nueva página del texto (beta)
Inglés (pdf)
Artículo en XML
Referencias del artículo
Traducción automática
Enviar artículo por email
Citado por SciELO 
Similares en
SciELO 
Permalink
1. Introduction
The principles of good corporate governance were outlined by the OECD in 1999, however, in Colombia direct steps towards these principles began to be taken since 1995 with the promulgation of Law 222 regarding the replacement of administrators and Law 550 of 1999 in regarding the protection of shareholders and their equal treatment (Pachon, 2007). A crucial step in Colombian regulations towards the voluntary adoption of good corporate governance practices was the promulgation of Law 275 of 2001, which established that companies should establish committees in charge of ensuring compliance with voluntarily adopted corporate governance standards. Likewise, Law 964 of 2005 established that the number of independent directors should be equal to or greater than 25% (Pachon, 2007)
In 2002 McKinsey prepared a survey commissioned by the Colombia “Compite” network in 8 regions of the country and with a sample of 88 companies and found serious deficiencies in terms of the formation of boards of directors and the rights of shareholders (McKinsey and company, 2002). Furthermore, Langebaek and Ortiz (2007) conducted a study on the adoption of a good corporate governance code in 57 companies listed on the Colombian Stock Exchange and found that although all had adopted a corporate governance code, and heir degree of compliance was low especially in the equitable treatment of shareholders (38%) and the functioning of the boards of directors (63%).
Given this evidence, some authors argued against the complexity of the Colombian regulation related to corporate governance and the poor performance of companies. Hence, although the adoption of a corporate governance code was very high, compliance with it was very poor (Gaitan, 2009). Hence, between October 2005 and December 2006, the Colombian government designed and elaborated the Country Code of corporate governance. This project was led by the Colombian Financial Superintendency (CFS), which supervises companies listed on the Colombian Stock Exchange. The main objective of this project was to unify the different efforts to establish guidelines, policies, and requirements for a code of good corporate governance (CFS, 2007). Furthermore, this Country Code was refined and improved in a second version in 2014 (CFS, 2014) (Gaitan and Saravia, 2017).
The Country Code consists of a survey that measures the fulfillment of a group of concrete requirements that all security issuers should follow. These requirements are voluntary for the issuers; however, if they do not follow them, they should annually inform the Superintendency and the public about the reasons why. The survey covers four different topics: a) General Shareholders’ Meetings, b) Board of Directors (size, composition, internal regulation, rights and duties of the members, functions), c) financial and non-financial information disclosure, and d) dispute resolution.
The introduction of the survey as an official way to track progresses in terms of corporate governance (CG) made by Colombian firms was supposed to encourage the adherence to the Code and improve all other firm’s policies related to shareholders such as the dividend policy. This should happen even more after the introduction of the “fulfill or explain” model in 2007, according to which - although companies are still able to not adopt the Country Code - they must explain their reasons for not honoring each of the points expressed in the survey. Firms are, therefore, expected to have significantly improved their CG mechanisms and policies, including their dividend payout, after 2007 and then even more since 2014.
Furthermore, since the introduction of the survey by the Superintendency, it has been common in literature to base the measures of CG on indexes that consider the results of the survey. For instance, Lagos and Vecino (2011) introduced the IGCCP index (Country Code Corporate Governance Index), which is based on the survey made by the Superintendency each year. The study uses a sample of 660 surveys filled by Colombian listed companies from 2007 to 2010, observing a positive trend in the adherence to CG practices since the introduction of the survey. This has also a positive effect on the investors, who feel safer investing in companies that adopt CG practices (Lagos, 2013).
Regarding the dividend policy in Colombia, Aguirre-Ríos et al. (2021) indicate that companies must grant their shareholders at least 50% of their profits in the form of dividends. However, if the shareholders who hold more than 78% of the shares decide not to distribute dividends, the company is not obliged to give them according to article 155 of the Colombian Commercial Code. Hence, CG practices must encourage firms to distribute even higher dividends.
What about the relationship between CG practices and dividend policy? Mohapatra and Panda (2022) conducted an extensive literature review including 66 papers that have covered the relationship between corporate governance and dividend policy, published between 2000 and 2021, and found that there is a positive relationship between better corporate governance practice and higher dividend payout, but there is scarce evidence of this relationship on emerging markets.
Given the above, we aim to fill the current gap by analyzing the relationship between the adoption of a corporate governance code and the dividend payout in companies that report to the Superintendency of Companies in Colombia. Furthermore, we also analyze whether the quality of the code plays a role and whether the reforms of the Country Codes (2007 and 2014) had an impact on the companies’ dividend payouts. To the best of our knowledge this is the first study that relates CG practices with dividend policies in Colombia.
In the next section we discuss the current literature, while in the third section we describe the data and the variables. We present and discuss our results in the fourth and fifth sections. The last section concludes the study.
2. A brief literature review
Early studies about dividends argue that this is one of the ways in which the companies communicate to the market information about their future profit and about its cash flows (Bhattacharya, 1979 and Miller and Rock, 1985). Nevertheless, when it comes to dividend policy, it is mandatory to refer to the agency conflicts theory. Berle and Means (1932) introduced the agency problem theory, stating that the agency problem relates to the conflict of interests that is present between insiders and outsiders. Basically, what this theory proposes is that insiders, which are the ones who have control over the company assets, could use their controlling position for purposes that could be against of the outsiders’ interest and that is where the dividends policy starts to play a significant role.
