Instrumental Variables Analysis and the Role of National Culture in Corporate Finance

• Published date: 01 June 2019
• Author: Ajay Patel,Robert Nash
• DOI: http://doi.org/10.1111/fima.12248
• Date: 01 June 2019

Instrumental Variables Analysis and the

Role of National Culture in Corporate

Finance

Robert Nash and Ajay Patel∗

In this article, we identify and review instrumental variables used to study the relation between

national culture and finance. Recognizing that an effective instrument must be correlated with

the endogenous variable (relevance condition) and must not be correlated with the outcome

variable(exclusion condition), we conduct analysis and provide theoretical argumentsto ascertain

the ability of each candidate instrument to meet these criteria. We further document whereto locate

the data to form the actual instruments. Overall, we design this study to help futureauthors weigh

the strengthsand weaknesses of using specif ic instruments and to better informempirical strategies

for addressing the endogeneity concerns inherentin the “culture and finance” literature.

Academic interest in examining the relation between national culture and economic outcomes

(or the “culture and finance” literature) has increased significantly over recent years. This is be-

cause national culture affects an individual’s beliefs and preferences, which then drivesdecisions,

affecting economic outcomes (Guiso, Sapienza, and Zingales, 2006). The findings of papers

comprising the culture and finance literature may be affected by mismeasurement of national

culture, or by omitted variables that may be correlated both with measures of national culture

and with the economic outcome being studied. To mitigate concerns about mismeasurement,

or omitted variables, and demonstrate causality, Guiso et al. (2006), Licht, Goldschmidt, and

Schwartz (2007), Ahern, Daminelli, and Fracassi (2015), Eun, Wang, and Xiao (2015), and many

other of the culture and finance papers use instrumental variables analysis (IV analysis hereafter).

The purpose of this study is to review those instruments, consider whether each satisfies the

requirements of a good instrument, and provide the sources of data to form the respective instru-

mental variables. Overall, our survey is designed to allow authors to evaluate the strengths and

weaknesses of recently used empirical instruments. Also, as wereview the theoretical foundations

for the recently used instruments, we separately recognize potential new instruments (which we

propose for possible use in future IV analyses).

This article is important because, as exemplified in a recent (Spring 2016) issue of Financial

Management (e.g., El Ghoul et al., 2016), the finance literature has increasingly argued that

differences in national culture are potential sources of variation in cross-national decision making.

More important, there appears to be much more that can be learned. Karolyi (2016) notes that

relative to other business disciplines, finance has been slow to embrace the explanatory power

provided by a rigorous examination of the impact of national culture and concludes that “there is

Wethank Rachel Pownall, and participants at the 2017 Financial Management Association (FMA) Asia/PacificMeeting,

the 2017 FMA EuropeMeeting, and the 2017 FMA Annual Meeting. We also thank Raghu Rau (Editor) and an anonymous

refereefor their very helpful comments.

∗Robert Nash is the Thomas K. Hearn Jr. Professor of Finance in the School of Business at Wake Forest University in

Winston-Salem, NC. Ajay Patel is a Professor and the Thomas S. Goho Chair in Finance in the School of Business at

WakeForest University in Winston-Salem, NC.

Financial Management •Summer 2019 •pages 385 – 416

386 Financial Management rSummer 2019

much potential yet for national culture to help us understand cross-border activities of firms and

investors” (p. 624).

However, this culture and finance research, like many empirical endeavors in financial eco-

nomics, is subject to endogeneity concerns. The ability to fulfill the research potential postulated

by Karolyi (2016) requires that these issues be addressed (such as through the use of IV analysis).

Our survey is designed as a guide to help future authors decide which instrumental variables may

be most appropriate given specific empirical settings.

If IV analysis helps resolve endogeneity concerns by forging a causal chain between national

culture and economic outcomes, the weakest link may be the economic justification for the

selection of particular instruments. Specifically, Larcker and Rusticus (2010) evaluate a 10-year

sample of top accounting journals and find that 80% of the studies using IV techniques provide no

discussion regarding the choice of instruments. Accordingly, another contribution of this article

is to help future authors overcome this methodological weaknessby not only identifying potential

instruments, but also by providing theoretical and empirical justification for selecting specific

instrumental variables.

Additionally, we offer the more utilitarian contribution of presenting detailed sources of data

for each variable. Although we should select potential instruments based primarily on the ability

to satisfy the econometric criteria of a good instrument (i.e., the relevance condition and the

exclusion condition), data availability is another factor to consider. Therefore, in our subsequent

discussion and in the Appendix, we provide numerous source citations to access the data we

discuss.

