Determinants of Small Firm Debt Ratios:

Nico van der Wifst

An Analysis of Retail Panel Data Roy Thurik

ABSTRACT. In this paper, the relevance of some debt ratio by, among others, Robichek and Myers (1965),
determinants from the recent theory of finance is empirically Jensen and Meckling (1976), Ross (1977), Leland
investigated in a small business sector. The data used in this
and Pyle (1977) and Myers (1977). The general
study consist of average financial data of 27 shoptypes in 20
different years, covering a period of 24 years. The panel result of these extensions is that the combination
character of the data facilitates the use of analytical tech- of leverage related costs (such as bankruptcy and
niques aimed at reducing or avoiding the biasing effect of agency costs) and a tax advantage of debt pro-
omitted variables on the outcomes. The main conclusion is, duces an optimal capital structure at less than a
that the theoretical determinants appear indeed to be relevant
100% debt financing, as the tax advantage is
for the small business sector investigated here, but the
influences encountered in the analyses are far less straight- traded off against the likelihood of incurring the
forward than the hypothesized effects in the theory. Influ- costs.
ences on total debt are frequently found to be the net effects This theoretical result is now widely accepted in
of opposite influences on long and short term debt and some the profession. The main criticism is of empirical
variables show large time and industry specific effects.
nature, i.e. whether or not the various leverage
Further, distinct patterns in the time specific effects were
found. related costs and benefits are economically signifi-
cant enough to have an appreciable impact on
optimal leverage. This criticism gave rise to a
I. Introduction number of empirical studies in which either the
magnitude of leverage related costs and benefits is
The capital structure of the firm has been a directly estimated (e.g. Warner, 1977) or in which
contentious issue in the theory of business finance observed capital structures are related to operat-
since the publication of Modigliani and Miller's ing characteristics which are assumed to reflect
(1958, 1963) seminal papers. Their initial conclu- these costs and benefits (e.g. Scott, 1972, Rem-
sions of capital structure irrelevance and, subse- mers et al. 1974, Scott and Martin, 1975, Ferri
quently, of optimal capital structure at a 100% and Jones, 1979, Flath and Knoeber, 1980,
debt financing are clearly incompatible with Castanias, 1983, Bradley et al. 1984). Most of
observed capital structures, so their findings these studies use data of firms that would be
initiated a considerable research effort to identify classified as large by any definition of small busi-
costs of debt financing that would offset the ness.
corporate tax advantage. Since then, extensions of The empirical implications of the theory of
the Modigliani-Miller theory have been provided finance for small business are seldom discussed
(McConnell and Pettit, 1984, and Van der Wijst,
1989, are exceptions) and even less frequently em-
pirically tested in a quantitative manner (Keasey
Final version accepted on November 19, 1991 and McGuinness, 1990, is a noteworthy example).
This means that there is a large discrepancy
Research Institute for Small and Medium-Sized Business in between the importance of small business in most
the Netherlands
Department of Fundamental Research
western economies and the attention devoted to it
P. 0. Box 7001 by the finance profession. Financial research in the
NL-2701 A.A Zoetermeer field of small business is usually confined to
The Netherlands descriptive analyses using either case studies (e.g.

SmallBusiness Economics 5: 55--65, 1993.

9 1993 KluwerAcademic Publishers. Printed in the Netherlands.
56 Nico van der Wifst and Roy Thurik

