Risk management and financial constraints in small U.S. non-financial firms Pauline Loijson 10587896 July 2018 Master Thesis Thesis supervisor Dhr. Dr. J.E. Ligterink University of Amsterdam Amsterdam Business School
MSc. Finance
Track: Quantitative Finance
Abstract
This thesis researches the trade-off between financial constraints and risk management for small U.S. non-financial firms. The effect of several measures of current net worth on both next quarter’s -, and next year’s fraction hedged by derivatives is measured, by use of time series fixed effect OLS regressions with standard errors clustered at industry level. Empirical support provides a significant positive relation between net worth and hedging, indicating that firms that are more financially constrained, hedge less. Furthermore, the effect of previous year’s derivatives usage on current credit rating is measured, by use of time series probit models with clustered standard errors at industry level. Contradicting results show only a significant positive impact of past derivatives usage on B-rated firms and a negative impact on B+-rated firms. Multidimensional panel data is used to conduct these analyses.
Table of contents
1. Introduction ... 4
2. Literature review ... 6
2.1. Risk management in small firms ... 7
2.2. Hypotheses ... 10
3. Data ... 11
3.1. Derivatives usage analysis ... 11
3.2. Financial constraints analysis ... 13
3.3. Summary statistics ... 14
4. Methodology ... 16
4.1. Derivatives usage analysis ... 16
4.2. Financial constraints analysis ... 19
4.3. Econometric limitations ... 21
5. Results ... 22
5.1. Derivatives usage analysis: one-quarter prediction ... 22
5.2. Derivatives usage analysis: one-year prediction ... 25
5.3. Financial constraints analysis ... 27
6. Robustness ... 29
7. Conclusion ... 35
8. Reference List ... 38
Statement of originality
This document is written by, Pauline Loijson, who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business of the University of Amsterdam is responsible solely for the supervision of completion of the work, not for the contents.
1. Introduction
Contradictory theories exist on the trade-off between risk management and financial constraints for small firms in the existing literature. It is often theorized that less capitalized firms would benefit from risk management, as it supposedly adds value to the firm (Froot, Scharfstein, & Stein, 1993: Tufano, 1996). However, as these firms are often financially constrained, it appears more difficult to hedge risk (Rampini & Viswanathan 2010). The contradictory trade-off between the need of hedging risk to reduce financial constraints, but potential limitations in executing risk management because of these financial constraints, leads to the following question: “What is the relation between financial constraints and the
hedging behavior of small U.S. non-financial firms?”
Managing risk by derivatives usage is said to enable firms to increase their debt capacity, ease informational asymmetries, increase tax shields, and mitigate future underinvestment, therefore reducing financial distress costs and financial constraints (Allayannis, Lel, & Miller, 2012: Bodnar, Giambona, Graham, & Harvey, 2016). However, risk management is also said to only be beneficial when firms are able to credibly commit to their hedging strategy (Ligterink, 2001). Recent empirical findings suggest that more financially constrained firms hedge less (Rampini & Viswanathan, 2010: Rampini, Sufi, & Viswanathan 2014). This decline in hedging is said to be a result of collateral constraints. Less capitalized firms have less collateral, leading to a trade-off between risk management and financing. Rampini et al. (2014) imply that in situations of low net worth, firms rather suffice investment needs and forcefully downsize less, than hedge risk.
The findings of Rampini et al. (2014) indicate that more constrained firms hedge less. Their research however applies to the airline industry, and does not support empirical evidence for small firms. Moreover, extensive literature on risk management and financial constraints exists for large firms, but research on this trade-off for small firms appears to be scarce. Generally, smaller firms tend to be less capitalized, and therefore often more financially constrained (Vickery, 2008). Hence, a deeper insight into the trade-off between risk management and financial constraints is relevant for small firms, as they could potentially improve their hedging strategy based on these findings.
In order to create a deeper insight into this trade-off, the core question is divided into two sub questions: (1) “To what extent do financial constraints influence the future hedging behavior of small U.S. non-financial firms?”, and (2) “To what extent does the previous hedging of risk by use of derivatives influence current financial constraints that small U.S. non-financial firms experience?”. These questions are answered by use of two sub analyses.
The first analysis consists of time series ordinary least squares regressions, controlling for fixed effects at firm level and clustering standard errors at industry level. It measures the effect of financial constraints on next period’s hedging by derivatives usage. The tests produce both a one-quarter, and one-year forecast. Three measures of net worth are used to proxy for financial constraints, based on the theory that during periods of collateral constraints, risk management is low when the marginal value of net worth is high (Rampini et al., 2014). The latter is the case when current net worth is low. Derivatives usage is measured as the net position in derivatives scaled by total assets at book value. Firm fixed effects are controlled for in order to mitigate potential dependence at firm level.
The second analysis consists of time series probit regressions, using clustered standard errors at industry level, and tests the effect of last year’s hedging behavior by derivatives usage on the current credit rating. Credit rating is used as a measure of financial constraints (Rampini et al., 2014). Hedging by derivatives usage is measured by the net position in derivatives scaled to the firm’s total assets at book value.
The multidimensional panel data used in this research consists of quarterly data and ranges from the last quarter of 2013 to the last quarter of 2016. The data consists of a sample of 5428 small U.S. non-financial firms that use derivatives to hedge risk, and 402 firms that do not. Firms that use derivatives for trading are excluded from the sample to reduce the probability of biased results from noise from trading effects (Bonaimé, Hankins and Harford, 2014). In order to mitigate a potential selection problem that arises from firms self-selecting by basing their derivatives choice on expected benefits that Allayannis et al. (2012) mention, the firms are matched on net worth between the derivatives users and non-users categories. In this research, firms are considered to be “small” when they earn total revenue below 250 million U.S. dollars quarterly (Bodnar et al., 2016). Although a lot of literature considers “small” to be much smaller, this measure is used due to lack of availability of data for the smallest firms. In order to mitigate omitted variable bias, control variables are included in both analyses. Standard errors are clustered at industry level to mitigate potential dependence on the industry level.
The empirical results of this thesis indicate that a positive relation exists between net worth and next year’s fraction hedged by derivatives to total assets, implying that more financially constrained firms hedge less next year. This relation appears strongest among distressed firms. The general relation between net worth and next quarter’s fraction hedged appears insignificant, but significance does exist for distressed firms. Firms rated CCC+ or worse are significantly proven to reduce next quarter’s hedging more than firms rated
otherwise. The results of the one-year prediction support the evidence that Rampini et al. (2014) provide. However, the validity of the results is arguable, as they do not appear to be robust in a first differences specification. Furthermore, contradicting results show only a significant positive impact of past derivatives usage on B-rated firms and a negative impact on B+-rated firms, leading to question the validity of these results.
The thesis proceeds as follows. Section 2 discusses the existing theories on risk management and financial constraints in general, and focused on small firms by use of empirical and theoretical literature. Section 3 provides information on the data used for this research, and explains how the variables used are constructed. Section 4 provides the models used for the analyses. Section 5 provides empirical evidence from the conducted tests on the relation between risk management and hedging. Section 6 provides results from robustness checks in order to test the validity of the empirical results. Section 7 concludes this thesis.
2. Literature review
Firms encounter many financial risks during their existence. In the absence of market imperfections, firm value would not improve by managing risk (Modigliani & Miller, 1958). However, many theories on corporate risk management prove that in the existence of market imperfections, corporate risk management may add value (Guay & Kothari, 2003). Within the category of financial risks that firms encounter, currency risk and interest rate risk appear the two most important types of risk for many corporations.
