An Empirical Analysis of the Relationship between Cash Flow and Dividend Changes in Nigeria

Transcription

1 An Empirical Analysis of the Relationship between Cash Flow and Dividend Changes in Nigeria Olatundun J. Adelegan* The purpose of this study is to re-evaluate the incremental information content of cash flows in explaining dividend changes, given earnings. I carry out an 882 firm-year study by analysing the dividend changes cash flow relationship on a sample of 63 quoted firms in Nigeria over a wider testing period from 1984 to Despite the fact that I used a wider testing period than previous studies and more refined cash flow measures than previous studies, I also introduced dummy variables to capture economic policy changes in the economy. The association of cash flows with dividend changes is tested using the modified Lintner Brittain model as adopted in Charitou and Vafeas (1998) on pooled cross sectional/ time series data from the full sample of observations from The models are estimated using the ordinary least squares (OLS) method and I do find a significant relationship between dividend changes and cash flow unlike previous studies. The empirical results reveal that the relationship between cash flows and dividend changes depend substantially on the level of growth, the capital structure choice, size of each firm and economic policy changes. 1. Introduction The association between earnings and dividend changes has been established for the past four decades (Lintner, 1956). Subsequent research attempts to study the relationship between cash flow and dividend changes, given earnings, has not been successful (Hagerman and Huefner, 1980, Crum et al., 1988; Simons, 1994; Charitou and Vafeas, 1998). Notwithstanding the outcome of previous studies, two reasons have been advanced to explain the superiority of cash flow over earnings in explaining dividend changes. First, managers may manipulate earnings to maximize their bonus awards (Healy, 1985) or to side step restrictive debt covenant violations. The fact that accrual components of earnings can be manipulated makes the cash flow component a more reliable indicator of corporate performance than the accrual component. Secondly, cash flows are a more direct measure of liquidity and liquidity is likely to be a contributing factor in setting dividend policy. Therefore, even if there is no difference between the market valuation of accruals and cash flows in measuring corporate performance, cash flows are expected to be more useful than accruals in determining dividend changes. Prior studies concentrate on developed markets and have not been able to find a significant relationship between dividend changes and operating cash flows. However, because of differences in institutional contexts, the empirical relationship between dividend changes and cash flow in an emerging market like Nigeria may be different from what obtains in the developed market. This study therefore carries the debate to Nigeria. An 882 firm-year study was carried out by analysing the dividend changes cash flow relationship on a sample of 63 quoted firms in Nigeria over a wider testing period from 1984 * Lecturer in Accounting and Finance, Department of Economics, University of Ibadan, Nigeria and Research Fellow, Centre for Econometrics and Allied Research (CEAR), University of Ibadan, Nigeria; adelegan@ pop.skannet.com R&D Management 15, 1, # Blackwell Publishing Ltd, Published by Blackwell Publishing Ltd Garsington Road, Oxford OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA.

2 O.J. Adelegan to In addition to using a wider testing period than previous studies and more refined cash flow measures than previous studies, dummy variables were also introduced to capture economic policy changes in the economy. Moreover, drawing from the traditional finance theory Jensen (1986) and Charitou and Vafeas (1998) I hypothesize that the relationship between cash flows and dividend changes depend on each firm s growth opportunities. I also improved on prior research by suggesting two additional factors that potentially mitigate the dividend changes and cash flow relationship, namely capital structure and size. I therefore partition the data into three classes each based on growth opportunities, level of gearing and firm size. The empirical results reveal that the relationship between cash flow and dividend changes depend substantially on the level of growth, the capital structure choice and size of each firm and economic policy changes such as the Structural Adjustment Programme. The rest of the paper is organized as follows. The next section gives a background description of the Nigerian environment, followed by a review of literature and a theoretical development of the four hypotheses. The fifth section describes the methodology and the sample. The penultimate section contains the analyses and results, followed by the conclusions of the study in the final section. 2. Background Description of the Nigerian Environment In Nigeria, the capital market is an engine of growth which is still classified by the International Finance Corporation as underdeveloped and emerging (Inanga, 1999). The thinness of trading, low market capitalization, low percentage of turnover level and illiquidity of the market can be attributed to barriers to foreign investors, imposition of price caps on share price movement and political instability among others (Inanga and Emenuga, 1997; Ogwumike and Omole, 1997). The decision between paying dividend and retaining earnings as a source of further investment has been taken seriously by both investors and management and has been the subject of considerable research and scrutiny by equity analysts and economists in developed nations. As a result, a fairly detailed, if incomplete picture of dividend policy is available for the major developed countries (Glen et al., 1995). However, with the exception of Izedonmi and Eriki (1996) who used data from 1984 to 1989, studies of dividend policy on Nigerian corporate firms have been limited and clustered around , which is the period surrounding the first indigenization decree (Uzoaga and Alozieuwa, 1974; Inanga, 1975, 1978; Soyode, 1975; Oyejide, 1976; Odife, 1977). There has been several economic policy changes since