## _wp12210 - 1. Correlation Matrix of National Cultural Dimensions

## Source details

**Canonical URL:** [_wp12210 - 1. Correlation Matrix of National Cultural Dimensions](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2012/_wp12210.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2012/_wp12210.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2012/_wp12210.pdf.json)

---

### Introduction and scope
- Sample and focus:
  - International sample of 50,000 firms across >400 SIC industries in 51 countries.
  - Uses four Hofstede (2001) national cultural dimensions (UAI, IDV, PDI, MAS) to analyze effects on corporate risk-taking.
  - Period covered: 2000–2012.
- Methodology:
  - Hierarchical linear mixed model (HLM) to model firm-, industry-, and country-level influences and cross-level interactions.
  - Dependent variables: z-score (natural log of Z), σ(ROA), R&D expenditures to market capitalization.

### Core findings (direct effects)
- Culture exerts direct effects on corporate risk-taking even after controlling for institutional, economic, industry, and firm-level factors.
- Direction and significance of primary cultural dimensions:
  - Uncertainty Aversion (UAI): negative and significant across models (supports H1).
  - Individualism (IDV): positive and significant across models (supports H2).
  - Power Distance (PDI): negative and significant for z-score and R&D, less robust for σ(ROA) (partial support for H3).
  - Masculinity (MAS): generally not significant in HLM specifications (H4 rejected).
- Economic magnitudes (effects reported using one standard deviation changes):
  - One standard deviation in uncertainty aversion (21.7) associated with:
    - Change in z-score: 0.1085 (0.005*21.7); mean z-score = 0.019; stdev = 1.12.
    - Change in σ(ROA): 4.1447 (0.191*21.7); mean σ(ROA) = 23.367; stdev = 51.18.
    - Change in R&D index: 0.2387 (0.011*21.7); mean R&D index = 0.02; stdev = 0.04.
  - One standard deviation in individualism (28.9) associated with:
    - Change in z-score: 0.2312 (0.008*28.9).
    - Change in σ(ROA): 6.5314 (0.226*28.9).
    - Change in R&D index: 9.9127 (0.343*28.9).
  - One standard deviation in power distance (18.7) associated with:
    - Change in z-score: 0.0187 (0.001*18.7).
    - Change in R&D index: 2.6741 (0.143*18.7).

### Theoretical channels and expectations
- Potential channels through which culture affects risk-taking:
  - Macro channels: institutional, legal, and economic development (creditor rights, investor protection, judicial efficiency, corporate governance, bankruptcy protection, transparency, corruption).
  - Market-structure channels: industry informational opacity, concentration/diversification.
  - Micro channels: managerial decision-making, board composition, leadership structure.
- Expectation highlighted:
  - Cultural traits (uncertainty aversion, individualism) expected to have stronger effects in industries with greater informational opacity (e.g., IT, financial services, oil extraction, chemicals).

### Data and measurement
- National culture:
  - Hofstede dataset based on IBM employee survey (1967–1973) across 66 countries; four dimensions: UAI, IDV, PDI, MAS.
  - Scale: 1 to 100.
  - Mean differences by income group (selected):
    - UAI: low income = 70; high income = 64 (mean-comparison t-test at the 5 percent level: low income 70 is six points higher than high income).
    - IDV mean: high income = 57; low income = 25; 27 point mean-difference significant at the 5 percent level.
    - PDI mean: low income = 67; high income = 48; 19 point mean-difference significant at the 5 percent level.
    - MAS mean: high income = 51; low income = 42; 9 point difference not statistically significant at the 5 percent level.
  - Pair-wise correlations (Table 1 excerpts):
    - Individualism vs. Power Distance: -0.66* (partial corr controlling for economic development: -0.27).
    - Uncertainty Aversion vs. Individualism: -0.21 (-0.30).
    - Power Distance vs. Uncertainty Aversion: 0.15 (0.16).
    - Masculinity correlations: with UAI 0.05 (0.05); with IDV 0.09 (0.14); with PDI 0.07 (0.09).
    - (*) denotes significance at the 1 percent level.
- Firm risk-taking measures (CVU database):
  - Z-score = (ROA + CAR) / σ(ROA); z-score denotes ln(Z).
  - σ(ROA): standard deviation of ROA (EBIT/total assets), computed as deviation from industry-year-country mean; requirement: earnings and assets available for at least eight years over 1999–2012.
  - R&D index = R&D expenditures / market capitalization.
  - CVU coverage: annual indicators 2000–2012, ~50,000 publicly traded firms, >400 SIC industries, 51 countries.
- Industry informational opacity:
  - Follows Huang (2008) using U.S. weekly returns 2000–2005; opacity = ln(1- R^2/ R^2).
  - Selected industry entries (R-squared; Informativeness Index; Opacity Index):
    - Finance, Insurance, And Real Estate: 0.0034; 0.9966; 5.684
    - Wholesale Trade: 0.0039; 0.9961; 5.550
    - Mining: 0.0042; 0.9958; 5.477
    - Manufacturing: 0.0085; 0.9915; 4.756
  - United States excluded from cross-country regressions since U.S. informativeness used as proxy.

