1
Speed of Leverage Adjustment: Multinationality
and Adjustment Costs
Erik Snijders | S3540375
Supervisor: Dr Peter Smid
Master Thesis
International Financial Management, MSc
2
Executive summary
Increasing global diversification has caused multinationality to become less of an option, and more of a necessity for many expanding companies. The revelations of speed of leverage adjustment (SOA) research and necessity of data on multinationality motivated the research question, “What is the effect of multinationality on the Speed of Leverage Adjustment (SOA)?”. Further research was conducted to uncover the interactions of earnings fluctuations and creditor rights in relation to the aforementioned relationship.
3 flexibility in an external market by removing some of the adjustment costs, so as to increase the speed at which firms adjust leverage. The two aforementioned hypotheses were viewed in consideration of the main independent variable, multinationality, found to have conflicting effects on speed of leverage adjustment.
These three hypotheses were measured using fixed-effects OLS regression, with a broad, unbalanced, panel dataset consisting of 51,463 observations. The results showed lacking evidence of a relationship between multinationality and SOA. Earnings deviations, however, was found to have a significant positive effect on SOA, attenuated by multinationality. Creditor rights was found to have a significant negative impact on SOA, also attenuated by multinationality.
4
Abstract
5
1. Introduction
The capital structure puzzle is becoming more complex with every effort made to solve it. In their attempt at understanding the differential leverage adjustment capabilities of Multinational Corporations (MNCs) and Domestic Corporations (DCs), Park et al., (2013) find that leverage differentials between DCs and MNCs are not significant if you control for factors that commonly differentiate MNCs from DCs. However, the authors also find that the speed at which MNCs adjust their leverage is faster than their domestic counterparts in non-US countries. Another study on the relationship of multinationality and speed of leverage adjustment found contradicting evidence, implying a potential negative relationship (McMillan & Camara, 2012). Efforts to answer this, and other capital structure related questions, have been made by ways of one of two theories: the trade-off theory, and the pecking-order theory pioneered by two seminal articles (Miller, 1977; Modigliani & Miller, 1958). The authors call for further investigation into this point of contention: What is the effect of multinationality on the Speed of Leverage Adjustment (SOA)? The answer to this question will be made more apparent by asking some related questions, which this thesis will cover extensively: What are the effects of earnings fluctuations and creditor rights on SOA, and its relationship with multinationality? Answering these questions will solve another piece of the capital structure puzzle; financial managers will benefit from a deeper understanding of their capital structure, and the external factors that influence it. Furthermore, the scarcity of research on the effects of multinationality has created an increasingly widening gap, as globalization continues to force firms to face internationalization.
6 review. Speed of leverage adjustment and multinationality are measured and regressed using OLS regression.
The results present conflicting implications. There is lacking evidence to support the theory that multinationality has a positive effect on SOA, likely as a result of their differential ability to overcome financial constraints and increased financing flexibility. A clear causal relationship could not be established. Additional models regress a cash flow fluctuation variable (proposed by Faulkender, Flannery, Hankins, & Smith (2012)), and a creditor rights variable, to uncover more information about the multifaceted effects of multinationality on capital structure decisions. Earnings fluctuations has a significant and positive effect on SOA, which supports the conclusions drawn by Faulkender et al. (2012), and the dynamic trade-off theory described in the literature review. However, the variable’s mediating effects on the relationship between multinationality and SOA goes on to contradict the trade-off theory. The final hypothesis concerns creditor rights. Literature predicts creditor rights induce flexibility in an external financial market, while the results point towards an alternative theory of creditor rights, which postulates that creditor rights instead make firms more weary of debt financing. Taken in aggregate, these results have important implications for the Trade-Off Theory, creditor rights policies, and multinationality and capital structure research in general.
7
2. Literature Review
Research on speed of leverage adjustment has largely focused on the trade-off theory of capital structure. To develop a comprehensive understanding of the factors involved in determining speed of leverage adjustment, the following text will outline a brief summary of its origins.
It wasn’t until 1958 when the world’s understanding of capital structure graduated from infancy, when Modigliani and Miller (1958) released their seminal paper on the topic. The debate that ensued from their irrelevance proposition resulted in two main theories: The trade-off theory and the pecking-order theory, first put into contest by Myers (1984). The former argues managers will balance their debt and equity so that an equilibrium is found between the tax savings from debt and the present value of the deadweight costs of bankruptcy. The latter views capital structure differently, putting the focus not on a cost-benefit analysis, but instead arguing that managers make capital structure decisions on the basis of information asymmetry. In other words, Myers and Majluf (2011) argue that the decision to issue new shares as a means of financing over debt signals to investors that management considers their stock to be overvalued, resulting in a downward adjustment. In sum, the pecking-order theory considers adverse selection and agency costs, while the trade-off theory proposes a cost-benefit relationship in understanding capital structure decisions.
8 certain, bankruptcy is rare. He compares the trade-off theory with the “fabled horse-and-rabbit stew—one horse, one rabbit” (Miller, 1977, p. 264).
Indeed, Warner (1977) indicated in his research that the direct costs of bankruptcy are insignificant. For his example of the railroad industry, direct bankruptcy costs are valued at about a percent of the value of the firm. While this may be the underpinning of much of the contemporary literature on capital structure, even Warner warns his readers that the indirect costs of bankruptcy may be far more substantial, but simply too difficult to measure or observe. Miller argues that bankruptcy is rare, so whether direct or indirect, costs of bankruptcy are rare. He would be forgetting that indirect bankruptcy costs are made up of costs preceding even potential bankruptcies. In other words, indirect bankruptcy costs are not necessarily followed by bankruptcy. They are the costs of financial distress (e.g. forgone profits), which overlaps with the costs associated with adjusting leverage, adjustment costs. Since indirect bankruptcy costs are not necessarily followed by bankruptcy, a measurement of indirect bankruptcy costs that is ex-ante, must account for the probability of financial distress. This different point of view is provided by Molina, (2016), who argues that default potential is endogenous to the leverage decision. To compare the benefits of debt to the indirect costs of bankruptcy, the probability of default must be accounted for. Strikingly, Molina finds that the indirect costs of bankruptcy are comparable to the tax savings benefit of debt, increasing in significance at higher leverage levels. In sum, while direct bankruptcy costs are indeed insignificant, indirect bankruptcy costs are not.