As mentioned by Easterbrook (1984) and Jensen (1986), it can be used to relieve the agency problem between managers and shareholders because the distribution of dividends reduces free cash flow in the companies and its discretionary use. These dividends, when retained, could be used by the insiders for personal use or could be reinvested in low profit projects, even with negative present value that would only benefit the insiders. This is known as the “Opportunist Management” theory and consists in the fact that managers of firms with weak shareholders’ protection prefer to retain cash flows, instead of distributing them. One way to avoid opportunistic management is to improve the level of shareholder protection, but this implies an increase in the probability of paying dividends, at least for those companies that are controlled by insiders (Esqueda, 2016).
Another risk for minority shareholders lies in the excessive unbalance between majority and minority shareholders reflected in a high concentration level of power of the former in the decision-making process, with the risk of expropriation of the minority shareholders (Battacharya, 2017). Dyck (2001) argues that companies should pay higher dividends to gain a reputation of fair treatment of the minority shareholders. Other authors that studied the role of insiders consider the impact of the shares’ concentration in their analyses. They find that high concentration is associated with low levels of dividends distributed (Jensen et al., 1992). Therefore, high concentration might increase the risk of expropriation via less dividends distributed. If the main shareholders are in control of the administration, the small shareholders are not protected from expropriation (Dyck, 2001).
Similarly, Rozeff (1982) finds that the dividends payout is a significantly negative function of the percentage of shares owned by the insiders. Easterbrook (1984) explained that dividends expose companies to a recurrent poll by the market because the dividend payout increases the probability that the company would have to issue common shares more frequently. This type of control, or vigilance, that the market does, helps to prevent inadequate behaviors by the administration and to avoid projects of “not common benefit” with the company cash flows. It benefits minority shareholders and leads to a reduction of the agency costs.
Gompers et al. (2003) link the agency costs to the strengths of shareholders’ rights and these, in turn, to the CG practices. La Porta et al. (2000) affirm that managers prefer to retain profits to increase their own benefit or reduce the risk of human capital loss, stating that low dividend payout ratios are associated to poor government standards and poor protection for shareholders.
La Porta et al. (2000) compares two agency models: the “Outcome Model” and the “Substitute Model”. The outcome model establishes that, for countries with a good legal system that protects the shareholders’ rights, the dividend ratios are higher, keeping everything else constant. Instead, for companies with better investment opportunities, the dividend payout ratios are lower, since the shareholders can sacrifice the current cash flow in exchange for future cash flows, while their rights are very well protected by regulations.
The substitute model establishes that, in countries with minimum protection for shareholders, the companies with the best investment opportunities pay higher dividends to maintain a good reputation in terms of shareholders’ expropriation. This reputation lets them easily access the financial market and obtain foreign capital, when necessary. Their results were in favor of the outcome model and suggest that, in absence of growing opportunities, the dividend ratio should be higher in countries with weaker legal protection compared to countries with a stronger one (negative relationship). The theory behind this study is that dividends act as a reward for minority shareholders, which do not have the possibility to participate in corporate decisions.
According to Black (2001), the effects of the adoption of CG Codes are easier to detect in emerging economies, because they commonly have weaker regulations and present wide differences in the CG practices between different firms. Nevertheless, the evidence from emerging markets is limited (Mohapatra and Panda, 2022).
The last two decades have witnessed several publications of the relationship between better CG practices and higher dividend payout. Sawicki (2009) tested the outcome and the substitute models in five Asian countries (Hong Kong, Indonesia, Malaysia, Singapore, and Thailand) during the period 1994-2003 and found evidence in favor of the substitute model before the Asian crisis and evidence in favor of the outcome model later on.
Al-Armaneh and Yassen (2014) studied the relationship between ownership concentration and dividend yields for 47 firms listed in Aman Stock Exchange (Jordan). They found a non-linear relationship as ownership concentration increases, so the dividend yields up to some point (50%) where the relationship started to be negative.
Baker et al. (2020) studied the interplay between corporate governance and the dividend payout and dividend yield of 201 firms listed at the Colombo Stock Exchange in Sri Lanka. They found evidence in favor of the outcome model, so a positive relationship between a corporate governance index and the firms’ dividend payout. In an opposite result, Hammed et al. (2021) studied the same relationship for listed firms at the Pakistan Stock Exchange and found a negative result.
In the case of Australia, Shamsabadi et al. (2016) found a positive effect between a corporate governance index and the dividend payout. Using a random effect panel data model and 11174 firm-year observations, they tested two versions of their corporate governance index and several governance variables as robustness check and found consistently the same result.
In the case of Latin America, evidence about the relationship between corporate governance practices and the dividend policy is scarce. Bebczuk (2005) by sing a sample of 65 non-financial Argentinian companies, among the years 1996-2003, establishes that the distribution of dividends is higher with the introduction of CG improving the company’s results.
Gonzalez et al. (2017) focuses on six Latin American countries (Argentina, Brazil, Chile, Colombia, Mexico, and Peru) from 2007 to 2014 when exploring the connection between ownership concentration and dividend policy. Particularly, they found out that if the ownership concentration is high and there is a large shareholder, the dividend payout ratio is lower than otherwise. Moreover, the dividend payout ratio might further decrease due to the presence of a second large shareholder.
