1
The Value of R&D and Dividends for NASDAQ
Firms
Bert-Jan van Leeuwen1 Supervisor: Dr. J.H von Eije
June 7, 2018
ABSTRACT
This paper investigates both the individual and combined value of research & development (R&D) and dividend payment for NASDAQ firms. Panel data regression of 2,552 NASDAQ firms proves that R&D and dividends are both positively related to stock performance, with every dollar invested in R&D valued at $2.66, while the signal of an extra dollar of dividends is valued at $1.43. The difference between these two effects proves to be statistically
significant. Furthermore, hypothesized is that there is a negative relationship between R&D (growth-seeking) and dividends (maturation), and consequently, there exists a negative interaction effect of R&D and dividends on stock performance. However, no evidence can be presented in support of these hypotheses.
Keywords: Payout policy, dividend signaling, research and development JEL classification: G10, G30, G35, O32
1 Student number s2563150. This is a thesis for the Master Finance at the Faculty of Economics and Business of
2
1. Introduction
In the perfect-world models of Miller & Modigliani (1961), payout policy contains no relevant information for investors about firm value or performance. In the real financial world, however, investors have to deal with, among others, asymmetry of information. This phenomenon causes payout policy to be a relevant signal for investors in the process of forming expectations on a firm’s future performance.
Black (1976) describes the Dividend Puzzle, which focuses on why firms actually pay
dividends. He proposes two factors that may play a role in a firm’s payout policy decision: (1) signaling of future profitability or creditworthiness by increasing the dividend ratio, and (2) the choice between reinvesting and paying out financial resources. These factors have different effects on the investor’s interpretation of firm value, which is why yet there is no straightforward, theory-substantiated, empirical evidence on the relationship between payout policy and firm valuation. Furthermore, with the presence of dividend taxation, firms with lower dividend ratios should be valued higher because of avoided taxation.
The interpretation of these signals, however, is likely to differ between different types of firms. A full-grown company that has already depleted its main growth opportunities and has started to pay out its excess cash through higher dividends cannot be treated the same as a high-tech start-up firm in Silicon Valley that wants to retain as much cash as possible in order to invest. The difference between these two extremes may be partly explained by Research and Development (R&D) expenditure. Firms that expect to have much (yet undiscovered) growth opportunities can be expected to spend more on R&D than so-called cash cow firms. Consequently, R&D expenditure may serve well as a proxy for growth opportunities and could thus be an important factor in the determination and valuation of payout policy. The proposition that R&D as well as New Product Development, which can be seen as the interaction between R&D and Marketing, adds value to a firm in the search for a sustainable competitive advantage, is backed by previous literature (Rein, 2004; Sivadas and Dwyer, 2000; Ayers and Dahlstrom, 1997; Salimi and Rezaei, 2018). Following the Resource Based View of Barney (1991), firms are thus investing in a potential sustainable competitive advantage via R&D, which is likely to be acknowledged by investors.
3
combination of both have hardly been made. What should the manager decide when choosing between dividend payment and R&D investment? What are the differences in the effect of payout policy on value, when considering R&D levels? This paper will try to analyze and understand the individual and combined value of R&D and dividends.
This paper presents evidence that both R&D and dividends are positively related to stock performance, with the dollar-value of R&D expenditure being significantly higher than that of dividend payment. The results from this paper are helpful to many types of agents in the market. Investors can better interpret the combination of payout policy and R&D investment. Dividends are a costly signal in an informational asymmetric world, and an investment in R&D is an investment in firm value. Furthermore, managers can incorporate this information into their decision-making process, but also anticipate to the interpretation of the market. The remainder of this paper is as follows: section 2 presents an overview of previous literature on the subject and derives the theoretical framework and hypotheses from it. Section 3
describes the methodology of the research. Section 4 discusses the data used in the regressions. In section 5, the empirical results are presented. Finally, section 6 concludes.
2. Literature and Theoretical Framework
This section discusses previous literature as well as the theoretical framework of this study. The first two subsections discuss the theory and empirics of payout policy and R&D, respectively. The last subsection discusses the combination of dividends and R&D and their effects on firm value.
2.1 – Payout policy
The process of understanding payout policy basically starts with the Dividend Irrelevance Theory of Miller & Modigliani (1961), which theorizes that payout policy of a company does not matter to the value of a company. The theory is based on the assumption of perfect capital markets, i.e. frictionless markets with equal access for all firms and people.
4
to buy back shares. Second, there are no signaling effects, because all information is already publicly available and reflected in the prices.
If capital markets were perfect, there would be no need to perform any research into payout policy effects on firm value. In reality, however, there exists no such thing as market perfection, and asymmetry of information is one important cause for this. Fama (1970) theorizes the Efficient Market Hypothesis, stating that only new information can cause movement in the market, because all known information is already reflected in the prices. Unexpected changes in dividend policy thus can signal information to outsiders.
Miller & Rock (1985) and Bhattacharya (1979), among others, theorize the Cash Flow Signaling theory, which states that a dividend increase contains new information in the form of a signal of growth of expected future cash flows, because the firm thereby signals to be able to commit to a higher payout ratio in the future. This should then be appreciated by the market through higher stock prices.
Dividend increases can also be interpreted as a positive signal in light of agency costs. Jensen & Meckling (1976) define agency costs as the costs the principal (shareholder) bears due to the asymmetry of information between him/her and the agent (manager). For example, a manager might invest in negative-NPV investments for private benefits. Jensen (1986) follows this theory and proposes that dividend may be a means to reduce agency costs, because dividends reduce cash levels and thereby the “private benefit possibilities” of the agent. The reduction of agency costs should then be recognized by the market as a value-enhancing event and therefore, stock prices would increase.
5
There are several papers that object to the proposed relationship between payout policy and firm value. Watts (1973) does find empirical evidence of a positive effect of unexpected dividend changes and future earnings changes, but at the same time, he argues that the absolute values of these effects are so small, that there is no reason to assume that companies’ payout policy conveys crucial information. Benartzi et al. (1997) argue that, although they acknowledge the positive relationship between dividends and stock prices, there is no evidence for actual future earnings growth. Furthermore, they argue that changes in dividends tell us more about the past than about the future, by showing that earnings growth was significantly high in the year before the dividend increases. Lie (2004) finds no significant effect of dividend omissions and decreases on earnings.
Ghosh & Woolridge (1989) argue that cuts in dividend could be a signal of higher investment, which would result in higher future earnings. However, they only present evidence of less negative returns after dividend cuts or omissions instead of actual positive returns. Nevertheless, their paper could give a possible explanation of the ambiguity of whether payout policy contains information about future earnings growth. There could, namely, exist an opposing effect to the positive effect described before. Regarding payout policy, there are in principle two opposing forces “fighting” each other. In the case of dividend increases, this would implicate that there is the signal of higher future profitability (Miller & Rock, 1985; Bhattacharya, 1979) on the one hand, and the signal of a lack of further growth opportunities on the other hand.