9 and bankruptcy costs. Indeed, the motivation to tend to a target at a particular speed may be dependent on additional factors exceeding such limited considerations. Accordingly, the traditional trade-off theory has been dubbed the static trade-off theory. And while Miller’s argument may have carried some validity against the static version, dynamic versions considering a broader set of factors have proven their predictive power through extensive empirical evidence (e.g.: Cook & Tang, 2010; Flannery & Hankins, 2013; Flannery & Rangan, 2006).
The dynamic trade-off theory accepts the possibility that many factors may be influential in determining the capital structure, target capital structure, and the speed at which firms tend to their target. Such factors include taxes, transaction costs, direct and indirect bankruptcy costs, agency conflicts, information asymmetry (Eckbo, 2008). Especially transaction, or adjustment, costs are found to be significant in determining speed of leverage adjustment.
11 Hypothesis 1: Multinationality has an effect on the speed of bi-directional leverage adjustments.
While the importance of internal capital markets in capital structure decisions is manifest, it does not remedy all MNC capital structure ailments. As any other firm accessing external capital markets, MNCs will face market imperfections. Extant literature on the topic finds that, in accordance with the dynamic trade-off theory of capital structure, adjustment speeds are likely to be determined in part by an economically meaningful concept: adjustment costs (Faulkender et al., 2012). These findings may go a long way in explaining the differential adjustment speeds between MNCs and DCs in different countries found by Park et al. (2013). Indeed, Faulkender et al. (2012) find that firms that realize relatively large cashflows in absolute value (both positive and negative) observe more aggressive changes in capital structure towards their target leverage. Their reasoning is striking. The rationale posits that firms with large positive cashflows have the financial liquidity required to confront adjustment costs associated with leverage changes, whereas firms with large negative cashflows have reason to access external capital markets to cover their losses, permitting the corresponding adjustment costs to be shared among the desire of the firm to cover its losses and to reach its target leverage, increasing their motivation to do so.
12 by Faulkender et al. (2012). The term ‘bi-directional’ ought be reiterated here; the hypothesis drawn from Faulkender et al. (2012) entails a positive relationship between severe deviations from break-even cash flows and speed of leverage adjustment. Strikingly, similar bi-directional earnings fluctuations are observed in the case of the currency depreciations investigated by Faulkender et al. (2012). Indeed, they show that in the year of the depreciation event, sales remained stable for domestic firms while increasing for multinationals, indicating multinationals were able to exploit the situation to their benefit, using differential abilities to overcome financial constraints. In the year following the depreciation event, not only did sales increase even further for multinationals, sales decreased for domestic firms. The year after that, sales continued to decrease for domestic firms, while stabilizing for multinationals, indicating the disadvantage of domestic firms inflicts more lasting damage, while the advantage of multinationals is of a more temporary nature. The results also indicate that leverage levels increased for domestic firms in the years after the event, while remaining relatively stable for multinationals. Faulkender et al. argue this is likely due to the fact that multinationals have access to internal capital markets, while the domestic firms were forced to confront the external capital market. Furthermore, the effects of currency depreciation constitute earnings fluctuations, which were found to have a positive impact on the speed of leverage adjustment, substantiating the relationships hypothesized in here.
Hypothesis 2.1: Large deviations from expected performance has a positive effect on speed of leverage adjustment.
Hypothesis 2.2: Large deviations from break-even in a firm’s net cash flows accentuate the relationship between multinationality and speed of leverage adjustment.
13 structure (Warner (1977) provides excellent empirical evidence of the insignificance of direct bankruptcy costs). In their seminal paper contributing to this theoretical discussion, Haugen & Senbet (1978) argue for the insignificance of bankruptcy costs. They acknowledge the significance of the indirect costs of bankruptcy, often represented by a disruption of supplier-customer relationships (Haugen & Senbet, 1978), lost profits, and an inability to obtain credit (Warner, 1977). However, they also argue that “any costs associated with bankruptcy, or the transfer of ownership (from stockholders to creditors), must be limited to the lesser of (a) the cost of bankruptcy and (b) the cost of avoiding the transfer entirely” (Haugen & Senbet, 1978, p. 894). They go on to argue that the (indirect and direct) costs of bankruptcy are significant, but the cost of avoiding the transfer entirely is limited to the transaction costs associated with the share issuance and the repurchase (at fair market value) of the asset claims of the firm. Essentially, they argue that the costs of bankruptcy are limited to the probability of an actual transfer of control (something they call a ‘formal reorganization’ and characterize as being quite costly) The probability of this happening is quite low if you consider the firm’s ability to repurchase the claims on their assets. Their arguments do rest on the assumption that all actors involved act rationally, as well as not being substantiated by empirical evidence. Furthermore, these findings also contradict the empirical evidence of Molina (2005), who accounts for the endogeneity inherent in the relationship between leverage and default probabilities.
14 have much of an impact on the firm’s capital structure decisions. In reality, however, debt is in many cases not nearly as popular as the irrelevance assumption would advise (Haugen & Senbet, 1978; Miller, 1977; Molina, 2005; Warner, 1977).
This contradiction is resolved if we consider the supply-side view of the relationship between creditor rights and leverage. While the demand-side view argues firms’ costs of accessing capital markets go up, so their tendency to borrow must go down, the supply-side view argues that creditor rights protect creditors to the point that their tendency to lend goes up (Akbel & Schnitzer, 2011; Cho, El Ghoul, Guedhami, & Suh, 2014; Djankov, McLiesh, & Shleifer, 2007). In his empirical study, Ameer (2013) shows that strong creditor rights have a positive impact on speed of adjustment. When analyzing his findings, Ameer (2013) argues along the supply-side view. He argues that the borrowing capacity of firms becomes less constrained when accessing capital markets that are regulated, which supposedly results in curtailing non-performing loans, and increasing financing flexibility of firms accessing external capital markets.
