Small and Medium Size Enterprise Financing: a note on some of the empirical implications of a pecking order

Transcription

1 Small and Medium Size Enterprise Financing: a note on some of the empirical implications of a pecking order by ROBERT WATSON Department of Accounting & Finance, University of Glasgow, Glasgow G12 8LE & NICK WILSON Leeds University Business School, University of Leeds, Leeds LS2 9JT Acknowledgement Both authors contributed equally to this paper. Address for correspondence: R. Watson, Department of Accounting & Finance, University of Glasgow, Glasgow G12 8LE, UK R.Watson@ACCFIN.GLA.AC.UK 1

2 Small and Medium Size Enterprise Financing: a note on some of the empirical implications of a pecking order ABSTRACT Asymmetric information models predict a pecking order which reflects a combination of owner-manager preferences and external capital supply constraints whenever insiders know more about the true value of the firm s prospects than outsiders. The pecking order results in retained earnings being the most preferred source of finance, then debt and finally the issue of new shares to outsiders. Using a sample of 629 UK SMEs over the five-year period from 1990 to 1995 we find evidence consistent with a pecking order in which retained equity is preferred over debt. As expected, the evidence of a pecking order was particularly strong in respect of the closely-held firms in our sample. Key Words: Asymmetric information, pecking order, capital structure, SME finance 2

3 Small and Medium Size Enterprise Financing: a note on some of the empirical implications of a pecking order INTRODUCTION It is now widely recognised that information asymmetries can give rise to significant financial contracting and agency costs with the potential to influence the riskiness and size (and not, as in the Modigliani and Miller (1958 and 1963) framework, merely the distribution) of a company s future cash-flows. Essentially, the financial claims underpinning any particular capital structure are seen as capable of influencing corporate value via the impact upon managerial incentives and therefore operational and financial decision making (see Brennan, 1995 and Harris and Raviv, 1991, for reviews). Capital structure models, such as Myers (1984), focus on information asymmetries between insiders (i.e., managers, who are assumed to always act in current shareholder interests) and outsiders, primarily potential new debt and equity suppliers. Insider s superior knowledge regarding the true worth of the business and the additional contracting costs associated with credibly signalling this proprietary information to outsiders reduces both the demand for and supply of external debt and equity finance. The demand from owner-managers will be adversely affected because the higher perceived risks and monitoring costs will lead potential external suppliers to demand an unreasonably high premium on any funds advanced. By making all forms of external finance, particularly new share issues, appear excessively expensive, the Myers (1984) and Myers and Majluf (1984) models predict the existence of a pecking order whereby internal cash flows (i.e., retained earnings) are the preferred form of financing new investments. Of course, in the absence of cost-effective risk assessment, monitoring and intervention technology, external suppliers of debt finance may also be reluctant to finance some forms of risky investments if they are unable to raise interest rates sufficiently without encountering severe adverse selection problems (Stiglitz and Weiss, 1981). Thus, financial pecking orders may also arise due to external debt finance being subject to rationing and other supply constraints. This suggests that, in practice, the interaction of owner-manager and external suppliers preferences can be expected to produce firm-level financial structure decisions consistent with a pecking order whereby internally-generated sources of finance will be preferred over external sources. In this paper we empirically test some of the implications of the pecking order model using a sample of UK small and medium size enterprises (SMEs). One of the main empirical implications of the pecking order models is that any pecking order will be positively related to both the commonality of interests between current shareholders and 3

4 corporate decision makers (managers) and the degree of information asymmetry between insiders and outsiders. In order to investigate these issues empirically, our analyses of the pecking order hypotheses includes separate estimates and comparisons based on a partitioning of the sample into closely-held (i.e., owner-managed) and other companies. 1 The remainder of the paper is structured as follows. The next section examines the existing theoretical and empirical literature with respect to SME capital structure decisions. This is followed by a section which details our main empirical hypotheses and the empirical model used in the analyses. The fourth section describes the sample and data and the fifth section presents the main empirical findings of the paper. The final section summarises the paper, examines the implications of the findings and suggests some further areas of research into the topic. CAPITAL STRUCTURE THEORY AND EMPIRICAL RESEARCH The Modigliani and Miller (1958 and 1963) capital structure model shows that in a perfect and complete market setting, how the firm is financed will have no effect upon its value. What determines firm value is the size and riskiness of the cash-flows arising from its investment and operating activities. As these cashflows are deemed to be independent of the financing decision, with the exception of tax effects, capital structure will therefore be irrelevant for determining firm value. In the intervening period, there have been many attempts to show that the MM result is itself irrelevant to understanding the capital structure choices of actual firms operating in realistic market settings. The risk characteristics of different types of debt and the nature of the firms assets, the increasing financial distress/failure costs and agency-induced value losses at high debt levels have all been suggested as creating firm-specific optimal capital structure (see chapter 13 of Grinblatt and Titman, 1998, for a review). Thus, many of the early post MM models of optimal capital structure, because they modified the MM propositions to reflect increasing agency costs and bankruptcy/distress costs at higher debt levels, are often referred to as (static) trade-off models (Jensen and Meckling, 1976). These models suggest that the optimal capital structure for any particular firm will reflect the balance (at the margin) between the tax shield benefits of debt and the increasing agency and financial distress costs associated with high debt levels. The trade-off models have intuitive appeal because they confirm casual observations of actual corporate policies, namely, as the vast majority of firms have some debt, it is clear that that for most firms both no debt and alldebt cannot be value-maximising, while a moderate amount of debt is probably optimal. Empirically then, as each firm will have an optimal target debt ratio which maximises its 4

