In order to measure the effect of corporate governance practices and investor

protection on dividends, we obtained annual payout ratios for over 200 listed
companies since 1994 to 2002. The data was obtained from Fecus Plus and
complemented using Economatica. We used as an indicator of the dividend policy of
the company the ratio between dividend payments (including non cash payments) on
year t and after-tax earnings on year t-1. W e used this traditional measure, even
though many times paid dividends may come from earnings attained in different years.
In Chile, the law establishes a minimum dividend requirement of 30% of annual
earnings. The rationale for such a compensatory measure is to protect minority
shareholders as indicated in La Porta et al. (1997) and implies that Chilean controllers
have less freedom in order to determine and use their dividend policies. In theory, a
company could pay less than 30% of earnings if unanimously approved by
shareholders.3 However, in practice, a company may pay less than 30% by declaring
the dividend and postponing the payment for the upcoming years. 4 Hence, despite
the legal restriction is possible to observe effective payout ratios of less than 30% of
earnings.
Table 3 summarizes the data collected. Payout ratios in Chile were over 53%
by 1994 and steadily declined to 36% in 2002. There is wide dispersion of payout
ratios in our sample, with som e companies paying over 150% of last year earnings.
Negative ratios indicate generally, that a company paid dividends even when last year
earnings where negative. The table also shows that affiliated companies to a
conglomerate have, on average, higher payout ratios than non-affiliated firms.

4. Agency problems.
Corporate governance deals foremost with agency problems inside the firm. In
highly concentrated Chilean firms, agency problems take mainly the form of conflict of
interest between controlling shareholders and minority shareholders. In this paper, we
explore several dimensions of this relation and study their impact on firm valuation and
payout policies.
Several of the theories previously discussed maintain that agency problems
between controllers and minority shareholders are more severe in firms affiliated to

3
Corporations Law, rule 79.
4
Corporations Law, rule 84 establishes that if dividends are postponed the amount finally paid must be
adjusted by inflation and interest.

15
conglomerate structures. However, the effect of affiliation on firm valuation is not clear
as indicated by the different competing hypothesis with respect to it. For instance,
after controlling for separation of rights, affiliation to a conglomerate in emerging
economies could be value enhancing due to internal capital markets, information
sharing and other synergies. Following Lefort and Walker (2000) we identify over 50
conglomerate structures in the Chilean economy operating between 1990 and 2002
and construct a dummy variable taking the value 1 when a company is affiliated to any
of those structures in any given year.
A key indicator of the potential existence of agency problems is the degree of
separation between the cash flow rights accrued by the controller and the control
rights he or she is exercising. We measure separation from cash to control rights at
the firm level considering direct and indirect holdings of controllers and the existence
of dual class shares. Under agency theory, we hypothesize that higher separation is
associated with lower valuation and lower payout. We constructed two different
indicators of the degree of coincidence between cash and control rights, under the
assumption that the largest shareholder effectively controls all company assets.5 W e
will analyze the validity of this assumption later on the paper. First, we measure
separation as the ratio between equity directly and indirectly owned by the largest
shareholder and total consolidated assets under control of the company. The ratio
captures the proportion between the amount of cash flows accrued by the controllers
and the total amount of cash flow potentially generated by the company including debt
payments. The second measure considers only the ratio between cash flows to
controllers and cash flows to all shareholders. In both cases, the assumption of total
effective control by the controllers means that the percentage of cash flow rights is a
direct indicator of coincidence. Perfect coincidence is achieved as these variables
approach to one.
Institutional investors have had an important role in helping to develop Chilean
capital markets.6 Specifically, pension fund managers can buy shares of Chilean
companies that reach specific levels of ownership dispersion and are approved as
investing vehicles by the Risk Classification Commission. Hence, the presence of

5
Rule 67 of Corporations Law establishes that the approval of major company decisions require the
support of two thirds of voting shares during a shareholder meeting. The Rule also establishes a
mandatory tender offer requirement whenever a shareholder reaches the two thirds threshold through
an acquisition.
6
See Walker and Lefort (1999).

