4
. Methodology and Data
4.1 Sample
Guiso (2003) highlighted the fact that Italy has many more small businesses than found in
countries at similar stages of development. Kumar et al. (1999) noted that the 3.2 million firms in Italy
have an average of 4.4 employees whereas the average firm size, measured by the number of
employees, in Germany, France, and the UK is respectively 10.3, 7.1, and 9.6. In Italy, firms with less
than 100 employees account for close to 70% of total employment, while in Germany, France, and the
UK such firms do not contribute more than 30% to employment.
As discussed above, small businesses strictly depend on external finance and, at the same time,
are vulnerable to asymmetric information problems. They can thus be constrained by leverage
decisions. For this reason, capital structure is a relevant topic for small firms,
because it influences their
growth patterns. In Italy, credit availability has a strong impact on the growth potential of small firms
and on the creation of new ones (Zingales et al. 2004). Therefore, small and medium-sized Italian firms
provide an interesting case-study to analyze the relationship between asymmetric information, growth
cycle, and capital-structure decisions. The sample employed in the study was stratified according to the
definition of small and medium-sized firms, defined by EU criteria and based on information obtained
from the database13. The AIDA (Analisi Informatizzata Delle Aziende) database, collected by Bureau Van Dijk,
was used in selecting the companies comprising the study sample. A panel-data analysis was
carried out to empirically examine the previously described hypotheses14.
The panel sample was made
up of 10242 Italian non-financial small and medium-sized firms not involved in a bankruptcy process;
the period studied was from 1996 to 2005. The dataset was restricted to observations that embodied all
essential variables for which a record of at least 5 years over the study period was available. The
number of firm-year observations was well-balanced across the sample. All of the variables used in the
study were based on book values 4.2 Methodology and Dependent Variable
To understand corporate financing decisions concerning the capital structure of small firms,
sensitivity to asymmetric information along the life cycle was verified. This was done with an
empirical procedure that considered the above-described research hypotheses. To account for
information opacity across the different stages of the life cycle, the following model was estimated:
Leverage = f [Age, Age2
, Profitability, Local financial development indicator, Size, Tangibility, Group
participations, Ownership structure, Growth opportunities, industry dummies]
Empirically we applied a least-squares dummy variable (LSDV) approach, as Michaelas et al.
(1999) have done.
Since the sample was quite large (69,694 observations), there were no problems
concerning degrees of freedom in the application of a fixed-effects model estimated in the least-squares
dummy variable (LSDV) form. This approach introduces firm type (industry) and time-specific effects
into the regression equations, which, in turn, reduces or avoids bias with respect to omitted variables.
As a result, the firm type (industry) and the time-specific effects of the omitted and the included
variables are captured (Showalter 1999). The econometric technique used in the model included the
computation of heteroskedasticity-consistent standard errors.
The dependent variable used as a proxy of
capital structure was financial leverage. This was calculated as the ratio of financial (or interestbearing) long-term and short-term debt (excluding trade debt) divided by the total financial debt plus
equity (as in Titman et al. 2003, Giannetti 2003, Rajan and Zingales 1995).
The model previously expressed is intended to analyze capital structure decisions along the life
cycle, through the use of the variable age. Firm’s age was calculated as the natural logarithm of the
number of years since the date of its incorporation.
This number was used to determine the stage of the
firm during its life cycle, its development of a reputation, and the amount of available information concerning the firm and its quality. A positive (or negative) relationship between this variable and debt
levels was expected according to hypotheses 2.a, and 2.b (3.a and 3.b). In addition, capital-structure
variation at specific threshold points, as argued by Berger and Udell (1998), was taken into account.
Specifically, changes in capital structure can be a non-linear function of a firm‘s age, as considered by
Brewer III et al. (1996). Thus, to account for non-linearity in the model, the variable AGE2
was
included15.
The model consider and control explicitly for the existence of a non-homogeneous financial
growth patterns in the capital structure determinants.
Concerning industry affiliation, as shown by Harris and Raviv (1991), the relevance of industryspecific features on the capital structure decisions requires the inclusions of industry dummies in the
model, in our analysis based on the classification’s first two-digits. Instead to present in tables the
coefficients for all the dummies the analysis showed the existence of an homogeneous financial life
cycle through industries comparing the results for high-tech and traditional sectors.
However,
considering that Holmes and Cassar (2001) indicate that the control of industry grouping in the
regressions had limited effect on the inferences, although industry effects were generally found to be
significant, it is noteworthy to verify whether small and medium-sized firms may experience different
financing life cycles, both within and across industries. Specifically, at the empirical level, the analysis
compare the general results with the output provided considering a sample of firms operating in highgrowth industries and a sample of firms operating in traditional and mature industries.
The role of the institutional context, fundamental in the provision of funds to small and mediumsized firms, is considered in the model by considering the well-known difference in the South of Italy,
characterized by a poor developed and inefficient institutional context, against other Italian macroareas. Regions that are financially better developed can offer credit to firms at a reasonable price;
by
contrast, in the regions of the South of Italy, with a low level of financial development and a low
protection by the court, the large amount of asymmetric information makes it unlikely that a small firm
will have access to reasonably priced external financing. For these reasons, the dummy South, a
dummy equal to one for regions south of Rome, is expected to reflect the negative influence of poor
and inefficient institutional context on the access to debt, particularly in the case of young firms which
are typically in need of external finances.
Several other proxies that have been mainly used in the empirical literature were selected as
explanatory variables. The variable size, measured by the log of total assets, is included in the model
(Sogorb-Mira 2005, Michaelas et al. 1999). Since large firms tend to have more easily collateralized
assets and more stable cash flows, a company’s size is inversely related to the probability of default,
allowing to carry more debt. Diamond (1991) also noted that large established firms have better
reputations in the debt markets, which allows them to carry more debt. In the model we considered also
asset tangibility, measured as the ratio of the property, plant, and equipment to total book assets, that
can be used as collateral (Van der Wijst and Thurik 1993, Ang 1992). Similar to the issue of size,
tangible assets or collateral, by conveying information to investors about the quality of a firm, reduce
the degree of information asymmetry and opaqueness (Bonaccorsi di Patti and Dell’Ariccia 2004). In
the model a dummy group is considered;
belonging to a business group can mitigate the information
asymmetry problems that plague firms going externally to obtain credit (Verschueren and Deloof 2006,
Deloof and Jegers 1999). Growth opportunities, defined as the percentage change in sales from the
previous to the current year, is also considered (Sogorb-Mira 2005, Michaelas et al. 1999, Titman and
Wessels, 1988). Due to the fact that the governance of a firm, and thus its financial decision-making, is
strictly influenced by the ownership structure (Jensen and Meckling 1976), our analysis contained a
variable that addressed different levels of ownership control, with 0 defined as ownership concentration
less than 5%, and 8 representing greater than 50% direct or indirect control of a firm.