Rapoport, 1990) or a behavioural approach 2. Debt ratio determinants

(Norton, 1990) or qualitative analyses (Walker,
1989). From the combined theoretical and empirical
In this paper, the relevance of some debt ratio studies of business finance, that have been pub-
determinants from the recent theory of finance is lished over the past three decades, some broad
empirically investigated in a small business sector. categories of debt ratio determinants can be seen
These determinants usually refer to factors that to emerge. These determinants make no explicit
can not or not adequately be measured by outside distinction between small and large business ~ and
researchers. Consequently, proxy variables have they refer, of course, to the costs and benefits
to be used in any empirical investigation and this associated with financial contracting. In the ab-
study is no exception. Although the theoretical sence of costs and benefits, i.e. in the perfect
assumptions of completeness and perfectness may capital markets for which financial economists
be less valid for the small firm's capital market have such a pronounced bias, the market value of
than they are for the large firm's, no attempts are the firm is independent of its capital structure, as
made to capture the effects, if any, of the alleged the famous Modigliani-Miller irrelevance theorem
imperfections of the small firm's capital market in demonstrates. The market imperfections that have
testing our hypotheses. Also, the determinants are been brought forward as determinants of the firm's
tested as such, i.e. we are not considering the joint optimal capital structure can be collocated in the
determinants of the investment and the financing categories taxes, bankruptcy costs, agency costs
decision. Hence, this paper is aimed at a rather and signalling effect}
straightforward test of the relevance of theoretical The influence of taxes on financial structure
determinants. Such tests are scarce in small busi- rests on differences in fiscal treatment between
ness, perhaps because empirical analyses of finan- debt and equity: since interest payments are tax
cial structure in small business are frequently deductible and dividends are not, the fiscal regime
hampered by a lack of available data. For the is generally considered to favour debt. This simple
majority of small firms, publication of the annual tax effect can be complicated by several factors.
report is not mandatory and, hence, simply not First of all, non-debt tax shields, such as deprecia-
done. This means that laborious fieldwork or tion charges and investment tax credits, can
questionnaires have to be used to collect small reduce the expected tax benefit from interest
firm financial statements and this, in turn, means payments. Secondly, personal taxes can offset the
that data collection is usually restricted to some favourable treatment of debt at the corporate
key figures on financial structure without much level, as Miller (1977) pointed out. Thirdly,
room for explanatory variables. As a consequence, aspects of the tax regime can create differences in
in empirical analyses important determinants of investors' preferences. For instance, in many
financial structure often have to be approximated countries the tax rates applying to capital gains are
in a rather arbitrary manner or, even worse, simply lower than those applying to dividend and interest
have to be omitted. This may, of course, bias the income. This creates a preference for capital gains
results of the analyses. The analytical technique over dividends from stock ownership, but other
used here reduces or avoids the omitted variables aspects can cancel out this effect. Some investors
bias. The panel character of the data, which are not taxed, e.g. pension funds or very small
consist of the average financial statements of 27 investors, who benefit from the fact that small
industries in retailing (shoptypes) in 20 years, amounts of dividend are usually exempt from
permits the use of this technique. taxation. Other investors, such as retired people,
The paper is organized as follows. In section 2, may depend on dividend income for their liveli-
the theoretical determinants of financial structure hood# Given the wide variety among investors' tax
are briefly summarized. The data used in this situations, the combined effect of all these aspects
study are described in section 3 and section 4 of the fiscal environment can become quite com-
contains the empirical analysis, i.e. the empirical plex. Therefore, it is usually assumed that the net
proxy variables, the model specifications and the effect of all fiscal rules creates, over a certain
estimation results. Section 5 concludes the paper. range, a positive ta'. incentive for corporate debt
 Small Firm Debt Ratios 57

financing, albeit that this effect is smaller than in monitoring devices in order to protect themselves.
the original Modigliani-Miller theory and that it Monitoring is generally facilitated by using cove-
may not apply equally to all investors. nants that restrict the entrepreneur's activity,4
Without some costs to offset the tax advantage, thereby increasing financing costs. The costs of
this incentive will lead to a maximal use of debt writing and enforcing these covenants and of
financing, which is not observed in practice. The employing other monitoring devices may well be
first counterbalancing costs to be suggested by, substantial. These costs are likely to be compara-
among others, Robichek and Myers (1965) and tively high for small and young firms, because of
Baxter (1967), were the dead-weight losses asso- the lack of formal financial control and the firms'
ciated with bankruptcy and reorganization. Bank- flexibility to change its assets. This makes it
ruptcy costs refer to the added costs (such as legal difficult to gain insight into their financial per-
fees, reduced sales, increased production costs, formance. Monitoring and bonding costs are likely
etc.) that arise because the firm cannot meet its to increase with the degree of protection they
obligations to creditors without changing its offer. Therefore, security holders will not seek
operating or external financing activities (Haley perfect protection, but these agency costs, as
and Schall, 1979). The theoretical influence of Jensen and Meckling call them, will be traded off
bankruptcy costs is not contentious but from an against the likelihood of a loss resulting from
empirical point of view the importance is some- management actions. If agency costs also increase
what controversial. The evidence collected in with the proportion of financing supplied by any
different studies varies from direct bankruptcy type of security holder, conformably to Jensen and
costs averaging a trivial 1% of the market value of Meckling's suggestion, it is possible to argue that a
the firm prior to bankruptcy to an approximate capital structure exists in which total agency costs
20% of the estate. Most studies agree that some are minimal. In this way, an optimal capital struc-
direct costs decline with firm size, stressing the ture may exist even in the absence of taxes and
importance for smaller firms. Substantial indirect bankruptcy costs.
costs can arise from the imperfection of secondary Additionally, it can be noted that some, if not
markets for physical assets. To illustrate, Kim all, of the firm's financial relationships have
mentions that in a distress sale, assets may bring as information value, as is suggested by Ross' (1977)
little as 30% of their wholesale price (Kim, 1978, signalling hypothesis. The financial obligations a
p. 47). All these effects contribute to a reduction firm enters into may serve as signals to the public
in firm value if the possibility of "financial distress" about the future performance of the firm. Dividend
exists, even if bankruptcy in the legal sense is policy is another obvious candidate for a signalling
virtually impossible. Hence, increasing the ex- device. Still other signals may spring from the
pected value of these costs by choosing high debt owner/manager's personal financial involvement
levels which increase the probability of bankruptcy in the firm, as is suggested by, among others,
will, at a certain point, cancel out the tax advan- Leland and Pyle (1977). Signalling activities arise
tage. in situations in which information is unevenly
Perhaps the most powerful concept of leverage distributed over investors and management (i.e.
related costs was introduced by Jensen and when "information asymmetry" exists) and when
Meckling (1976), who describe the firm as a nexus moral hazard hampers the direct transfer of
of relationships that can be characterized as information between market participants. Gener-
principal-agent relationships. The firm's security ally, the firm's management are most knowledge-
holders (debtholders as well as stockholders) can able of its prospects, but they cannot be expected
be seen as principals and the firm's management as to be entirely straightforward about these pro-
the agent, managing the principals' assets. This spects for two reasons. First, there may be
principal-agent relation is costly, because there is a substantial rewards for exaggerating the firm's
possibility that the agent will not always conduct positive qualities and obscuring the negative ones,
business in a way that is consistent with the best when ascertaining the true qualities by outside
interest of the principals. Hence, the latter may parties is costly or impossible. Second, managers
insist on various types of protective covenants and can be reluctant to disclose the full merit of the
58 Nico van der Wijst and Roy Thurik