### Controls and constructed indices
- Country-level controls (13 listed): credit to GDP; policy rate; aggregate economic volatility; rule of law; regulatory quality; government effectiveness; voice and accountability; control of corruption; political stability; efficiency of debt enforcement; creditor rights; shareholder protection; bankruptcy costs.
- Formal Institutions Index (PCA-derived):
  - Formal Institutions Index = Rule of Law*0.41 + Efficiency of Debt Enforcement*0.34 + Control of Corruption*0.40 + Regulatory Quality*0.39 + Government Effectiveness*0.40 + Voice and Accountability*0.36 + Political Stability*0.31.
- Industry-level controls:
  - Herfindahl index (industry concentration), industry informational opacity.
- Firm-level controls:
  - Firm size (log total assets), leverage (total liabilities/total assets), profitability (ROA = EBIT/total assets), sales growth (annual log growth), dependence on external finance (Rajan-Zingales index).
- Data treatment:
  - Winsorize firm-level controls at the 1% level in both tails; sales growth and leverage also winsorized at top 5% of distribution.
  - Final sample reported: 50,000 firms across 400 industries in 51 countries.

### Key empirical findings (indirect effects, institutions, and industry links)
- Formal Institutions Index:
  - Positive and significant across models — better formal institutions associated with more firm risk-taking.
- Shareholders' rights:
  - Positive and significant in several specifications (examples: 17.687** in stdev(ROA) HLM column).
- Creditor rights:
  - Associated with lower corporate risk-taking and lower σ(ROA) (example: -3.525**).
- Industry effects:
  - Industry competition (Herfindahl): positive association with firm risk-taking.
  - Industry informational opacity: negative association with firm risk-taking on average.

### Moderating / interaction effects (H5–H7 results)
- H5 (Individualism × Formal Institutions; Power Distance × Formal Institutions):
  - IDV × Formal Institutions: positive and significant (examples: 0.006** inverse z-score; 0.155* σ(ROA); 0.259* R&D) — individualism strengthens the positive relationship between formal institutions and corporate risk-taking.
  - PDI × Formal Institutions: interaction not significant in most models.
- H6 (Individualism/Power Distance via bankruptcy/creditor/shareholder rights):
  - No consistent pattern; mixed interaction results; authors reject H6 overall.
  - Selected interactions: IDV × Creditor Rights = -0.009*** (inverse z-score); IDV × Shareholders' Rights = 0.037*** (inverse z-score); PDI × 1/Eff. of Debt Enforcement = -0.118* (inverse z-score).
- H7 (Culture × Industry Opacity):
  - UAI × Industry Opacity: negative and significant across models (examples: -0.001* inverse z-score; -0.006** σ(ROA); -0.058* R&D) — higher UAI in opaque industries → even less risk-taking.
  - IDV × Industry Opacity: highly positive and significant across models (examples: 1.185* inverse z-score; 27.323** σ(ROA); 94.36* R&D) — in highly opaque industries, firms in highly individualistic societies take even more risk (authors note this is surprising and may reflect competitive dynamics).
  - PDI and MAS interactions with industry opacity: generally not significant.

### Foreign versus domestic firms
- Foreign-firm sample limited to ~500 foreign firms; results tentative.
- Key finding:
  - Risk-taking by foreign subsidiaries is better explained by parent (origin) country cultural norms than by host country culture.
  - Example coefficients (foreign firms, origin country effects): Individualism (origin) positive and significant for σ(ROA) and R&D (e.g., 0.554*** σ(ROA)); Power Distance (origin) negative and significant across models (examples: -0.031*** inverse z-score; -0.778** σ(ROA); -0.231*** R&D).
- Caveats: small sample; does not control for tenure of foreign presence or cultural composition of management.