More specifically, when firms seek lower interest rates, borrowing in a foreign currency could be beneficial. However, foreign borrowings accompanied by potential fluctuating exchange rates create risk for a firm (Brown, Ongena, & Yesin, 2010). When taking on loans domestically, empirical evidence proves that high interest rates lead to a lower availability of both internal and external financing relative to possible investments. When interest rates increase, both expected future profits, and the maturity mismatch of long-lived assets and short-term liabilities weaken the firm’s financial position resulting in lower investment, and therefore output (Vickery, 2008).
Firm investment is not only sensitive to fluctuating interest and currency rates, but also to credit-market conditions. Firms benefit from the availability of external financing, as it is one way of enabling them to finance investments. However, it is a necessity for firms to have sufficient collateral when they want to borrow in credit markets, because informational asymmetries between firms and their creditors create agency costs. In absence of sufficient collateral, the availability of external financing will lower for these firms. This is reflected in
the fact that firms experiencing a greater loss in collateral obtain a lower amount of bank credit (Gan, 2007). In the occurrence of general tighter credit-market conditions, collateral will naturally be lower, also resulting in a lower external capital availability (Perez-Quiros & Timmermann, 2000). A lower availability of external financing might disable firms to carry out their investment strategies.
Without market imperfections, hedging these risks would not add value. However, several market imperfections can cause volatility to be costly. The most important imperfections are costly external financing, the existence of taxes, costs of risk averse managers, and financial distress costs. Existing theories of risk management argue that “the primary goal of risk management is to eliminate the probability of costly lower-tail outcomes—those that would cause financial distress or make a company unable to carry out its investment strategy” (Guay & Kothari, 2003).
When it is more costly to raise external financing than internal financing, hedging potentially adds value by matching capital inflow and outflow more closely, thereby lowering the probability that firms need to enter the credit-markets to raise external capital (Froot et al., 1993). Lowering cash flow volatility by hedging can increase debt capacity, and create tax benefits, hence adding value to the firm (Graham & Rogers, 2002). Hedging risk also potentially lowers the required risk premium of risk averse managers, when having their compensation tied to the firm’s, hereby adding value. Lastly, hedging can add value by reducing financial distress costs of a levered firm, by lowering the probability of distress (Guay & Kothari, 2003).
One way of managing the financial risks that firms face –now considering hedging adds value in the existence of market imperfections- is by use of interest rate, currency or commodity derivatives. Conclusively, managing risk by derivatives usage is said to enable firms to increase their debt capacity, ease informational asymmetries, increase tax shields, and mitigate future underinvestment, therefore reducing financial distress costs and financial constraints (Allayannis et al., 2012: Bodnar et al., 2016).
2.1. Risk management in small firms
Certain risks have different effects on smaller firms than on larger firms. A key rationale for managing risk is financial constraints. Small firms are less capitalized than larger firms, and more often financially constrained (Rampini & Viswanathan, 2010: Vickery, 2008). Therefore, interest rate risk and currency risk amongst other risks have a different effect on small firms than larger firms.
More specifically, small firms are relatively more affected by interest rate shocks than large firms. In periods of rising interest rates, credit constrained firms find that the availability of both internal and external financing lowers relatively to investment opportunities. When interest rates increase, both expected future profits, and the maturity mismatch of long-lived assets and short-term liabilities weaken the small firm’s financial position. As a result, the present value of the firm’s assets declines more as opposed to the present value of their liabilities. Additionally, the net current cash flow decreases. The former causes the firm’s creditworthiness to decline, ultimately lowering their ability to find external financing. This results in lower output and investment (Vickery, 2008).
Furthermore, theories on currency risk in small firms suggest that small firms can be expected to borrow in foreign currencies when having revenues in foreign currencies. However, highly leveraged small firms are less likely to borrow in a foreign currency, due to the higher costs of distress. Foreign debt usage by small firms appears relatively stronger related to firm-characteristics than macroeconomic conditions. Exchange rate fluctuations are known to affect growth performance of financially constrained firms (Brown et al., 2010). Exchange rate fluctuations do not only affect growth performance, but also earnings. An exchange rate appreciation causes a reduction in current earnings, and therefore a reduction in the firm’s borrowing ability in order to overcome idiosyncratic liquidity shocks. Subsequently, a depreciation causes an increase in current earnings and hence, borrowing ability. It is stated that in the existence of financial constraints, generally the negative effect of an appreciation does not outweigh the positive effect of depreciation, therefore imposing the constrained firm to risk (Aghion, Bachetta, Rancière, & Rogoff, 2009). As small firms are more likely to be financially constrained, they are more likely exposed to currency risk when taking on foreign debt (Géczy, Minton, & Schrand, 1997).
Although foreign currency might be less sensitive to macro-economic factors, not all risk for small firms is unaffected by macro-economic conditions. It is argued that informational asymmetries are most important for firms that are less collateralized, younger and more exposed to idiosyncratic risk. Firms with these characteristics tend to be smaller firms. These informational asymmetries, just like the previously argued rising interest rates, increase the firm’s cost of external financing. Public information about small firms is less available, and therefore such firms are more reliable on credit that is insensitive to information, like bank loans. Subsequently, small firms are more sensitive to tightening credit market conditions, especially in financial recessions (Perez-Quiros & Timmermann, 2000). As tighter credit market conditions imply lower collateral, and therefore lower ability of
raising external financing, smaller firms might benefit from securing future cash flows by hedging, in order to be protected against the potential future inability to raise external financing alongside insecure internal financing.
Lastly, distress costs are said to be an important indicator of priced risk exposure. Low-leveraged firms in distress have a stronger deteriorating operating performance than high-leveraged distressed firms. Distress costs emerging from low payoff states contribute to systematic risk, as low asset payoffs are partially systematic. Systematic risk cannot be diversified, but it can be hedged (George & Hwang, 2010).
When taking into consideration that small firms bear these risks arisen from financial constraints, and that less capitalized firms are more risk averse, it could be concluded that these firms are therefore more likely to hedge all these mentioned systematic risks, as managing these risks potentially adds value to the firm (Froot et al., 1993: Tufano, 1996).
Although it is found that smaller firms use different risk management strategies than larger firms, it is not yet researched to what extent managing risk by use of derivatives influences small firm’s financial constraints (Howton & Perfect, 1998). Large non-financial corporations use derivatives to dampen cash flow volatility, as it will lower the likelihood of costs of using external financing, thus lowering the likelihood of becoming financially constrained (Guay & Kothari, 2003). Small firms are more likely to encounter high volatilities, thereby incentivized to use derivatives.
Considering these potential value-adding effects of hedging, it could be concluded that small firms benefit from using derivatives to manage the risks faced. However, controversy emerges by research proving that more financially constrained firms hedge less (Bodnar et al., 2016: Rampini & Viswanathan, 2010: Rampini et al., 2014). Firms have an incentive to conserve debt capacity as it allows for exploiting future investment opportunities. However, conserving debt capacity and engaging in risk management bears opportunity costs. When engaging in risk management, the conserving of both net worth and debt capacity lowers the availability of net worth to use for investment opportunities, resulting in these opportunities foregone. Small, constrained firms with less internal funds are more productive on the margin as they operate on a smaller scale. Therefore, the opportunity costs tend to be larger for less capitalized firms (Rampini & Viswanathan, 2010).
