑖+𝛽2� ×𝐶𝐻�����𝑖+𝛽3� ×𝐶𝐹����𝑖×𝐶𝐹����𝑖+𝛽41 � ×𝐿𝑎𝑟𝑔𝑒����������𝑖+𝛽5� ×𝑃𝑢𝑏𝑙𝚤𝑐����������𝑖+𝛽6� ×𝑁𝐵𝐷𝑒𝑝�����������𝑖
In fact, this equation converts the outcome of the product (the Z-Value) of the estimated coefficients and their accompanying mean values to the probability that a firm has a line of credit. The mean probabilities for having a credit line are 0.60 and 0.62 for 2005 and 2008 respectively11. The estimated mean outcomes of this model seem to confirm that the specification fits the data accurately, since the values are not far off what was found in the descriptive statistics. Increasing cash flow by 1% increases the probability of having a line of credit by 0.44% for 2005 and 0.18% for 200812, includes the interaction effect. Increasing cash holdings by 1% decreases the probability of having a line of credit by 0.09% for 2005 and 0.44% for 2008. Furthermore, a negative interaction between cash flow and cash holdings is significantly present with an interaction of -0.699 in 2005 and -0.851 in 2008. Additionally, the probability is higher for large, public and non bank-dependent firms.
The outcomes of the probit regressions provide insights into how the discussed variables add to the probability of a firm having a revolving credit facility at its disposal. The mean probability outcomes of each year confirm that the specified model fits the data well. Moreover, the results confirm one of the relations revealed by the correlation matrix by several aspects. According to the correlation matrix (table 3), cash flow might be negatively correlated to the size of credit lines, but here they add to the accessibility of them. The outcomes of this section advocate that cash flow contributes to the probability of having a credit line. Perhaps the results in this section display that firms with high cash flows have a higher probability of having a credit line, while the accompanying cash holdings show a lower dependency on them.
11) The appendix to table 4 displays a visual explanation to how these figures are calculated.
12) Calculated the following way: 0.60((0.01*0.543) + (-0.0028*-0.699)) = 0.0044, as to include the interaction effect. Also see the appendix to table 4.
Table 4
Dependent variable: Line of credit Column 1. Column 2. Column 3.
2005 2008 2005 2008 2005 2008 Cash flow 0.562*** 0.470*** 0.367*** 0.283*** 0.543*** 0.505***
(0.08) (0.09) (0.07) (0.10) (0.07) (0.09) Cash holdings -2.659*** -2.814*** -2.907*** -3.240***
(0.22) (0.22) (0.16) (0.15)
Cash flow x cash holdings -0.699*** -0.851***
(0.07) (0.09) Large 0.441* 0.667*** 0.206 0.356** 0.201 0.332** (0.24) (0.22) (0.17) (0.15) (0.16) (0.13) Public firm 0.204** 0.177* 0.603*** 0.649*** 0.623*** 0.664*** (0.09) (0.10) (0.08) (0.10) (0.07) (0.10) Non bank-dependent 0.937*** 0.934*** 0.634*** 0.752*** 0.608*** 0.732*** (0.16) (0.14) (0.16) (0.15) (0.17) (0.16) Investment grade -0.154 0.005 -0.194 -0.078 -0.203 -0.101 (0.24) (0.33) (0.21) (0.35) (0.21) (0.36) Observations 2,382 1,932 2,382 1,932 2,382 1,932
This table reports the outcomes of a probit regression where Line of credit is taken as the dependent variable. Line of credit is an indicator
variable equal to 1 when a firm has a line of credit in place and 0 otherwise. The regressions in this table are performed on the full sample. All regressions include a constant term (not reported). Cash flow is EBITDA relative to total assets. Cash holdings are cash and
short-term investments as a percentage of total assets. Large is an indicator variable equal to 1 when a firm’s turnover equals or
exceeds $1 billion and 0 otherwise. Public is an indicator variable equal to 1 when a firm is listed on a stock exchange and 0 otherwise. Non bank-dependent is an indicator variable equal to 1 when a firm has a S&P credit rating and 0 otherwise. Investment Grade is an
indicator variable equal to 1 if its credit rating is BBB- or higher and 0 otherwise. T-statistics reported in parentheses are based on
clustered standard errors. Note: ***, **, and * indicate statistical significance at 1%, 5%, and 10% (two-tail) test levels, respectively. Appendix to table 4 2005 2008 β Mean β Mean Constant 0.680 - 0.644 - Cash Flow 0.543 -0.220 0.505 -0.240 Cash Holdings -2.907 0.280 -3.24 0.260 Interaction term -0.699 -0.0616 -0.851 -0.063 Large 0.201 0.200 0.332 0.250 Public 0.623 0.710 0.664 0.790 Non bank-dependent 0.608 0.210 0.732 0.220 Z-Value 0.40 0.50 PR(LC-Dummy = 1) 0.60 0.62
4.3.2. Credit line size
Where the probit regression of the previous section examined what variables added to the probability of having a credit line, this section wants to determine how credit lines and cash holdings behave relative to assets. One of the inferences drawn from the univariate analysis was that the height of cash flows could be a determinant for a bank to grant a line of credit; on the other hand they also reduce the need for these facilities. To further scrutinize this finding, the model in this section focuses on the effect cash holdings and cash flows have on the height of the total loan commitment available. To stay in line with one of Sufi’s (2009) main comments, their interaction effect will be taken into account as well.
