Political institutions, connectedness, and corporate risk-taking · 2017-09-09 · Political institutions, connectedness, and corporate risk-taking Narjess Boubakri1, Sattar A Mansi2

Political institutions, connectedness, and

corporate risk-taking

Narjess Boubakri1,Sattar A Mansi2 andWalid Saffar3

1School of Business and Management, American

University of Sharjah, Sharjah, United Arab

Emirates; 2Pampline College of Business, VirginiaTech, Blacksburg, USA; 3School of Accounting

and Finance, The Hong Kong Polytechnic

University, Hong Kong, China

Correspondence:N Boubakri, American University of Sharjah,School of Business and Management, Office2162, Sharjah 26 666, United Arab Emirates.Tel: + 971 6 515 2587;email: [email protected]

Received: 18 September 2011Revised: 23 December 2012Accepted: 6 January 2013

AbstractWe investigate the impact of political institutions on corporate risk-taking.Using a large sample of non-financial firms from 77 countries covering the

period from 1988 to 2008, we find that sound political institutions are

positively associated with corporate risk-taking, and that this relation is strongerwhen government extraction is higher. In a subsample of 45 countries, we also

find that politically connected firms engage in more risk-taking, which suggests

that close ties to the government lead to less conservative investment choices.

Our results are economically significant, and are robust to alternative risk-takingmeasures, various political institution proxies, cross-sectional and country-level

regressions, and endogeneity concerns of political institutions. Our results have

important implications for governments and corporate managers by providingdirect relevance of political institutions to the corporate decision-making

process. To encourage investment at the firm level, and hence innovation and

overall growth, governments need to undertake the necessary reforms tocontrol corruption and enforce contracts better, and thus decrease government

predation and extraction.

Journal of International Business Studies (2013) 44, 195–215. doi:10.1057/jibs.2013.2

Keywords: corporate governance; political institutions; risk-taking strategies; politicalconnections

INTRODUCTIONIn a continuously globalized environment, it has become impor-tant on both theoretical and practical grounds to foster our know-ledge of managerial decision-making and choices. Specifically,managerial risk choices and risk-taking are fundamental todecision-making, and have important implications for firm growth,performance, and survival (Bromiley, 1991; Shapira, 1995). Indeed,as indicated by Baird and Thomas (1985), “risk is embedded inmost long-range decisions,” and thus “studying the risk-takingpropensities of the decision-makers as they interact with particulardecision situations” will help us understand firms’ strategiesaround the world better.

A wide array of aspects related to the firm and to its environmenthave been shown to affect risk-taking. The environmental factorsthat are external to the firm are also beyond the control of themanager. Among such factors, political institutions that define thepolitical environment of the firm are material in shaping or con-straining managerial incentives to take risk and thus firm growthopportunities. As reported by Tan (2001), the political climate of

Journal of International Business Studies (2013) 44, 195–215& 2013 Academy of International Business All rights reserved 0047-2506

www.jibs.net

a country and overall political conditions arerelated to the degree to which the environmentfosters entrepreneurial growth.

Motivated by the growing body of work on thedeterminants of risk-taking, we take a firm-centeredapproach and examine whether political institutionsaffect managerial risk choices in the context ofcorporate investment decisions. In doing so, weadd to our understanding of why firms around theworld behave as they do, and of how to formulateappropriate strategic risk policies. This issue is ofparticular interest to foreign investors and multi-national corporations (MNCs) seeking joint-venturepartners abroad. A better understanding of theimpact of political constraints on firms’ strategicchoices is likely to help MNCs in their own decisions.

We posit that a country’s political institutions,or the extent of political constraints on the gov-ernment, affect corporate risk-taking through bothdirect and indirect mechanisms.1 For instance, thefirst-order effects of political institutions (i.e., thedegree of political constraint) may stem from theirimpact on firms’ operations, which in turn affectscorporate risk-taking. In support of this conjecture,Stulz (2005) in his “twin agency” model argues thatin countries with authoritarian regimes (i.e., thosewith weak political institutions that have fewerchecks and balances) governments are likely toaffect firm operations through over-regulation,solicitation of bribes, confiscatory taxation, andoutright expropriation of firm assets. More specifi-cally, government rent-seeking in such uncon-strained political environments, and the resultingoutright expropriation of firm assets (Caprio, Faccio,& McConnell, 2011), discourage corporate risk-taking. In addition, government policies drivenby the political objectives of policymakers, whotypically seek to maximize social stability andemployment (Fogel, Morck, & Yeung, 2008), alsoconstrain the firms’ ability to undertake riskyinvestments. Qi, Roth, and Wald (2010) presentrelated evidence that firms have a higher cost ofdebt financing in countries with relatively weakpolitical rights. This will eventually lead firms toborrow less, and to engage in less corporate risk-taking. Authoritarian governments are also likelyto reduce managerial risk-taking, as high policyrisk, or the likelihood of policy reversals, will leadcareer-concerned managers to overweight theuncertainty in the political environment by choos-ing sub-optimal investments.

The second-order effects of political institutionson risk-taking may alternatively stem from their

effect on the legal system, and on firm corporategovernance. Indeed, political institutions are shownto influence a country’s political stability, andtherefore the constancy of the legal system (e.g.,Rajan & Zingales, 2003; Roe & Siegel, 2011). Thelatter has in turn been shown by John, Litov andYeung (2008) to affect the extent of corporate risk-taking, as well as the firm’s information environ-ment shaping the balance of power between thefirm’s insiders and outsiders, all of which willultimately impinge on the extent and effectivenessof monitoring of corporate insiders. Additionally,Roe (2003) shows that political institutions affectthe degree of ownership concentration, and hencethe extent of monitoring they exert on managerialactions and decisions. Finally, the degree of influ-ence of interest groups such as labor unions – whotypically impose lower risk-taking on managers –depends on the extent of political constraints onthe government.2 In sum, we identify two potentialeffects of political constraints on corporate risk-taking: a direct mechanism that stems from theeffect of political institutions on the operatingenvironment of the firm, and an indirect mechan-ism that runs through the legal environment andfirms’ corporate governance. In our analysis, weseek to identify whether political institutions havea significant and sizable direct effect beyond theindirect “legal channel” effect documented by Johnet al. (2008).

A potential issue with our specifications is thatthe results may be driven by endogeneity. Thecurrent framework treats political institutions asa random variable, even though political insti-tutions are not randomly distributed around theworld, nor are firms randomly assigned to thetreatment group (strong political institutionregimes). To remedy these concerns, we examinethe results using an instrumental variable approachbased on country fragmentation and the individu-alism–collectivism dimension of national culture,and we find results that confirm the associa-tion between political institutions and corporaterisk-taking.

We also analyze whether the extent of govern-ment corruption and predation in the countryaffects the association between political insti-tutions and corporate risk-taking. Specifically,when political extraction and corruption prevail,the impact of political institutions will be stronger,because the latter affect not only resource allo-cation at the firm level but also the level of politicalextraction and enforcement of contracts and rules

Political institutions, connectedness, and corporate risk-taking Narjess Boubakri et al

196

Journal of International Business Studies

(Rajan & Zingales, 2003; Roe, 2006; Roe & Siegel,2011). Existing studies indeed show that when statecorruption and expropriation are high, managersbecome more risk averse (Caprio et al., 2011; Stulz,2005). This implies that the relation between poli-tical institutions and corporate risk-taking is likelyto be stronger in environments with high govern-ment extraction.

Lastly, we investigate the role of political connec-tions in corporate risk-taking. Faccio (2006) showsthat political connections are widespread aroundthe world, and that politically connected firmshave higher leverage, lower tax rates, and strongermarket power than non-politically connectedfirms. The benefits of close ties with politiciansinclude lower budget constraints, easier access tocredit, government rescue in times of distress, andlower cost of debt, despite the fact that politicallyconnected firms also exhibit lower earnings quality(Chaney, Faccio, & Parsley, 2011; Faccio, Masulis, &McConnell, 2006). This implies that politicallyconnected firms are more likely to undertake riskyinvestments, and therefore may exhibit higherearnings volatility than non-connected firms.

To test the above conjectures on the link betweenpolitical institutions, connectedness, and corporaterisk-taking, we consider all non-financial firmsfrom 77 countries available in Compustat over theperiod 1988 to 2008. We capture risk-taking usingthe volatility of firm-level profitability over a 5-yearoverlapping period. Our measure of political insti-tutions is the proxy of political constraints con-structed by Henisz (2010), which covers severalcountries over a long period of time, and incorpo-rates the political interactions among independentbranches of the government, thus capturing differ-ent dimensions of government and political con-straints. Our analysis provides evidence that poli-tical institutions with effective political constraintsare positively correlated with corporate risk-taking.We also find that this relation strengthens whengovernment extraction is high. In a subsample of45 countries, we further show that politically con-nected firms engage in more risk-taking, whichsuggests that close ties to the government lead toless conservative investment choices. Our resultsare not only statistically significant but also eco-nomically meaningful, and are robust to usingalternative risk-taking measures and various politi-cal institution proxies.

Our paper extends the literature on the deter-minants of firm-level managerial risk choices byexamining the impact of the prevailing country

political institutions. We provide complementaryevidence to Qi et al. (2010), who focus on theex post effect of political institutions on the firms’operating environment. In this respect, whereas Qiet al. (2010) look at the cost of debt financing,we rely on the firms’ earnings’ volatility. We alsoadd to extant evidence on the strategic behaviorof corporate insiders in response to governmentpredation and diversion of corporate resources(Stulz, 2005). Our results further contribute to theliterature on political connections by identifyinganother effect of these connections on corporatedecisions. Finally, going beyond John et al. (2008),who establish that shareholder rights protectionand rule of law affect the extent of risk-taking bycorporations around the world, we show thatpolitical institutions also matter above and beyondthese institutional constraints.

Our results also have important implications forgovernments and corporate managers, by pointingto the direct relevance of political institutionsto the corporate decision-making process. Toencourage investment at the firm level, and henceinnovation and overall growth, governments needto undertake the necessary reforms to controlcorruption and enforce contracts better, and thusdecrease government predation and extraction.Sound political institutions need also to be put inplace to improve the investment environment ofthe firms. These institutions will not only compen-sate for the weakness in legal institutions but also,ultimately, drive future institutional change.