Mongrut et al. (2017) examines the effects of the adoption of the CG Code on the dividend payout ratio for companies listed in the Lima Stock Exchange, finding that companies that adhere to the code and especially those that have a high CG quality tend to pay higher dividends. Moreover, the sector that is paying the higher dividends in Peru - pension funds - is the one in which investors are the most active.
Given the above, we state the following hypothesis:
H1: There is a positive relationship between the firm’s adoption of a CG Code and the dividend payout.
Concerning the relationship between CG quality and the dividend payout, several authors have found a positive relationship too. For instance, in the case of Africa, Abor and Fiador (2013) studied how good corporate governance practices influenced dividend policy in 525 firms from Ghana, Nigeria, Kenya, and South Africa. They found that better CG practices have a positive relationship with the firms’ dividend payout in all countries except for Nigeria where the relationship was negative.
Rodrigues et al. (2020) studied the relationship between CG quality and the dividend policy of firms from Continental European Countries, Great Britain, and Ireland. They found that several governance variables had a positive relationship with the firms’ dividend yields and dividend payouts.
Lin and Lin (2020) studied the influence of corporate governance quality on dividend policy using a sample of Canadian firms listed in the S&P/TSX Index. The results showed that better governance quality was related to higher dividend payouts and higher propensity to pay dividends.
Kanojia and Bathia (2022) contrasted the outcome versus the substitution hypothesis in Indian and US listed firms. They found evidence of the outcome model, so there is a positive relationship between strong (good) CG practices and the firms’ dividend payout.
Given the above our second hypothesis is that not only the voluntary adoption of the Country Code impacts the dividends policy, but also the quality of the CG practices influences positively the dividend payout, hence:
H2: There is a positive relationship between the quality of the CG Code and the dividend payout.
Finally, given the fact that in Colombia the Country Code was refined two times in 2007 and in 2014, we state the following hypothesis:
H3: There is a positive relationship between the Country Code’s refinements of 2007 and 2014 and the dividend payout.
3. Methodology
We collect information from 605 firms that report their financial information to the Superintendency of Companies in Colombia. Our data corresponds to January 1997 until December 2022. It is important to notice that from 1997 to 2006 there was no survey of corporate governance practices, and neither was the Country Code. The first version of the Country Code was designed between 2005 and 2006 and launched in 2007, while a refinement of this Country Code was launched in 2014.
The financial information comes from the website of the Colombian Financial Superintendency (CFS): financial statements, annexes, the ownership composition (up to the first 20 stockholders), dividends and all the CG information. To avoid self-selection bias, the total sample includes both companies that pay dividends and companies that do not, companies with CG Codes and without it. Of the 605 firms, 232 (38%) pay dividends and 292 (48%) have adopted a CG code.
In addition, we do not consider negative values except for GDP growth, EBIT and Profit and we do not consider outliers outside the 95% level of confidence. Furthermore, companies which have more than 25% of missing values during the entire period were dropped from the sample to avoid biased estimators. After removing these data, we obtained 3484 firm-year observations for the total sample.
In 2005, the Colombian Financial Superintendency (CFS) included several entities with the aim of defining a Good Corporate Governance Code (Country Code). The topics of the Code are: (a) General Shareholders’ Meeting, (b) Board of Directors, (c) Financial and non-Financial Information Disclosure and (d) Mediation. Moreover, the CFS established an annual survey to monitor the firms’ degree of compliance. The survey includes 80 questions, grouped by the previous four topics.
Our goal is to measure the impact of the CG Code adoption, quality, and the Country Code on the dividend payout. With respect to the CG Code adoption, it is a dummy variable and the firms self-declare to have adopted (1) or not (0) a CG Code during 1997-2006 and the Country Code during 2007-2022. With respect to the CG Code quality the situation is a bit complex because between 1997 and 2006 there was not a survey to measure its quality. Finally, to measure the impact of the Country Code in 2007 and its refinement in 2014 is straightforward with the aid of a dummy variables.
With respect to the CG Code quality, the survey started in 2007, so between 1997 and 2006 there was no survey and we do not have a rating to measure the quality of the CG Code. To measure the CG quality, we assume that 2007’s rating was constant for the years 1997-2006. This is because the ratings’ increase between 2007 and 2008 was very low. Table 1 compares the average rating for 2007 and 2008, the annual increase of these years, as well as the average rating (see Table 1).
This assumption means that the firms’ CG Code quality was constant during the years 1997-2007 and equal to their performance in 2007. As this assumption is arbitrary, we estimate our models with four different weights for the topics during the period 1997-2007 and we obtain the same significant results although with different coefficients (see Table 2). Finally, we kept the weight that provided us the best fitted regression.
Table 1 Comparison of the average ratings between 2007 and 2008 per topic
| Rating 1 | Rating 2 | Rating 3 | Rating 4 | |
|---|---|---|---|---|
| Year 2007 | 42.6% | 45.5% | 43.3% | 43.7% |
| Year 2008 | 45.2% | 49.0% | 45.8% | 46.4% |
| Rating Increase | 2.6% | 3.50% | 2.50% | 2.70% |
| Average | 43.9% | 47.2% | 44.5% | 45.0% |
Source: Own elaboration
Table 2 Alternative weights for the four topics considered for the period 1997-2007
| Topic | Questions | Rating 1 | Rating 2 | Rating 3 | Rating 4 |
|---|---|---|---|---|---|
| General Shareholders’ Meeting | 20 | 24.1% | 25.0% | 9.0% | 10.0% |
| Board of Directors | 37 | 44.6% | 25.0% | 60.0% | 75.0% |
| Information Disclosure | 23 | 27.7% | 25.0% | 30.0% | 10.0% |
| Mediation | 3 | 3.6% | 25.0% | 1.0% | 5.0 % |
Source: Own elaboration
All the remaining years (2008-2022) entered our models with the corresponding survey result to measure quality. Hence, we run four different regressions per model version, the first regression included the rating weights 1 from years 1997-2006 and then the assigned rating from surveys of 2007-2022; the second regression included the rating weight 2 for the firms during years 1997-2006 and then their assigned rating from the surveys of 2007-2022; and so on.