5 value, in the absence of changes in the firms business risk characteristics, empirically the best predictor of time t capital structure will be the t-1 capital structure, i.e., Debt/Total Assets t = α + β(debt/total Assets t-1 ) + u t (1) where α = 0 and β = 1. Generally however, despite their intuitive appeal, the empirical testing of some of the more interesting implications of the trade-off models has been disappointing since their explanatory power is typically low and the results in respect of specific propositions, such as the tax rate relationship, are not always supportive. Miller (1977), also declared himself to be unimpressed by these types of model. Indeed, he dismissed them as a recipe for a horse and rabbit stew, namely one horse and one rabbit because the estimated financial distress/failure costs were simply far too small to account for the significant trade-off with corporate tax gains implied by these models. Nevertheless, in the words of Brennan (1995), the theoretical focus had undoubtedly shifted from exploring the valuation of alternative ways of allocating given cash flow streams to exploring the implications for the cash flow stream itself of the allocation mechanism used to distribute it. Empirical studies have continued to be published which are based on the idea of some increasing cost component which is associated with high levels of debt to counter the corporate tax shield benefits of debt, e.g., the amount of debt a firm carries is related to business risk characteristics, the nature of the assets the firm invests in and the firm s tax rate (see Jordan et al (1998) for a recent example using UK SME data). More recently, empirical researchers have tended to focus on what causes changes in a firm s capital structure over time. In first-difference form, which overcomes many of the problems associated with unobservable firm-level fixed effects, the dependent variable is normally some measure of the change in capital structure over the relevant time period and the right-hand-side variables usually consist of factors which capture changes in firm-level risk over the same period. Dhawan (1997), is a good recent example of this empirical approach, which also reflects the contribution of finance theorists to the continued relaxation of some of the ceteris paribus assumptions of the Modigliani and Miller model. The new theories of capital structure have tended to emphasize the effect that the structure of financial claims has upon the incentives and behaviour of the decision makers that determine the income stream. Myers (1984) and Myers and Majluf (1984) developed a pecking order theory which was derived by assuming 5

6 a commonality of interests between current shareholders and managers (insiders) but asymmetric information and therefore heterogeneous expectations between insiders and potential new investors (outsiders). The basic idea here is that, due to information asymmetries, outsiders know less about the firms prospects than the owner-manager. The owner-managers will attempt to maximise their (i.e., the current shareholders) value, not the value of new investors. Thus, if the firm has good investment prospects, the owner-manager will not want to issue new shares because some of the benefits of the investment projects will have to be shared with the new shareholders. 2 The conclusion is that retaining internally generated earnings will be the preferred option, followed by raising debt finance for any additional funding needs, since this ensures that the increased value from the future investment projects accrue only to the existing shareholders. If, however, prospects are poor, the owner-manager will wish to issue new equity since this would benefit existing shareholders. Apart from the case of new and rapidly growing firms, new share issues are, therefore, seen by outside investors as a signal of poor prospects. As a result, share prices tend to decline when new share issues are announced, which makes new issues an extremely expensive method of obtaining additional equity finance. It is worth noting that both the Myers (1984) and Myers and Majluf (1984) models assume that the decision makers (i.e., the corporate board) always act in current shareholders interests. Such an assumption is uncontentious in the context of an owner-managed or closely held firm. In the case of widely-held firms, however, where managers have discretionary decision making powers and where governance and incentive mechanisms may not succeed wholly in aligning shareholder and manager interests, managers may exploit their information advantage to the detriment of the shareholders. Another issue that the pecking order model does not address is that capital structure choices are themselves typically constrained by information asymmetries and other market imperfections which influence the supply-side, i.e., availability and costs of different types of financial arrangement (Stiglitz and Weiss, 1981). The inability of potential lenders to reliably distinguish between good and bad credit risks and the optimistic bias typically associated with owner-manager assessments of their projects, exposes them to the danger of over-investment, that is, investing in firms which subsequently default or otherwise generate a negative net present value return to the lender (De Meza and Webb, 1990). Outside investors fears regarding possible over-investment can be expected to raise the cost of capital in sectors where it is particularly difficult to distinguish between good and bad credit risks. As Akerlof (1970) has argued, in the extreme, information sensitive markets may collapse altogether in the absence of credible commitments/signalling mechanisms. In the 6