While the assumption of irrelevance is not entirely reliable, strikingly, not only do Ameer’s findings not contradict the insignificance of bankruptcy costs, they actually support this view. Creditor rights may therefore serve as an indication of the flexibility of an external capital market. Similarly, firms in countries with low creditor rights are likely to face higher transaction costs when adjusting their leverage, resulting in the decreased speed of adjustment observed by Ameer (2013).
Hypothesis 3.1: Creditor rights has a positive effect on speed of leverage adjustment.
15 preferentially (Desai et al., 2004). Domestic firms are left without a choice, being forced to access domestic external financial markets, facing the full effect of poor creditor rights regulation. From previously reviewed literature one can deduce that lower leverage levels (see BDR in table 3), access to internal capital markets, and other ways to avoid domestic external markets (equity infusions, external capital markets in other countries, etc.), may allow multinationals to issue debt and equity more flexibly while facing lower costs, thereby expediating the process. As such, the expectation is that in the absence of creditor rights, multinationals will have the differential ability to overcome corresponding adjustment costs, compared to domestic companies.
Hypothesis 3.2: Creditor rights will attenuate the relationship between multinationality and speed of leverage adjustment.
3. Data Data Collection
16 of the dependent variable require at least two consecutive years of data, or the lagged variables would be invalid. Firm-year observations with missing data for any leverage factor are removed. Firm-year observations with negative values for book equity, debt, or liabilities have also been removed (Flannery & Hankins, 2013). This process resulted in an unbalanced sample of panel data comprised of 6,126 firms from 71 countries, amounting to a total of 51,463 firm-year observations. The sample is dominated by the United States (24%), Japan (16%), and Great Britain (8%). More information on the sample, including distributions according to multinationality, may be found in table 1.
The regressions are executed in three separate models, one for each hypothesis. They are ran using Fixed-Effects Ordinary-Least-Square regression in STATA. All firm-specific variables have been winsorized at 1%, in order to reduce the effect of outliers on the final outcome of the regression.
Table 1 – Industry distribution
The table reports the number of observations that comprises the sample, organized by industry. The first and second column shows industry code and name classification according the NACE Rev. 2. MNC20 is the multinationality dummy variable that takes on the value 1 when international sales make up 20% or more of a firm’s total sales. The remainder make up DC, domestic companies.
Code Industry Name MNC DC Total
A Agriculture, forestry and fishing 290 250 540
B Mining and quarrying 1,185 951 2,136
C Manufacturing 13,324 16,358 29,682
E Water supply; sewerage, waste management and remediation activities 39 30 69
F Construction 1,452 589 2,041
G Wholesale and retail trade; repair of motor vehicles and motorcycles 3,168 1,412 4,580
H Transportation and storage 1,383 811 2,194
I Accommodation and food service activities 827 305 1,132
J Information and communication 2,726 2,201 4,927
M Professional, scientific and technical activities 664 934 1,598 N Administrative and support service activities 666 573 1,239
P Education 63 27 90
Q Human health and social work activities 488 48 536
R Arts, entertainment and recreation 465 85 550
S Other service activities 125 24 149
17 Country specific variables such as creditor rights and GDP-growth are assumed not to have significant outliers.
4. Empirical Methodology
The culmination of the preceding hypotheses and corresponding variables lead to a unified model, where the identified determinants of speed of leverage adjustment are tested for their explanatory powers. The full model in its mathematical form is calculated as
𝑆𝑂𝐴𝑖,𝑡 = 𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡 + 𝑀𝑁𝐶𝑖,𝑡+ 𝐶𝐹𝑑𝑒𝑣𝑖,𝑡 + 𝐷𝐹𝑑𝑒𝑣𝑖,𝑡 ∗ 𝑀𝑁𝐶𝑖,𝑡+ 𝐶𝑅𝑖𝑔ℎ𝑡𝑠𝑖
+ 𝐶𝑅𝑖𝑔ℎ𝑡𝑠𝑖 ∗ 𝑀𝑁𝐶𝑖,𝑡+ 𝑐𝑜𝑛𝑡𝑟𝑜𝑙𝑠𝑖,𝑡+ 𝑒𝑖,𝑡. (1) Here, SOA is speed of leverage adjustment, regressed against a constant, three main independent variables that resulted from the literature review, their corresponding interaction variables, the controls, and the error term. The main relationship comprises the speed of leverage adjustment (the dependent variable) and multinationality (the independent variable). The measurement of the speed of leverage adjustment relies on a special technique outlined by (Flannery & Hankins, 2013), and fundamental leverage data. The measurement of the dependent variable will be covered in the next section.
Multinationality
18 Cash flow deviation
The second hypothesis seeks to analyze the impact of cash flow deviations on capital structure decisions (depending on multinationality). As the hypothesis touches on an ex-ante topic, the measurement of the deviation ought to rely on an ex-ante measure of cash flows as well. Therefore, the deviation will be based on the degree to which the net cash flows deviate from the expected results. Expected performance is measured as the average net cash flows of the past three years. Should any firm-year observation part of said measurement be missing, the observation is omitted. The deviation of net cash flows from this expected result will be measured. Research indicates that poor financial performance raises the need for external financing, while good financial performance grants firms the financial leeway to face adjustment costs and move closer to their target leverage. Therefore, the resulting data will be transformed to represent the absolute value, as both negative as well as positive deviations are expected to have the same effect on the main relationship. To reiterate, this is measured in absolute value because the literature indicates that a positive deviation may give a firm the financial liquidity required to face adjustment costs, while a negative deviation may give a firm the motivation it needs to optimize their leverage and face adjustment costs, despite illiquidity. Correspondingly, a break-even with expected results would encourage passivity. Finally, to ensure the data is represented relative to the size of the firm, the variable is scaled by total sales.