3.1 Variables
In the literature, there are several ways to measure the dividend policy: dividend payout, dividend yield, dividends/cash flow, dividends/sales, dividends/net profit, and so on. La Porta et al. (2000) recommend using dividends/sales for two reasons: 1) it does not depend on the accounting principles, which might vary among countries and 2) it is not easily manipulated by accounting practices. However, in this study those assumptions can easily be violated, in fact: 1) it is not cross-country research, so differences in the accounting principles do not affect our results and 2) although the Colombian Tax Authority (DIAN) makes efforts to avoid accounting manipulation, they are still quite far from a solution.
We consider also that the economic interpretation of the ratio dividend/sales is not clear. Due to these reasons we decided to use the dividends/net profit or dividend payout as our dependent variable. Furthermore, many studies used the same ratio (Carvalhal-da-Silva and Leal, 2003).
To measure the adoption of the Code we use a dichotomous variable Dycode that takes value one (1) since the year of adoption of the Code and zero (0) before it. We also include an interaction term between the adoption of the Code and the quality of it measured by the total rating of each company and we name this variable Dum_Rat. Finally, to measure the impact of the Country Code we include two dummy variables CC2007 and CC2014 that take the value of 1 on this year and afterwards and zero before it.
The Herfindahl- Hirschman Index (HER) is defined as follows Cubbin & Leech (1983).: N is the number of shareholders, H is the concentration index, Si the number of shares owned by shareholder i, T N the total number of outstanding shares owned by N shareholders.
In Colombia, around 74% of the ownership is concentrated in hands of the first five shareholders, we did consider the ownership concentration up to the fifth (5) shareholder:
Other control variables are profitability, measured through the Return on (Total and Financial) Assets (ROA and ROAf), the Operational Margin (EBIT), the Net Margin (PROFIT), two different measures of leverage, total (LEV) and financial (LEVf), two proxies for the size, (SIZE) and tangibles (TANG), and the Gross Domestic Product growth as a macro control variable (GDP) (see Table 3).
3.2 Descriptive statistics
Table 4 offers the descriptive statistics of the total sample of 605 companies and the sub samples of companies with and without a Code (292 and 313), and the ones that pay dividends (232 companies). Although not reported, the dividend payout ratio is positive for 51% of the total sample. The average dividend payout was 27.70% for the total sample and 48.5% for the dividend paying firms. It is interesting to note the huge difference in the average dividend payout for the firms with and without a CG Code. Here, the ones with CG Codes pay twice as many dividends (41.5%) than firms without the CG Code (21.50%).
In Latin America firms present high ownership concentration levels from the first to the fifth shareholder. Specifically, 53% of the ownership is concentrated by the first shareholder, 73% up to the third shareholder and 79% up to the fifth for the Latin American region; while 44% for the first, 65% up to the third and 73% up to the fifth shareholder for Colombia (Fuenzalida et al., 2008; Gutiérrez et al., 2005; Pombo and Gutiérrez, 2007). Although not reported in Table 4, we encountered a similar pattern in the ownership concentration in our sample of 605 firms in Colombia: 49% for the first, 71% the third and 75% the fifth.
Table 3 Estimation of variables
| Variable | Symbol | Measure |
|---|---|---|
| Dividend Payout | Dpay | Dividends/Net Profit |
| Country Code 2007 | CC2007 | Dichotomous variable |
| Country Code 2014 | CC2014 | Dichotomous variable |
| Gross Domestic Product | GDP | Ln(GDPt/GDPt-1) |
| Return On Assets | ROA | Net Profit/ Total Assets |
| Return On Financial Assets | ROAf | Net Income/ External Financial Needs |
| Total leverage | LEV | Liabilities/Assets |
| Financial leverage | LEVf | External Financial Needs/Total Assets |
| Tangibility | TANG | Fixed Assets/Total Asset |
| Sales | SIZE | Ln(Sales) |
| Operating profit | EBIT | Operating Profit/Sales |
| Net Profit | PROFIT | Net Profit/Sales |
| Ownership Concentration | HER | Equation (2) |
| Adoption of the Code | Dycode | Dichotomous variable |
| Quality of the Code | Rat | Rating |
| Interaction between the Adoption of the Code and its Quality | Dum_Rat | Dycode x Rat |