16
pension funds as shareholders of a company is an indication that the firm is less risky
and that its governance mechanisms are more mature. In addition, once the pension
funds reach a given level of ownership in the company, they may elect a board
member and become an active minority shareholder. Under the assumption that
important institutional investors improve governance, their presence can improve
performance. They can also be seen as the second important shareholder as in
Gurgler and Yurtoglu (2002).

5. Corporate Governance Practices in Chile

Corporate governance has many more dimensions than just the affiliation to a
conglomerate and the degree of coincidence between cash flow rights and control
rights. In order to complement those measures, we conducted a survey on corporate
governance practices at the firm level, through a 67 questions questionnaire handed in
to principal officers and board members of over 200 listed companies in Chile. The
annex at the end of the paper shows the questions and the answers per question
obtained. Table 4 summarizes the main results. The survey was conducted between
May and September, 2004. Questions were referred to firm practices as of the end of
2003.
The response rate was moderately low. W e received 59 completed
questionnaires, representing less than 30% of the firms contacted. The low response
rate was relatively expected given the type of survey we were conducting. However,
the companies that answered the questionnaire tended to be the largest in terms of
market capitalization. They account for 42% of total market capitalization in Chile.
Around one third of the questions in the survey could be directly completed using
public information available at the financial statements and Annual Book of the
companies or using information made public by the SVS (Superintendence of
Securities and Exchanges). W e have compiled information through those
mechanisms for an additional 47 firms, achieving a total coverage of 106 companies
amounting to 76% of total market capitalization in Chile.
The questionnaire was divided in four sections: (i) about general principles; (ii)
about officers and the board; (iii) about shareholders; and (iv) about disclosure and
information. Most questions could be answered by a simple yes or no. For all those
questions (55 approximately) we used and indicator variable that took the value of 1
whenever the answer could be associated with best practices and 0 otherwise. In

17
many cases, the answer was in fact a “don’t know/don’t answer”.  This was the case, 
for instance, for companies for which the questionnaires were filled using sources
other than officers or board members. In fact, one third of the questions could not be
answered using public information as detailed above. We then normalized each
answer between 0 and 7. A score of seven would correspond to questions where all
respondents got a 1 in the indicator regardless of the size of the company responding
the questionnaire. W e then averaged the results for each section. This procedure
implicitly considers that each question has the same relative importance in order to
measure the quality of a company corporate governance practices. This assumption
is not necessarily true and, hence, the average results summarized en Table 4 must
be interpreted carefully. The annex provides the precise result for each question
included in the questionnaire.
Table 4 shows that, by 2003, Chilean companies scored relatively well in their
corporate governance practices obtaining an overall score (non-weighted) of 4.12 out
of 7. Not surprisingly, the worse results were obtained in the first category: general
principles (2.63). Most Chilean companies do not have a code or a mission statement
that gives any explicit importance to governance practices. The best scores are
obtained in the category disclosure and information (5.14). Chilean companies
adequately disclose information on control, ownership and related party transactions.
The weakest aspects of information disclosure are promptness and the lack of
announced targets with respect to future performance of the company. The second
category on the questionnaire was about officers and the board. The average score
obtained was 4.54. The weakest aspect on this category was the low participation of
independent directors on company boards and the absence of special committees
such as audit and governance committees. Corporation law in most Latin American
countries explicitly indicates that boards are the main decision making body of a
company and that board members owe duty of loyalty and duty of care to all
shareholders. However, as a consequence of the high ownership concentration
observed in most firms in the region, boards in Latin American countries tend to be
much weaker than in the US or UK, and constitute a poor governance mechanism. In
general terms, boards in Latin America have mainly an advisory character to
controllers, have very few independent board members and few if any functioning
committees. Lefort and Walker (2000c) and the report prepared by Spencer Stuart-
PUC (2000) have looked preliminary to board composition and functioning in Chile