MOTIVATION AND HYPOTHESESThe agency theory developed by Jensen andMeckling (1976) suggests that, lacking incentivesto the contrary, managers try to divert corporateresources so as to maximize their own privatebenefits at the expense of shareholders. Theseprivate benefits condition managers’ choices ofinvestment risks (John et al., 2008), leading them toreject positive net value projects because, comparedwith shareholders, their investment horizon isshorter (as it is linked to their job tenure), andtheir wealth less diversified (Wright, Ferris, &Awasthi, 1996).

In addition to these agency problems, recentstudies have underlined a potential link betweenpolitical institutions and managers’ choices ofinvestment risks. Stulz (2005), in his “twin agencymodel”, offers a theoretical framework toexplain the link between government expropri-ation (predation) and managerial diversion and


197


decision-making. Stulz argues that governmentsplay an important role in managerial decision-making by affecting the firms’ operating environ-ment through, among other things, over-regula-tion, solicitation of bribes, and expropriation of thefirms’ assets. When state expropriation (under weakpolitical constraints) is high, managers are likely tobecome more risk averse.

In an international study, Durnev and Fauver(2011) capture the multidimensional nature of stateinterference (including factors such as expropria-tion risk, degree of property rights, political con-straints on policymakers, and corruption), andshow that government predatory policies (e.g.,profits expropriation and bribes extraction) interactwith managerial incentives in shaping firms’ poli-cies, thus affecting managerial risk-taking behavior(John et al., 2008). Caprio et al. (2011) furtherprovide firm-level evidence from 109 countries onthe impact of government expropriation on firms’propensity to hold liquid assets. Specifically, theyshow that firms’ owners are more likely to channeltheir cash into less liquid assets, such as investmentin property, plant and equipment, to make extrac-tion by politicians and bureaucrats more costly.Under such conditions, firms’ managers are morelikely to avoid risk-taking.

Whether political institutions increase or decreasemanagers’ risk-taking in making corporate invest-ment decisions is still an open question, since theseinstitutions are likely to create two potential chan-nels of transmission with conflicting predictions.On the one hand, political institutions have a first-order effect on the risk-taking behavior of managersby affecting the firm’s operating environment (Stulz,2005). Indeed, under an authoritarian govern-ment characterized by few checks and balances orpolitical constraints, entrenched managers will tendto reduce their exposure to government over-regulation, extraction of bribes, and expropriationby undertaking less risky investments (Caprio et al.,2011). In addition, under weak political institutions,the firms’ cost of debt capital can be higher (e.g.,Qi et al., 2010), which is likely to result in reducedrisk-taking. These arguments are consistent with theview that stronger political institutions are positivelyrelated to corporate risk-taking.

On the other hand, political institutions have asecond-order effect channeled to risk-taking thro-ugh their impact on the firms’ corporate govern-ance/ownership structure. For instance, Roe (2003)shows that, under governments characterized byfew checks and balances, ownership concentration

is higher, which in turn will lead controllingowners to pressure managers to increase risk-taking.Additionally, because labor unions and other inter-est groups can impose lower risk-taking on mana-gers, and since these interest groups are lessinfluential under less constrained governments,risk-taking by managers is expected to be higherunder more authoritarian governments (e.g., Pagano& Volpin, 2005; Roe, 2003). These arguments areconsistent with a negative relation between poli-tical institutions and corporate risk-taking.

Disentangling the effect of political institutionson corporate investment decisions is an empiricalissue, but for exposition purposes we state ourhypothesis on this relation as follows:

Hypothesis 1a: Strong political institutions leadto higher risk-taking in corporate investmentdecisions, all else being equal.

Hypothesis 1b: Weak political institutions leadto lower risk-taking in corporate investmentdecisions, all else being equal.

We also analyze whether the relation betweenpolitical institutions and corporate risk-takingdepends on the extent of government predationin the country. We argue that resource allocation atthe firm level is likely to depend on the politicalinstitutions in place, as well as on the level ofgovernment extraction and corruption. The latterare indications of the lack of enforcement of rulesand contracts, and depend on the country’spolitical institutions (Rajan & Zingales, 2003; Roe,2006; Roe & Siegel, 2011). For instance, Knack andKeefer (1995) observe that:

Repudiation of contracts by governments is another indi-

cator of contract enforcement. It is likely if private actors

cannot count on the government to respect the contracts it

has with them, they will also not be able to count on the

government enforcing contracts between two private

parties. This restriction on economic activity severely limits

the universe of possible Pareto-improving exchanges that

would otherwise be undertaken.

We thus conjecture that when governmentextraction is high, political institutions are likelyto affect resource allocation strongly at the firmlevel, and hence affect the incentives of mana-gers to engage in risk-taking activities. Existingstudies indeed show that when state expropria-tion (under weak political constraints) prevails,corporate insiders either try to shelter their assets(Caprio et al., 2011) or collude with the government


198


to minimize any future extraction of private benefits(Stulz, 2005). Both possibilities are likely to depressrisk-taking as managers become more risk averse,and will tend to be more pronounced when politicalinstitutions are weak (and hence when legal institu-tions are unlikely to change). This leads to oursecond hypothesis:

Hypothesis 2: The association between politicalinstitutions and corporate risk-taking is strongerin environments where government extraction ishigher.

Under extreme agency problems, managers maycollude with the state to divert corporate resourcesfrom shareholders (Shleifer & Vishny, 1997), mak-ing expropriation by the state more likely. Onemechanism through which such collusion canoccur is political connectedness. Politically connec-ted firms are shown to benefit from lower budgetconstraints, government bailouts in times of dis-tress, and lower cost of debt (Faccio et al., 2006).Politically connected firms also exhibit higher per-formance and leverage than their non-connectedpeers after establishing a connection with thegovernment – either through a politician joiningthe board or the management team of the firm, orthrough the nomination of a corporate insider to apolitical position (Boubakri, Cosset, & Saffar, 2012).

For our purposes, of particular interest is thebailout protection that politically connected firmsenjoy in times of financial distress. In typical firms,the state of nature affects managerial risk aversion,because it also affects the magnitude of privatebenefits that can be extracted. As Durnev andFauver (2011) explain:

In low-cash flow states, there are less corporate resources to

siphon and the action is more readily detectable. To play

safe, insiders may even avoid some firm value enhancing

risky projects to preserve their private benefits. They will

undertake a risky project only if its outcomes in high-cash

flow states are high enough to compensate for the lower

level of diversion in less profitable states.

This raises the question as to whether links withthe government reduce risky investment by mana-gers. We conjecture that because politically connec-ted firms are insulated from bankruptcy in worsestates of nature (Faccio et al., 2006), they will bene-fit more from this implicit guarantee.3 This leads toour third hypothesis:

Hypothesis 3: Politically connected firms aremore likely to undertake risky projects.

SAMPLE, VARIABLE MEASUREMENTS, ANDDESCRIPTIVE STATISTICS

SampleWe consider non-financial firms covered by Com-pustat Global and Compustat North America overthe period from 1988 to 2008. We exclude financialfirms with SIC codes between 6000 and 6999because their profitability ratios, leverage ratios,and growth rates are calculated differently fromthose of non-financial firms, and because they areheavily regulated and hence highly sensitive to thedesign of a country’s political institutions. Thesample includes both active and non-active firmsfrom 77 countries, to mitigate concerns regardingsurvivorship bias of less risk-taking firms.

We begin the sample period in 1988 becauseCompustat Global covers non-North Americanfirms starting in 1987, and 1 prior year is neededto calculate growth rates. To be included in thesample we require the volatility of earnings to beavailable over 5 consecutive years. Our final samplecomprises 211,794 firm-year observations for26,513 firms from 77 countries covering the periodfrom 1988 to 2008. The distribution of our sampleacross countries shows that it is dominated by firmsfrom the United States (36.53%), Japan (14.89%),United Kingdom (6.51%), China (5.87%), andCanada (4.88%).4 To the best of our knowledge,our sample covers the largest number of countriesand the longest period to date in the literature oncorporate risk-taking.

Variable Measurements

Risk-taking variablesWe follow Faccio, Marchica, and Mura (2011) andcompute corporate risk-taking (RISK1) as the coun-try-adjusted standard deviation of the firm profit-ability (ROA) over 5-year overlapping periodsstarting in 1988 and ending in 2008 (i.e., 1988–1992, 1989–1993 etc), where ROA is computed asthe ratio of earnings before interest, taxes, depre-ciation, and amortization (EBITDA) to total assets.Similar to John et al. (2008), we take into account acountry’s economic cycle by computing the differ-ence between a firm’s ROA and the country’saverage ROA across non-financial firms in acorresponding given year. This adjustment isdesigned to remove country-level influences anddeliver a clean proxy for corporate risk-taking. Forrobustness, we also estimate three other measuresof corporate risk-taking (RISK2, RISK3, and RISK4).


199


RISK2 is the difference between the maximum andminimum ROA reported over the 5-year interval.RISK3 is the country-adjusted volatility of earn-ings for each firm over the entire sample period(1988–2008), requiring a minimum of five obser-vations in the cross-sectional regressions, as inJohn et al. (2008). RISK4 is the country-levelaverage of all the firm-level observations of RISK1.To mitigate concerns about outliers and data entryomissions we winsorize ROA at the 1% level onboth sides of the sample distribution. A briefdescription of each of these variables, as well asother variables used in the study and their datasources, is provided in the Appendix.

Political institutions and connectedness variablesTo properly examine the relation between politicalinstitutions and corporate risk-taking, we selecta political institutions proxy that covers a largeset of countries over a long time period, and onethat captures investors’ ex ante expectations. Weuse Henisz’s (2010) political constraints index(POLITICAL) as a measure of political institutions,where POLITICAL ranges from 0 to 1, with higherscores indicating greater political constraints andhence stronger political institutions.5 This measureis extensively used in the literature, and has severaladvantages.6 First, Henisz (2000) suggests that amajor determinant of political institutions is the“government’s ability to credibly commit not tointerfere with private property rights”. Thereforean ideal measure of political institutions shouldnot involve any constraints on policy change.Henisz’s index takes into account various charac-teristics of political institutions, including theextent of constraints on veto players in the system,and their political preferences. Second, Henisz’sindex captures investors’ ex ante views of restric-tions on government behavior rather than ex postgovernment performance (e.g., Qi et al., 2010).Third, Henisz’s index is available for more than 200countries, and covers all our sample period. Finally,Henisz’s index compares well with other proxiesthat have relatively small country and/or yearcoverage, while not exhibiting the drawbacks thatcharacterize other widely used political indices.7

To ensure the accuracy of our measure, we alsoreran our results using three alternative proxies ofpolitical institutions:

(1) government fractionalization, computed as thechance that two random draws will producelegislators from different parties (GOVFRAC);

(2) maximum difference of orientation amonggovernment parties (POLARIZ); and

(3) how responsive a government is to people(DEMOCRATIC).

