Income smoothing with unlimited liability firms

*
Income smoothing with unlimited liability firms
Jochen Bigus**, Nadine Georgiou and Philipp Schorn
Abstract
We analyze income smoothing in private small and medium-sized enterprises (SMEs) in Germany
using a unique database of the German central bank containing around 18,000 firm-year observations.
Due to reduced agency problems of debt, we expect and find that unlimited liability firms (sole
proprietors, partnerships) have less need to disclose stable net income than limited liability firms
(corporations). Income smoothing with unlimited liability firms is approximately 20% to 30% lower.
Further, we expect and find that unlimited liability firms have stronger incentives to smooth income
for tax reasons. Income smoothing significantly increases with bank debt, but only in corporations. To
sum up, income smoothing of private firms is not homogeneous but is significantly affected by the
legal status and its various motives including information provision, contracting and tax reduction.
Keywords: Unlimited liability firms, private firms, income smoothing, taxes and financial accounting,
debt and financial accounting
JEL: M41, G32, G35, K34
*
We thank the research center of the Deutsche Bundesbank, especially Dr. Liebig and Dr. Heid, for giving us
access to the Bilanzdatenbank and Dr. Stein and Dr. Memmel (both from the Deutsche Bundesbank) for the
valuable advice they provided on how to use it. We thank Hans B. Christensen (Chicago), Harris Dellas (Bern),
Joachim Gassen (Humboldt, Berlin), Al Ghosh (Baruch College, New York), Igor Goncharov (WHU Vallendar),
Urska Kosi (Humboldt, Berlin), Jörg Werner (Frankfurt), and seminar participants at the European Accounting
Association Conference, Rome 2011; the EUFIN Conference, Bamberg 2011;the IAAER Conference, Eschborn
2013; and those at the universities of Bern, Humboldt (Berlin), Leipzig and Leuven for their valuable comments.
**Corresponding author: Jochen Bigus, Freie Universität Berlin, Garystraße 21, 14195 Berlin, Germany,
[email protected]; T: +49-30-83852509, F: +49-30-838455113. Nadine Georgiou is at Freie Universität
Berlin; Philipp Schorn is at Rhine-Waal University of Applied Sciences, Germany.
1
1
Introduction
The literature on the determinants of income smoothing includes many factors such as increasing
manager wealth, reducing information uncertainty, decreasing agency costs, lowering cost of debt,
decreasing taxes and addressing manager and shareholder risk aversion (Grant et al. 2009; Gassen et
al. 2006). With private firms, cost of debt, taxes and shareholder risk-aversion are important issues:
Private firms do not have access to public equity markets and have few and suboptimally diversified
shareholders, and they often provide single accounts for both financial and tax accounting. Income
smoothing is a suitable accounting device to simultaneously address all three issues: tax reduction,
information provision to creditors and contracting with creditors and shareholders.
This paper looks at income smoothing of private firms in detail. We do this for two reasons. First,
there is virtually no literature that addresses income smoothing in private firms at the firm level.
Second and more importantly, we want to go deeper and investigate the question of whether private
unlimited liability firms, such as (sole) proprietorships and partnerships, have different incentives to
smooth income than private limited liability firms (corporations) do. Even though unlimited liability
firms account for a large proportion of all firms (USA: 53%, UK: 73%, Germany: 75%),2 virtually no
accounting research has been conducted on this type of firm. We investigate the link between
unlimited liability and income smoothing incentives for a unique dataset of German private firms.
We expect unlimited liability firms to have different needs to smooth income than do corporations for
three reasons. First, due to unlimited liability and the fact that owners are at risk of losing virtually all
their fortune, agency problems of debt are less severe, such that there is less need to smooth income in
order to signal low default risk to creditors (Trueman and Titman 1988; Tucker and Zarowin 2006;
Gassen et al. 2006). For the same reason, there is less need to write debt contracts based on accounting
data, e.g. covenants on debt to EBIT/EBITDA or interest coverage (Dichev and Skinner 2002;
Nikolaev 2010; Christensen and Nikolaev 2012). The latter argument especially refers to contracts
with banks, which are considered experts in writing and enforcing covenants with private firms
2
See Mach and Wolken (2006), BIS (2011) and Statistisches Bundesamt (2009) (German Statistics Agency)
respectively.
2
(Berger and Udell 1995; Boot 2000; Elsas 2005). Second, according to German company law, payouts to owners of unlimited liability firms might exceed net income, which is not allowed for
corporations. Given that owners are risk-averse, corporations have a stronger incentive to smooth
income in order to smooth dividend pay-outs. The first argument addresses the information and
contracting role of accounting in debt financing; the second argument stresses the role of financial
accounting in determining shareholders’ dividends. The third argument is related to taxes: Because
owners of unlimited liability firms often invest a large part of their fortune in the firm, they have
limited opportunities to influence personal taxable income by other income sources. Further, German
unlimited liability firms must not separate ownership and control whereas separation is possible and
more common with corporations. Professional managers are supposed to pursue their own goals with
financial accounting and thus might be less interested in tax management than owner-managers. As a
result, we expect that unlimited liability firms have stronger incentives to decrease expected tax
payments by income smoothing.
We obtain the following results. The total level of income smoothing is about 20% to 30% lower with
unlimited liability firms than with corporations controlling for size, leverage, profitability, growth,
riskiness, industry and year effects. Income smoothing is positively associated with tax avoidance
incentives, the association being stronger with unlimited liability firms. Income smoothing
significantly increases with bank debt, but only in corporations. These results are robust to several
modifications of the econometric model that address, for instance, different measures of income
smoothing, cross-sectional dependence and potential endogeneity problems.
We use a unique database run by the German central bank (Deutsche Bundesbank). This database
contains standardized non-consolidated financial statements of private German firms that sold bills of
exchange to commercial banks. These commercial banks then proceed to sell the bills to the Deutsche
Bundesbank. Since the firms were obligors, the Deutsche Bundesbank asked for financial accounting
data prior to buying the bills, even for unlimited liability firms. Because German company law does
not generally require unlimited liability firms to disclose their financial statements, this is the only
comprehensive database containing reliable financial accounting data of German unlimited liability
firms.
3
Since unlimited firms are relatively small but corporations can be large, we limit our analysis to small
and medium-sized firms (SMEs). This allows us to better separate the differences in liability status
from differences in size and agency problems related to size. With larger firms, agency problems of
equity due to the separation of ownership and control become more pronounced, and financial
accounting choices are likely to be significantly affected by them. Our definition of SMEs is based on
that given by the European Commission. A firm qualifies as a SME if it has fewer than 250 employees
and, additionally, if sales are below €40 million and/or if total assets do not exceed €27 million
(European Commission 1996, 2003). Our final sample consists of 17,982 firm-year observations on
SMEs for the time period 1996-2004, 6,451 of which relate to unlimited liability firms.
There are several contributions to the literature. First and most importantly, to our knowledge this is
the first study that explicitly focuses on income smoothing in unlimited liability firms. In doing so, it
contributes to the small amount of literature on income smoothing in private firms. Gassen and Fülbier
(2010) find for a large sample of European private firms that income smoothing depends on firm-level
and country-level determinants. At the firm level, income smoothing increases with debt financing and
with tax-book-conformity. They do not explicitly consider the liability status, though, and neither do
they address the question as to how its interaction with tax saving incentives and different forms of
debt financing affects income smoothing. Burgstahler et al. (2006) analyze the institutional drivers of
earnings management including income smoothing for public and private firms at a country-industrylevel. Van Tendeloo and Vanstrealen (2008) find that private firms with Big4 auditors show lower
levels of earnings management and also lower levels of income smoothing. Other papers dealing with
private firms focus on other financial accounting patterns.3 Ball and Shivakumar (2005) find that
private firms in the UK show significant lower levels of conditional conservatism than public firms.
Goncharov and Zimmermann (2006, 2007) look at the incentives of Russian private and public firms
to reduce earnings and to avoid losses in order to save tax payments without deterring lenders. Peek et
al. (2010) find that asymmetry in earnings timeliness becomes more important for public firms than
3
Coppens and Peek (2005) find that private firms avoid losses less often in countries with high financial and tax
accounting alignment (e.g. Belgium, France, Germany, Italy). Garrod et al. (2008) show that more profitable and
larger private firms are more likely to write-off.
4
for private firms in countries with a higher degree of creditor protection. To sum up, the general notion
is that there is more earnings management in private firms than in public firms and that these
differences are related to institutional, country-specific factors (e.g., tax-book conformity, investor and
creditor protection, rule of law). In contrast, we keep the institutional setting constant and provide a
“micro-analysis” within the group of private firms. Further, since agency problems of equity are
supposedly less pronounced in our sample we believe that income smoothing is not an earnings
management device. Rather, income smoothing might be an efficient tool to simultaneously address
information and contracting needs in debt financing, to address shareholder risk aversion and to reduce
taxes.
Second, we are able to disentangle the driving forces of income smoothing dependent on the private
firm’s legal status. Corporations generally smooth income more than unlimited liability firms do. Tax
avoiding incentives matter but more so for unlimited liability firms. Bank financing is positively
associated with income smoothing, but for corporations only. In contrast, with other forms of debt
financing we do not find differences. Hence, we also provide a more detailed understanding of the
relationship between different forms of private debt and income smoothing.
There is a large body of literature addressing income smoothing with publicly listed firms. In contrast
to many private firms, ownership and control are separated with public firms and information
asymmetries between managers and (minority) shareholders are much more pronounced.
Consequently, there is a well-developed literature suggesting that income smoothing serves managers
(and/or controlling shareholders) by masking the expropriation of (minority) shareholders, especially
within a weak governance setting (DeFond and Park 1997; Leuz et al. 2003; LaFond et al. 2007;
Huang et al. 2009), or by masking the underlying risk of the firm (Grant et al. 2009). On the other
hand, income smoothing is found to signal managers’ private information on future earnings (Trueman
and Titman 1988; Tucker and Zarowin 2006) which tends to reduce information uncertainty in stock
markets (Bowen et al. 2008; Chen 2012, more critically; Jayaraman 2008) and bond markets (Gu and
Zhao 2006; Soderstrom et al. 2012).
5
There is much less work for public firms on the relationship between income smoothing and private
debt, between income smoothing and tax incentives or between income smoothing and shareholder
risk aversion. These are issues which are important for private firms but are certainly less relevant for
public firms because the latter have access to the organized capital market and their shareholders are
easily able to diversify. Moreover, the financial statement of public firms is only mildly influenced, if
at all, by tax considerations since they use a tax statement for this purpose. There are papers
acknowledging that publicly listed firms’ level of income smoothing is also affected by the country’s
level of tax-book-conformity (LaFond et al. 2007; Burgstahler et al. 2006); however, evidence on the
firm level is scarce. With regard to private debt, Amiran and Owens (2012) find that income
smoothing decreases (increases) the public firm’s cost of private debt in countries with a strong (weak)
governance setting. Abdel-Khalik (2007) shows that income smoothing increases with CEO risk
aversion as measured by CEO wealth and the risky component of CEO compensation. We were unable
to find any empirical work addressing shareholder risk aversion.
