Does market timing drive capital structures? A panel data study for Dutch firms

Transcription

1 DNB WORKING PAPER Does market timing drive capital structures? A panel data study for Dutch firms Tijs de Bie and Leo de Haan No. 16/November 2004

2 Does market timing drive capital structures? A panel data study for Dutch firms Tijs de Bie and Leo de Haan * * Views expressed are those of the individual authors and do not neccessarily reflect official positions of De Nederlandsche Bank. Working Paper No. 016/2004 November 2004 De Nederlandsche Bank NV P.O. Box AB AMSTERDAM The Netherlands

3

4 Does market timing drive capital structures? A panel data study for Dutch firms Tijs de Bie * Leo de Haan November, 2004 ABSTRACT This empirical study revisits the determinants of firms' capital structures. The main focus thereby is on the 'market timing theory', according to which the current level of the capital structure is the cumulative outcome of past attempts to time the market, i.e. issuing shares when equity is overvalued and repurchasing shares in case of undervaluation. Since the positive evidence for this theory found by Baker and Wurgler (2002) for the US, this strand of empirical literature is growing. This paper presents evidence for a sample of 135 Dutch listed non-financial firms over the period as well as for a sub-sample of 45 Dutch firms that did an initial public offering (IPO). The research methodology follows Kayhan and Titman (2004), who model capital structure as a mix of market timing, pecking order and capital structure targeting behaviour. The findings for the Dutch sample do not find strong and persistent effects of market timing on capital structures. JEL codes: C23, G32 Key words: Market timing, Capital structure * De Nederlandsche Bank, Financial Markets Division and Research Division, respectively. The first author was intern at the Research Division during the research. t.m.de.bie@dnb.nl and l.de.haan@dnb.nl. This paper has benefited from useful comments from Lex Hoogduin, Frank de Jong, Jeroen Ligterink and seminar participants at De Nederlandsche B ank (July 2004).

5 2 1 INTRODUCTION Since the publication of the Modigliani and Miller s (1958) irrelevance theory of capital structure, the theory of corporate capital structure has been a topic of great interest to finance economists. Over the years two major theories of capital structure emerged which diverge from the assumption of perfect capital markets under which the irrelevance model is working. The first is the static trade-off theory which assumes that firms trade off the benefits and costs of debt and equity financing and find an optimal capital structure after accounting for market imperfections such as taxes, bankruptcy costs and agency costs. The second is the pecking order theory (Myers, 1984, Myers and Majluf, 1984) that argues that firms fo llow a financing hierarchy to minimize the problem of information asymmetry between the firm s managers-insiders and the outsiders-shareholders. Recently, Baker and Wurgler (2002) have suggested a new theory of capital structure: the market timing theory of capital structure. This theory states that the current capital structure is the cumulative outcome of past attempts to time the equity market. Market timing implies that firms issue new shares when they perceive they are overvalued and that firms repurchase own shares when they consider these to be undervalued. Market timing issuing behaviour has been well established empirically by others already, but Baker and Wurgler show that the influence of market timing on capital structure is highly persistent 1. Baker and Wurgler s persistency effect raised considerable attention among finance researchers. Other studies followed examining whether market timing indeed has a persistent influence on capital structures. Kayhan and Titman (2004) model capital structure as a mix of market timing, pecking order and capital structure targeting (i.e. trade-off) behaviour. They find pecking order, market timing and targeting behaviour to be aspects of the short-term capital structure choice. The long-term effect for market timing- and pecking order behaviour is not found to be strong though. Firms tend to move towards target leverage ratios as predicted by the trade-off theory of capital structure. This study applies Kayhan and Titm an s (2004) research methodology to a sample of Dutch listed firms. The variable to be explained is the change in leverage and the explanatory variables are proxies for pecking order, market timing and trade-off financing behaviour. More specifically, both financial deficit and internal cash flow are used as proxy variables for pecking order behaviour. When the pecking order theory holds, a negative relation between these variables and the change in leverage is expected. The effect of equity market timing on the capital structure of firms is examined by investigating the relationship between changes in leverage on the one hand and fluctuations in stock prices and market-to-book values on the other. Finally, capital structure targeting is examined by 1 See section 2.2 for an overview of evidence on market timing.

6 3 testing the relationship between changes in leverage and deviations from optimal leverage ratios. As in Kayhan and Titman, the analysis focuses both on the relationships in the short and in the long term, i.e. its persistence. The paper proceeds as follows. Section 2 presents an overview of the traditional theories of capital structure and discusses the recently developed market timing theory of capital structure. Furthermore, this section gives an overview of empirical studies related to the latter theory. Section 3 describes the sample selection and characteristics of the Dutch firms in the data set. The fourth section follows with an outline of the methodology and the model specificat ion. The fifth section sets out the results of the empirical analysis, section 6 analyses the robustness of these results. Sec tion 7 closes the paper with a summary and conclusions.