7 case of corporate credit markets, whereby SME owner-managers are unable to convincingly signal their inside information to potential new investors, systematic under-investment (i.e., credit rationing whereby some firms with positive net present value projects are unable to obtain finance) may result. Thus, empirically a firm s capital structure is likely to reflect trade-offs between external financing constraints and owner-manager preferences. Once made, capital structure decisions are important determinants of firm value through their impact on the incentive structures, managerial behaviour and financial risk exposure of the enterprise (Keasey and Watson, 1995). Hence, unlike in the (symmetrically distributed information) Modigliani and Miller model, in which capital structure decisions are only of second-order distributional and/or tax-related importance, in an asymmetric information framework capital structure decisions may have far reaching economic consequences. To date, the evidence regarding the relative explanatory power of the static trade-off and pecking order models is somewhat weak, not least because the trade-off model will often fail to be empirically rejected even when false (see Shyam-Sunder and Myers, 1999). Helwege and Liang (1996), using a US sample of (post) IPO firms found limited evidence of a pecking order, though arguably the asymmetric information problem that gives rise to pecking order preferences may be less severe in the case of post IPO firms. However, a study by Griner and Gordon (1995), also based on US data, examined the alternative empirical predictions of the pecking order and managerial (self-interest) hypotheses and uncovered evidence in favour of the pecking order. Jordan et al (1998), using a capital structure levels model on data from, what appeared to be, a fairly representative sample UK SME s, also reported that they had found evidence consistent with a pecking order. HYPOTHESES AND EMPIRICAL MODELLING The static trade-off models that focus on the determinants of optimal capital structure usually make the simplifying assumption that there are only two types of finance: equity and debt. In empirical work, such as the recent Jordan et al (1998) study, this allows for the construction of a dependent variable such as the debt to equity ratio or the ratio of debt to total assets. With the pecking order model, finer distinctions between different sources of finance are required to test predictions such as that retained earnings are preferred to new share issues (even though both are classed as equity). Thus, the full range of pecking order predictions are difficult to test using a single financial leverage dependent variable. Moreover, as Shyam- Sunder and Myers (1999) have argued, most empirical specifications incorporate a partial adjustment mechanism which means that, even if firms behave in a pure pecking order manner, it is still extremely difficult to empirically reject the static trade-off model as a valid description of firms capital structure policies. 7

8 In order to mitigate these problems, in this paper we undertake two forms of analysis. First, we analyse the actual one-period changes in each major category of financing. Based on the static trade-off model predictions and each firm s actual growth rate over the period, these actual changes are further decomposed into their expected and unexpected components. For example, in the case of total debt, the actual change (as a proportion of total assets) can be decomposed into its expected and unexpected components as follows: Actual change = in total debt Expected change in total debt + Unexpected change in total debt (D t -D t-1 )/TA t-1 = (g x D t-1 /TA t-1 ) + ((D t -D t-1 )/TA t-1 - g x D t-1 /TA t-1 ) (2). where g = (TA t -TA t-1 )/TA t-1 If all of the average expected changes in financing derived from the static trade-off model are statistically indistinguishable from the actual changes (or, equivalently, the average unexpected change for each category of financing = 0), then this will be evidence which is consistent with the static trade-off model, but inconsistent with the pecking order model. However, if: (a) the average unexpected change in retained earning is statistically greater than zero, (b) the average unexpected change in debt is statistically less than the unexpected change in retained earnings and (c) the average unexpected change in new share issues to outsiders is statistically less than the average unexpected change in debt, then this will provide statistical evidence consistent with the pecking order model but inconsistent with the static trade-off model. The second set of tests involve regression analyses whereby the dependent variable consists of the actual growth rate of each firm in each period and the independent variables consist of the relative changes in each of the different category of financing used. By focusing on how the changes in each type of financing impacts upon total financing we are able to investigate whether the pattern of estimated coefficients are consistent with a pecking order. We begin from the accounting balance sheet identity, which equates the total assets of the enterprise (the uses of funds) with different types of financial claims used to fund these assets (the sources of funds), i.e., 8

9 Total Assets (TA t ) = Equity (E t ) + Debt (D t ) + Other Liabilities (OL t ) (3) Initially, we focus on changes in the relative proportions of debt and equity. If changes in other liabilities (OL) for each firm are assumed to randomly fluctuate around its average growth rate, then the following empirical model can be estimated: (TA i t TA i t-1)/ta i t-1 = α i + β 1 (E i t-e i t-1)/ta i t-1 + β 2 (D i t-d i t-1)/ta i t-1 + u i t (4) where α i = a vector of fixed effects representing firm i s average growth in (OL i t-ol i t- 1)/TA i t-1, and β 1 = β 2 if the target capital structure model is correct, i.e., the proportionate change in the requirement of finance is exactly matched by the same proportionate changes in equity and debt. If, however, the pecking order model, which assumes that retained earnings are preferred to debt, is valid, then (ignoring new equity issues for the moment) it is to be expected that the coefficient on the change in equity variable (β 1 ) will be greater than the coefficient on the change in debt variable (β 2 ), i.e., β 1 > β 2. Of course, the pecking order model also assumes that taking on more debt will normally be preferable to issuing new equity to outsiders since the latter will be expensive and will lead to a dilution of ownership. In order to include this into the model, we can define the change in equity to be as follows: E it = E it-1 + P it Div it + NE it, (5) where NE it = net new equity issued over the period, P it = profit available for distribution to shareholders, Div it = dividend payments and, therefore, P it Div it = retained profit for the period. Substituting (P it Div it ) + NE it into equation (4) gives the following: (TA it TA it-1 )/TA it-1 = α i + β 1 ((P it - Div it )/TA it-1 ) + β 2 (NE it /TA it-1 ) + β 3 (D it -D it-1 )/TA it-1 + u it (6) 9