then, such as the Structural Adjustment Programme (1986), the Second Indigenization Decree (1989) and the Investment Promotion Decree (1995). This study carries the dividend debate further, with a view to determining whether or not the dividend policy of corporate firms in Nigeria conform with the acceptable norms in developed countries. Furthermore, the size of dividend payments relative to corporate profit after tax, are very important because they negatively affect the level of business savings, which is in turn, an important source of capital formation and one of the most critical determinants of a country s economic performance. An understanding of the issues that drive dividend decisions of the corporate firms will enhance a proper evaluation of the private sector s contribution to Nigeria s economic performance. This will assist in furnishing the government with the required information around which appropriate policy can be designed to guide the pattern and process of its economic growth (Oyejide, 1976; Ariyo, 1983). For years, it has been over-emphasized to developing countries that financial liberalization is necessary for prosperity. Instead of discouraging foreign investors, they were advised to open up so as to have access to global savings that can be invested to accelerate their growth. The Nigerian capital market is now more liberalized than before. The bottlenecks to inflow of foreign capital such as the Nigeria Enterprise Promotion Decree (1989), the Exchange Control Act (1962) and the Capital Gains Tax have been abolished, and the Nigeria Investment Promotion Decree 16 and Foreign Exchange (Monitoring and Miscellaneous Provision) Decree 17 (1995) was promulgated. Nigeria is now open to international capital, and dividend policy has now increased in importance in the post economic liberalization period. Restrictions on foreign ownership of shares of limited liability companies registered and/or incorporated in Nigeria have been abolished. It is therefore imperative for corporate 36 R&D Management 15, 1, 2003 # Blackwell Publishing Ltd, 2003

3 Cash Flow and Dividend Changes in Nigeria firms to pursue sound dividend policies in order to attract and retain local and foreign investors. Foreign and local equity holders and managers are now more concerned about their dividend policy now than they were in the past. This is further augmented by the increased role of the government, which acts as a protector of both minority shareholders and creditors in dividend policy decisions (ss of the Companies and Allied Matters Decree 1990). This study, therefore, helps to educate foreign and local investors and economists about the dividend policy of Nigerian companies and through that process, promote further interest in the Nigerian stock market. 3. Literature Review Dividend payment modelling work starts from Lintner s (1956) ground-breaking study, which argued that the main determinants of changes in dividend are current earnings and preceding dividend level. The model further indicates that year to year changes in dividend payments are explained by current earnings levels in conjunction with a target pay-out ratio. However, yearly adjustment towards the target pay-out ratio is partial considering the reluctance of management to reduce dividend. Prior attempts to determine the association between dividend changes and cash flows have not been successful (Fama and Babiak, 1968; Hagerman and Huefner, 1980; Crum et al., 1988; Simons, 1994; Charitou and Vafeas, 1998). Using Lintner s model and Lintner Brittain modified versions, Fama and Babiak (1968) and Hagerman and Huefner (1980) observed significant empirical support for Lintner s model. They conclude that historical cost is a better predictor of dividend changes than cash flows and cash flows are found to be insignificant in predicting changes in dividend. However, their findings do not conclusively preclude the ability of cash flows to explain dividend changes because they define cash flow as income plus depreciation. This measure has been shown to be a proxy for profitability and not liquidity (Largay and Stickney, 1980; Gombola and Ketz, 1983; Bowen et al., 1986; Charitou and Vafeas, 1998). Simons (1994), Crum et al. (1988) and Charitou and Vafeas (1998) analysed the dividend changes cash flow relationship using more refined cash flow measures. Crum et al. (1988) conclude that the most important predictors of dividend changes are prior year dividends, current net income plus depreciation and working capital from operations. Simons (1994) studied the dividend cash flow relationship by isolating firms for which the relationship between dividend changes and cash flow is weak, but none of the three liquidity measures proposed in the study had an association with changes in dividend, given earnings, and he concluded that the relationship between dividend changes and cash flow remains elusive. Researchers in the United Kingdom have also advocated support for cash flow reporting because it avoids arbitrary allocation of funds and are therefore useful to users of financial statements for estimating future dividend flows (Lawson and Stark, 1975, 1981). Using aggregate data for German companies, Lawson and Moeller (1996) challenge the view that historical cost retained earnings constitute internally generated finance and opined that periodic changes in earnings may not necessarily be accompanied by an equal liquidity change. Notwithstanding the fact that Lawson (1996) obtained empirical evidence to show that dividend policies are based on accrual earnings, he suggests that such policies are not consistent with an ex ante shareholder value creation (SVC) model because organizations should invest in the projects with positive net present values (NPV) and consider firm liquidity in attempting to maximize firm value. This view was upheld in Lawson and Stark (1981) and Lee (1983). Appropriation of after-tax earnings of corporate firms did not receive any serious attention among academic scholars in Nigeria until Uzoaga and Alozienwa (1974) highlighted the pattern of dividend policy pursued by Nigerian firms, particularly since and during the period of indigenization and participation defined in the decree (i.e., ). They claimed that dividend levels were raised by corporate firms in the wake of indigenization and they checked but found very little evidence to support the classical influences that determine dividend policies in Nigeria during those periods. Inanga (1975) and Soyode (1975) challenged the fear and resentment theory put forward in the study of Uzoaga and Alozienwa. Among others, Inanga (1975, 1978) identified under-pricing of the new issues of companies affected by the indigenization decree by the Capital Issue Commission as a major contributory factor to the significant upward change in the rates and level of dividend distribution # Blackwell Publishing Ltd, 2003 R&D Management 15, 1,