Our findings relate to corporations and unlimited liability firms in Germany. Since the basic agency
problems of limited and unlimited liability firms should not differ across countries, we expect that
corporations smooth income more in other countries as well. The country-specific regulatory
environment as well as the corporate governance and financing patterns may affect other variables
though. For instance, we would expect less pronounced results where payouts to shareholders do not
depend on the firm’s legal status, as is the case in some (but not all) US states. We would also expect
less pronounced results in countries with weaker tax-book-conformity and where private firms rely to
a lesser extent on bank financing and more on internal or equity financing. Bank debt is quite
important to SMEs in Germany (Agarwal and Elston 2001; Franks and Mayer 1998) and also to SMEs
in the USA (Berger and Udell 1998).
Our study suggests that private firms are not homogeneous, which is reflected in their accounting
choices. Moreover, it is important to note that the financial statement of private firms does not only
serve to reduce information uncertainty but needs to address several, sometimes even conflicting,
goals. Consequently, one may ask whether the regulators’ efforts to make financial reporting of private
firms more informative (e.g., IASB, 2009) makes it more difficult to pursue the other goals.
6
The paper is organized as follows. Section 2 reviews the literature in more detail and develops the
hypotheses. Section 3 provides a description of the data and methodology used. Section 4 presents the
results of the regression analyses on income smoothing and section 5 concludes.
2
Literature review and hypotheses
2.1
Unlimited liability, income smoothing, and payouts to shareholders
Theoretical background. Risk-averse shareholders prefer to reduce the volatility of the net income
stream (Ronen and Yaari 2008; Gassen et al. 2006) if two conditions hold: First, that the shareholders
find it difficult to perfectly diversify risk and second, that dividends must be tied to earnings. The
former condition is usually met with private firms because the equity market is not perfect there and
the transaction costs in trading shares of private firms are substantial. Whether the latter condition
holds depends on the legal setting. According to German company law, dividends paid out by
corporations must not exceed the accumulated net income (Sections 58, 150 Aktiengesetz and Section
29 GmbH-Gesetz). However, manager-owners of unlimited liability firms are allowed to distribute
more than net income (Section 122 Handelsgesetzbuch).
Prior empirical findings. To the best of our knowledge, there is no evidence on the relationship
between shareholder risk aversion and income smoothing. Abdel-Khalik (2007) shows for a sample of
US public firms that income smoothing increases with CEO risk aversion as measured by CEO wealth
and the risky component of CEO compensation. Similar to managers of publicly listed firms,
shareholders of private firms suffer from imperfect risk diversification.
Hypothesis. There is evidence that risk aversion induces incentives to smooth payouts. Thus, income
smoothing matters if net income determines the pay-out to shareholders. In Germany, dividends are
tied to net income in corporations, but less so in unlimited liability firms.
7
Hypothesis 1:
Unlimited liability firms exhibit lower levels of income smoothing than firms with limited liability
(corporations).
2.2
Unlimited liability, income smoothing, and bank financing
Theoretical background. Income smoothing is beneficial when it conveys valuable information to
banks on future performance (Ronen and Yaari 2008). Managers have private information on future
earnings and income smoothing is considered a device to publicly disclose this information (Trueman
and Titman 1988; Tucker and Zarowin 2006). To make signaling work, income smoothing must
impose certain costs. Such costs could include the fact that early recognition of book income also
causes early recognition of tax income (Dou et al. 2012). Further, there are effort costs related to the
smoothing process itself and, potentially, in convincing the auditor that smoothing is in line with local
standards. Effort costs are especially high in “bad” years, but might be worthwhile in order not to lose
the reputation built up in the previous years. Reliable disclosure of private information might also
include proprietary costs. In corporations, which can be run by professional managers, income
smoothing might arise with the cost of reduced or delayed bonus compensation.
Further, due to their concave pay-off function, creditors prefer a lower volatility of income since this
reduces the probability of default. Creditors assess default risk based on both debt ratios and
performance indicators (Christensen and Nikolaev 2012). Income smoothing reduces fluctuations in
net income but may also reduce fluctuations in debt ratios.
While these arguments are related to the information role, income smoothing also serves efficient debt
contracting. With debt contracts, interest coverage (EBIT/interest expenses) and the ratio of debt to
EBIT or to EBITDA are commonly used as accounting-based covenants (Nikolaev 2010: 146; Dichev
and Skinner 2002: 1101). Covenant violations, such as on interest coverage, may trigger certain
contractual rights to banks, for instance permitting them to require more collateral or a higher interest
rate or even to recall the loan. Income smoothing helps to reduce the probability of accounting-based
8
covenants being violated. Usually, failure to meet the covenants triggers special rights to the lender
which are costly to the firm, such as an increase in cost of debt or the provision of additional
collateral.
We expect that agency problems of debt are less severe with unlimited liability firms, because owners
are at risk of losing virtually all their fortune. Consequently, there is less need to smooth income in
order to signal low default risk to creditors. For the same reason, there is less need to write covenants.
Thus, reduced agency problems of debt are another important argument as to why we would expect
lower levels of income smoothing with unlimited liability firms (see Hypothesis 1).
Prior empirical findings. Income smoothing tends to reduce information uncertainty in stock markets
(Tucker and Zarowin 2006; Bowen et al. 2008; Chen 2012) and bond markets (Gu and Zhao 2006;
Soderstrom et al. 2012). There are ambiguous results regarding the association between income
smoothing and leverage: LaFond et al. (2007) find a positive association, Dou et al. (2012) report
mainly negative associations. Amiran and Owens (2012) find that income smoothing decreases the
public firm’s cost of private bank debt in countries with strong governance. In countries with weak
contract enforceability Dou et al. (2012) show that income smoothing is more pronounced in
industries with a greater need for relationship-specific investments by, for instance, suppliers.
However, in Germany contract enforceability is considered to be strong. Gassen and Fülbier (2010)
find a positive association between debt financing and income smoothing for private firms. We are not
aware of any evidence regarding income smoothing which differentiates between bank debt and other
forms of debt.
There is evidence that bank debt is the single most important source of outside financing with small
and medium-sized German firms (Elsas and Krahnen 2004). Further, banks are considered experts in
writing and enforcing covenants with private firms (Berger and Udell 1995; Boot 2000; Elsas 2005).
We are not aware of any evidence that other lenders use accounting-based covenants in debt
contracting.
Hypothesis: We think that in the context of bank financing the contracting role of income smoothing is
more important for corporations than for unlimited liability firms. Owners who can be held liable for
9
their private assets should have stronger incentives to avoid default. This holds even if the private
assets are small in size. The important fact is that owners are at risk of losing their entire private
fortune. This threat may reduce agency problems of debt significantly, especially since there is
evidence that an individual’s loss aversion increases with the size of loss relative to their total fortune
(Holt and Laury 2002). Due to the bank’s access to private information, the information role is
supposed to be less relevant.
As a result, we believe that agency problems of debt and default risk are higher with corporations,
such that there is a greater need for debt covenants and thus for income smoothing in order to avoid
covenant violation. Because banks are experts in debt contracting we expect that income smoothing
increases with bank debt, but that the relationship is stronger for corporations.
Hypothesis 2:
The association between income smoothing and bank financing is stronger for corporations than for
unlimited liability firms.
Debt financing. The information role and contracting aspect of income smoothing generally applies to
all types of debt financing, not only to bank financing. Other debt mainly consists of trade credit,
advanced payments and note payables.4 Suppliers may wish to see a stable income of the firm just as
much as banks do. We are not aware, however, of any evidence of suppliers of SMEs systematically
using financial reporting data for contracting purposes. In contrast to bank loans, suppliers’ claims are
rather short term. Due to the fixed costs of contracting, it does not pay to set up contractual provisions
if the claim size is too small. Thus, we do not necessarily expect differences related to liability status.
In summary, we believe there may be an association between income smoothing and debt financing
based on information aspects and to a lesser extent on contracting purposes. Since the contracting
aspect matters less with debt financing than with bank financing, we may not find that debt financing
4
We do not consider provisions because they do not result from debt contracting.
10
affects income smoothing substantially differently with unlimited liability firms than with
corporations.
2.3
Unlimited liability, income smoothing, and tax avoidance incentives
Theoretical and legal background: From a tax perspective, income smoothing is beneficial when two
conditions hold: first, when financial accounts are used for tax purposes and second, when income
smoothing lowers the present value of tax payments.
In Germany, tax accounts and financial accounts generally do not differ much. In fact, most private
firms provide a single set of accounts. Owners of both corporations and unlimited liability firms are
taxed on their personal income, including income generated from owning the firm. With low and
medium levels of income, the tax rate increases with income. For sufficiently high levels of income,
the tax rate becomes flat. Apparently, with low and medium income levels, there is a stronger
incentive for income smoothing to decrease the average tax rate over time. With a flat tax rate there is
generally no tax benefit from income smoothing because the average tax rate does not change and is
already at the maximum.5
Thus, for sufficiently high levels of income, individuals have no incentive to smooth income over time
whereas this may be the case for lower income levels. Rather than being affected by the legal status, a
tax-based incentive on income smoothing may depend on (1) the actual net taxable income of the firm
(2) the individual owner’s shares and (3) the other opportunities available to influence their personal
taxable income. We are only able to observe the net income of the firm. We lack firm-specific
5
For instance, in 2000 the maximum marginal income tax rate was 51% with an income of about €60,000 or
higher. For €30,000 it was 35%. The effective tax rates were 33.5% and 23.1% respectively. Assuming the “real”
income is €0 and €60,000 in two subsequent years, implying a tax of €0 and €20,100, there is an incentive for
income smoothing by “assigning” €30.000 to each year because the tax is €6,900 in each year and €13,800 in
total. Note that smoothing does not pay with a “real” income distribution of, e.g., €200,000 and €400,000 (taxes:
€91,474 and €193,474). Having an income of €300,000 in each year implies taxes of €142,474 in each year.
11
ownership and tax data. The German Statistics Agency states that the median number of owners with
partnerships and limited partnerships is 2 and the mean is 2.4 and 4.4, respectively (Statistisches
Bundesamt, 2012: 13). Moreover, about 90%-95% of partners’ individual taxable income is from the
business activity; only a small part stems from other income sources (capital income, income from
rents and leases etc., Statistisches Bundesamt, 2012: 20). We were unable to find ownership and tax
data on the owners of corporations.6
We predict that income smoothing is positively associated with tax avoidance incentives, that is, with
a sufficiently low net income of the firm.