10 where, if a pecking order exists with retained equity, then debt and finally new equity to outsiders as the preferred ranking of finance, β 1 > β 3 > β 2. In Table 2, we present panel data empirical results from estimating equation (6) that are consistent with the pecking order model. In order to further test some of the implications of the pecking order model, we partition the sample into high and low information asymmetry and managerial ownership groups, i.e., closely-held and other firms respectively. The pecking order pattern of the coefficients should be more apparent in the case of the closely-held (high information asymmetry and managerial ownership) group than in the case of the other firms group (i.e., the relatively low information asymmetry and managerial ownership firms). Also included in Table 2 are the separately estimated regressions for each group, plus an augmented version of equation (6) which includes an additional interaction term to measure the differential slope coefficient on the retained earnings variable in respect of the high information asymmetry firms. If the estimated coefficient on this interaction term is significantly positive then this will indicate that the closely-held firms have a significantly greater preference for using retained earnings over other sources of finance. The next stage of the analysis focuses on the composition of debt, since not all types of debt may be equally preferred. The total change in debt finance can be decomposed into categories that reflect its costs, availability and owner-manager preferences. Four categories of debt are considered: hire purchase liabilities, long term debt, short term debt (including overdrafts) and intra-group debt balances. Intra-group balances may be governed by external factors such as head office credit policies that are unrelated to information asymmetries. Hire purchase liabilities are essentially a form of fixed term finance tied to a specific asset purchase in which the legal title to the asset remains with the lender until the final instalment has been made. Thus during the term of the loan, the finance is secured on the basis of the second-hand value of the asset. This may be a significant source of finance for very small owner-managed firms that are either already highly geared and/or have insufficient cash or collateral to finance the purchase of general purpose items in which there is a wellorganised second-hand market, e.g., vehicles. Due to the relatively higher information requirements and greater monitoring costs associated with long term finance, it is reasonable to assume a likely preference amongst owner-managers for short-term rather than long term debt finance. Thus, equation (6) can be further augmented as follows: (TA it TA it-1 )/TA it-1 = α i + β 1 ((P it -Div it )/TA it-1 )) + β 2 (NE it /TA it-1 ) + β 3 (GRP it -GRP it-1 )/TA it-1 10

11 + β 4 (HP it -HP it-1 )/TA it-1 + β 5 (SD it -SD it-1 )/TA it-1 + β 6 (LD it -LD it-1 )/TA it-1 + u it (7) where (GRP it - GRP it-1 )/TA it-1, (HP it -HP it-1 )/TA it-1, (SD it -SD it-1 )/TA it-1 and (LD it -LD it-1 )/TA it-1 are respectively the change in intra-group balances, hire purchase, short term debt and long term debt. Empirical estimates of the model described by equation (7) are presented in Table 3. As with the previous table, additional estimates of the model are presented which include an interaction term to measure the differential slope coefficient on the closely-held firms retained equity variable, plus separately estimated models for the closely-held firms and the other firms samples. Our data also allows an exploration of the relationships between gross operating profits, directors fees, corporate taxes and dividends. Given that firms have a preference for retained earnings, how do they manage their capital structures, what discretion do they have in regard to dividends and/or directors fees? Non-listed, smaller, owner-managed firms clearly have more discretion over these matters (and have greater incentives to exercise their discretion) than listed and widely-held firms where outside investors may rely upon/expect a certain level of dividends each year and where directors remuneration is determined via remuneration committees and related to external managerial labour market factors (see Ezzamel and Watson, 1998). So, for example, if the profit before directors remuneration is defined as, π it = P it + DR it, this will represent the total funds available for reinvestment from internal sources and the discretionary outflows will be (Div it + DR it ) and, therefore, the actual amounts reinvested (i.e., the retained earnings at time t) will be: π it - (Div it + DR it. ). We can replace the retained earnings variable (P it -Div it )/TA it-1 used in the previous two tables with the above three variables (divided by TA it-1 ) that together constitute retained earnings. A significant improvement in the explanatory power of the empirical estimates vis-à-vis Table 2 is to be expected if these discretionary payments across the firms in the sample are of any systematic importance in explaining their relative need for new finance. The positive coefficient on the profit before directors fees variable (π it ) should also be larger in absolute 11