While this relationship has been analyzed extensively, and its impact on SOA has become increasingly clear, the novel aspect of this model is that it assumes 𝐶𝐹𝑑𝑒𝑣𝑖,𝑡 has a moderating
19 differential ability to overcome financial constraints, due to a fundamental difference in how they interact to external financial markets.
Creditor Rights
The final hypothesis constitutes another moderating variable; the 𝐶𝑅𝑖𝑔ℎ𝑡𝑠𝑖 variable, which measures the extent to which the country in which the company 𝑖 is located protects creditors by ways of imposing laws.
20 The diverging nature of the discussion on the topic of creditor rights and their impact on capital structure has been discussed extensively. Generally, the supply-side view trumps the demand-side view empirically, yet how it relates to speed of leverage adjustment is unclear. As with the second hypothesis, this variable is interacted with 𝑀𝑁𝐶𝑖,𝑡 to investigate its moderating effect on the
relationship between 𝑀𝑁𝐶𝑖,𝑡 and SOA.
Control Variables
Some econometric issues can be predicted and controlled for using certain observable control variables. Most prominently, Rygh & Benito (2018) point out that intangibility of assets is an important control for any research including leverage as a variable. They argue that debt financing
Table 2
Description of variables and source of associated data. Independent
Variable
Data Source
Market Debt Ratio Book value of interest-bearing debt (used to compute MDR and SOA, outlined in the equations below)
Datastream
Market Debt Ratio Shares outstanding (used to compute market value of equity)
Compustat 199
Market Debt Ratio Price per share (used to compute market value of equity) Datastream Multinationality International sales as a percentage of total sales Datastream Cash Flow Deviation CFdev (or Cash Flow Deviation) relates to the extent to
which the results deviated from the expected performance. This includes both a positive as well as a negative deviation. It is calculated as the sum of Operating,
Financial, and Investment Cash Flows. Expected cash
flow is calculated as the average net cash flows of the past three years. Missing values are not accepted into the calculation and are removed from the variable. The variable is made relative to firm size by scaling it to total
sales.
Datastream
Creditor Rights Four dimensions of creditor rights have been identified (La Porta et al., 1998). For each that is present in a country, the variable increases by 1, resulting in a range of 0-4.
(La Porta et al., 1998) Dartmouth dataset
Control variable
21 frequently requires collateral in ways of tangible assets. For firms with predominantly intangible assets, either higher adjustment costs would be incurred for debt financing, or they would omit debt altogether and opt for equity. This necessitates the inclusion of a variable controlling for intangibility. Intangible assets is the only variable that will be scaled by total assets as opposed to total sales. This is because an intangibility ratio is more sensible than a ratio of intangible assets to total sales. In order to isolate the effects of multinationality on speed of leverage adjustment, additional controls common in literature will be considered. These include firm-specific variables that may affect capital structure decisions in a firm. As such, the following variables will be controlled for: firm size (proxied by total net sales), EBITDA scaled by total sales (Park et al., 2013). Cook & Tang (2010) also argue that any analysis including leverage should include a macro-economic indicator. They assert that the (static) trade-off theory predicts that leverage is determined by tax-savings and bankruptcy cost considerations, both of which are influenced by the macro-economic conditions. As such, GDP-growth will be added as a control. GDP data has been accessed through the OECD library. Finally, industry dummies will be calculated to control for industry-specific variations.
Speed of Adjustment
The leverage measure used throughout this analysis is the market debt ratio. The book ratio is omitted as it is considered to lack significance (Fama & French, 2002; Flannery & Rangan, 2006). The market debt ratio is defined as
𝑀𝐷𝑅𝑖,𝑡 = 𝐷𝑖,𝑡
22 where Di,t denotes the book value of firm i’s interest-bearing debt at time t, Si,t denotes the number of common shares outstanding at time t, and Pi,t denotes the price per share at time t. (Flannery & Rangan, 2006)
The second component of the partial adjustment model is the target leverage measure. As cross-sectional variation is likely to occur, most contemporary literature advocates a firm-specific, time-varying, target leverage measure (Eckbo, 2008; Flannery & Rangan, 2006). Accordingly, the target will be measured as the 3-year average MDR of the firm. Longer averages would suffer under data unavailability, shorter would lack representativeness.
The final component of the partial adjustment model is the measurement of the variable of interest: speed of adjustment. For the purpose of this research, the traditional partial adjustment model will be altered somewhat: a firm-specific, time-varying measure for speed of adjustment will be formulated using the model postulated by Flannery & Rangan (2006), calculated as
𝑀𝐷𝑅𝑖,𝑡+1− 𝑀𝐷𝑅𝑖,𝑡 = 𝑆𝑂𝐴𝑖,𝑡(𝑀𝐷𝑅𝑖,𝑡+1∗ − 𝑀𝐷𝑅𝑖,𝑡), (3)
which may be rearranged to
𝑆𝑂𝐴𝑖,𝑡 =
𝑀𝐷𝑅𝑖,𝑡+1− 𝑀𝐷𝑅𝑖,𝑡 𝑀𝐷𝑅𝑖,𝑡+1∗ − 𝑀𝐷𝑅𝑖,𝑡
. (4)
23
5. Results Summary Statistics
24 The remaining variables are more in line with predictions. The cashflow deviation variable measures the extent to which the net cash flows deviated from the expected performance, scaled to total sales.