Prior empirical evidence. There is only limited evidence on the link between income smoothing and
tax avoidance incentives. LaFond et al. (2007) and Burgstahler et al. (2006) report that publicly listed
firms’ level of income smoothing is positively affected by the country’s level of tax-book-conformity
(multiplied by the average country-specific corporate tax rate in Burgstahler et al. 2006); however,
evidence at the firm level is scarce.7 Gassen and Fülbier (2010) find that consolidated statements of
European private firms show lower levels of income smoothing than unconsolidated ones.
Hypothesis. We expect that income smoothing generally increases with tax avoidance incentives. For
two reasons, we expect tax avoidance incentives to matter more to unlimited liability firms than to
corporations. First, since owners of unlimited liability firms receive about 90%-95% of their income
from their business activity it may be reasonable to assume that they have limited opportunities to
influence personal taxable income using other income sources. Second, unlimited liability German
firms must be run by the shareholders. Corporations are more likely to be run by professional
managers, who are supposed to be less interested in tax management.
6
The NSSBF database includes ownership data on US private firms (Mach and Wolken 2006: A171). The
average number of owners is 1.2 and 2.9 with proprietorships and partnerships respectively; it is 2.0 and 10.2
with S corporations and C corporations respectively.
7
Unlike publicly listed firms, small and medium-sized private firms are less likely to have production facilities
in many different countries. Consequently, tax management rather has to rely on income smoothing rather than
on transfer pricing mechanisms.
12
Hypothesis 3:
The association between income smoothing and tax avoidance incentives is stronger for unlimited
liability firms than for corporations.
3
Data and research methodology
3.1
Data
We use the database Unternehmensbilanzstatistik run by the Deutsche Bundesbank. This database
contains financial accounting data, i.e. balance sheets and income statements with about 183,000 firmyear observations on SMEs for the fiscal years 1992 to 2004. We do not have complete financial
accounting data for the years after 2004.
We discounted financial institutions and firms with bonds outstanding in order to focus on private
non-financial firms.8
We also discounted any firms that do not meet the definition of SMEs according to the European
Commission (1996). In order to be able to compare the effects of legal status more effectively, we
want unlimited and limited liability firms to differ not too much in size. Size affects financial
accounting choices, not only directly but also indirectly since agency problems of equity tend to
become more pronounced as size increases. According to the European Commission, a firm qualifies
as a SME when it has fewer than 250 employees and when either sales are below €40 million and/or
total assets do not exceed €27 million. In addition to the size criteria, a SME must be independent in
the sense that no other firm holds more than 25% of the shares. Since we do not have any data on the
shares and shareholders, we assumed dependence on another firm if at least one of the following
criteria was met: (1) there is an obligation to provide consolidated statements (2) there is a contractual
8
Bonds outstanding are a balance sheet position. Unfortunately, we were unable to observe whether the firm’s
shares are listed because the data is anonymous. However, listed firms in the time period investigated do not
meet the size criteria of our SME definition, such that we are sure that there are no listed firms in our sample.
13
obligation to transfer losses or profits to another firm and (3) there are liabilities or financial claims
with regard to group firms.
The final sample of our study contains 17,982 observations. Most of the observations lost were due to
our research design since the dependent variable (income smoothing) and independent variables are
measured over a five-year horizon. We also had to exclude firms that changed legal status within this
five-year period or where data on equity, sales or total assets were (partly) missing or had implausible
negative values. In order to account for outliers, all variables are winsorized at the 1% and 99%
percentiles.
--insert Table 1 about here--
--insert Table 2 about here--
The database is unbalanced for three reasons. First, larger firms are likely to be overrepresented
compared to the total population of German private firms. Second, the number of balance sheets
decreases over time because financing via bills of exchange have declined in importance in recent
years (see Table 2). Third, there may be a selection bias, indicating that economically sound firms are
overrepresented.
The first bias is relevant since size is related to legal status. This is why we focus on SMEs only;
moreover, we control for firm size. The median firm in our sample is quite small and has total assets
of about €1.4 million (as a comparison: with Burgstahler et al. (2006), the median German private firm
has total assets of €38.2 million).
The second bias may be irrelevant for our study because the research questions are defined
independently of time constraints. The third bias has little impact on our results since we are mainly
interested in the differences between unlimited liability firms and corporations. We may even expect
the differences (and the results of our study) to be more pronounced for firms in a weaker financial
condition because agency problems of debt then become even more severe. Hence, there is a greater
need to signal low default risk by smoothing income, especially for corporations.
14
The basic model is a pooled OLS regression with robust standard errors clustered at the firm level. We
account for both year fixed effects and industry fixed effects; firms are designated to one of six
industries9. We also run a Fama-MacBeth analysis to account for potential serial correlation. Finally,
we address potential endogeneity problems.
3.2
Measurement of variables
Limited Liability
Within the period of investigation, German company law defines the following legal categories.
●
Sole proprietorship (one-person business, Einzelkaufmann), Section 1 German Commercial
Code (Handelsgesetzbuch, HGB): The owner provides equity to the firm, must run it, and is
unlimitedly liable with private assets for any (debt) obligations.
●
Partnership (Offene Handelsgesellschaft), Sections 105-160 HGB: The owners provide equity
to the firm, must run it (Sec. 114 HGB), and are unlimitedly liable with their private assets for
any obligations (Sec. 128 HGB).
●
Limited partnership (Kommanditgesellschaft), Sections 161-177a HGB: There are two types
of shareholders. One type (Komplementäre) is unlimitedly held liable with private assets and
has to run the firm. The other shareholders (Kommanditisten) are not allowed to run the firm
and can only be held liable for share capital that has not yet been paid in (Sec. 161, 164 HGB).
●
Corporation (Gesellschaft mit beschränkter Haftung, GmbH), GmbH-Gesetz: Shareholders
are not held liable with their private assets. There is a minimum capital requirement of
€25,000. The firm can be run by one or several professional managers, who do not need to be
shareholders.
9
The industries are: manufacturing; construction; accommodation and catering; transport and communication;
real estate, renting and business activities; wholesale and retail trade, garages.
15
●
Stock corporation (Aktiengesellschaft), Aktiengesetz: Shareholders are not held liable with
their private assets. There is a minimum capital requirement of €50,000. The firm is run by a
management board. A supervisory board is required. Managers do not need to be shareholders.
We considered the first three types of firms as unlimited liability firms and the latter two as
corporations. Limited partnerships are supposed to share similar accounting incentives as partnerships
since both are run by shareholders, who are unlimitedly liable.
There are also mixed legal categories. One quite common legal form in Germany is the so-called
GmbH & Co. KG, which is similar to a limited partnership but defines a corporation to be the
shareholder with unlimited liability. Consequently, the liability status cannot be clearly assigned,
which is why we excluded these firms from our analysis.10
Income smoothing
Following the literature, we measure income smoothing by dividing the variability of earnings over
time by the variability of income in economic terms, such as the variability of cash flow from
operations or of sales (e.g. Leuz et al. 2003; Bao and Bao 2004; Burgstahler et al. 2006). Our
smoothing measure is calculated using firm-level data:
(1)
SMTHi,t 
SD(net incomei,t / total assetsi,t 1 )
SD(operative cash flow i,t / total assetsi,t 1 )
 (1)
SD stands for standard deviation. Standard deviations are computed on the basis of financial data for
five fiscal years in order to mitigate the effect of abnormally high or low values on income or
operative cash flows. Following the literature (Burgstahler et al. 2006), we multiply by −1 so that
higher values correspond to more income smoothing. We scale by total assets, following, for example,
10
The same argument applies for another much less common mixed form (Kommanditgesellschaft auf Aktien,
KGaA), which is similar to a limited partnership where a stock corporation represents the shareholder with
unlimited liability.
16
LaFond et al. (2007) and Leuz et al. (2003). Operative cash flow is defined as the difference between
net income and total accruals (see Daske et al. 2006). Total accruals equal: ∆Inventory (or: ∆current
assets - ∆cash and cash equivalents) - ∆current liabilities + ∆short-term debt included in current
liabilities - depreciation and amortization - ∆provisions.
Bank debt, contractual debt, and tax avoidance incentives
Bank debt is defined as the ratio of bank debt to total assets, averaged over five years. Contractual debt
is defined as the ratio of contractual liabilities (= total liabilities minus provisions) to total assets,
averaged over five years. We have to define all control variables on a five-year time horizon due to the
measurement of income smoothing.
It is rather difficult to specify tax avoidance incentives because we have no firm-specific data on the
number of owners and their shares, single owners’ marginal tax rates or any other income sources. As
argued above, with low and medium levels of shareholder income the effective tax rate increases with
income whereas beyond a certain individual threshold income level the tax rate becomes flat. Hence,
we expect a stronger tax avoidance incentive for low and medium levels of income.
We assume that an individual’s income is roughly the firm’s income divided by a certain number of
owners. Other things being equal, the higher the assumed number of owners, the more likely the
owners are to earn a small or medium income where there are high tax avoidance incentives.
The data from the German Statistics Agency suggests that the mean number of owners is 2.4 with
partnerships and 4.4 with limited partnerships (Statistisches Bundesamt, 2012: 13). One-person
businesses have one owner; corporations usually have more owners than partnerships though we lack
data to corroborate this claim. We run pre-tests to determine which assumed number of owners
separates low and high tax avoidance incentives. Regardless of the firm’s legal status, a threshold level
17
defined as three times the individual threshold income separates high tax avoidance incentives (due to
non-flat tax rates) from low tax avoidance incentives (due to the flat tax rate) quite well.11
Other control variables
With small and medium-sized private German firms, there is hardly any firm-specific (or even
aggregated) data on management compensation, ownership structure or other corporate governance
data available because firms do not have to disclose this information. We also believe that agency
problems of equity are (much) less pronounced than with publicly listed firms and thus this data would
be of limited use.
There are other firm-specific variables included because they are likely to affect income smoothing
incentives quite independently of listing status (Dechow and Dichev 2002; Leuz et al. 2003; Francis et
al. 2004; Burgstahler et al. 2006; Gassen et al. 2006; LaFond et al. 2007; Dou et al. 2012).
Operating Risk. From an economic point of view, with higher operating risk comes a greater need to
smooth income in order to smooth dividend payouts. Thus we expect a positive association. We
measure operating risk by the standard deviation of sales divided by lagged total assets.
Firm Size. Even though we construct the sample with SMEs such that unlimited and limited liability
firms do not differ much in size, size is likely to matter. Larger firms may show higher levels of
income smoothing because they are expected to have a wider array of discretionary expenditures.
Since problems of asymmetric information are more pronounced with larger firms, there is a greater
need to indicate low default risk by income smoothing.
Firm Profitability. Firms with good performance have less need to impress creditors and therefore less
need to smooth income. We should, therefore, expect higher firm profitability to imply less income
smoothing.