We capture political connections using a dummyvariable, CONNECTED (drawn from Faccio, 2006),that equals 1 if the firm is politically connected:that is, if “at least one of its large shareholders(anyone controlling at least 10% of voting shares)or one of its top officers (CEO, president, vice-president, chairman, or secretary) is a member ofparliament, a minister, or is closely related to a toppolitician or party.”8

Control variablesWe control for firm and country characteristics thathave been shown to impact managerial risk-takingas in John et al. (2008) and Faccio et al. (2011).Country- specific controls include the quality ofthe institutional environment (LAWORDER) andthe real GDP growth (GDPG) in 1995 constant US$from the World Bank’s Development Indicators. Wefollow prior work by La Porta, Lopez-de-Silanes,Shleifer and Vishny (1998) and John et al. (2008)and use the rule of law index measure (LAWORDER)from the International Country Risk Guide (ICRG)to control for the quality of the legal environment(i.e., the effectiveness of the country’s regulatoryenforcement). The index LAWORDER ranges from 0to 6, with higher values indicating higher quality oflegal institutions. It covers more than 180 coun-tries, and varies over the sample period, whichallows us to control for changes in the quality ofthe countries’ legal environment over time.

Firm-specific controls include growth in assets,firm size, profitability, and leverage. Asset growth(GROWTH), a proxy for the influence of firm-specific growth opportunities on corporate risk-taking, is computed as the firm’s average growth intotal assets, denominated in US$ over the 5 yearsover which RISK1 is calculated. Firm size (SIZE),a proxy for economies of scale, is computed as thenatural log of total assets in millions of US$. Firmprofitability (ROA), a proxy for performance, ismeasured as earnings before interest, tax, deprecia-tion, and amortization scaled by total assets. Firmleverage (DTA), a proxy for financial health, ismeasured as the ratio of total long- and short-termdebt, scaled by assets. Finally, we include country,year, and industry dummies as in Campbell (1996)to control for the different fixed effects of thesevariables. We winsorize the firm-level variables at


200


the 1% level in each tail of the sample distribution,to reduce the potential influence of outliers.9 Allthe independent variables enter the regression atthe first year-end of the sample period over whichthe corporate risk-taking proxy (RISK1) is measured,as in John et al. (2008).

Descriptive StatisticsTable 1 reports the descriptive statistics of thevariables RISK1, POLITICAL, and CONNECTED bycountry for all the firms in the sample. In terms ofcorporate risk-taking, the data indicate that firmsincorporated in Western countries such as Canada,the United States of America, Sweden, and theUnited Kingdom along with Australia exhibit thehighest earnings volatility, while firms in Slovenia,Japan, and Colombia have the lowest volatility.A similar pattern exists for the political and con-nectedness variables. Firms in Western countries

Table 1 Descriptive statistics by country

Country Observation RISK1 POLITICAL CONNECTED

Argentina 431 0.05 0.63 0

Australia 7474 0.21 0.87 2

Austria 777 0.04 0.74 1

Bahrain 28 0.03 0.00 NA

Bangladesh 11 0.05 0.35 NA

Belgium 922 0.04 0.89 4

Botswana 5 0.06 0.71 NA

Brazil 1984 0.06 0.79 0

Bulgaria 13 0.06 0.75 NA

Canada 10,331 0.15 0.86 3

Chile 989 0.03 0.75 0

China 12,431 0.06 0.00 NA

Colombia 110 0.02 0.37 0

Croatia 52 0.03 0.68 NA

Cyprus 10 0.05 0.76 NA

Czech Republic 65 0.04 0.75 0

Denmark 1154 0.06 0.77 3

Egypt 65 0.05 0.71 NA

Estonia 75 0.05 0.77 NA

Finland 1115 0.06 0.77 2

France 5630 0.05 0.74 10

Germany 834 0.05 0.85 5

Greece 618 0.03 0.47 1

Hungary 129 0.04 0.75 0

Iceland 29 0.03 0.76 NA

India 4143 0.05 0.73 3

Indonesia 1660 0.06 0.25 21

Ireland 629 0.07 0.76 1

Israel 630 0.08 0.78 1

Italy 1639 0.04 0.70 12

Table 1 Continued

Country Observation RISK1 POLITICAL CONNECTED

Jamaica 63 0.04 0.34 NA

Japan 31,530 0.02 0.76 28

Jordan 81 0.04 0.72 NA

Kenya 32 0.03 0.47 NA

Korea South 2556 0.04 0.75 4

Kuwait 81 0.04 0.78 NA

Latvia 50 0.05 0.78 NA

Lithuania 77 0.06 0.77 NA

Luxembourg 173 0.05 0.77 0

Malaysia 6113 0.06 0.64 50

Mexico 805 0.03 0.36 5

Morocco 76 0.04 0.74 NA

Netherlands 1650 0.06 0.74 0

New Zealand 655 0.08 0.76 0

Nigeria 30 0.06 0.44 NA

Norway 1252 0.08 0.77 0

Oman 127 0.04 0.00 NA

Pakistan 532 0.05 0.18 NA

Papua 50 0.10 0.61 NA

Peru 295 0.05 0.48 0

Philippines 925 0.07 0.52 2

Poland 526 0.07 0.73 0

Portugal 383 0.04 0.75 1

Qatar 26 0.05 0.00 NA

Romania 17 0.05 0.76 NA

Russia 335 0.07 0.19 3

Saudi Arabia 48 0.03 0.00 NA

Singapore 3488 0.07 0.67 11

Slovakia 24 0.07 0.77 NA

Slovenia 95 0.02 0.77 NA

South Africa 1578 0.07 0.71 0

Spain 1214 0.03 0.75 1

Sri Lanka 91 0.03 0.44 0

Sweden 2383 0.10 0.77 3

Switzerland 2041 0.04 0.88 3

Taiwan 4128 0.04 0.76 5

Thailand 2574 0.05 0.66 11

Trinidad 8 0.03 0.81 NA

Tunisia 7 0.04 0.00 NA

Turkey 401 0.06 0.73 0

Ukraine 2 0.04 0.69 NA

United Arab

Emirates

84 0.04 0.00 NA

United Kingdom 13,791 0.10 0.74 82

United States of

America

77,358 0.13 0.85 12

Vietnam 7 0.05 0.14 NA

Zambia 7 0.07 0.76 NA

Zimbabwe 42 0.07 0.34 0

Total 211,794 0.09 0.73 290

Note: This table reports summary descriptive statistics by country for thekey variables used in the empirical tests to examine the impact of politicalconstraints on corporate risk-taking for a maximum sample of 211,794firm-year observations from 77 countries. The definitions and datasources for the variables are provided in the Appendix. NA denotes notavailable.


201


along with Australia and New Zealand have strongpolitical constraints and sound institutions, andcontain several politically connected firms.

Panel A of Table 2 reports descriptive statistics forthe variables used in the analysis. Included are themean, median, standard deviation, minimum, andmaximum values for RISK1, political, legal, andfinancial variables. The dependent variable RISK1has mean, median, and standard deviation of0.091, 0.041, and 0.166, respectively. This statisticis similar in magnitude to that reported by Johnet al. (2008) and Faccio et al. (2011). The largestRISK1 value is reported in Australia, and the lowestis in Japan. This evidence is consistent with Johnet al. (2008), who find that Japanese firms havethe least volatile earnings in their respectivesamples. This result has a national culture root,

and is consistent with Hofstede (2001), who ratesJapanese individuals as low risk-takers.

In terms of country-level variables, the mean,median, and standard deviation values of ourmain proxy for political institutions (POLITICAL)are 0.74, 0.77, and 0.21, respectively. This suggeststhat the majority of the sample contains countriesthat have strong political constraints and soundinstitutions. The variable POLITICAL varies from 0in countries such as Bahrain, Qatar, Tunisia, China,and United Arab Emirates to 0.89 in Belgium,which indicates that political institutions are nothomogeneous across our sample countries. Theresults further show that our sample includescountries with varying degrees of investor pro-tection, as measured by the ICRG assessment ofrule of law. Indeed, the mean and median values

Table 2 Descriptive statistics

Mean Median Standard deviation Minimum Maximum

Panel A Descriptive statistics of key regression variables (n¼211,794)

RISK1 0.091 0.041 0.166 0.000 2.049

POLITICAL 0.738 0.772 0.215 0.000 0.894

LAWORDER 5.374 6.000 0.957 0.500 6.000

SOCIOECON 7.681 7.542 1.788 0.917 11.000

EXPROP 8.557 8.167 2.196 2.167 12.000

PRESSFREEDOM 1.684 2.000 0.670 0.000 2.000

CORRUPTION 1.925 1.917 1.215 0.000 6.000

GOVFRAC 0.147 0.000 0.239 0.000 0.893

POLARIZ 0.869 0.000 0.976 0.000 2.000

DEMOCRATIC 5.198 6.000 1.370 0.000 6.000

GDPG 7.010 7.125 1.600 0.000 10.000

CONNECTED 0.015 0.000 0.121 0.000 1.000

ROA 0.028 0.089 0.332 �2.444 0.416

DTA 0.562 0.532 0.411 0.024 3.669

GROWTH 0.266 0.087 0.771 �0.316 6.416

SIZE 4.932 4.907 2.411 �2.486 11.143

RISK1 POLITICAL LAWORDER GDPG CONNECTED ROA DTA GROWTH

Panel B Selected sample correlation (n¼211,794)

POLITICAL 0.121

LAWORDER 0.135 0.471

GDPG 0.067 �0.292 �0.276

CONNECTED �0.033 �0.120 �0.066 �0.022

ROA �0.645 �0.089 �0.102 �0.052 0.026

DTA 0.244 0.061 0.032 �0.037 0.003 �0.399

GROWTH 0.396 0.043 0.055 0.067 �0.019 �0.376 0.173

SIZE �0.468 �0.022 �0.054 �0.109 0.087 0.414 �0.067 �0.342

Panel A reports summary statistics for the key variables for a sample of 211,794 from 77 countries over the period 1988–2008. The data cover the periodfrom 1988 through 2008. The definitions and data sources for the regression variables are provided in the Appendix.Panel B reports Pearson correlations for the regression variables for a sample of 211,794 from 77 countries over the period 1988–2008. Boldfaceindicates statistical significance at the 1% level. The definitions and data sources for the variables are provided in the Appendix.