The range is -1.1245 to 1.2692 in both cases (again, due to winsorization). The minimum (maximum) indicates that negative (positive) cashflows never exceeded 112% (127%) of the company’s total net sales in that year. On average, the accumulated annual net cashflows of the MNCs in the sample added up to about .76% of the firm’s total revenue, and 1.39% in the case of domestic companies. The mean indicates the average firm in the sample was profitable, albeit moderately. The location of the mean, size of the sample, and moderate SD, indicates the variable comprises a normal distribution with long tails. CF Dev (abs) presents the absolute value of the same variable. This is how it will be taken into the regression. The mean is clearly far closer to the min compared to the max, indicating that the distribution is positively skewed; it has a long tail on the right side of the median. This is to be expected of a distribution of a variable whose tails have been moved to one side as a result of the absolute value transformation. One can infer that the
Table 3 – Summary Statistics
MDR represents the Market Debt Ratio and is calculated as indicated by equation 1. BDR is the book debt ratio and is calculated as equation 1, except the market value of equity is substituted by the book value of equity, downloaded from Datastream. SOA represents speed of leverage adjustment and is calculated as indicated by equation 2 and 3. CF Dev is calculated as indicated by the entry in table 2. CF Dev (abs) shows the absolute value of the same variable (this is the variable that will be used in the regression) C. Rights is the creditor rights variable, refer to table 2 for specifics on the calculation. Intangible is the intangible assets variable, scaled by total assets. EBITDA is the performance variable, scaled by total sales. Size is proxied by total sales, presented in millions. GDP-G is the GDP growth rate variable.
MNC DC Total
Obs Obs Mean SD Min Max Obs Mean SD Min Max
25 majority of the firm-year observations congregate around the mean, while outliers make up the upper quartiles of the distribution.
The mean of the creditor rights variable implies that on average, the countries represented in the sample are relatively protective of their creditors, having at least one creditor right. The similarity between the two means of MNCs and DCs hints at similarities in the countries represented between both types of firms, which is fortunate.
26 growing faster. This is not entirely surprising, as emerging economies with fast growth rates are more likely to be home to firms that have yet to tap into international markets.
Summary Statistics – Speed of Adjustment
The dependent variable is quite troubling, it will be analyzed separately in this section, with corresponding summary statistics shown in table 4. The dependent variable speed of adjustment (SOA) shows a plausible mean, but a range that seems unlikely. The mean indicates that on average, firms tend to close the gap between current and target leverage by about 58% to 57%, for multinationals and domestic firms respectively. This does not support the scarce literature on the topic (McMillan & Camara, 2012), which indicates that domestic firms have an edge of multinational firms in terms of speed of leverage adjustment. Winsorization has caused the tails of the distribution to become equal. Most disturbingly, the range of SOA is very large. The minimum of -4.35 indicates that some firms may adjust their leverage away from the target, as opposed to tending toward it. Essentially, those firms would be extending the gap 4.35-fold. Similarly, the max of 6.23 is an indication of firms that do in fact tend towards the target leverage, but ‘overshoot’ by a factor of 6.23. This sounds less absurd in the example of a firm being very close to the optimal leverage level, and for any particular reason, it decides to move away from, or beyond it somewhat. Relative to the size of the gap, the SOA variable would measure a significant deviation from the target.
27
Table 4 – Summary Statistics (Dependent Variable)
This table shows the summary statistics for the SOA variable. SOA represents speed of leverage adjustment and is calculated as indicated by equation 3 and 4.
MNC Obs Mean SD Min Max
SOA 19,868 0.5788 2.0046 -4.3557 6.2303
DC
SOA 17,694 0.5667 2.0525 -4.3557 6.2303
28 The issue with the methods proposed by the literature is that it artificially increases the SOA depending on the proximity of the current leverage level to the target. Equation 4 is altered to reconcile, so that
𝑆𝑂𝐴𝑖,𝑡 = 𝐷𝑖,𝑡+1− 𝐷𝑖,𝑡
𝐷𝑖,𝑡+ 𝑆𝑖,𝑡𝑃𝑖,𝑡. (5)
Essentially, the change in leverage, as calculated according to the MDR equation, is the SOA. This way, SOA isn’t measured relative to the target, but to the capital structure itself. Imagine a firm with a market debt ratio of 20%, and a target of 21%. Next year its leverage is 30%. According to equation 4, its SOA is 10, according to equation 5, it is 0.5. 50% is a more realistic representation of speed of leverage adjustment, since the firm did in fact take one year to increase its leverage with 50%, not 1000%.
29 regardless of target - a measure where raw all values are made positive - is included in Appendix A.
Table 5 presents a clear picture of how the distribution has changed using the new method. Most prominently, a very high kurtosis indicates that the distribution is very steep. This leptokurtic distribution is a sign that the severe variance of the previous measure has disappeared (reflected by the accordant change in the SD), and the vast majority of firms only adjust their leverage incrementally, as in accordance with literature. The mean indicates the average firm (both MNC and DC) adjust their leverage at about 18.5%-points per year. The max of 180% can only be explained by rare occasions where a firm goes from a very low debt position to taking on a relatively large amount of debt, in the same year where their firm’s market value was quite low. The skewness has changed rather dramatically as well. It has remained positively skewed, meaning a long tail exists on the right side of the mean. This tail has increased in length, amplified by the increased compression of the interquartile range. Remember, a positive value for the SOA variable indicates that the firm has adjusted towards the target. The fact that large values are only common on the positive side of the distribution is expected, because it implies that aggressive adjustments are more commonly in the direction of the target. Considering the SD is 37%, one could infer that
Table 5 – Summary Statistics (Dependent Variable)
This table shows the summary statistics for the two versions of the SOA variable. SOA represents speed of leverage adjustment. SOA (old) is calculated as indicated by equation 3 and 4. SOA (new) is calculated as indicated by equation 5.
MNC Obs. Mean SD Min Max Skewness Kurtosis
SOA (old) 19,868 0.5788 2.0046 -4.3557 6.2303 .5799 5.1792 SOA (new) 24,598 0.1856 0.3682 -0.3193 1.8089 2.5870 10.7058
DC
30 firms may tend to ‘overshoot’ their leverage target by closing the gap and going beyond the target. For further analysis, we move on to the regressions.
This measure of SOA, for the purposes of this thesis, is considered superior to the SOA (old) measurement. SOA (old) is an adjustment to the original measurement (Flannery & Hankins, 2013). This adjustment clearly amplified the econometric issues inherent in the method, as indicated by the absurd summary statistics, and regressions results (Appendix B).