11
With US private firms included in the NSSBF database, the average number of owners is 3, see Mach and
Wolken (2006: A171).
18
Losses. Firms that report losses may have a greater need to impress creditors with stable income. On
the other hand, firms that are unable to avoid losses may already have suffered reputation damage. We
may also observe a negative correlation due to the fact that a loss drives down the smoothing variable.
Overall, the sign of the association between income smoothing and losses is not clear.
Other reasons for income smoothing. Burgstahler et al. (2006) mention other potential reasons for
variations in accruals such as firm growth, the length of the operating cycle and audit quality. We
considered the length of the operating cycle as defined in Burgstahler et al. (2006), as well as firm
growth. Generally, high growth firms exhibit higher levels of operating risk, implying a greater need
for income smoothing. They also usually have higher accruals, e.g. due to higher investment levels,
and more discretion for income smoothing. We therefore expect a positive sign.
It transpires that audited financial statements did not show any significantly different income levels to
non-audited ones, leading us to omit an audit variable in the regressions. The audit variable might not
have a significant effect for two reasons. First, less than 12% of all financial statements in our sample
are audited. In Germany, there is no legal obligation for statutory audits to be performed for either
unlimited liability firms or small corporations. Second, auditor liability is very lenient in Germany
(Gietzmann and Quick 1998). Thus, an auditor is likely to accept discretionary accruals to a larger
extent than in other countries that have rather strict auditor liability, such as in the USA.
Table 3 depicts all control variables. Most variables are computed on a five-year average since the
variable on income smoothing also requires five years.
--insert Table 3 about here--
19
4.
Results
4.1
Univariate test statistics
As a first step, we divide the total sample into two subsamples in order to compare income smoothing
between the two subsamples. The subsamples are divided by (1) legal status, (2) tax avoidance
incentives and (3) importance of total bank debt. D_BANK is a dummy variable with value 1 if bank
debt to total assets (averaged over five years) exceeds the industry median; otherwise it is 0. Table 4
provides an overview.
--insert Table 4 about here--
As already stated the specification of income smoothing in (1) implies that higher values represent
more income smoothing. The statistics strongly suggest that income smoothing is significantly
stronger with (i) corporations than with unlimited liability firms and (ii) firms that have a strong tax
avoidance incentive. For instance, with corporations the volatility of net income amounts to only 32%
of the volatility in operative cash flow whereas it is 43% for unlimited liability firms. That is, the level
of income smoothing is about 33% lower with unlimited liability firms – in the absence of control
variables. Firms with strong tax avoidance incentives and with a lot of bank debt exhibit levels of
income smoothing that are about 41% and 14% higher respectively.
4.2 Multiple regression analysis (without interaction terms)
We expect limited liability firms to exhibit more income smoothing. We also expect a positive
association between income smoothing and contractual (bank) debt. Further, we expect a positive sign
for the variables on tax avoidance. The regression on hypothesis 1 is reflected by:
2 ,
,
,
,
,
,
,
,
,
,
,
,
20
Table 5 reports summary statistics.
--insert Table 5 about here--
The smoothing variable SMTH indicates that on average the standard deviation of scaled net income
is 36% of the standard deviation of scaled operative cash flows. Burgstahler et al. (2006) report a
mean ratio of 47% for German private firms excluding unlimited liability firms; Gassen and Fülbier
(2010) report a mean ratio of 56.7% for a large sample of European firms. About 64% of the firms in
the sample are corporations; thus, roughly one third are unlimited liability firms. The average ratio of
contractual debt to total assets is about 69%, less than half of which (29%) is bank debt.12 There is
relatively more variation in bank debt than in contractual debt. However, mean and median do not
differ much with either variable. Since we focus on relatively small firms, about 87% of the firms have
strong tax avoidance incentives.
Summary statistics also indicate that we have quite small firms in our sample with mean total assets of
€1.43 million. Since we restricted the sample to SMEs, the variation in the size variable is relatively
low. Note that other studies include bigger private firms.13 The average ROA (EBIT/lagged total
assets) is about 10.4%. However, the ROA variable spreads fairly widely. Although ROA is usually
positive over a five-year horizon, about 41% of the five-year series show a loss in at least one year.
Table 6 reports Pearson pair-wise correlations among the variables for our lead measure of income
smoothing SMTH. Spearmen correlation coefficients are similar (not tabulated). The univariate
analysis suggests that income smoothing increases in corporations with strong tax avoidance
12
(Bank) debt financing is quite common in Germany. The average contractual debt ratio of small US business
firms with less than 500 employees (based on the NSSBF database) is about 49%; the average bank debt ratio is
19% (Berger and Udell 1998: 620).
13
For instance, with Burgstahler et al. (2006) the median book value of private firms is €38.2 million. The mean
book value of private firms is £3.7 million with Ball and Shivakumar (2005), €28.5 million with Peek et al.
(2010) and €4.97 million with Gassen and Fülbier (2010).
21
incentives, with higher levels of contractual debt and with bank debt. As predicted, income smoothing
also increases with firm size, growth and lower ROA. Income smoothing is lower when there is a loss.
The correlation with the volatility of sales is not significant. Correlation coefficients do not indicate
severe multicollinearity. The correlation between the BANK and CONTRACTDEBT variables is
relatively high, but not critical (0.5631), which prompts us to include both variables in the basic
regression.14
--insert Table 6 about here—
Table 7 depicts the results of the basic multiple regression analysis. It also includes variations of the
basic pooled OLS regression model such as a regression based on a three-year definition of the income
smoothing variable and other variables. The table also depicts the results of an annual cross-sectional
Fama-Macbeth regression.
There is significantly more income smoothing with corporations than with unlimited liability firms. As
predicted, the LIM variable has a positive sign and is highly significant. With corporations, the level
of income smoothing is higher by 0.0939, that is, about 26% of the mean value (-0.3626, see Table 5).
This finding supports hypothesis 1.
The results also suggest that firms with more debt financing and stronger tax avoidance incentives
exhibit more income smoothing. Firms with strong tax avoidance incentives show income levels that
are 0.0459 higher, which is about 13% above the mean income smoothing level. If the debt ratio
increases by 10 percentage points, the level of income smoothing increases by 0.01875. The BANK
variable covers the marginal effect of bank debt on income smoothing. This marginal effect is weak
and statistically not significant indicating that income smoothing generally does not substantially
14
As a robustness test we also run regressions where CONTRACTDEBT2 only includes non-bank contractual
debt. The correlation with the BANK variable is still high (−0.5121) and qualitative results are correspondingly
similar. The negative correlation suggests that bank debt and other forms of contractual debt are substitutes to a
considerable degree.
22
differ between banks and other creditors. This finding is corroborated by the fact that the BANK
variable becomes highly significant (with positive sign) when we define CONTRACTDEBT2 as
contractual debt excluding bank debt (see Table 10, Panel C).
As predicted, firm size and growth are significantly positively associated with income smoothing;
ROA and LOSS have a negative sign. In contrast to the findings of Burgstahler et al. (2006), we find
that the length of the operating cycle and the variance of sales is not significant.15
--insert Table 7 about here –
Table 7 also exhibits results with a pooled OLS regression where income smoothing and other
variables are defined on a three-year basis. As a consequence, the number of observations increases
from about 18,000 to more than 31,000. The statistical and economic significance of the LIM variable
increase, while the results with BANK, TAX and CONTRACTDEBT remain similar.
We accounted for serial correlation by annual cross-sectional Fama-MacBeth regressions exhibiting
the mean coefficients of nine different annual cross–sections, starting with the 1992–1996 period.
Compared to the pooled OLS regression, the economic significance of the main variables LIM,
BANK, TAX and CONTRACTDEBT is similar, and statistical significance increases considerably.
4.3 Robustness checks
We ran several robustness checks. Most importantly, we used different specifications of the income
smoothing variable. We obtain similar results in both statistical and economic terms with regard to the
main variables LIM, BANK, TAX and CONTRACTDEBT (not tabulated) when adopting the
alternative measures of income smoothing suggested in the literature (Leuz et al. 2003; Bao and Bao
2004; Burgstahler et al. 2006; Gassen et al. 2006; LaFond et al. 2007; Dou et al. 2012):
15
However, we find a positive association when we measure risk as the variance of cash flows. We do not adopt
this risk measure due to the mechanical relation to the income-smoothing variable.
23
•
SD(net income i,t )
•
SD(operating income i,t / total assets i,t 1 )
•
  TACCi,t ; oCFi,t  ( 1) and
•
SD(operating income i,t )
SD(sales i,t )
(  1) ,
SD(sales i,t / total assets i,t 1 )
SD(oCFi,t )
(  1) ,
 (  1)
where SD, TACC and oCF stand for standard deviation, total accruals and cash flow from operations
respectively.
More recent research suggests distinguishing between the “normal” and the “discretionary” level of
income smoothing where only the latter is considered to be subject to management choice and to be
informative (Barth et al. 2012). The normal level is determined by industry and year effects. We
estimated discretionary smoothing by the residual of the following regression:
3 ,
,
where INDi and YEARt reflect industry and year dummies respectively. We ran regressions
considering discretionary income smoothing as a dependent variable and obtain similar qualitative
results (see Table 10, Panel A): corporations show significantly higher levels of income smoothing. We performed several additional robustness tests (results not tabulated).16 We accounted for an
income tax reform in 2002, which does not change the qualitative results. We defined the tax
avoidance variable differently by changing the threshold number of shareholders, which separates
firms with high tax avoidance incentives from other firms. When we set the threshold level to five
times the threshold of individual income that triggers a flat tax rate, both LIM and CONTRACTDEBT
show a significantly positive association with income smoothing. However, the TAX variable then
16
We also controlled for whether the firm is registered in West or East Germany. After the fall of the Berlin
Wall, both the economic environment and tax incentives remained different in East Germany. It transpires that
the qualitative results are not affected by the region in which the firm is registered.
24
becomes insignificant, which indicates that the initial definition of TAX is better able to discriminate
weak and strong tax avoidance incentives.
We defined BANK and CONTRACTDEBT as a dummy variable with a value of 1 if bank debt
(contractual debt, respectively) exceeds the industry-year median and zero otherwise. We also ran a
regression where CONTRACTDEBT does not include bank debt, resulting in a strongly negative
correlation with BANK (−0.51). With either specification, the LIM variable remains highly significant
both in economic and statistical terms (coefficients: 0.0925 and 0.0941, p-values: < 0.001). With
CONTRACTDEBT2 not including bank debt, the BANK variable becomes highly significant as
indicated above.
Further, we defined unlimited liability firms in a different way. There is anecdotal evidence that
owners of small corporations who receive most of their financing from one bank are often asked to
provide personal unlimited guarantees. This scenario is similar, albeit not identical, to the case of an
unlimited liability firm. We ran the regressions by adding very small corporations to unlimited liability
firms. We define very small firms as a subgroup of SMEs which either have total assets below €1
million or sales not exceeding €1.4 million. With this definition the qualitative results remain the same
and the four main variables are highly significant.