4 O.J. Adelegan by affected companies shortly before the [indigenization] decree was implemented. Soyode finds excess liquidity as a result of cash inflows from Nigerianized shares responsible for the high dividend pay-out during the indigenization decree. Nevertheless, Oyejide (1976) examined empirically the factors driving dividend policy of companies in Nigeria using Lintner s (1956) model as modified by Brittain (1964); he observed that there is substantial support in Nigeria for Lintner s model. Moreover, Odife (1977) criticized Oyejide s study for failing to adjust for stock dividend while Izedonmi and Eriki (1996) using data from 1984 to 1989 found support in Nigeria for Lintner s model. However, studies of dividend policy in Nigeria, though few, have resulted in mixed, controversial and inconclusive results. 4. Hypotheses Modern finance theory argues that the value of the firm depends on its stream of future cash flows. However, earnings are widely used by investors and creditors as a summary measure of firm performance, as well as in executive compensation contract and in debt covenant agreements. The primary appeal of earnings over cash flows is that they mitigate timing problems in revenue recognition and in matching revenues with appropriate costs in time. Cash flows are considered a noisier measure of firm performance than earnings because of timing and matching problems and accrual components of earnings is incrementally important in measuring firm performance. To the extent that firm performance determines dividend changes as in Lintner s model, both cash flows and aggregate accruals should be significantly associated with dividend changes (Charitou and Vafeas, 1998). However, while cash flows and accruals are important in explaining dividend changes, two plausible reasons exist to expect cash flows to be significantly more important than accruals. The first pertinent issue is that accruals are recognized partly at the discretion of the managers. Managers have incentives to manipulate accruals to their advantage because earnings are often used as a performance criterion in compensation contracts where managers have personal interests (Healy, 1985, Charitou and Vafeas, 1998). Therefore, manipulated accruals may be less important than cash flows in explaining dividend changes. Nevertheless, recent studies document mixed evidence as to the extent of accruals manipulation by managers (Dechow, 1994; Dechow et al., 1995; Gaver and Gaver, 1993). Secondly, it appears that pay-out policy is dependent on cash availability. An organization s decision to reduce, increase or maintain dividend partly reflects its liquidity position, therefore operating cash flow should reflect firm liquidity for cash flow to be a significant determinant of dividend changes, given earnings. These arguments suggest operating cash flows are likely to be a better predictor of dividend changes than accruals. Hypothesis 1 is as follows: H1: Operating cash flows are positively related to dividend changes, given earnings. Initially I argued that cash flows should be significant in setting dividend policy both as a performance and as a liquidity measure. Prior studies have rejected hypothesis 1 and these results stimulate further research on the issue. I replicate the tests documented in earlier empirical studies on Lintner s (1956) model. However, my empirical results are in contrast to the notion that the Lintner (1956) model and its variants explain the dividend level of corporate firms in Nigeria. This therefore points towards further testing on this issue. In particular, are there any conditions under which aftertax earnings, preceding dividend level and economic policy changes are significant determinants of dividend policy in Nigeria? Following prior capital market studies, my research approach on the assessment of the usefulness of the modified Lintner (1956) model in explaining dividends has been to use aggregate data, assuming that the relationship between these variables is homogenous across firms. As previously argued, the underlying assumption that investors react identically to earnings and preceding dividend level of all firms implying constant response coefficients is not realistic. Previous studies provided empirical evidence that the response coefficients are not constant, but they are affected by firm-specific, industryspecific and economic factors, such as firm size, industry classification, magnitude of accruals, quality of earnings and economic policy changes (Easton and Zmijewski, 1989; Collins and Kothari, 1989; Dechow, 1994; Charitou and Vafeas, 1998; and Adelegan, 2000). In line with this argument, I investigate contextual factors that are potentially important in mitigating the relationship between current 38 R&D Management 15, 1, 2003 # Blackwell Publishing Ltd, 2003

