Original Research

Threshold capital structure and predictors of choice between distribution strategies

Mpinda F. Mvita, Leon M. Brummer, Hendrik P. Wolmarans

Received: 08 Apr. 2021; Accepted: 07 July 2021; Published: 29 Oct. 2021

Copyright: © 2021. The Author(s). Licensee: AOSIS.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Abstract

Orientation: The determination of a threshold capital structure and company specific attributes as predictors of choice between distribution strategies is crucial in the creation of shareholders’ wealth.

Research aim: To investigate whether the change in regimes given a threshold capital structure maximises distribution strategies over the period 1990–2017 and 1999–2017. In addition, the study examined how the capital ratio and company specific attributes were used in the process of choosing between distribution strategies.

Motivation for the study: The need to determine the impact of the capital ratio within different regions on distribution strategies motivated this study. In addition, the majority of studies on predictors of choice between distribution strategies have ignored the dual and the no distribution policy alternatives relative to share repurchases.

Research approach/design and method: all the data used in this research were sourced from the Iress data bases. The research employed an advanced panel threshold regression estimation and a multinomial logistic regression (pooled and fixed effects using the generalised structural equation model).

Main findings: Firstly, over the period 1990–2017 the empirical results revealed the existence of a single threshold effect between the debt-to-equity ratio and the dividend payments, and a double threshold effect between the total debt based on the book value and the dividend payment. Secondly, the choice between distribution strategies was driven by company specific attributes.

Practical/managerial implication: These findings provide useful insights to South African managers for formulating and maximizing pay-out decisions.

Contribution/value-add: The study contributes to the scant body of knowledge on the effect of threshold capital ratio and company specific attributes on distribution strategies.

Keywords: threshold capital structure; predictors of choice; distribution strategy; South African context; generalised structural equation.

Introduction

Financing decisions determine pay-out decisions and according to Noronha, Shome and Morgan (1996), optimal debt ratio has corresponding to it an optimal pay-out rate, at which point the sum of transactions and equity agency costs are minimised for that debt ratio. Furthermore, at optimum leverage, the value of the company is maximised whilst the Weighted Average Cost of Capital (WACC) is minimised. If the company has too little debt, it loses the leverage benefits, because it does not maximise its tax shields. Such a company will pay higher taxes if it is profitable. To move its leverage to the optimal level, the company should either issue more debt or increase its capital distribution to shareholders by either dividend payments or share repurchases. On the one hand, if a company decides not to distribute cash or distribute less in the form of dividends or share repurchases, the company will have more cash available, thus reducing its reliance on external financing. On the other hand, if a company distributes more cash, it will have less internal earnings, thus increasing the company’s dependence on debt or other external financing. This implies that the decision to raise funds is directly associated with the dividend payments (Yusof & Ismail 2016).

Given the importance of a distribution policy, an optimal distribution policy is crucial because it is driven by the decision to issue debt or equity. In addition, prior studies have reported that the decision to distribute cash could be affected by various factors, including profitability, company size, growth opportunities and cash flow (CF). Over the years, academic research has also systematically examined the predictors of choice between distribution strategies. Despite the abundance of research on the effect of capital structure on distribution policies (the dividend payments and share repurchases) and on the predictors of choice between distribution strategies, the evidence reported remains inconclusive. Furthermore, the majority of South African studies ignored the different natures of capital structure in explaining pay-out decisions.

In this research, we argue that the decision to distribute cash to shareholders is driven by the capital structure and company-specific variables. The research empirically investigates whether the change in the regimes of capital structure (given a threshold) maximises distribution strategies. Furthermore, the research examines how Johannesburg Stock Exchange (JSE) listed companies used company-specific variables in choosing between distribution strategies. Advanced threshold regressions provide us with a platform to test for the effect of threshold capital structure on distribution policies and a multinomial logistic regression (pooled and fixed effects using structural equation models) provides us with a platform to test for the company-specific variable as a determinant of choice between distribution strategies.

The main contributions of this research to the existing knowledge of capital structure and distribution policies decisions are twofold. Firstly, to the best knowledge of the authors, this is the first study that applies threshold regression models to investigate how the change in regimes given a threshold capital structure affects pay-out decision for companies listed on the JSE over the period 1990–2017 and 1999–2017 because share repurchases were introduced in South Africa in 1999. The model is computed in line with Hansen (1999, 2000) and Chan (1993). Thus, the study provides new insight on the effect of threshold debt-to-equity and total debt based on the book value on pay-out decisions in South Africa. Secondly, this study is novel as it investigates how the capital ratio and company-specific variables such as profitability, company size, quick ratio, market volatility, liquidity and CF were used as predictors of choice between distribution strategies in line with Jagannathan, Stephens and Weisbach (2000). The pay-out decisions in this research are not only limited to the dividend payments and share repurchases but also extended to dual decisions (the dividend payments and share repurchases) and the no distribution alternative (neither the dividend payments nor the share repurchases). The research used pooled and fixed effect multinomial logistic regression using the generalised structural equation method.