18
and reach similar conclusions regarding the shortage of truly independent directors in
Chilean corporations. In areas related to shareholders rights, Chilean companies
scored relatively well (4.18). That was the case in the applicability of the one share-
one vote rule, the general voting rights of minority shareholders and the absence of
formal sanctions applied by the SVS to board members and officers.
In order to empirically ascertain the importance of corporate governance
practices in Chilean firms’ market valuation and payout policies, we focused on a 
subset of the questionnaire. We selected 20 questions according to the following
three criteria. They had been answered directly or indirectly by most companies in the
sample, they capture a relevant feature of corporate governance practices in an
emerging economy such as a Chile, and they can be answered by a simple yes or no.
Table 5 summarizes the results for this subset of 20 questions that comprise our
simple index of corporate governance practices in Chile (CGI).
Questions on the CGI were grouped in four categories: disclosure, board
functioning and independence, shareholders rights, and conflicts of interest. The
Table presents the questions and the original number of each question in the full
length questionnaire. The CGI had a maximum score of 20. The average score of the
106 companies surveyed was 11.3, indicating an only mediocre performance. Not
surprisingly, the worst areas on the survey were board functioning and conflict of
interest. We detected a very low level of board involvement in committees. Less than
5% of the largest firms of the country have a corporate governance committee and
only 14% had compensation or nomination committees. In only 21% of the companies
the chairman of the board is and independent and non-affiliated board member. In
70% of the boards of the largest Chilean companies there are board members that are
also executives or board members of other companies of the same group, indicating a
high degree of board interlocking and lack of independence of board members. This
result is consistent with the findings on Lefort and Walker (2000) and indicates a high
likelihood of conflict of interest at the group level.
On the other hand, Chilean large firms score relatively well in disclosure and
shareholders rights. Chilean legislation and, specially, the OPA Law approved on the
year 2000 is on a large part responsible for the rigorous disclosure of related party
transactions by listed Chilean companies.

19
 V. EMPIRICAL ANALYSIS OF THE EFFECTS OF AGENCY PROBLEMS ON
FIRM’S MARKET VALUATION 

The empirical research on corporate governance boosted up after the seminal

work of LLSV during the nineties. The original research investigated whether specific
legal arrangements related to investor protection on different countries affected capital
markets development. The focus of this paper is a related question. Here we ask
whether corporate governance practices at the firm level within a single country affect
these firms’ market valuation.  This question is crucial to asses the potential benefits
for firms to change their own practices, even though, they cannot affect their country’s 
rules.  As clearly stated by Black et al (2003), “to what extent can a firm increase its 
market value by upgrading its corporate governance practices, and to what extent is it
tied to its home country’s rules and reputation?” 
As the list of empirical studies trying to asses this question increases, the
understanding of the difficulties entangled in the task improves. Even if firm level
corporate governance practices correlate with share prices, we cannot be sure that
these practices cause investors to value firms more highly. Alternative explanations
related to different forms of endogeneity and omitted variable bias are also consistent
with such empirical findings.
In this section of the paper, we perform regression analysis of measures of firm
performance and payout policy on corporate governance indicators at the firm level
and a series of control variables. Among the indicators of firm performance, we
consider firm market valuation using Tobin’s q and market to book ratios, ROA and 
the dividend-earnings ratio. The empirical model tries to capture the hypothesis
previously discussed regarding the control structure of the company, the extent of the
agency problem at the firm level and the market value of the company.
The empirical model is, therefore, of the type:

y it 1 ( dgroup it ) 2 ( concent it ) 3 ( coincid it ) ZFit 1 ZG it 2 it

Where:
y: a firm performance and value indicator such as Tobin’s q, ROA, and dividend 
payout ratio.