Subsequently, the need for current investments of small firms in order to satisfy growth tends to override the need for hedging risk. Conclusively, small, constrained, firms with less internal funds might rather exhaust their debt capacity and abstain from risk management
instead of bearing the opportunity costs that rise from conserving (Rampini & Viswanathan, 2010).
It follows that exhausting debt capacity is considered to be efficient under the condition of financial constraints. Not conserving is said to be unrelated to the inability of hedging. However, when constrained firms are collateralized, their borrowing ability is raised ex ante, creating allowance for exhausting their debt capacity even more, which will lower their net worth ex post (Rampini & Viswanathan, 2010). Finally, they remain with a lower collateralizability and net worth, resulting in a stronger inability to hedge risk.
Rampini et al. (2014) build upon the idea that in practice less financially constrained firms hedge more. They use a sample of airlines in order to find evidence to support the prediction that more financially constrained firms hedge less than unconstrained firms. It is found that airlines with a higher credit rating indeed hedge more, which is consistent with their previous findings (Rampini & Viswanathan, 2010). They empirically prove that in the year before entering distress, airlines’ hedging strongly decreases from 25% of its expected fuel expenses to less than 5%. In the two years after, the hedging ratio slowly increases back to 20%. They explain this decline to be a result of collateral constraints. The collateral constraints lead to a trade-off between risk management and financing, implying that in situations of low net worth firms rather suffice investment needs and forcefully downsize less, than hedge risk.
2.2. Hypotheses
Despite the fact that many motives for hedging risk exist for small, constrained firms, hedging is only expected to be beneficial when a firm is able to credibly commit to its hedging strategy (Ligterink, 2001). The collaterization problem, along with the need to suffice investment needs, and avoid large downsizing, overrides the willingness to hedge risk (Bodnar et al., 2016: Rampini & Viswanathan, 2010; Rampini et al., 2014). Therefore, it is hypothesized that small firms that are more financially constrained, hedge less than small firms that are less financially constrained. Specifically, a decrease in net worth, implicating larger constraints and higher collateral, is expected to decrease the percentage hedged with derivatives for small firms.
Furthermore, assuming that in general small constrained firms hedge less, it can be argued that small firms that do decide to hedge have carefully assessed the benefits and costs of their hedging strategy, and will only apply a hedging strategy when its benefits will outweigh its costs. Hence, it is hypothesized that small firms that do decide to apply a hedging
strategy will improve their financial situation. More specifically, it is expected that an increase in the amount hedged by derivatives, increases the credit rating of firms that do apply a hedging strategy.
3. Data
In this section information on the data used for this research, and on construction of variables is provided. The data consists of a sample of 5,428 small U.S. non-financial firms that use derivatives to hedge risk, and 402 firms that do not. Firms are considered to be “small” when they earn total revenue below 250 million U.S. dollars quarterly (Bodnar et al., 2016). Although a lot of literature considers “small” to be much lower than 250 million U.S. dollars quarterly, this measure is used due to a lack of availability of data for the smallest firms.
It is determined whether firms hedge or not by use of data on derivatives, retrieved from the Compustat database. The matching principle is applied, in order to reduce the self-selection problem arising from firms self-selecting by basing their derivatives choice on expected benefits (Allayannis et al., 2012). By matching firms on net worth, it is intended to assure hedging firms and non-hedging firms have similar firm characteristics.
Derivatives cannot only be used for hedging, but also for trading. By excluding financial firms, the sample will less likely include firms that use derivatives to hedge risk, creating less biased results from noise of trading effects (Bonaimé, Hankins and Harford, 2014). Hence, financial firms are excluded from this sample. The hedging purposes will manually and sample-wised be verified by use of statements in 10-K forms from the derivative users, which can be found on the Thomson One’s database. The data is quarterly and ranges from the last quarter of 2013 until the last quarter of 2016. Two analyses are conducted, each with their own set of variables.
3.1. Derivatives usage analysis
In the first analysis, the effect of financial constraints on hedging is researched. The dependent variable in this analysis is derivatives usage. To capture scale differences between firms, the amount hegded by derivatives is scaled by the firms’ total assets at book value. The variable is measured by the net position held in derivatives relative to total assets (3.1). The data on derivatives usage is retrieved from Compustat database.
The explanatory variables in this analysis are proxies of financial constraints. The proxies are the net worth to assets at book value ratio (3.2), the net worth to assets at market value ratio (3.3) and credit rating, which is measured by the firm’s Standard & Poor’s
domestic long-term issuer credit rating. Net worth is used as proxy, considering that during periods of collateral constraints, risk management is low when the marginal value of net worth is high. The latter is the case when the level of net worth is low, hereby resembling a sufficient proxy. To capture scale differences between firms, the book and market values of net worth are scaled by the firms’ total assets (Rampini et al., 2014). The issuer credit rating measures to what extent firms are expected to be able to meet their long-term financial obligations from the current perspective (Compustat S&P Ratings database, 2018). When firms are unable to meet their financial obligations, they are considered to be financially constrained. Therefore, the S&P domestic long-term issuer credit rating is a suited proxy for financial constraints. The data is retrieved from the Compustat database and the Compustat S&P’s rating database, respectively.
The control variables used in this analysis are leverage, growth opportunities, cash flow volatility and the cash to assets ratio, resembling collateral. All data is retrieved from the Compustat database. Leverage is measured by the ratio of the book value of debt to market value of assets (3.4). The market value of assets exists of the market value of equity and the total liabilities (3.5). Highly levered firms face higher costs of distress, creating a potential incentive to reduce these costs by hedging. Several studies have found a positive relation between leverage and hedging (Guay & Kothari, 2003). Therefore, leverage is included as a control variable.
Cash flow volatility is measured by the volatility of the cash flows from operational activities. It is included in as a control variable, because a higher volatility increases costs of external financing, therefore creating an incentive to hedge (Guay & Kothari, 2003).
Growth opportunities is included as a control variable as it is expected that investment needs override hedging needs, and growth opportunities resemble investment opportunities (Rampini & Viswanathan, 2010). Growth opportunities are measured by the market to book ratio of assets (3.6) (Guay & Kothari, 2003).
Lastly, the amount of collateral a firm has to offer is measured by their cash to assets ratio (3.7). Collateral is included as control variable as collateral constraints are said to reduce the ability of hedging (Rampini et al., 2014).
(3.1) 𝐷𝑒𝑟𝑖𝑣𝑎𝑡𝑖𝑣𝑒𝑠 𝑢𝑠𝑎𝑔𝑒 = !"# !"#$%$"& !!"# !" !"#$%&'$%"( !"#$% !""#$"
(3.3) 𝑁𝑒𝑡 𝑤𝑜𝑟𝑡ℎ 𝑡𝑜 𝑎𝑠𝑠𝑒𝑡𝑠 (𝑚𝑣) = !"!#$ !"#$%& !"#$% !" !""#$"!"!#$ !""#$"!!"!#$ !"#$ (3.4) 𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒 = !"#$%& !"#$% !""#$"!"!#$ !"#$ (3.5) 𝑀𝑎𝑟𝑘𝑒𝑡 𝑣𝑎𝑙𝑢𝑒 𝑠𝑠𝑒𝑡𝑠 = 𝑚𝑎𝑟𝑘𝑒𝑡 𝑣𝑎𝑙𝑢𝑒 𝑒𝑞𝑢𝑖𝑡𝑦 + 𝑡𝑜𝑡𝑎𝑙 𝑑𝑒𝑏𝑡 (3.6) 𝐺𝑟𝑜𝑤𝑡ℎ 𝑜𝑝𝑝𝑜𝑟𝑡𝑢𝑛𝑖𝑡𝑖𝑒𝑠 = !"#$%& !"#$% !" !""#$"!""#$%&'( !" !""#$" (3.7) 𝐶𝑜𝑙𝑙𝑎𝑡𝑒𝑟𝑎𝑙 = !"!#$ !""#$"!"#!