The study provides useful insight to the board of South African directors for formulating and revising distribution strategies by taking into consideration the change in regimes given a threshold capital ratio and company-specific attributes that have been evidenced to exercise significant influence on the dividend payments and share repurchases.

The remaining article is structured as follows: the next section discusses the theoretical framework of the study, reviews relevant literature on capital structure and companies’ specific attributes as determinant of dividend payments and outlines the research hypotheses. Then the research methodology presents the data sources and the method employed to achieve the objective of the study. A further section presents the research findings and discussions, with the conclusion, implications for managers, limitations and suggestions for future research drawn in the final section.

Literature review and hypotheses development

A study’s background and hypothesis development are elaborated through the literature review. The literature review is described as the drive to gain first-hand knowledge of what has been investigated within a particular field of the study (Bless, Higson-Smith & Sithole 2013). In this section, the research reviews, firstly, prior research on threshold capital structure and distribution policy. Secondly, it reviews prior research on predictors of choice between distribution strategies.

Threshold capital structure and distribution policy

There is a vast literature on the relationship between financing decisions and pay-out decisions, mainly spurred by the seminal contribution of Modigliani and Miller (1958) who argued that in a frictionless and perfect market, company value is independent of the capital structure and there is no optimal capital structure for a specific company. However, the assumption of perfect capital markets with no transaction costs, no taxes, homogeneous expectations and asymmetric information is unrealistic and not applicable because taxes, friction, agency costs and the differences in information all exist in reality (Jensen & Meckling 1976; Modigliani & Miller 1963; Myers 1977; Myers & Majluf 1984).

Most researchers have considered leverage as an explained variable in the optimisation process. This variable, however, varies amongst researchers, as some of the conflicting results are because of the variation and the measurement of this variable. For instance, some authors investigated the optimisation of the capital structure using company-specific variables and confirmed that companies had an optimal capital structure (Barclay, Smith & Morellec 2006; Effendi 2017; Harford, Klasa & Walcott 2009; Hovakimian, Opler & Titman 2001; Ozkan 2001). Furthermore, the above-mentioned studies have used quantitative descriptive analysis, static and dynamic models in the optimisation process. Highlighting one major limitation of a single period model in the optimisation of the capital structure, Fischer, Heinkel and Zeckner (1989) argued that the model ignores the company’s optimal restructuring choices in response to fluctuations. Furthermore, in the absence of transaction cost, a company can carry large amounts of debt and in terms of an appropriate share repurchase strategy, the company can capture large tax shields whilst keeping the debt essentially riskless. Companies allow their financial structure to change over time because of the costs of recapitalising, therefore, any ratio lying within a set of specific boundaries could be optimal. As a result, similar companies could have different leverage ratios at any point in time. According to the boundary conditions, when the value-to-debt ratio increases, the leverage ratio drops and when the value-to-debt ratio decreases over time, the leverage ratio increases. Furthermore, in terms of the trade-off theory of capital structure, when the debt ratio increases, the interest tax shield increases. However, leverage-related costs increase to offset the positive effects of the debt ratio on the company value and subsequently on the distribution strategies.

Investigating the effects of financing decisions on the distribution policy, some authors indicate that there is a positive relationship between leverage and dividend payments. They infer that because of the positive relation between leverage and dividend payments, a negative relationship between leverage and the retention rate exists, suggesting the preference of retention to debt financing (Al-Najjar 2011; Chang & Rhee 1990; Gill, Biger & Tibrewala 2010; Sim 2011). In contrast, other authors argue that when companies borrow capital, they commit themselves to the payment of fixed commitment charges, which include principal repayments. Failure to meet these obligations may result in the companies facing the risk of liquidation and bankruptcy, indicating a negative relationship between the leverage and dividend payments (Arko et al. 2014; Banerjee & De 2015; Ben Amar, Ben Salah & Jarboui 2018; Benavides, Berggrun & Perafan 2016; Frank & Goyal 2009; Kaźmierska-Jóźwiak, 2015; Moon, Lee & Dattilo 2015; Nizar Al-Malkawi 2007; Yusof & Ismail 2016). However, some of the above-mentioned researchers have overlooked the effect of changes in regimes given a threshold capital ratio in explaining pay-out decisions (dividends and share repurchases). Hence, the need for a robust approach in line with Fischer et al. (1989) to investigate how the company’s change in regimes given a threshold capital ratio restructuring choices in response to fluctuations that affect pay-out policies is warranted.

Examining the effect of debt on distribution policies, Rozeff (1982) argued that riskier companies’ pay-out policy will lower dividends, indicating a negative statistical relationship amongst dividends, bankruptcy cost and the amount of debt used by a company. This argument is also in line with the narrative that a company’s capital structure that is below the threshold increases its distribution policies, suggesting a positive correlation between the capital structure and distribution policy. However, when the company’s capital structure is above the threshold, the maximisation of wealth creation becomes unclear as the two policies will move in opposite directions.