Lang & Litzenberger (1989) prove that average stock returns after dividend changes depend on whether a firm is over-investing or value-maximizing. Over-investing firms experience higher returns on dividend increases, since these outflowing funds would otherwise be invested in value-decreasing investments with a negative net present value (NPV). On the other hand, value-maximizing firms still have growth opportunities, and therefore dividend increases are seen as a less positive phenomenon, since these firms could still maximize value by investing in positive-NPV investments.
6
Besides its relationship with returns, payout policy is also related to firm value through the concept of risk. Many papers confirm the negative relationship between dividend payouts and firm risk (Grullon et al., 2002; Grullon and Michaely, 2004; Pastor and Veronesi, 2003; Bartram et al., 2012). All these papers report evidence that dividend initiations or increases significantly reduce idiosyncratic and/or systematic risk. Consequently, firm value should also increase with the risk-reduction of dividend increasing policies through a lower discount rate.
Following this subsection, I derive the following hypothesis:
Hypothesis 1: There exists a positive relationship between firm dividends and stock returns
2.2 – Research and Development
In order to outperform competitors in the long-run and consequently increase firm value, firms have to gain a sustainable competitive advantage. This can be achieved through two channels: cost leadership and differentiation (Porter, 1985). The former implies that firms are able to produce the same good as others at lower cost, providing an advantage. The latter means that a firm creates an advantage by producing better products or providing better services than its competition.
The Resource Based View (e.g., Wernerfelt, 1984; Barney, 1991) is an important theory in understanding the determinants of competitive advantages. This view proposes that firms can pursue different strategies due to heterogeneity in resources between firms. By obtaining and/or preserving specific internal resources, a firm could create a competitive advantage over its competition. Previous literature assumed that the potential advantages of heterogeneity will be cancelled out because of mobility of resources. Barney (1991), however, proposes that there can be obtained a sustainable advantage when resources meet four key characteristics. They should be 1) valuable, 2) rare, 3) imperfectly imitable, and 4) non-substitutable.
R&D could be a channel through which these kinds of internal resources are created. R&D as a determining factor in firm performance is widely acknowledged, standing alone as well as a part of New Product Development (Rein, 2004; Sivadas and Dwyer, 2000; Ayers and Dahlstrom, 1997; Salimi and Rezaei, 2018, among others). The R&D-levels of firms could be interpreted by investors as a signal about future performance.
7
higher total risk for firms with higher R&D spending. This leads to a conflicting effect in stock valuation, because an increase in risk leads to a higher discount factor, which subsequently decreases stock value. However, rational managers should be aware of the increase in risk through R&D, and therefore only invest in R&D when the increase in profitability is expected to offset the increase in risk.
From the findings presented in this subsection, the second hypothesis is derived:
Hypothesis 2: There exists a positive relationship between firm’s R&D levels and stock
returns
2.3 – Payout policy and R&D
Until the present, hardly anything has been written about the combined effect of R&D levels and dividend policies with regard to stock price behavior. However, when taking into account what is presented in the previous section – R&D as a means to pursue future performance growth – one could assume R&D to be a proxy or representative factor for the exploration of growth opportunities. Following the results of Lang & Litzenberger (1989), as discussed in the dividend subsection, it may be that the investor interprets high R&D as an indicator of a firm being growth pursuing, and low R&D as an indicator of maturity. The assumption is then made that firms with higher R&D levels are considered better at finding positive-NPV investment opportunities than low-level R&D firms.
With both the relationships of dividends and R&D with stock returns expected to be positive, it is interesting for managers to know how to spend every dollar of cash that can either be assigned to R&D or paid out as dividend. Since the sample is constructed from data of NASDAQ firms, which are generally more technology-oriented, it seems rational to assume that R&D delivers more value than dividend payment. Therefore, I hypothesize:
Hypothesis 3: R&D expenditure has a more positive individual effect on stock performance
than dividend payment
8
for the firm. When the firm has little opportunities to invest, then excess cash should be paid out in order to reduce or at least restrict agency costs. However, in case of a firm with many investment opportunities, shareholders will demand less dividends, because the firm can invest its funds in those opportunities.
When combining this model (La Porta et al., 2000) with the assumption that R&D is a proxy for growth opportunities, there would exist a negative relationship between R&D expenditure and dividend changes. In a simplified setting, firms would treat the combination of R&D and dividends as a trade-off. For every dollar, the manager has to decide whether his firm will use it for R&D (positive-NPV investments), or pay dividends. In line with this expectation, I propose the following hypothesis:
Hypothesis 4: Trade-off Hypothesis
There exists a negative relationship between firm’s dividend levels and R&D levels
9
Figure 1 reflects this paper’s application of the Outcome model graphically. In line with a negative relationship between dividends and R&D, there exists some optimal combination between the two. However, when a firm deviates from this line, it would be in a suboptimal position. A firm above the line uses a value-decreasing combination of activities, namely spending too much on R&D and paying too high dividends. The former would indicate that it might be over-investing; the latter is a sign of paying out funds that could be used to fund profitable projects (missed opportunity). On the other hand, when a firm moves below the line, it could be a sign of under-investment and/or agency costs. The firm possesses a certain amount of cash funds, but does not spend it on either R&D (correcting for the under-investment) or dividends (reducing agency costs). Since both positive and negative deviations are value-decreasing, the interaction effect might be best estimated by a quadratic relation, with a global maximum point at the optimal combination point. This is graphically shown by Figure 2. Since both being above and below the Ideal R&D-Dividend Line should have a negative effect on value and thus stock performance, I propose the final hypothesis of this paper:
Hypothesis 5: Interaction Hypothesis
Deviation from the Ideal R&D-Dividend Line has a negative impact on a firm’s stock returns
10
3. Methodology
This paper tries to investigate the stand-alone as well as interaction effects of dividend and R&D expenditure on the value of the firm, by applying both dividend and R&D to the dividend outcome model of La Porta et al. (2000). Using panel data of over 2500 NASDAQ-firms over the period 1998 to 2017, multiple regression models are derived. In all regressions, I use cross-section fixed effects as well as time fixed effects. Furthermore, White diagonal standard errors and covariance are applied in order to correct for potential heteroscedasticity. The first subsection discusses the regression model, to test for hypotheses 1, 2, 3, and 5, is presented and discussed. Thereafter, the methodology of the robustness tests, which also tests for hypothesis 4, is presented.