20
 dgroup: affiliation to a conglomerate dummy.
coincid: degree of coincidence between cash and control rights at the firm level.
concent: ownership concentration at the firm level.
ZF: a set of control variables at the firm level, including Tobin’s q in the payout 
equation, and time and industry dummies.
ZG: a set of control variables at the group level.

For estimation purposes we will consider three different samples because of

restrictions on data availability. First, we use an annual panel data comprising all
listed companies with a fair amount of trading (around 200) over a 13 years’ time 
horizon (1990-2002). On average, this data base supplies over 1800 year-firm
observations allowing obtaining robust estimates using different estimation procedures
of the relationship between agency problems and firm market valuation. Secondly, we
constructed a similar annual panel for the period 1994-2002, because no information
on dividend payments was available for the period prior to 1994. This panel provide
over 1100 year-firm observations. Finally, in next section, we also analyzed a cross
section sample for 106 large companies for the year 2003. Although, this is a smaller
data base, we use it to capture the effect of other dimensions of corporate governance
practices affecting firm valuation and payout policies and to provide estimates of the
incidence of corporate governance practices at the firm level on company valuation
that are robust to the endogeneity problem. We use the CGI (Corporate Governance
Index) and its components for that purpose.

1. Econometric concerns
Endogeneity:
A key concern in this type of studies has to do with the potential endogeneity
problem as discussed by Klapper and Love (2003) and Black et al (2003) among
others. In the context of this paper, the endogeneity problem would arise, for
instance, if firms with high market valuation tended to adopt good governance
practices in order to further improve their share prices. In that case, part of the
correlation captured in the regressions would respond in fact to reverse causation.
Black et al (2003) refers to a slightly different type of endogeneity referred as “optimal 

21
differences” which occurs when firms endogenously and optimally choose different 
governance practices in the sense of Demsetz and Lehn (1985). 7
A related problem of spurious correlation could arise due to omitted variable
bias. In equilibrium, corporate governance likely correlates with various economic
variables. A study that omits some economic variables, which predict both
governance and share price, could wrongly conclude that governance is directly
associated with share price. This problem can be described by noticing that corporate
governance practices at the firm level could be determined by the firm’s contracting 
environment. For instance, firms with more tangible assets or more growth
opportunities would want to improve corporate governance mechanisms in order to
raise external finance. In such a situation, they may decide to reduce, for instance,
separation of control and cash flow rights or transfer control to other, maybe foreign,
companies. Hence, if we do not adequately control by these variables, the
governance factors will capture the effect of the contracting environment on the firm
on its market value.
Panel data estimation:
The use of a panel data base increases the number of observations but
introduces potential biases in the estimation. In order to account for unobservable
individual effects we provide fixed and random effects estimations in addition to the
traditional pooled least squares. Moreover, we also provide GLS heteroskedasticity -
consistent estimators in case that observations of different companies present
different variances. We performed Hausman tests of specification in order to chose
the best estimations obtained.
Censored data:
Traditionally, payout ratio data are censored at cero since companies do not
pay negative dividends even if they were willing to. In addition, Chilean legislation
requires companies to pay dividends of at least 30% of last year profits. Hence, we
estimated panel Tobit regressions in the case of payout ratios due to the censored
nature of the dependent variable, and computed Hausman tests to evaluate the
importance of the censoring problem.

7
Black et al (2003) give an alternative explanation for the potential correlation: quality signaling. The
idea is that firms may adopt good governance rules to signal its good behavior. In that case, the signal
rather than the firm’s governance practices, affects share prices.   

