Family Business Review OnlineFirst, published on March 30, 2010 as doi:10.

1177/0894486510365508

Family Business Review

Family Ownership and Firm XX(X) 1­–7

© 2010 Family Firm Institute, Inc.
Reprints and permission: http://www.
Performance in Chile: A Note sagepub.com/journalsPermissions.nav
DOI: 10.1177/0894486510365508

on Martinez et al.’s Evidence http://fbr.sagepub.com

Claudio A. Bonilla1, Jean Sepulveda1, and Mariela Carvajal2

Abstract
The authors revisit the evidence presented in Martinez et al. using new data and estimation techniques that take
into account unobserved firm heterogeneity. The results of the earlier study are found to be robust to the new
procedures because performance of family-controlled firms continues to be superior to that of nonfamily firms. The
authors then add the risk dimension to the earlier analysis using a risk-adjusted return on assets (ROA) variable, and
family-controlled firms again performed better. A test of the standard deviations of ROA for both firm categories
revealed that family-controlled firms not only perform better but also show less volatility in their returns.

Keywords
family-controlled firms, performance, ROA, risk, returns volatility

Various authors have maintained that family-con- These extensions to the Martinez study generate a
trolled firms are more profitable than nonfamily ones, number of interesting results. First, by using return on
basing their assertion on certain advantages such as assets (ROA) as a profitability measure, we found that, as
the lower agency costs of dealings between family in Martinez, family-controlled firms are more profitable
members and a longer-term management perspective. than nonfamily ones in Chile. Second, after controlling for
Others, however, take the opposite view, arguing that AFP (private pension fund managers) investment in fam-
family businesses have significant disadvantages stem- ily-controlled firms and nonfamily ones, the outcome does
ming from their limited capacity to hire external not change. In other words, the fact that a firm is “AFPable”
managers for key executive positions and the always- (i.e., its characteristics are such that AFPs may legally
present possibility that the family may expropriate value invest in it) does not explain the difference in profitability
from minority stakeholder. Either of these phenomena despite the influence AFPs have in the Chilean market.
could presumably damage market confidence in compa- Furthermore, even after adjusting ROA for risk, by type of
nies controlled by families and negatively affect their firm, we found that family-controlled businesses per-
share prices. formed better than nonfamily ones.
This research note is an extension of Martinez, Stöhr, Finally, and most importantly, we found that in family-
and Quiroga (2007; hereafter simply Martinez) in which controlled businesses profitability not only is higher but
we incorporate certain methodological improvements,
more recent data, and a new conceptual aspect that 1
Universidad del Desarrollo, Santiago, Chile
extends the comparisons beyond profitability to include 2
University of Chile, Santiago, Chile
the risk dimension. These innovations address a signifi-
Corresponding Author:
cant weakness in that comparing profitabilities of the Claudio A. Bonilla, Universidad del Desarrollo, Av. La Plaza 700, Las
two types of businesses is of little use without some idea Condes, Santiago, Chile
of the risk differential. Email: cbonilla@udd.cl
2 Family Business Review XX(X)