Source: Own elaboration
Table 4 Descriptive Statistics
| A. Total Sample (605) | C. Subsample with CG Code (292) | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Variable | Mean | Std Dev | Min | Max | Variable | Mean | Std Dev | Min | Max | |
| year | 1997 | 2022 | year | 1997 | 2022 | |||||
| Dpay | 27.70% | 47.50% | 0.00% | 310.15% | Dpay | 41.50% | 37.50% | 0.00% | 303.50% | |
| CC2007 | 0 | 1 | CC2007 | 0 | 1 | |||||
| CC2014 | 0 | 1 | CC2014 | 0 | 1 | |||||
| GDP | 3.36% | 3.56% | -7.25% | 11.01% | GDP | 3.36% | 3.56% | -7.25% | 11.00% | |
| ROA | 2.70% | 21.35% | 0.10% | 57.00% | ROA | 4.50% | 7.40% | 0.80% | 57.40% | |
| ROAf | 1.75% | 20.70% | 0.05% | 48.30% | ROAf | 4.07% | 7.15% | 0.12% | 48.34% | |
| LEV | 35.50% | 51.45% | 0.00% | 65.67% | LEV | 38.20% | 22.40% | 0.54% | 90.34% | |
| LEVf | 27.20% | 72.65% | 0.00% | 125.58% | LEVf | 24.70% | 20.10% | 0.00% | 93.56% | |
| TANG | 25.70% | 22.30% | 0.00% | 95.60% | TANG | 21.50% | 21.10% | 0.00% | 91.10% | |
| SIZE | 13.4 | 311% | 0 | 21.1 | SIZE | 14.7 | 189.00% | 7 | 17.9 | |
| EBIT | 1.30% | 142.70% | -3915% | 100% | EBIT | 8.40% | 157.00% | -2534.00% | 98.80% | |
| Profit | 10.40% | 299.10% | -4530% | 7750% | Profit | 7.50% | 166.00% | -2589.00% | 415.00% | |
| Dycode | 46.50% | 47.60% | 0.00% | 100.00% | Dycode | 67.50% | 45.70% | 0.00% | 100.00% | |
| Her | 30.70% | 30.10% | 0.40% | 100.00% | Her | 36.60% | 33.25% | 0.75% | 100.00% | |
| Rat | 56.70% | 20.50% | 14.50% | 100.00% | Rat | 61.60% | 14.50% | 100.00% | ||
| B. Subsample without CG Code (313) | D. Subsample that paid dividends (232) | |||||||||
| Variable | Mean | Std Dev | Min | Max | Variable | Mean | Std Dev | Min | Max | |
| Year | 1997 | 2022 | year | 1997 | 2022 | |||||
| Dpay | 21.50% | 35.50% | 0.00% | 195.70% | Dpay | 48.50% | 43.60% | 0.00% | 299.10% | |
| CC2007 | CC2007 | 0 | 1 | |||||||
| CC2014 | CC2014 | 0 | 1 | |||||||
| GDP | 3.36% | 3.56% | -7.25% | 11.01% | GDP | 3.36% | 3.56% | -7.25% | 11.01% | |
| ROA | 1.50% | 24.30% | 0.10% | 57.00% | ROA | 5.70% | 6.35% | 0.86% | 45.76% | |
| ROAf | 1.55% | 26.80% | 0.05% | 48.30% | ROAf | 5.50% | 8.57% | 0.84% | 43.56% | |
| LEV | 35.40% | 65.40% | 0.00% | 65.67% | LEV | 30.20% | 22.40% | 0.00% | 93.50% | |
| LEVf | 25.40% | 86.40% | 0.00% | 125.58% | LEVf | 17.50% | 18.50% | 0.00% | 90.05% | |
| TANG | 24.30% | 21.20% | 0.00% | 92.50% | TANG | 23.50% | 20.60% | 0.00% | 90.40% | |
| SIZE | 11.1 | 245% | 0 | 15.4 | SIZE | 13.5 | 205.10% | 6.1 | 18.1 | |
| EBIT | -5.20% | 165% | -3915% | 100% | EBIT | 15.60% | 125.30% | -2415.00% | 100.00% | |
| Profit | 11.80% | 370.10% | -4436% | 7468% | Profit | 15.80% | 165.30% | -2489.00% | 3017.30% | |
| Dycode | Dycode | 47.30% | 44.70% | 0.00% | 100.00% | |||||
| Her | 29.80% | 29.40% | 0.30% | 100.00% | Her | 24.30% | 27.40% | 0.00% | 100.00% | |
| Rat | Rat | 58.30% | 19.40% | 13.50% | 100.00% | |||||
Source: Own elaboration
This Table offers the descriptive statistics (mean, standard deviation, maximum and minimum values) for four subsamples: the first subsample (A) includes the total number of companies; the second subsample (B) corresponds to all the companies without a CG Code; the third subsample (C) includes companies who have adopted the CG code, and the last subsample (D) includes all companies that have paid dividends at least once during the period (1997-2022).
With respect to the CG Code quality, we used the rating, and this only applies to those firms that adopted the CG Code and presented the survey from 2007 until 2022. These companies got an average rating of 77.6% with 33.6% as the lowest and 100.0% as the highest qualification.
There is a relevant difference in terms of dividend distribution. Companies that pay dividends had, on average, a rate that is 23.9 percentage points higher, achieving a value of 78.1% during the period 2007-2022. In 2008, out of 71 firms that presented the survey, 31 improved their rating with respect to 2007 and only 12 kept the same one of 2007. Meanwhile, 7 companies had decreased their rating from 2007 to 2008 by 13.3%.
The general leverage ratio is 35.5% and it is similar for companies with (38.2%) and without a Code (35.4%). This evidence is in contrast with the one founded by Benavides (2009), who found a positive relationship between debt and CG Codes. This might be explained by thinking that the adoption of the CG Code in Colombia does not reflect higher leverage possibilities. However, there is an important difference between companies that pay dividends (30.2%) and companies that do not (54.1%).