202


for the variable LAWORDER are 5.37 and 6,respectively, with a country average varying from1.38 in Bangladesh to 6 in several countries (e.g.,Luxembourg, Australia, and Finland). In terms ofpolitical connectedness, 1.5% of the firms in thesample are politically connected, with a standarddeviation of 12.1%.

As for our firm-specific characteristics, the resultsshow that our sample includes small and large firms,as well as high- and low-leverage firms. For example,the mean and median values of leverage in oursample are 0.56 and 0.53, respectively. Companies inthe sample appear to be relatively profitable, withmean and median ROA of 0.028 and 0.089, respec-tively, and exhibit a relatively high level of growth,with mean and median 5-year average assets growthrate of 0.266 and 0.087, respectively.

Panel B of Table 2 provides the correlationcoefficients between the political variable, RISK1,and various control measures. In general, RISK1is positively correlated with the political institu-tions variable (POLITICAL), law and order of thecountry (LAWORDER), gross domestic product(GDP) growth, firm leverage, and firm asset growth,and is negatively related to profitability, firm size,and political connections. Consistent with ourexpectations, the analysis indicates that firms loca-ted in countries with sound political institutionsare associated with more risk-taking. However,because of possible confounding effects by othervariables, we use a multivariate framework toexplore our hypotheses.

MULTIVARIATE ANALYSISIn this section, we report our results on the impactof political institutions (as well as firm and countrycharacteristics) on risk-taking, using a pooled, multi-variate regression framework. Panel observationshelp shed light on how the volatility of earningschanges over time in response to changes in thepolitical environment. Also, because our studyspans a long period (1988–2008), a panel estima-tion is more appropriate than a cross-sectionalestimation, since the control variables enter theregression at the first year-end of the sample periodover which the volatility of the earnings iscalculated.10 Nevertheless, in a robustness test weconsider a cross-sectional estimation. We estimateordinary least square (OLS) regressions using robuststandard errors corrected for clustering at the firmlevel. Because the number of firms varies acrosscountries, the individual observations are weightedwith the inverse of the number of firms from the

corresponding country. Specifically, we estimatethe following model (subscripts are suppressed fornotational convenience):

RISK1 ¼ aþ g1POLITICAL þ g2COUNTRYCONTROLS

þ g3FIRMCONTROLSþXY�1

Y¼1

YEARþXK�1

K¼1

IND

þXC�1

C¼1

CNT þ Z

ð1Þ

where RISK1 is the country-adjusted volatilityof firms’ ROA over 5-year overlapping periods;POLITICAL is Henisz’ (2010) index of politicalconstraints; COUNTRY CONTROLS include realGDPG and LAWORDER, the proxy for rule of law;FIRM CONTROLS refers to the set of firm-levelcontrol variables (SIZE, DTA, GROWTH, and ROA);YEAR, IND, and CNT are dummies that controlfor year, industry, and country fixed effects,respectively; and Z is an error term. Our focus inthe analysis is the coefficient g1, which measuresthe sensitivity of corporate risk-taking to thequality of the political institutions prevalent inthe country. A positive value indicates that soundpolitical institutions are associated with highercorporate risk-taking.

Evidence on the Relation between PoliticalInstitutions and Corporate Risk-takingTable 3 reports the results of regressing corporaterisk-taking on different proxies for political institu-tions, while controlling for firm- and country-levelcharacteristics. Model 1 presents our primary specifi-cation, and estimates our main variable, POLITICAL,while Models 2 through 4 examine alternativepolitical institution proxies.

The result from Model 1 suggests a positive andsignificant relation (at the 1% level) between poli-tical constraints and corporate risk-taking. Thisresult is consistent with Murphy, Shleifer andVishny (1991), who show that a lack of crediblecommitment on the part of government not tointerfere with private property rights leads, amongother things, to delayed investment and hencelower corporate risk-taking. In addition, becauseweak political institutions put fewer constraints onpredatory governments, the extent of expropriationis so high that there are fewer corporate resources todivert, driving managers away from high-riskprojects. Our results are economically significant.The coefficient estimate for POLITICAL suggests


203


that increasing this proxy by one standard deviationtranslates into increasing risk-taking by 9.11% (from0.0911 to 0.0994).

The remaining control variables have theirexpected signs. We find that firm leverage, firm-level growth, and growth in GDP are positively andsignificantly related to risk-taking, while firm sizeand profitability are negatively related. We also findthat the variable LAWORDER loads negative and isstatistically significant at the 1% level, consistentwith the idea that countries with law and ordersystems are associated with lower corporate risk-taking.11

Models 2 through 4 examine alternative measuresof political institutions. We use two alternativemeasures of political constraints from the WorldBank’s database of political institutions by Beck,Clarke, Groff, Keefer and Walsh (2001). Model 2replaces POLITICAL with GOVFRAC, a measure that

captures the political constraints on the country’slegislature. Consistent with the notion that strongchecks and balances in a country are expected toaffect the volatility of corporate earnings positively,we find a positive and significant relation (at the1% level) between GOVFRAC and corporate risk-taking. Model 3 replaces POLITICAL with POLARIZ,which captures the maximum difference of orienta-tion among government parties. Consistent withthe idea that greater polarization indicates strongerpolitical constraints, we find a positive and sig-nificant relation (at the 1% level) between POLARIZand corporate risk-taking. Model 4 replicates Equa-tion (1) using the ICRG assessment of democracy(DEMOCRATIC) as a measure of the country’spolitical environment. We find that democraticcountries are associated with higher corporate risk-taking. Overall, our results indicate that, in coun-tries with strong political institutions, firms aremore likely to undertake risky activities.

Sensitivity Tests

Corporate risk-taking, institutional environment andownership structureTo test the robustness of our results, we provideadditional specifications that include alternativeproxies for corporate risk-taking and control vari-ables for the institutional environment, and for theownership concentration. The results are providedin Table 4.

We consider three additional measures of corpo-rate risk-taking. First, we test the difference betweenthe maximum and minimum ROA over the 5-yearoverlapping periods (RISK2). Model 1 reports theresults. For our main political constraints variable,we find that the coefficient on POLITICAL loadspositively and is statistically significant at the 1%level. We also repeat this regression using differentproxies for political constraints and find similarresults. In an unreported regression, instead ofadjusting the risk measure for the economy-wideaverage, we consider a firm-specific unadjustedproxy. We continue to find support for our previousresult, that it is the firm-specific rather than theeconomy-wide volatility that drives our results.

Second, we follow John et al. (2008) and calculatethe country-adjusted volatility of earnings foreach firm for at least 5 years over the 1988 to2008 period (RISK3), and consider the average assetgrowth over the same period. In this case, eachfirm is included in our regression only once, andthe number of observations drops from 211,794

Table 3 Political institutions and corporate risk-taking

Variable Primaryspecification

GOVFRAC POLARIZ DEMOCRATIC

(1) (2) (3) (4)

Intercept 0.298*** 0.313*** 0.319*** 0.307***(17.935) (19.380) (19.708) (19.035)

POLITICAL 0.039***(4.688)

GOVFRAC 0.009***(4.220)

POLARIZ 0.002***(3.961)

DEMOCRATIC 0.005***(4.772)

GDPG 0.004*** 0.004*** 0.004*** 0.003***(8.574) (8.214) (7.642) (6.687)

LAWORDER �0.008*** �0.008*** �0.009*** �0.008***(�7.132) (�7.582) (�8.323) (�7.212)

ROA �0.222*** �0.222*** �0.222*** �0.222***(�47.085) (�47.074) (�46.995) (�47.065)

DTA 0.006* 0.006* 0.006* 0.006*(1.752) (1.754) (1.737) (1.759)

GROWTH 0.027*** 0.027*** 0.027*** 0.027***(13.912) (13.916) (13.877) (13.904)

SIZE �0.018*** �0.018*** �0.018*** �0.018***(�33.219) (�33.222) (�33.064) (�33.232)

Years FE Yes Yes Yes YesIndustry FE Yes Yes Yes YesCountry FE Yesp-value 0.000 0.000 0.000 0.000R2 0.516 0.516 0.516 0.516Observations 211,794 211,471 201,050 211,794

This table provides estimated coefficients from regressing the corporaterisk-taking variable on political institution proxies and various country-and firm-specific control variables. The data cover the period from 1988through 2008. Each estimate is reported using robust t-statisticsclustered at the firm level. The complete definitions and data sourcesfor the variables are outlined in the Appendix. The notation ***, **, and *denotes statistical significance at the 1%, 5%, and 10% levels,respectively.


204


firm-years to 26,513 firms. All the independentvariables enter the regressions at the first year-endof the sample period over which the corporate risk-taking proxy (RISK3) is measured. The results areprovided in Model 2, and indicate that politicalinstitutions are positively related to corporate risk-taking. The variable POLITICAL loads positively,and is statistically significant at the 1% level.

Third, since panel regression results may bedriven by countries with more firm-level observa-tions (those in more developed countries withstronger political institutions), we use the averageof the firm-level observations in each country, thus

giving each country an equal weight in theregression. This approach avoids overweightinglarge economies. We calculate RISK4 as the averageof RISK1 within a given country. Model 3 providesthe results, and again suggests that politicalinstitutions are positively and significantly relatedto corporate risk-taking. The variable POLITICALremains positive, and is statistically significant atthe 1% level.