3.1 – Dividend, R&D, and interaction versus return
The framework of the main regression model is defined as follows: ∆𝑀𝑖,𝑡 𝑀𝑖,𝑡−1 = 𝛾0+ 𝛾1 𝐷𝑖,𝑡 𝑀𝑖,𝑡−1 + 𝛾2 𝑅𝐷𝑖,𝑡 𝑀𝑖,𝑡−1 + 𝛾3( 𝐷𝑖,𝑡 𝑀𝑖,𝑡−1 𝑅𝐷𝑖,𝑡 𝑀𝑖,𝑡−1 ) + 𝛾4( 𝐷𝑖,𝑡 𝑀𝑖,𝑡−1 𝑅𝐷𝑖,𝑡 𝑀𝑖,𝑡−1 ) 2 + 𝛾5 𝐶𝑖,𝑡−1 𝑀𝑖,𝑡−1 + 𝛾6𝐿𝑖,𝑡−1+ 𝛾7𝑀𝐵𝑖,𝑡−1+ 𝛾8ln(𝑆𝑖,𝑡−1) + 𝛾9 𝑅𝑖,𝑡−1 𝑀𝑖,𝑡−1 + 𝜀𝑖,𝑡 (1)
In line with Faulkender & Wang (2006), all variables that represent absolute values (instead of ratios or natural logarithms) are divided by the lagged market capitalization (𝑀𝑖,𝑡−1) of firm 𝑖 in year 𝑡 − 1, in order to correct for size differences between firms. Without this deflation, the results would be affected stronger by larger companies, while it is preferable that every company has the same relative influence. The regression model is used to test for both the first three hypotheses of R&D and dividend related to returns, as well as hypothesis 5. In this regression model, the dependent variable is the company’s stock performance ∆𝑀𝑖,𝑡
𝑀𝑖,𝑡−1, measured
by the relative growth of the firm’s Market Capitalization, with dividends fictionally re-invested in order to control for the normal one-to-one price-decreasing effects of dividend payment. This means that ∆𝑀𝑖,𝑡
𝑀𝑖,𝑡−1 is the dividend re-invested return. The first two main explanatory variables are 𝐷𝑖,𝑡
𝑀𝑖,𝑡−1 and 𝑅𝐷𝑖,𝑡
𝑀𝑖,𝑡−1, representing total dividends paid
11
reflected by an increase in the market value of equity (Pettit, 1972, 1976; Aharony and Swary, 1980; Benartzi et al., 1997, among others). Moreover, 𝛾2 should be positive as well, since hypothesis 2 states that the relationship between firm value and R&D is positive, based on previous literature (Rein, 2004; Sivadas and Dwyer, 2000; Ayers and Dahlstrom, 1997; Salimi and Rezaei, 2018, among others). Following hypothesis 3, expected is that 𝛾3 > 𝛾2, indicating that the value-increasing effect of R&D is greater than that of dividend payment.
In order to test for hypothesis 5, the interaction term and the square of the interaction term are added to the equation. Adding these variables to the regression gives the ability to measure if there indeed is a negative interaction effect of R&D and dividend for firms that deviate from the ideal combination between the two. Below, for readability concerns, the interaction term is written as 𝐼𝑖,𝑡 = 𝐷𝑖,𝑡
𝑀𝑖,𝑡−1
𝑅𝐷𝑖,𝑡
𝑀𝑖,𝑡−1 below. The first difference of the annual return with respect to the
interaction term 𝐼𝑖,𝑡 should have a global extreme point of the interaction effect of R&D and dividends on returns. When the second derivative is negative, the extreme point is a maximum, indicating that the extreme point is the optimal combination between R&D and dividends, as graphically shown in Figure 2. Should the company move below the optimal line, i.e. reduce dividends and/or R&D, this would be value-decreasing, as well as moving above the line – paying too high dividends or spending too much on R&D. The equations below show the process of finding the maximum.
𝑑(𝑀∆𝑀𝑖,𝑡 𝑖,𝑡−1) 𝑑𝐼𝑖,𝑡 = 𝛾3+ 2𝛾4𝐼𝑖,𝑡 = 0 (2) 𝑑2(𝑀∆𝑀𝑖,𝑡 𝑖,𝑡−1) 𝑑𝐼𝑖,𝑡2 = 2𝛾4 < 0 (3)
Equation (2) is the first difference of annual return with respect interaction, which is set to zero in order to find an extreme point. Next, equation (3) represents the second difference, which is conditioned to be smaller than zero, in order for the extreme point to be a maximum. Consequently, expecting 𝛾3 > 0 and 𝛾4 < 0, the optimal combination of R&D and dividends should correspond with 𝐼𝑡𝑚𝑎𝑥 = 𝛾3/−2𝛾4, and everything that deviates from this should be lower-valued by the market.
12
one of the main explanatory variables. These control variables are cash holdings (𝐶𝑖,𝑡−1 𝑀𝑖,𝑡−1), leverage (𝐿𝑖,𝑡−1), the market-to-book ratio (𝑀𝐵𝑖,𝑡−1), firm size, measured by the natural logarithm of book value of assets (ln (𝑆𝑖,𝑡−1)), and share repurchases (𝑅𝑖,𝑡−1
𝑀𝑖,𝑡−1), of firm 𝑖 in year 𝑡 −
1. To cope with possible endogeneity arising from current interactions, all control variables are lagged. In line with Faulkender & Wang (2006), no instrumental variables are used in the regressions. Although this could be more helpful in preventing endogeneity, this kind of variables is very difficult to find and therefore not added to the methodology.
3.2 – Robustness: dividend versus R&D
In order to test for hypothesis 4, which states that dividends and R&D should be negatively related, I construct the following two regression models:
𝑅𝐷𝑖,𝑡 𝑀𝑖,𝑡−1 = 𝜃0+ 𝜃1 𝐷𝑖,𝑡 𝑀𝑖,𝑡−1+ 𝜀𝑖,𝑡 (4) 𝐷𝑖,𝑡 𝑀𝑖,𝑡−1 = 𝛿0+ 𝛿1 𝑅𝐷𝑖,𝑡 𝑀𝑖,𝑡−1+ 𝜀𝑖,𝑡 (5) 𝑅𝐷𝑖,𝑡
𝑀𝑖,𝑡−1 is defined as R&D spending of firm 𝑖 in year 𝑡.
𝐷𝑖,𝑡
𝑀𝑖,𝑡−1 is the amount of dividends paid by firm 𝑖 in year 𝑡. Both variables are deflated by the lagged market capitalization to correct for size differences. Since the hypothesis proposes that there exists a trade-off between R&D and dividends, expected is that 𝜃1 < 0 and 𝛿1 < 0.