firm since this might allow competitors to appro- Both series of surveys are claimed to be represen-
priate key features of the firm's projects. Informa- tative of the German retail trade for all variables
tion asymmetry is particularly pronounced for, except size variables (sales size, total assets). From
again small and young firms in a dynamic industry. the publications, sales size (and sample size) can
These firms usually must incur high costs to be seen to differ between both surveys, but the
generate information for use by outside investors, only ratio they have in common, the inventory
because much of their expertise and knowledge is turnover, correlates almost perfectly between the
intangible and hard to quantify. So, like bank- surveys (simple correlation coefficient always
ruptcy and monitoring costs, information costs are > 0.94, usually > 0.97). Hence, it seems safe to
also likely to be a more important financial struc- use ratios from the income statement survey in the
ture determinant for small businesses than for analysis of balance sheet data. The combined
large ones. surveys contain concise balance sheets, income
statements and information on labour volume,
shop surface, prices, etc. This information is
3. Data
available for 27 shoptypes over the entire period
The data used in this study refer to the retail trade and for 15 other shoptypes over a part of the
in the former Western Germany and are publicly period. Only the former 27 shoptypes are used in
available: the information has been published by this study. A more detailed description of the data
the collectors for purposes such as interfirm can be found in the publications referred to
comparisons. The data are based on information earlier. A list of the 27 shoptypes involved and
of individual firms, but the unit of observation some key financial data per shoptype are ap-
used here is the published "industry average", i.e. pended to this paper. In Figure 1 the development
the averaged data for narrowly defined shoptypes of some leverage ratios over time (averaged across
such as supermarkets and shoe shops. The data shoptypes) is depicted. Figure 1 shows, that the
have a panel character: information on shoptypes average financial structure in retailing over the
is available for a number of years.5 period is largely stable. The share of equity in total
The data were collected by the "Institut fiir assets rises slowly in the fifties, is more or less
Handelsforschung an der Universitiit zu Krln" in stable in the sixties and slowly diminishes in the
two separate series of surveys. The first series seventies. Total debt shows, of course, the oppo-
consists of regular inquiries into the financial site picture but the ratio of long term debt to total
situation of the retail trade, made every alternate assets slowly increases over almost the entire
year over the period 1955 through 1969 and the period. Taken as a whole, however, the changes
years 1973 and 1977. The balance sheet data for occurring in the period of almost a quarter of a
this study have been taken from two ten year century are remarkably small.
summaries and a periodical, both brought out by
the research institute mentioned above. 6 Because o . ei - . .
. . . . I . . . . . . .

beginning and ending balance sheets were col-

lected, an uninterrupted row of ending balance
o.m
sheets over the period 1954 through 1969 can be
constructed, as well as ending balance sheets for
the years 1972, 1973, 1976 and 1977. o.4
The second series of surveys consists of yearly
inquiries into the sales, costs, margins and profits
of the retail trade over the period 1949 through o~

1986. In these surveys, information on income

statements is collected and published for largely i
the same shoptypes as in the balance sheet survey.
196e lSllll ~,lliSl[ ~.i'l'l ~Lll'llg s
The income statement data for this study have yZI'IE

been taken from three ten year summaries brought

out by the Cologne institute mentioned earlier. 7 Fig. 1. Leverage ratios over time
 Small Firm Debt Ratios 59