also has lower variance. This result contradicts the notion conclusion was that family-controlled businesses were
of a risk–return trade-off, which states that if a given more profitable. The same study obtained ordinary least
asset or asset portfolio has a higher rate of return than squares (OLS) estimates of a model that included time
some other one, its risk, in equilibrium, should also be fixed effects and dummy variables representing the
higher. It therefore also constitutes indirect proof that industrial classification of the firms included in their
the Chilean equity market does not satisfy the efficient panel data. These dummies captured unobservable het-
market hypothesis, as was recently described in the erogeneity at the industry level but not at the level of the
finance literature by Romero, Bonilla, and Hinich (2007), individual firm. Indeed, unobservable factors such as
because it is possible to build a portfolio with family firm management quality and organizational structure that
stocks and obtain abnormal returns for a given level of could affect the profitability of the companies in both
risk. How to form such a portfolio is not the concern of categories are not captured by incorporating dummies
this article; that such a possibility exists, however, is at the industry level. Furthermore, the unobserved firm
interesting in itself and indirectly corroborates empirical heterogeneity could be correlated with explanatory
findings in the literature on the Chilean market. variables. As an example, management quality, which
affects profitability, might be correlated with the type of
firm (family controlled or nonfamily), size, age, and
Related Empirical Literature indebtedness as well as other firm characteristics. Not
Comparisons of the financial performance of family- picking up these correlations meant losing valuable infor-
controlled firms and nonfamily businesses have been mation, and the OLS estimators could then be biased.
the subject of some recent studies. Using data for 1992 Martinez did not measure the effect of institutional
through 1999 on the ROA for companies included in the (AFP) investors on the Chilean capital market, nor did
Standard & Poor’s 500, Anderson and Reeb (2003) found Martinez incorporate the risk levels associated with the
that family-controlled firms performed significantly returns of family-controlled firms versus nonfamily
better than nonfamily ones and that the highest profit- ones. Both of these phenomena are included in the model
ability was reported by firms in which a family member we present below. In the case of AFPs, this reflects our
was also the CEO. They thus concluded that family view that the impact of the AFPs is significant and should
ownership is an effective organizational structure. From be taken into account.
a similar perspective, Lee (2006) utilized data for 1992
through 2002 to confirm that family-controlled firms in
the United States generated more employment and rev- Method and Data
enue growth and displayed higher profitability. These Hypotheses to Be Tested
results bolster the theory that firm performance grows
when founder members belonging to the family are Our first purpose is to investigate whether the results of
involved in management. Another recent investigation previous studies on profitability are in fact true for Chile
by Allouche, Amann, Jaussaud, and Kurashina (2008) on using a more robust methodology than Martinez, new
the Japanese economy also showed that family-controlled data, and specific measures to control for the presence of
firms perform better and have a stronger financial struc- AFPs. Therefore, the hypotheses we test are as follows:
ture than do nonfamily firms. Their evidence was based
on financial profitability indicators such as ROA, return Hypothesis 1: Family-controlled firms listed on
on equity (ROE), and return on invested capital. the Chilean stock market are more profitable
On the family business sector in Chile, little work has than nonfamily firms.
yet been published, an exception being the 2007 Marti-
nez study. Its authors examined the impact of family The AFPs are major institutional investors that
ownership on the performance of Chilean firms listed on administer Chileans’ pension funds and play a major role
the Bolsa de Comercio de Santiago, the country’s prin- in the market because of the large amount of resources
cipal stock exchange. Three performance measures were they manage. Their demand for a particular asset usually
used—ROA, ROE, and a proxy for Tobin’s q—and the influences the stock price of it. Also, the financial assets
results for family-controlled firms and nonfamily firms in their portfolios reveal which firms they are investing
were subjected to difference of means testing. Their in and may be an indicator of those firms’ financial
Bonilla et al. 3

solidity and attractiveness. It is not uncommon for other Table 1. Number of Companies in Sample, by Year
market players to follow the investment strategies of the
Year Total
AFPs. Therefore, when an AFP decides to buy a particular
asset or increase the asset’s portfolio weight, it is very 1998 253
likely that the stock price of that asset will be positively 1999 255
affected. Thus, it is interesting to see if the difference in 2000 259
profitability between family-controlled firms and nonfam- 2001 251
2002 246
ily firms remains when controlling for the effect of AFPs.
2003 241
2004 237
Hypothesis 2: If the effect of AFPs is controlled 2005 260
for, the difference in profitability between 2006 257
family-controlled firms and nonfamily ones 2007 246
declines significantly or disappears. Total 2,505
Source: Derived from databases (see text).
In addition to studying the differences in profitability,
we incorporate an analysis of the variance in the returns
of the two types of businesses. Previous studies have not firm if it was controlled at the senior management
carried out such an analysis. We include it here to better level by one or more members of a family-controlled
explain the results and determine whether this variability firm on the SVS list.
is an important attribute in the Chilean market. This Third, a company not in any business group was clas-
leads us to a third hypothesis: sified as a family-controlled firm if its board of directors
was controlled by one or more members of a family on
Hypothesis 3: The returns to family-controlled the SVS list. For both this and the second criterion, we
firms in the Chilean market exhibit greater vari- used information from credit rating agencies, company
ance than returns to nonfamily firms and thus financial reports, market data, and other company
are consistent with the equilibrium risk–return sources. Nonfamily firms were defined as all companies
trade-off. not fitting these three family-controlled-firm criteria.1