In order to test for hypothesis 5 in another way, the squared residuals of the regression of equations (4) and (5) are taken as replacement for the two interaction terms of equation (1). Since the residuals of the trade-off regression represent the deviations from the Ideal R&D-Dividend Line as shown in Figure 1, these regressions test for the existence of this value-decreasing effect of these deviations. The regression equations are as follows:
13 ∆𝑀𝑖,𝑡 𝑀𝑖,𝑡−1 = 𝜇0+ 𝜇1 𝐷𝑖,𝑡 𝑀𝑖,𝑡−1 + 𝜇2 𝑅𝐷𝑖,𝑡 𝑀𝑖,𝑡−1 + 𝜇3𝑉𝑖,𝑡2 + 𝜇4 𝐶𝑖,𝑡−1 𝑀𝑖,𝑡−1 + 𝜇5𝐿𝑖,𝑡−1+ 𝜇6𝑀𝐵𝑖,𝑡−1 + 𝜇7ln(𝑆𝑖,𝑡−1) + 𝜇8 𝑅𝑖,𝑡−1 𝑀𝑖,𝑡−1 + 𝜀𝑖,𝑡 (7)
Besides the newly added terms 𝑈𝑖,𝑡2 and 𝑉𝑖,𝑡2, all variables are defined the same as in regression equation (1). 𝑈𝑖,𝑡is defined as the residual of observation 𝑖 in year 𝑡 from regressing equation (4). 𝑉𝑖,𝑡is defined as the residual of observation 𝑖 in year 𝑡 from regressing equation (5). Expected is that 𝛽3< 0 and 𝜇3< 0, since that would indicate that, the more a firm deviates from the norm in the relationship between R&D and dividends, the greater is the negative impact on stock performance. An overview of all the variables and subsequent symbols that are presented in this section is provided in Table 1.
TABLE 1
Overview of symbols of variables used in different regressions
This table presents all the symbols and definitions of all variables that are used in the regressions of this paper. All variables are presented for firm 𝒊 in year 𝒕. ∆𝑴𝒊,𝒕, 𝑫𝒊,𝒕, 𝑹𝑫𝒊,𝒕, 𝑪𝒊,𝒕−𝟏 and 𝑹𝒊,𝒕−𝟏 are divided by 𝑴𝒊,𝒕−𝟏 in order to correct for size differences between observations.
Symbol Variable Definition
∆𝑴𝒊,𝒕
𝑴𝒊,𝒕−𝟏 Annual return
Change in Market Capitalization (Market Value of Equity) of firm 𝑖 between year 𝑡 and 𝑡 − 1
𝑫𝒊,𝒕
𝑴𝒊,𝒕−𝟏 Total dividends Total amount of dividends paid to shareholders by firm 𝑖 in fiscal year 𝑡
𝑹𝑫𝒊,𝒕
𝑴𝒊,𝒕−𝟏 R&D expenditure R&D expenditure of firm 𝑖 in fiscal year 𝑡
𝑪𝒊,𝒕−𝟏
𝑴𝒊,𝒕−𝟏 Cash
Cash holdings plus marketable securities of firm 𝑖 at the end of fiscal year 𝑡 − 1
𝑳𝒊,𝒕−𝟏 Leverage Leverage ratio (Debt / Assets) of firm 𝑖 at the end of fiscal year 𝑡 − 1 𝑴𝑩𝒊,𝒕−𝟏 Market-to-Book ratio Market-to-Book ratio (Market Capitalization / Book Value of Equity) of firm 𝑖 at end of fiscal year 𝑡 − 1 𝒍𝒏(𝑺𝒊,𝒕−𝟏) Natural logarithm of Size Natural logarithm of the Book Value of Assets of firm 𝑖 at end of fiscal year 𝑡 − 1
𝑹𝒊,𝒕−𝟏
𝑴𝒊,𝒕−𝟏 Repurchases Shares repurchased by firm 𝑖 in fiscal year 𝑡 − 1
𝑼𝒊,𝒕𝟐
Squared residuals of trade-off regression
Squared deviation of firm 𝑖 in fiscal year 𝑡, with respect to the regression results of equation (4)
𝑽𝒊,𝒕𝟐
Squared residuals of trade-off regression
14
4. Data
This section discusses the collecting, organizing and verification of the data that is used for the empirical research. Section 4.1 discusses the collected data. Thereafter, section 4.2 presents and debates some sample descriptive statistics. Section 4.3 ends with a discussion of the sample’s correlation statistics.
4.1 – Data collecting and organization
This paper’s data was obtained via DataStream. The selected sample initially consisted of all equities on DataStream’s constituent list of the NASDAQ. The data contains balance sheet and income statement information on all companies that were traded on the NASDAQ in the period of 1998 to 2017. In total, information on 2,552 listed stocks was retrieved, of which 2,524 remained after the elimination of some companies of which the amount of information available was not sufficient. Observations of negative equity are not taken into account, causing the observations of a Debt-to-Assets ratio of greater than one, or a negative Market-to-Book ratio, to be filtered out. Furthermore, observations with ∆𝑴𝒊,𝒕
𝑴𝒊,𝒕−𝟏> 10 are eliminated from the sample as
outliers, since firms with an annual return of over a thousand percent are very rare and of excessive influence on the regression results. Moreover, the ratio variables (variables that are deflated by the lagged market capitalization) are constrained in the following way, in order to prevent for excessive outliers: 𝐷𝑖,𝑡
𝑀𝑖,𝑡−1 ≤ 1, 𝑅𝐷𝑖,𝑡 𝑀𝑖,𝑡−1 ≤ 1 , 𝐶𝑖,𝑡−1 𝑀𝑖,𝑡−1≤ 1, 𝑅𝑖,𝑡−1 𝑀𝑖,𝑡−1 ≤ 1. After automatic elimination of unavailable data points by econometric software (Eviews), the complete sample consists of 12,086 observations, whereas full availability would imply 50,480 (20 years times 2,524 firms) observations. An important explanation for this large difference is the relative lack of data on R&D spending. Of the 50,460 potential R&D-observations, 31,376 are unavailable. It could be that the majority of the firms with unavailable R&D does not spend on R&D at all. This would be a source of bias, and is therefore recognized as a limitation to this study.
4.2 – Descriptive statistics
15
the firm did not pay dividends. The observations of negative share repurchases are likely to be best interpreted as previously repurchased stock that is either re-sold or paid out as executive compensation. Since there are only 14 observations with negative repurchases, it is not likely that there is some constructional error in the variable.