4. Empirical analyses likely to reflect the liquidity of, and thus, the
security offered by the inventories. Both variables
4.1 Variables are expected to have a positive effect on leverage,
but they can also influence the maturity structure
From the data described in the previous section, of debt, i.e. a large fixed asset component can be
the following variables are adopted for the empiri- associated with a comparatively heavy use of long
cal analyses. The obvious choice of variable to be term debt and a high inventory turnover with
explained is the debt ratio, since all theoretical relatively much short term debt.
determinants are formulated in terms of more or The fifth variable to be included is return on
less leverage. Estimating separate relations for investment, which reflects the possibilities to
long and short term debt ratios (long and short retain earnings. The importance of these internally
term debt to total assets) allows for influences on generated funds is stressed by Myers (1984) and
the maturity structure of debt as well as on Myers and Majluf (1984), following the observa-
leverage. Total assets is included as size variable in tion that firms seem to prefer raising capital by
the explanation of the debt ratios in order to test retaining earnings in the first place, by borrowing
for scale effects in the relation between debt and in the second place and by issuing new equity in
total assets. the last place. An explanation of this "pecking
The tax effect on debt has to be incorporated in order" in financing alternatives, as Myers calls it,
an indirect manner, because no direct measures can be based on the asymmetrical distribution of
are available. Following Bradley et al. (1984), information between potential outside investors
depreciation charges are used to indicate non-debt and the firm's management. When the investors
tax shields? The ratio of depreciation charges to are less knowledgeable of the firm's prospects than
total costs is included in the analyses to indicate its management, a situation may arise in which
the tax advantage. Since depreciation charges can firms face the dilemma of either passing by
reduce the expected tax benefits from interest projects with a positive net present value or issuing
payments, this variable is expected to be nega- stock at a price they think is too low. This situation
tively related to leverage. can be avoided if a firm can retain enough
Two variables pertaining to the firm's assets are internally generated funds to cover its positive
included to capture the effects of bankruptcy and NPV opportunities or if it can maintain financial
agency costs. Agency costs are, to a large extent, slack in the form of "reserve borrowing power". In
made to avoid bankruptcy costs, so a reduction in this view, observed debt ratios will reflect the
the latter also diminishes the need to incur the cumulative requirement for external financing
former. Hence, both determinants are usually over an extended period, and will be negatively
approximated with the same variables. Bankruptcy related to profitability) I
costs can be thought of as the difference between
the firm's operating value and its liquidation value,
so a high liquidation value makes debt financing 4.2 Analyses
more attractive. Since fixed assets are generally The hypotheses formulated above as a priori
considered to offer more security (i.e. a higher expectations are tested by including the five
liquidation value) than current assets, asset struc- variables in a Least-Squares Dummy Variable
ture (i.e. the ratio of fixed to total assets) is used to (LSDV) regression analysis of the pooled cross
indicate the liquidation v a l u e . 9 The second vari- section and time series data. Autocorrelation and
able to be included, inventory turnover, is not heteroskedasticity are accounted for in the estima-
frequently encountered in the literature, but in tion procedure. Beforehand, the variables ex-
retailing inventories are a financially (and com- pressed as money amounts have been deflated
mercially) important part of the assets. The with the industry specific price indices included in
characteristics of the inventories determine wheth- the data, 12 and separate regression equations are
er or not they are accepted by banks as collateral estimated for long and short term debt. Since the
for loans 1~ and this can have a substantial influ- analysis concentrates on debt ratios as such (and
ence on financial structure. Inventory turnover is not their changes over time), neither leads and lags
60 Nico van der Wijst and Roy Thurik

nor the simultaneities, if any, in the financing and part of total retailing, so they cannot be considered
investment decisions are considered. A few re- a small sample from a much larger population of
marks may serve to illustrate this research design. shoptypes. In this situation, the fixed-effects model
The LSDV technique is chosen because the five seems more appropriate than its random-effects
variables could very well exhibit important indus- counterpart.~3 Hence, time and industry effects are
try and/or time commonalities and, more seri- introduced in this study by estimating the coeffi-
ously, they can be correlated with industry and cients of a fixed-effects model using the least-
time specific omitted variables. In the latter case, squares dummy variable (LSDV) approach. For
the results of an OLS analysis would partly reflect all shoptypes and all but the first period a separate
the influence of the omitted variables rather than dummy variable is included in the regression
the pure effects of the included variables. By equation, replacing the intercept. Note that these
introducing shoptype and/or time specific var- dummy variables will not only capture the time
iables into the regression equation, it is possible to and industry specific effects of omitted variables,
reduce or avoid the omitted variable bias (see but also the time and industry commonalities in
Hsiao, 1986). The panel character of the data the included variables. 14
permits these variables to be included. Time and Autocorrelation and heteroskedasticity, which
industry specific variables can be included in are likely to occur in panel data, are accounted for
basically two ways. First, since all variables consist by using an estimated generalized least squares
of observations per shoptype repeated over time, estimator for the coefficients of the LSDV analy-
they can be averaged across shoptypes per period sis. 15 This estimator is obtained by re-running the
and across periods per shoptype. If these averages LSDV regression after a transformation of the
would be added to the regression equation, they data. The transformation is done separately for
would provide insight into the cross-sectional long and short term debt. The elements of the
(within group) and time series (between group) transformation are obtained as follows. First order
characteristics of the variables themselves. This autocorrelation coefficients (p) are estimated ~6for
procedure would only capture effects of omitted the residuals (ui3 of the original LSDV regressions
variables in so far as they are related to the (i and t are subscripts for shoptype and time
fluctuations in the shoptype and period averages resp.). Residuals corrected for autocorrelation
of the included variables. (u*) are calculated as uTt --- ui, - p u i , - v 17 The
The second way to include time and industry squares of the corrected residuals are regressed on
specific effects is based on the more likely assump- the shoptype dummies: (u*) 2 ---- a 2 i D u m i + ~,. T h e
tion that the effects of shoptype specific omitted coefficient of this regression, o2~ provides an
variables stay constant through time for a given estimate of the shoptype specific variance. Using
shoptype but vary across shoptypes. Similarly, the these results, the data (Z) are transformed by:
effects time specific omitted variables are likely to Z~t = (Zit - P Z i t - 1) '-- Dri fts The thus transformed
be the same for all shoptypes, but will vary in time data are then used to re-estimate the coefficients
without necessarily showing any pattern. A simple of the LSDV model.
way to take account of these effects is to use The results of the LSDV analyses are reported
variable-intercept models. Given the above as- on in three different ways. First of all, the esti-
sumptions regarding the effects of omitted var- mated GLS results regarding the non-dummy
iables, they can be absorbed into the intercept variables are presented in Table I.
term of a regression model as a means to explicitly From the coefficients for total assets in Table I
allow for the individual and time heterogeneity it can be seen, that no scale effects in the long and
contained in the pooled cross section and time short term debt ratio are found: both coefficients
series data (see Hsiao, 1986 or Dielman, 1989). are not significantly different from zero. The
Within the variable-intercept models, the shop- hypothesis regarding the tax effect is not sup-
type and period specific effects can be treated as ported: the coefficients of depreciation charges are
fixed constants (in fixed-effects models) or as not significantly different from zero and the coeffi-
random variables (in random-effects models). The cient for short term debt is even positive, contrary
27 shoptypes involved in this study cover a large to the expected influence of non-debt tax shields.
 Small Firm Debt Ratios 61