In addition to testing for differences in variance, we

adjust the returns for risk and type of firm to determine Data Set
if the results in Martinez continue to hold. This leads us The data set for our test samples of listed companies cov-
to our fourth hypothesis: ered the period between January 1998 and December
2007. They were obtained from Economática (a Latin
Hypothesis 4: After adjusting for risk, the differ- American database vendor), the SVS, the Chilean Super-
ence in profitability between family-controlled intendant of Pension Fund Administrators (SAFP), and
firms and nonfamily firms disappears. the Santiago Stock Market. The number of enterprises in
the sample for each year is shown in Table 1. As in Mar-
tinez, we have excluded nonprofit entities and holding
Definition of Family Firm in the Chilean Context companies, whose financial statements were a composite
To classify a given company as a family firm, we of their also-public subsidiaries. However, we did not
employed three criteria. First, we examined the list of exclude financial institutions as Martinez did. Financial
business groups published by the Chilean Superinten- institutions are an active part of the Chilean economy
dant of Securities and Insurance (SVS). As of year-end and may help to better explain the findings of the refer-
2007, there were 117 such groups. In each case, if the ence study.2
group was clearly associated with a business family, Based on the criteria described above, an average of
the firms constituting it were considered to be family- 68% of the businesses in the sample were classified as
controlled firms. family-controlled firms and 32% as nonfamily firms for
Second, if a company was not a member of any of these the period covered. The figures for individual years, given
corporate groups, we categorized it as a family-controlled in Table 2, reveal that the percentage of family-controlled
4 Family Business Review XX(X)

Table 2. Percentages of Family-Controlled Firms and • Age: Age of the firm, that is, the number of
Nonfamily Firms years since it was founded; this information
Year Family (%) Nonfamily (%) Total (%) was found on the companies’ Web sites or
through direct consultation by telephone
1998 70.8 29.2 100.0 • Industry: The firm’s industrial classification;
1999 67.1 32.9 100.0 the Economatica database uses a classification
2000 68.0 32.0 100.0
system of 19 sectors
2001 68.5 31.5 100.0
2002 68.7 31.3 100.0
2003 69.3 30.7 100.0
2004 69.6 30.4 100.0
Method
2005 66.2 33.8 100.0 Our sample embraces information on a number of firms
2006 66.5 33.5 100.0 over a period of years (1998–2007) and thus constitutes
2007 68.7 31.3 100.0 a panel data set. As in Martinez, we first test for differ-
Source: Derived from databases (see text). ences between the means of ROAs for family-controlled
firms and for nonfamily ones.3 We then estimate the
model using the same method but with our new data.
firms remained much higher than the percentage of non- And because there exist nonobservable effects that are
family ones throughout the 10-year period. probably correlated with the independent variables (e.g.,
the type and quality of management may be different for
firms of different size and age), we opted to estimate the
Model Variables relationship using a fixed-effects regression model
Our regression model contains a single dependent and six before going into the risk comparison.
independent variables. The dependent variable is ROA, Nevertheless, we allow for the possibility that the
which was chosen for two reasons: first, because it is the unobservable factors are random and thus employ the
most commonly used variable in this type of analysis and, Hausman test to decide which method is most appropri-
second, to ensure our results would be comparable to ate. The basic model is written as follows:
those of the Martinez study. ROA is a measure of finan-
cial performance that indicates how the firm’s assets were ROAit = ai + b1SIZEit + b2AGEit + b3DEBTit +
managed during the period under study. The data for this b4DFAMit + Uit
variable were drawn from the Economatica database. t = 1998, 1999, . . . 2007 (1)
The independent control variables, measured for i = firm
each firm, are the following:
Our first step is to determine whether the incorporation
• Family dummy (DFAM): Dummy variable that of the individual unobservable effects (ai) changes
equals 1 for family-controlled firms the results found by Martinez. We then incorporate
• AFP ownership dummy (DAFP): Dummy vari- the DAFP dummy variable to establish whether the
able that equals 1 for firms whose ownership presence of AFPs in the ownership structure helps
structure includes institutional (AFP) inves- explain the profitability differences between family-
tors; the values for this variable were obtained controlled firms and nonfamily firms.
from SAFP reports on the investment portfo- Finally, to ascertain the effect of risk, we carry out
lios of Chilean pension funds during the period two procedures. First, we perform a difference of vari-
under study ance test on the two firm types. Second, we estimate
• Debt/assets: Leverage, defined here as the Equation (1) in which the dependent variable ROA is
debt/assets ratio, extracted from the Economat- adjusted for risk. We obtain the standard deviation of
ica database ROAs, by firm type, for each year in the sample. We use
• Size: Size of the firm, measured as the natural it as a proxy for risk and divide each firm’s ROA by
logarithm of total assets; the raw values were this measure, as explained in the Comparison of Risk-
extracted from the Economatica database Adjusted ROA Values section below.
Bonilla et al. 5