TABLE 2
Descriptive statistics for the regression sample
This table provides descriptive statistics on the sample used in this research, consisting of all available observations except outliers (annual returns ∆𝑀𝑖,𝑡
𝑀𝑖,𝑡−1 > 10, 𝐷𝑖,𝑡 𝑀𝑖,𝑡−1≤ 1, 𝑅𝐷𝑖,𝑡 𝑀𝑖,𝑡−1≤ 1 , 𝐶𝑖,𝑡−1 𝑀𝑖,𝑡−1≤ 1, 𝑅𝑖,𝑡−1 𝑀𝑖,𝑡−1≤ 1) and observations of firms with negative equity. The dependent variable ∆𝑀𝑖,𝑡
𝑀𝑖,𝑡−1
reflects annual return, defined as relative change in market capitalization, with dividend fictionally re-invested. The explanatory variables are total dividends 𝐷𝑖,𝑡
𝑀𝑖,𝑡−1, R&D expenditure
𝑅𝐷𝑖,𝑡
𝑀𝑖,𝑡−1, the interaction term between dividends and R&D
𝐷𝑖,𝑡
𝑀𝑖,𝑡−1∗
𝑅𝐷𝑖,𝑡
𝑀𝑖,𝑡−1, the square of the interaction term ( 𝐷𝑖,𝑡
𝑀𝑖,𝑡−1∗
𝑅𝐷𝑖,𝑡
𝑀𝑖,𝑡−1)
2, the squared residuals of the trade-off regressions 𝑈
𝑖,𝑡2 and 𝑉𝑖,𝑡2, lagged cash holdings plus marketable securities𝐶𝑖,𝑡−1
𝑀𝑖,𝑡−1, the lagged leverage ratio 𝐿𝑖,𝑡, the lagged market-to-book ratio 𝑀𝐵𝑖,𝑡, the lagged natural logarithm of size (book value of assets) ln (𝑆)𝑖,𝑡 and lagged share repurchases𝑅𝑖,𝑡−1
𝑀𝑖,𝑡−1.
Variable Mean Median Maximum Minimum St. Dev. Observations
∆𝑀𝑖,𝑡 𝑀𝑖,𝑡−1 0.231 0.094 9.788 -0.998 0.782 23,969 𝐷𝑖,𝑡 𝑀𝑖,𝑡−1 0.012 0.000 0.928 0.000 0.028 23,983 𝑅𝐷𝑖,𝑡 𝑀𝑖,𝑡−1 0.069 0.040 0.983 0.000 0.092 12,580 𝐷𝑖,𝑡 𝑀𝑖,𝑡−1 ∗𝑅𝐷𝑖,𝑡 𝑀𝑖,𝑡−1 +0.000 0.000 0.295 0.000 0.004 12,551 ( 𝐷𝑖,𝑡 𝑀𝑖,𝑡−1 ∗𝑅𝐷𝑖,𝑡 𝑀𝑖,𝑡−1 )2 +0.000 0.000 0.087 0.000 0.001 12,551 𝑈𝑖,𝑡2 0.004 +0.000 0.650 0.000 0.019 12,098 𝑉𝑖,𝑡2 +0.000 +0.000 0.595 0.000 0.008 12,098 𝐶𝑖,𝑡−1 𝑀𝑖,𝑡−1 0.190 0.133 1.000 0.000 0.186 24,090 𝐿𝑖,𝑡−1 0.153 0.091 12.571 0.000 0.207 24,109 𝑀𝐵𝑖,𝑡−1 4.588 1.940 4002.530 0.000 42.800 24,109 ln (𝑆)𝑖,𝑡−1 12.810 12.833 19.589 2,773 1.984 24,109 𝑅𝑖,𝑡−1 𝑀𝑖,𝑡−1 0.160 0.000 0.896 -0.127 0.044 23,238
16
TABLE 3 Correlation matrix
This table presents the correlation matrix of all the explanatory variables used in this paper’s regressions. The variables are total dividends 𝐷𝑖,𝑡
𝑀𝑖,𝑡−1, R&D expenditure
𝑅𝐷𝑖,𝑡
𝑀𝑖,𝑡−1, the interaction term
between dividends and R&D 𝐷𝑖,𝑡
𝑀𝑖,𝑡−1∗ 𝑅𝐷𝑖,𝑡
𝑀𝑖,𝑡−1, the squared interaction term ( 𝐷𝑖,𝑡 𝑀𝑖,𝑡−1∗
𝑅𝐷𝑖,𝑡 𝑀𝑖,𝑡−1)
2, the residuals of the trade-off regression 𝑈
𝑖,𝑡2, cash holdings plus marketable securities 𝐶𝑖,𝑡−1 𝑀𝑖,𝑡−1, the leverage ratio 𝐿𝑖,𝑡−1, the market-to-book ratio 𝑀𝐵𝑖,𝑡−1, the natural logarithm of size (book value of assets) ln (𝑆)𝑖,𝑡−1 and share repurchases
𝑅𝑖,𝑡−1
𝑀𝑖,𝑡−1. All variables that represent absolute values are deflated by the lagged market capitalization 𝑀𝑖,𝑡−1 to correct for size differences.
𝐷𝑖,𝑡 𝑀𝑖,𝑡−1 𝑅𝐷𝑖,𝑡 𝑀𝑖,𝑡−1 𝐷𝑖,𝑡 𝑀𝑖,𝑡−1 ∗ 𝑅𝐷𝑖,𝑡 𝑀𝑖,𝑡−1 ( 𝐷𝑖,𝑡 𝑀𝑖,𝑡−1 ∗𝑅𝐷𝑖,𝑡 𝑀𝑖,𝑡−1 )2 𝑈 𝑖,𝑡2 𝑉𝑖,𝑡2 𝐶𝑖,𝑡−1 𝑀𝑖,𝑡−1 𝐿𝑖,𝑡−1 𝑀𝐵𝑖,𝑡−1 ln (𝑆)𝑖,𝑡−1 𝑅𝑖,𝑡−1 𝑀𝑖,𝑡−1 𝐷𝑖,𝑡 𝑀𝑖,𝑡−1 1.000 𝑅𝐷𝑖,𝑡 𝑀𝑖,𝑡−1 -0.060 1.000 𝐷𝑖,𝑡 𝑀𝑖,𝑡−1 ∗ 𝑅𝐷𝑖,𝑡 𝑀𝑖,𝑡−1 0.629 0.075 1.000 ( 𝐷𝑖,𝑡 𝑀𝑖,𝑡−1 ∗ 𝑅𝐷𝑖,𝑡 𝑀𝑖,𝑡−1 )2 0.467 0.048 0.921 1.000 𝑈𝑖,𝑡2 -0.027 0.611 0.021 0.013 1.000 𝑉𝑖,𝑡2 0.744 0.025 0.786 0.764 0.006 1.000 𝐶𝑖,𝑡−1 𝑀𝑖,𝑡−1 0.036 0.461 0.083 0.048 0.187 0.060 1.000 𝐿𝑖,𝑡−1 -0.004 -0.062 -0.020 -0.011 0.018 -0.015 -0.099 1.000 𝑀𝐵𝑖,𝑡−1 -0.010 0.007 -0.002 -0.001 0.002 -0.002 -0.025 0.058 1.000 ln (𝑆)𝑖,𝑡−1 0.056 -0.212 -0.002 -0.017 -0.132 -0.018 -0.068 0.186 -0.039 1.000 𝑅𝑖,𝑡−1 𝑀𝑖,𝑡−1 0.059 -0.030 0.032 0.009 -0.014 0.017 0.031 0.050 -0.009 0.185 1.000
17
4.3 – Correlation statistics
Correlation statistics between the dependent variables are presented in Table 3. Most of the relatively large correlation coefficients are those including the interaction terms or residuals, which are logical to be correlated to each other or the two first variables. Furthermore, the correlation between lagged cash and R&D is explained by the need for cash in order to invest in R&D. The last large coefficient is the one of R&D and the squared residuals of the trade-off regression. Since there are no further large coefficients, there does not seem to be the need to assume that multicollinearity plays an important role in the regressions of the data.