22
2. Empirical results on agency problems and firm valuation
In the first part of the empirical analysis we want to explore the information
contained on the panel data regarding the effect of agency conflicts at the firm level on
its market valuation and payout ratios. We take the lack of coincidence between cash
flow rights and control rights as an indicator of conflict of interest and potential agency
problems and, thus, as a proxy for bad corporate governance practices at the
company level. In the next subsection, we complement our analysis by considering
indicator variables of the quality of corporate governance constructed from index
variables derived from the questionnaire previously described.
In order to construct our proxy for the potential existence of agency problems
we calculated for each company the market value of the consolidated equity held by
the controlling shareholders. We then divide this value by the market value of assets
calculated as the sum of the market value of total equity plus debt. As explained
before this ratio indicates the percentage accrued by the controllers of each dollar of
assets created by the company. We take this variable as an indication of the
coincidence between cash flow rights and control rights and we call it Coincid . We
also computed a simple measure of ownership concentration as the fraction of total
equity held by the three largest shareholders (concent).
In order to measure firm valuation we consider three variables. Following most
of the empirical literature, we use Tobin’s q measured as the ratio between the sum of 
the market value of equity and book value of debt, and the book value of assets. We
also calculated the market to book ratio of equity and ROA.
From a long list of control variables, we selected group affiliation dummy,
pension fund dummy, debt-equity ratios (at market values), log of firm size, investment
ratios, cash flow available, average traded volumes, time dummies and 11 industry
dummies. Tables 6 and 7 explain the construction of these variables and summarize
descriptive statistics for them including cross correlations.
We want to study the effect of agency problems in the firm and its value. A
simple look at the correlation matrix of the variables, presented in Table 7.B, shows
that higher ownership concentration is negatively correlated with firm valuation and
that a higher coincidence between cash flow and control rights is positively correlated
with firm valuation. Although group affiliation is not correlated with the proxies to
market valuation of companies, affiliation is positively correlated with firms’ ROA.  

23
Figure 1 complements this evidence presenting scatter plots of these relationships
indicating that these results are not likely due to few outliers.

Table 6
Description of Main Variables

Var Description
Tobin’s q  Market book of assets / book value of assets at the end of each calendar year.  We 
estimate market value of assets as the book value of debt + book value of preferred
stock + market value of common stock.
M-to-B Ratio Market to book ratio. Market value of common stock / book value of common stock
ROA Return on assets. Net income divided by book value of total assets.
Concent Percentage of common stock owned by the three largest shareholders
Coincid Degree of coincidence between cash flow rights and control rights. Market value of
total shares directly and indirectly owned by the controller divided by market value of
total assets (book value of debt + book value of preferred stock + market value of
common stock).
Dgroup Dummy of group affiliation. 1 if the firm is affiliated to a conglomerate.
Dafp Dummy of pension fund investment. 1 if company shares may be acquired by a
pension fund.
Laec Log of market value of company assets.
DE Ratio Capital structure of company. Book value of debt divided by market value of equity.
Cash Cash flow available. Net income plus depreciation divided by market value of assets.
Invest Investment rate of company. Change in fixed assets plus depreciation divided by last
period fixed assets.
Turnover Number of shares traded during the year divided by total number of shares at the end
of the year.
Industry Dummy variables for membership in one of 11 industries.
dummies
Time dummies Dummy variables for each of 12 years.

However, as previously discussed, the correlations do not necessarily indicate

causation because of potential endogenous relations and omitted variable bias. We
tackle the second problem by running multiple regressions using the set of control