5 Cash Flow and Dividend Changes in Nigeria earnings, preceding dividend level and dividend yield. This study hypothesized that the homogeneity of the dividend policy equation across firms does not hold as a result of cross-firm variation in certain contextual factors such as firm growth, magnitude of debts (that is the level of leverage) and firm size. In this study I argue that (1) the cash flow dividend relationship is stronger for firms with moderate growth rates, (2) cash flows are a more important predictor of dividend changes when firms are highly levered and (3) cash flows are a better predictor of dividend changes for small-sized firms. My reasoning for (1) above is as follows. Future growth is expected to mitigate the relationship between dividend changes and cash flows, therefore the cash flow dividend changes relationship depends on the extent of future growth opportunities. As operating profits are generated, firms have to choose between reinvesting versus returning these funds to shareholders. Optimally, corporate firms will invest in all projects that have positive net present values and return a portion of the remaining after-tax earnings to shareholders in the form of dividends or share repurchases. The amount of money invested depends mainly on each firm s growth opportunities. High growth companies prefer to capitalize on their favourable investment prospects and have clear disincentives in paying operating cash flows and profits as dividend (Gaver and Gaver, 1993; Charitou and Vafeas, 1998; Adelegan, 2000). Managers in business firms with low growth may also prefer to invest after-tax earnings rather than paying additional dividends due to personal incentives (Jensen, 1986). Therefore we expect the relationship between profit after tax and dividend behaviour to be the strongest in moderate growth firms because these firms have lower opportunity cost of dividends than high growth firms. Furthermore, managerial and shareholder interests are more closely aligned, compared to these parties interests in low growth companies (Charitou and Vafeas, 1998). From the above discussion one can infer that cash flow coefficient is expected to be higher for moderate-growth firms. The research hypothesis to be tested is: H2: Cash flows are a better predictor of dividend changes for firms with moderate growth prospect, given earnings. Alternatively, cash flows are a more important predictor of dividend changes when firms are highly levered. A firm s ability to alter its dividend policy sometimes depends on its liquidity position. When there is adequate liquidity, a firm can set its dividend policy according to its performance. However, when cash flows are inadequate, the ability of the firm to change its dividend policy is constrained. Hence, when operating cash flows are low, cash flows are expected to play a significant role in setting dividend policy (Charitou and Vafeas, 1998). The relationship between after-tax earnings and dividend behaviour is expected to be significantly positive for business firms that are highly levered. Size has a positive influence on leverage and growth. A company that is highly levered has more debt obligations and interest to pay. Large firms have better access to debt (Marsh, 1982; Baskin, 1989; Chang and Rhee, 1990; Bennets and Donnelly, 1993; Charitou and Vafeas, 1998; Adedeji, 1998) and is likely to be less liquid when compared with small firms. Shareholders are likely to expect more dividends and the debtholders will also expect interest and the principal on the other hand. During periods of inflation, firms employ more debt in their capital structure because the real cost of debt falls. As more debt is introduced the company will be able to pursue more profitable ventures and the cost of equity will rise in response to an increase in debt. A highly levered firm will therefore be expected to change its dividend policy in line with its liquidity position coupled with its performance. We expect Lintner s model and structural and economic policy changes such as the Structural Adjustment Programme to be significant in explaining the dividend behaviour of highly geared firms. The research hypothesis to be tested is as follows: H3: The relationship between operating cash flow and dividend changes is significantly positive for firms that are highly levered, given earnings. Furthermore, the cash flow dividend changes relationship is conditioned on the size of firms. Size is expected to have a positive influence on leverage and growth. Large firms have better access to debt and should be highly geared on the one hand (Marsh, 1982; Baskin, 1989, Chang and Rhee, 1990; Bennets and Donnelly, 1993; Charitou and Vafeas, 1998; Adedeji, 1998). On the other hand, small firms pay more to issue new equity and also somewhat more to issue long-term debt. This suggests that small firms may be more leveraged than large firms and # Blackwell Publishing Ltd, 2003 R&D Management 15, 1,

6 O.J. Adelegan may prefer to borrow short term (through bank loans) rather than issue long-term debt because of the lower fixed cost associated with this alternative. Small-sized firms are therefore expected to have low cash flows because of repayment of short-term loans and interest and dividend to shareholders. Cash flows are therefore expected to be significant in setting their dividend policies, leading to the following research hypothesis: H4: Cash flows are a better predictor of dividend changes for small-sized firms, given earnings. This hypothesis is investigated by regressing dividend yield on the traditional model and its variants. Hypothesis 1 examines the usefulness of cash flows in explaining dividend changes using aggregate data in accordance with prior capital market studies assuming that the relationship between these variables is homogenous across firms. However, it has been argued that the underlying assumptions that investors react identically to earnings and cash flows of all firms are unrealistic. Empirical evidence that the response coefficient is affected by firm-specific, industry-specific and economic factors such as firm size, industry classification, magnitude of accruals and quality of earnings abound (Easton and Zmijewski, 1989; Collins and Kothari, 1989; Dechow, 1994; Charitou and Falas, 1996; Charitou and Vafeas, 1998; Adelegan, 2000). In line with this argument this study hypothesized that the homogeneity across firms does not hold due to cross-sectional differences in firm growth opportunities (hypothesis 2) and the magnitude of leverage (hypothesis 3) and firm size (hypothesis 4). In this paper, investigation of the dividend changes cash flow relationship was performed through the examination of four hypotheses: H1: Operating cash flows are positively related to dividend changes, given earnings. H2: Cash flows are a better predictor of dividend changes for firms with moderate growth prospect, given earnings. H3: The relationship between operating cash flow and dividend changes is significantly positive for firms that are highly levered, given earnings. H4: Cash flows are a better predictor of dividend changes for small-sized firms, given earnings. 