Table 3. Return on Assets (ROA) Sample Characteristics Table 5. Estimation Using the Method of Martinez et al.
(2007)
Variable Obs. M SD
Variable Return of Assets
ROA 2,446 0.0437126 0.134452
Intercept –0.2945
(.000)***
Table 4. Difference of Means Test for Return on Assets Size 0.0174
(ROA) of Family-Controlled Firms and Nonfamily Ones (.000)***
Age –0.0002
Means Statistics (.001)***
Debt/assets –0.0015
Family (%) Nonfamily (%) t Significance Level (.000)***
DFamily 0.0354
ROA 4.79 3.46 2.1745 .0149
(.000)***
Year dummies (9) Yes
Industry dummies (18) Yes
Results and Discussion Adjusted R2 .1175
Comparison of ROA Values F statistic 10.74
(.000)
The sample data relating to the ROA variable are sum- Obs. 2,269
marized in Table 3. Note:Values in parentheses indicate p values.
A difference of means test was then performed on the ***Significant at the 1% level.
mean ROA values for family-controlled firms and nonfa-
mily enterprises, the results of which are given in Table 4. variables. The results are presented in Table 6. The
As can be seen, family-controlled firms had a mean second column contains the equivalent of the estimates
ROA of 4.79% over the 10 years under study and thus in Martinez with unobservable heterogeneity permitted.
performed better than nonfamily firms, whose mean The results do not change; the Dfamily variable contin-
ROA was 3.46%. A t test yielded a value of 2.1745 (p = ues to be significant at conventional levels. In other
.0149), demonstrating that this difference is statistically words, under the new methodology and with new data,
significant and thus corroborating the first result from family-controlled firms still perform better.
Martinez despite the use of a different sample. An important variable not included in the Martinez
We then estimate Equation (1) with the same method analysis is the presence of the AFPs, Chile’s largest
(OLS) as Martinez using similar independent variables investors. We have incorporated it in our model to deter-
with our new data. The results, set out here in Table 5, mine the impact of these institutional investors on the
are comparable to those of Martinez both qualitatively performance of family-controlled firms and nonfamily
and quantitatively. The Dfamily dummy variable is pos- companies. Existing studies have already described the
itive and statistically significant, implying that even major influence they wield in the small Chilean capital
when size, age, and debt are controlled for, family- market (Romero et al., 2007).
controlled firms perform better than nonfamily firms. The figures in the third column of Table 6 demon-
As noted earlier, the model specification in Martinez strate that despite the inclusion of a dummy variable to
could not capture unobservable heterogeneity in the distinguish between firms that do and do not have
various firms, whether family controlled or not. Factors AFP investors, the previous results remain valid.
such as management quality and organizational struc- Family-controlled firms continue to show statistically
ture, though they clearly influence performance, are not significant higher average returns than nonfamily ones.
observable. To incorporate them into the estimation, we
can add dummies for each firm in the study, an approach
that would limit the degrees of freedom, or estimate Comparison of Risk-Adjusted ROA Values
them using a fixed-effects panel model. To investigate the possibility that the family-controlled
Thus, Equation (1) was reestimated allowing unob- firms’ superior performance is accompanied by higher
servable fixed effects correlated with the independent levels of risk, the ROA dependent variable in Equation
6 Family Business Review XX(X)