3.2. Financial constraints analysis
In the financial constraints analysis the effect of derivatives usage on a firm’s future financial constraints is measured. The dependent variable in this analysis is a measure of financial constraints. The level of financial constraint is again measured by the firm’s Standard & Poor’s domestic long-term issuer credit rating. The ratings available in this dataset range from A+ until SD. To enable a numerical analysis, all ratings are transformed into a numerical rating, ranging respectively from 22 until 1. The credit ratings are retrieved from Compustat S&P Ratings database.
The explanatory variable in this analysis is derivatives usage. The variable is constructed as the net position held in derivatives relative to total assets (3.1), similar to section 3.1. The data is retrieved from the Compustat database.
The control variables in this analysis are the net worth to assets ratio at book value (3.2), and cash flow volatility. Net worth is said to have adverse effects on firms with high distress costs (i.e. financially constrained firms) (Denis & Sibilkov, 2010). Again, the firms’ total assets scale net worth, in order to capture potential scale differences between firms. Cash flow volatility is measured by the volatility of the cash flows from operational activities. It increases costs of external financing, therefore possibly increasing financial constraints (Guay & Kothari, 2003). Hence, these variables are suited control variables. The data is retrieved from the Compustat database.
3.3. Summary statistics
The descriptive statistics of the variables included in models are presented in table 1. The matching principle is applied to the net worth to book value of assets ratio. After matching on net worth, the means of credit rating, cash to assets ratio and cash flow volatility do also no longer significantly differ.
In both categories the average credit rating lies between 9 and 10. This corresponds to values between BB- and B+. The average credit rating is non-investment grade, indicating that on average, firms are fairly at risk of being unable to meet their financial obligations (Compustat S&P Ratings database, 2018). The average ratings in both categories do not significantly differ (table 2). The standard deviation of the derivatives users category is slightly higher than the standard deviation of the non-users category, meaning the ratings in the former category are slightly more widespread than in the latter category.
The average net position in derivatives held relative to total assets at book value is 2.58% with a standard deviation of 4.24%. This is very small compared to the research of Guay and Kothari (2003), where 17% of total assets at book value are hedged on average. However, in contrast to their research, which is on large U.S. non-financial firms, this research is on small U.S. non-financial firms. The median of market value of equity is 2,376 million U.S. dollars for the derivatives users, while it is only 466 million U.S. dollars in this research, which is roughly 5 times as large. It is in line with the theory that small firms hedge less than large firms. The standard deviation is larger than the mean, indicating widespread data points.
The averages of the net worth to assets ratio at book value in both categories do not significantly differ, because the matching principle was applied, and the categories are purposely matched on the net worth to book value of assets ratio (table 2). However, the average net worth to market value of assets ratio differs quite strongly between the two categories. The net worth to market value is roughly 35% larger for derivatives users, compared to non-users. The standard deviation in the non-user category is as large as the mean, but it is almost twice as large for the user category, indicating the data in the latter category is almost twice as spread out as in the former category.
Finally, the control variables are compared between categories. Table 1 shows that the average growth opportunities appear larger for the non-derivatives users. This is in line with the theory, as derivatives users use part of their net worth to hedge risk, rather than to invest, therefore reducing their growth opportunities. The standard deviations are larger than the means, indicating data is widespread. Secondly, the average leverage ratio is significantly
larger in the user category. This insinuates a potential positive relation between leverage and hedging, which is also proven in several previous studies (Guay & Kothari, 2003). Lastly, the means of the user and non-user categories do not significantly differ for both the cash to assets ratio, and the cash flow volatility (table 2). Both variables show standard deviations larger than the mean, indicating data is widespread. The standard deviation of the cash flow volatility of the derivatives users is approximately twice as large as the non-user category’s standard deviation, indicating even wider spread data.
Table 1
Summary statistics
This table presents the summary statistics for the categories of derivatives users and derivatives non-users.
Descriptive statistics ($ thousands) No. of firms Mean Standard
Deviation Minimum Maximum
Panel A: Derivatives users
Net position in derivatives to total assets (bv) 1,042 0.026 0.042 2.41e-06 0.436
Net worth to assets (bv) 4,734 0.674 0.304 -3.755 1
Net worth to assets (mv) 4,670 0.688 1.303 -0.764 36.975
Credit rating 1,780 9.570 2.726 1 18
Growth opportunities 4,670 1.745 2.148 0.027 44.956
Cash flow volatility 4,381 0.049 0.091 0 1.605
Leverage 4,670 0.301 0.280 0 0.996
Cash to assets (bv) 5,428 0.139 0.196 0 0.994
Panel B: Derivatives non-users
Net position in derivatives to total assets (bv) 402 0 0 0 0
Net worth to assets (bv) 402 0.677 0.308 -1.784 0.990
Net worth to assets (mv) 402 0.507 .502 -0.633 5.820
Credit rating 89 9.416 2.571 2 17
Growth opportunities 402 2.670 4.677 0.162 80.833
Cash flow volatility 312 0.0460 0.051 0 0.332
Leverage 402 0.223 0.236 0 0.994
Table 2
Results mean comparison test
This table presents results from the mean comparison tests, using the t-test statistic. Standard errors are in parenthesis. Coefficients that are statistically different from zero at the, 1%, 5% and 10% significance levels are denoted by ***, **, *, respectively.
4. Methodology
In this section the methodologies of the two core analyses are presented separately. The models used, each with their corresponding variables, are discussed in detail. The hypotheses on the additional variables follow. Lastly, several econometric problems that arise with conducting this research are discussed.
In this research, two main analyses are conducted. In the derivatives usage analyses, the effect of current financial constraints on future derivatives usage is researched. In the financial constraints analyses, the effect of past derivatives usage on current financial constraints is measured. The dataset consists of multidimensional panel data. The debt, assets, derivatives, cash flow from operations, and market value of equity values are winsorized at 1% in order to reduce the chance of spurious outliers.
4.1. Derivatives usage analysis
This analysis is constructed to test the hypothesis that small firms that are more financially constrained hedge less than small firms that are less financially constrained. Specifically, the effect of net worth and credit rating on next period’s percentage hedged by derivatives is tested. Rampini et al. (2014) test the same hypothesis, in a sample of airlines. As many differences between large airlines and the sample of small firms in this research exist, the core variables are similar. Therefore, their methodology is largely followed in this research.
Variable Absolute mean
difference
Number of observations
Net worth to assets (bv) 0.002*** 5,136
(0.016)
Credit rating 0.154*** 1,869
(0.295)
Cash flow volatility 0.003*** 4,693
(0.005)
Cash to assets (bv) 0.013*** 5,830
Similar to Rampini et al. (2014), a fixed effects ordinary least squares (OLS) time series regression model is used, as it captures the effects of financial constraints on hedging behavior. Fixed effects are specifically controlled for, as possible dependence at firm level exists, resulting from the likelihood that derivatives use is correlated from year to year (Allayannis et al., 2012). A fixed effects model partially solves this problem by absorbing the dependence at firm level.