5. Methodology 5.1 Data Sources, Nature and Scope The sample contains 63 companies quoted on the first and second tier securities market of the Nigerian stock exchange (NSE) as at There are currently 170 companies listed on the first tier market and 16 companies listed on the second tier securities market of the NSE. Companies listed on the first tier securities market are expected to have a trading record of at least five years, not less than 25 per cent shares in the hand of the public, and not less than 500 shareholders. Such companies are also expected to give quarterly, half-yearly and yearly reports and there is no limit to the amount of money that they can raise from the securities market. The second tier securities market was introduced in 1986 with less stringent conditions to accommodate more companies. Companies listed on the second tier securities market of the NSE are expected to have a track record of at least three years, not less than 25 per cent shares in the hands of the public, not less than 100 shareholders and they cannot raise more than 20 million naira from the securities market. Sample firms cover all sectors according to the NSE s classification, namely agriculture, automobile and tyre, banking, breweries, building materials, chemical and paints, conglomerates, commercial services, computer and office equipment, construction, engineering technology, food, beverage and tobacco, footwear, healthcare, industrial/domestic products, insurance, investment companies, machinery (marketing), packaging, petroleum (marketing), publishing and textiles. We have a full sample of 882 firmyear observations for the period 1984 to A firm is included in the sample if the following financial and market information necessary to estimate the various pooled cross-sectional/ time series models are available in the summarized annual reports in the Nigerian Stock Exchange Factbooks for 1984 to 1997: profit after tax (PAT), cash flow (CF), total distributable earnings (TDE), dividend per share (DPS), market price per share (MPS), market value of equity (MVE) and book value of assets (BVA) at the year end for the period 1984 to R&D Management 15, 1, 2003 # Blackwell Publishing Ltd, 2003

7 Cash Flow and Dividend Changes in Nigeria Data used in this study are mainly from secondary sources, namely the Nigerian Stock Exchange Factbooks of 1989/90 to 1998, annual reports of companies, and daily official lists of the NSE for the last day of trading in each of the years covered in the study. 5.2 Measurement of Variables The analysis focuses on the relationship between changes in dividend yield (DV) and profit after tax (PAT), cash flow (CF), total distributable earnings (TDE), and preceding year dividend per share (DVL). The variables are defined as follows: * Dividend changes (DV): current year dividend per share deflated by the end of current year market price per share minus prior year dividend per share deflated by the end of previous year market price. * Dividend par share (DPS): current year cash dividend plus current year stock dividend. * Profit after tax (PAT): profit after tax but before extraordinary items, discontinued operations, special and non-operational items divided by the end of the year market value of equity. * Cash flow (CF): working capital from operations deflated by the end of the year market value of equity. * Lagged dividend yield (DVL): previous year dividends deflated by the preceding year market value of equity. * Total distributable earnings (TDE): profit after tax plus distributable/revenue reserves. * Growth (market-to-book ratio, MBV): market value of equity divided by the book value of total assets. * Size (TA): natural logarithm of total assets deflated by the current year market value of equity. * Leverage (LEV): total debt deflated by the current year market value of equity. The regression models were structured using the ordinary least square (OLS) method. A market value deflator was used in the regression model because it avoids historical cost bias that is inherent in other deflators such as the book value of equity and total assets. Moreover, it is widely believed that dividend policy is driven by market performance; any study of changes in dividend policy should therefore be measured in the backdrop of each firm s market value (Christie, 1987; Kothari, 1992; Alford et al., 1993; Ali and Pope, 1995; Charitou and Vafeas, 1998). 5.3 Model Specification The following regression equations were estimated using the ordinary least squares (OLS) method to provide bearing on the hypothesis indicated above. DV it ¼b0 þ b1pat it þ b2dvl it þ: b3cf it þ b4tde it þ U it DV it ¼ b0 þ b1pat it þ b2dvl it þ U it DV it ¼ b0 þ b2dvl þ b3cf it : þ U it DV it ¼ b0 þ b1pat it þ b2dvl it þ b3cf it : þ U it Da vit ¼b0 þ b1pat it þ b2dvl it þ b3f it : þ D1 þ D2 þ U it ð1þ ð2þ ð3þ ð4þ ð5þ where DV it is the change in dividend; PAT it is the profit after tax; DVL it is the preceding dividend level; TDE it is the total distributable earnings; CF it are cash flows; D1 is assigned the value 1 for the period , and 0 for ; D2 is assigned the value 1 for the period , and 0 for From hypotheses 1 we expect the coefficient of profit after tax (b1) to be positive and cash flows (b3) to be positive and statistically significant. We also expect the coefficient of DVL (b2) to be negative. To test for the effect of growth opportunities in dividend changes cash flows relationship (hypothesis 2), the sample firms are ranked according to market-to-book ratio and partitioned into three portfolios based on average growth rate of each firm. We expect the cash flow regression estimate (b3) to be positive and statistically significant for the group of moderate growth firms (sub-sample 2). We then estimate models 1 to 3 above separately for each portfolio. In testing the effect of leverage on the association between cash flows and dividend changes (hypothesis 3), we partitioned firms into three observation portfolios based on the average level of each firm s total debt. We expect the cash flow regression estimate (b3) to be positive and statistically significant for the group of highly levered firms (sub-sample 3). We then estimate models 1 to 3 above separately for each portfolio. # Blackwell Publishing Ltd, 2003 R&D Management 15, 1,