The model is engaged over three separate stages. The first stage of the regression analysis is executed comprising the full sample of firms13. The second stage narrows down the sample size to all firms who have a credit line higher than 0; in other words where the ‘Line of credit’ indicator
variable equals 1. The third stage compares the data of 25% of the firms with the highest cash holdings to the lower 75%. The threshold separator in that stage is suggested by Campello et al (2011)14. All three stages will apply the same model where the total loan commitment available to total assets is taken as the dependent variable. Adding explanatory variables in each column until the complete linear equation is defined as creates the model:
𝐶𝑟𝑒𝑑𝑖𝑡 𝑙𝑖𝑛𝑒
𝐴𝑠𝑠𝑒𝑡𝑠𝑖 =𝑐 + 𝛽1𝐶𝑎𝑠ℎ 𝑓𝑙𝑜𝑤𝑖+ 𝛽2𝐶𝑎𝑠ℎ ℎ𝑜𝑙𝑑𝑖𝑛𝑔𝑠𝑖+ 𝛽3(𝐶𝑎𝑠ℎ 𝑓𝑙𝑜𝑤 × 𝐶𝑎𝑠ℎ ℎ𝑜𝑙𝑑𝑖𝑛𝑔𝑠)𝑖+ 𝛾𝑿𝒊+ 𝜀𝒊
‘Credit line to total assets’ is regressed on ‘cash flow’, ‘cash holdings’ and their interaction. Furthermore the
variable X exists of all the firm specific control variables: ‘Large’, ‘Public’, ‘Non bank-dependent’ and
‘investment grade’. Again, the constant in the regression equation is denoted by 𝑐 and 𝜀 is the error
term.
13) Meaning that firms without a credit line have the amount 0 in the total loan commitment available instead of a missing value.
14 ) Campello et al (2011) separate firms at the median level. To search for more extreme outcomes, the analysis of this paper checks the top 25% relative to the bottom 75%
4.3.2.1. Stage 1
In this stage the complete sample of firms is taken, the results of which are presented in table 5. The regression analysis starts off in the first column where the effect of cash flow on the height of credit lines to total assets is examined. In 2005, a 1% increase in cash flow would increase the height of credit lines by 0.019%, for 2008 this would be 0.023%. The effect of this coefficient is small but the positive linear relation is comparable with the results of Sufi (2009) and Campello et al. (2011). Also, the coefficient is statistically significant both in 2005 and 2008.
The analysis is expanded in the second and third columns where the effects of cash holdings are introduced to the model. When holding all other variables constant the second column verifies the results of the correlation matrices, which suggested the presence of a negative linear relation between cash holdings and the height of credit lines. In the third column the model is augmented with an interaction term, which substantially changes the interpretation of the coefficients. The third column is also the most important column of this table to draw inferences from. It not only has the highest R-squared value (always gets higher when adding more values if the adjusted R-squared is not used) but it also has the highest number of statistically significant coefficients.
In column three, there still is a positive relationship between cash flow and the size of credit lines, insinuating that a firm’s income helps raise the level of a credit line. However, the effect of cash flow is only significant during 2008 and the size of this coefficient is miniscule (0.023**) compared to cash holdings. Adding the interaction term amplifies the effect cash holdings have on the height of credit lines; this is implied by the rise in the coefficients for cash holdings. E.g. the coefficient for the year 2008 went from -0.268 to -0.299. The interaction term is significant through both periods with a coefficient of -0.038 in 2005 and -0.055 during 2008. This could indicate that the effect of cash flow is different at different values of cash holdings and vice versa. The effect of which further comes to light in the third stage of this analysis. The firm-specific control variables do not display any consistent behavior throughout the model, except for the size of the company. The control variable large in column three suggests a negative linear relation in both years. This suggests that larger firms generally have smaller lines of credit relative to assets.