In Model 4, we control for firm ownershipstructure (OWNERSHIP) using observations fromthe Osiris database (Bureau Van Dijk). Jensen andMeckling (1976) suggest that large shareholders

Table 4 Political institutions and corporate risk-taking: Additional tests

Variable RISK2 Max(ROA) –

Min(ROA)

RISK3 Cross-sectional

regression

RISK4 Country-level

regression

OWNERSHIP Additional

control

(1) (2) (3) (4) (5)

Intercept 0.717*** 0.065** 0.083** 0.113*** 0.316***

(19.055) (2.381) (2.416) (7.111) (17.310)

POLITICAL 0.092*** 0.147*** 0.018*** 0.027** 0.049***

(4.832) (5.476) (3.011) (2.032) (5.757)

SOCIOECON 0.001***

(2.850)

EXPROP �0.000

(�0.649)

CORRUPTION �0.004***

(�3.990)

PRESSFREEDOM 0.003***

(4.279)

OWNERSHIP 0.000

(0.587)

GDPG 0.009*** 0.013*** 0.004** 0.001 0.002***

(8.832) (7.847) (2.001) (0.643) (5.329)

LAWORDER �0.020*** �0.015*** �0.001 0.001 �0.010***

(�8.400) (�3.059) (�0.503) (0.424) (�9.422)

ROA �0.507*** �0.010*** 0.087* �0.275*** �0.222***

(�47.298) (�4.043) (1.814) (�30.687) (�47.016)

DTA 0.003 �0.001 �0.032 0.008 0.006*

(0.372) (�1.480) (�1.415) (1.351) (1.760)

GROWTH 0.058*** 0.063*** 0.005** 0.016*** 0.027***

(13.767) (7.291) (2.149) (4.864) (13.905)

SIZE �0.042*** �0.043*** �0.001 �0.007*** �0.018***

(�34.681) (�38.966) (�0.682) (�12.359) (�33.205)

Years FE Yes Yes No Yes Yes

Industry FE Yes Yes No Yes Yes

Country FE Yes Yes No Yes Yes

p-value 0.000 0.000 0.000 0.000 0.000

R2 0.518 0.410 0.427 0.522 0.516

Observations 211,794 26,513 77 24,546 211,794

This table provides estimated coefficients from regressing the corporate risk-taking variable on alternative corporate risk-taking measures. Model 4controls for OWNERSHIP. Model 5 controls for additional country-level variables. Each estimate is reported using robust t-statistics clustered at the firmlevel. The complete definitions and data sources for the variables are outlined in the Appendix. The notation ***, **, and * denotes statistical significanceat the 1%, 5%, and 10% levels, respectively.


205


have the ability and the incentive to shapecorporate risk-taking. Despite the reduction in thenumber of observations, we continue to find apositive and significant relation between politicalinstitutions and corporate risk-taking after control-ling for ownership concentration. We also find apositive, albeit insignificant, relation betweenOWNERSHIP and our main variable RISK1.

Finally, in unreported regressions, we examinesome possible channels through which politicalinstitutions may impact corporate risk-taking beha-vior by adding sequentially measures of socio-eco-nomic condition, expropriation risk, corruption,and freedom of press as possible explanations forthe impact of political institutions on risk-taking.Overall, we find limited evidence that politicalconstraints can be subsumed by our other insti-tutional variables. Indeed, limited levels of risk ofexpropriation and corruption, higher press free-dom, and a good socio-economic environmentdo not eliminate the impact of sound politicalinstitutions in shaping corporate risk-taking. InModel 5 of Table 4, when these variables are jointlyincluded, we find that political constraints con-tinue to be an important determinant of corporaterisk-taking, even if all the other measures areincluded.

Other alternative specificationsWe also perform a series of additional tests inunreported regressions. To ensure that our resultsare not driven by any particular country or timeperiod, we consider several subsample analyzes.The descriptive statistics show that our sample isdominated by American firms, followed by Britishfirms and Japanese firms. When we exclude thesecountries from the sample, which amount to morethan half of the observations, our results continueto hold. This evidence suggests that our models arenot driven by any particular country.

Prior research suggests that political rights haveimproved in recent years (e.g., Roe & Siegel, 2011).Therefore we limit our sample to include data from1996 and onward. The coefficient on POLITICALcontinues to load positively, and is statisticallysignificant. Hence our results are robust to theexclusion of early years, and they are not affectedby recent improvements in political rights.

Since our measure of political institutions isex ante, non-strategic firms – that is, those thatcan adapt their strategies and policies more easilythan strategic industries can – could react a prioriand adjust their activities accordingly. In this case,

our results may be driven by strategic industries(such as mining, steel, telecommunications, trans-portation, utilities, oil, and military-related pro-duction) lacking such flexibility. To mitigate thisconcern, we exclude strategic industries, and con-tinue to find a positive and significant rela-tion between political institutions and corporaterisk-taking.

In separate tests, we also control for the quality ofaccounting information measured by the earningsmanagement score (EARNINGSMGN) from Leuz,Nanda and Wysocki (2003). EARNINGSMGN iscomputed as the average rank across of:

(1) the country’s median ratio of the firm-levelstandard deviations of operating income andoperating cash flow;

(2) the country’s Spearman correlation betweenthe change in accruals and the change in cashflow from operations;

(3) the country’s median ratio of the absolute valueof accruals and the absolute value of cash flowsfrom operations; and

(4) the number of “small profits” divided by thenumber of “small losses” for each country.

We find that our main evidence remainsunchanged. Indeed, POLITICAL enters this regres-sion positive and statistically significant at the 1%level. As in John et al. (2008) and Faccio et al.(2011), we find that EARNINGSMGN loads negativeand is statistically significant. Overall, the resultsfrom these sensitivity tests reinforce our mainconclusion on the positive effect that politicalconstraints have on managerial risk-taking.

EndogeneityTo address the issue of endogeneity, we perform thefollowing tests. First, we use an instrumentalvariable approach for the main panel regression(Model 1 of Table 3) and for the country-levelregression (Model 6 of Table 4). The instrumentmust satisfy the conditions of relevance andexogeneity (Reeb, Sakakibara, & Mahmood, 2012).Following Larcker and Rusticus (2010), we searchfor a potential instrumental variable from thetheory of incomplete institutional contracts.Aghion, Alesina and Trebbi (2004) build on thistheory and model the delegation of power “as theshare of votes that can block the leader ex post whenhe tries to implement legislation”. Using a large setof developed and developing countries, the authorsfind strong empirical evidence that checks andbalances, for which the political constraints proxy


206


is an appropriate substitute, are more likely inhighly fragmented countries. Therefore we usecountry fragmentation as the first instrumentalvariable for political institutions. To measure thisinstrument, we use the country fractionalizationvariable (CNT_FRAC), which is equal to the princi-pal component of three fractionalization variables:ethnic, linguistic, and religious. All the fractionali-zation variables are drawn from Alesina, Dev-leeschauer, Easterly, Kurlat and Wacziarg (2003).The results of the first-stage regressions presentedin Models 1 and 5 of Table 5 confirm the evidencein Aghion et al. (2004) that political constraints arepositively related to country fractionalization. Thesecond-stage regressions in Models 2 and 6 ofTable 6 also show that the fitted values of politicalconstraints are positively related to corporate risk-taking.12

Second, we use another instrument from theHofstede (2001) study. Hofstede links politi-cal systems and institutions to the individualism–collectivism dimension of national culture. Thisdimension is considered by cross-cultural psychol-ogists to be the most significant and fundamentaldriver of cultural differences across societies (Triandis,2001). In high individualism societies, the politicalpower is exercised by voters, and is more balanced.Consistent with this argument, we find in Models 3and 7 that individualism (INDIVIDUALISM) is posi-tively related to political constraints.13 In Models 4and 8, we find that the fitted values of politi-cal constraints are positively related to corporaterisk-taking.

For the different models presented above, weconduct two tests to assess the appropriateness ofthe instrument. In particular, we perform the Stock

Table 5 Political institutions and corporate risk-taking: Endogeneity

Variable Panel regressions (RISK1) Country-level regressions (RISK4)

First

stage

Second

stage

First

stage

Second

stage

First

stage

Second

stage

First

stage

Second

stage

(1) (2) (3) (4) (5) (6) (7) (8)

Intercept 0.767*** �0.077*** 0.807*** �0.036*** 0.724 0.074* �0.122 0.079**

(56.437) (�7.641) (74.469) (�3.701) (1.506) (1.954) (�0.450) (2.069)

POLITICAL 0.248*** 0.204*** 0.059** 0.151***

(34.986) (42.726) (2.413) (3.783)

GDPG �0.075*** 0.022*** �0.057*** 0.018*** �0.031 0.000 0.036 0.001

(�54.170) (40.699) (�58.471) (39.499) (�0.808) (0.105) (1.116) (0.219)

LAWORDER 0.112*** �0.014*** 0.020*** �0.010*** 0.078 �0.001 0.044* �0.010***

(74.477) (�11.939) (16.546) (�11.451) (1.641) (�0.313) (1.821) (�2.826)

ROA �0.025*** �0.216*** �0.016*** �0.219*** �1.030 �0.253** �0.426 �0.332***

(�14.355) (�46.367) (�10.038) (�46.767) (�1.165) (�2.313) (�1.003) (�3.991)

DTA 0.012*** 0.003 0.003** 0.005 0.373 �0.070** 0.272 �0.048

(7.805) (0.867) (2.306) (1.465) (0.853) (�2.078) (0.880) (�1.172)

GROWTH 0.004*** 0.026*** �0.004*** 0.027*** �0.177*** 0.020* �0.151*** 0.021**

(6.172) (13.604) (�6.682) (14.381) (�5.384) (1.839) (�2.861) (2.218)

SIZE 0.007*** �0.020*** 0.004*** �0.019*** �0.034 0.000 0.014 �0.005*

(21.095) (�38.093) (12.540) (�36.337) (�1.298) (0.160) (0.950) (�1.958)

CNT_FRAC 0.053*** 0.093**

(58.400) (2.446)

INDIVIDUALISM 0.005*** 0.004**

(70.342) (2.476)

Years FE Yes Yes Yes Yes No No No No

Industry FE Yes Yes Yes Yes No No No No

Country FE No No No No No No No No

p-value 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000

R2 0.428 0.514 0.514 0.515 0.634 0.551 0.538 0.704

Observations 206,497 206,497 210,675 210,675 61 61 54 54

This table addresses the endogeneity of the political institutions using instrumental variables for both the panel regression and the country-levelregression. Each estimate is reported using robust t-statistics. The complete definitions and data sources for the variables are outlined in the Appendix.The notation ***, **, and * denotes statistical significance at the 1%, 5%, and 10% levels, respectively.