7 4 2 MARKET TIMING: REVIEW OF THEORY AND EVIDENCE In order to position the new market timing theory of capital structure in the existing theory of capital structure, in this section a short account is given of the most influential theories to date: the static trade-off theory and the pecking order theory. 2 Subsequently, the empirical evidence until now is discussed. 2.1 Capital structure theory The static trade-off theory argues that firms have optimal capital structures, which they determine by trading off the costs against the benefits of the use of debt and equity. One of the benefits of the use of debt is the advantage of a debt tax shield. One of the disadvantages of debt is the cost of potential financial distress, especially when the firm relies on too much debt. Already, this leads to a trade-off between the tax benefit and the disadvantage of higher risk of financial distress. But there are more cost and benefits involved with the use of debt and equity. One other major cost factor consists of agency costs. Agency costs stem from conflicts of interest between the different stakeholders of the firm and because of ex post asymmetric information (Jensen and Meckling (1976) and Jensen (1986)). Hence, incorporating agency costs into the static trade-off theory means that a firm determines its capital structure by trading off the tax advantage of debt against the costs of financial distress of too much debt and the agency costs of debt against the agency cost of equity 3. Many other cost factors have been suggested under the trade-off theory, and it would lead to far to discuss them all. Therefore, this discussion ends with the assertion that an important prediction of the static trade-off theory is that firms target their capital structures, i.e. if the actual leverage ratio deviates from the optimal one, the firm will adapt its financing behaviour in a way that brings the leverage ratio back to the optimal level. The pecking order theory does not take an optimal capital structure as a starting point, but instead asserts the empirical fact 4 that firms show a distinct preference for using internal finance (i.e. retained earnings and excess liquid assets or slack ) over external finance. If internal funds are insufficiently available to finance investment opportunities, firms may or may not acquire external financing, and if they do, they will choose among the different external finance sources in such a way as to minimise additional costs of asymmetric information. The latter costs basically reflect the lemon premium (Akerlof, 1970) that outside investors ask for the risk of failure for the average firm in the market. This 2 The theory of capital structure has been documented extensively in the finance literature of the past decades and is well known to finance researchers; in this section only a short review is provided of the most important and most empirically tested theories. 3 Among others, Titman and Wessels (1988), Rajan and Zingales (1995), Hovakimian, Opler and Titman (2001) support the believe that firms maintain, or try to keep up with, a certain target leverage. 4 The pecking order theory was first proclaimed by Donaldson (1961), in a survey study among US firms.

8 5 implies that above-average quality firms are confronted with a premium which they do not deserve, and it is this wedge driven between the cost level for internal capital on the one hand and external financing on the other which may induce such high quality firms to retreat from the capital market. This adverse selection problem forms the foundation of the pecking order theory of finance. The resulting pecking order of financing is as follows: internally generated funds first, followed by respectively (low-risk) debt financing and share financing (Myers (1984) and Myers and Majluf (1984)). The market timing theory of capital structure argues that firms time their equity issues in the sense that they issue new stock when the stock price is perceived to be overvalued, and buy back own shares when there is undervaluation. Consequently, fluctuations in stock prices affect firms capital structures. Two theories explain the existence of market timing behaviour by firms. The first assumes economic agents to be rational. Companies are assumed to issue equity directly after a positive information release which reduces the asymmetry problem between the firm s management and stockholders. The decrease in information asymmetry coincides with an increase in the stock price. Hence, firms create their own timing opportunities (Lucas and McDonald (1990); Korajczyk, Lucas and McDonald (1992)). The second theory assumes (one or both of) the economic agents to be irrational (Baker and Wurgler, 2002). Due to irrational behaviour there is a time-varying mispricing of the stock of the company. When managers perceive the stock to be mispriced (and when they act in the interest of the existing shareholders) they will issue equity when the stock is overvalued and buy own shares back when it is undervalued. 2.2 Existing empirical evidence Empirical evidence in general supports both the pecking order and the trade-off theory. Empirical tests to examine whether the pecking order or the trade-off theory is a better predictor of observed capital structures yield mixed results (e.g.: Shyam -Sunder and Myers, 1999; Hovakimian, Opler and Titman, 2001; Fama and French, 2002; Frank and Goyal, 2003). Most of these papers find support for both theories of capital structure. For the Netherlands, empirical evidence also gener ates mixed results. De Haan en Hinloopen (2003) for example find evidence in favour of pecking order financing behaviour while De Jong and Veld (2001) do not; they find more evidence for capital structure targeting. Baker and Wurgler (2002, hereafter BW ) provide evidence that equity market timing has a persistent effect on the capital structure of the firm. More specifically, they define a market timing measure, which is a weighted average of external capital needs over the past few years, where the weights used are market to book values of the firm. By construction, this measure has a high value when firms use relatively much external capital while market values are high. BW indeed find that leverage changes