Determining a threshold debt-to-equity ratio which could maximise corporate value and using the method of quantitative descriptive analysis for the period 2011–2015, Effendi (2017) found that companies with optimal capital structures were in line with the trade-off theory. The capital structure is optimal if the debt levels are such that corporate value will be at its maximum. However, if the debt limit passes a certain level, profit and corporate value will decrease. This finding also suggests that a decrease in profitability and corporate value may lead to a decrease in a company’s willingness to pay dividends or repurchase shares, because these distributions are paid out of earnings.

Non threshold variables and distribution policy

Prior studies have also evidenced that the company size, profitability, CF, market volatility, growth and investments could influence the distribution policy. Based on the agency cost theory, the wide spread of ownership in larger companies decreases the shareholders’ ability to monitor the internal and external financing of the company, which leads to greater information asymmetry, thereby increasing agency costs. The dividend payment is an approach to mitigate the problem. In other words, larger companies tend to distribute more cash than smaller companies. The positive relationship between the company size and the dividend payment is evidenced in a number of previous studies (Juma’h & Pacheco 2008; Mehrani, Moradi & Esk 2011; Yusof & Ismail 2016). At the same time, the negative effect of company size on the distribution policy was also reported in several prior studies (Ahmed & Javid 2009; Huda & Farah 2011).

Based on the signalling theory, the dividend announcements convey some information about the company’s performance that would cause shareholders to react to the announcement (Miller & Modigliani 1961; Lintner 1956). In particular, the ability of companies to pay a dividend signals a company’s profitability. In addition, the higher the amount of the distribution policy in the form of dividends or share repurchases, the greater the profitability, suggesting a positive correlation between profitability and the distribution policy (Baker et al. 2019; Ho 2003; Yusof & Ismail 2016). The announcement of the dividend payments also signals the stability of the company’s future CF suggesting a positive correlation between CF and the dividend payments. This finding is supported by some scholars (Chen & Dhiensiri 2009; Reyna 2017). On the contrary, a negative but significant relationship was reported by prior research (Baker et al. 2019; Imran 2011; Utami & Inanga 2011). Yusof and Ismail (2016) found an insignificant effect of cash on the dividend payments.

In terms of risk, high dependence on external financing reflects higher volatility of a company’s CF, which subsequently increases the company’s risk (Yusof & Ismail 2016). In minimising the company’s risk because of external financing, a company will pay lower dividends (Rozeff 1982). This is consistent with the findings by some scholars who found a negative correlation between risk and the distribution policy (Al-Shubiri 2011; Juma’h & Pacheco 2008; Ramli 2010).

Growth and investment as determinants of dividend payments are in line with the agency cost theory, whereby companies with no growth or fewer investment opportunities have greater exposure to agency costs, which are related to free CFs (Yusof & Ismail 2016). This argument suggests that to reduce the agency costs, these companies will pay higher dividends to the shareholders as compared to the companies with high growth and greater investment opportunities. Rozeff (1982) hypothesised that the relationship between anticipated investment opportunities and dividend payments is negative because companies prefer to avoid transaction costs related to external financing. Evidence from various studies supports the narrative that companies distribute lower dividends when they are experiencing higher growth opportunities because this growth seemingly involves higher investment expenditures.

Against this background, the authors have resorted to a relatively different approach, namely, threshold regression, to explain the effects of a dynamic capital structure on the pay-out decisions. There are several advantages in using threshold regression. Firstly, threshold regression distinguishes between the characteristics of companies’ dynamic capital structure, which collectively defines mean leverage ratio, because it separates companies’ lower and upper refinancing thresholds and targets leverage ratios, as proposed by fundamental contributions to the dynamic trade-off theory (Fischer et al. 1989). Secondly, the implication of the model is that the debt ratio is discontinuous and monotonic in the bankruptcy cost parameter. Thirdly, building on the traditional tax/bankruptcy cost theory of capital structure irrelevance as argued by Modigliani and Miller (1958), the authors argued that the model provides distinct predictions for company-specific properties regarding the range of dynamic leverage ratios: smaller, riskier and lower bankruptcy cost companies exhibit wider swings in their debt ratios over time.

Although studies on the effect of company-specific attributes on pay-out decisions abound, studies on the effect of a dynamic capital structure on pay-out policies in South Africa are scarce. The majority of studies have used the capital structure as a static variable in explaining pay-out decisions, ignoring the different fluctuations in the capital structure.

Based on the theoretical framework of the trade-off theory and prior studies, despite their mixed findings, the following hypotheses stated in alternative form were developed for determining how the threshold capital structure is used in the process of distribution policies and their choices thereof:

H1: There is a positive capital structure threshold effect for the payment of dividends over the periods 1990–2017 and 1999–2017.

H2: There is a positive capital structure threshold effect for share repurchases over the period 1999–2017.