24
variables listed before. Table 8 presents this first set of results using standard OLS
pooled m ultivariate regressions. We tried several specifications in order to see
whether our results are robust. We found that including a large set of control variables
does not alter the preliminary results. In all specifications, firm valuation is negatively
and significantly correlated with ownership concentration and positively and
significantly correlated with the degree of coincidence of cash flow and control rights.
We also found that changing the set of control variables did not affect the signs and
significance of these coefficients.
These results tend to support the hypothesis that agency problems,
characterized by a lesser degree of coincidence of cash and control rights in hands of
company controllers, are penalized by the market. Holding ownership concentration
constant, more aligned incentives increase company value. On the other hand,
holding the relation between cash and control rights constant, an increase in
ownership concentration can be associated to more power on controllers’ hands and, 
potentially, more agency conflicts between controlling and minority shareholders.
However, the negative coefficient in the ownership concentration value could also be
related to liquidity problems. W e will explore this possibility later in the paper using a
measure of turnover.
There is another important result regarding corporate governance practices and
firm valuation. We find that the presence of pension funds as minority shareholders
increases the market value of listed companies. This result is robust across different
estimation procedures and means that for a given level of separation of cash from
control rights, institutional investors tend to mitigate agency problems between
controlling and minority shareholders.
The results indicate that under this type of model specification group affiliation
does not significantly affect firm value. Both the time dummies and the industry
dummies were statistically significant as a group in all specifications where they were
included.  Also, we find that larger firms have a higher Tobin’s q indicating higher 
market valuation, while more indebted firms present lower market valuation after
controlling for other factors. Both coefficients were statistically significant in most
specifications.

3. Robustness Checks

25
 Tables 9 and 10 present additional results that confirm the robustness of the
findings. In Table 9 we replicate the last regression of Table 8 using different
measures of firm valuation. The evidence shows that it is highly unlikely that our
previous results may be due to spurious correlations arising from measurement error
in the construction of Tobin’s q and the concentration and coincidence variables. The
Table shows that the results hold when we substitute Tobin’s q with the Market to 
book ratio or the firm’s ROA.  In Table 10 we show cross section regressions for each 
of the 12 years included in the sample. Again, the coefficient on concentration is
negative and significant while the coefficient on coincidence is positive and significant
in each of the 12 cross sections.
Additional tests for robustness are presented in Table 11. In this table we show
the results for different econometric methods. Among other things, we run fixed
effects panel regressions that take care of potential unobservable firm effects that
might be correlated with ownership concentration and rights coincidence, hence
biasing our previous results. The coefficients obtained are, again, very similar to
those previously obtained and, almost certainly, rule out the possibility that the results
could be due to omitted variable biases.
We also checked for the existence of non-linearities on the relationship
between concentration, coincidence and valuation. We find that introducing these
variables squared does not substantially change the results. We find an inverse U
shaped relation for concentration but no significant non-linear relation between
coincidence and value after we control for ownership concentration. The results are
presented in Table 12.

4. Endogeneity checks
Because of all these safeguards we are very confident that our results are not
induced by some omitted variable bias. However, we still have to tackle the potentially
endogenous nature of the correlations obtained. Both ownership concentration and
the degree of coincidence between cash flow and control rights could be
endogenously determined by the firm’s market valuation and/or performance. 
The positive coefficient obtained for the coincidence between cash and control
rights variable can also be explained under the reverse causation story. Endogeneity
of this type would imply that, for a given level of control rights, the owners of firms with
high Tobin’s q are more likely to increase their rights over the company cash flows,

26
hence increasing the degree of coincidence between cash flow and control rights.
Under this reverse causation, there could still be a causal connection between
coincidence and firm value, but the OLS coefficient would overstate it. Furthermore, if
an endogenous relation of the type predicted by Demsetz’s hypothesis is present, 
even if there is a causal relation for some firms between coincidence and valuation
one could not imply that other firms can improve their market values by increasing the
degree of coincidence between rights.
In the case of ownership concentration, it could also be argued that the
controllers of companies more valued by the market also tend to increase the
concentration of their holdings. That would be the case if they increased ownership
concentration through a pyramid scheme without increasing necessarily the
coincidence between their cash and control rights. However, in this case since the
coefficient on concentration is negative, the reverse causation correlation would run in
the opposite direction and, hence, would not reinforce the direct effect.
In order to adequately solve the endogeneity problems we should find suitable
instruments to run some type of instrumental variable or simultaneous equations
model. A suitable instrument should ideally be exogenous and not influenced by the
dependent variable of interest. It should be strongly correlated with the independent
variable for which there is suspicion of endogeneity (conc and coincid), and it should
predict the dependent variable only indirectly, through its effect on the independent
variable, but not directly. Given those restrictions, it should be difficult to get suitable
instruments unless some restrictive assumptions are made or an exogenous condition
on corporate governance practices is imposed on firms. The last regression in Table
11 intends to solve the endogeneity problem under some restrictive assumptions. For
that purpose, we run an Arellano and Bond Dynamic Panel GMM regression. This
econometric procedure takes care of unobserved firm specific effects and potential
endogeneity of the explanatory variables under the assumption that there is no second
order serial correlation on the error term. The estimated coefficients are very similar
to those obtained before but a Sargan test of the validity of the instruments (lagged
values of the control variables) largely rejects the null hypothesis.
Because the concentration and coincidence measures present much less time
series variation that firm valuation and performance variables, we conjecture that the
endogeneity problem is not likely very important in this case. However, we directly
tackle the endogeneity problem using an instrument related to control concentration.