9 6 are strongly and positively related to their market timing measure. BW therefore conclude that the capital structure of a firm is the cumulative outcome of past attempts to time the equity market. The BW paper received considerable attention. The practice of market timing had already been documented before BW in many empirical studies. 5 Hence, that was not the issue. It was the long term persistence that market timing would have on capital structures, which drew considerable attention. The BW study initiated studies test ing the consequences of changes in equity market valu ation on capital structure. Up till now, these studies are only performed for the US market. These US-studies are discussed next 6. Welch (2004) follows the idea of BW and concludes that capital structure ratios (in market values) are strongly determined by past stock prices. Welch concludes that firms do not compensate for the effects of stock prices on capital structure. In his view, changes in leverage are not caused by market timing efforts, but are the result of the unwillingness of firms to counteract the effects of stock price changes on capital structure. According to Welch s inertia theory, the motives for firms to issue remain a mystery. Proxy variables as the market-to-book ratio, profitability and market timing fail to explain market based capital structure dynamics. When looking at variations in the market based capital structure ratios, stock price effects are the main determinants and explain about 40% of capital structure ratio dynamics. Frank and Goyal (2004) examin e the influence of market conditions on capital structure changes using a vector auto regression (VAR) framework. They find that in the long term there is a capital structure ratio to which firms seem to revert. High market-to-book ratios have a short term influence on debt issuance (debt is reduced), but there is no clear relationship between market valuations and equity issue activity. They conclude that the evidence is in favour of the trade-off theory. Hovakimian (2003) finds that the importance of historical average market-to-book in leverage regressions is not due to past equity market timing. He finds that only equity issues may be timed to 5 Other studies which show that seasoned equity offerings coincide with high market valuations come from: Taggart (1977), Marsh (1982), Asquith and Mullins (1986), Korajczyk, Lucas and McDonald (1991), Jung, Kim and Stulz (1996), Hovakimian, Opler and Titman (2001). Share repurchases with low market valuations are given proof of in: Ikenberry, Lakonishok and Vermaelen (1995). Long term underperformance following seasoned equity offerings is indirect evidence for firms timing the market, among others this is shown in: Loughran and Ritter (1995, 1997), Spiess and Affleck-Graves (1995), and Eckbo, Masulis and Norli (2000). A third evidence for market timing is the survey interview of Graham and Harvey (2001) in which CFO s admit that the market value of the stock (and the recent price run-up) is an important consideration for the timing of the equity issue. 6 This paragraph focuses on studies which examine the long term influence of changes in stock market values on capital structure ratios and which appeared after BW s publication in We omit studies that are based on different definitions for market timing than BW use (for instance: Alti, 2003) and on studies that primarily focus on rebalancing behaviour in accordance with an optimal target (for instance: Leary and Roberts, 2004, Hovakimian, 2002).

10 7 conditions in equity market, but they do not have significant long-lasting effects on capital structure. Other transactions (equity repurchases, debt issues and debt reductions) exhibit timing patterns that are unlikely to induce a negative relation between market-to-book and leverage. Kayhan and Titman (2004) analytically split BW s market timing measure into two components, one short-term and one long-term timing measure. They do confirm that leverage changes are driven by market-timing (next to pecking-order and trade-off behaviour). However, in contrast to BW, they do not confirm the long term persistency of market timing effects. All of the above mentioned studies are for US firms. For the Netherlands De Haan and Hinloopen, (2003) find evidence that firms are more likely to issue new shares after stock price increases, De Jong and Veld (2001) observe negative excess returns after seasoned equity issues which could be indirect evidence for market timing. However, there are no studies for the Netherlands on the consequences of market timing for capital structures. This is the aim of this paper. In the next section the data are discussed.

11 8 3 DATA Data are taken from financial statements of Dutch listed firms, as published in the Jaarboek van Nederlandse Ondernemingen. The initial sample consists of 156 Dutch non-financial companies listed on the Amsterdam Stock Exchange (AEX) and covers the period from 1983 to We use two samples, one regular calendar time sample and one special IPO-time sub-sample consisting only of firms that did an initial public offering (IPO). This facilitates exploitation of any distinctive characteristics of firms that have just gone public 7. It is conceivable that financing decisions are made differently by a firm which has already been publicly listed for a long time than by a firm which has just done an initial public offering. Companies are selected for the calendar-time data set if at least eight consecutive years of annual account data are available. This leaves us with a calendar time (unbalanced) panel of 135 companies. The second data set is reorganized into IPO-time. The first observation for each firm is the year of its initial public offering. To be included in this sample a firm s IPO year should be identifiable, be equal to or greater than 1983, and be followed by at least eight consecutive years of annual accounts data. Dates for initial public offerings are identified from the Amsterdam Exchange annual reports and other sources 8. This results in an IPO-time data set of 45 companies. Definitions for all variables are provided in Appendix A. Noticeable is the definition for leverage and its components. The definitions of the variables follow the definitions as used by the US studies with CompuStat 9 data such as BW (2002) and KT (2004) as closely as possible. Due to differences in accounting standards between the US and the Netherlands, some corrections had to be made in the Dutch balance sheet data in order to construct comparable debt ratios. 10 The Dutch annual accounts of firms specify provision items on the liabilities side of the balance sheet that do not exist in the US. Therefore, pension provisions and revaluation provisions are excluded when calculating debt ratios. Another reason for this adjustment is to avoid the often severe yearly fluctuations in the revaluation provision to affect the leverage ratio. Table 1 gives a first impression of capital structure-, financing-, and firm performance measures. Panel A lists results for the calendar time data set, while panel B provides results for the IPO data set. 7 In this we follow Kayhan and Titman (2004) who use a calendar time data set and Baker and Wurgler (2002) who use the IPO time data set to study changes in capital structure. 8 Additional information on initial public offerings is taken from IPO -lists provided by the Euronext Amsterdam. 9 COMPUSTAT is provided by Standard and Poor s and offers all kinds of financial information on publicly traded companies in the U.S. and around the world. 10 See Nobes and Parker (2000), sections 7 and 8, for a detailed explanation of differences in accounting practices between the US and the Nether lands.

