Do Business Groups Change With Market Development?

• Published date: 01 September 2016
• Date: 01 September 2016
• DOI: http://doi.org/10.1111/jems.12165
• Author: Borja Larrain,Francisco Urzúa I.

Do Business Groups Change With Market

Development?

BORJA LARRAIN

Escuela de Administraci´

on and Finance UC

Pontificia Universidad Cat´

olica de Chile

Avenida Vicu˜

na Mackenna 4860 Macul, Santiago, Chile

borja.larrain@uc.cl

FRANCISCO URZ ´

UA I.

Rotterdam School of Management

Erasmus University

the Netherlands

urzuainfante@rsm.nl

Khanna and Yafeh hypothesize that business groups should be more common in economies with

less developed markets and institutions. We test the time-series version of this hypothesis by

looking at changes in Chilean groups over 20 years (1990–2009). In this period, Chile experi-

enced a deep economic transformation as measured by common proxies of market development

(e.g., per capita income doubled). Despite this dramatic transformation, groups remained mostly

unchanged in terms of relative size, industrial diversification, vertical integration, control struc-

tures, internal capital markets, and reliance on external funds (minority equity plus debt). Only

leverage increased. Also, groups’ initial conditions were uncorrelatedwith market development at

the time of formation. This evidence casts doubts on the institutional-voids hypothesis, although

more subtle institutional voids, not captured by the type of macro proxies we use, might explain

the existence and resilience of business groups.

Business groups are prevalent in emerging markets (e.g., Brazil, India, China, South

Korea) and developed markets (e.g., Italy, Sweden). Yet, despite their extended pres-

ence, we know little about why they form and how they evolve. Khanna and Yafeh 2007

put forward several hypotheses in this respect. One hypothesis is that business groups

should be more common in economies with less developed markets and institutions.

Basically, business groups act as substitutes for capital, goods, or labor markets when

frictions are severe. A first approach to test this idea is to make cross-country compar-

isons on the prevalence of business groups as a function of variables that proxy for

market development. A second approach is to use within-country, time-series data to

see whether business groups correlate with market dynamics. The time series approach

has the advantage of controlling for country unobservables that blur cross-country

comparisons.

In this paper we contribute to the literature on business groups precisely by per-

forming a country-study in the style of the second approach described above. We use

a relatively long time series (20 years) of Chilean groups and test to what extent the

We would like to thank the co-editor, an anonymous referee, Yishay Yafeh, Steven Ongena and participants

of the First Latin American Workshop in Law and Economics (2014) for comments and suggestions. Carla

Castillo provided excellent research assistance. Larrain acknowledges partial financial support provided by

Programa Fondecyt (Proyecto Fondecyt Regular #1141161).

C2016 Wiley Periodicals, Inc.

Journal of Economics & Management Strategy, Volume25, Number 3, Autumn 2016, 750–784

Business Groups 751

structure of these groups correlates with proxies of the development of markets and in-

stitutions. A crucial ingredient to our experiment is that Chile’s economy experienced a

deep transformation during this period (1990–2009): per capita GDP doubled, the stock

market tripled in size, international trade kept expanding after liberalization, and so

on. This transformation was the result of macroeconomic reforms (e.g., trade and fi-

nancial liberalization, pension reform, independent central bank and subdued inflation,

among others) pushed by the government in the 1970s and 1980s rather than by the

business groups themselves. These reforms, together with a peaceful transition back to

democracy in 1990, made Chile one of the success stories in Latin America, and hence

a useful laboratory to test whether business groups lose prevalence once markets de-

velop.1In this way we are following Khanna and Yafeh(2007)’s recommendation: “Many

more historicalstudieswith explicithypotheses[...]whosetestableimplicationscan

be contrasted in time-series data, could shed further light on the evolution of groups, on

path dependence (ways in which “history matters”), and on the raison d’ˆ

etre of group

formation and development.” (p. 364).

By focusing on a particular country, instead of using a cross-country sample, we

lose some potentially interesting dimensions. However, identification (absent good in-

struments) is harder with purely cross-sectional data. In our case, identification is driven

by the massive size of the shock—the reforms and subsequent economic development—

that affected business groups in this period and which allows us to perform a quasi-

natural experiment. Several papers look at the historical evolution of business groups

in a country or perform a case study of the evolution of a particular group (see, for

instance, the recent country studies of Cuervo-Cazurra, 2014 (Spain); Ferreira da Silva

and Neves, 2014 (Portugal); Larsson and Petersson, 2014 (Sweden); or the co-evolution

of Indonesia and the Salim group in Dieleman and Sachs, 2008). In the case of Chilean

groups, Khanna and Palepu (1999) study the evolution of business groups in 1987–1997

with field data based on interviews. Khanna and Palepu (2000) study the benefits of

group affiliation in Chile with a 9-year sample, and conclude that the evolution of the

institutional environment affects the benefits of group affiliation, albeit slowly.