27
In the previous analysis, we had used a measure of ownership concentration and of
coincidence of cash and control rights. Rule 67 of Chilean Corporations Law
establishes that major company decisions must be taken with the support of two thirds
of voting rights. It could be argued, then, that effective control requires two thirds of
voting rights. Figures 2.A and B show the relationship between a control dummy
variable and the coincidence between cash and control rights. Notice that although
both variables are related, the relationship is not obvious. For instance, a company
may have a controlling shareholder that controls with two thirds of the votes a holding
company that owns two thirds of the company shares. Therefore, the control
concentration dummy for this company would be 1, although, even in the zero debt
case, the coincidence between cash and control rights would be only four ninths.
Figure 2.A shows that, in any case, the degree of coincidence tends to be higher in
firms that surpass the two thirds threshold. Figure 2.B shows that there is no obvious
clustering of firms around the two thirds threshold, and that company market value
does not respond in an evident way to the threshold.
Therefore, in the following analysis we will use a dummy variable that takes the
value of one whenever the controller of a firm holds, directly or indirectly, more than
two thirds of voting rights as an instrument for the coincidence of cash and control
rights variable. Table 13 presents OLS regressions of Tobin’s q on this dummy, the 
coincidence variable and a series of control variables under different specifications.
The results are straightforward. When we include the dummy variable in our standard
specification the previous results do not change. Lower ownership concentration and
higher coincidence are still related to higher valuation. We then present a first stage
2SLS regression of the degree of coincidence on the concentration dummy and a
series of controls. We find that, as expected, firms with control concentration over two
thirds present higher coincidence. A hypothesis for this result is that as a company
reaches such a level of concentration in voting rights, it becomes very difficult for the
controller to attract external investors in order to separate cash from control rights.
We then run a second stage regression of Tobin’s q in all control variables and 
the instrumented coincidence variable. Table 13 shows that the 2SLS regressions
indicate that, after controlling for endogeneity and omitted variable bias, the coefficient
on the degree of coincidence remains positive although the economic and statistic
significance is notoriously reduced. Table 13 shows that the results hold for a similar

28
set of regressions run for a smaller sample of firms after controlling for traded
volumes.

4. Dividend Payout Ratios

Tables 14 and 15 present the results for the regressions using the dividend
payout ratio as the explanatory variable. They are structured in a similar way as those
for Tobin-s q. In the case of dividend payout ratios we find inverse U shaped
relationships similar to the one obtained for the German case by Durtouglu et al. The
results are the following. First, firms affiliated to conglomerates, firms with pension
funds as minority shareholders and larger firms present higher payout ratios. Second,
more debt implies less dividends. Third, separation of cash from control rights affects
payout ratios in a non/linear way. Similar to the German case, we find that there is a
threshold around 45% where the effect of higher controlling shareholder participation
changes the sign of the marginal effect of separation on payout ratios. We find that for
low values of the coincidence variable, increases in ownership concentration, as
expected, increase payout policy. However, when concentration in terms of equity
reaches over 70% of shares own by the controller, payout ratios start to decrease. A
hypothesis for that result might be related to tax incentives. As we already discussed,
once the controller has achieved such a high level of ownership, he/she can make
almost anything without opposition and there may be better (less expensive in terms
of taxes) ways of getting his/her money back.