TABLE I
Estimated LSDV regression coefficients (pooled data, corrected ror autocorrelation and heteroskedasticity, standard errors
between parentheses)

Total Asset Invent. Depr. ROI R2

assets struct, turnov, charg (adj.)

Shortterm debt 0.122" -0.410 0.005 0.542* -0.016 0.959

(0.098) (0.054) (0.002) (0.276) (0.004)
Long term debt 0.032* 0.394 -0.004 -0.174* -0.008 0.953
(0.076) (0.041) (0.001) (0.199) (0.003)

Note: * meansnot significantlydifferentfrom zero at a 5% levelof significance.

As regards the effects of bankruptcy and agency from the dummy variables. These differences can
costs, Table I shows that a high fixed asset com- be split up (decomposed) into a time specific and
ponent is associated with more long term debt and an industry specific component by regressing the
less short term debt, while the reverse is true for a time and industry dummies (separately) on the
high inventory turnover. Hence, debt categories same set of explanatory variables that was used in
apparently are related to the collateral value of the the OLS and LSDV analyses.E~ The results of
asset categories corresponding to turnover rate, these exercises are included in Table II, which also
i.e. the maturity structure of debt strongly relates contains a set of estimation results regarding total
to the "maturity structure" of assets. These results debt. Since the corrections for autocorrelation and
could indicate that rules of thumb, proclaiming heteroskedasticity are done separately for long,
that the time span of investments and their financ- short and total debt, the results do not add up
ing should be matched (the "golden balance sheet vertically, only horizontally. In Table II, the OLS
rule" as it is called in The Netherlands), have results are labelled "total effect' because they
played an important role in the financial behavior represent the joint effects of the variables and the
of small firms over the period. 19 The effect of time and industry specific effects. The latter two
profitability, based on Myers' pecking order argu- are removed from the LSDV results which are,
ment, is supported for both short and long term hence, labelled "pure variable effect".
debt: the coefficients for ROI are negative and The first column of Table II confirms, that a
significantly so in both cases. large part of the influences pertain to the maturity
Taken as a whole, the determinants considered structure rather than the overall level of debt: the
here (except the tax effect) seem indeed to be effects on the short and long term debt ratio are
relevant for the shoptypes involved, although they partly of opposite signs. The on balance results of
appear to influence the maturity structure of debt the total debt ratio reflect the large proportion of
rather than its overall level. Finally, it is noted that short term debt in total debt. The decomposition
the results obtained here are largely equivalent to of these results in the second, third and fourth
those of earlier analyses of the same data set in column shows, that in the relations for long and
which less elaborate estimation techniques are short term debt the pure variable effects are
used (Van der Wijst, 1989, Van der Wijst and dominant for most variables, except the insignifi-
Thurik, 1990). cant size variable (total assets) and depreciation
The second way in which the results are charges in the case of long term debt. The industry
reported on is presented in Table II, in which a specific effects are usually larger than the time
"decomposition analyses" is provided. This de- specific effects. For total debt, however, the
composition analysis involves the estimation of an industry specific effects are dominant for almost
OLS regression using the same transformed var- all variables, while the pure variable effects are
iables that were used in the LSDV analysis largely insignificant because the effects on long
(except, of course, the dummies). The differences and short term debt tend to cancel out. The
between the OLS and LSDV estimates spring composition of the effects of depreciation charges
62 Nico van der Wijst and Roy Thurik

TABLE II e.ee

Decomposition of regression analyses results

e.e4 -"

Pure Industry Time

Total variable specific specific
effect effect effect effect

Short term debt ~ -o.e

intercept 0.405 0.000 0.430 -0.026 1 1

total assets 0.245 0.122" 0.200 -0.077* -0,oo "
asset struct
invent, turnov.
depr. charg.
-0.590
0.009
0.615
-0.410
0.005
0.542*
-0.100
0.003
0.333
-0.081
0.001
--0.260*
..... :ig5
ROI --0.030 --0.016 --0.009 --0.005 II41I JLII~I loee lge$ lrto lo'rl twloe
"rz~