Predictors of choice between distribution strategies

The conventional literature on corporate finance has overlooked the effects of risk and return in the choice between dividend payments and share repurchases. Recent studies seeking to explore the determinants of choice between dividend payments and share repurchases highlighted the importance of the different measures of capital structure and financial distress when choosing between distribution strategies (Renneboog & Trojanowski 2011; Wesson et al. 2018). Several channels through which the capital structure and financial distress might influence the choice between dividend payments and share repurchases were identified and examined (Caudill et al. 2006). Nonetheless, the findings on capital structure, financial distress and company-specific variables as predictors of choice between the distribution strategies remain theoretically ambiguous and empirically inconclusive. The authors argue that the omission of the different natures of capital structure, different measures of capital structure, financial distress and company-specific variables in the choice between distribution policies is likely to generate misleading results and lead to inappropriate inferences, casting doubt on the conclusion drawn from the existing literature.

According to Wesson et al. (2018) and Caudill et al. (2006), the choice between dividend payments and share repurchases is also expected to be influenced by company-specific characteristics. Furthermore, industries and companies face different macroeconomic and microeconomic risk and returns. Companies and industries differ in terms of profitability, growth option, legal and tax frameworks of countries where they operate, asset structures, calibre of management and operational risks. The design of the company’s choice between the dividend payments and share repurchases must incorporate all these factors and enable it to minimise risk and maximise return. The choice between dividend payments and share repurchases is expected to vary between countries, industries and companies’ sizes. The dividend payments and share repurchases are expected to be greatly influenced by both macroeconomic and microeconomic factors, as well as by the uniqueness of each company (companies’ heterogeneity). Past empirical research has identified several factors that ultimately determine the choice between dividend payments and share repurchases as follows: company size, institutional ownership and number of shareholders, officers’ and directors’ ownership, level of debt, dividend payment history, size of distribution, level of company undervaluation, share performance before distribution, takeover threats and executive share options. Theories on distribution strategies use these variables to explain the pay-out choices of companies.

Developing an empirical model of choice amongst all four one-time cash disbursement methods in the US, Caudill et al. (2006) reported that company-specific variables such as the ownership structure, current pay-out level, the size of the distribution and the share price performance prior to the announcement date were the significant determinants of a company’s choice between alternative pay-out methods. The basis of Caudill et al. (2006) argument is on the expected relationship between different pay-out methods and the company-specific variables. However, the capital structure as a determinant of choice was not considered.

Using logistic regression, Wesson et al. (2018) found that the level of debt per sector was statistically significant, whereas in the total sample, this variable was not found to be statistically significant. This is an indication that the financing decisions could have a strong implication for the choice between dividend payments and share repurchases. Their findings are not in line with international empirical evidence, which postulates that lower debt levels are associated with the choice of open-market share repurchases, mainly because share repurchases are usually financed through debt, hence resulting in increased financial leverage for companies with below-target leverage levels. Furthermore, in the South African regulatory environment, the reported results may indicate that open-market share repurchases are not financed through debt, as is the case globally, but rather financed by utilising cash reserved. It is worth noting that Wesson et al. (2018) did not use the different measures of capital structure and the different natures of the capital structure. The study used a single debt ratio, namely the long-term debt to total assets.

Renneboog and Trojanowski (2011) investigated the decision to distribute funds as well as the choice of pay-out channel (e.g. dividend payments, share repurchases, both dividend payments and share repurchases and neither dividend payments nor share repurchases). Their findings demonstrate that the importance of share repurchases is increasing, but that dividend payments still constitute a vast proportion of the total pay-out. Using company-specific variables as determinants of choice between dividend payments, share repurchases, dividend payments and share repurchases and no pay-out (neither dividend payments nor share repurchases) for United Kingdom companies (UK), the findings showed that UK companies that were profitable and large were more likely to pay than not to pay dividends. They found that leverage was a decreasing predictor of dividend payments and companies that were highly leveraged were less likely to pay dividend. The UK companies that were profitable and large were more likely to repurchase shares than not to pay (neither the dividend payments nor the share repurchases). Highly leveraged companies were less likely to repurchase shares than not to pay. The UK companies that were profitable and large were more likely to engage in both dividend payments and share repurchases than not to pay. It is worth noting that Renneboog and Trojanowski (2011) did not distinguish between the natures of the capital structures.

Jagannathan et al. (2000) used a multinomial logistic regression to investigate financial flexibility and choice between dividend payments and share repurchases. They state that share repurchases and dividend payments are used at different times, by different kinds of companies. Share repurchases are very pro-cyclical, whilst dividend payments tend to be steady over time. Dividends are paid by companies with higher permanent operating CFs, whilst repurchases are used by companies with higher temporary, non-operating CFs. Repurchasing companies also have more volatile CFs and distributions. Finally, companies repurchase shares following poor market stock performance and increase dividends following good market performance. These results are consistent with the view that the financial flexibility inherent in repurchase programmes is one reason why they are sometimes used instead of dividend payments. It is worth noting that using company-specific variables (company size, operating income, institutional ownership and prior pay-outs) as predictors of choice, Jagannathan et al. (2000) compared the following decisions: firstly, firm-years with no increase in dividend payments to firm years in which distribution through the use of repurchases increased. Secondly, increased dividend payment or both increasing dividend payments and share repurchase. Thirdly, company-years with a new share repurchase programme or expansions to the other possible outcomes. Fourthly, company-years with both share repurchases and dividend increases to the other alternatives and company-years with dividend increases compared to the other alternatives. They did not investigate the effect of the capital structure as a predictor of choice.