VI. DO CORPORATE GOVERNANCE PRACTICES AFFECT FIRM MARKET

VALUATION AND PAYOUT POLICY?

1. Correlations with firm market valuation

In this section we focus on cross section data for the year 2003 in order to
include other dimensions of corporate governance in the empirical analysis. We
include in the analysis the corporate governance indicator variables constructed
starting from the questionnaire and summarized in the CG Index described above. As
explained above, we divided the corporate governance index in four sections:
disclosure, board practices, shareholders rights and conflict of interest. In addition we
consider a dummy variable that takes the value of 1 whenever we had to fill the

29
answers to the questionnaire without company assistance and consider, also
separately, the pension fund dummy. Of course, one could expect that the different
measures of corporate governance are highly correlated. In Table 16 we look at the
correlations between the control variables used in the previous regressions and the
corporate governance indicators. In general, the CGI index and sub-indices are
positively correlated among them and with market valuation. The sub-index
shareholders rights is the exception and presents negative, but low, correlation with
the other sub-indices. Firm size is positively correlated with good corporate
governance practices and less potential for agency problems, while the opposite is
true for debt.
In Table 17 we present multivariate Logit regressions between the CGI
components and the control variables. Larger firms, less indebted firms, firms with
more growth opportunities and with more cash flows available tend to present better
corporate governance practices.
Finally, we look at the effect of better corporate governance practices as
measured by the CG index and its components on market valuation. The regressions
presented in Tables 18 are estimated over a cross section sample of 85 companies for
the year 2003. The results are not very encouraging. We find that after controlling for
the list of control variables previously used only the sub-index Conflict of Interest
appears to be statistically significant in explaining firm value, and only few questions
have individual significance in the regressions. The overall index is not significant
and, the shareholders rights sub-index although significant, appears with negative sign
in the regressions. The lack of explanatory power can be attributed to the limited
sample use, because most of the variables that were significant in the previous panel
data regressions are statistically insignificant in these cross section regressions. In
addition, the high correlation observed among the different measures of corporate
governance and the control variables may, of course, imply multicolinearity in the
regressions, and hence low individual explanatory power.

2. Controlling for Endogeneity

Of course, as previously discussed, even the positive and significant coefficient
on the Conflict indicator could be due to endogeneity of the reverse causation type.
Without knowing if that is the case, we can not affirm that these type of better
corporate governance practices at the firm level are valued by the market. Following

30
Black et al (2003) we look for exogenous determinants of corporate governance
practices not directly caused by firm market valuation. Similarly to the Korean case,
Chilean Corporations Law requires that all firms with market capitalization above 45
million dollars form an audit committee composed by a majority of independent
directors.8 Presumably, firms with market capitalization above this value will tend to
have better corporate governance practices in order to be sure that they comply with
the law. Figure 3.A shows that, in fact, firms above this threshold present an audit
committee while the smaller ones do not. As a simple way to look at the validity of this
instrument, Figure 3.B shows that there is no apparent relationship between this size
threshold and company’s market valuation. 
The last two columns of Table 18 show 2SLS regressions using the size
dummy as an instrument for the CGI index in a regression of Tobin’s q on the agency 
problem variables and the set of control variables. The results show that the second
stage coefficient of conflict of interest on Tobin’s q remains positive and similar in 
value to the one obtained in an OLS regression indicating no evidence of endogeneity
in an important way. However, the result is only significant at the 15% level.

8
Rule 50 bis, Corporations Law.

31