### Robustness, specification, and regression evidence (Appendix C highlights)
- Estimation framework:
  - HLM models reported alongside OLS with clustered SEs; significance notation: ***, **, * for 1, 5, 10 percent levels respectively.
- Selected coefficient patterns across dependent variables:
  - Inverse z-score regressions (Table C1): UAI coefficients range from -0.001*** to -0.005*** across specifications; IDV coefficients range from 0.002*** to 0.008***; PDI generally negative and significant; MAS generally small/insignificant.
  - σ(ROA) regressions (Table C2): UAI strongly negative (examples: Basic OLS -0.400*** [0.011]; Complex OLS -0.577*** [0.012]); IDV strongly positive (examples: 0.453*** [0.011] to 0.726*** [0.015]); shareholder/creditor rights display very large magnitudes (e.g., Shareholders' Rights 45.436*** [1.158]; Creditor Rights -17.255*** [0.313]).
  - R&D regressions (Table C3): UAI generally negative and significant in many specs; IDV positive across specs; PDI negative and significant across specifications.
- Interaction tests (Table C5) confirm heterogeneous moderation by institutions and industry opacity; many interactions significant with varying magnitudes.

### Limitations and caveats
- Cultural measures:
  - Derived from IBM employee survey (mid-level employees) from 1967–1973; may not perfectly reflect entire national populations.
  - Authors argue UAI, PDI, and IDV “do not change significantly over time” and cite replication studies, but acknowledge potential cultural change due to globalization or extreme shocks.
  - Potential measurement heterogeneity: different cultures may define “risk” differently.
- Data limitations:
  - Industry opacity constructed using U.S. stock return informativeness; U.S. excluded from cross-country regressions because of that construction.
  - Foreign-firm analysis limited by small sample (~500 firms) and lacks controls for length of foreign presence or management composition.

### Policy-relevant implications and recommendations
- Policy design:
  - The same institutional rules can produce different economic outcomes in culturally different societies; policymakers should account for cultural values and norms when designing policies to promote competitive business environments and when addressing risk-taking in informationally opaque sectors.
- Multinational firms:
  - Should account for parent- and host-country cultural influences on internal decision-making to reduce “noisy” executive interactions and decision errors (e.g., risky R&D projects).
- Research directions suggested by authors:
  - Examine cultural composition of firms’ boards and managers to refine firm-level cultural measures.
  - Decompose formal versus informal institutions in more depth and assess whether cultural dimensions reflect “as is” practices or “as should be” values.
  - Investigate regional and organizational heterogeneity within countries.
  - Decompose risk-taking into managerial decisions versus institutional/regulatory constraints and analyze cross-industry regulatory effects.

### Key descriptive and summary statistics (selected entries from Appendices D–E)
- Sample and observations:
  - Firm — 471820 observations (unique firms 50702).
  - Year — 471820 observations (unique years 13); Mean: 2005; Std. Dev.: 5; Min: 2000; Max: 2012.
  - Country — 471820 observations (unique 51).
- Variable moments (selected):
  - z-score — Obs: 209161; Mean: 0.01; Std. Dev.: 1.12; Min: -11; Max: 12.
  - σ(ROA) — Obs: 413159; Mean: 18.35; Std. Dev.: 44.32; Min: 0; Max: 1081.
  - R&D Index — Obs: 324142; Mean: 0.02; Std. Dev.: 0.07; Min: 0.00; Max: 1.00.
  - Uncertainty aversion — Obs: 471820; Mean: 56.20; Std. Dev.: 21.70; Min: 8; Max: 112.
  - Individualism — Obs: 471820; Mean: 58.82; Std. Dev.: 28.99; Min: 12; Max: 91.
  - Power distance — Obs: 471820; Mean: 53.18; Std. Dev.: 18.78; Min: 11; Max: 104.
  - Masculinity — Obs: 471820; Mean: 59.46; Std. Dev.: 17.18; Min: 5; Max: 95.
  - Industry informational opacity — Obs: 471820; Mean: 4.65; Std. Dev.: 1.03; Min: 0; Max: 8.
  - Formal institutions index — Obs: 471820; Mean: 2.24; Std. Dev.: 1.77; Min: -3.18; Max: 4.87.
  - Shareholder rights — Obs: 448254; Mean: 0.63; Std. Dev.: 0.20; Min: 0.09; Max: 1.00.
  - Creditor rights — Obs: 451908; Mean: 1.98; Std. Dev.: 1.04; Min: 0.00; Max: 4.00.
- Country list (selected): Argentina; Australia; Austria; Belgium; Brazil; Chile; China; Colombia; Czech Republic; Denmark; Egypt; Finland; France; Germany; Greece; Hong Kong; Hungary; India; Indonesia; Ireland; Israel; Italy; Japan; Kenya; Kuwait; Lebanon; Malaysia; Mexico; Netherlands; New Zealand; Nigeria; Norway; Pakistan; Peru; Philippines; Poland; Portugal; Saudi Arabia; Singapore; South Africa; South Korea; Spain; Sweden; Switzerland; Taiwan; Thailand; Turkey; United Arab Emirates; United Kingdom; United States*; Venezuela.  
  - Note: * United States excluded from regressions to avoid econometric biases.