12 size than the (expected) negative coefficients on the dividend (Div it ) and directors fees variables (DR it. ). These results, because they confirm the main findings discussed later in the paper, are not reported in detail but are available from the authors upon request. In addition, the following forms of partitioning, reflecting other possible causes of any one-period changes in capital structure, have also been undertaken, but not reported since these results did not materially alter the conclusions to be drawn from the reported results: the firm being debt constrained, i.e., firms with prior period high debt/equity ratios can be expected to use relatively more retained equity and other sources of equity than low D/E ratio firms. firms with unexpectedly high earnings (and/or unexpectedly low Dividends) relative to last year, since these may prefer retained earnings over debt. firms experiencing positive growth, since the pecking order model coefficients may only apply to growing firms as firms experiencing negative growth may have different preferences. young firms, age less than 10 years old, since information asymmetries may be expected to decrease with time small firms, less than 50 employees, may be expected to have different pecking order preferences irrespective of ownership concentration. THE SAMPLE AND DATA Data for this study is from 655 firms that responded to a mail survey on trade credit practice conducted in late 1994, constituting a useable response rate of 18% 3. The survey collected detailed data on the firm itself; its ownership and organisational structure, products, markets, supplier relationships, and credit management issues. The firms mailed were a randomly selected stratified sample of manufacturing companies originally drawn from the UK FAME database and all had single digit SIC Codes 2,3 or 4. Details of the survey instrument can be obtained from the authors. This was then matched with the financial and behavioural data contained on a full credit reference report, provided by Dun & Bradstreet International. The sample reduces to 626 if we include only those firms for which we have more than one year of financial data. Section A of Table 1 shows the number of observations per year and the (almost-equal) split between closely-held businesses (i.e., those in which the directors collectively own a majority 12

13 stake) and other firms. It is to be expected that the pecking order models will have greater explanatory power in respect of the closely-held firms than for the other firms. Section B of Table 1 indicates that the mean total asset size of the closely-held firms, at slightly less than 4m, is less than 40% that of the other firms ( 10.3m). In terms of the relative proportions of different types of financing, Section B indicates that though both groups of firms have approximately the same proportions of equity finance (share capital plus retained profits), the closely-held firms have a smaller proportion of issued share capital and a correspondingly higher proportion of retained earnings vis-à-vis the other firms. In terms of the non-equity forms of finance, it is also clear that the closely-held firms make greater use of short-term debt (e.g., overdrafts) and other liabilities (e.g., trade credit, provisions for future tax, other accrued charges and directors loans, unpaid fees, etc.), whilst the other firms obtain proportionately more of their funds from other group company members. In Section C of Table 1, the profit and loss account variables are expressed in terms of their percentage of annual sales. For both the closely-held and other firms, the total appropriations of profit in the form of directors fees and dividends is about 5% of sales. In the case of the other firm sample, the average appropriations of profit as directors fees and dividends are approximately equal. However, as should be expected for the closely-held firms, directors fee payments are some 3.5 times the size of dividend payments. Section D of Table 1 provides a summary of the variables used in the empirical models in which all items are expressed in terms of the annual balance sheet opening total assets figures. As can be seen from the table, the average percentage change in total assets over the five year period from to , is a little over 3.6%, slightly higher at 4% in the case of the closely-held firms. Retained profits at 2.41% represents approximately 60% of the additional finance for the growth in total assets in respect of the closely-held firms whilst in the case of the other firms retained profits account for only one-third of the additional finance. THE EMPIRICAL RESULTS In section E of Table 1, the actual changes in retained earnings, total debt and new share issues are each decomposed into their expected and unexpected changes. It will be recalled from equation (1) that the static trade-off model implies that all of the unexpected changes will be statistically indistinguishable from zero. The pecking order model predicts that the average unexpected retained earnings will be positive, the average unexpected debt either negative or significantly less positive than the average level of unexpected retained earnings, and the average unexpected new share issues significantly less than the average unexpected debt. Moreover, these differences in unexpected changes will be most apparent in 13

14 the case of the closely-held firms, due to the predicted greater responsiveness to ownermanager preferences, greater information asymmetries between insiders and outsiders and the greater costs associated with external financing characteristic of these firms. As can be seen from section E of the table, the overall sample unexpected changes in retained earnings, though positive, is not statistically significantly so at conventional levels of confidence. However, as predicted by the pecking order model, for the sub-sample consisting of closely-held firms, the average positive unexpected retained earnings is statistically significant at 1% levels. The unexpected change in debt is statistically significantly negative as predicted by the pecking order model at 5% for the full sample and for the other firms, though not for the closely-held sub-sample. As regards new share issues, though the overall sample average is statistically significantly positive at the 5% level of confidence, it is clear from the sub-sample results that only in the case of the other firms is the average positive unexpected new share issues of any statistical importance. T-tests for the equality of means between the closely-held and other firms conducted on the three unexpected changes in finance variables show that with the exception of the debt component, the differences between the two sub-samples are statistically significant at 1% levels of confidence. On the basis of these results, it seems safe to conclude that the pecking order model provides a more empirically plausible explanation of the changes in financing than the static trade-off model, particularly, as expected, in relation to the closely-held firm subsample. We now turn to the regression estimates detailing the relative influence of each of the financing components in relation to overall financial changes. Table Two presents the simplest models in which we distinguish between retained earnings and changes in share capital and total debt. The first two columns, labelled Models 1 and 2, are estimated using all observations whilst the final two columns, closely-held and other firms, have been estimated using only the respective sub-samples. In all 4 models, however, the pattern of coefficients are the same in terms of relative size: the retained profits coefficient is largest, followed by the change in total debt coefficient and, finally, the change in share capital coefficient being in all cases the smallest. Model 2 includes an additional interaction term to measure the differential coefficient on the retained profits variable for the closely-held firms and, as expected, this is significantly positive. The separately estimated sub-sample models, unlike model 2 which allows only the coefficient on the retained profits variable to vary, allows all of the coefficients in the model to differ across the two groups of firms. As can be seen from the table, all of the estimated coefficients for the closely-held firms are higher than those estimated on the other firms sample. The separately estimated model is also able to explain a 14