5. Results
This section presents results of the regressions that are performed to test for the hypotheses stated in section 2. In section 5.1, results from the main regression model are discussed with regard to the individual effects of R&D and dividends on performance. Section 5.2 discusses the same regression, but focuses of the interaction hypothesis. Thereafter, section 5.3 discusses the robustness tests.
5.1 – Individual effects
The results of the regression that links annual returns to dividends and R&D are shown in Table 4. The first regression is performed in order to test for hypotheses 1, 2, and 3. Both coefficients are significantly positive, which is in line with the first two hypotheses. In regression II, the signs are in line with the hypotheses as well, but in case of dividends, the coefficient is not significant. This is explained by the fact that much explaining power of dividends is taken away by the interaction terms in regression II.
18
TABLE 4
Regression results for the individual and interaction hypotheses
This table presents the results of the panel data regression performed to test for hypotheses 1, 2 (individual hypotheses), and 3 (interaction hypothesis). In both regressions, the dependent variable is annual
returns ∆𝑀𝑖,𝑡
𝑀𝑖,𝑡−1, defined as the change in market capitalization for firm 𝑖 in year 𝑡, with dividends re-invested fictionally. The explanatory variables are dividends paid ( 𝐷𝑖,𝑡
𝑀𝑖,𝑡−1), R&D expenditure ( 𝑅𝐷𝑖,𝑡
𝑀𝑖,𝑡−1) in regression I. In regression II, their interaction term as well as squared interaction term are added. 𝐶𝑖,𝑡−1
𝑀𝑖,𝑡−1 is lagged cash
holdings plus marketable securities, 𝐿𝑖,𝑡−1 is leverage, 𝑀𝐵𝑖,𝑡−1 is lagged market-to-book ratio, ln (𝑆𝑖,𝑡−1) is the lagged natural logarithm of book value of assets (size), 𝑅𝑖,𝑡−1
𝑀𝑖,𝑡−1 is lagged total share repurchases. All variables that reflect absolute (dollar) values are deflated by the lagged market capitalization 𝑀𝑖,𝑡−1, in order to control for size differences. In order to test for the existence of a significant difference between the effects of 𝑅𝐷𝑖,𝑡
𝑀𝑖,𝑡−1 and 𝑅𝐷𝑖,𝑡
𝑀𝑖,𝑡−1 (hypothesis 3), a Wald test is performed, of which the p-value is presented in the table.
Fixed time effects and fixed cross-section effects are applied to the regressions. Furthermore, White diagonal standard errors are used in order to correct for potential heteroscedasticity.
I II Coefficient (White SE) Coefficient (White SE) Intercept 3.635 *** (0.216) 3.645 *** (0.216) 𝐷𝑖,𝑡 𝑀𝑖,𝑡−1 1.431 ** (0.599) 0.440 (0.477) 𝑅𝐷𝑖,𝑡 𝑀𝑖,𝑡−1 2.663 *** (0.239) 2.632 *** (0.240) 𝐷𝑖,𝑡 𝑀𝑖,𝑡−1 ∗ 𝑅𝐷𝑖,𝑡 𝑀𝑖,𝑡−1 13.211 (14.058) ( 𝐷𝑖,𝑡 𝑀𝑖,𝑡−1 ∗ 𝑅𝐷𝑖,𝑡 𝑀𝑖,𝑡−1 )2 -27.028 (48.736) 𝐶𝑖,𝑡−1 𝑀𝑖,𝑡−1 0.354 *** (0.087) 0.354 *** (0.087) 𝐿𝑖,𝑡−1 0.422 *** (0.085) 0.421 *** (0.085) 𝑀𝐵𝑖,𝑡−1 -0.000 *** (0.000) -0.000 *** (+0.000) ln (𝑆)𝑖,𝑡−1 -0.300 *** (0.018) -0.301 *** (0.018) 𝑅𝑖,𝑡−1 𝑀𝑖,𝑡−1 -0.319 * (0.182) 0.332 * (0.177) Observations 12,086 12,086 P(Wald test) 0.0495 0.000 Adj. 𝑅2 0.237 0.238
19
exists a positive relationship between R&D expenditure and dividend payment, because R&D adds value as a means to create new growth opportunities. The significant coefficient of 2.663 implies that every dollar invested in R&D on average delivers $2.66 of value. Because a dollar investment in R&D is a reduction in value, the “gross benefit” of R&D is even $3.66 for every dollar invested.2 The magnitude of this coefficient is striking: every R&D-expense more than triples the value of the money spent. Following these results, I do not reject hypotheses 1 and 2, and conclude that there are positive individual effects of R&D and dividends on stock performance.
A Wald test is performed in order to test for significance in 𝛾3 > 𝛾2, i.e. the effect of R&D on stock performance is greater than the effect of dividends on stock performance. For regression I, the p-value of the Wald test of 0.049 implies that the effect of R&D on performance is statistically significantly higher than the effect of dividends on performance at the 5 percent confidence level. These results are sufficient to conclude that the value-effect of R&D expenditure is on average larger than that of dividend payment.
5.2 – Interaction effects
With regard to the interaction hypotheses, both coefficients of the interaction terms in regression II of Table 4 show the expected sign, but without significance. Calculation of the maximum of the interaction effect (see section 3.2) gives
𝐼
𝑡𝑚𝑎𝑥=
𝛾3−2𝛾4 = 13.211
−54.056= 0.265, which means that the maximum value of interaction between R&D and dividend is achieved when the multiplication of dividends and R&D as fractions of the market capitalization equals 0.265. However, this number does not say very much without knowledge of the magnitude of R&D and dividends separately.