In relation to the firms’ size those with a CG Code are 27% bigger than firms without a CG Code and, the ones that pay dividends are 22% bigger than firms without a CG Codes. In general, the growth in sales that the firms of the sample had in the period has been steady.
The ROA, EBIT and profit are affected by the low involvement that Colombian firms had during the period 1997-2002 probably due to the economic trends of these years. However, there are still some important observations. The ROA of the firms with a CG Code is higher (4.5% against 1.5%), and firms that pay dividends follow a similar pattern (5.70% against -2.7% of firms that do not pay dividends). Moreover, the 232 companies that pay dividends have double the EBIT of those that do not pay. Among the latter ones, there is a big difference between those with a CG Code (that achieved a margin of 8.5%) and those without (-23.3%). Besides, companies with a CG Code have a lower profit (7.5%) than those without a CG Code (11.8%).
4. Empirical results
4.1 Total sample (605 firms)
In this section we discuss our main results for the total sample of 605 firms during the period 1997-2022. First, we describe all the analysis we did for all regressions reported in Table 5 and then we discuss our results.
The error structure U it in panel data models tend to be correlated through time for a given individual, hence we correct the model specifying firm's cluster for the standard errors, which produces consistent estimators when the errors are not identically distributed between the panels or there is serial correlation on it (Cameron and Trivedi, 2009).
We applied the Breusch & Pagan Lagrange multiplier test to the error component model, whose null hypothesis is that the variance is equal to zero (0) for the error component that doesn't vary across time but across individuals, i.e., Var(α i )=0. The null hypothesis was strictly rejected so it was necessary to model for the non-observed heterogeneity. We also test for the presence of contemporary and serial correlation. Our results led to a rejection of the null hypothesis, so it was necessary to correct the model for heteroscedasticity, serial correlation, and contemporary correlation.
There are two methodologies that help to correct for serial correlation in panel data: Feasible Generalized Least Square (FGLS) and Panel Corrected for Standard Errors (PCSE). Because our sample is composed of 605 firms and 26 years, we should not use the FGLS estimators because it is appropriate for samples where T is three times as large as N or more (Beck and Katz, 1995; Beck, 2001). We ran the Hausman test for all regressions resulting in a strong rejection of the null hypothesis of the random effects' estimators and, therefore, we chose the fixed effect model. We also test for the presence of heteroscedasticity in the residuals for a fixed-effect regression through a Modified Wald Test for Fixed-effect Models. The result was strict rejection of the null hypothesis, according to which the variance is equal for all the individuals. The advantage of this test is that it works well for unbalanced panels, even when the normality assumption is violated.
We also measure the impact of including time effects, a dummy variable for the year, to model observable characteristics that do not vary among individuals but across time. The joint significance test of these variables was rejected, so it was not necessary to include them in the models. Finally, we analyze our data according to four different samples to deal with the risk of having the incidental parameters problem: total sample, the subsample of firms without CG Code, the subsample of firms with the CG Code and the subsample of firms that paid dividends with or without the CG Code.
In all regressions we considered only the independent variables that were significant at least in one regression except for the Country Code dummies (CC2007 and CC2014). The models we applied for the total sample data are the following ones (see Table 3 for the definition of variables):
with i= 1, …, N and t= 1, …, T
Columns 1 and 2 in Table 5 show the results of the previous two models. In both cases we estimate the models with 605 firms during the period 1997-2022 and we report the best fitted regression with respect to the possible weights given in Table 2 for the period 1997-2006. It is important to note that variables ROA, LEVf, EBIT and PROFIT were not significant in any of the regressions, so we excluded them from the analysis.
Table 5 Results of the Panel Data Regression Models
| Variables | Model 1 | Model 2 | Model 3 | Model 4 | Model 5 | Model 6 | Model 7 |
|---|---|---|---|---|---|---|---|
| (1) | (2) | (3) | (4) | (5) | (6) | (7) | |
| CC2007 | 0.05 | 0.04 | 0.10 | 0.11 | 0.07 | 0.08 | |
| (0.05) | (0.04) | (0.05) | (0.06) | (0.10) | (0.10) | ||
| CC2014 | 0.1 | 0.11 | 0.18 | 0.16 | 0.15 | 0.16 | |
| (0.07) | (0.08) | (0.05) | (0.05) | (0.11) | (0.11) | ||
| GDP | -0.413*** | -0.451*** | -0.512*** | -0.453** | -0.443** | -0.511** | -0.497** |
| (-3.01) | (-3.25) | (-2.75) | (-2.51) | (-2.43) | (-3.43) | (-3.47) | |
| ROAf | 0.068* | 0.071* | 0.063** | 0.156 | 0.175 | 0.201* | 0.211* |
| (1.18) | (2.03) | (2.53) | (0.57) | (0.61) | (1.51) | (1.72) | |
| LEV | -0.029* | -0.017* | -0.001 | -0.313** | -0.321*** | -0.201* | -0.213* |
| (-1.71) | (-1.72) | (-0.72) | (-2.02) | (-2.34) | (-1.32) | (-1.34) | |
| TANG | 0.161** | 0.172*** | 0.131* | 0.435*** | 0.436*** | 0.245*** | 0.251*** |
| (2.35) | (2.24) | (1.85) | (4.01) | (3.99) | (2.12) | (2.18) | |