De Jong, Van Dijk and Veld (2003) used single and multinomial logistic regression models to investigate dividend payments and share repurchases by Canadian companies listed on the Toronto Stock Exchange. Their findings are firstly consistent with the structure in which the company first decides whether it wants to pay out cash to its shareholders or not, and secondly, the company decides on the form of pay-out: dividend payments, share repurchases or both dividend payments and share repurchases. Pay-out is determined by free CF and the behavioural and tax preferences of company management. Furthermore, the pay-out is less likely to be divided between dividends and share repurchases if the company has executive stock option plans. Finally, in line with the model by Brennan and Thakor (1990), the empirical evidence of De Jong et al. (2003) validates the existence of asymmetric information amongst outsiders based on the narrative that it is associated with the preference for dividend payments over share repurchases. It is worth pointing out that they did not use capital structure as one of the predictors. They also did not find evidence of free CF and overinvestment as determinants of dividend payments relative to share repurchases.

Using a logistic regression estimation to define a comprehensive life cycle model of the likelihood of dividend payments, Hauser and Thornton (2017) found that companies that were profitable, larger and had higher retained earnings were more likely to pay dividends. Furthermore, the results also revealed that companies that had growth opportunities were less likely to pay dividends. Finally, the results showed that companies that were equity financed were more likely to pay dividends. Hauser and Thornton (2017) did not investigate share repurchases.

To fill the gap of using financing decisions in the choice between pay-out decisions, this research uses two different measures of the capital structure (total debt based on the book value and the debt-to-equity ratio) and extends the list of company-specific factors of choice between the decision to pay a dividend, to repurchase shares, to engage in both dividend payments and share repurchases and to engage in neither the payment of dividends nor the repurchase of share.

Methodology

Data

All the data used in this research were sourced from the IRESS database. The sample consisted of 68 companies listed on the JSE for the periods 1990–2017 and 1999–2017. A combination of judgement and convenience sampling was used to draw a sample from four industries. The considered industries are basic material, consumer goods, consumer service and industrial. Financial companies were excluded from the study. Because share repurchases were only authorised in 1999 as a distribution policy, the study is conducted over two periods. To minimise the problem of outliers, the data were winsorised at the fifth and 95th percentiles.

Threshold regression specification

To capture the threshold effects, the researchers used the following single set-up threshold model in line with Hansen (1999, 2000) and Chan (1993):

where:

• ds_i,t: represents the distribution strategies (cash dividend paid [ordinary share dividend scaled by total assets, share repurchases (share repurchases scaled by total assets)) and total payouts (the sum of cash dividend and share repurchases scaled by total assets)] of a company i in period t;
• de_i,t: represents the debt-to-equity ratio ([Total long-term loan capital + total current liabilities]/Total owners’ interest) and is also a threshold value;
• γ: represents the specific estimated threshold value;
• h_i,t: represents the five control variables, namely size (Logarithm of sales and total assets), profitability (Return on assets [ROA], [Profit before interest and tax {EBIT} – Total profit of extraordinary nature]/Total assets × 100), (CF, Sum of net income plus depreciation expenses/Total assets or CF from operating activities/Total assets), INVEST (investment in fixed assets acquired or the growth in sales) and (VO, standard deviation of market price) of a company i in period _t;
• μ_i,t: represents the fixed effect, which represents the heterogeneity of companies under different operating conditions and
• ε_i,t: represents the error terms, assuming that they are independent and identically distributed with mean zero. The finite variance is σ² (ε_i,t ~ i.i.d.(0,σ²).

If there is a double threshold, model 1 can be modified as:

where the threshold value γ₁ < γ₂. This can be extended to multiple thresholds (γ₁, γ₂, γ₃,…,γ_n).

Testing for a threshold

To test for the significance of the model, the null and the alternative hypotheses can be, respectively, represented as:

On the one hand, when the null hypothesis (the coefficient ω₁ = ω₂) holds, it indicates that the threshold effect does not exist. On the other hand, when the alternative hypothesis (the coefficient ω₁ ≠ ω₂) holds, it indicates that the threshold effect between financing decisions and distribution strategies exists.

Under the null hypothesis of no threshold, the model is:

After fixed-effects transformation is done, we have:

The regression parameter is estimated using OLS, which yields estimated , residual and the sum of the squared errors .

Hansen (1999) recommends that the relevant F-test approach and the sup-Wald statistic be used to test for the threshold effect and to test the null hypothesis, respectively. That is:

F = SupF(γ); where:

Under the null hypothesis, some coefficients (e.g. the pre-specified threshold [γ]) do not exist; therefore, a nuisance parameter exists.