15 significantly higher proportion of the variance in the dependent variable for the closely-held firms than is the case for the other firms sample model 4. The models estimated in Table 2 do not distinguish between different types of debt. In Table 3 we distinguish between four types of debt: hire purchase, debts owing to group members, long-term debt and short-term debt. As can be seen from the table, the inclusion into the models of these additional variables produces stronger results in terms of the adjusted R 2 s, and in terms of the relative size and statistical significance of all the (new and old) independent variable coefficients. The pattern of coefficients in all cases confirm the previous results but also suggest that in the case of the closely-held firms there is also a stronger preference for HP debt than for other types of debt, but that all types of debt are preferred over the issuing of new share capital. The greatly increased explanatory power of all of the empirical estimates vis-à-vis Table 2 suggests that there may be a pecking order within debt finance, particularly in regard to the other firms sub-sample since these firms appear to have a greater variety of choices of finance open to them than the closely-held firms. It is also evident that for many closely-held firms there is a greater preference for additional hire purchase finance than for other types of debt. As indicated earlier, it is possible to replace the retained profit variable with the 3 variables which together constitute the retained profits measure: profit before directors fees, directors fees and dividend payments. If, as expected, closely-held firms have both greater opportunities and incentives to retain profits in the business then we would expect to find that the model including the separate components of retained earnings would provide higher explanatory power than the model which included only the retained profits variable. Also, the (negative) coefficients on the directors fees and dividend variables should be smaller (in absolute terms) than the (positive) total profit coefficient. The estimate for the closely-held firms, not reported here, is indeed able to explain a significantly higher proportion of the variance in the dependent variable vis-à-vis the Table 2 estimates. Moreover, the (negative) coefficients on the directors fees and dividend variables are, as expected, smaller than that of the total profit variable. These findings were, however, also equally applicable in respect of the empirical results of the other firms. It seems, therefore, that irrespective of the level of insider ownership all SMEs have the necessary incentives and opportunities to vary their profit distribution policies in a manner which appears to increase the explanatory power of the pecking order model. 5 When we further partition both samples by age, we find some evidence that younger firms have a stronger preference for financing via retained earnings and short-term debt. This latter 15

16 finding, though entirely consistent with the pecking order explanation may, however, be more to do with supply-side constraints than owner-manager preferences. Finally, as can be seen from tables 1 and 2, as expected, the firm-level fixed effects are only of any statistical significance in respect of the other firms sample. Indeed, it is the explanatory power of the fixed effects that largely accounts for the greater explanatory power of the other firms model relative to that of the closely-held firms. We interpret this finding as being consistent with the pecking order model as it confirms that the changes in other liabilities (mainly trade credit and unpaid tax balances) for the closely-held firms are relatively less stable than these of the other firms. Though, once again, this finding could equally reflect supply-side constraints. CONCLUDING REMARKS The paper has made a preliminary attempt at empirically testing the pecking order model implication that when SMEs require additional finance the use of retained earnings will be preferred over debt and that debt will be preferred over new share issues to outsiders. We examined the relative size of the estimated coefficients on the various equity and debt variables in a regression model that attempted to explain how the annual changes in firm assets had been financed. The pattern of coefficients was found to be consistent with the pecking order model predictions and this was found to be particularly strong in relation to the closely-held firms where information asymmetries and a commonality of interests between shareholders and managers, and therefore the suggested pecking order preferences, would be most apparent. The results also suggest that there may be a pecking order within debt types since the explanatory power of all the estimated models increases significantly when the change in debt is decomposed into its various components. Clearly, much more work has to be undertaken prior to firm empirical conclusions can be drawn. Our analysis does not take into account differences in the use of the funds, i.e., the asset side of the balance sheet, changes in perceived risks and nor does it address the cost and availability supply-side factors which, rather than owner-manager (demand-side) preferences, may be driving the results obtained. 16