Long term debt

intercept 0.067 0.000 0.066 0.002*
Fig. 2. Coefficientsof time dummies
totalassets 0.198 0.032* -0.057 0.223
asset struct 0.545 0.394 0.091 0.061
invent, turnov. -0.003 -0.004 0.001 0.000" An almost monotonous decline of the values over
depr.charg. -0.504 -0.174" -0.119" -0.211 the period 1955 through 1967 is followed by a
ROI -0.002* -0.008 0.001" 0.004 similarly characteristic increase over the rest of the
Total debt time span. The largely insignificant time dummy
intercept 0.497 0.000 0.516 -0.019* coefficients for long term debt show a much less
total assets 0.502 0.203* 0.193 0.107' clear development over time. A check on relevant
asset struct -0.132 -0.037* -0.072 -0.022*
macro economic variables (such as interest rates,
invent, turnov. 0.007 0.002* 0.004 0.001"
depr. charg. 0.119" 0.172" 0.428 -0.482 inflation, economic growth, etc.) did not produce a
ROI -0.034 --0.025 -0.009 0.000" variable with a corresponding pattern, so there is
no ready explanation for the observed time struc-
Note: * means not significantly different from zero at a 5% ture in the values of the time dummies for short
level of significance. term and total debt. It should be noted however,
that although the pattern in the time dummies is
on short term and total debt is particularly com- unmistakable, their contribution to the explana-
plex, consisting of large, opposite time and indus- tion of debt categories is limited. So the time
try specific effects. We have no ready explanation structure in Figure 2 is a distinct but not a
for this result. On the other hand, the influence of dominant factor in the analyses.
asset structure on long and short term debt is
comparatively straightforward, consisting almost
entirely of pure variable effects with minor indus-
5. Conclusions
try and time specific components. Summarizing, it The present analyses of financial structure in a
can be concluded that the two way decomposition German panel of shoptype averages give rise to
of the effects on total debt shows the influences on several conclusions. First of all, most of the
leverage to be much more complex than the determinants of financial structure presented by
coefficients of the simple test in the lower left hand theory of finance appear indeed to be relevant for
corner of Table II suggest. the small business sector investigated here. Non-
The third way in which results of the LSDV debt tax shields, approximated by depreciation
regression are reported on is presented in Figure charges, of which no significant influences were
2. In this Figure, the coefficients of the 19 time found, are an exception. Secondly, however, the
dummies, obtained in the analyses of long and influences encountered in the analyses are far less
short term and total debt, are plotted against the straightforward than the hypothesized effects in
years to which they refer. Figure 2 shows, that the theory. Most variables influence the maturity
there is a distinct pattern in the values of the time structure of debt rather than leverage: the effects
dummy coefficients for short term and total debt. on long and short term debt tend to cancel out.
 Small Firm Debt Ratios 63

This makes the influences on total debt not only schtmg an der Universit/it zu KSln" as Sonderheft 17 respec-
smaller, but also more susceptible to industry and tively 25 by Westdeutscher Verlag, KSln respectively Verlag
time specific effects. Thirdly, time and industry Otto Schwartz, GSttingen and: "Die VermSgens -- und
Kapitalsituation des Einzelhandels in den Jahren 1976 und
specific omitted variables, or commonalities in the 1977" in: "Mitteilungen des Instituts fiir Handelsforschung an
variables used, also play an important role in the der Universit/it zu KSln', Jahrg. 31, no. 4, April 1979.
analyses of short and long term debt: the influence 7 "Umsatz, Kostern, Spannen und Gewinn des Einzelhandels
of a variable can consist for 50% or more of time in der Bundesrepublik Deutschland in dem Jahrzehnt 1949
and industry specific effects, but the pure variable bis 1958" respectively "1959 bis 1968" respectively "1969
bis 1978" published in 1962 respectively 1970 respectively
effects are dominant in these analyses. Further, the 1980 in the series "Sehriften zur Handelsforschung" as Schrift
time specific effects in the analyses of short term no. 22 respectively 44 respectively 63 by the Westdeutscher
and total debt show a distinct pattern of steady Verlag, KSln, (n. 22 and 44) and Verlag Otto Schwartz,
decrease until 1967, followed by a steady increase GSttingen (no. 63).
over the rest of the period. We have no ready s Bradley et al. include investment tax credits in their
measure of non-debt tax shields and use R&D expenses as a
explanation for this phenomenon. As a concluding separate indicator of non-debt tax shields; these variables are
remark it can be noted that quantitative tests of the not available in the data at hand.
empirical implications o f the theory of finance in 9 This variable is also used by, among others, Marsh (1982).
small business are an extensive and promising area 10 The inventories of shoptypes selling fast moving consumer
of research, although it is as yet largely neglected goods (e.g. supermarkets) typically are accepted as collateral
while the inventories of shoptypes selling goods subject to
in the literature on both finance and small busi- fashion, such as clothes shops, are not.
ness. 11 The share of retained earnings in equity, reflecting the
possibility to retain earnings over an extended period of time,
would be a better variable to capture the pecking order effect
Acknowledgements but unfortunately this variable is not available.
12 Information on prices is not available for all shoptypes; if
An earlier version of this paper was presented at information is missing, price indices of adjacent shoptypes or
the 8th meeting of the E U R O Working Group on of total retailing are used.
Financial Modelling, November 1--3, 1990. We 13 See Hsiao (1986) p. 41--47, for a discussion of the pros
thank working-group members, Prof. Dr. J. and cons of fixed and random-effects models.
14 Including e.g. both the slioptype dummies and the averages
Spronk and our colleagues at the Department
across periods per shoptype would, of course, produce
of Fundamental Research for their helpful com- perfect multicollinearity.
ments, and Erik M. Vermeulen and Ren6 G. J. den ~5 See e.g. Judge et al., 1982, chp. 14--15 for a description of
Hertog for computational assistance. this technique.
16 The estimation procedure involves regressing the residuals
on the residuals one period lagged.
No~s ~7 For the first ycar of observation u.t _ (,[l _ p2)ui, is used.
is For the first year of observation Z~ - ((~/1 p2)Zit ) - t7i
-