Model of choice specification

To determine how the JSE-listed sample companies in the four main sectors (basic material, consumer goods, consumer service and industrial) chose between dividend payments, share repurchases, the engagement in both share repurchases and dividend payments and the engagement in neither share repurchases nor dividend payments using the nature of the capital structure and company-specific variables, the research used multinomial logistic regression estimation. Multinomial logistic regressions (pooled and fixed effect using the generalised structural equation model) are a straightforward extension of logistic models. Supposing that a dependent variable has j categories, one value (typically the first, the last or the value with the highest frequency) of the dependent variable is designed as the reference category. The probability of membership in other categories is compared with the probability of membership in the reference category (e.g. dividend payments relative to share repurchases).

For a dependent variable with J categories (dividend payments, repurchasing shares, engaging in both or engaging in none), this requires the calculation of J-1 equations, one for each category relative to the reference category, to describe the relationship between the dependent variable (distribution strategies) and the independent variables (capital structure and company-specific variables). Hence in line with Jagannathan et al. (2000), if the first category is the reference category, then for j = 2, …, J, the probability that the y takes, j can be written as:

The methodology provides insight into the sequential decision-making moments concerning distribution strategies (the dividend payments, share repurchases, both and none). For the multinomial logistic regression across time with unobserved heterogeneity (fixed effect), the following model is used:

The parameter α_i is an individual effect, β_j, is the coefficient vector, which is constant for the given company in the sample. The advantage of the multinomial logit model with fixed effects is that it allows for individual unobserved heterogeneity with respect to the intercepts.

Research findings and discussion

The objective of research was to investigate, firstly, whether the change in the regimes of capital structure (given a threshold) maximises distribution strategies. Secondly, to examine how JSE listed companies used company-specific variables in choosing between distribution strategies. The following sections present the findings and discussion of the research.

Summary statistics

Table 1 presents the descriptive statistics for the threshold regression model used in this study. As shown in Table 1, the average actual dividend paid and average change in debt-to-equity ratio of the companies are 0.028779 and 9.17E-05, respectively. Whilst the average change in total debt based on the book value of companies is 0.000541 and the average size is 6.708634 in the form of the natural logarithm of total assets. The mean value for profitability is 11.35141 and the mean CF is 0.101436. The average investment opportunities and the average VO are 0.070579 and 38.37514, respectively. Over the period 1999–2017, the average actual dividend paid, share repurchases, change in the debt-to-equity ratio and change in total debt based on the book value are 0.033016, 0.004406, 0.000585 and 0.000394, respectively. The mean value of profitability is 11.27745 and the mean value of CF is 0.108711

Threshold capital structure results of dividend payments: 1990–2017

The results in Table 2 show that there is a single threshold effect for the debt-to-equity ratio and a double threshold effect for the debt-to-asset ratio. The regimes are distinguished by the different slopes and, and the findings suggest that there is a level of capital ratio beyond which the relationship between financing and pay-out decision becomes unclear.

Threshold coefficient of dividend payments for the debt-to-equity ratio

In the first regime, where the debt-to-equity ratio was less than 1.6654999, the estimated coefficient of the upper bound was positive and significant at 5%. The results indicated that a one-unit change in the debt-to-equity ratio was associated with a 0.005195 increase in the actual dividend paid. In the second regime where the debt-to-equity ratio was greater than or equal to 1.6654999, the estimated coefficient was positive and insignificant. The results indicated that when the debt-to-equity ratio of the JSE-listed companies in the sample was greater than 1.665499, the relationship between financing decisions and dividend payment decisions became insignificant. Two classes of companies shown by the point estimates were those with low debt-to-equity ratio and those with high debt-to-equity ratio. Comparing the two regimes, companies that were not at risk or lowly geared were likely to distribute dividend and companies that were at risk or highly leveraged would not pay dividend. In addition, the findings suggest that below the threshold, the return was higher and above the target, the risk was higher. Therefore, Hypothesis 1 was supported.

Threshold coefficient of dividend payments for total debt based on the book value

To validate the above results, the research uses a different measure of the capital structure (total debt based on the book value). The results reveal that there is a double threshold effect of the long-term debt based on the book value. In the first regime where the total debt based on the book value is strictly less than 0.5585192, the upper bound coefficient is positive and significant at the 5% level. In the second regime where the total debt based on the book value is between 0.5585192 and 0.8331859, the estimated coefficient of the upper bound is positive and significant at the 10% level. In the third regime where LTB is greater or equal to 0.8331859, the coefficients are both positive and insignificant. The research concludes that there is an increasing trend in the threshold effect between long-term debt based on the book value and the dividend payments for the period 1990–2017. Therefore, Hypothesis 1 is accepted.

Non-threshold coefficients and dividend payments

The coefficient of the lagged company size is positive and statistically significant. A one-unit change in the lagged size is associated with a 0.002392 increase in the actual dividend paid (with a DE used as a threshold) and with 0.002989 increase in the actual dividend (with a LTB used as a threshold) at the 5% and 1% levels, respectively.