| SIZE | 0.047*** | 0.048*** | 0.03*** | 0.051*** | 0.057*** | 0.064*** | 0.065*** |
| (11.41) | (9.85) | (5.35) | (5.53) | (5.75) | (6.32) | (6.39) | |
| HER | -0.173*** | -0.162*** | -0.231*** | -0.011* | -0.015* | -0.211*** | -0.223* |
| (-4.11) | (-4.21) | (-3.11) | (-1.19) | (-1.21) | (-3.31) | (-3.41) | |
| Dycode | 0.072** | 0.051* | 0.08*** | ||||
| (3.11) | (1.81) | (3.85) | |||||
| Dum_Rat | 0.331*** | 0.099* | 0.491*** | ||||
| (3.68) | (1.91) | (3.91) | |||||
| Constant | -0.311*** | -0.256*** | -0.225*** | -0.09 | -0.11 | -0.314*** | -0.323*** |
| (-5.81) | (-4.91) | (-3.75) | (-0.91) | (-0.88) | (-6.38) | (-6.41) | |
| Observations | 3484 | 3484 | 2090 | 2021 | 2021 | 1834 | 1834 |
| Wald Chi2 | 249.11*** | 152.04*** | 65.49*** | 65.05*** | 64.80*** | 74.34*** | 75.42*** |
| R-Squared | 0.25 | 0.27 | 0.22 | 0.28 | 0.31 | 0.26 | 0.28 |
Robust standard errors in parenthesis, *** p<0.01, ** p<0.05, * p<0.1
Source: Own elaboration
The dependent variable of the model is the dividend payout ratio (Dpay) and the independent variables are the ones reported in Table 3. Model 1 and Model 2 consider the total sample, while the first one calculates the impact of the adoption of the Code (Dycode) and the second one measures the joint effect of the adoption of the Code and its quality (Dum_Rat). Model 3 considers the subsample of companies that haven’t adopted the Code yet. Models 4 and 5 refer to the subsample of companies with the Code (adoption and quality) and models 6 and 7 refer to the subsample of companies that paid dividends without and with the code (adoption and quality)
In model 1, we test for the adoption of the CG Code (Dycode) (hypothesis 1) and for the Country Code dummies (hypothesis 3); while in model (2) we test for the quality of the CG Code (DumRat) and the Country Code dummies (hypothesis 3). Our coefficients show that the impact of CG Code adoption, on average, is 3.3% (7.2% × 46.50%), so the adoption of the CG Code will increase the dividend payout by 3.3%. The impact of the CG quality on the dividend payout is far more important, on average, 18.8% (33.1% x 56.7%). That is, the positive impact of the CG Code adoption if magnified by the quality of the CG Code by more than 15%.
Our results for the total sample are highly significant for the adoption of the CG Code (Dycode in Model 1) and the CG Code quality (DumRat in Model 2), but the two Country Code dummies were not significant (CC2007 and CC2014). Hence, overall, we find support for our first and second hypotheses, but not for the third one. It seems that the decision to adopt a CG Code was not related to the launch of the Country Code in 2007 and neither to its refinement in 2014.
It is also interesting to note that the dividend payout is related positively to the profitability (ROAf) and size (SIZE and TANG) of the firms and negatively related to the GDP growth, leverage (LEV) and ownership concentration (HER). Hence, the larger and more profitable the firm is, the higher the dividend payout and the less indebted and the less concentrated is the higher the dividend payout.
The inverse relationship between ownership concentration and the dividend payout might be related to the expropriation risk that minority shareholders are facing in emerging economies with weak shareholders' protection. This tendency is disadvantageous for minority shareholders in highly concentrated companies because majority shareholders can get cash flows for their own benefit using means that are different from dividend distribution. The magnitude of this coefficient becomes one of the most relevant of the models, being the one of highest impact among control variables.
Besides, the GDP has a negative coefficient, a fact that might be counterintuitive if not considering that, to measure the impact of the variable, it is necessary to bring the function back to levels that are comparable to the Dpay, i.e., to apply the reverse logarithm function used to calculate the GDP, after which the expected positive relation is evident.
4.2 Subsample without CG Code (313 firms)
In this section we test for the determinants of the dividend payout for the subsample of companies that did not adopt a CG Code (see Model 3 in Table 5). We test the following model (see Table 3 for the definition of variables):
In general, we obtain the same results for the GDP growth, ROAf, TANG, SIZE and HER except for the variable leverage (LEV) that seems not to have a role in the dividend payout for this sample.
Another interesting result is the higher impact of the ownership concentration on the dividend payout. On average, an increase of 1% in the ownership concentration will decrease the dividend payout by -6.9% (-23.1% × 29.8%) as opposed to the decrease by -5.3% (-17.3% × 30.7%) according to the total sample. Hence, ownership concentration becomes an important indicator for this subsample of firms.
4.3 Subsample with CG Code (292 firms)
In this section we test our three hypotheses in the subsample of firms that have adopted a CG Code. We test the same models as equations (3) and (4) and our results are shown in columns 4 and 5 in Table 5, respectively.
From Table 5, we can see that the effect of the CG Code adoption is nearly the same as with the total sample 3.4% (5.1% × 67.5%), but the effect of the CG Code quality is lower than with the total sample 6.1% (9.9% × 61.6%). Hence, the CG Code quality remains more important than the mere CG Code adoption, but quality is far less important for this subsample than for the total sample of firms. The Country Code dummies (2007 and 2024) are also not significant.