4.4
Endogeneity
Our main variable LIM is likely to be endogenous. The owners choose the legal status of the firm
depending on many criteria such as
● the owners’ risk or loss aversion
● the expected tax burden,
● owners’ preferences for hiring a manager,
● dependence on external (debt) financing sources,
● disclosure requirements,
● and legal restrictions on dividend pay-outs.
25
Unfortunately, we are unable to observe owners’ risk aversion, which is probably an important motive
for choosing limited liability. We cannot observe preferences for hiring a manager either. Disclosure
requirements are very similar across the firms in our sample because we focus on smaller corporations.
We already account for tax incentives and sources of financing in the structural equation such that we
cannot select these variables as instruments.
Thus, we looked for suitable instrumental variables that are correlated with the LIM variable but
supposedly uncorrelated with the error term in the structural equation. We chose the following
variables: COLLATERAL and LIQUIDITY.
COLLATERAL is defined as the account receivables deflated by lagged total assets averaged over
five years. Account receivables are assets that are often assigned to banks or other creditors in order to
secure a loan. We believe that a firm with more potential collateral has less need to set up an unlimited
liability firm in order to signal creditworthiness. The argument is similar for another measure of
default risk that relates to liquidity. LIQUIDITY is cash in hand, cash in banks and short-term
securities deflated by lagged total assets averaged over five years. The correlation between LIM and
COLLATERAL (LIQUIDITY) is 21.01% (14.16%). The correlation with the income-smoothing
variable is considerably weaker (4.44% and -6.65% respectively) than between LIM and SMTH
(16.6%). However, a remaining non-negligible correlation with the structural error cannot be ruled
out.
--insert Table 8 about here--
We used COLLATERAL and LIQUIDITY as instrumental variables. Due to missing data, we
computed the OLS and the 2SLS regressions based on a smaller sample size. It transpires that the
Hausman test strongly rejects the exogeneity of LIM with either instrument.17 However, the Hausman
test alone is not sufficient to conclude that the 2SLS estimates are preferable to the OLS estimates
17
Hausman test: χ2 = 16.76 and = 52.07 with COLLATERAL and LIQUIDITY respectively.
26
(Larcker and Rusticus 2010: 201). While the partial F-statistic of the first-stage regression is
encouraging with either instrument, the partial R2 is quite low, namely 1.64% with COLLATERAL
and 1.52% with LIQUIDITY (see Table 8). The low partial R2 is indicative of weak instruments.
When the instruments have low explanatory power at the first stage, estimates can become highly
biased and “the estimated coefficients on the instrumented variable will become unreasonably large or
small in the second stage” (Larcker and Rusticus 2010: 197). Indeed, the coefficient of the LIM
variable increases by about 188% (about 341%) when using COLLATERAL (LIQUIDITY) as an
instrument.18 This indicates that the IV estimates are not reliable enough to replace the OLS estimates
(Larcker and Rusticus 2010: 197).19 Note also that the standard error of the LIM variable increases
from 0.8% with the OLS regression to 6.3% and 7.4% with the 2SLS regressions. Moreover, the
variable SIZE loses significance in 2SLS regressions even though size is considered the most
important driver of the firm’s accounting choices in the empirical accounting literature. Even though
LIM and CONTRACTDEBT remain highly significant with 2SLS regressions, the instruments may
have too weak an explanatory power for the results to be reliable.20 In summary, the above results do
not suggest that IV estimates are clearly more reliable than OLS estimates.
18
For instance, the IV estimation with LIQUIDITY implies that corporations show an average income
smoothing level of -0.4772 + 0.4099 = -0.0673, which clearly exceeds the 75% quartile (-0.1481).
19
The same arguments apply when using both COLLATERAL and LIQUIDITY as instrumental variables. We
also tested for variations of these instruments.
20
Another reason for endogeneity is omitted variables, such as the owner’s level of risk or loss aversion, which
probably affects the level of income smoothing. A firm-fixed model or a first-difference analysis are suitable for
accounting for unobservable omitted variables based on firm-specific heterogeneity. However, fixed effects
estimation removes all cross-sectional variation between the firms such that it requires sufficient over-time
changes within the firm (Nikolaev and van Lent 2005: 692). Unfortunately, two important variables (LIM and
TAX) and LOSS are defined as dummy variables, leading to too little firm-specific variation.
27
4.5 Multivariate analysis with interaction terms: main regression
So far, we have found support for hypothesis 1. However, we have not yet addressed the question
whether the association between income smoothing and bank financing/debt financing and between
income smoothing and tax avoidance incentives are driven by the firm’s liability status (hypotheses 2
and 3). Consequently, we introduce an interaction term with these variables. The regression is as
follows:
4 ,
∗
,
,
,
,
,
∗
,
,
,
,
,
,
∗
,
,
,
,
,
,
In order not to inflate variance inflation factors we interact the variable on liability status (LIM) with
either BANK debt or CONTRACTDEBT, but not with both simultaneously. Table 9 depicts the
results of the regression and contrasts them with the findings without interaction terms (see Table 7).
--insert Table 9 about here--
Table 9 reveals why corporations show significantly more income smoothing than unlimited liability
firms: bank financing. A corporation with a bank debt ratio of 30% (which is close to the mean of
29%) exhibits a level of income smoothing that is about 0.019 (30%*0.0628) units higher than with an
unlimited liability firm with the same bank debt ratio. This is about 5% of the mean (absolute) income
smoothing level of 0.36.
Table 9 also suggests that the positive association between bank debt and income smoothing only
refers to corporations and not to unlimited liability firms, which supports hypothesis 2. With debt
financing, however, it is different. Here we find more income smoothing with higher debt levels
regardless of the liability status. One explanation for these surprising results could be that the
information aspect of income smoothing matters to all creditors, but the contracting role is only
28
important for banks. For banks it is relatively cheap to write and enforce (accounting) covenants and
they obviously do so for corporations where agency problems of debt are thought to be more severe.
Another finding is that stronger tax avoidance incentives lead to more income smoothing. This
association is stronger for unlimited liability firms than for corporations, which could be explained by
stronger incentives and better opportunities for income smoothing. Owners of unlimited liability firms
receive between 90%-95% of their income from their firm, whereas owners of corporations may also
hold other investments, giving them more opportunities to save taxes. Unfortunately, we lack
individual owners’ investment data to support this claim. In addition, unlimited liability firms have
more choices and discretion in financial and tax accounting.
Overall, it is interesting to see that the association between liability status and income smoothing is
based on two countervailing effects. Consequently, we find support for hypotheses 2 and 3.
4.6 Multivariate analysis with interaction terms: robustness checks
We performed several robustness checks, which generally corroborate our findings (see Table 10).
Panel A in Table 10 shows the results with different measures of income smoothing. The LIM variable
is always highly significant (p < 0.001), suggesting that corporations show higher levels of income
smoothing than unlimited liability firms. Income smoothing of corporations increases with the
importance of bank debt. Note that with CONTRACTDEBT there is no specific effect attached to the
firm’s liability status. Income smoothing is higher with tax-avoidance incentives; the effect tends to be
weaker with corporations.
We obtain the same qualitative results when we use different econometric models, more specifically
annual cross-sectional Fama-MacBeth regressions and a random effects model (see Table 10, Panel
B). Since our important exogenous variables LIM and TAX are binary, we cannot run a firm-fixed
effects model.
29
Table 10, Panel C reports findings with different specifications on the TAX variable and on the
CONTRACTDEBT variable. When we set the threshold level to five times the threshold of individual
income that triggers a flat tax rate, the liability status (LIM) and its interaction with BANK debt
remains highly significant and positive. However, the TAX5 variable becomes insignificant and,
consequently, its interaction with LIM is not always significant even though the coefficient is
negative. The TAX variable of the basic model seems better able to discriminate weak and strong tax
avoidance incentives.
The right side of Panel C shows the regression results where CONTRACTDEBT2 does not include
bank debt, resulting in a strongly negative correlation with BANK (-0.51). Consequently, BANK
becomes highly significant with a positive sign. The LIM variable remains highly significant as do its
interactions with BANK debt and TAX avoidance incentives.
5.
Summary
We analyze income smoothing with private small and medium-sized enterprises (SMEs) in Germany
focusing on differences between unlimited liability firms and corporations. We obtain the following
results. First, corporations smooth income more than unlimited liability firms do (e.g. partnerships,
sole proprietorships). We believe this is the case for two reasons: (i) Due to owners’ limited liability,
agency problems of debt are more severe with corporations, such that there is more need for income
smoothing (ii) There is a provision under German company law that in the case of corporations
dividend payouts are tied to net income, while payouts to owners of unlimited liability firms might
exceed net income.
Second, we find there is more income smoothing with higher levels of (bank) debt. Creditors may
perceive default risk to be lower when the volatility of net income decreases. Income smoothing also
increases with bank debt, but only for corporations. We believe this is the case because in contrast to
other contractual debtholders, banks are experts in writing and enforcing covenants in debt contracts.
Banks may ask for covenants when they contract with corporations because agency problems of debt
30
are thought to be more pronounced than they are with unlimited liability firms. Corporations will
engage in income smoothing because it reduces fluctuations in net income and helps to reduce the
likelihood of covenant violations, such as interest coverage.
Third, income smoothing also increases with higher tax avoidance incentives, the effect being stronger
for unlimited liability firms than for corporations. Owners of corporations may be more likely to invest
in several firms, giving them additional ways to manage their taxable income. Moreover, unlike
unlimited liability firms, corporations are allowed to be run by professional managers, which may give
less weight to the owners’ tax issues.
It is important to note that there seem to be differences in accounting behavior between different
subgroups of private firms while the literature tends to focus on differences between public and private
firms, implicitly treating private firms as a homogeneous group. Our findings suggest that private
firms are not homogeneous and that the legal status of private firms matters for income smoothing.
Moreover, our study points out that the financial statements of private firms seem to address several,
and even conflicting, goals. Thus, one may need to carefully reconsider the costs and benefits of
regulators’ efforts to make financial reporting of private firms more informative to investors (e.g.,
IASB, 2009) when the financial statement needs also to address tax, dividend and debt contracting
issues.
Our findings relate to corporations and unlimited liability firms in Germany. Even though the basic
agency problems of limited and unlimited liability firms should not differ across countries, they are
likely to be affected by the country-specific regulatory environment as well as corporate governance
and financing patterns. We need to leave this issue for future research.
31
References
Abdel-Khalik, A.R. (2007). An empirical analysis of CEO risk aversion and the propensity to smooth
earnings volatility. Journal of Accounting, Auditing, and Finance, 22 (2), 201-235.