8 O.J. Adelegan In order to test for the effect of size on the cash flows dividend relationship (hypothesis 4), we similarly rank firms according to the natural logarithm of their total assets. We then assign firms into three portfolio classes and estimate models 1 to 3 for the three sub-samples separately. The coefficient of cash flows (b3) is expected to be significantly positive for the low size portfolios, primarily portfolio 1. In testing for hypotheses 1 the data is pooled across firms and years. This pooling approach assumes that the dividend policy equation is the same for all the firms in the sample; it therefore assumes that firm-specific independent variable coefficients are not systematically different from one firm to the other. By partitioning the data in testing for hypotheses 2, 3 and 4, also assumes that each sub-sample firm-specific variable coefficients are the same across firms. 6. Empirical Analysis and Results 6.1 Sample Summary Statistics The summary statistics of the sample are in table 1. The table indicates that on average the rate of dividend changes is 1.93 per cent and the mean lagged dividend yield is 7.6 per cent, while the market-to-book ratio and cash flows have a mean of 1.72 and 0.61 respectively. Earnings after tax (PAT), total distributable earnings (TDE), total debt (LEV) have higher standard deviation than other variables signifying that they are noisier measures. The correlations of the variables in table 1 are summarized in table 2. The correlation between CF and DV is positive and the lagged dividend yield (DVL) is negatively correlated with cash flows. CF is also positively correlated with PAT. Lagged dividend yield (DVL) is negatively correlated with earnings. The correlations of the variables are generally low. However, these correlations coefficients primarily have descriptive values and conclusions about our hypothesis rely on the regression (multivariate tests). 6.2 Regression Results The empirical results relating to the four research hypotheses are discussed here. From hypothesis 1, we expect a positive relationship between dividend changes and cash flows, given earnings. The reasons given for these expectations are, first, managers have incentives to manipulate accruals to their advantage because earnings are used in measuring performance in Table 1. Sample summary statistics for 882 firm-year observations. Variable a Mean Std. dev Minimum Maximum DVL PAT TDE CF MBV SZ LEV DV a DVL ¼ preceding dividend level; PAT ¼ profit after tax; TDE ¼ total distributable earnings; CF ¼ cash flows; MBV ¼ market to book value; SZ ¼ natural logarithm of fixed assets as a proxy for firm size; LEV ¼ total debt; DV ¼ change in dividend. PAT, CF, TDE, LEV and DVL are deflated by the stock price at the end of the fiscal year. Table 2. Correlation matrix for 882 firm year observations from 1984 to 1997 a. DVL PAT CF TDE MBV SZ LEV DV DVL PAT CF TDE MBV SZ LEV DV 1 a For an explanation of the variables, see table R&D Management 15, 1, 2003 # Blackwell Publishing Ltd, 2003

9 Cash Flow and Dividend Changes in Nigeria compensation contracts. Manipulation of accruals is expected to reduce its usefulness in measuring firms performance (Healy, 1975; Charitou and Vafeas, 1998). Secondly, it appears that dividend pay-out policy is dependent on cash availability and a firm s ability to increase or decrease dividend partly mirrors its liquidity position. This hypothesis is examined by regression dividend changes (DV) on profit after tax (PAT), cash flows (CF), lagged dividend yield and total distributable earnings (TDE). Dummy 1 and 2 are always introduced into the hypothesis in equation (4). Dummy 1 (D1) is used to reflect the impact of the Structural Adjustment Programme introduced in 1986 on dividend policy from 1984 to 1985 (value zero) and for 1986 to 1999 (value one) while Dummy 2 (D2) is used to reflect the impact of investment promotion decree introduced in 1995 (from 1984 to 1994, D2 is assigned the value zero and the value one from 1995 to 1997). The regression results for hypothesis 1 are presented in table 3 for the 882 firm year study. Consistent with previous studies, PAT is positively associated to dividend changes and DVL is negatively related with dividend changes. The model s adjusted R 2 is 21.2 per cent. These results reaffirm the importance of earnings and preceding year dividends in determining changes in dividend pay-out, but it also shows that only 21.2 per cent of the variation in dividend changes is explained by PAT and DVL (Model 3). We tried to introduce more variables into the equation by bringing in total distributable earnings (TDE), cash flows (CF) and D1 and D2. However, this further reduced the adjusted R 2 to 20.4 per cent (Model 5); while the changes in economic policy, D1 (effect of the SAP) on dividend changes is found to be significant. In Model 1, the dummy variables are excluded from the equation and the adjusted R 2 become better at 62.6 per cent. On the basis of low adjusted R 2 values one may be tempted to conclude that the models are not suitable in terms of the explanatory power. This temptation should be resisted since it is not unusual for the R 2 values which result from regression equations dealing with the differences in variables (rather than level of variables) to be generally low. A reason advanced for this is that by using change rather than level data, we omit the variance to be explained by trend thus reducing R 2 leaving only the cyclical and random components. The random component is actually magnified, because the change data add the random elements in two adjacent level observations. As the equation is not expected to explain random movement, this further reduces the R 2 (Keran and Riordan, 1976; Oyejide, 1976). Model 2 excludes TDE from the equation and the adjusted R 2 dropped to 58.5 per cent. Model 4 substitutes PAT for CF in the Lintner model. The coefficient of CF is positive and highly significant, while the explanatory power, R 2 increases to 56.9 per cent. In all the five equations specified, the coefficients of PAT and DVL have the expected positive and negative relationships respectively with dividend changes. In equations (1) (4), CF is positively significant with DV at 1 per cent. The t-statistics are Table 3. Cross-sectional OLS regression results for 882 firm-year observations from 1984 to 1997 a. Variable Model 1 Model 2 Model 3 Model 4 Model 5 Intercept (1.5623) (2.0403)** (2.222)** (1.8162)* ( ) PAT (1.6628) (0.3786) (1.1595) ( ) DVL ( 3.251)*** ( 2.145)** ( 3.630)*** ( 3.653)** ( 2.120)** CF (3.259)*** (2.778)*** (3.014)*** (1.5078) TDE ( ) D (3.108)*** D (0.4625) Adj R % 21.20% 56.90% 58.50% 20.40% F-Stat a t-values are in parentheses. ***, ** and * indicate values are significant at the 1%, 5% and 10% level. The estimates reported here are obtained by using OLS regression procedure in E-views (1995 version). # Blackwell Publishing Ltd, 2003 R&D Management 15, 1,