12 9 Table 1 Firm characteristics (arithmetic averages) Lev BV LevMV e d RE DIV NetE ROA M/B TANG GRT SAL INV COSTS PANEL A: CALENDAR TIME DATA SET Mean PANEL B: IPO TIME DATA SET IPO = n/a n/a n/a 11.7 n/a 1.40 IPO IPO IPO IPO IPO IPO IPO IPO Mean Explanatory note. For variable definitions see Appendix A. RE, e, d, and DIV are reported in percentage points. Total net earnings (NetE) are the sum of dividends plus retained earnings. TANG and COSTS are scaled on the end of year value of total assets. GRT and INV are scaled on the beginning of the year value of total assets. From panel A it becomes clear that average book leverage (LevBV) moves around a ratio of 0.60 and is rather stable over the years. In terms of market value, leverage (LevMV) has declined from 0.64 in 1984 to 0.39 in This decline can be explained by the increase of the capitalization of firms over those years. In this period the stock exchange climate was favourable and many companies experienced high stock price increases. The year 1987 shows a peak for the market leverage. In this year stock market prices collapsed because of the crash on the stock exchange.

13 10 Figure 1 Financing flows for the calendar group of firms. Percentages of total assets Year (equity issues) (debt issues) (retained earnings) Figure 1 represents the financing flows for the firms in the calendar time data set; figures are taken from Table 1. Of the financing activities, net equity issues (e) are quite stable over the years and amount one to two percent of total assets on average each year. For the Dutch market, the years 1986 and 1989 are known as hot issue years in which many firms went public and there were high stock turnovers. There was a favourable issue climate in both of these years which was reflected in relatively high market-to-book ratios. Net debt issues (d) are more volatile over the years. On average debt issues comprise 2.6 percent of total assets, which is more than two and a half times the figure for net equity issues. Retained earnings (RE), on average, are almost equal in size to net debt issues and appear to be a major source of financing. ROA and net earnings, which are measures of profitability and performance, indicate that since 1993 firms experienced good operational years, during which net earnings were above average. The year 1989 is the most profitable year in the sample period. The IPO data shows a relatively high market-to-book ratio, especially in the first four years after the introduction on the exchange. Taking market to book as a proxy of growth opportunities (Marsh, 1982), this suggests that IPO firms have above-average growth opportunities. This is also what could be expected. The firm seeking an intitial listing on the public stock exchange is often doing so because it needs large amounts of funds for its investment plans. The investment rate of IPO firms is also relatively high in the first, second and third year after the IPO. The same can be observed for equity financing and debt financing in the years following the IPO. Next to these external funds, internally generated funds are also an important source of financing in the first four years after the stock quotation. In the IPO data set, financing with retained earnings on average even exceeds debt- and external equity financing and it exceeds equity financing in all the years. Book leverage on average seems to be around the 60 percent leverage ratio as it was in the calen dar time data set. After a rise from 0.38 in the year of the IPO to 0.47 in year five, market value leverage

14 11 becomes fairly stable over the subsequent years and reverts to its mean ratio of 0.43 in year eight. The market value leverage for the IPO firm is lower than that for the average firm in panel A, reflecting higher market-to-book ratios.