From the descriptive statistics in Table 2 can be seen that the average R&D spending is approximately 0.069 and the average dividend payment is only 0.012, both as fractions of the lagged market cap. An average firm would thus have an interaction of 0.000828, which is valued at 0.0109, while the optimal interaction is valued at 1.603.3 This means that the
2 E.g.: firm value = 100, 1 dollar R&D expense, firm value decreases to 99, but R&D-effect causes value to increase to
102.663. Therefore, gross benefit = 102.663 – 100 – 1 = 3.663
3 Average interaction = 0.000828 = 0.069 ∗ 0.012
Average value of interaction = 0.0109 = 13.211 ∗ 0.265 – 27.028 ∗ 0.2652
Maximum interaction = 0.265
20
average firm would miss out on 1.5921 (159 percent) return annually because of suboptimal dividend-R&D ratio setting. However, in order to live up to this “optimal” ratio, firms would have to invest around its whole market capitalization in the combination of dividend and R&D, a scenario that seems quite unrealistic. Graphically, the real world, with firms that work with much lower ratios than the maximum of 0.265, is on the left side of the line in Figure 2, indicating that there is a positive relationship between annual returns and the interaction of R&D and dividends.
Based on both the unrealistic coefficients and the insignificance of the results, hypothesis 4, proposing a negative interaction effect of dividends and R&D on performance, is rejected.
5.3 – Robustness
Based mainly on the Outcome model of La Porta et al. (2000), this paper hypothesizes that there should be a negative relationship between firm’s dividends payment and R&D expenditure. Table 5 presents the results of regressing R&D and dividends to each other. Although the relationship is estimated to be negative, the coefficient is not significant. Apparently, managers do not treat dividends and R&D as substitutes. Therefore, the Trade-Off Hypothesis is rejected. Although the relationship between R&D and dividends is not found to be significant, it is interesting to measure what are the value effects of deviating from the estimated regression line between R&D and dividends. Table 6 presents the results of two regressions that are performed, which replace the interaction terms of R&D and dividends by the squared residuals of the trade-off regressions. The coefficient of this new term tests whether it is true that, the more firms deviate from the estimated regression line, the larger the negative effect on value should be. The results are shown in Table 6. In both regressions, the squared residual coefficient reflects the correct, but insignificant sign. This insignificance is in line with the results from the main regression model and therefore confirms the rejection of hypothesis 4
21
TABLE 5
Regression results for the trade-off setting
This table presents the results of the panel data regression performed to test for the relationship between R&D expenditure 𝑅𝐷𝑖,𝑡
𝑀𝑖,𝑡−1 and dividend payment 𝐷𝑖,𝑡
𝑀𝑖,𝑡−1. In regression I, R&D is the dependent variable and in regression I, dividend payment is the dependent variable. Both variables are deflated by the lagged market capitalization 𝑀𝑖,𝑡−1, in order to control for size differences. The table presents coefficients, standard errors and probability values.
Fixed time effects and fixed cross-section effects are applied to the regressions. Furthermore, White diagonal standard errors are used in order to correct for potential heteroscedasticity.
Dependent variable 𝑹𝑫𝒊,𝒕 𝑴𝒊,𝒕−𝟏 𝑫𝒊,𝒕 𝑴𝒊,𝒕−𝟏 I II Coefficient (White SE) Coefficient (White SE) Intercept 0.068 *** (0.001) 0.006 *** (0.000) 𝐷𝑖,𝑡 𝑀𝑖,𝑡−1 0.047 (0.036) 𝑅𝐷𝑖,𝑡 𝑀𝑖,𝑡−1 0.005 (0.004) Observations 12,551 12,551 Adj. 𝑅2 0.502 0.248
22
TABLE 6
Regression results for the regression with residual terms
This table presents the results of the panel data regression performed to test for hypotheses 1, 2 (individual hypotheses), and 3 (interaction hypothesis). In both regressions, the dependent variable is annual
returns ∆𝑀𝑖,𝑡
𝑀𝑖,𝑡−1, defined as the change in market capitalization for firm 𝑖 in year 𝑡, with dividends re-invested
fictionally. The explanatory variables are dividends paid ( 𝐷𝑖,𝑡
𝑀𝑖,𝑡−1), R&D expenditure ( 𝑅𝐷𝑖,𝑡
𝑀𝑖,𝑡−1) in regression I. In regression I, the squared residuals of regression I (𝑈𝑖,𝑡2) of table 5 are added. Regression II adds the squared residuals of regression II (𝑉𝑖,𝑡2) of table 5.
𝐶𝑖,𝑡−1
𝑀𝑖,𝑡−1 is lagged cash holdings plus marketable securities, 𝐿𝑖,𝑡−1 is
leverage, 𝑀𝐵𝑖,𝑡−1 is lagged market-to-book ratio, ln (𝑆𝑖,𝑡−1) is the lagged natural logarithm of book value of assets (size), 𝑅𝑖,𝑡−1
𝑀𝑖,𝑡−1 is lagged total share repurchases. All variables that reflect absolute (dollar) values are deflated by the lagged market capitalization 𝑀𝑖,𝑡−1, in order to control for size differences. In order to test for the existence of a significant difference between the effects of 𝑅𝐷𝑖,𝑡
𝑀𝑖,𝑡−1 and 𝑅𝐷𝑖,𝑡
𝑀𝑖,𝑡−1 (hypothesis 3), a Wald test is performed, of which the p-value is presented in the table. Fixed time effects and fixed cross-section effects are applied to the regressions. Furthermore, White diagonal standard errors are used in order to correct for potential heteroscedasticity. I II Coefficient (White SE) Coefficient (White SE) Intercept 3.605 *** (0.216) 3.634 *** (0.216) 𝐷𝑖,𝑡 𝑀𝑖,𝑡−1 1.429 ** (0.600) 1.732 ** (0.813) 𝑅𝐷𝑖,𝑡 𝑀𝑖,𝑡−1 2.983 *** (0.241) 2.666 *** (0.239) 𝑈𝑖,𝑡2 -1.692 (1.085) 𝑉𝑖,𝑡2 -0.956 (2.410) 𝐶𝑖,𝑡−1 𝑀𝑖,𝑡−1 0.325 *** (0.086) 0.351 *** (0.087) 𝐿𝑖,𝑡−1 0.422 *** (0.083) 0.422 *** (0.085) 𝑀𝐵𝑖,𝑡−1 -0.000 *** (0.000) -0.000 *** (0.000) ln (𝑆)𝑖,𝑡−1 -0.299 *** (0.018) -0.300 *** (0.018) 𝑅𝑖,𝑡−1 𝑀𝑖,𝑡−1 -0.330 * (0.181) -0.320 * (0.181) Observations 12,086 12,086 P(Wald-test) 0.013 0.258 Adj. 𝑅2 0.238 0.237
23
6. Discussion and conclusion
The purpose of this article is to understand and analyze the individual and combined value of two types of firm expenditures, namely dividends and R&D. Using regressions from panel data on all firms listed on the NASDAQ in the period 1998-2017, I find evidence that both dividends and R&D are positively related to stock returns, and the dollar-impact on dividend re-invested share value is significantly higher for R&D ($2.66) than for dividends ($1.43). Furthermore, I find no evidence for the hypothesis that a trade-off exists between R&D and dividends. Finally, no evidence is found in favor of the hypothesis that dividend and R&D expenditure have a negative interaction effect on stock performance, despite their individual positive relationships to performance.