207


and Yogo weak identification test and the Kleiber-gen and Paap under-identification test. First, weconduct an F-test of the excluded exogenousvariables in the first-stage regressions. Specifically,we test the null hypothesis that the instrumentdoes not explain differences in political constraints.We reject this null hypothesis at the 1% level in allspecifications. Hence our instruments are notweakly correlated to the endogenous variable.Second, in the under-identification test, the Klei-bergen and Paap rk LM statistic produces a zerop-value, indicating that the excluded instruments areappropriate for the suspected endogenous regressor.

Subsample AnalysisA natural extension of our primary specification inTable 3 is to assess whether the relation between poli-tical institutions and corporate risk-taking dependson the extent of government extraction (our secondhypothesis). We follow Caprio et al. (2011) andDurnev and Fauver (2011) and consider two variablesreflecting the extent of government extraction (pre-dation), namely the ICRG assessment of the countrylevel of corruption (CORRUPTION), and the ICRGassessment of the country investment profile(EXPROP). CORRUPTION is an assessment of corrup-tion within the political system, and EXPROP is an

Table 6 Political institutions and corporate risk-taking: Subsample analysis

Variable Low

corruption

High

corruption

CORRUPTION �POLITICAL

Low

expropriation

High

expropriation

EXPROP �POLITICAL

(1) (2) (3) (4) (5) (6)

Intercept 0.202*** 0.313*** 0.276*** 0.274*** 0.287*** 0.314***

�(3.546) (16.452) (15.478) (7.080) (15.156) (17.035)

POLITICAL �0.013 0.031*** 0.083*** 0.000 0.053*** 0.007*

(�0.608) (3.348) (7.606) (0.011) (5.567) (1.659)

CORRUPTION �POLITICAL

0.014***

(6.277)

EXPROP � POLITICAL 0.005**

(2.559)

CORRUPTION �0.014***

(�8.854)

EXPROP �0.002

(�1.075)

GDPG 0.004*** 0.005*** 0.222*** 0.195*** 0.268*** 0.222***

(4.415) (8.729) (4.007) (3.061) (4.739) (4.037)

LAWORDER �0.014*** 0.003* 0.006* 0.001 0.015*** 0.006*

(�8.276) (1.686) (1.755) (0.183) (3.839) (1.762)

ROA �0.239*** �0.200*** 0.027*** 0.031*** 0.019*** 0.027***

(�48.353) (�30.919) (13.903) (12.393) (7.950) (13.899)

DTA 0.007** 0.005 �0.018*** �0.022*** �0.014*** �0.018***

(2.052) (1.082) (�33.248) (�27.135) (�24.348) (�33.135)

GROWTH 0.024*** 0.030*** 0.004*** �0.002** 0.002*** 0.003***

(11.452) (11.072) (7.602) (�1.996) (2.903) (5.610)

SIZE �0.016*** �0.021*** �0.009*** 0.000 �0.015*** �0.007***

(�29.177) (�26.050) (�8.305) (0.083) (�10.415) (�7.012)

Years FE Yes Yes Yes Yes Yes Yes

Industry FE Yes Yes Yes Yes Yes Yes

Country FE Yes Yes Yes Yes Yes Yes

p-value 0.000 0.000 0.000 0.000 0.000 0.000

R2 0.506 0.530 0.516 0.519 0.511 0.516

Observations 110,953 100,841 211,794 102,598 109,196 211,794

This table provides estimated coefficients from regressing the corporate risk-taking variable on the political institution proxy and various country- andfirm-specific control variables for different subsamples. The complete definitions and data sources for the variables are outlined in the Appendix. Eachestimate is reported using robust t-statistics clustered at the firm level. The notation ***, **, and * denotes statistical significance at the 1%, 5%, and 10%levels, respectively.


208


assessment of factors such as expropriation andcontract repudiation that affect the risk to invest,other than the typical political, economic and finan-cial risk components. Both variables are scaled so thathigh values reflect high government extraction.

To test our second hypothesis, we first split thesample at the median value for CORRUPTION (Models1 and 2) and for EXPROP (Models 4 and 5). We nextinteract POLITICAL with CORRUPTION in Model 3,and with EXPROP in Model 6. Our results show thatPOLITICAL loads positively and is statistically sig-nificant at the 1% level only for the subsample ofhigh-corrupt countries and for countries with highexpropriation risk. Thus, in high-corrupt and –expropriation environments, strong political institu-tions can boost managerial risk-taking in corporateinvestment decisions. When we interact POLITICALwith CORRUPTION and with EXPROP, the interactionterms are both positive and statistically significant,suggesting that sound political institutions compen-sate for the weakness of the institutional environmentin encouraging corporate risk-taking, and support theresults of the subsample analysis discussed above.Overall, the results corroborate our earlier findings ofa positive and significant relation between politicalinstitutions and corporate risk-taking, but also showthat this relation is more pronounced when govern-ment extraction is high.

Evidence on the Relation between PoliticalConnections and Corporate Risk-TakingIn this section, we investigate the effect of politicalconnections on corporate risk-taking. The literatureprovides ample evidence on the sensitivity ofpolitically connected firms to the political environ-ment prevailing in the country. For example, Faccio(2006) finds that politically connected firms repre-sent almost 8% of the world’s stock marketcapitalization, and that these firms are pervasivein both democratic and non-democratic countries.

We follow the manager–state collusion model ofShleifer and Vishny (1994) to gain an understand-ing of whether politically connected firms have amore or less conservative approach to risk-taking.Specifically, we estimate the following model (sub-scripts suppressed for notational convenience):

RISK1 ¼ aþ mCONNECTEDþ g1POLITICAL

þ g2FIRMCONTROLSþ g3COUNTRYCONTROLS

þXY�1

Y¼1

YEARþXK�1

K¼1

INDþXC�1

C¼1

CNT þ Z

ð2Þ

where CONNECTED is a dummy variable thatequals 1 if the firm is politically connected. Theremaining variables are defined as in Eq. (1). Ourmain focus is the coefficient m, which tests thesensitivity of corporate risk-taking to political con-nections. Similar to Faccio (2006), we restrict ouranalysis to the period after 1996, and to the 45countries included in her study. We manuallymatch her database of politically connected firmswith our database. The final sample of politicallyconnected firms consists of 290 firms and 2922firm-year observations from 29 countries.

Table 7 provides our results. Model 1 is ourprimary specification. The results show that CON-NECTED is positively and significantly associatedwith corporate risk-taking (at the 1% level), sug-gesting that politically connected firms have highearnings volatility.14 This result is also economic-ally significant. Moving the variable CONNECTEDfrom 0 to 1 increases risk-taking by 30%, from 0.10to 0.13. The variable POLITICAL continues to loadpositively, and is statistically significant at the 5%level.

The remaining models in Table 7 present addi-tional analyses and robustness of our main find-ings. One potential concern with the regressionreported in Model 1 is that the dummy variableCONNECTED may not be exogenous. Specifically,some unobserved determinants of firms’ risk-takingmay also explain political connections, causing ourreported OLS estimates to be biased and incon-sistent. We address the issue of endogeneity usingan instrumental variables estimation approach. Inparticular, we specify the firm’s location as aninstrument for political connections. This choice ismotivated by studies that provide evidence on theinfluence of firm location on political connections(e.g., Boubakri et al., 2012). In the first-stage regre-ssion, we predict political connections via probitestimation using the location of a company’sheadquarters (CAPITAL), where CAPITAL is a dum-my variable that equals 1 if the company isheadquartered in the capital of its country. Thefirst-stage fitted values for political connectionsare then used in the second-stage OLS regression.The results, unreported owing to space constraints,show that, in the first stage, the presence of a firm’sheadquarters in the capital city is a good predictorof political connections. Model 2 shows that, in thesecond-stage regression, the instrumented valueof connections is positive and statistically signifi-cant. This result reinforces our main inferences onthe impact of political connections on corporate


209


risk-taking, and suggests that politically connectedfirms have more volatile earnings than their non-connected counterparts.15

Next, in unreported regressions, we attempt toaccount for the fact that it is difficult to identifyexactly when connections are established (Chaneyet al., 2011; Faccio, 2006). We address this issueby running the analysis in different time periods.First, we run our analysis using the period 1997 to2001, as in Chen, Ding and Kim (2010), and findqualitatively similar results. Second, we repeat theanalysis for the period after 1991, as some connec-tions may have been established before 1997. Wefind that the coefficient on CONNECTED continuesto load positive, and is statistically significant at the1% level. Hence our results are not driven by thetime period considered.

Models 3 and 4 utilize alternative dependentvariables to study the impact of political connec-tions on corporate risk-taking. In Model 3 we use

the variable RISK2 (Max (ROA) – MIN (ROA)), and inModel 4 we follow John et al. (2008) and sacrificethe panel structure of our data. Specifically, as inChaney et al. (2011), we calculate the volatility ofearnings for each firm for the 1996 to 2005 period(cross-sectional regression).16 We also consider theaverage asset growth over the same period. Thisprocedure allows each firm to enter only once inour regression, and the number of observationsdrops from 122,804 firm-year observations to19,090 firm observations. All the independentvariables enter into the regression at the first year-end of the sample period over which the corporaterisk-taking proxy is measured. The results confirmour main finding that politically connected firmstake more risk than their non-connected counter-parts.17

In Models 5 and 6 we split our sample at themedian level of POLITICAL. We find that CON-NECTED loads positive and is statistically significant

Table 7 Political connectedness and corporate risk-taking

Variable Primary

specification

Second

stage

RISK2 Max(ROA)

– Min(ROA)

RISK3

Cross-sectional

regression

POLITICAL

Low

POLITICAL

High

BAILOUT

Yes

BAILOUT

No

(1) (2) (3) (4) (5) (6) (7) (8)

Intercept 0.354*** 0.279*** 0.846*** 0.255*** 0.390** 0.167*** 0.374*** 0.036**

(16.917) (14.900) (18.603) (7.012) (2.361) (10.470) (16.678) (2.407)

CONNECTED 0.029*** 0.305** 0.064*** 0.041*** 0.046*** 0.003 0.040*** 0.001

(5.412) (2.049) (5.283) (6.634) (3.215) (0.801) (5.955) (0.370)