Regression Results
31 The second hypothesis, shown in table 7, concerns the impact of the extent to which a firm’s net CF deviated from expected CF on the relationship between multinationality and speed of leverage adjustment. Both the multinationality and CFdev variable, as well as the interaction term between the two are included into the OLS regression. Firstly, the second model presents a slightly more sizable coefficient at the same significance for the MNC variable. Meanwhile, the CFdev measures in at a sizable, positive, and very significant coefficient. The interaction term on the other hand is only slightly significant, and negative. The significance from the size control has also been subsumed by the newly added variables. These results point at several striking relationships. Firstly, there seems to be a direct and positive relationship between CFdev and SOA. In other words, the more a firm deviates from their expected performance, the higher their propensity to adjust leverage towards the target.
Table 6 – Regression Results (Primary Relationship)
This table presents the regression results of the main relationship. MNC represents the multinationality variable; a dummy that takes on the value of 1 at international sales > 20%. LN_TA, LN_IA, and LN_EBITDA represent the natural logarithm of total assets, intangible assets, and EBITDA, respectively. GDPgrowth represents the GDP-growth variable, in percentages.
Variable SOA SOA
MNC 0.0056* 0.0070* (0.033) (0.0037) LN_TA 0.0035** (0.0009) LN_IA 0.0000 (0.0000) LN_EBITDA 0.0000 (0.0001) GDPgrowth 0.0035*** (0.0006) Constant 0.1924*** 0.1167*** (0.0159) (0.0252)
Industry fixed effects Yes Yes
Adj-R2 0.0046 0.0054
Observations 51,463 47,166
32 CFdev has been scaled to net sales, so should a firm deviate from their expected performance, measured as the average net CF of the past three years, by an amount equal to 40% (CFdev (abs) mean + 1 SD; see table 3) of their net sales, the speed at which said firm would adjust towards their leverage would increase with 1.7%-points. Hypothesis 2.1 is supported by these results. Secondly, the interaction term points to the existence of a mediating relationship where the effect of multinationality through CFdev is negative. Since CFdev and MNC have a positive effect on SOA when isolated, the interaction works as a counterweight. A multinational that deviates from
Table 7 – Regression Results
This table presents the regression results of all three hypotheses. MNC represents the multinationality variable; a dummy that takes on the value of 1 at international sales > 20%. MNC * CFdev represents the Cash Flow deviation variable, the independent variable for the second hypothesis. MNC * Crights represents the Creditor Rights variable, the independent variable of the third and final hypothesis. Size is proxied by total net sales, to represent size. Intangible Assets is the total intangible assets scaled by the size variable. EBITDA is the performance proxy, scaled by the size variable. GDP-growth represents the GDP growth of the country in which the firm is incorporated, in percentages.
Variable SOA SOA SOA
MNC 0.0070* 0.0086* -0.0050 (0.0037) (0.0049) (0.0073) CFdev 0.0427*** (0.0128) MNC * CFdev -0.0328* (0.0193) Crights -0.0079*** (0.0027) MNC * Crights 0.0066* (0.0034) Size 0.0035** 0.0001 0.0033*** (0.0009) (0.0011) (0.0009) Intangible Assets 0.0000 0.0000 0.0000 (0.0000) (0.0001) (0.0000) EBITDA 0.0000 0.0002 0.0000 (0.0001) (0.0002) (0.0001) GDP-growth 0.0035*** 0.0017** 0.0036*** (0.0006) (0.0007) (0.0006) _cons 0.1167*** 0.1948*** 0.1288*** (0.0252) (0.0313) (0.0264)
Industry fixed effects Yes Yes Yes
Adj-R2 0.0054 0.0047 0.0055
Observations 47,166 33,739 47,080
33 their expected performance with 40% will move away from their target 1.3%-points more quickly than a domestic firm. In conjunction, these effects cancel each other out for a large part, in the case of multinationals. This means a domestic corporation will speed up their leverage adjustment toward their target faster than a multinational when their deviation from expected performance is high. This goes directly against the expectations from the second hypothesis. Therefore, hypothesis 2.2 is not supported. To substantiate this conclusion, Appendix A presents the regression results for the absolute value of SOA. In other words, the coefficients represent the actual speed of leverage adjustment, regardless of whether or not the adjustment is in the direction of the target. Even here we see similar results. MNC and CFdev isolated are significant and speed up leverage adjustments with a rather sizable effect. The interaction term is negative, but also insignificant. The conclusions that hypothesis 2.1 is supported and hypothesis 2.2 is not, is robust to the alternative measurement of SOA.
34 on speed of adjustment are somewhat negated if the firm in question is internationally active. Besides the isolated effects of MNC and Crights, a multinational will increase their speed of adjustment towards the target .7% more quickly per creditor right. The interaction term shows the likelihood of multinationals’ differential ability to overcome financial constraints such as creditor rights. However, it should be noted that the coefficient is only significant to the 10% level. Furthermore, hypothesis 3.2 predicts an attenuation of the relationship between multinationality and SOA by creditor rights. While that is in fact what may be observed from the results, the main relationship is insignificant and negative. Hypothesis 3.2 cannot in good faith be supported.
6. Discussion and Concluding Remarks
While the isolated relationships described above alone are quite interesting, the overarching picture presents information relevant to the dynamic trade-off theory around which the concepts described in the literature review revolve. The dynamic trade-off theory was a development of the static trade-off theory which was predicated on the assumption that costs of bankruptcy and tax benefits were the two main factors impacting capital structure decisions. The dynamic trade-off theory postulated an expanded view of the field, introducing a variety of factors that could have an equally, if not more, significant impact on such decisions. For researchers, the question then became how one could prove that concept empirically. The theory predicts that factors beyond those comprising the static trade-off theory would have an impact. Factors such as adjustment costs. The purpose of this thesis is to shed light on that concept and try to support or disprove these predictions.