The coefficient of the lagged company profitability is positive and statistically significant. A one-unit change in the lagged profitability is associated with a 0.000932 increase in the actual dividend paid (when the DE is used as a threshold) and with a 0.000980 increase in the actual dividend (when the LTB is used as a threshold) at the 1% significance level.

The coefficient of the lagged company’s CF is positive and statistically significant. A one-unit change in the lagged CF is associated with a 0.139429 increase in the actual dividend paid (when the DE is used as a threshold) and with a 0.135828 increase in the actual dividend (when the LTB is used as a threshold) at the 1% significance level.

The coefficient of the lagged VO is negative and statistically significant. A one-unit change in the lagged volatility is associated with a 9.71E-05 decrease in the actual dividend paid (when the DE is used as a threshold) and with a 0.000100 decrease in the actual dividend (when the LTB is used as a threshold) at the 5% significance level.

Threshold capital structure results of only share repurchases and total pay-out: 1999–2017

Table 3 presents the threshold effects of the capital structure (the debt-to-equity ratio) for only share repurchases, dividend payments and distribution strategies (the sum of CD and SRP) for the period 1999 to 2017.

Threshold effects of the debt-to-equity ratio and distribution strategies (SRP, actual dividend paid and DS)

In the first regime over the period 1999–2017, where the debt-to-equity ratio was less than 0.56019999, the estimated coefficients of the upper bound and the lower bound were both negative and statistically insignificant. The results indicated that there was a threshold effect of the debt-to-equity ratio for share repurchases. However, this effect was insignificant because share repurchases were not of a big magnitude. Likewise, in the second regime where the debt-to-equity ratio was greater or equal to 0.56019999, the estimated coefficients of the debt-to-equity ratio were insignificant. The existence of the threshold of the debt-to-equity ratio did not have any effects on share repurchases of a smaller magnitude. Size, profitability and CF positively and significantly correlated with share repurchases. Investment and VO were negative and insignificant. Cash flow appeared to be the biggest non-threshold determinant of share repurchases.

The introduction of share repurchases in 1999 did not deter JSE-listed companies in the sample from paying dividends. The results revealed that over this period (1999–2017), in the first regime where the debt-to-equity ratio was strictly less than 1.4144999, the estimated coefficients of the debt-to-equity ratio were both positive and significant at the 1% and 5% significance levels, respectively. The results indicated that a one-unit change in the debt-to-equity ratio was associated with a 0.006364 increase in the actual dividend paid at the 1% significance level for the upper bound and with a 0.003641 increase for the lower bound. In the second regime where the debt-to-equity ratio was greater or equal to 1.4144999, the estimated coefficient of the lower bound of the debt-to-equity ratio was positive and significant at 5%. The results indicated that there was an existing significant threshold effect of the capital structure for the dividend payments even after the introduction of share repurchases. The estimated coefficients of profitability and CF were positive and highly significant. The coefficient of VO was negative and highly significant. The coefficients of size and investment were negative and insignificant.

Extending the threshold effect to the distribution strategies (the sum of share repurchases and the actual dividend paid), the results revealed that in the first regime where the debt-to-equity ratio was strictly less than 1.113899, the estimated coefficient of the debt-to-equity ratio was positive and significant at 5%. In the second regime where the debt-to-equity ratio was greater than or equal to 1.113899, the estimated coefficients were positive but insignificant. Therefore, Hypothesis 2 could be accepted.

The findings in Table 4 suggest that there is no threshold effect of the total debt based on the book value, on the dividend payments and the sum of both the dividend payments and share repurchases over the period 1999–2017. This finding indicates that the threshold effect does not exist and ω₁ = ω₂. The findings on the effects of company-specific attributes (non-threshold variables) are similar to the results in Table 3.

Predictors of choice between distribution strategy results

The results of the debt-to-equity ratio, the total debt based on the book value and company-specific variables as predictors of choice between distribution strategies are presented in Tables 5 (pooled model) and 6 (fixed effect model). The first set of coefficients in Table 5 and 6 represents comparison between the decision to pay dividends and the decision to repurchase shares. Only profitability and company size were the significant coefficients in the model in the fixed-effect model. More profitable companies were more likely to pay dividends and less likely to repurchase shares. This finding is similar to the finding by Renneboog and Trojanowski (2011). Company size was a negative and significant predictor, indicating that large companies were less likely to pay dividends. This finding contradicts the finding by Renneboog and Trojanowski (2011), who found that in the UK, large companies were more likely to pay dividends. But the finding is similar to the finding by Dittmar (2000), Liu and Mehran (2016). Dittmar (2000) argued that if size and information available are positively correlated, large companies are less likely to be valued incorrectly. Thus, the conjunction of these two results illustrates that large companies may also be miss valued and use share repurchases to take advantage of possible undervaluation. In the pooled model, cash was a significant and positive predictor whilst growth and VO were negative and significant predictors.