The point of this analysis is to find determinants for the size of credit lines. Overall, the main findings in the first stage on the size of credit lines are similar to previous empirical research. Just as in the article of Sufi (2009) and Campello et al. (2011) the outcomes show that a positive linear relationship between cash flow and the size of credit lines is present. Furthermore, it can be said that the negative relation between cash holdings and credit lines suggested by the same authors is established. Even when analyzing the full sample in which not all firms employ revolving credit facilities, the results are in line with the hypotheses. Nevertheless, the outcomes in table 5 are a good start but they are not sufficient to draw any strong inferences from. In the second stage the sample size will be narrowed down which probably exposes more detailed results.
Table 5 (stage 1)
Dependent variable: Credit line to assets 2005 Column 1. 2008 2005 Column 2. 2008 2005 Column 3. 2008
Cash flow 0.019** 0.023*** 0.004 0.007 0.017 0.023** (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Cash holdings -0.221*** -0.268*** -0.240*** -0.299***
(0.01) (0.02) (0.01) (0.02)
Cash flow x cash holdings -0.038** -0.055***
(0.02) (0.02) Large -0.013 0.008 -0.032*** -0.025** -0.034*** -0.027** (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Public firm -0.020** -0.035*** 0.001 -0.003 0.002 -0.002 (0.01) (0.02) (0.01) (0.02) (0.01) (0.02) Non bank-dependent 0.036*** 0.013 0.002 -0.016 -0.001 -0.019* (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Investment grade 0.004 -0.009 0.007 -0.005 0.007 -0.005 (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Observations 2,382 1,932 2,382 1,932 2,382 1,932 R-squared 0.015 0.014 0.124 0.130 0.128 0.136
This table reports the coefficients of the OLS regressions on the size of credit lines for stage 1. The OLS regressions in this table are performed on the full sample. All the columns use as the dependent variable Credit line to Assets, which is total loan commitment
available relative to total assets. All regressions include a constant term (not reported). Cash flow is EBITDA relative to total assets. Cash holdings are cash and short-term investments as a percentage of total assets. Large is an indicator variable equal to 1 when a firm’s
turnover equals or exceeds $1 billion and 0 otherwise. Public is an indicator variable equal to 1 when a firm is listed on a stock
exchange and 0 otherwise. Non bank-dependent is an indicator variable equal to 1 when a firm has a S&P credit rating and 0 otherwise. Investment grade is an indicator variable equal to 1 if its credit rating is BBB- or higher and 0 otherwise. T-statistics reported in
parentheses are based on clustered standard errors. Note: ***, **, and * indicate statistical significance at 1%, 5%, and 10% (two-tail) test levels, respectively.
4.3.2.2. Stage 2
In this stage the same model is used as in the previous stage, only the sample used to analyse the size of the loan commitment is narrowed down to all firms who actually have a line of credit. The outcomes of this model with the altered sample size are presented by table 6. The sample now consists of 1481 and 1245 firms for the years 2005 and 2008 respectively. The effect this has on the size and the significance of the coefficients is immediately visible. Furthermore, without changing any of the variables the R-squared for every regression increased impressively, ensuring a better fit of the model to the data.
In contrast to the full sample analysis, the effect cash flows have on the height of the total loan commitments increased significantly. In all three columns the size of these coefficients more than quadrupled and they are all significant at the 1% level. Opposed to stage one the coefficients no longer exhibit a positive linear relation, instead a negative one is observed. In the first column cash flow now has a coefficient of -0.125 and -0.099 for 2005 and 2008 respectively. Augmenting the model with cash holdings and the interaction term to the model increases the effect cash flow has on the size of credit lines even more. In column three a 1% increase in cash flow would result in a 0.14% decrease in the size of loan commitments. Contrary to the positive correlation in the article of Campello et al. (2011), the relation found here suggest that firms with high cash flows have smaller loan commitments on their balance sheets.
When examining the effect of cash holdings for firms with credit lines, Columns two and three again confirm the findings of Sufi (2009) and Campello et al. (2011). Once more the negative relation from cash holdings to the size of credit lines is established. Compared to the results from stage 1 the size of the coefficients decreased to some extent, e.g. column three 2005 dropped from -0.240 to -0.193. Furthermore, adding the interaction term for cash flow and cash holdings in column does not improve the model in this stage. Narrowing down the sample size seems to remove the interaction effect between cash flow and cash holdings.