35 there is truth to that theory, adjustment costs would have to be a significant factor, and for that to be true, multinationality would have a significant and positive impact on speed of leverage adjustment. While there is some significance, results are not significant enough to conservatively state that this is the case. Therefore, a causal link between multinationality and speed of leverage adjustment is far from being established. For instance, the interaction terms of the second and third models of the regressions (table 7) show that the effects of multinationality may be more ambiguous than previously thought. A conclusion supporting a significant relationship between multinationality and speed of leverage adjustment cannot be made.
36 the data, domestic firms may in fact speed up their leverage adjustment at a faster rate than multinationals, when facing deviations from the expected cash flow. This may mean that adjustment costs may be more easily mitigated by domestic firms than multinationals, which would go against everything we know about multinationals. An alternative explanation could be that earnings deviations are perceived as being more serious for domestic firms than for multinationals, as multinationals are frequently larger and more complex, and have internal capital markets to fall back on. In other words, perhaps adjustment costs are simply not significant enough for multinational firms experiencing earnings volatility to become motivated enough to face said costs.
37 but also have important implications for the trade-off theory. The dynamic, but especially, the static trade-off theory postulate that bankruptcy costs are a significant determinants of capital structure ratios. The results of the final hypothesis support this view and implies that bankruptcy costs may be less irrelevant than some have argued.
Future Research
The empirical methodologies applied to speed of leverage adjustment research have evolved tremendously over the past decade, and strides are being made in verifying the reliability of the best methods. Flannery & Hankins (2013) outline excellent models. The preliminary SOA measurement, results of which are available in appendix B, was based on those models. The adjustments made to it were, however, not reliable. For this reason, the alternative SOA measurement was devised. Future research would benefit from applying the complete and proven method.
38 Future researchers would benefit from researching relationships that include any of the main variables included in this thesis. However, I advise future researchers to be weary of the main relationship in this thesis. The relationship between multinationality and SOA suffers under the econometric difficulties of measuring SOA and interpreting multinationality. Furthermore, there is a very real possibility that the relationship between the two main variables is in fact insignificant enough to constitute a non-causal relationship.
39
Bibliography
Akbel, B., & Schnitzer, M. (2011). Creditor rights and debt allocation within multinationals.
Journal of Banking and Finance, 35(6), 1367–1379.
https://doi.org/10.1016/j.jbankfin.2010.10.015
Ameer, R. (2013). Financial liberalization and firms’ capital structure adjustments evidence from Southeast Asia and South America. Journal of Economics and Finance, 37(1), 1–32. https://doi.org/10.1007/s12197-010-9158-3
Burgman, T. (1996). An Empirical Examination of Multinational Corporate Capital Structure, 27(3), 553–570. Retrieved from http://www.jstor.org/stable/155438
Cho, S. S., El Ghoul, S., Guedhami, O., & Suh, J. (2014). Creditor rights and capital structure: Evidence from international data. Journal of Corporate Finance, 25, 40–60. https://doi.org/10.1016/j.jcorpfin.2013.10.007
Çolak, G., Gungoraydinoglu, A., & Öztekin, Ö. (2018). Global leverage adjustments, uncertainty, and country institutional strength. Journal of Financial Intermediation, 35(February), 41–56. https://doi.org/10.1016/j.jfi.2018.01.010
Cook, D. O., & Tang, T. (2010). Macroeconomic conditions and capital structure adjustment
speed. Journal of Corporate Finance, 16(1), 73–87.
https://doi.org/10.1016/j.jcorpfin.2009.02.003
Creal, D., Robinson, L. A., Rogers, J. L., & Zechman, S. L. C. (2011). The Multinational Advantage. Ssrn, (11). https://doi.org/10.2139/ssrn.1933777
40 Desai, M. A., Foley, C. F., & Hines Jr., J. R. (2004). A Multinational Perspective on Capital
Structure Choice and Internal Capital Markets. The Journal of Finance, 59(6), 2451–2487. Djankov, S., McLiesh, C., & Shleifer, A. (2007). Private credit in 129 countries. Journal of
Financial Economics, 84(2), 299–329. https://doi.org/10.1016/j.jfineco.2006.03.004
Eckbo, B. E. (2008). Handbook of Corporate Finance: Empirical Corporate Finance Vol. 2. North Holland.
Fama, E. F., & French, K. R. (2002). Fama-2002-Testing Trade‐Off an.pdf, 15(1), 1–33. https://doi.org/10.2307/2696797
Faulkender, M., Flannery, M. J., Hankins, K. W., & Smith, J. M. (2012). Cash flows and leverage adjustments. Journal of Financial Economics, 103(3), 632–646. https://doi.org/10.1016/j.jfineco.2011.10.013
Flannery, M. J., & Hankins, K. W. (2013). Estimating dynamic panel models in corporate finance. Journal of Corporate Finance, 19(1), 1–19. https://doi.org/10.1016/j.jcorpfin.2012.09.004 Flannery, M. J., & Rangan, K. P. (2006). Partial adjustment toward target capital structures, 79,
469–506. https://doi.org/10.1016/j.jfineco.2005.03.004
Gande, A., Schenzler, C., & Senbet, L. W. (2009). Valuation effects of global diversification. Journal of International Business Studies, 40(9), 1515–1532. https://doi.org/10.1057/jibs.2009.59
Gao, W., & Chou, J. (2015). Innovation efficiency, global diversification, and firm value. Journal
of Corporate Finance, 30(September), 278–298.
41 Gao, W., Ng, L., & Wang, Q. (2008). Does geographic dispersion affect firm valuation? Journal
of Corporate Finance, 14(5), 674–687. https://doi.org/10.1016/j.jcorpfin.2008.08.004
Haugen, R. A., & Senbet, L. W. (1978). the Insignificance of Bankruptcy Costs To the Theory of Optimal Capital Structure. The Journal of Finance, 33(2), 383–393. https://doi.org/10.1111/j.1540-6261.1978.tb04855.x
La Porta, R., Lopez-de-Silanes, F., Shleifer, A., & Vishny, R. W. (1998). Law and Finance. Journal OfPolitical Economy, 106(6), 1113–1155.