The firm’s choice of derivatives may differ on the industry level, as corporations from different industries encounter different organizational structures (Pennings & Garcia, 2004). Different industries may face different economic conditions and constraints, possibly leading to different choices in derivatives usage. Therefore, the motives for firms to use derivatives might not be homogenous. The dependence at firm and industry level tends to inflate standard errors in fixed effects models (Allayannis et al., 2012). This could be partially solved by use of robust standard errors, as these take the possibility of dependence within clusters into account. Since the dependence at firm level is already absorbed as fixed effects, clustering standard errors at firm level would be redundant, and is therefore not conducted. Hene, this research model only uses clustered standard errors at industry level.
In order to create a testable hypothesis, the variables are transformed into testable variables. In order to predict derivatives usage the dependent variable in this research is next period’s amount hedged by use of derivatives as a fraction of total assets. To capture scale differences between firms, their total assets at book value scale the amount hedged.
To measure the level of financial constraints that firms experience, the same three variables are constructed and used as Rampini et al. (2014) use. Net worth is used as proxy, considering that during periods of collateral constraints, risk management is low when the marginal value of net worth is high. The latter is the case when the level of net worth is low, hereby resembling a sufficient proxy. Again, to capture scale differences between firms, net worth is scaled by the firms’ total assets. Therefore, the effect of both net worth at book value of assets, and net worth scaled by the market value of assets is measured.
When testing the prediction of net worth and credit rating on derivatives usage, omitted variable bias is a concern (Rampini et al., 2014). Leaving variables out of the regression while they are determinants of the dependent variable, and correlated with the explanatory variable, could cause omitted variable bias. One way to address this problem is by including control variables (Stock & Watson, 2015). Therefore, the models include growth opportunities, cash flow volatility, cash to assets at book value -resembling collateral-, and
leverage as control variables. Following from the described literature (Guay & Kothari 2003), and as explained in the data section, these variables appear to be suited control variables.
Following from the described literature, it is expected that growth opportunities have a significant negative effect, as investment needs are expected to override the need for hedging (Rampini & Viswanathan, 2010). It is expected that cash flow volatility has a positive significant effect on hedging, as hedging is used to dampen cash flow volatility (Guay & Kothari, 2003). Collateral is expected to have a significant positive effect, as larger collateral allows for more hedging (Rampini et al., 2014). Lastly, it is expected that leverage has a positive effect on hedging. High leverage is accompanied by high costs of distress, possibly incentivizing firms to hedge this risk. Several studies have found a significant positive effect of leverage on hedging. The expectation of the significance of this effect is debatable as empirical findings on this estimate are mixed (Guay & Kothari, 2003).
Four regression models are used to test the hypothesis. The first models include net worth to the book value of assets and net worth to market value as core explanatory variables (4.1). The third model includes credit rating as core explanatory variable (4.2). In the fourth model three dummy variables are included instead of the non-binary credit rating variable, indicating in with credit rating category the firm is placed (4.3) (Rampini et al., 2014). The dummies are added to the model to represent certain levels of financial constraints, separating the general effect of changes in credit rating. This partially mitigates the concern that firms in distress drive the results. The results from this test will simplify the comparison between different levels of constraints. The first dummy includes credit ratings BB+, BB, and BB-. The second dummy includes ratings B+, B, and B-. The last dummy includes ratings of CCC+ or worse. The models are specified as follows:
(4.1) 𝑁𝑒𝑡 𝑎𝑚𝑜𝑢𝑛𝑑 ℎ𝑒𝑑𝑔𝑒𝑑 𝑤𝑖𝑡ℎ 𝑑𝑒𝑟𝑖𝑣𝑎𝑡𝑖𝑣𝑒𝑠!,!!!= 𝛽!+ 𝛽!𝑛𝑒𝑡 𝑤𝑜𝑟𝑡ℎ 𝑡𝑜 𝑎𝑠𝑠𝑒𝑡𝑠 !"+ 𝛽!𝑔𝑟𝑜𝑤𝑡ℎ 𝑜𝑝𝑝𝑜𝑟𝑡𝑢𝑛𝑖𝑡𝑖𝑒𝑠!"+ 𝛽!𝑙𝑒𝑣𝑒𝑟𝑎𝑔𝑒!"+ 𝛽!𝑐𝑎𝑠ℎ 𝑓𝑙𝑜𝑤 𝑣𝑜𝑙𝑎𝑡𝑖𝑙𝑖𝑡𝑦!"+ 𝛽!𝑐𝑜𝑙𝑙𝑎𝑡𝑒𝑟𝑎𝑙!"+ 𝜆!+ 𝜀! (4.2) 𝑁𝑒𝑡 𝑎𝑚𝑜𝑢𝑛𝑑 ℎ𝑒𝑑𝑔𝑒𝑑 𝑤𝑖𝑡ℎ 𝑑𝑒𝑟𝑖𝑣𝑎𝑡𝑖𝑣𝑒𝑠!,!!!= 𝛽!+ 𝛽!𝑐𝑟𝑒𝑑𝑖𝑡 𝑟𝑎𝑡𝑖𝑛𝑔!"+ 𝛽!𝑔𝑟𝑜𝑤𝑡ℎ 𝑜𝑝𝑝𝑜𝑟𝑡𝑢𝑛𝑖𝑡𝑖𝑒𝑠!"+ 𝛽!𝑙𝑒𝑣𝑒𝑟𝑎𝑔𝑒!"+ 𝛽!𝑐𝑎𝑠ℎ 𝑓𝑙𝑜𝑤 𝑣𝑜𝑙𝑎𝑡𝑖𝑙𝑖𝑡𝑦!"+ 𝛽!𝑐𝑜𝑙𝑙𝑎𝑡𝑒𝑟𝑎𝑙!"+ 𝜆!+ 𝜀!
(4.3) 𝑁𝑒𝑡 𝑎𝑚𝑜𝑢𝑛𝑑 ℎ𝑒𝑑𝑔𝑒𝑑 𝑤𝑖𝑡ℎ 𝑑𝑒𝑟𝑖𝑣𝑎𝑡𝑖𝑣𝑒𝑠!,!!!=
𝛽!+ 𝛽!𝑛𝑒𝑡 𝑤𝑜𝑟𝑡ℎ 𝑡𝑜 𝑎𝑠𝑠𝑒𝑡𝑠 (𝑏𝑣)!"+ 𝛽!𝑔𝑟𝑜𝑤𝑡ℎ 𝑜𝑝𝑝𝑜𝑟𝑡𝑢𝑛𝑖𝑡𝑖𝑒𝑠!"+ 𝛽!𝑙𝑒𝑣𝑒𝑟𝑎𝑔𝑒!"+
𝛽!𝑐𝑎𝑠ℎ 𝑓𝑙𝑜𝑤 𝑣𝑜𝑙𝑎𝑡𝑖𝑙𝑖𝑡𝑦!"+ 𝛽!𝑐𝑜𝑙𝑙𝑎𝑡𝑒𝑟𝑎𝑙!"+ 𝛽!𝐵𝐵!, 𝐵𝐵, 𝐵𝐵!𝑟𝑎𝑡𝑖𝑛𝑔!"+ 𝛽!𝐵!, 𝐵, 𝐵!𝑟𝑎𝑡𝑖𝑛𝑔!"+
𝛽!𝐶𝐶𝐶! 𝑜𝑟 𝑤𝑜𝑟𝑠𝑒 𝑟𝑎𝑡𝑖𝑛𝑔!"+ 𝜆!+ 𝜀!
n indicates the period forecasted. The models are tested for n=1 and n=4, modeling a
one-quarter and one-year forecast. The one-quarter forecast is used to predict a fairly immediate effect, and the one-year forecast to predict a longer-term forecast. Due to a lack of availability of long-term data on derivatives usage, longer-term predictions cannot be conducted. 𝜆!represents the firm fixed effects.