It is very interesting to notice that profitability doesn’t matter for the dividend payout, the ROAf is not significant in both regressions (columns 4 and 5) meaning that the CG Code adoption generates a firm’s “commitment” to keep its dividend payout. It is also remarkable the decrease of the negative impact of the ownership concentration upon the dividend payout from -5.3% (-17.3% × 30.7%) for the total sample to -0.4% (-1.10% × 36.6%) for this subsample. That is, the adoption of the CG Code nearly wipes out the ownership concentration impact.
The CG Code quality is also having a lower impact on the dividend payout than in the total sample. The positive impact of CG Code quality goes down from 18.8% (33.1% x 56.7%) to 6.1% (9.9% × 61.9%), so quality is less important for the subsample of firms that adopted the CG Code than for firms in the total sample. The firm size seems to have a more important role than for the total sample, the impact of the variable TANG more than doubles from the total sample 4.1% (16.1% × 25.7%) to this subsample 9.4% (43.5% x 21.5%). Hence, larger firms with a CG Code have larger dividend payouts than firms in the total sample.
The impact of leverage for this subsample is also more important because an increase of 1% in leverage will decrease the dividend payout by -12% (-31.3% × 38.2%) as opposed to the decrease in -1% (-2.9% × 35.5%) for the total sample.
4.4 Subsample of positive dividend payout ratio (232 firms)
Finally, we test out three hypotheses in the subsample of firms that distribute dividends during the years 1997-2022 (see columns 5 and 6 in Table 5). We use the same equations (3) and (4) to test these hypotheses.
In this subsample is where we find the strongest impact of the CG Code adoption 3.8% (8% x 47.3%) as compared to the subsample with CG Code 3.4% (5.1% × 67.5%) and the total sample 3.8% (7.2% × 46.5%). In other words, firms with a CG Code pay dividends more regularly than firms in the other groups.
Another important result is that CG Code quality matters more for dividend paying firms than firms in the other subsamples. The impact of CG Code quality goes from 6.1% for firms with a CG Code to 28.6% (49.1% × 58.3%) for dividend paying firms, this result is even higher than 18.8% for firms in the total sample. Hence, firms with a high rating in the survey offer a higher dividend payout ceteris paribus. The Country Code dummies are not significant again.
The effect of leverage is half of the effect of the previous subsample because it goes down from -12% to -6.1% (20.1% × 30.2%), so the risk of cutting dividends due to a higher leverage diminishes. The effect of firm’s size also decreases from 9.4% (previous subsample) to 5.8% (24.5% × 23.5%), so firms with smaller sizes can distribute dividends.
The ownership concentration becomes important again with respect to the previous subsample because it goes up from -0.4% to -5.1% (-21.1% × 24.3%). This effect is nearly the same as the impact of -5.3% (-17.3% × 30.7%) in the total sample, hence the CG Code adoption matters.
5. Conclusions
We find compelling evidence of the positive impact of the CG Code adoption into the firm’s dividend payout for the total sample as well as for the sample of firms that adopted the CG Code and for the sample of the dividend paying firms. The CG Code adoption is important for the three samples, but especially for the dividend paying firms. We also conclude that CG Code quality has an important impact on the dividend payout because and increase in the rating would yield an increase in the dividend payout of 18.8% for the total sample and 28.6% increase for the dividend paying firms.
We were unable to find a significant impact of the two Country Code dummies (2007 and 2014), so we conclude that these two efforts from the Colombian government while helped to structure the CG survey it didn’t impact the dividend payout of the firms. Hence, it is far more important the firm’s conviction to adopt a CG Code rather than a regulation.
Our findings are in line with the results of other authors such as Kowalewski et al. (2007), Bebczuk (2005) and La Porta et al. (2000). Our results are robust in the sense that we conduct the same regressions in three samples. In relation to our fourth sample of firms that didn’t adopt a CG Code we conclude that ownership concentration act as a signaling instrument to shareholders, so the higher ownership concentration is the lower will be the dividend payout. Besides, it is very likely that these firms do not have a well-established dividend policy and/or they distribute dividends according to the Colombian Law.
The firm’s size and leverage are two important features but especially for firms that adopted the CG Code and probably some of them do not distribute dividends. Profitability is important for firms that didn’t adopt a CG Code and for dividend paying firms, while ownership concentration is important for firms that didn’t adopt a CG Code and dividend paying firms.
A couple of implications appear from our analysis: firms that adopt a CG Code and pay dividends will lower the negative effect of leverage and ownership concentration into the dividend payout, and the Colombian Superintendencies should focus more on fostering the adoption of CG Codes beyond regulating the disclosure of its compliance.
Clearly higher ownership concentration signals the possibility of lower dividend distributions, so the adoption of a CG Code signals the opposite and may help to diminish the negative impact of the ownership concentration on firm value. Another negative signal comes from leverage, so firms that adopt distribute dividends more regularly must be able lower the negative effect of leverage upon firm value too.
The Colombian Superintendencies of listed and non-listed firms are putting efforts to integrate the Country Code with the Colombian regulation and to monitor the firms’ effort to comply with the Country Code and disclose the information to investors. These efforts are good, but not enough to entice new companies to adopt the Country Code and to disclose what would be the possible impact of the implementation and compliance to good corporate governance practices in the firm’s value. It would be interesting to test the same hypotheses in different industries and in relation to the CG Board characteristics and these will provide more and interesting evidence of the different policies that may be applied to foster the adoption of the Country Code in Colombia.