17 REFERENCES Akerlof, G. (1970), The Market for Lemons: quality, uncertainty and the market mechanism, Quarterly Journal of Economics, Vol. 89, pp Brennan, M.J. (1995), Corporate Finance over the past 25 years, Financial Management, Vol.24, pp De Meza, D. and D. Webb (1990), Risk, Adverse Selection and Capital Market Failure, Economic Journal, Vol.100, pp Dhawan, R. (1997), Asymmetric Information and Debt Financing: the empirical importance of size and balance sheet factors, International Journal of the Economics of Business, Vol.4, pp Ezzamel, M. and R. Watson (1998), Market Comparison Earnings and the Bidding-Up of Executive Cash Compensation: evidence from the UK, Academy of Management Journal, Vol. 41, pp Grinblatt, M. and S. Titman (1998), Financial Markets and Corporate Strategy, (Macgraw-Hill publishers). Griner, E.H. and L.A. Gordon (1995), Internal Cash Flow, Insider Ownership, and Capital Expenditures: a test of the pecking order and managerial hypotheses, Journal of Business Finance and Accounting, Vol. 22 (March), pp Harris, M. and A. Raviv (1991), The Theory of Capital Structure, Journal of Finance, Vol. 46, pp Helwege, J. and N. Liang (1996), Is there a pecking order?: Evidence from a panel of IPO firms, Journal of Financial Economics, Vol.40, pp Jensen, M. and W. Meckling (1976), Theory of the Firm: Managerial Behaviour, Agency Costs and Ownership Structure, Journal of Financial Economics, Vol. 3, pp Jordan, J., J. Lowe, and P. Taylor (1998), Strategy and Financial Policy in UK Small Firms, Journal of Business Finance and Accounting, Vol.25, pp Keasey, K. and R. Watson (1995), The Bank Financing of Small Unlisted Firms in the UK: an analysis of recent conflicts, Journal of Small Business Finance, Vol.4, pp Miller, M. (1977) Debt and Taxes, Journal of Finance, Vol. 32, pp Modigliani, F. and M. Miller (1958), The Cost of Capital, corporation Finance, and the Theory of Investment, American Economic Review, Vol. 49, pp Modigliani, F. and M. Miller (1963), Taxes and the Cost of Capital: A correction, American Economic Review, Vol. 53, pp Myers, S.C. (1984), The Capital Structure Puzzle, Journal of Finance, Vol. 39, pp Myers, S. and N. Majluf, (1984), Corporate Financing and Investment Decisions when 17

18 firms have information that investors do not have, Journal of Financial Economics, Vol.13 pp Stiglitz, J.E. and A.Weiss (1981), Credit Rationing in Markets with Imperfect Information, Amercian Economic Review, Vol. 71, pp Shyam-Sunder, L. and S.C. Myers(1999), Testing Static Tradeoff against Pecking Order Models of Capital Structure, Journal of Financial Economics, Vol. 51, pp

19 Table One Descriptive Statistics Section A: Observations by Year and Firm Type Year All Firms Closely-Held Other Firms Total Section B:Capital Structure Variables All Firms Closely-Held Other Firms Mean (SD) Mean (SD) Mean (SD) Total Assets ( m) 7.14 (10.86) 3.98 (4.70) (14.65) Share Capital (%) (24.26) (17.35) (29.64) Retained profits (%) (35.80) (28.58) (41.23) Long-term Debt (%) 4.12 (10.10) 4.74 (8.22) 3.50 (11.65) Hire Purchase (%) 2.24 (5.30) 2.95 (6.03) 1.53 (4.33) Short-term Debt (%) 9.04 (12.72) (12.38) 7.55 (12.88) Group Debt (%) (20.05) 5.28 (13.42) (29.40) Other Liabilities (%) (20.27) (20.48) (19.48) All section B variables expressed in terms of their % of the end of year balance sheet total asset values. Section C: Profit & Loss Account Variables All Firms Closely-Held Other Firms Mean (SD) Mean (SD) Mean (SD) Sales ( m) (12.29) 6.57 (7.06) (14.80) Profit before DF(%) 6.95 (21.06) 7.21 (14.32) 6.70 (25.86) Directors Fees (%) 3.09 (11.02) 3.75 (10.55) 2.44 (11.43) Profit after DF (%) 3.77 (19.11) 3.27 (9.57) 4.24 (11.43) Tax (%) 1.20 (2.82) 1.07 (2.17) 1.33 (3.33) Dividends (%) 1.83 (8.04) 1.04 (3.52) 2.62 (10.77) Retained profits (%) 0.61 (19.31) 1.02 (8.28) 0.22 (25.70) All section C variables, with the exception of sales, are expressed in terms of their % of annual turnover 19