Agarwal, R., & Elston, J. A. (2001). Bank-firm relationships, financing and firm performance in
Germany. Economics Letters, 72, 225-232.
Amiran, D., & Owens, E. (2012). Private benefits extraction and the opposing effects of income
smoothing on private debt contracts. Working paper, New York University, University of Rochester,
April 2012.
Ball, R., & Shivakumar, L. (2005). Earnings quality in UK private firms. comparative loss recognition
timeliness. Journal of Accounting & Economics, 39, 83-128.
Bao, B.-H., & Bao, D.-H (2004). Income smoothing, earnings quality and firm valuation. Journal of
Business Finance and Accounting, 31, 1525-1557.
Barth, M. E., Landsman, W. R., Lang, M., & Williams, C. (2012). Are IFRS-based and US GAAPbased accounting amounts comparable?. Journal of Accounting and Economics, 54, 68-93.
Berger, A. N., & Udell, G. F. (1995). Relationship lending and lines of credit in small firm finance.
Journal of Business, 68, 351-381.
Berger, A. N., & Udell, G. F. (1998). The economics of small business finance: the roles of private
equity and debt markets in the financial growth cycle. Journal of Banking & Finance, 22, 613-673.
BIS Department for Business Innovation & Skills (2011). Business population estimates for the UK
and Regions 2011. Sheffield.
Boot, A. W. A. (2000). Relationship banking: what do we know?. Journal of Financial
Intermediation, 9, 7-22.
Bowen, R. M., Rajgopal, S., & Venkatachalam, M. (2008). Accounting discretion, corporate
governance and firm performance. Contemporary Accounting Research, 25 (2), 351-405 .
Burgstahler, D., Hail, L., & Leuz, C. (2006). The importance of reporting incentives. earnings
management in European private and public firms. Accounting Review, 81, 983-1016.
Chen, L, H. (2012). Income smoothing, information uncertainty, stock returns, and cost of equity.
Working paper, Washington State University, October 1, 2012.
Christensen, H. B., & Nikolaev, V. V. (2012). Capital versus performance covenants in debt contracts.
Journal of Accounting Research, 50, 75-116.
32
Coppens, L., & Peek, E. (2005). An analysis of earning management by European private firms.
Journal of International Accounting, Auditing & Taxation, 112, 32-53.
Daske, H., Gebhardt, G., & McLeay, S. (2006). The distribution of earnings relative to targets in the
European Union. Accounting and Business Research, 36, 137-167.
Dechow, P. M., & Dichev, I. D. (2002). The Quality of Accruals and Earnings: The Role of Accrual
Estimation Errors. Accounting Review, 77, 35-59.
DeFond, M. L., & Park, C. W. (1997). Smoothing income in anticipation of future earnings. Journal of
Accounting and Economics, 23, 115-139.
Dichev, I. D., & Skinner, D. J. (2002). Large-Sample Evidence on the Debt Covenant Hypothesis.
Journal of Accounting Research, 40, 1091-1123.
Dou, Y., Hope, O.-K., & Wayne, T. (2012). Relationship-specificity, contract enforceability, and
income smoothing. Accounting Review, forthcoming.
Elsas, R. (2005). Empirical determinants of relationship lending. Journal of Financial Intermediation,
14, 32-57.
Elsas, R., & Krahnen, J.-P. (2004). Universal banks and relationships with firms. In J. P. Krahnen, R.
H. Schmidt (Ed.), The German Financial System (pp. 197-232). Oxford: Oxford University Press.
European Commission (2003). Commission Recommendation of May 2003 concerning the definition
of micro, small, and medium-sized enterprises (203/361/EC). Official Journal L 124 (20.05.2003).
European Commission (1996). Commission Recommendation of 3 April 1996 concerning the
definition of small and medium-sized enterprises (96/280/EC). Official Journal L 107 (30.04.1996).
Francis, J., LaFond, R., Olsson, P. M., & Schipper, K. (2004). Costs of equity and earnings attributes,
Accounting Review, 79, 967-1010.
Franks, J., & Mayer, C. (1998). Bank control, takeovers and corporate governance in Germany.
Journal of Banking and Finance, 22, 1385-1403.
Garrod, N., Kosi, U., & Valentincic, A. (2008). Asset write-offs in the absence of agency problems.
Journal of Business Finance and Accounting, 35, 307-330.
Gassen, J., Fülbier, R. U., & Sellhorn, T. (2006). International differences in conditional conservatism
– The role of unconditional conservatism and income smoothing. European Accounting Review, 15,
527-564.
33
Gassen, J., & Fülbier, R. U. (2010). The contracting role of income smoothing. evidence of European
private firms, Working paper, Humboldt University Berlin and University of Bayreuth, November
2010.
Gietzmann, M. B., & Quick, R. (1998). Capping auditor liability. the German experience. Accounting,
Organizations & Society, 23, 81-103.
Goncharov, I., & Zimmermann, J. (2006). Earnings management when incentives compete: the role of
tax accounting in Russia. Journal of International Accounting Research, 5, 41-65.
Goncharov, I., & Zimmermann, J. (2007). The supply and demand for accounting information – The
case of bank financing in Russia. Economics of Transition, 15, 257-283.
Grant, J., Markarian, G., & Parbonetti, A. (2009). CEO risk-related incentives and income smoothing.
Contemporary Accounting Research, 26 (4), 1029-1065.
Gu, Z., & Zhao, J. Y. (2006). Accruals, income smoothing and bond ratings. Working paper, Carnegie
Mellon University, 2006.
Holt, C. A., & Laury, S. K. (2002). Risk aversion and incentive effects. American Economic Review,
97, 1644–1655.
Huang, P., Zhang, Y., Deis, D. R., & Moffitt, J. S. (2009). Do artificial income smoothing and real
income smoothing contribute to firm value equivalently?. Journal of Banking and Finance, 33, 224233.
IASB (International Accounting Standards Board) (2009). IFRS for small and medium-sized entities,
London, 2009.
Jayaraman, S. (2008). Earnings volatility, cash flow volatility, and informed trading. Journal of
Accounting Research, 46 (4), 809-851.
LaFond, R., Lang, M., & Skaife, H. A. (2007). Earnings smoothing, governance and liquidity.
international evidence, Working paper, MIT/University of North Carolina/ University of Wisconsin,
March 2007.
Larcker, D. F., & Rusticus, T. O. (2010). On the use of instrumental variables in accounting research.
Journal of Accounting and Economics, 49, 186-205.
Leuz, C., Nanda, D., & Wysocki, P. D. (2003). Earnings management and investor protection: an
international comparison. Journal of Financial Economics, 69, 505-527.
Mach, T. L., & Wolken, J. D. (2006). Financial services used by small businesses: evidence from the
2003 Survey of Small Business Finances. Federal Reserve Bulletin, October 2006, A167-A195.
34
McNichols, M. F. (2000). Research design issues in earnings management studies. Journal of
Accounting and Public Policy, 19, 313-345.
Nikolaev, V. V. (2010). Debt covenants and accounting conservatism. Journal of Accounting
Research, 48, 137-175.
Nikolaev, V. V., & Lent, L. van (2005). The endogeneity bias in the relation between cost of debt
capital and corporate disclosure policy. European Accounting Review, 14, 677-724.
Peek, E., Cuijpers, R., & Buijink, W. (2010). Creditors’ and shareholders’ reporting demands in public
versus private firms: evidence from Europe. Contemporary Accounting Research, 27, 49-91.
Ronen, J., & Yaari, V. (2008). Earnings Management. New York: Springer.
Soderstrom, N. S., Jong, B., & Yang, Y. S. (2012). Earnings smoothing activities of firms to manage
credit ratings. Contemporary Accounting Research, forthcoming.
Statistisches Bundesamt (2009). Unsamtzsteuerstatistik, Wiesbaden.
Statistisches
Bundesamt
(2012).
Lohn-
und
Einkommensteuer
–
Statistik
über
die
Personengesellschaften / Gemeinschaften, Wiesbaden.
Trueman, B., & Titman, S. (1988). An explanation for accounting income smoothing. Journal of
Accounting Research, 26, 127-139.
Tucker, J. W., & Zarowin, P. A. (2006). Does income smoothing improve earnings informativeness?.
Accounting Review, 81, 251-270.
Van Tendeloo, B., & Vanstraelen, B. (2008). Earnings management and audit quality in Europe.
evidence from the private client segment market. European Accounting Review, 17, 447-469.
35
Table 1: Sample selection
Data selection
Firm-year
observations
SME
183,497
Number discounted due to missing variables & certain industries
(e.g. financial institutes)
(16,449)
Number discounted due to implausible values
(31,366)
Number discounted due to non-identifiable legal status
(4,872)
Number discounted due to firm-years with bonds outstanding
(1,251)
129,559
Number discounted due to missing values over three years and
mixed legal status
Final sample (all variables defined over three years)
Number discounted due to missing values over five years and
mixed legal status
(98,317)
31,242
(111,577)
Final sample (all variables defined over five years)
17,982
Number of unique firms
6,354
36
Table 2: Observations per year and legal status
Year
Firm-year
observations
Firm-year observations
for LIM=1
(corporations)
Firm-year observations for
LIM=0
(unlimited liability firms)
1996
3,239
1,879
1,360
1997
2,915
1,792
1,123
1998
2,357
1,503
854
1999
2,051
1,338
713
2000
2,030
1,382
648
2001
1,780
1,233
547
2002
1,513
1,030
483
2003
1,227
811
416
2004
870
563
307
Total
17,982
11,531
6,451
37
Table 3: Definition of independent variables
LIM
Dummy variable: 1, if limited liability firm in year t and the four preceding
years, 0 if unlimited liability firm in year t and the four preceding years.
Other firms have been excluded from the sample.
BANK
Ratio of bank debt to total assets, averaged over five years.
TAX
Dummy variable: 1, if net firm income before tax in year t and in the four
preceding years is less than three times the threshold individual income
that triggers a flat tax rate, 0 if net firm income before tax in year t and in
the four preceding years equals or exceeds three times the threshold
individual income that triggers a flat tax rate. Other firms have been
excluded from the sample.
CONTRACTDEBT Ratio of contractual liabilities (= total liabilities minus provisions) to total
assets, averaged over five years.
SIZE
Ln (total assets), averaged over five years.
ROA
Ratio of EBIT (without extraordinary items) to lagged total assets,
averaged over five years.
LOSS
Dummy variable: 1, if there was a loss in t or the four preceding years; 0
otherwise.
GROWTH
Annual percentage change in revenue, averaged over five years.
OPCYCLE
Length of operating cycle in days according to Burgstahler et al. (2006):
(yearly average accounts receivable)/(total revenue/360) + (yearly average
inventory)/(cost of goods sold/360), averaged over five years.
RISKSALES
Standard deviation of (sales/lagged total assets), computed over five years.