4.2. Financial constraints analysis
Assuming that in general small constrained firms hedge less, it can be argued that small firms that do decide to hedge have carefully assessed the benefits and costs of their hedging strategy, and will only apply a hedging strategy when its benefits will outweigh its costs. In this analysis the effect of past derivatives usage on current credit rating is measured. Therefore, the hypothesis that small firms that hedge improve their credit rating is tested in this analysis.
The dependent variable is a credit rating dummy, representing the level of financial constraints a firm embodies. As the dependent variable is a binary variable, a time series probit model is used. Regressions with binary dependent variables model the probability that the binary variable equals 1. Hence, a nonlinear model that forces dependent variables to be between 0 and 1 is adopted (Stock & Watson, 2015). The probabilities between 0 and 1 are produced by cumulative probability distributions, and in the probit model the standard normal cumulative distribution function is used.
The coefficients are estimated using the maximum likelihood method (MLM). This method creates efficient estimators with minimum variance. In large samples the MLM estimators are consistent and normally distributed (Stock & Watson, 2015).
Following the same reasoning as in the previous analysis, robust standard errors are used, as these take the possibility of dependence within clusters into account. Standard errors are clustered at industry level, as firm levels are panels, and panels are not nested within clusters.
The analysis consists of several probit regressions. Several dummy variables with corresponding sample ranges are the dependent variables. This separates the effects of derivatives usage on credit rating between different categories of credit ratings. This method
simplifies the comparison of the effect of derivatives usage between different credit ratings. The dummies equal to one if the probability that the firm is above the corresponding credit rating threshold equals to one. Therefore, a significant positive effect on the dummy can be interpreted as an increase in the probability to be rated above the threshold rating, resembling an increase in the probability to become less financially constrained.
These credit rating thresholds used as dependent variables vary from the S&P domestic long-term issuer credit ratings BB+ until CCC+. The BB+ rating indicates that a firm’s ability to meet its financial obligations is fairly at risk. A change in circumstances or economic conditions could lead to a lower capability of meeting financial commitments. However, the firm is less vulnerable on the short-term than lower-rated obligors. Every step down in credit rating indicates a higher level of financial constraints. The CCC+ rating, which is the lowest threshold used, indicates that a firm is currently vulnerable, and dependent on favorable changes in financial circumstances and economic environment to meet its financial obligations (Compustat S&P Ratings database, 2018). Firms with credit rating CCC+ are considered to be in financial distress.
The explanatory variable in the analyses is the net position in derivatives held relative to total assets at book value four quarters ago. The lagged value of four quarters is used, as managing risk by use of derivatives is not expected to have an immediate effect on a firm’s credit rating. Due to a lack of availability of longer-term derivatives data, a longer-term forecast cannot be modeled.
Following the same reasoning as the previous analysis, control variables are included to reduce the chance of omitted variable bias. Based on the literature, it is expected that the control variable net worth to book value of assets ratio has a significant positive effect, as a larger net worth supposedly lowers the chance of being financially constrained (Rampini et al., 2014). The control variable of cash flow volatility is expected to have a significant negative effect, as more volatile cash flows imply less security to meet future financial obligations (Guay & Kothari, 2003).
The first probit analyses consist of samples of firms with credit ratings ranging from one credit rating above and below the thresholds, four quarters ago. Credit ratings larger or smaller than this sample are taken out of consideration, as the likelihood of a firm changing its credit rating by more than one rating within one year is assumed to be low. The dependent binary variable equals one when the current credit rating is above the threshold rating. The explanatory variable is the net position in derivatives relative to assets four quarters ago. The
control variables are the net worth to book value of assets ratio and net worth. The probit model is as follows:
(4.4) Pr 𝐶𝑟𝑒𝑑𝑖𝑡 𝑟𝑎𝑡𝑖𝑛𝑔 𝑡ℎ𝑟𝑒𝑠ℎ𝑜𝑙𝑑!,!,!" =
1 𝑑𝑒𝑟𝑖𝑣𝑎𝑡𝑖𝑣𝑒𝑠 𝑢𝑠𝑎𝑔𝑒, 𝑛𝑒𝑡 𝑤𝑜𝑟𝑡ℎ 𝑡𝑜 𝑎𝑠𝑠𝑒𝑡𝑠 𝑏𝑣 , 𝑐𝑎𝑠ℎ 𝑓𝑙𝑜𝑤 𝑣𝑜𝑙𝑎𝑡𝑖𝑙𝑖𝑡𝑦) = 𝜙(𝛽!+ 𝛽!∗
𝑑𝑒𝑟𝑖𝑣𝑎𝑡𝑖𝑣𝑒𝑠 𝑢𝑠𝑎𝑔𝑒!,!!!+ 𝛽!∗ 𝑛𝑒𝑡 𝑤𝑜𝑟𝑡ℎ 𝑡𝑜 𝑎𝑠𝑠𝑒𝑡𝑠 𝑏𝑣 !,!+ 𝛽!∗ 𝑐𝑎𝑠ℎ 𝑓𝑙𝑜𝑤 𝑣𝑜𝑙𝑎𝑡𝑖𝑙𝑖𝑡𝑦!.!)
tr indicates the threshold credit rating, varying from BB+ until CCC+. A significant positive
effect of the explanatory variable implicates that an increased amount hedged with derivatives four quarters ago, increases the chance that a firm improved its credit rating.
In the second analysis, the same model (4.4) is used as previously, but the sample changes. To properly capture and compare the effects of hedging on credit rating, the firms’ previous state of financial constraints is considered. Hence, this analysis separates the sample of the first probit analyses. Hereby, the effect of derivatives usage on credit rating can be tied to the firms’ previous credit rating. In the second set of probit analyses the sample consists of firms with a credit rating equal to the threshold credit rating The previous financial state is determined as four quarters ago, which is the same period the percentage hedged by derivatives is measured. This test indicates whether the probability that a firm improves its credit rating next year increases, if it increases its percentage hedged by derivatives.
4.3. Econometric limitations
Several econometric issues arise in both analyses. Firstly, the previously described potential dependence at firm and industry level is only partially solved by controlling for fixed effects and clustering standard errors at the industry level. Secondly, a self-selection problem occurs. The sample of derivatives users is not random, as firms tend to self-select by basing their derivatives choice on expected benefits (Allayannis et al., 2012). This self-selection problem is partially solved by matching the categories on net worth at book value. After matching on net worth at book value, the categories match on credit rating, cash to assets ratio and cash flow volatility as well. When matched on net worth, the self-selection bias is reduced, as it is less likely that only the larger firms within the sample use derivatives. However, the self-selection problem is not completely solved. Lastly, omitted variable bias is also a problem occurring in this research. This is partially solved by including control variables into the regressions (Stock & Watson, 2015).
A way of reducing these biases to a larger extent is by using derivatives as a function of other variables that influence the amount hedged (Allayannis et al., 2012). Due to a lack of
availability on derivatives data, the dataset used in this research is too small to be able to conduct a research like such, hence certain selection- and omitted variable biases do limit this research.