20 Section D: Variables in the Empirical Models All Firms Closely-Held Other Firms Mean (SD) Mean (SD) Mean (SD) % in Total Assets 3.62 (16.50) 4.00 (16.12) 3.24 (16.88) % share capital 0.92 (7.67) 0.30 (3.04) 1.55 (10.39) % Long-term Debt (7.19) 0.17 (6.11) (8.13) % Short-term Debt (10.21) (9.39) (10.98) % HP Debt 0.11 (3.53) 0.19 (3.49) 0.02 (3.57) % Group Debt 0.41 (12.26) 0.15 (8.00) 0.67 (15.41) Profit before DF (16.37) (17.13) 9.98 (15.59) Directors Fees 4.96 (9.40) 6.55 (12.11) 3.37 (4.97) Tax 2.11 (3.36) 1.99 (3.10) 2.22 (3.60) Dividends 2.46 (6.43) 1.71 (4.71) 3.22 (7.72) Retained Profits 1.73 (10.63) 2.41 (8.68) 1.07 (12.20) All section D variables expressed in terms of their % of start-of-period total assets. Section E: Unexpected Changes in Financing All Firms Closely-Held Other Firms Mean (SD) Mean (SD) Mean (SD) % unexpected in Retained Profits t-value for equality of means = % unexpected in share capital t-value for equality of means = % unexpected in Total Debt t-value for equality of means = 0.24 (10.85) 2.38 *** 0.56 (8.57) ** 2.71 *** (11.67) ** (8.40) *** (12.77) 0.11 (3.91) 1.02 (11.47) *** (10.68) (12.58) *** All section E variables expressed in terms of their % of start-of-period total assets. *** = significant at 1%; ** significant at 5%; * significant at 10%. 20

21 Table Two Empirical Estimates:Basic Model Dependent Variable = (TA t -TA t-1 )/TA t-1 All Firms All Firms Closely-Held Other Firms Model 1 Model 2 Variable Name Coefficient Coefficient Coefficient Coefficient Retained Profit (15.65) *** (11.27) *** (12.62) *** (10.32) *** Share Capital (4.19) *** (4.08) *** (2.96) *** (3.02) *** Total Debt (17.14) *** (17.17) *** (16.07) *** (9.23) *** Retained Profit x closely-held dummy (3.37) *** Constant (8.47) *** (7.81) *** (4.05) *** (6.45) *** Year dummies Yes *** Yes *** Yes *** Yes *** Significant? Fixed Effects Yes * Yes * No Yes *** Significant? Adjusted R 2 (%) N = *** = significant at 1%; ** significant at 5%; * significant at 10%. 21

22 Table Three Empirical Estimates:Types of Debt Model Dependent Variable = (TA t -TA t-1 )/TA t-1 All Firms All Firms Closely-Held Other Firms Model 1 Model 2 Variable Name Coefficient Coefficient Coefficient Coefficient Retained Profit (22.62) *** (17.50) *** (15.42) *** (16.62) *** Share Capital (11.00) *** (11.00) *** (4.65) *** (9.53) *** Short-term Debt (21.67) *** (21.72) *** (15.38) *** (14.98) *** Long-term Debt (15.60) *** (15.50) *** (9.93) *** (11.48) *** HP Debt (8.99) *** (8.96) *** (7.21) *** (5.15) *** Group Debt (23.08) *** (22.97) *** (12.22) *** (18.56) *** Retained Profit x closely-held dummy (2.86) *** Constant (6.17) *** (5.64) *** (2.71) *** (5.00) *** Year dummies Significant? Fixed Effects Significant? Yes *** Yes *** Yes *** Yes *** No No No Yes *** Adjusted R 2 (%) N = *** = significant at 1%; ** significant at 5%; * significant at 10%. 22

23 Endnotes 1 Closely-held is a binary variable, coded 1 when the current directors have a majority equity stake and 0 otherwise. It would be desirable to measure ownership concentration as a continuous variable i.e. percentage of equity owned by current directors in order to test hypothesis concerning the precise relationship between funding sources and principal/agent compatibility. Moreover, there are other potential proxies, the age of the firm, for instance, that might be used to test whether informational asymmetries decrease over time. Also, as size and ownership concentration are correlated, a number of tests for age/time and size impacts on the pecking order hypothesis were also undertaken. 2 Also, there are no issue costs associated with retained earnings. 3 The response rate was as expected given the detailed nature of the questionnaire survey. However, a response rate of 18% requires that tests for non-response bias should be undertaken. The comprehensive data-base from which the sample frame was drawn allowed thorough tests to be completed. Two tests were carried out. The first involved tests for differences in the sample characteristics between early respondents (returned questionniares within one month) and later respondents (returned > 1 month). No significant differences were found. The second test involved t-tests on the full sample of respondents and non-respondents in four areas: size (e.g. total assets, employees, turnover); profitability (ROCE, ROTA); working capital structure (debtor days, creditor days) and age. There were no differences in the means of these variables at 5% significance. Pearson s chi square was used to test for geographical and industry bias and differences in legal status. 4 Various experiments were conducted with variables representing both age and size of firm. Age was measured as a continuous variable, a quadratic function, a linear spline and a binary variable for firms less than 10 years old and used as both additional control variable and an alternative partitioning mechanism. The sample was partitioned by age (less than 10 years old) and size (less than 50 employees). A pattern of results consistent with a pecking order was always maintained. In respect of age partitions it was evident that younger firms pay lower dividends and are more reliant on short-term debt finance. The latter result, however, may be more consistent with supply-side financial constraints than firm level preferences. 5 Also, the lack of any significant ownership effect either within or between the two groups of firms appears to be consistent with Griner and Gordon s (1995) inability to uncover any significant evidence in favour of the managerial (self-interest) hypothesis. 23