We organize the remainder of this article as follows. Section I describes the requirements

of a good instrument. Section II reviews primary measures of national culture. In studies

of the impact of national culture on economic outcomes, these measures of national cul-

ture represent the potentially endogenous variables for which IV analysis is designed to con-

trol. In Section III, we identify potential instrumental variables. During our discussion, we

provide economic justification for each instrument, assess its ability to fulfill the require-

ments of a valid instrument, and analyze its potential shortcomings. Section IV presents our

conclusion.

I. Requirements for Good Instrumental Variables

For an instrument to provide reliable estimates, it must meet two important requirements.

First, the instrument must be correlated with the endogenous variable (relevance condition)

after controlling for all other exogenous variables. Second, the instrument should only affect the

outcome variable through its effect on the endogenous variable (exclusion condition).

A. Relevance Condition

In our subsequent analysis, we empirically test the relevance condition by examining the

correlations between each candidate instrument and each potentially endogenous measure of

national culture. The pairwise correlations presented throughout this article provide an initial

(though incomplete) test of the relevance condition. Our univariatetests offer some evidence of an

instrument’sability to satisfy the relevance condition by demonstrating whether the instrumental

variable is correlated with the endogenous treatment variable. However, this is not a definitive

test because a relevant instrument must be correlated with the endogenous treatment variable

Nash and Patel rInstrumental Variables Analysis and the Role of National Culture 387

after accounting for all control variables.1Nevertheless, we feel that the pairwise correlations we

discuss are valuable because an instrument that is not correlated with a treatment variable in a

univariate setting is also less likely to be correlated with the treatment variable after accounting

for all other control variables.

B. Exclusion Condition

Conditioned on controls included in the models, the instrument should have no direct impact

on the outcome under study (other than through the effect on the endogenous factor). The exclu-

sion restriction is violated if the outcome variable is affected by additional factors that are also

correlated with the instrument. Because the exclusion condition mandates that the instrument be

uncorrelated with any unobserved variables influencing the outcome under study,empirical tests

of the exclusion restriction must be inferred because observing the unobserved variable is not

possible. Prior scholars havefrequently conducted overidentification tests. However, these are not

tests of exogeneity, but rather tests of whether the additional restrictions imposed by having addi-

tional instruments are valid (Parente and Santos Silva, 2012).The most popular overidentification

tests are from Hansen (1982) and Sargan (1958).

Because the exclusion restriction is untestable, it is imperative that the researcher build a

strong theoretical case for the use of a specific instrument. Section III is designed to provide the

theoretical justification for each potential instrument and thus address each instrument’s ability

to satisfy the exclusion restriction.

II. Measuring National Culture

A primary challenge in measuring national culture is first being able to define national culture.

Hofstede (1980) regards culture as “the collective programming of the mind which distinguishes

the members of one human group from another” (p. 25). Guiso et al. (2006) define culture

as “those customary beliefs and values that ethnic, religious, and social groups transmit fairly

unchanged from generation to generation” (p. 23). Synthesizing commonalities from these and

other definitions of culture, Taras, Rowney, and Steel (2009) observe that most depictions of

culture involve complex, multilevel constructs that are cast over extensive periods of time and are

therefore relatively stable. Zingales (2015) further notes that cultural constructs are frequently

couched in terms of beliefs (i.e., priors) and values (i.e., preferences). As such, the concepts of

culture may be incorporated into neoclassical economic models driven by individual preferences.

Another important step is differentiating national culture from corporate culture. Weber,

Shenkar, and Raveh (1996) note that national culture (following a Hofstede-esque specifica-

tion) is predicated on broad and deeply held values whereas corporate culture is defined by

specific operational practices. Hofstede (1991) similarly observes that national cultures are based

on values; corporate cultures are based on practices within companies. Corporate culture is

therefore considered unique to a particular organization and its market and environment (Schein,

1985; O’Reilly and Chatman, 1996; Tharp, 2009; Martin and Frost, 2012; Ferrell, Fraedrich,

1For the instrument to be “relevant,” the coefficient on the instrumental variable in the first-stage regression must be

statistically significant. The weakness of instruments may be tested using the F-statistic of Cragg and Donald (1993) and

the rk Wald statistic of Kleibergen and Paap(2006). Stock and Yogo (2005) provide rules of thumb regarding the use of

these metrics. We recommend that the researcher investigate Stata for supplemental information regarding tests of the

relevance restriction.