The second set of coefficients in Tables 5 and 6 represents the comparison between companies that engaged in the dual distribution (dividend payments and repurchase of shares) relative to share repurchases repurchased only. The only significant coefficients were profitability (in the pooled and fixed effects model) CF and VO (in the pooled model). Surprisingly, profitability appears to be a negative predictor in the pooled model and a positive predictor in the fixed model. As a result, when accounting for companies’ differences, the findings suggest that more profitable companies were more likely to engage in both the dividend payments and share repurchases. Renneboog and Trojanowski (2011) had a similar result. Companies with more CF were less likely to engage in a dual distribution strategy relative to share repurchases. Companies faced with high uncertainty in the market were more likely to engage in both dividend payments and share repurchases.

The final set of coefficients in Table 5 and Table 6 represents a comparison between the no distribution alternative (choosing not to engage in dividend payments and share repurchases) relative to share repurchases only. The only significant coefficients were profitability, CF, size, growth opportunities and VO. Company size and the CF were significant and negative predictors, indicating that large companies with high levels of CF were less likely to engage in the no distribution alternative and more likely to only repurchase shares. Companies faced with higher VO were more likely to engage in the no distribution policy relative to share repurchases. Companies with higher debt levels were more likely to engage in the no transaction alternative relative to share repurchases. It is worth pointing out that the results of the pooled model should be interpreted with caution because this model assumes that companies are not different, and this is not always the case because of cross section differences.

Conclusion, implications, limitations and suggestions for future research

Threshold capital structure and distribution strategies

Capital structure as a determinant of distribution strategies amongst other variables has been investigated in the existing literature. However, empirical evidence on the effect of the change in regimes (given a threshold capital structure) on distribution strategies and empirical evidence of the capital structure as a predictor of choice between the decision to pay dividends, repurchase shares, engage in both (dual distribution) and engage in neither dividend payments nor share repurchases are still scant in developing economies. The aim of this study was to extend the empirical literature by providing new evidence from South African markets.

Firstly, the research used an advanced threshold regression approach to capture the threshold effects of two alternatives measures of the capital structure (debt-to-equity ratio and the total debt based on the book value) for dividend payments and share repurchases. The results of the study revealed the existence of a threshold capital structure for the payment of dividends to South African shareholders. Surprisingly, the threshold effect on share repurchases appeared to be insignificant over the period 1999–2017. The researchers argued that the non-existence of the threshold effects on share repurchases could be explained by the narrative that South African companies in the sample did not use share repurchases as the main distribution policy.

Predictors of choice between distribution strategies

Using a multinomial logistic regression model (pooled and fixed effect using the generalised structural equation model), the results revealed that the choice between dividend payments, the engagement in both dividend payments and share repurchases and the engagement in neither dividend payments and share repurchases relative to share repurchases was driven by profitability, company size, cash flow, liquidity ratio and volatility. The results suggested that for every one-unit increase in profitability as a predictor of choice, JSE-listed companies in the sample were more likely to pay dividends only or pay dividends and repurchase shares in the same financial year. The results showed that during a period of high market volatility, South African managers of companies in the sample would choose not to pay dividends nor repurchase shares at all. Large companies were less likely to pay dividends and less likely to engage in none. Finally, the results suggested that companies with a higher debt-to-equity ratio in the sample were more likely to engage in a dual distribution policy than to choose only the repurchase of shares.

The study provides useful insight to the board of South African directors for formulating and revising distribution strategies by taking into consideration the change in regimes given a threshold capital ratio and company-specific attributes that have been evidenced to exercise significant influence on the dividend payments and share repurchases. In particular, if the board of South African directors of companies listed on the JSE is considering increasing the dividend payments to the shareholders, they must do so bearing in mind that the optimal debt ratio has corresponding to it an optimal pay-out rate, at which point the sum of transactions and equity agency costs is minimised for that debt ratio. Furthermore, the factors of profitability, investment opportunities, size, CF and VO must be given careful attention when choosing between distribution strategies.

This study is not without limitations. The present study focussed solely on JSE-listed companies in the main four sectors, namely, the basic materials, the industrials, the consumer goods and consumer services sector. For greater generalisability of the findings and to better reflect on capital structure and company-specific attributes as determinants of distribution strategies in South Africa, future research may be required to include other listed companies in South Africa. Despite its limitations, this study contributes to the existing literature regarding the important issue of the change in regimes, different measures of the capital structure and company-specific variables affecting distribution strategies in South Africa.

Acknowledgements

The authors declare that they have no financial or personal relationships that may have inappropriately influenced them in writing this article.

M.F.M. was responsible for conceptualisation, methodology, formal analysis review and editing whilst L.M.B. and H. P.W. were responsible for supervision, review and writing.

This article followed all ethical standards for research without direct contact with human or animal subjects.

This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.

The data are available as per request from the corresponding author.

The views and opinions in this article are those of the authors and do not necessarily reflect the official position of any affiliated agency of the authors.

References

Appendix 1