38
Table 4: Mean levels of income smoothing with limited and unlimited liability firms, with strong and
weak tax avoidance incentives and with and without total bank debt exceeding the mean
Group
Mean (income smoothing) Mean (income smoothing)
(Group Variable = 1)
(Group Variable = 0)
Difference
(t-value)
LIM
− 0.3244
(obs.: 11,531)
− 0.4310
(obs.: 6,451)
0.1066***
(21.25)
D_BANK
− 0.3393
(obs.: 8,972)
− 0.3858
(obs.: 9,010)
0.0465***
(10.16)
TAX
− 0.3443
(obs.: 15,635)
− 0.4850
(obs.: 2,347)
0.1407***
(19.25)
LIM is a dummy variable with value 1 (0) for corporations (unlimited liability firms). TAX is a dummy variable
with value 1 for firms with high tax avoidance incentives and with value 0 for low tax avoidance incentives.
D_BANK is a dummy variable with value 1 if bank debt to total assets (averaged over five years) exceeds the
industry median; otherwise it is 0. Two-sample t-test with unequal variances.
39
Table 5: Summary statistics (N = 17,982)
Variable
Mean
Standard
deviation
25%quartile
Median
75%quartile
SMTH
-0.3626
0.3079
-0.4763
-0.2708
-0.1481
LIM
0.6412
0.4796
0
1
1
CONTRACTDEBT
0.6865
0.2038
0.5697
0.7353
0.8460
BANK
0.2922
0.2154
0.1022
0.2734
0.4505
TAX
0.8695
0.3369
1
1
1
SIZE: ln(total assets
in € 1,000)
7.2678
1.0203
6.5980
7.2351
7.8552
SIZE: total assets in
€ 1,000
1,433.4
733.6
1,387.3
2,579.1
ROA
0.1042
0.0978
0.0472
0.0749
0.1261
LOSS
0.4115
0.4921
0
0
1
GROWTH
0.0063
0.0901
-0.0489
0.0032
0.0518
OPCYCLE
106.64
85.99
58.73
83.04
124.87
RISKSALES
0.5090
0.4765
0.2129
0.3735
0.6337
For a definition of variables see Table 3.
40
Table 6: Pearson correlations (p-values in brackets)
pred.
sign
SMTH
SMTH
LIM
CONTRACT
DEBT
BANK
TAX
SIZE
ROA
LOSS
GROWTH
OPCY-CLE
RISK
SALES
1
LIM
+
0.1661
(0.0000)***
1
CONTRACT
DEBT
+
0.2260
(0.0000)***
0.0023
1
BANK
+
0.0968
(0.0000)***
-0.1817***
0.5631***
1
TAX
+
0.1540
(0.0000)***
0.1332***
0.2935***
0.1661***
1
SIZE
?
0.0927
(0.0000)***
0.0980***
0.0986***
0.1544***
-0.4839***
1
ROA
-
-0.2913
(0.0000)***
-0.2456***
-0.3629***
-0.2131***
-0.4970***
-0.1083***
1
LOSS
?
-0.0546
(0.0000)***
0.2456***
0.1895***
0.1450***
0.3240***
-0.0389***
-0.4729***
1
GROWTH
?
0.0469
(0.0000)***
0.1182***
0.0403***
-0.0280***
-0.0876***
0.0970***
0.1251***
-0.1207***
1
OPCYCLE
+
0.0153
(0.0405)**
-0.1149***
0.1453***
0.0975***
0.0878***
0.0210***
-0.1513***
0.0633***
-0.1154***
1
RISK SALES
+
0.0107
(0.1532)
0.1468***
-0.0218***
-0.1661***
0.0535***
-0.2282***
0.1347***
-0.0033
0.0986***
-0.2876***
*, **, and *** indicate significance at the 10%, 5%, and 1% levels, using a two-tailed test. For a definition of variables see Table 3.
1
41
Table 7: Pooled OLS regressions, dependent variable: income smoothing (N=17,982)
Dependent
variable:
SMTH
Predicted
sign
Intercept
SMTH (5 years)
SMTH (3 years)
SMTH (5 years)
(annual cross-sectional
Fama-MacBeth
regressions)
DISCRET_SMTH
(5years)
Coeff.
t-value
(p-value)
Coeff.
t-value
(p-value)
Coeff.
t-value
Coeff.
t-value
(p-value)
-0.5372
-11.54
(0.000)***
-0.6240
-13.09
(0.000)***
-0,5335
-16.80***
-0.1580
-3.47
(0.001)***
LIM
+
0.0939
11.84
(0.000)***
0.1273
14.14
(0.000)***
0.0910
15.45***
0.0952
12.00
(0.000)***
BANK
+
0.0102
0.56
(0.576)
-0.0175
-0.85
(0.395)
0.0149
1.32
0.0208
1.16
(0.245)
TAX
+
0.0459
3.11
(0.002)***
0.0290
1.90
(0.057)*
0.0462
6.32***
0.0407
2.74
(0.006)***
CONTRACT
DEBT
+
0.1875
8.30
(0.000)***
0.1977
8.15
(0.000)***
0.1852
15.63***
0.1651
7.48
(0.000)***
SIZE
+
0.0153
3.35
(0.001)***
0.0182
3.80
(0.000)***
0.0165
8.39***
0.0155
3.38
(0.001)***
ROA
-
-0.9778
-16.55
(0.000)***
-0.9266
-17.22
(0.000)***
-0.9702
-18.67***
-0.9720
-16.54
(0.000)***
LOSS
?
-0.1675
-23.25
(0.000)***
-0.2681
-30.31
(0.000)***
-0.1712
-25.93***
-0.1670
-23.02
(0.000)***
GROWTH
+
0.2139
5.32
(0.000)***
0.1874
5.61
(0.000)***
0.1963
6.10***
0.1601
5.08
(0.000)***
OPCYCLE
?
-0.00004
-0.96
(0.335)
-0.00007
-1.34
(0.181)
-0.00009
-1.36
-0.00005
-1.07
(0.283)
RISKSALES
+
0.0062
0.79
(0.428)
-0.0089
-0.97
(0.334)
-0.0104
-1.37
0.0073
1.02
(0.309)
Industry fixed
effects
included
included
included
--
Year fixed
effects
included
included
--
--
17,982
31,242
17,982
17,982
VIF (LIM)
1.31
1.28
--
1.28
Mean/max
VIF
2.03/4.26
2.01/4.06
--
1.58/2.36
Adj. R2 in %
18.22
8.78
18.81
16.67
F-Stat.
61.50
77.91
--
122.64
Prob(F-Stat.)
0.0000
0.0000
--
0.0000
N=
*, **, and *** indicate significance at the 10%, 5%, and 1% levels, using a two-tailed test. T-statistics are based
on standard errors, which are adjusted for heteroscedasticity and clustering at the firm level. VIF: variance
inflation factor. DISCRET_SMTH: Discretionary smoothing is estimated by the residual of the regression
SMTHi,t = β0 + INDi + YEARt + εi,t where INDi and YEARt are industry and year dummy variables respectively.
For a definition of variables see Table 3. With SMTH (3 years), all variables are defined on a three-years basis.
42
Table 8: OLS and 2SLS regressions, dependent variable: income smoothing
Dependent
variable: SMTH
Predicted
sign
SMTH (5 years), OLS
Coeff.
Intercept
t-value
(p-value)
SMTH (5 years), 2SLS
IV: Collateral
Coeff.
z-value
(p-value)
SMTH (5 years), 2SLS
IV: Liquidity
Coeff.
z-value
(p-value)
-0.5349
-11.22
(0.000)***
-0.5031
-9.91
(0.000)***
-0.4772
-8.56
(0.000)***
LIM
+
0.0929
11.60
(0.000)***
0.2676
4.23
(0.000)***
0.4099
5.52
(0.000)***
BANK
+
0.0061
0.33
(0.742)
0.1133
2.60
(0.009)***
0.2005
3.98
(0.000)***
TAX
+
0.0452
2.99
(0.003)***
0.0052
0.24
(0.809)
-0.0274
-1.08
(0.280)
CONTRACTDEBT
+
0.1867
8.14
(0.000)***
0.1719
6.99
(0.000)***
0.1599
5.82
(0.000)***
SIZE
+
0.0157
3.36
(0.001)***
-0.0022
-0.27
(0.787)
-0.0167
-1.84
(0.066)*
ROA
-
-0.9854
-16.33
(0.000)***
-0.8499
-10.89
(0.000)***
-0.7395
-8.64
(0.000)***
LOSS
?
-0.1675
-22.80
(0.000)***
-0.1983
-14.55
(0.000)***
-0.2234
-14.63
(0.000)***
GROWTH
+
0.2273
5.52
(0.000)***
0.1349
2.51
(0.012)**
0.0597
0.99
(0.322)
OPCYCLE
?
-0.00005
-0.98
(0.328)
0.00005
0.88
(0.380)
0.0001
1.95
(0.051)*
RISKSALES
+
0.0058
0.73
(0.463)
-0.0203
-1.64
(0.101)
-0.0416
-3.00
(0.003)***
Industry fixed
effects
included
included
included
Year fixed effects
included
included
included
17,406
17,406
17,406
First stage: partial
R2 in %
1.64
1.52
First stage: partial
F-Stat.
73.37
79.94
N=
*, **, and *** indicate significance at the 10%, 5%, and 1% level, using a two-tailed test. T-statistics are
standard errors, which are adjusted for heteroscedasticity and clustering at the firm level. Due to missing data on
instrumental variables, we computed the 2SLS regressions on a smaller sample size. IV: Instrumental variable.
Collateral: Account receivables deflated by lagged total assets averaged over five years. Liquidity: cash in hand,
cash in banks and short-term securities deflated by lagged total assets averaged over five years. For a definition
of other variables see Table 3.