Our approach complements and improves upon these previous papers in several

respects. First, we assemble a long time-series (20 years) of quantitative measures of

group characteristics for 30 different groups. This panel structure, which is quite unique

in the business group literature, allows us to perform statistical analysis that is hard to

do with purely historical or qualitative data. Second, we cover a wide range of group

characteristics such as size, industrial structure (diversification and vertical integration),

control structures (pyramidal characteristics), and financial structure (leverage, external

funding, internal capital markets), which gives a general characterization of business

groups instead of focusing on one particular dimension (e.g., industrial diversification

as in Ferreira da Silva and Neves, 2014). Our characterization of internal capital markets

is particularly novel because data on intragroup lending is extremely hard to get in other

countries. Finally, we explicitly test the institutional-voids hypothesis by looking at the

relationship between Chile’s impressive development path and group characteristics.

More precisely, we compute the sensitivity of groups’ characteristics to different proxies

for market development such as per capita GDP, trade openness, bank credit, stock

1. Data availability makes Chilean business groups a frequent focus of study in relation to their perfor-

mance (Khanna and Palepu, 1999), their organization and structure (Khanna and Palepu, 1999; Lefort and

Walker, 2000), interlocking and stock returns (Khanna and Thomas, 2009), internal capital markets (Buchuk

et al., 2014), and board compensation (Urz´

ua, 2009).

752 Journal of Economics & Management Strategy

market capitalization, and regulatory improvements. For instance, given that financial

markets became deeper and more sophisticated throughout this period, the internal

capital markets of groups should become less active since, in theory, their purpose is

to provide risk-sharing agreements that are absent from formal markets (Khanna and

Yafeh, 2005; Belenzon et al., 2013).

Our results show that, despite the changes that Chile experienced in these two

decades, the structure of groups remains very similar to what it was in 1990 or when

they start. Although there are some changes, most changes in group characteristics do

not appear to be systematic (i.e., correlated with market dynamics). The sole exception to

the stability of group structure is leverage, which increases significantly from an average

of 30% in 1990 to 44% in 2009. This overall increase in leverage is explained by old groups

increasing leverage, and not simply by new groups being formed with higher leverage.

Business groups take advantage of the expansion of domestic credit by funding more

of their operations with debt, in a way that is perhaps analogous to the private equity

industry in developed countries (see Axelson et al., 2013).

Initial conditions (i.e., group characteristics at the beginning of the sample period or

when groups start) explain the lion’s share of variation in groupstructure throughout the

sample. Simply put, the structure of groups is more closely related to the starting point

of each group rather than to subsequent market dynamics. Initial conditions themselves

could be a function of market development and institutional voids when each group

is formed. In fact, looking at initial conditions is a stronger test of the institutional

voids hypothesis, because they are free from the interference of adjustment costs. In

particular, the absence of major changes that we mention above could simply be the

result of adjustment costs once groups are set up. However,in the data we do not find a

clear association between initial conditions and proxies of market development either.

In other words, groups’ initial structure does not seem to be designed to tackle different

levels of market development or institutional voids. Again, the sole exception is group

leverage: new groups start off with much higher leverage than the leverage that old

groups had when they started.

Our paper contributes, first and foremost, to the literature on business groups’

formation and evolution. One strand of the existing theoretical literature is based on the

idea that groups compensate for underdeveloped markets and institutional voids (see,

for instance, Kali, 2003; Almeida and Wolfenzon, 2006; Khanna and Yafeh 2007). Our

results show that the substitutability between groups and markets is non-existent in the

medium run, and even perhaps in the long run. Another view of our results is that they

help to quantify what “long run” means for business groups. In simple words, twenty

years does not seem to be enough to trigger a significant change in business groups.

This is despite the transformation of an economy that doubles in per capita income or

triples in stock market capitalization during the same period. This resilience of business

groups is not unheard of, as shown by empirical evidence (e.g., the survival of British

trading companies through the market development of the nineteenth and twentieth

century; see Jones and Colpan, 2010) and recent theoreticalinterest on this matter (Colpan

and Hikino, 2014). Perhaps the underlying relationship between business groups and

markets is nonlinear. It could be the case that Chile in these 20 years did not cross the

particular development threshold that triggers a change in group structure. Although

this is a plausible alternative, our results suggest that these thresholds, if they exist, are

far apart in the development path. For example, a country like Chile that doubles per

capita income does not seem to cross such a threshold.