15 12 4 MODEL SPECIFICATION The methodology used in this study closely follows KT (2004). KT explore the effects of market timing for the development of the capital structures of firms, controlling for any effects of alternative behavioural assumptions, particularly capital structure targeting and pecking order financing. They explore both short and long term effects of market timing. If market timing has both short and long term effects, KT conclude that market timing effects are persistent. The variable to be explained is the change in leverage during a certain period, where leverage can be measured either in market values or book values. Hence, these are the variables LevMV and LevBV introduced in Section 3. The explanatory variables in the model are proxy variables representing market timing, targeting, and pecking order financing, respectively. We first introduce the proxy variables for pecking order, market timing and targeting behaviour, then discuss the specification of the model Proxy variables Market timing As discussed in Section 2.2, the market timing theory of capital structure claims market valuation of companies to be an important factor for the decision to issue debt or equity. When firms experience high market-to-book ratios or recent stock price increases they are expected to issue more equity than debt, resulting in a lower leverage ratio. The stock return variable (r), measured as the cumulative log return on the stock over the years t-4 till year t, is the first proxy used to account for the impact of market timing behaviour on capital structure changes. We expect a negative relationship between leverage changes and stock price increases 11. The second proxy of market timing is BW s timing measure. BW use a measure for equity market timing, the socalled market-to-book external finance weighted average ratio (MB efwa ), defined as : MB efwa FD*( MB / ) =, (1) FD 11 Welch (2004) examines the relationship between stock price changes and market leverage ratios. He claims that firms do nothing to counteract the influence of stock price changes on leverage rat ios. As a consequence, changes in stock prices are negatively and closely related to changes in market value lever age. This is called Welch s inertia theory.

16 13 where FD is the financial deficit of the firm, i.e. the external funds actually used, defined as the sum of equity and debt issues, respectively: FD= e+ d (2) and M/B is the market-to-book ratio. Σ accumulates the financial deficits over a range of past years, in BW being all the years since the initial public offering. For a firm being at time t, the MB efwa is the weighted average of past market-to-book ratios, starting at the year of the IPO till year t-1. The weight for each year is the ratio of the financial deficit in that particular year to the total financial deficit accumulated over all the years since the IPO. The resulting relationship between this measure and leverage is supposed to be negative, becaus e high market-to-book values are supposed to coincide with a higher share of equity issues than debt issues. BW confirm this relationship. KT criticise the use of the BW timing variable as a proxy for equity timing. They decompos e MB efwa to elucidate their reason for this criticism: MB efwat, 1 = cov( FD, MB / ) ( MB / ) FD + (3) The first term in this decomposition, i.e. the covariance between external financing and market to book (scaled by the average level of external financing over the sample period), indeed reflects BW s idea of market timing. The second term, however, i.e. the average market to book ratio over the sample period ( MB / ), has nothing to do with timing attempts. In the regression analysis, this second term can have a negative relationship with leverage because firms that experience high growth opportunities may choose not to finance with debt, but with equity, to maintain their financial flexibility (Myers, 1977). A high market-to-book ratio is often considered to be a proxy for high growth opportunities (Marsh, 1982). KT redefine BW s timing measure into two components: a Yearly Timing and a Long-Term timing measures. To arrive at these measures, KT multiply both terms in the former equation by FD to get the so-called Yearly Timing measure: YT = cov( FD, MB / ) (4) and the Long-term Timing measure:

17 14 LT = ( MB / )*( FD) (5) Both measures are calculated over year t-4 to year t. The YT measure, the covariance betw een external financing and market valuation, is a direct measure of market timing. YT is expected to be negatively related to leverage. The LT measure is also expected to have a negative relationship with leverage, but that may reflect other factors than market timing. As already mentioned, high market-to-book firms often have many growth opportunities and for that reason may avoid using too much debt (Myers, 1977). This has nothing to do with equity market timing behaviour. LT also captures the fact that the cost of external equity capital is inversely related to the company s stock price Capital structure targeting The trade-off theory of capital structure suggests that firms have a target capital structure which is determined by balancing the costs and benefits of debt financing. If a firm is overleveraged compared to its target capital structure, trade-off theory predicts the firm to lower its debt ratio in subsequent periods. Therefore, one of the explanatory variables used in the analysis is the deviation of the actual leverage from the target leverage level (Lsurp). The target leverage is approximated by regressing leverage on a set of firm specific variables and using the residuals from this regression as deviations from the target capital structure. We have chosen four variables as determinants of the target leverage. These factors, which have been found to be significant by Rajan and Zingales (1995), are the size of the company, asset tangibility, the market-tobook ratio and profitability. Industry dummies are also included in the target leverage regression to correct for industry specific effects 12. Appendix B provides detailed results for this regression Pecking order financing The pecking order theory predicts that firms with fin ancial deficits tend to increase their debt ratios, because they prefer debt over equity when acquiring external capital. We use a financial deficit measure (FD) to control for pecking order behaviour. The financial deficit is defined by the sum of net equity issues (e) and net debt issues (d) 13. In their studies, KT (2004), Frank and Goyal (2003) and 12 For an overview of industry specification, see Appendix C. 13 BW (2002) and KT (2004) use this balance sheet approach to calculate the financial deficit of the firm. Frank and Goyal (2003) and Shyam -Sunder and Myers (1999) calculate this variable from the cash flow statements.