43
Table 9: Pooled OLS regressions with interaction terms
Dependent
variable: SMTH
SMTH (5 years)
SMTH (5 years)
SMTH (5 years)
SMTH (5 years)
Coeff.
t-value
(p-value)
Coeff.
t-value
(p-value)
Coeff.
t-value
(p-value)
Coeff.
t-value
(p-value)
Intercept
-0.5372
-11.54
(0.000)***
-0.5231
-11.08
(0.000)***
-0.5353
-11.33
(0.000)***
-0.5518
-11.12
(0.000)***
LIM
0.0939
11.84
(0.000)***
0.0739
5.34
(0.000)***
0.1150
5.09
(0.000)***
0.1339
4.26
(0.000)***
BANK
0.0102
0.56
(0.576)
-0.0294
-0.99
(0.321)
-0.0439
-1.50
(0.135)
0.0080
0.44
(0.658)
TAX
0.0459
3.11
(0.002)***
0.0456
3.09
(0.002)***
0.0694
3.90
(0.000)***
0.0650
3.65
(0.000)***
CONTRACTDEBT
0.1875
8.30
(0.000)***
0.1890
8.36
(0.000)***
0.1960
8.64
(0.000)***
0.1951
5.88
(0.000)***
0.0628
1.91
(0.056)*
0.0820
2.50
(0.012)**
-0.0562
-2.53
(0.011)**
-0.0445
-2.00
(0.046)**
-0.0037
-0.10
(0.923)
LIM*BANK
LIM*TAX
LIM*CONTRACT
DEBT
SIZE
0.0153
3.35
(0.001)***
0.0153
3.35
(0.001)***
0.0145
3.16
(0.002)***
0.0147
3.20
(0.001)***
ROA
-0.9778
-16.55
(0.000)***
-0.9851
-16.54
(0.000)***
-1.0015
-16.77
(0.000)***
-0.9888
-16.64
(0.000)***
LOSS
-0.1675
-23.25
(0.000)***
-0.1682
-23.23
(0.000)***
-0.1673
-23.18
(0.000)***
-0.1666
-22.93
(0.000)***
GROWTH
0.2139
5.32
(0.000)***
0.2120
5.28
(0.000)***
0.2120
5.28
(0.000)***
0.2145
5.34
(0.000)***
OPCYCLE
-0.00004
-0.96
(0.335)
-0.00004
-0.97
(0.333)
-0.00005
-1.09
(0.275)
-0.00005
-1.05
(0.293)
0.0062
0.79
(0.428)
0.0069
0.87
(0.382)
0.0079
1.00
(0.317)
0.0069
0.87
(0.382)
RISKSALES
Industry fixed
effects
included
included
included
included
Year fixed effects
included
included
included
included
17,982
17,982
17,982
17,982
1.31
3.54
8.37
15.31
mean/max VIF
2.03/4.26
2.29/4.26
2.82/9.57
3.56/15.58
2
Adj. R in %
18.22
18.25
18.33
18.26
F-Stat.
61.50
58.39
56.81
57.77
Prob(F-Stat.)
0.0000
0.0000
0.0000
0.0000
N=
VIF (LIM)
*, **, and *** indicate significance at the 10%, 5%, and 1% levels, using a two-tailed test. T-statistics are based
on standard errors, which are adjusted for heteroscedasticity and clustering at the firm level. VIF: variance
inflation factor. For a definition of other variables see Table 3.
44
Table 10: Robustness tests with interaction terms
Panel A: Different specifications of the income smoothing variable
Dependent
variable: SMTH
DISCRET_SMTH (5 years), OLS
SMTH (3 years), OLS
Coeff.
(p-value)
Coeff.
(p-value)
Coeff.
(p-value)
Coeff.
(p-value)
Coeff.
(p-value)
Coeff.
(p-value)
Coeff.
(p-value)
Coeff.
(p-value)
0.0952
(0.000)***
0.0755
(0.000)***
0.1176
(0.000)***
0.1364
(0.000)***
0.1273
(0.000)***
0.1031
(0.000)***
0.1252
(0.000)***
0.1265
(0.000)***
0.0208
(0.245)
-0.0182
(0.535)
-0.0334
(0.353)
0.0184
(0.301)
-0.0175
(0.395)
-0.0662
(0.053)*
-0.0739
(0.032)**
-0.0173
(0.402)
TAX
0.0407
(0.006)***
0.0403
(0.007)***
0.0648
(0.000)***
0.0603
(0.001)***
0.0290
(0.057)*
0.0279
(0.067)*
0.0438
(0.029)**
0.0419
(0.039)**
CONTRACTDEBT
0.1651
(0.000)***
0.1664
(0.000)***
0.1738
(0.000)***
0.1731
(0.000)***
0.1977
(0.000)***
0.1995
(0.000)***
0.2030
(0.000)***
0.1790
(0.000)***
0.0620
(0.059)*
0.0816
(0.013)**
0.0747
(0.044)**
0.0845
(0.024)**
LIM
BANK
LIM*BANK
LIM*TAX
-0.0575
(0.010)***
LIM*CONTRACT
DEBT
-0.0460
(0.040)**
-0.0308
(0.168)
-0.0038
(0.920)
Control variables
0.0318
(0.444)
included
included
Industry fixed
effects
--
included
Year fixed effects
--
included
17,982
31,242
N=
VIF (LIM)
-0.0255
(0.263)
1.28
3.49
8.31
15.03
1.28
3.42
7.50
14.06
1.58/2.36
2.17/4.13
3.29/9.55
4.80/15.34
2.01/4.06
2.27/4.20
2.75/8.68
3.43/14.50
Adj. R2 in %
16.63
16.67
16.75
16.68
8.78
8.80
8.80
8.78
F-Stat.
122.64
111.25
103.27
103.51
77.91
74.95
72.10
72.21
Prob(F-Stat.)
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
mean/max VIF
*, **, and *** indicate significance at the 10%, 5%, and 1% levels, using a two-tailed test. T-statistics are based
on standard errors, which are adjusted for heteroscedasticity and clustering at the firm level. Discretionary
smoothing is estimated by the residual of the regression SMTHi,t = β0 + INDi + YEARt + εi,t where INDi and
YEARt are industry and year dummy variables respectively. For a definition of other variables see Table 3.
45
Panel B: Fama-MacBeth regressions and random effects model
Dependent
variable: SMTH
SMTH (5 years) (Random effects)
SMTH (5 years), (annual cross-sectional Fama-MacBeth
regressions)
Coeff.
(z-value)
Coeff.
(z-value)
Coeff.
(z-value)
Coeff.
(z-value)
Coeff.
(t-value)
Coeff.
(t-value)
Coeff.
(t-value)
Coeff.
(t-value)
0.1137
(0.000)***
0.0937
(0.000)***
0.1336
(0.000)***
0.1494
(0.000)***
0.0910
(15.45)***
0.0724
(7.18)***
0.1046
(8.63)***
0.125
(6.08)***
0.0044
(0.809)
-0.0345
(0.228)
-0.0470
(0.103)
0.0026
(0.886)
0.0149
(1.32)
-0.0215
(-0.87)
-0.0323
(-1.22)
0.0130
(1.21)
TAX
0.0635
(0.000)***
0.0632
(0.000)***
0.0880
(0.000)***
0.0838
(0.000)***
0.0462
(6.32)***
0.0461
(6.29)***
0.0633
(7.40)***
0.0598
(6.70)***
CONTRACTDEBT
0.1718
(0.000)***
0.1733
(0.000)***
0.1789
(0.000)***
0.1750
(0.000)***
0.1852
(15.63)***
0.1867
(15.95)***
0.1919
(16.16)***
0.1976
(6.98)***
0.0623
(0.051)*
0.0787
(0.015)**
0.0563
(2.36)**
0.0714
(2.66)***
LIM
BANK
LIM*BANK
LIM*TAX
-0.0537
(0.008)***
LIM*CONTRACT
DEBT
-0.0438
(0.035)**
-0.0438
(-3.38)***
0.0016
(0.967)
Control variables
-0,0126
(-0,39)
included
included
Industry fixed
effects
--
included
Year fixed effects
--
--
17,982
17,982
N=
2
R overall in %
(adj. R2)
F-Stat. (Wald Chi2)
2
Prob(F-Stat./Chi )
17.97
18.01
18.09
18.02
1581.35
1579.96
1589.07
1590.79
0.0000
0.0000
0.0000
0.0000
18.81
18.85
18.88
*, **, and *** indicate significance at the 10%, 5%, and 1% levels, using a two-tailed test. T-statistics are based
on standard errors, which are adjusted for heteroscedasticity and clustering at the firm level. For a definition of
variables see Table 3.
-0.0331
(-2.76)***
18.85
46
Panel C: Alternative specifications of the variables TAX and CONTRACTDEBT
Dependent
variable: SMTH
SMTH (5 years), OLS with TAX5
SMTH (5 years), OLS with CONTRACTDEBT2
Coeff.
(p-value)
Coeff.
(p-value)
Coeff.
(p-value)
Coeff.
(p-value)
Coeff.
(p-value)
Coeff.
(p-value)
Coeff.
(p-value)
Coeff.
(p-value)
0.0960
(0.000)***
0.0667
(0.000)***
0.1184
(0.000)***
0.1330
(0.000)***
0.0941
(0.000)***
0.0735
(0.000)***
0.1157
(0.000)***
0.1717
(0.000)***
BANK
0.0073
(0.680)
-0.0492
(0.079)*
-0.0595
(0.033)**
0.0060
(0.737)
0.1964
(0.000)***
0.1573
(0.000)***
0.1495
(0.000)***
0.1992
(0.000)***
TAX
0.0125
(0.486)
0.0129
(0.473)
0.0410
(0.083)*
0.0338
(0.150)
0.0489
(0.001)***
0.0486
(0.001)***
0.0730
(0.000)***
0.0674
(0.000)***
0.1928
(0.000)***
0.1947
(0.000)***
0.1993
(0.000)***
0.1908
(0.000)***
0.1881
(0.000)***
0.1895
(0.000)***
0.1968
(0.000)***
0.2692
(0.000)***
0.0919
(0.003)***
0.1050
(0.001)***
0.0644
(0.049)**
0.0840
(0.010)***
LIM
CONTRACTDEBT
LIM*BANK
LIM*TAX
-0.0609
(0.046)**
LIM*CONTRACT
DEBT
-0.0463
(0.129)
-0.0577
(0.009)***
0.0081
(0.824)
-0.1135
(0.001)***
Control variables
included
included
Industry fixed
effects
included
included
Year fixed effects
included
included
20,014
18,191
N=
VIF (LIM)
-0.0429
(0.051)*
1.30
3.51
15.03
21.10
1.32
3.55
8.39
10.89
mean/max VIF
1.97/4.20
2.22/4.20
3.30/16.53
4.01/21.10
2.03/4.24
2.28/4.24
2.81/9.59
3.08/9.18
2
Adj. R in %
18.18
18.27
18.32
18.20
18.14
18.18
18.27
18.30
F-Stat.
69.64
66.80
64.82
64.74
62.24
58.90
57.35
57.93
Prob(F-Stat.)
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
*, **, and *** indicate significance at the 10%, 5%, and 1% levels, using a two-tailed test. T-statistics are based
on standard errors, which are adjusted for heteroscedasticity and clustering at the firm level. VIF: variance
inflation factor. TAX5 is a dummy variable with value 1 if net firm income before tax in year t and in the four
preceding years is less than five times the threshold individual income that triggers a flat tax rate, 0 if net firm
income before tax in year t and in the four preceding years equals or exceeds five times the threshold individual
income that triggers a flat tax rate. Other firms have been excluded from the sample. CONTRACTDEBT2:
contractual debt excluding bank debt. The correlation between CONTRACTDEBT2 and BANK is −0.512. For a
definition of other variables see Table 3.