5. Results
In this section the results of the analyses are presented and interpreted. The estimates of the core explanatory variables on the dependent variables are extensively discussed and placed within the existing literature. The results tied to the control variables will be only briefly discussed and placed within existing literature.
5.1. Derivatives usage analysis: one-quarter prediction
Table 3 reports the results of the test of the hypothesis that small firms that are more financially constrained hedge less than firms that are less financially constrained. The tests contained cross-sectional time series fixed effects ordinary least squares regressions, controlling for firm fixed effects, and clustering standard errors at the industry level. Table 3 reports the predictions of the effects of net worth and credit rating on next quarter’s percentage hedged by derivatives. It appears that no statistically significant relation exists between current net worth and next quarter’s derivatives usage. This is not in line with the hypothesis, nor the empirical findings of Rampini et al. (2014). However, Rampini et al. (2014) forecast one-year predictions, and table 3 presents results of one-quarter forecasts, hence it is arguable to what extent these results are comparable. Implementing a new hedging strategy that is adjusted by a firm’s new financial situation might be not be feasible within three months, considering it could be both time consuming and possibly costly.
The effect of credit rating on next quarter’s percentage hedged does appear to be a robust and statistically significant positive effect at the 1% significance level (table 3). A one-step increase in credit rating is related to a 0.4% increase in next quarter’s percentage hedged. This is in line with the hypothesis that more financially constrained firms hedge less. Moreover, it supports the results of Rampini et al. (2014). However, this significant estimate is not in line with the insignificant estimates of the other two proxies of financial constraints: the measures of net worth. The empirical results of Rampini et al. (2014) show the effect of credit rating to be more significant than the measures of net worth. It is possible that a one-step change in credit rating is perceived as a more convincing valuation of financial constraints in quarterly decisions than a 1% change in net worth.
The positive effect of credit rating on the percentage hedged in the third regression presented in table 3, is also reflected in the fourth regression. Although the credit rating categories of BB+, BB, BB- and B+, B, B- appear to have no significant effect on the percentage hedged, the last category containing ratings of CCC+ or lower has a robust and statistically significant negative effect on next quarter’s percentage hedged. This means that firms rated CCC+ or worse hedge 2.95% less than firms rated higher than CCC+. The category of CCC+ or lower implicates that firms are in distress (Rampini et al., 2014). These findings support the hypothesis that more financially constrained firms hedge less. Moreover, they imply that distressed firms hedge less than non-distressed firms. This does however insinuate that the significant positive effect of credit rating in the previous regression was driven by the effect of distressed firms.
The results from table 3 implicate that generally becoming more financially constrained, does not lead to less hedging the next quarter. However, in case of reaching the worst state of constraints -distress-, the results show that firms do tend to decline their percentage hedged, even though hedging could potentially reduce financial distress costs (Allayannis et al., 2012: Bodnar et al., 2016: Guay & Kothari, 2003). When a firm is already in distress, the value adding effect of hedging might already be diminished, as the firm can no longer credibly commit to its hedging strategy (Ligterink, 2001). The investment needs, and unwillingness to downsize seem to override the will and ability to hedge (Rampini et al., 2014). It could be concluded that changing a hedging strategy within three months could potentially be more costly than beneficial to a firm. Therefore, firms would generally not change their strategy on such short notice. However, when in distress, the benefits from reducing the percentage hedged might override the costs of implementing this on such short-term, leading distressed firms to reduce their percentage hedged nevertheless.
Table 3
One-quarter prediction of net worth on hedging with derivatives: firm fixed effects
This table presents coefficient estimates of time series fixed effect ordinary least squares regressions relating the next quarter's net position in derivatives hedged scaled by total assets (bv) to measures of net worth in the current quarter. The dependent variable is the fraction of next quarter's net position in derivatives scaled by total assets (bv). Standard errors are in parenthesis. Coefficients that are statistically different from zero at the, 1%, 5% and 10% significance levels are denoted by ***, **, *, respectively.
The control variables are all statistically significant at least at the 10% significance level, except for growth opportunities (table 3). No significant effect on next quarter’s percentage hedged is found for the latter. Table 3 shows a significant positive relationship between leverage and next quarter’s percentage hedged at the 1% significance level. This is in line with the hypothesis. A statistically significant positive effect is shown for cash flow volatility at the 10% significance level, conform the hypothesis. Lastly, the results show a significant relation between the cash to assets ratio resembling collateral and next quarter’s percentage hedged. This does not meet the expectation, as an increase in collateral was expected to increase hedging.
Net worth to assets (bv)
Net worth to
assets (mv) Credit rating
Credit rating dummies
(1) (2) (3) (4)
Measure of net worth 0.000533 -0.00111 0.00404***
(0.00411) (0.00171) (0.000597) Rating = BB+, BB, or BB- -0.00340 (0.00443) Rating = B+, B, or B- -0.00374 (0.00460) Rating = CCC+ or worse -0.0295*** (0.00357) Control variables Growth opportunities -0.00180 -0.00196 -0.000203 -0.000817 (0.00165) (0.00174) (0.00167) (0.00108) Leverage 0.0501*** 0.0494*** 0.0645*** 0.0657*** (0.00834) (0.00587) (0.00578) (0.00794)
Cash flow volatility 0.185* 0.185* 0.350*** 0.164*
(0.107) (0.106) (0.0631) (0.0973) Cash to assets -0.0319** -0.0327** -0.0288** -0.0188** (0.0133) (0.0146) (0.0119) (0.00756) Constant 0.00305 0.00460 -0.0465*** -0.00155 (0.00832) (0.00517) (0.00920) (0.00809) Number of observations 1000 1000 495 1000 Adjusted R-squared 0.077 0.078 0.129 0.104
5.2. Derivatives usage analysis: one-year prediction
Table 4 also reports the results of the test of the hypothesis that small firms that are more financially constrained hedge less than firms that are less financially constrained. Again, the tests contained fixed effect ordinary least squares regressions, controlling for firm fixed effects and using clustered standard errors at the industry level. Table 4 however reports the predictions of the effect of net worth and credit rating on next year’s percentage hedged by derivatives. From the first two columns it appears a robust and statistically significant positive relation exists between net worth and next year’s percentage hedged at the 1% significance level. Although a 1% increase in net worth to book value of assets, shows an increase of 6.91% in next quarter’s percentage hedged, and the same increase in the net worth scaled by the market value only shows a 1.54% increase, both results confirm the hypothesis that more financially constrained firms hedge less than less financially constrained firms. The positive nature of these results are conform the results of Rampini et al. (2014).
Although the general effect of credit rating on next year’s percentage hedged is insignificant, significance is found in the regression with the credit rating categories. Column 4 shows that a downgrade from the credit rating category of B+, B, B- to CCC+ or worse, reduces next year’s percentage hedged by 3.42%. Again, this indicates that distressed firms hedge less.
The results mostly seem to confirm the hypotheses that more constrained firms hedge less, and the effect seems strongest for distressed firms. The results discredit previous theories of Froot et al. (1993), Tufano (1996), and others, explaining that managing risk adds value to a firm, and that small and financially constrained firms would therefore be more likely to hedge. It is in line with the empirical findings of Rampini et al. (2014) and supports the theory that the lack of collateral that constrained firms experience, along with the need to suffice investment needs and avoid downsizing, overrides the willingness to hedge risk (Bodnar et al., 2016: Rampini & Viswanathan, 2010; Rampini et al., 2014). Moreover, it supports the theory that hedging is only expected to be beneficial when a firm is able to credibly commit to its hedging strategy (Ligterink, 2001). When firms are distressed, the ability to credibly commit to their hedging strategy lowers, resulting in an incentive to reduce hedging.