18 15 Shyam-Sunder and Myers (1999) find that FD is positively related to changes in leverage which indicates pecking order financing behaviour. A financial deficit may have a different effect on leverage than a financial surplus, for example because issuing shares may encounter more asymmetric information problems than buying back own shares (KT, 2004). For this reason, we interact a dummy variable d with the financial deficit variable : FDd. The dummy takes the value of 1 when the financial deficit is positive and the value of zero otherwise. The second pecking order variable used in the model is a profitability measure. If firms generate a lot of internal cash, investment can more easily be financed with retained earnings. According to the pecking order theory, firms prefer to use internal finance anyway, implying a negative relationship between profitability and leverage. The profitability measure is defined as earnings before interest, taxes and depreciation (EBITD), scaled by total assets at the start of the year Specification of the model The equation to be estimated relates the change in the leverage ratio to the explanatory variables mentioned above. Following KT, we specify the equation for two time horizons: a short run and a long run. We define the short run or period 1 as a five year episode running from t-4 to t. 15 Flow variables such as FD are accumulated over these five years. Hence, the equation for period 1 reads: β r + + β L - L = α + [, -4] β YT itit [ itit, -4] + β LT[ itit, -4] LSurp it it 4 it it-4 + β FD + β FDd β, -4, -4 7 EBITD itit itit [ itit, -4] + ε it (6) where L is the leverage ratio in book or in market values (LevBV, LevMV ), i denotes the individual firm and t denotes the year. α it is the firm and time specific constant, and ε it is the residual term. Following KT, the same equation is specified for the change in leverage in period 2 which runs from t-8 and t and hence is twice as long as period 1. KT test whether the market timing variables (or any other variable) over a five year period t-8 to t-4 still affect the change in leverage over a nine year period running from t-8 to t. They call this the test for persistence. The equation reads: 14 The beginning period asset value in book leverage regressions is the book value of total assets. In the market leverage regression the beginning period asset value is the market value of total assets. For definitions see Appendix A. 15 Period 1 covers five years, instead of the six years used by KT. This is done to save observations. The KT sample ranges over 29 years ( ), whereas our sample has 14 years ( ). The robustness analysis in section 6 discusses the outcomes of the study when using a five year instead of a four year period.

19 16 β r L - L = α + [ 4, -8] β YT it it [ it 4, it-8] + β LT [ it 4, it-8] LSurp it it 8 it it-8 + β FD + β FDd + + β β 4, -8 4, -8 7 EBITD it it it it [ it 4, it-8] + ε it (7) Finally, KT also specify a third version of the equation, which differs from (7) in that the change in leverage, to be explained, is taken between t-4 and t: βr + + β LSurp L - L = α + β YT + β LT [ it 4, it-8] [ it 4, it-8] [ it 4, it-8] it it 4 it it-4 + β FD + β FDd 5 it 4, it-8 6 it 4, it -8 β 7 EBITD[ it 4, it-8] ε + + it (8) KT denote this a test for reversal effects, that is the possibility that the direction of the effects switches sign from equation (6) to (8). That would indicate that the effect of pecking order and timing variables on the debt ratio is later reversed, which would be a reason for non-persistence.

20 17 5 RESULTS In this section we present the estimation results for equations (6), (7) and (8) for the sample of Dutch firms. The equations have been estimated using a Generalized Least Square regression (GLS) with random effects. In order to get consistent standard errors bootstrapping techniques are used (Efron, 1979), as did KT. This corrects for biases resulting from using multiple observations and moving lagged variables. We draw 200 bootstrapped samples by repeated sampling, with replacement from the population. The sample drawn during each replication is a bootstrap sample of clusters, which preserves the time-series structure of the data. A cluster is formed by observations of one firm (it is assumed there is no cross-section correlation). Table 2 presents the results of the estimation of equation (6), which tests whether the change of leverage over the past five year period can be explained by market timing, while controlling for targeting and pecking order financing during the same period. 5.1 What drives capital structure developments? Table 2 Estimation results for equation (6): L t - L t-4 = α 0 + β 1 r tt,-4 + β 2 YT tt,-4 + β 3 LT tt,-4 + β 4 LSurp t-4 + β 5 FD tt,-4 + β 6 FDd tt,-4 + β 7 EBITD tt,-4 + ε t Book leverage Market leverage I II III I II III Estimate SE Estimate SE Estimate SE Estimate SE Estimate SE Estimate SE r * * * * * * 0.01 YT * * LT * * Lsurp * * * * * * 0.05 FD * 0.06 FDd 0.149* * EBITD * * * * * * 0.03 Explanatory notes. The it subscripts in the model have been suppressed for reasons of space. Industry dummies have been included (not reported). SE is the standard error, estimated using 200 bootstrapping replications. * Denotes significance at the 5% level. The significantly negative coefficient estimate for the stock return r in Table 2 indicates that firms use relatively more equity after periods of a stock price increase, which indicates market timing effects on capital structure. This outcome is consistent with the empirical evidence of KT for the US and earlier evidence for the Netherlands. Both De Jong en Veld (2001) and De Haan and Hinloopen (2003) found evidence for equity market timing in the issuing policy of Dutch firms.