Mansi, A. S., & Reeb, M. D. (2002). Corporate international activity and debt financing. Journal of International Business Studies, 33(1), 129. Retrieved from http://proquest.umi.com/pqdweb?did=112180161&Fmt=7&clientId=44986&RQT=309&V Name=PQD
McMillan, D. G., & Camara, O. (2012). Dynamic capital structure adjustment: US MNCs & DCs. Journal of Multinational Financial Management, 22(5), 278–301. https://doi.org/10.1016/j.mulfin.2012.10.001
Miller. (1977). Debt and Taxes. The Journal of Finance, 40(3), 637–657. https://doi.org/10.1111/j.1540-6261.1985.tb04986.x
Modigliani, F., & Miller, M. H. (1958). The Cost of Capital, Corporation Finance and the Theory of Investment. British Medical Journal, 48(3), 261–297. https://doi.org/10.1136/bmj.2.3594.952
42 Myers, S. C. (1984). The Capital Structure Puzzle, 39(3).
https://doi.org/10.1111/j.1540-6261.1984.tb03646.x
Myers, S. C., & Majluf, N. S. (2011). Brealey, Richard A., Stewart C. Myers, and Franklin Allen. Principles of Corporate Finance. New York: McGraw-Hill/Irwin, 2011. Print., 13(2), 187– 221. https://doi.org/10.1016/0304-405X(84)90023-0
Öztekin, Ö., & Flannery, M. J. (2012). Institutional determinants of capital structure adjustment speeds. Journal of Financial Economics, 103(1), 88–112. https://doi.org/10.1016/j.jfineco.2011.08.014
Park, S. H., Suh, J., & Yeung, B. (2013). Do multinational and domestic corporations differ in their leverage policies? Journal of Corporate Finance, 20(1), 115–139. https://doi.org/10.1016/j.jcorpfin.2012.08.001
Rygh, A., & Benito, G. R. G. (2018). Capital Structure of Foreign Direct Investments: A Transaction Cost Analysis. Management International Review, 58(3), 389–411. https://doi.org/10.1007/s11575-017-0335-x
Shen, C. H. han. (2014). Pecking order, access to public debt market, and information asymmetry. International Review of Economics and Finance, 29, 291–306. https://doi.org/10.1016/j.iref.2013.06.002
43
Appendix A
Table 7 – Regression Results (Alternative Measurement)
This table presents the regression results of all three hypotheses. The measurement of the dependent variable is the same as the one used in the main body, as per equation 5, except this measure does not alter the sign to indicate the direction relative to the target, but instead provides the absolute value, to give an indication of sheer speed. MNC represents the multinationality variable; a dummy that takes on the value of 1 at international sales > 20%. MNC *
CFdev represents the Cash Flow deviation variable, the independent variable for the second hypothesis. MNC * Crights represents the Creditor Rights variable, the independent variable of the third and final hypothesis. Size is
proxied by total net sales, to represent size. Intangible Assets is the total intangible assets scaled by the size variable. EBITDA is the performance proxy, scaled by the size variable. GDP-growth represents the GDP growth of the country in which the firm is incorporated, in percentages.
Variable SOA SOA SOA
MNC 0.0425* 0.0544** 0.0299 (0.0235) (0.0273) (0.0466) CFdev 0.1589** (0.0706) MNC * CFdev -0.1154 (0.1068) Crights 0.0150 (0.0177) MNC * Crights 0.0062 (0.0221) Size -0.0137** -0.0102 -0.0121** (0.0061) (0.0063) (0.0061) Intangible Assets 0.0000 0.0000 0.0000 (0.0003) (0.0003) (0.0003) EBITDA 0.0006 0.0008 0.0006 (0.0008) (0.0008) (0.0008) GDP-growth -0.0016 -0.0010 -0.0016 (0.0038) (0.0039) (0.0038) _cons 0.9241*** 0.8333*** 0.8193*** (0.1659) (0.1725) (0.1745)
Industry fixed effects Yes Yes Yes
Adj-R2 0.0009 0.0010 0.0009
Observations 34,773 33,739 34,712
44
Appendix B
Table 7 – Regression Results (Preliminary Measurement)
This table presents the regression results of all three hypotheses. The measurement of the dependent variable is presented in equation 4. MNC represents the multinationality variable; a dummy that takes on the value of 1 at international sales > 20%. MNC * CFdev represents the Cash Flow deviation variable, the independent variable for the second hypothesis. MNC * Crights represents the Creditor Rights variable, the independent variable of the third and final hypothesis. Size is proxied by total net sales, to represent size. Intangible Assets is the total intangible assets scaled by the size variable. EBITDA is the performance proxy, scaled by the size variable. GDP-growth represents the GDP growth of the country in which the firm is incorporated, in percentages.
Variable SOA SOA SOA
MNC -0.0005 0.0093 -0.0292 0.0475 (0.0489) (0.0530) (0.0617) (0.1053) CFdev 0.1626 (0.1599) MNC * CFdev 0.1883 (0.2418) Crights -0.0130 0.0472 (0.0144) (0.0401) MNC * Crights 0.0000 -0.0220 (0.0008) (0.0500) Size -0.0205 0.0009 -0.0177 (0.0138) (0.0019) (0.0139) Intangible Assets 0.0000 -0.0071 0.0000 (0.0008) (0.0088) (0.0008) EBITDA 0.0007 0.0006 (0.0019) (0.0019) GDP-growth -0.0060 -0.0062 (0.0087) (0.0087) _cons 0.5700** 0.9326** 0.7657** 0.7445* (0.2361) (0.3751) (0.3907) (0.3948)
Industry fixed effects Yes Yes Yes Yes
Adj-R2 -0.0000 -0.0000 -0.0000 -0.0000
Observations 37,562 34,773 33,739 34,712