*Source: _wp12210 - 1. Correlation Matrix of National Cultural Dimensions (appendices A–E and regression appendices summarized).*

### 1. Correlation Matrix of National Cultural Dimensions ..................................................................

### _wp12210 - 1. Correlation Matrix of National Cultural Dimensions

### Introduction and scope
- Study employs four dimensions of national culture identified by Hofstede (2001) and an international sample of 50,000 firms spread across 400 industries in 51 countries to analyze the effects of cultural differences on corporate risk-taking.
- Aim: identify channels through which cultural values influence corporate risk-taking at macro (institutional/economic), market-structure (industry concentration/diversification), and micro (corporate/individual decision-making) levels.
- Uses a hierarchical linear mixed model to simultaneously model firm-, industry-, and country-level influences and their cross-level interactions.

### Core findings
- Culture impacts corporate risk-taking directly and not merely through indirect channels such as the legal and regulatory frameworks.
- Corporate risk-taking is higher in societies with:
  - low uncertainty avoidance,
  - low tolerance for hierarchical relationships,
  - societies which value individualism over collectivism.
- These effects are even more accentuated in societies with better formal institutions.
- Firms in countries ranking high in uncertainty-aversion and low in individualism take significantly less risk in informationally opaque industries (examples: finance, IT, oil refinery and mining) compared to firms in countries lower in uncertainty-aversion and higher in individualism.
- Risk-taking by foreign firms is best explained by the cultural norms of their country of origin.
- The cultural dimensions studied are not proxying for legal constraints, economic development, bankruptcy costs, insurance safety nets, or many other factors.

### Literature context and gaps addressed
- Previous studies largely focused on the banking and financial sectors; some on manufacturing. This paper extends analysis across all industries to capture cross-industrial differences.
- Prior work examples:
  - Kanagaretnam et al. (2011): banks in low uncertainty avoidance and high individualism societies show aggressive risk-taking.
  - Griffin et al. (2012): uncertainty avoidance negatively and individualism positively associated with firm-level riskiness in manufacturing.
- This paper innovates by:
  - covering all industries to capture variation in informational opacity,
  - differentiating direct and indirect effects of national culture,
  - employing a hierarchical linear mixed model to model multi-level effects without aggregation/disaggregation,
  - modeling cross-level interactions among firm, industry, and country influences.

### Data and measurement: national culture
- National cultural dimensions derived from Geert Hofstede's psychological survey based on IBM employee values scores collected between 1967 and 1973 in 66 countries: 31 advanced economies and 35 emerging markets.
- Four Hofstede dimensions used: (1) tolerance of uncertainty (UAI); (2) degree of group integration (individualism versus collectivism, IDV); (3) immobility between social classes (power distance); and (4) division of roles between genders (masculinity/femininity).
- The first index, national uncertainty aversion (UAI), defined as “feeling uncomfortable with uncertainty and ambiguity, and therefore valuing beliefs and institutions that provide certainty and conformity (Hofstede 2001).”
- On a scale from 1 to 100:
  - a simple mean-comparison t-test at the 5 percent level shows that the average UAI score for low income countries is 70, six points higher than the average UAI score for high income countries.
- Subsequent studies validated and extended Hofstede’s results on diverse occupational populations (school teachers, airplane pilots, high school students, etc.).

### Theoretical channels and expectations
- Culture may affect firm risk-taking through:
  - macro channels: institutional, legal, and economic development (creditor rights, investor protection, judicial efficiency, corporate governance, bankruptcy protection, transparency, corruption),
  - market-structure channels: industry informational opacity, industry concentration/diversification,
  - micro channels: individual and managerial decision-making, board composition and leadership structure.
- Expectation: cultural traits like uncertainty aversion and individualism will have stronger effects in industries with greater informational opacity (e.g., IT, financial services, oil extraction, chemicals) due to higher information uncertainty and complexity.