10 O.J. Adelegan shown in parentheses under the parameter estimates. This could be explained by the peculiarities of Nigerian companies. Soyode (1978) and Oyejide (1987) have documented that most Nigerian companies rely on retained earnings for financing their activities because of the illusion of costlessness usually associated with retained earnings. The decision to allocate the available cash flow between worthwhile investment opportunities and payment of dividend makes it important for firms to consider cash flow in dividend decisions, more so when dividend can only be paid out of available cash flows. In summary, by employing a wider testing period than prior studies, our regression results show that there is a significant relationship between dividend changes and cash flows, given earnings unlike previous studies (Simon, 1994; Charitou and Vafeas, 1998). We therefore accept hypothesis 1. Our research approach on the relationship between dividend changes and cash flows in hypothesis 1 was based on pooled cross-sectional time series data; assuming that the relationship between these variables are homogeneous across firms. This assumption is based on prior capital market studies. However, empirical studies have provided evidence that the response coefficients are affected by firm-specific, industry-specific, and economic factors, firm size, industry classification, quality of earnings and magnitude of accruals (Easton and Zmijewski, 1989; Collins and Kothari, 1989; Dechow, 1994; Charitou and Vafeas, 1998). We therefore investigate contextual factors that are potentially important in mitigating the relationship between dividend changes and cash flows. This study hypothesized that the homogeneity of the dividend policy equation does not hold across firms because of variation in certain contextual factors such as (1) firm growth (hypothesis 2), (2) firm level of leverage (total debt hypothesis 3) and (3) firm size (natural logarithm of total assets, hypothesis 4). The results and discussions are presented below. Hypothesis 2 predicts that cash flows are better predictors of dividend changes for firms with moderate growth prospect. This is because as cash flows are generated, firms have to choose between reinvesting versus returning the funds to investors. Optimally, firms are expected to invest in all projects that will yield positive net present values (NPV) and distribute only free cash flow to shareholders in the form of dividend. The amount of money invested is expected to depend on each firm s growth opportunities (Charitou and Vafeas, 1998; Gaver and Gaver, 1993). However, managers in firms with low growth may also prefer to invest operating cash flows, rather than paying additional dividends (Jensen, 1986). Managerial and shareholders interests are more closely aligned in moderate growth firms than parties interest in low growth firms; we therefore expect the relationship between operating cash flows and dividend changes to be the strongest in moderate growth firms because these firms have a lower opportunity cost of dividends than high growth firms (Charitou and Vafeas, 1998). In table 4 hypothesis 2 was tested by empirically estimating the dividend policy equation separately for each of the three sub-samples, which are partitioned, based on firm growth. Consistent with prior studies, the coefficient of PAT is positive and that of DVL is negative, but while the coefficient of DVL and CF is significant for low growth firms, the coefficient of PAT is not significant. However, the coefficient of PAT is significant for moderate and high growth firms, while that of CF is not significant (Model 1). Model 1 has an adjusted coefficient of determination R 2 that is 38 per cent, which is low. Model 2 (Lintner s model) was also estimated for low growth firms and it revealed PAT as having a negative significant relationship with dividend changes, while DVL has a significant relationship (not reported here). CF was substituted for PAT in Lintner s model (Model 3) and the coefficient of CF revealed a positive and significant relationship at the 5 per cent level between dividend changes and cash flows for low growth firms with adjusted R 2 of 45 per cent. Cash flow is not statistically significant for moderate growth firms. The result revealed that cash flows have a positive statistically significant relationship with dividend changes of low growth and high growth firms. This is not consistent with prior findings where CF is not found significant in the extreme portfolios (Charitou and Vafeas, 1998). This can be partly explained by the overdependence of Nigerian firms on retained earnings for corporate financing (Soyode, 1978; Oyejide, 1987). Low growth firms are less liquid because they have less access to debt than moderate growth and large growth companies. If pay-out policy is dependent on cash availability, their decision to reduce or decrease dividends must partly reflect liquidity. High growth firms on 44 R&D Management 15, 1, 2003 # Blackwell Publishing Ltd, 2003