Table 4
One-year prediction of net worth on hedging with derivatives: firm fixed effects
This table presents coefficient estimates of time series fixed effect ordinary least squares regressions relating the next year's net position in derivatives hedged scaled by total assets (bv) to measures of net worth in the current year. The dependent variable is the fraction of next year's net position in derivatives scaled by total assets (bv). Standard errors are in parenthesis. Coefficients that are statistically different from zero at the, 1%, 5% and 10% significance levels are denoted by ***, **, *, respectively.
The effect of growth opportunities on next year’s percentage hedged is shown to be only significant and negative in the last two columns at 1% and 10% significance levels respectively, confirming the expectation. It is arguable to what extent the result is reliable considering the insignificant results in the first two columns. Leverage shifts both in significance and sign between the regressions, so no real conclusion can be drawn on whether or not it confirms the expectation. It could possibly explain the variation of results found in previous studies (Guay & Kothari, 2003). Cash flow volatility seems to have a robust and significant negative effect on next year’s percentage hedged. This is not in line with the hypothesis, nor with the results from table 3. Cash to assets has a negative effect on hedging behavior, and is significant in most regressions. The negative effect is not in line with the hypothesis, it is however consistent with the results from table 3.
Net worth to assets (bv)
Net worth to
assets (mv) Credit rating
Credit rating dummies
(1) (2) (3) (4)
Measure of net worth 0.0691*** 0.0154*** -0.00291
(0.0180) (0.00163) (0.00178) Rating = BB+, BB, or BB- 0.000245 (0.00150) Rating = B+, B, or B- -0.00257** (0.00129) Rating = CCC+ or worse -0.0316*** (0.00201) Control variables Growth opportunities 0.00147 -0.00299 -0.0196*** -0.00769* (0.00150) (0.00234) (0.00437) (0.00402) Leverage 0.0418*** -0.00315* -0.0241*** -0.00323 (0.0114) (0.00185) (0.00417) (0.00357)
Cash flow volatility -0.104*** -0.154*** -0.146*** -0.217***
(0.0237) (0.0166) (0.0124) (0.0247) Cash to assets -0.00470 -0.0126** -0.0202*** -0.0153*** (0.00475) (0.00539) (0.00561) (0.00401) Constant -0.0278 0.0321*** 0.104*** 0.0520*** (0.0180) (0.00522) (0.0228) (0.00886) Number of observations 479 479 284 479 Adjusted R-squared 0.062 0.029 0.037 0.028
5.3. Financial constraints analysis
Table 5 shows the results of the test of the hypothesis that small firms that do decide to apply a hedging strategy will improve their financial situation. More specifically, it is expected that an increase in the amount hedged by derivatives leads to an increase in the credit rating of firms that do apply a hedging strategy. The tests contain time series probit regressions with standard errors clustered at industry level. The sample includes firms with ratings ranging from one step above to one step below the threshold rating. Due to a lack of observations the tests with threshold ratings BB+, BB, and BB- could not be properly conducted, and are therefore excluded from the table. The only statistically significant results are found in the sample with threshold rating B. The effect of an increase in the percentage hedged in the previous year is robust, statistically significant at the 1% significance level and negative. The estimate implies that the probability that firms that were previously rated, B+, B or B- have a current credit rating above B, decreased by hedging. The effect is not consistent with the hypothesis, nor is it with the theory stating risk management adds value to a firm (Froot et al., 1993).
In order to control for firms’ previous credit rating, the sample was changed, and the same hypothesis was tested. Table 6 shows the results of the tests on firms with credit ratings equal to the threshold in the previous year. The results of the regression with threshold rating B contradict the results of table 5. A possible explanation for this could be that in the analysis of table 5, no distinction was made in the firms’ financial situation with regards to the threshold rating at the time of hedging. Separating the sample based on firms’ previous credit rating, gives a more trustworthy insight into the movement of the credit ratings by movements in derivatives usage. The sample of firms with previous credit rating B shows a robust and statistically significant positive effect of the percentage hedged by derivatives in the previous year on the probability that the current credit rating has improved at the 1% significance level. This means that firms that have a credit rating B and increase their percentage hedged by derivatives are more likely to improve their credit rating than similar firms that do not. This is in line with the hypothesis that firms that hedge improve their credit rating, and therefore reduce their financial constraints. This is also in line with the general theory on corporate risk management stating that risk management adds value to a firm (Allayannis et al., 2012: Bodnar et al., 2016: Guay & Kothari, 2003: Froot et al., 1993), taking into consideration that a firm only implements its hedging strategy when the firm can credibly commit to its strategy. It is assumed that firms possess the knowledge that hedging is only beneficial when they are able to credibly commit to strategy (Ligterink, 2001).
The conclusion drawn from the results of the regression on B-rated firms is however contradicted by the results from the regression on the credit rating dummy with threshold B+. Table 6 shows a robust and statistically significant negative effect of an increase of the previous year’s percentage hedged. This means that firms with a previous credit rating of B+ lower their chances of improving their credit rating when they increase their percentage hedged. This is not in line with the hypothesis, and also not in line with the previously discussed results. One explanation could be that the opportunity costs of investment foregone are larger than the benefits the hedging strategy generates (Rampini & Viswanathan, 2010). It is however unlikely that this effect would be larger for firms rated B+, in contrast to firms rated B, as B-rated firms are supposedly more financially constrained. The dependence at firm level is not controlled for, because probit regression do not allow for this. The dependence at firm level creates biased results when not controlled for (Allayannis et al., 2012). The previously discussed self-selection problem that arises from firms self-selecting by basing their derivatives choice on expected benefits might cause the inconsistency in these results (Allayannis et al., 2012). Including control variables only partially solved the problem. Future research could use derivatives as a function of other variables that influence the amount hedged to reduce this bias to a larger extent (Allayannis et al., 2012).
Table 5
Hedging with derivatives and credit rating: probit analysis
This table presents estimates of time series probit regressions relating a firm's credit rating to last year's net position in derivatives hedged scaled by total assets (bv). The sample includes firms with last year’s credit rating ranging from one rating above till one rating below the corresponding threshold rating. The dependent variable is binary and represents the probability that a firm's current credit rating is above the corresponding threshold rating. Standard errors are in parenthesis. Coefficients that are statistically different from zero at the, 1%, 5% and 10% significance levels are denoted by ***, **, *, respectively.
Probability that rating is above: Rating B+ Rating B Rating B- Rating CCC+
(1) (2) (3) (4)
Explanatory variable
-148.9 -13.42*** -32.74 5.975
(184.4) -1.075 (203.7) (15.29)
Control variables
Cash flow volatility -74.42 -22.22*** 157.3 37.97**
(577.4) -5.484 (730.2) (17.98)
Net worth to assets (bv) -3.212 5.910*** 24.76 4.628***
(50.24) -1.335 (116.8) -1.719
Constant -5.426 -4.902*** -10.38 -2.076
(73.58) -1.057 (46.23) -1.599
Chi-squared statistic 0.91 1245.36*** 1.96 (.)
Number of observations 227 248 184 41
Last year's net position in derivatives to total assets (bv)