POLITICAL 0.020** 0.024*** 0.054** 0.094*** 2.137* 0.036*** 0.022** 0.042***

(2.131) (3.296) (2.525) (2.921) (1.800) (4.243) (2.027) (3.719)

GDPG �0.000 0.003*** �0.000 0.007*** �0.001 0.002*** 0.000 0.003***

(�0.358) (5.901) (�0.221) (3.618) (�0.334) (4.300) (0.111) (5.616)

LAWORDER �0.003** �0.003*** �0.011*** 0.052*** �0.025*** �0.005*** �0.004*** �0.001

(�2.464) (�2.778) (�3.865) (11.721) (�4.265) (�3.751) (�2.726) (�0.809)

ROA �0.196*** �0.227*** �0.446*** �0.014*** �0.188*** �0.262*** �0.192*** �0.107***

(�32.609) (�47.257) (�32.952) (�2.653) (�29.616) (�38.397) (�31.177) (�6.081)

DTA 0.001 0.003 �0.009 �0.002 �0.000 0.009*** 0.001 �0.001

(0.192) (0.964) (�0.977) (�1.140) (�0.022) (2.865) (0.129) (�0.355)

GROWTH 0.031*** 0.025*** 0.067*** �0.000 0.031*** 0.027*** 0.031*** 0.029***

(12.610) (14.430) (12.408) (�0.275) (11.928) (10.198) (12.245) (8.310)

SIZE �0.021*** �0.017*** �0.050*** �0.039*** �0.024*** �0.007*** �0.023*** �0.004***

(�27.524) (�15.420) (�28.918) (�39.183) (�26.186) (�16.552) (�26.813) (�11.791)

Year FE Yes Yes Yes Yes Yes Yes Yes Yes

Industry FE Yes Yes Yes Yes Yes Yes Yes Yes

Country FE Yes Yes Yes Yes Yes Yes Yes Yes

p-value 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000

R2 0.523 0.480 0.526 0.354 0.505 0.529 0.507 0.228

Observations 122,804 113,250 122,804 19,090 55,885 66,919 87,930 34,874

This table provides estimated coefficients from regressing risk-taking measures on the political connectedness proxy and various country- and firm-specific control variables. The data cover the period from 1997 through 2008. The complete definitions and data sources for the variables are outlined inthe Appendix. Each estimate is reported using robust t-statistics clustered at the firm level. The notation ***, **, and * denotes statistical significance atthe 1%, 5%, and 10% levels, respectively.


210


at the 1% level for only the subsample of firmslocated in weak political institutions countries(Model 5). This finding suggests that, absent poli-tical constraints on policymakers, extraction ofbenefits and outright expropriation of corporateresources are unhindered in politically connectedfirms, leading the impact of these connections to belarger in such environments.

Finally, Faccio et al. (2006) show that politicallyconnected firms are more likely to be bailed out bythe government in cases of distress. Based on thisfinding, we expect that politically connected firmsin countries that rescue them in case of trouble willbe more likely to undertake risky projects. Indifferent specifications, and based on informationfrom Faccio et al. (2006), we split our sample intotwo groups: countries that bailed out connectedfirms, and those that never provided such support.The results are provided in Models 7 and 8, andindicate that politically connected firms have morevolatile earnings in the set of countries that providerescue in case of turbulence; indeed, CONNECTEDloads positive and significant at the 1% level inModel 7. Overall, the results reported in this sectionshow that politically connected firms have morevolatile earnings, and are thus associated withgreater risk-taking. These results are robust toendogeneity issues, using different time periods,different samples, and alternative dependent vari-ables. Our results also show that connected firms incountries that rescue them in case of trouble aremore likely to undertake risky activities.

CONCLUSIONFollowing recent evidence advocating the impor-tance of politics in finance, we adopt a microeco-nomic perspective and analyze the politicaldeterminants of managerial risk choices in corpo-rate investment decisions. Our investigation isgrounded in Stulz’s (2005) model of the strategicbehavior of corporate insiders in response topredatory government practices. We hypothesizethat political institutions influence managerial risk-taking. However, a priori it is unclear whetherpolitical institutions have a positive or negativeeffect on managers’ risk choices. Our empiricalanalysis is applied to the non-financial firms from77 countries covered by Compustat over the period1988 to 2008.

Our results show that the risk choices at the firmlevel are affected by the political institutions pre-vailing in the country. Specifically, we report strongand robust evidence that sound political institutions

(i.e., effective checks and balances) are positivelycorrelated with corporate risk-taking. We also findthat this relation strengthens when governmentextraction is high. Our results also show that poli-tically connected firms have relatively more volatileearnings, suggesting that close ties to the govern-ment lead to less conservative investment choices.

Our paper extends the literature on the determi-nants of firm-level managerial risk choices byexamining the effect of the prevailing country’spolitical institutions. Moreover, by providing anempirical test of Stulz’s (2005) theoretical twin-agency model, we add to the evidence on thestrategic behavior of corporate insiders to facegovernment predation. Finally, we contribute tothe literature on political connections by identify-ing yet another impact on firms’ behavior. We thuspresent evidence that corporate risk-taking ispolitically determined at the country level via thepolitical institutions prevailing in the country, andat the firm level through political connectedness.

ACKNOWLEDGEMENTSWe thank Najah Attig, John Cantwell (Chief Editor),Sadok El Ghoul, Omrane Guedhami, Jeffrey Pittman,Samir Trabelsi, and especially David Reeb (Area Editor)for their insightful comments and suggestions, whichgreatly improved the paper. We appreciate financialsupport from Canada’s Social Sciences and HumanitiesResearch Council, as well as excellent researchassistance from Heba Abu Ghazalah, and MiraMneimneh. All remaining errors are ours.

NOTES1Greater political constraints on policymakers reflect

stronger political institutions (Henisz, 2000).2Labor unions are found, for instance, to be less

influential under less constrained governments (e.g.,Pagano & Volpin, 2005; Roe, 2003). Therefore risk-taking by managers is more likely to be higher undermore authoritarian governments.

3Alternatively, since managers of politically con-nected firms are not as closely monitored by share-holders as those of their non-connected peers, they arelikely to be more entrenched and more powerful,owing to their connections. Consequently, managersof connected firms may have lower incentives toengage in risky projects.

4All 39 countries covered in the John et al. (2008)study are included in our sample.

5Henisz’ (2010) dataset used to estimate thepolitical constraint index is an updated version of hisearlier work (i.e., Henisz, 2000).


211


6See, for example, Stulz (2005) and Qi et al. (2010)for the use of the political constraints index.

7The Gastil indices measure the degree of democ-racy in a given country, but are not necessarilycorrelated with the degree of commitment to privateproperty rights, and the widely used political instabilitymeasures are not suitable when one considers auto-cratic regimes.

8For a detailed description of the political connec-tions database, see Faccio (2006). Examples of politicalconnections from developed economies include firmsconnected to the Italian prime minister, Silvio Berlus-coni. Connections in Malaysia and Indonesia relatemainly to the prime minister Mohamad Mahathir andPresident Suharto, respectively.

9Our results remain qualitatively similar when weinclude outliers. Indeed, in the primary specification ofTable 3, POLITICAL loads positive (¼ 0.042) and isstatistically significant at the 1% level (t-stat ¼ 4.52).

10The last year of entrance of the independentvariables is 2004 for RISK1 calculated over 2004–2008.

11Note that in some specifications in John et al.(2008) there is a negative but insignificant relationbetween LAWORDER and corporate risk-taking.

12We exclude the country fixed effect from theregressions, given that CNT_FRAC does not vary acrosstime for any given country. See also Faccio et al.(2011) for a similar approach.

13Hofstede’s individualism–collectivism index iswidely used in the international business literature(e.g., Brewer & Venaik, 2011; Lim, Leung, Sia, & Lee,2004; Morris, Davis, & Allen, 1994).

14Our results are qualitatively similar when weexclude countries without a connection.

15We perform the Stock and Yogo weak identifica-tion test and the Kleibergen and Paap under-identifi-cation test, and find that CAPITAL is an appropriateinstrument.

16Chaney et al. (2011) analyze the quality ofaccounting information in politically connected firms,and calculate the standard deviation of residualsduring 1996–2005.

17We also consider different country subsamples.Excluding the largest number of politically connectedfirms in the United Kingdom and politically connectedfirms from East Asian countries (Thailand, Indonesia,and Malaysia) in separate regressions, we find similarresults.

REFERENCESAghion, P., Alesina, A., & Trebbi, F. 2004. Endogenous political

institutions. Quarterly Journal of Economics, 119(2): 565–611.Alesina, A., Devleeschauer, A., Easterly, W., Kurlat, S., &

Wacziarg, R. 2003. Fractionalization. Journal of EconomicGrowth, 8(2): 155–194.

Baird, I., & Thomas, H. 1985. Toward a contingency model ofstrategic risk taking. Academy of Management Review, 10(2):230–243.

Beck, T., Clarke, G., Groff, A., Keefer, P., & Walsh, P. 2001.New tools in comparative political economy: The databaseof political institutions. World Bank Economic Review, 15(1):165–176.

Boubakri, N., Cosset, J., & Saffar, W. 2012. The impact ofpolitical connections on firms’ operating performanceand financing decisions. Journal of Financial Research, 35(3):397–423.

Brewer, P., & Venaik, S. 2011. Individualism–collectivism inHofstede and GLOBE. Journal of International Business Studies,42(3): 436–445.

Bromiley, P. 1991. Testing a causal model of corporate risktaking and performance. Academy of Management Journal,34(1): 37–59.

Campbell, J. 1996. Understanding risk and return. Journal ofPolitical Economy, 104(2): 298–345.

Caprio, L., Faccio, M., & McConnell, J. 2011. Shelteringcorporate assets from political extraction. Journal of Law,Economics, and Organization, advance online publication10 October. doi: 10.1093/jleo/ewr018.

Chaney, P., Faccio, M., & Parsley, D. 2011. The quality ofaccounting information in politically connected firms. Journalof Accounting and Economics, 51(1–2): 58–76.

Chen, C. J. P., Ding, Y., & Kim, C. 2010. High-level politicallyconnected firms, corruption, and analyst forecast accuracyaround the world. Journal of International Business Studies,41(9): 1505–1524.

Durnev, A., & Fauver, L. 2011. Stealing from thieves: Firmgovernance and performance when states are predatory. Work-ing Paper, McGill University.