Table 6. Fixed-Effects Estimation of Model 1 fifth column, indicating that the variable is itself not sig-
nificant but its inclusion causes the family-controlled
Variable ROA ROA ROARISK ROARISK
firm dummy to become marginally nonsignificant. It
Size 0.0545 0.0541 0.4081 0.4057 continues to carry a positive sign, however.
(.000)*** (.000)*** (.000)*** (.000)*** Although intuitively we would expect the unobserv-
Age 0.0001 0.0000 0.0015 0.0009 able effects to be fixed rather than random and correlated
(.837) (.937) (.785) (.864)
with the independent variables such as size, we applied
Debt/assets –0.0021 –0.0021 –0.0161 –0.0159
(.000)*** (.000)*** (.000)*** (.000)*** the Hausman test to check this assumption. The results
Dfamily 0.0332 0.0303 0.2454 0.2275 for each specification confirmed that the fixed-effects
(.066)** (.097)** (.077)** (.105) estimation was the correct one for the present cases. The
DAFP 0.0099 0.0610 test values and their corresponding p values are shown
(.277) (.386) in the lower panel of Table 6.
R2 (within) .1006 .1011 .0975 .0979 Finally, we tested whether the standard deviation of
F statistic 55.78 44.87 53.91 43.27
the returns (the risk proxy) for family-controlled firms
p value (.000) (.000) (.000) (.000)
Hausman (c2) 44.08 46.45 48.53 51.27 was greater than that for nonfamily ones. This was done
p value (.000) (.000) (.000) (.000) using a difference of standards deviations test, whose
Obs. 2,269 2,269 2,269 2,269 results are given in Table 7.
As can be seen, the standard deviation of the ROA
Note: ROA = return on assets.Values in parentheses indicate p values.
**Significant between 5% and 10%. ***Significant < 1%. values for nonfamily firms is greater than that of the
family-controlled firms. The null hypothesis that the
deviations for the two categories of firms are the same
Table 7. Variance Ratio Test is therefore rejected with a high level of statistical con-
fidence in favor of the alternative hypothesis according
Ratio = sd(ROAnofamiliy)/sd(ROAfamily)
Ho: Ratio = 1 F statistic = 1.3315 to which the variability of family-controlled firm
Ha: Ratio¹ 1 prob (F > f) = 0.000 returns is less than that of nonfamily enterprises.
Ha: Ratio > 1 prob (F > f) = 0.000 The foregoing implies that family-controlled firms
Ha: Ratio < 1 prob (F < f) = 1.000 not only perform better than nonfamily ones but
also are less volatile on average. This is significant
because it raises the question of whether it is possible
to build an investment portfolio of family-controlled
(1) is replaced with ROARISK, ROA adjusted for the firm stocks that earns abnormal returns above the
risk factor, market line. The results indicate that this would be
possible, indirectly suggesting that the Chilean market
ROARISKi,t = ROAi,t/σj,t does not satisfy the weak efficiency hypothesis, as other
for i = firm, t = year, j = family recent studies have already demonstrated (Romero
firms, nonfamily firms et al., 2007).

where ROA is as previously defined and σj,t is the

standard deviation of the returns on family-controlled Conclusions
businesses and nonfamily ones for the year t. Thus, we In this research note we studied the financial perfor-
reestimate Equation (1) using a measure of the dispersion mance of family-controlled firms and nonfamily ones in
of the returns as a proxy for risk.4 the Chilean capital market. We extended the evidence
The new estimates with risk-adjusted ROA are shown presented in Martinez et al. (2007) and applied new data
in Table 6. The results obtained when the AFP dummy and a different and more robust methodology that took
variable is excluded, given in the fourth column, are sta- into account the unobservable heterogeneity in each
tistically significant and reveal that family-controlled firm. Our analysis also controlled for the effect of the
firms still perform better. The addition of the AFP private pension funds (AFPs) on the Chilean financial
dummy variable generates the results indicated in the market and adjusted for return risk.
Bonilla et al. 7