21 18 Neither of the timing measures YT and LT are significant. This outcome is in contrast to the results of KT (2004), who did find significant negative coefficients for both market timing measures for US firms. In view of the strong positive correlation betw een FD and LT 16, we re-estimate the KT model leaving out either FD (and FDd) or LT (regressions II and III, respectively). In regression II both timing variables YT and LT become highly significant, but have opposite signs compared to KT. This would imply that Dutch firms issue more debt instead of equity when stock market valuations are high. Re-estimatio n of the model while leaving LT out (regression III) does not alter the outcomes compared to the full model ones. Hence, the KT equation turns out to be not robust to the exclusion of the FD variable when applied to the Dutch sample. Therefore, our results are inconclusive as to the influence of market timing behaviour on capital structure. The significant negative sign found for the deviation from the target leverage Lsurp provides evidence that firms try to keep their leverage on target, in accordance with the trade-off theory. The negative sign of the Lsurp variable suggests that overleveraged firms bring back their leverage ratios towards the target level. These findings are consistent with KT. De Jong and Veld (2001) also found evidence that Dutch firms issue more shares than bonds when leverage ratios are above target. De Haan and Hinloopen (2003) confirmed their finding, but observed that firms at the same time counteract this targeting behaviour by taking on more private debt and using less internal equity. The coefficient estimate for the financial deficit variable FDd in regression I reveals a positive relationship between the occurrence of a financial deficit and a change in leverage, which is in line with the findings of KT for the US. This is consistent with pecking order financing, which predicts that firms prefer debt over equity when they are short of internal finance and acquire additional funds in the external capital market. The deficit variable FD itself is not significant though, so there is no clear inference to be made whether companies prefer to retire capital through either debt reductions or stock repurchases. This outcome may be related to the unpopularity of stock repurchases in the Netherlands during the sample period 17. The evidence for a preference for internal funds is consistent with earlier finding for the Netherlands. Cools (1993), De Haan, Koedijk and De Vrijer (1994), and De Haan and Hinloopen (2003) found that Dutch firms prefer to finance with retained earnings. 16 The correlation coefficient between FD and LT is 0.90 and between FDd and LT These high correlations are to be expected, as LT is the four year average market-to-book ratio multiplied by the four year average financial deficit ratio. 17 Till the end of 2001, the repurchase of own stock was unattractive because of taxes. The treasury in the Netherlands considered the difference between the repurchase price and the original issue price to be equivalent to a dividend payment and taxed this difference at a 25% rate. Some compani es made special agreements with the treasury though, or used special constructions to lower the equity funds of the company without paying taxes. For example, KLM bought back shares from the KLM pension funds thus avoiding dividend tax.

22 19 Another piece of evidence in favour of the pecking order theory is the significant ly negative coefficient estimate for EBITD. In accordance with the prediction of the pecking order theory, highly profitable firms have lower leverage ratios than low -profitable firms. KT (2004) find the same. Summing up, we find evidence for capital structure targeting and pecking order financing, but we find only weak evidence for market timing effects on capital structures of Dutch companies. 5.2 Are the effects persistent? This section examines whether the influence of the explanatory variables on leverage are persistent. Following KT, that is accomplished by estimating equation (7) where the influences are extended over nine years. The regression results in Table 3 show that financial deficit FDd and profitability EBITD have persistent effects on leverage. This suggests that pecking order behaviour has a long-term effect on capital structures. The cumulative stock return variable r looses its significance, indicating that equity market timing has no long-term influence on the leverage ratio. Table 3 Estimation results for equation (7): L t - t t 4, t-8 2 t 4,-8 t β 3 t 4,-8 t β 4 t-8 5 t 4,-8 t β 6 t 4,-8 t β 7 t 4, t-8 t Book leverage Market leverage I II III I II III Estimate SE Estimate SE Estimate SE Estimate SE Estimate SE Estimate SE r YT LT * * Lsurp * * * * * * 0.05 FD FDd 0.123* * * * 0.11 EBITD * * * Explanatory notes. The it subscri pts in the model have been suppressed for reasons of space. Industry dummies have been included (not reported). SE is the standard error, estimated using 200 bootstrapping replications. * Denotes significance at the 5% level. 5.3 Reversal effects In this section equation (8) is estimated, to test for reversal effects (see section 4.2). Reversal effects are present when the coefficients have significant and opposite signs in equation (6) and (8), or in Table 2 and 4, respectively. In both the book and the market leverage regressions reversal effects are found only for the stock return variable r. Hence, the non-persistency found earlier for this variable appears to be driven by