1 Some costs are reported to decrease with firm size, as is is used.

explained later on. 19 Recent research shows that this rule of thumb still influ-
2 This division is adopted from Van der Wijst (1989), which ences financial policy in small business; see Van der Wijst
contains a more comprehensive discussion of this subject; (1989, p. 93) for a more detailed discussion and references.
a short discussion of similar purport can be found in 20 For these analyses it must be assumed that the lime
McConnell and Pettit (1984). dummies are the same for all shoptypes and that the shoptype
3 A discussion and empirical evidence regarding this clientele dummies are the same for all periods. See Carree and Thurik
effect, as it is called, is provided by, among others, Copeland (1990) where a comparable method is used to estimate the
and Weston (1983). influence of entry barriers on retail profit margins.
4 Contract provisions may prohibit a borrower from incur-
ring additional debt, from paying dividends, from removing
assets from the business site, etc. References
s The data do not have a complete panel character: for some
shoptypes information is available for less years than it is for Baxter, N. D., 1967, 'Leverage Risk of Ruin and the Cost of
others. Capital', Journal o f Finance 22 (4), 395--403.
6 "Die Verm6gens -- und Kapitalsituation des Einzelhandels Bradley, M., G. A. Jarell and E. H. Kim, 1984, 'On the
in dem Jahrzehnt 1954 bis 1963" respectively "1964 bis Existence of an Optimal Capital Structure: Theory and
1973" published in 1966 respectively 1975 in the series Evidence', Journal o f Finance 39 (3), 857--880.
"Sonderhefte der Mitteilungen des Instituts fiir Handelsfor- Carree, M. and R. Thurik, 1990, Market Volatility, Entry
64 Nico van der W(ist and Roy Thurik

Barriers and Retail Profit Margins, Research paper 9002, Modigliani, F. and M. Miller, 1963, 'Taxes and the Cost of
Research Institute for Small and Medium-Sized Business, Capital: A Correction', American Economic Review
Zoetermeer. 53 (3), 433--443.
Castanias, R., 1983, 'Bankruptcy Risk and Optimal Capital Myers, S. C., 1977, 'Determinants of Corporate Borrowing',
Structure', Journal of Finance 38 (5), 1617--1635. Journal of Financial Economics 5, 147-- 175.
Copeland, T. E. and J. F. Weston, 1983, Financial Theory and Myers, S. C., 1984, 'The Capital Structure Puzzle', Journal of
Corporate Policy, Reading (MA): Addison-Wesley. Finance 34 (3), 575--592.
Dielman, T. E., 1989, Pooled Cross-Sectional and Time Series Myers, S. C. and N. Majluf, 1984, 'Corporate Financing and
Data Analysis, New York: Marcel Dekker. Investment Decisions When Firms Have Information
Ferri, M. G. and W. H. Jones, 1979, 'Determinants of Investors Do Not Have', Journal of Financial Economics
Financial Structure: A New Methodological Approach', 34 (3), 17--34.
Journal of Finance 34 (3), 631--644. Norton, E., 1990, 'Similarities and Differences in Small and
Flath, D. and C. R. Knoeber, 1980, 'Taxes, Failure Costs, and Large Corporation Beliefs about Capital Structure Policy',
Optimal Industry Capital Structure: An Empirical Test', Small Business Economics 2 (3), 229--245.
Journal of Finance 35 (1), 99--117. Rapoport, J., 1990, 'Financing of Technological Innovation
Haley, C. and L. Schall, 1979, The Theory of Financial by Small Firms: Case Studies in the Medical Device
Decisions, New York: McGraw-Hill. Industry', Small Business Economics 2 (1), 59--71.
Hsiao, C., 1986, Analysis of Panel Data, Cambridge: C a m - Remmers, L., A. Stonehill, R. Wright and T. Beekhuisen,
bridge University Press. 1974, 'Industry and Size as Debt Ratio Determinants in
Jensen, M. C. and W. H. Meckling, 1976, 'Theory of the Firm: Manufacturing Internationally', Financial Management
Managerial Behaviour, Agency Costs and Ownership (2), 24--32.
Structure', Journal of Financial Economics 3,305--360. Robichek, A. A. and S. C. Myers, 1965, Optimal Financing
Keasey, K. and P. McGuinness, 1990, 'Small New Firms and Decisions, Englewood Cliffs: Prentice Hall.
the Return to Alternative Sources of Finance', Small Ross, S. A., 1977, 'The Determination of Financial Structure:
Business Economics 2 (3), 213--222. The Incentive-Signalling Approach', Bell Journal of Eco-
Kim, E. H., 1'978, 'A Mean-Variance Theory of Optimal nomics 8, 23--40.
Capital Structure and Corporate Debt Capacity', Journal Scott, D. F., 1972, 'Evidence on the Importance of Financial
of Finance 33 (1), 45--63. Structure', Financial Management 2 (2), 45--50.
Leland, H. E. and D. H. Pyle, 1977, 'Informational Asymme- Scott, D. F. and J. D. Martin, 1975, 'Industry Influence on
tries, Financial Structure, and Financial Intermediation', Financial Structure', Financial Management 4 (1), 67--73.
Journal of Finance 32(2), 371--387. Walker, D., 1989, 'Financing the Small Firm', Small Business
Marsh, P., 1982, 'The Choice Between Equity and Debt: An Economics 1 (4), 285--296.
Empirical Study', Journal of Finance 37 (1), 121-- 144. Warner, J., 1977, 'Bankruptcy Costs: Some Evidence', Jour-
McConnell, J. J. and R. R. Pettit, 1984, 'Application of the nal of Finance 17 (2), 337--347.
Modern Theory of Finance to Small Business Firms', in P. Wijst, D. van der, 1988, Determinants of Financial Structure
M. Horvitz and R. R. Pettit (eds.), Small Business Finance in Retailing, Research Paper 8801, Research Institute for
1, Greenwich, Connecticut: JAI Press. Small and Medium-Sized Business, Zoetermeer.
Miller, M., 1977, 'Debt and Taxes', Journal of Finance 32 (2), Wijst, D. van der, 1989, Financial Structure in Small Business,
261--275. Theory, Tests and Applications, Berlin: Springer-Verlag.
Modigliani, F. and M. Miller, 1958, 'The Cost of Capital' Wijst, D. van der and A. R. Thurik, 1990, Determinants
Corporation Finance and the Theory of Investment', of Small Firm Debt Ratios, Research Paper 9005,
American Economic Review 48 (3), 261 --297. Research Institute for Small and Medium-Sized Business,
Zoetermeer.
 Small Firm Debt Ratios 65