The results we obtained confirmed those presented in 2. If we exclude financial institutions, our results do not change.
Martinez. Controlling for AFP also maintained their orig- 3. As in Martinez, we also perform difference in means tests
inal findings. We embrace Martinez’s explanation in the for return on equity and Tobin’s q, with results similar to
sense that market scrutiny and accountability to minority the ones in Martinez. Because of space limitations we only
shareholders put pressure over family-controlled firms show those for return on assets.
and make them overcome most of their traditional disad- 4. This is equivalent to suggest a possible source of heterosce-
vantages. However, we also hypothesize that the quality dasticity. We checked it by a Goldfeld and Quandt test, and it
of the institutional setup and the capital market regula- did show that the variance was not the same for the two types
tions improvements implemented in the past 20 of firms. That is, by deflating by the standard deviation, we
years—such as the laws that established that takeovers are implicitly correcting for heteroscedasticity. We thank an
must be made through a public offer of acquisition and anonymous referee for pointing this out.
that minority shareholders may form audit committees
and the mandatory regulation that requires the presence References
of independent members in the board of directors—make Allouche, J., Amann, B., Jaussaud, J., & Kurashina, T. (2008).
difficult the extraction of value from the minority stake- The impact of family control on the performance and finan-
holders and help to explain the better performance of cial characteristics of family versus nonfamily businesses
public family-controlled firms in Chile. Also, given the in Japan: A matched-pair investigation. Family Business
social structure of Chilean society, it is highly likely that Review, 21, 315-329.
the family members’ human capital is among the highest Anderson, R., & Reeb, D. (2003). Founding family ownership
in the country and that they benefit from the social capital and firm performance: Evidence from the S&P 500. Jour-
and networking they have built over generations. All of nal of Finance, 58, 1301-1328.
these factors reinforce the advantages in reducing the Lee, J. (2006). Family firm performance: Further evidence.
agency problem of family-controlled firms over the dis- Family Business Review, 19, 103-114.
advantages of this ownership structure. Martinez, J., Stöhr, B., & Quiroga, B. (2007). Family owner-
We also provide a new interesting result in our arti- ship and firm performance: Evidence from public compa-
cle, showing that family-controlled firms are not only nies in Chile. Family Business Review, 20, 83-94.
more profitable but also less risky. This result is espe- Romero, R., Bonilla, C., & Hinich, M. (2007). Nonlinear event
cially interesting because it implies that, theoretically, detection in the Chilean stock market. Applied Economics
a portfolio of family-controlled firm assets could be Letters, 14, 987-991.
built whose returns would be above the expected rate
for a given risk level. This constitutes indirect proof Bios
of the efficiency problems of the Chilean capital Claudio A. Bonilla is an associate professor of economics at
market, which have been recently documented in the the Universidad del Desarrollo specialized in financial eco-
literature. nomics and organizational economics. He holds a PhD in
economics from the University of Texas at Austin and has pub-
Declaration of Conflicting Interests lished in such journals as Macroeconomic Dynamics, Public
The author(s) declared no potential conflicts of interests with Choice, Applied Economics, and The Manchester School.
respect to the authorship and/or publication of this article.
Jean Sepulveda is an assistant professor of economics at the
Financial Disclosure/Funding Universidad del Desarrollo specialized in econometrics and
The authors thank Superintendencia de Valores y Seguros of financial economics. He recently graduated with a PhD in eco-
Chile for partial financial support. nomics from North Carolina State University and has worked
as a financial consultant for several private institutions.
Notes
1. These criteria are similar to the ones in Martinez, except Mariela Carvajal is a PhD student at the University of Chile.
for the fourth one in their study, which we believe is less She works as an instructor at the Faculty of Economics of the
objective than the first three. We chose not to follow that University of Chile teaching financial analysis and doing
criterion. research in financial economics.