11 Cash Flow and Dividend Changes in Nigeria Table 4. Cross-sectional OLS regression results for 882 firm-year observations from 1984 to 1997 a. Variable Sub-sample 1 Sub-sample 2 Sub-sample 3 Low growth Moderate growth Highest growth Panel A, Model 1: DV ¼ b0 þ b1pat þ b2dvl þ b3cf þ b4tde Intercept (2.8012)** (0.105) (0.4827) PAT (0.9204) (1.5981)* (4.728)*** DVL (2.2173)* ( ) ( ) CF (2.2172)** (0.1334) (1.5282) TDE (0.8662)* (2.405)** (1.0229) R2 63% 57% 92% Adj. R 2 38% 32% 87% F-Stat Panel B, Model 3: DV ¼ b0 þ b2dvl þ b3cf Intercept (3.881)*** (0.5295) (1.1726) DVL ( 3.144)*** (0.5075) (2.1761)** CF (2.1082)** (0.125) (1.397)* R 2 56% 3% 35% Adj. R 2 45% 18% 20% F-Stat a Firms are partitioned into three according to the firm growth. t-values are in parentheses. ***, ** and * indicate values are significant at the 1%, 5% and 10% level. The estimates reported here are obtained by using OLS regression procedure in E-views (1995 version). the other hand may have projects competing for the available fund. They therefore need to consider the available cash flow in deciding on the distribution of available cash between projects that will yield positive NPV and shareholders in the form of dividend. The positive significant relationship between cash flow and dividend changes in the extreme portfolios could also be partly due to our definition of cash flow as working capital, which is free cash flow. Hypothesis 2 is therefore rejected. Hypothesis 3 predicts that the relationship between operating cash flows and dividend changes is significantly positive for firms that are highly levered, given earnings. This is because a firm s ability to change its dividend policy sometimes depends on its liquidity position. A company that is highly levered has more debt obligations and interest to pay. Size has a positive influence on leverage and growth. Large firms have better access to debt (Marsh, 1982; Baskin, 1989; Chang and Rhee, 1990; Bennets and Donnelly, 1993; Charitou and Vafeas, 1998; Adedeji, 1998) and is likely to be less liquid when compared with small firms. Shareholders are likely to expect more dividend on the one hand and the company is expected to pay principal and interest to debtholders on the other hand. When there is inadequate cash flows (when a company is highly levered), the firm s ability to change its dividend policy in response to its performance is constrained. Therefore cash flows are expected to play a significant role in setting dividend policies. This hypothesis is examined by regressing dividend changes on profit after tax (PAT), cash flows (CF), total distributable earnings (TDE) and lagged dividends (DVL). In table 5, hypothesis 3 is investigated by estimating the empirical model separately for each of the three portfolios, which are partitioned, based on the magnitude of debt. The regression result in panel A of table 5 revealed a negative significant relationship between dividend changes and cash flows of lowly geared firms (at the 1 per cent level of significance) and positive significant relationships between cash flows and dividend changes of highly geared and moderately geared firms at the 5 per cent level of significance. PAT have negative coefficients that are not significantly # Blackwell Publishing Ltd, 2003 R&D Management 15, 1,

12 O.J. Adelegan Table 5. Cross-sectional OLS regression results for 882 firm-year observations from 1984 to 1997 a. Variable Sub-sample 1 Sub-sample 2 Sub-sample 3 Lowly geared Averagely geared Highly geared Panel A, Model 1: DV ¼ b0 þ b1pat þ b2dvl þ b3cf þ b4tde Intercept (0.553) ( ) PAT (114.0)*** ( ) ( ) DVL ( ) (0.2806) (1.7056)* CF ( 3.791)*** (2.2166)** (2.0727)** TDE (1.2728) ( ) (0.6930) R % 52.69% 51.39% Adj. R % 29.04% 27.08% F-Stat Panel B, Model 3: DV ¼ b0 þ b2dvl þ b3cf Intercept (1.7798) ( ) (0.5994) DVL ( ) (0.9579) (2.558)*** CF ( ) (2.568)*** ( ) R % 45.44% 41.55% Adj. R % 34.53% 29.86% F-Stat a Firms are partitioned into three according to the firm leverage. t-values are in parentheses. ***, ** and * indicate values are significant at the 1%, 5% and 10% level. The estimates reported here are obtained by using OLS regression procedure in E-views (1995 version). different from zero for moderately and highly levered firms. The negative coefficient of CF for dividend changes of lowly geared firms may be due to the fact that large firms are usually less levered than small firms. The cost of issuing debt and equity securities is related to firm size. Therefore small firms pay more to issue new equities and long-term debt than large firms do. If large firms are lowly levered than small firms (Smith, 1977; Titman and Wessel, 1988), then their dividend behaviour is likely to be influenced more by current earnings (PAT), which gives a significant positive relationship. Even when cash flow is reducing, lowly levered firms will be unwilling to reduce their dividend. In table 5 panel B, CF was substituted for PAT in Lintner s model; the result revealed that cash flow is only significant for averagely geared firms and not significantly different from zero for the extreme portfolios. We therefore reject hypothesis 3, which says that the relationship between dividend change cash flow is significantly positive for highly levered firms. Hypothesis 4 predicts that operating cash flows are a better predictor of dividend changes for small firms. Our reasoning for this hypothesis is as follows: small firms are more levered than large firms because the cost of issuing new equities and long-term debt are higher for small firms than large firms. They therefore fall back on short-term debt (Smith, 1977; Titman and Wessel, 1988). High leverage means high commitment to repayment of principal and interest and therefore low cash flows. Low size will bring about low earnings and low cash flows. When cash flows are low, cash flows (in addition to earnings) are expected to play a significant role in setting dividend policy. In table 6 hypothesis 4 is investigated by estimating empirically the dividend policy equation separately for each of the three portfolios, which are partitioned, based on firm growth. The coefficient of cash flow is significant for small firms and not significantly different from zero for average and large sized firms. Large firms and moderate firms, however, have significant positive coefficient for PAT. This reveals that large firms are lowly levered and their dividend changes is dependent on earnings after tax (PAT) and lagged dividend (DVL) and total distributable earnings. This is consistent with Lintner (1956). Small firms dividend change is determined more by the availability of cash flows. They are highly geared compared to large 46 R&D Management 15, 1, 2003 # Blackwell Publishing Ltd, 2003