Faccio, M. 2006. Politically connected firms. American EconomicReview, 96(1): 369–386.

Faccio, M., Marchica, M., & Mura, R. 2011. Large shareholderdiversification and corporate risk-taking. Review of FinancialStudies, 24(11): 3601–3641.

Faccio, M., Masulis, R., & McConnell, J. 2006. Political con-nections and corporate bailouts. Journal of Finance, 61(6):2597–2635.

Fogel, K., Morck, R., & Yeung, B. 2008. Big business stability andeconomic growth: Is what’s good for General Motors good forAmerica? Journal of Financial Economics, 89(1): 83–108.

Henisz, W. 2000. The institutional environment for economicgrowth. Economics and Politics, 12(1): 1–31.

Henisz, W. 2010. The political constraint index (POLCON) dataset. Philadelphia: University of Pennsylvania.

Hofstede, G. 2001. Culture’s consequences: Comparing values,behaviors, institutions and organizations across nations, (2ndedn), Thousand Oaks, CA: Sage.

Jensen, M. C., & Meckling, W. H. 1976. Theory of the firm:Managerial behavior, agency costs, and ownership structure.Journal of Financial Economics, 3(4): 305–360.

John, K., Litov, L., & Yeung, B. 2008. Corporate governance andrisk-taking. Journal of Finance, 63(4): 1679–1728.

Knack, S., & Keefer, P. 1995. Institutions and economicperformance: Cross-country tests using alternative institutionalmeasures. Economics and Politics, 7(3): 207–227.

La Porta, R., Lopez-de-Silanes, F., Shleifer, A., & Vishny, R.1998. Law and finance. Journal of Political Economy, 106(6):1113–1150.

Larcker, D., & Rusticus, T. 2010. On the use of instrumentalvariables in accounting research. Journal of Accounting andEconomics, 49(3): 186–205.


212


Leuz, C., Nanda, D., & Wysocki, P. 2003. Earnings managementand investor protection: An international comparison. Journalof Financial Economics, 69(3): 505–527.

Lim, K. H., Leung, K., Sia, C. L., & Lee, M. K. 2004. Is eco-mmerce boundary-less? Effects of individualism–collectivismand uncertainty avoidance on Internet shopping. Journal ofInternational Business Studies, 35(6): 545–559.

Morris, M. H., Davis, D., & Allen, J. W. 1994. Fostering corporateentrepreneurship: Cross-cultural comparisons of the impor-tance of individualism versus collectivism. Journal of Interna-tional Business Studies, 25(1): 65–89.

Murphy, K. M., Shleifer, A., & Vishny, R. 1991. The allocation oftalent: Implications for growth. Quarterly Journal of Economics,106(2): 503–530.

Pagano, M., & Volpin, P. 2005. The political economy of corporategovernance. American Economic Review, 95(4): 1005–1030.

Qi, Y., Roth, L., & Wald, J. 2010. Political rights and the cost ofdebt. Journal of Financial Economics, 95(2): 202–226.

Rajan, R., & Zingales, L. 2003. The great reversals: The politics offinancial development in the twentieth century. Journal ofFinancial Economics, 69(1): 5–50.

Reeb, D., Sakakibara, M., & Mahmood, I. P. 2012. From theEditors: Endogeneity in international business research. Journalof International Business Studies, 43(3): 211–218.

Roe, M. 2003. Political determinants of corporate governance.Oxford: Oxford University Press.

Roe, M. 2006. Legal origins, politics, and modern stock markets.Harvard Law Review, 120(2): 460–527.

Roe, M., & Siegel, J. 2011. Political instability: Effects on financialdevelopment, roots in the severity of economic inequality.Journal of Comparative Economics, 39(3): 279–309.

Shapira, Z. 1995. Risk taking: A managerial perspective. NewYork: Russel Sage Foundation.

Shleifer, A., & Vishny, R. 1994. Politicians and firms. QuarterlyJournal of Economics, 109(4): 995–1025.

Shleifer, A., & Vishny, R. 1997. A survey of corporate govern-ance. Journal of Finance, 52(2): 737–777.

Stulz, R. 2005. The limits of financial globalization. Journal ofFinance, 60(4): 1595–1638.

Tan, J. 2001. Innovation and risk-taking in a transitional economy:A comparative study of Chinese managers and entrepreneurs.Journal of Business Venturing, 16(4): 359–376.

Triandis, H. C. 2001. Individualism–collectivism and personality.Journal of Personality, 69(6): 907–924.

Wright, P., Ferris, S., & Awasthi, V. 1996. Impact of corpo-rate insider, blockholder, and institutional equity ownershipon firm risk-taking. Academy of Management Journal, 39(2):441–463.


213


Appendix

Table A1 Variable definitions

Variable Definition Data source

Corporate risk-taking

RISK1 Computed as the country-adjusted standard deviation of the firm’s

profitability (ROA) over 5-year overlapping periods starting in 1988 and

ending in 2008, where ROA is measured as the ratio of earnings before

interest, taxes, depreciation, and amortization to total assets.

Compustat

RISK2 Computed as the difference between the maximum and minimum ROA

reported over a 5-year interval.

Compustat

RISK3 Computed as the country-adjusted standard deviation of the ROA for

each firm over the entire sample period, requiring a minimum of five

observations in the cross-sectional regressions, as in John et al. (2008).

Compustat

RISK4 Considers the country-level average of the firm-level observations of RISK1. Compustat

Country-level variables

POLITICAL Measures the degree of political constraint of a country. Derived from a

model of political interaction that incorporates information on the number

of independent branches of government with veto power, and the

distribution of preferences across and within those branches. Government

branches considered are chief executives, lower house of legislature, higher

house of legislature, judiciary, and sub-federal branches. Higher scores

indicate stronger political constraints and sound political institutions.

Henisz (2010)

GOVFRAC Government fractionalization, computed as the chance that two random

draws will produce legislators from different parties.

Beck et al. (2001)

POLARIZ Maximum difference of orientation among government parties. This

variable ranges from 0 to 2, with higher scores indicating greater

polarization among parties.

Beck et al. (2001)

LAWORDER Assessment of the law and order tradition in the country. This variable

ranges from 0 to 6, with higher scores indicating greater rule of law in the

country.

ICRG (2008)

SOCIOECON Assessment of the socio-economic pressures at work in society that could

constrain government action or fuel social dissatisfaction. The risk rating

assigned is the sum of three subcomponents, each with a maximum score

of 4 points and a minimum score of 0 points. A score of 4 points equates to

“very low risk”, and a score of 0 points to “very high risk”. The

subcomponents are unemployment, consumer confidence, and poverty.

ICRG (2008)

PRESSFREEDOM An index of freedom of the press. Higher scores mean greater freedom of

the print and broadcast media in a country. The index is time-varying, and

ranges from 0 (not free) to 2 (free).

Freedom House

(2007)

EXPROP Assessment of factors affecting the risk to investment that are not covered

by other political, economic or financial risk components. The

subcomponents are contract viability/expropriation, profits repatriation,

and payment delays. This variable ranges from 0 to 12, with higher scores

indicating higher risks.

ICRG (2008)

INDIVIDUALISM Hofstede’s cultural index for individualism. Hofstede (2001)

CNT_FRAC The principal component of three measures of country-level

fractionalization (ethnic, linguistic, and religious).

Alesina et al.

(2003)

CORRUPTION Assessment of the corruption in government. This variable ranges from

0 to 6, with high scores indicating that high government officials are likely

to demand special payments, and illegal payments are generally expected

throughout lower levels of government in the form of bribes connected

with import and export licenses, exchange controls, tax assessment, policy

protection, or loans.

ICRG (2008)


214


ABOUT THE AUTHORSNarjess Boubakri is Professor of Finance at theAmerican University of Sharjah. Prior to that, shewas faculty member at HEC Montreal (Canada). Sheholds a PhD in finance from Laval University(Canada). Dr Boubakri’s work has been publishedin major finance, accounting and internationalbusiness journals. Her research interests include,among others, privatization, international corpo-rate governance, and political connections.

Sattar A Mansi is a Wells Fargo Professor of Financeat Virginia Tech’s business school. He has over20 years of research and private sector experience.

Dr Mansi is well published in the most influentialjournals in finance, accounting, law, and interna-tional business. His private sector experienceincludes working with Freddie Mac and UnisysCorporation. Dr Mansi has a PhD in finance fromthe George Washington University.

Walid Saffar is an Assistant Professor of Finance atThe Hong Kong Polytechnic University. Born inTunisia, he holds both Canadian and Tunisian citizen-ships. He holds a PhD in finance from HEC Montreal,Canada. His research interests cover primarily priva-tization, political connections, and corporate govern-ance. He can be reached at [email protected].

Accepted by David Reeb, Area Editor, 6 January 2013. This paper has been with the authors for four revisions.

Table A1 Continued

Variable Definition Data source

Country-level variables

DEMOCRATIC Measures how responsive a government is to its people, on the basis

that the less responsive it is, the more likely it is that the government

will fall, peacefully in a democratic society, but possibly violently in

a non-democratic one. This variable ranges from 0 to 6.

ICRG (2008)

BAILOUT A dummy variable that takes the value of 1 if the country bails out politically

connected firms, and 0 otherwise.

Faccio et al. (2006)

GDPG Growth rate of GDP. The latter is measured in 1995 constant US$. World Bank

Firm-level variables

CONNECTED Dummy variable; equals 1 for politically connected firms and 0 otherwise. Faccio (2006)

ROA Ratio of earnings before interest, tax, depreciation, and amortization to total

assets.

Compustat

DTA Ratio of total debt to total assets. Compustat

GROWTH Average assets growth over 5 years, where assets are converted into US$. Compustat

SIZE Log of total assets in millions of US$. Compustat

CAPITAL Dummy variable; equals 1 for firms located in the capital, and 0 otherwise. Compustat

OWNERSHIP Equity ownership of the largest shareholder. OSIRIS

Bureau Van Dijk

This Appendix provides definitions for the variables used in the analysis, along with their data sources. The data cover the period from 1988 through2008.


215


Political institutions, connectedness, and corporate risk-taking · 2017-09-09 · Political institutions, connectedness, and corporate risk-taking Narjess Boubakri1, Sattar A Mansi2

Documents