Appendix. Shoptypes and key financial data per shoptype.

Shoptype CURR EQ TOTAL

/TA /TA ASSETS SALES

1 LEBENSMITYELEINZELHND 0.625 0.202 120.57 816.73

2 TABAKWARENEINZELHND 0.851 0.098 70.30 512.22
3 TEXTILEINZELHND HERR~NOBERBEKLEIDUNG 0.741 0.089 522.39 1594.60
4 TEXTILEINZELHND DAM~NOBERBEKLEIDUNG 0.719 0.110 427.37 1553.76
5. TEXTILEINZELHND HERREN-, DAMEN-BEKL. 0.704 0.082 928.31 2656.20
6 TEXTILEINZELHND HERRENAUSSTATTUNG 0.758 0.115 223.63 674.88
7 TEXTILEINZELHND WAESCHE, WIRK-STRICKWAREN 0.779 0.088 234.59 716.34
8 TEXTILEINZELHND HAUS-UND BETTWAESCHE usw. 0.624 0.106 321.07 938.09
9 TEXTILEINZELHND MIT GEMISCHTEM SORTIMENT 0.700 0.081 687.61 1762.12
10 SHUHEINZELHND 0.749 0.092 303.66 770.54
11 MOEBELEINZELHND 0.675 0.055 1056.10 2610.20
12 GLAS-, PORZELLAN-UNK KERAMIKEINZELHND 0.728 0.100 337.13 855.03
13 ELSENWAREN-UND HAUSRAT VORW KUECHENGER. 0.726 0.073 294.14 764.67
14 ELSENW.-UND HAUSR. VORW BAUBESCHLAEGEN 0.785 0.053 778.05 1953.31
15 ELSENWAREN-UND HAUSRATHND GEMISCHT SORT. 0.736 0.060 671.63 1619.46
16 BELEUCHTUNGS-UND ELEKTROEINZELHND 0.711 0.076 471.65 1233.41
17 RADIO-LrND FERNSEHEINZELHND 0.778 0.092 414.92 1171.57
18 TAPETEN-UND BODENBELAGHND 0.756 0.082 555.28 1708.31
19 DROGERIEN 0.703 0.142 130.64 426.45
20 LEDERWAREN-UND GALANTERIEWARENHND 0.740 0.108 231.63 708.46
21 PHOTOEINZELHND 0.766 0.147 331.65 1143.48
22 UHREN-, JUWELEN-, GOLD-, UND SILBERWARENHND 0.821 0.083 389.56 689.57
23 SPORTARTIKELEINZELHND 0.778 0.095 369.24 1122.13
24 SORTIMENTSBUCHHND 0.830 0.103 231.28 1034.73
25 PAPIER-, BUEROBEDARF-UND SCHREIBWARENHND 0.762 0.106 250.87 936.06
26 BUEROMACHINEN-, BUEROMOEBELHND 0.805 0.079 1066.14 3897.77
27 BLUMENBINDEREIEN 0.592 0.262 62.20 425.82

Note: The first column (CURR/TA) contains current assets divided by total assets and the second column (EQ/TA) contains the
ratio of equity to total assets. The third and fourth column show total assets and sales respectively in 103 DM, current prices. All
figures refer to data averaged over 20 years of observation.