Since decades, it has been empirically confirmed that dividend payment as well as R&D spending on average positively relates to stock return. In the case of dividends, cash flow signaling is an important determinant in this positive relationship: a dividend increase is effectively a channel for a company to convince the market that future cash flows will be high enough to maintain the new dividend ratio. However, paying dividend may not be the most value-maximizing way of utilizing the company’s assets, as some companies might be able to deliver the shareholder a higher return to his/her funds by retaining earnings and re-investing them in profitable projects. Therefore, to really understand the effect of dividend policy, one needs to take into account the amount of profitable investment opportunities a firm will have in the future. This concept is mainly determined by the growth opportunities a firm has: a cash cow that pays dividends is probably reducing agency costs by paying out excess cash, while a star might be giving up major growth opportunities by distributing cash that could otherwise be invested in those opportunities.
24
and dividend payment, and (3) there exists a negative interaction effect of R&D and dividends on stock performance.
From regressions with 20-year panel data on 2,552 listed stocks from the NASDAQ constituent list, evidence is presented in favor of the individual hypotheses as well as the hypothesis that R&D is valued higher than dividend payment. Based on the regression, no evidence is found in favor of both the Interaction Hypothesis and the Trade-Off Hypothesis.
This paper confirms the hypothesis of the positive relationship between dividends and returns. Since the dependent variable is dividend re-invested return, only the signaling effects of dividends are measured. Therefore, the results are in line with the Cash Flow Signaling hypothesis (Miller & Rock, 1985; Bhattacharya, 1979): investors derive value from dividends, since this is a signal of expected future cash flows. Moreover, the common hypothesis that R&D and performance are related (Rein, 2004; Sivadas and Dwyer, 2000; Ayers and Dahlstrom, 1997; Salimi and Rezaei, 2018) is confirmed by this paper’s results.
A more innovative finding of this paper is the result that the dollar-value of R&D expenditure is significantly higher than that of dividend payment. This provides managers with new insights on how to spend every dollar of available cash. Apparently, one dollar of excess cash can on average deliver $1.23 of extra added value when investing it in R&D rather than paying it out as dividend.
25
References
Aharony, J., & Swary, I. (1980). Quarterly Dividend and Earnings Announcements and Stockholders' Returns: An Empirical Analysis. The Journal of Finance, 1-12. Asquith, P., & Mullins, D. W. (1983). The Impact of Initiating Dividend Payments on
Shareholders' Wealth. The Journal of Business, 77-96.
Ayers, D., Dahlstrom, R., & Skinner, S. (1997). Journal of Marketing Research. American Marketing Association, 107-116.
Barney, J. B. (1997). Gaining and Sustaining Competitive Advantage. Boston: Addison-Wesley Publishing Company.
Bartram, S. M., Brown, G., & Stulz, R. M. (2012). Why Are U.S. Stocks More Volatile. The Journal of Finance, 1329-1370.
Benartzi, S., Michaely, R., & Thaler, R. (1997). Do Changes in Dividends Signal the Future or the Past? The Journal of Finance, 1007-1034.
Bhattacharya, S. (1979). Imperfect Information, Dividend Policy, and "the Bird in the Hand" Fallacy. Bell Journal of Economics, 259-270.
Black, F. (1976). The Dividend Puzzle. The Journal of Portfolio Management, 5-8. Bulan, L., Subramanian, N., & Tanlu, L. (2007). On the Timing of Dividend Initiations.
Financial Management, 31-65.
Dittmar, A. K. (2000). Why Do Firms Repurchase Stock. The Journal of Business, 331-355. Fama, E. F. (1970). Efficient Capital Markets: A Review of Theory of Empirical Work. The
Journal of Finance, 383-417.
Faulkender, M., & Wang, R. (2006). Corporate Financial Policy and the Value of Cash. The Journal of Finance, 1957-1990.
Ghosh, C., & Woolridge, J. (1989). Stock-Market Reaction to Growth-Induced Dividend Cuts: Are Investors Myopic? Managerial and Decision Economics, 25-35. Grullon, G. M. (2002). Are Dividend Changes a Sign of Firm Maturity? The Journal of
26
Grullon, G., Michaely, R., Benartzi, S., & Thaler, R. H. (2005). Dividend Changes Do Not Signal Changes in Future Profitability. Journal of Business, 1659-1682.
Jensen, M. C. (1986). Agency Costs of Free Cash Flow, Corporate Finance, and Takeovers. The American Economic Review, 323-329.
Jensen, M., & Meckling, W. H. (1976). Theory of the Firm: Managerial Behavior, Agency Costs and Ownership Structure. Journal of Financial Economics, 305-360.
La Porta, R., Lopez-De-Silanes, F., Shleifer, A., & Vishny, R. W. (2000). Agency Problems and Dividend Policies around the World. The Journal of Finance, 1-33.
Lang, L. H., & Litzenberger, R. H. (1989). Dividend Announcements - Cash Flow Signalling vs. Free Cash Flow Hypothesis. Journal of Financial Economics, 181-191.
Lie, E. (2005). Operating performance following dividend increases and omissions. Journal of Corporate Finance, 27-53.
Miller, M. H., & Modigliani, F. (1961). Dividend Policy, Growth, and the Valuation of Shares. The Journal of Business, 411-433.
Miller, M. H., & Rock, K. (1985). Dividend Policy under Asymmetric Information. The Journal of Finance, 1031-1051.
Officer, M. (2010). Overinvestment, corporate governance, and dividend initiations. Journal of Corporate Finance, 710-724.
Pastor, L., & Veronesi, P. (2003). Stock Valuation and Learning about Profitability. The Journal of Finance, 1749-1789.
Pettit, R. (1972). Dividend Announcements, Security Performance, and Capital Market Efficiency. The Journal of Finance, 993-1007.
Pettit, R. (1976). The Impact of Dividend and Earnings Announcements: A Reconciliation. The Journal of Business, 86-96.
Porter, M. (1985). The Competitive Advantage: Creating and Sustaining Superior Performance. NY: Free Press.
27
Salimi, N., & Rezaei, J. (2018). Evaluating firms' R&D performance using best worst method. Evaulation and Program Planning, 147-155.
Sivadas, E., & Dwyer, F. (2000). An Examination of Organizational Factors Influencing New Product Success in Internal and Alliance-Based Processes. Journal of Marketing, 31-49.
Suurmeijer, M., Smid, P., & Eije, von, H. (2013). Research and Development and Firm Risk. Journal of Corporate Finance Research, 7-26.
Watts, R. (1973). The Information Content of Dividends. The Journal of Business, 191-211. Wernerfelt, B. (1984). A Resource-Based View of the Firm. Strategic Management Journal,