23 20 reversal effects. Our conclusion, that there are persistent pecking order effects on leverage, while at the same time there is targeting behavior, is enhanced by this test. Table 4 Estimation results for equation (8): L it - L it 4 = α it + β 1 r it 4, it-8 + β 2 YT it 4, it-8 + β 3 LT it 4, it-8 + β 4 LSurp it-4 + β 5 FD it 4, it-8 + β 6 FDd it 4, it-8 + β 7 EBITD it 4, it -8 + ε it Book leverage Market leverage I II III I II III Estimate SE Estimate SE Estimate SE Estimate SE Estimate SE Estimate SE r 0.025* * * * * * 0.01 YT LT Lsurp * * * * * 0.07 FD * * 0.07 FDd 0.180* * * * 0.10 EBITD * * * Explanatory notes. The it subscripts in the model have been suppressed for reasons of space. Industry dummies have been included (not reported). SE is the standard error, estimated using 200 bootstrapping replications. * Denotes significance at the 5% level. 5.4 Analysis of the IPO time data set impact of the history This section repeats the analysis for the IPO data set of firms (see section 3 for details). Table 5 presents estimates for changes in leverage over period one. A striking outcome for the IPO firms is that there is no evidence of pecking order behaviour. In the book leverage regressions the FDd variable is still significantly positive, but the coefficient estimate for EBITD is significantly positive instead of negative. Apparently, IPO firms have not come to the equity market for nothing; they indeed are there to attract external equity. IPO firms aim at a certain target capital structure, as shown by the highly significant negative sign for the deviation from the target leverage variable Lsurp in both the book and market leverage regression. Hence, in this respect IPO firms do not behave differently. The evidence for market timing effects on capital structure, already weak for the full sample, is absent for the sample of IPO firms. In the book leverage regression I, the coefficient estimate for the stock return r is insignificant, which means no additional equity capital is attracted when firms experience high stock price increases. The stock return variable r is significant in the market leverage regression, but this may also be due to the mechanical inverse relationship between stock price changes and the market value of equity. The market leverage ratio may fall when stock prices rise, even if no equity issue has occurred.

24 21 Table 5 Estimation results for equation (6): - IPO firms L t - L t-4 = α 0 + β 1 r tt,-4 + β 2 YT tt,-4 + β 3 LT tt,-4 + β 4 LSurp t-4 + β 5 FD tt,-4 + β 6 FDd tt,-4 + β 7 EBITD tt,-4 + ε t Book leverage Market leverage I II III I II III Coef SE Coef SE Coef SE Coef SE Coef SE Coef SE FD * 0.06 FDd 0.229* * YT * LT * * r * * * 0.01 EBITD * Lsurp * * * * * * 0.08 Explanatory notes. The it subscripts in the model have been suppressed for reasons of space. Industry dummies have been included (not reported). SE is the standard error, estimated using 200 bootstrapping replications. * Denotes significance at the 5% level. Table 6 presents the results for the persistency test. None of the explanatory variables, which were significant in equation (6), have lasting effects on leverage for the IPO firms. Only the LSurp variable keeps its significance, indicating targeting behaviour for recently listed firms. Table 6 Estimation results for equation (7): IPO firms L t - L t-8 = α 0 + β 1 r t 4, t-8 β 2 YT t 4,-8 t β 3 LT t 4,-8 t β 4 LSurp t-8 β 5 FD t 4,-8 t β 6 FDd t 4,-8 t β 7 EBITD t 4, t-8 ε t Book leverage Market leverage I II III I II III Coef SE Coef SE Coef SE Coef SE Coef SE Coef SE FD 0.935* FDd YT LT R EBITD Lsurp * * * * * * 0.19 Explanatory notes. The it subscripts in the model have been suppressed for reasons of space. Industry dummies have been included (not reported). SE is the standard error, estimated using 200 bootstrapping replications. * Denotes significance at the 5% level.

25 22 6 CONCLUSION In this paper we examine the effects of market timing on capital structures of Dutch firms during We allow for capital structure targeting and pecking order financing. We adopt Kayhan and Titman s (2004) methodology. We perform the analysis on a full sample of 135 firms and a subsample of 45 IPO firms. IPO firms differ from the average firms in several respects. Financing needs of IPO firms are especially high in the first four years after the first stock quotation. IPO firms also have more growth opportunities and better operating performance. IPO firms have in common with non-ipo firms that internally generated cash flow forms the most important source of funds, followed by debt and external equity. For the full sample we find only weak evidence for market timing effects on capital structures of Dutch companies. Firms use relatively more equity after periods of a stock price increase, which finding is consistent with earlier evidence for the Netherlands by De Jong en Veld (2001) and De Haan and Hinloopen (2003). However, we do not confirm the evidence of KT for the US for our Dutch sample, that a high stock market valuation of the firm significantly and persistently affects the capital structure. What we do find, however, is a strong confirmation of earlier evidence for capital structure targeting and pecking order financing by Dutch firms. Overleveraged firms tend to bring back their leverage ratios towards target levels. When internal cash flows are small and there is a need for external finance, firms allow their leverage ratios to rise.