### Implications for policy and practice
- Same institutional rules can produce different economic outcomes in culturally different societies; policy design should account for cultural values and norms.
- Policymakers should consider cross-cultural values when drafting policies that promote competitive business environments and when addressing risk-taking in informationally opaque sectors.
- Multinational firms should account for cultural influences on internal decision-making to reduce “noisy” interactions among executives and errors in decision-making (e.g., when pursuing risky R&D projects).

*Source: _wp12210 - 1. Correlation Matrix of National Cultural Dimensions.*

### Appendix A details the survey questions used by Hofstede (2001) to calculate the national cultural values.

### _wp12210 - Appendix A details the survey questions used by Hofstede (2001) to calculate the national cultural values.

### Hofstede dataset and sample
- Sample described: 117,000 IBM local employees of marketing and customer service positions from 66 countries around the world.
- Replication studies cited: Hoppe (1990), Schwartz (1992, 1994), the European Media and Marketing Survey (1995, 1997 and 1999), the Helmreich and Merritt Study (1998), and Hofstede et al. (2010).

### National cultural dimensions (definitions and descriptive differences)
- Four primary cultural variables used:
  - Uncertainty Aversion (UAI): tolerance for ambiguity; high UAI = fear of failure.
  - Individualism (IDV): emphasis on individual vs. group; high IDV linked to overconfidence/overoptimism.
  - Power Distance (PDI): extent to which less powerful accept unequal power distribution.
  - Masculinity (MAS): dominant gender role patterns (assertiveness/achievement vs. modesty/nurturance).
- Mean differences between income groups:
  - IDV mean: high income countries = 57; low income countries = 25; 27 point mean-difference significant at the 5 percent level.
  - PDI mean: low income countries = 67; high income countries = 48; 19 point mean-difference significant at the 5 percent level.
  - MAS mean: high income economies = 51; low income countries = 42; 9 point difference not statistically significant at the 5 percent level.
- Pair-wise correlations (Table 1 excerpts):
  - Individualism vs. Power Distance: -0.66* (partial correlation controlling for economic development in parentheses: -0.27).
  - Uncertainty Aversion vs. Individualism: -0.21 (-0.30).
  - Power Distance vs. Uncertainty Aversion: 0.15 (0.16).
  - Masculinity correlations with others: 0.05 (0.05) with UAI; 0.09 (0.14) with IDV; 0.07 (0.09) with PDI.
  - Note: (*) represents significance at the 1 percent level.

### Limitations of cultural variables
- Stability and cross-time replication:
  - National uncertainty aversion, power distance, and individualism “do not change significantly over time” per authors’ argument.
  - High correlations across decades cited: Hofstede (1984, 2001); Hoppe (1990); European Media and Marketing Survey (1995, 1997, 1999); Hofstede et al. (2010).
- Key caveats:
  - Cultural measures derived from mid-level IBM employees may not perfectly reflect entire national populations.
  - Changes in culture due to globalization or dramatic exogenous events are possible but would likely require longer periods or extreme shocks.
  - Different cultures may define “risk” differently, creating potential measurement heterogeneity.

### Measuring firm risk-taking (CVU database and risk proxies)
- Data source: Corporate Vulnerability Utility (CVU), annual indicators from 2000 to 2012, coverage ~50,000 publicly traded firms, >400 SIC industries, 51 countries with national culture data.
- Three firm-level risk-taking indicators:
  1. Z-score (natural logarithm of Z): Z = (ROA + CAR) / σ(ROA); ROA = earnings before taxes and loan loss provision divided by assets; CAR = capital-asset ratio; σ(ROA) = standard deviation of ROA over sample period. Higher z-score = more stable/financially healthy firm. Authors denote the natural logarithm as “z-score.”
  2. σ(ROA): standard deviation of ROA (EBIT / total assets), computed as deviation from industry-year-country average and then standard deviation across available years (requirement: earnings and total assets available for at least eight years over 1999-2012).
  3. R&D expenditures to market capitalization: research and development expenditures divided by market capitalization.

### Measuring industry informational opacity
- Methodology overview:
  - Follows Huang (2008) using relative firm-specific stock return variation (Durnev et al. 2004; Rajan and Zingales 1998).
  - Uses weekly stock returns of all U.S. public-listed firms from 2000 to 2005 to calculate stock price informativeness by industry; opacity is inverse of informativeness.
  - United States excluded from cross-country regressions because U.S. informativeness used as proxy.
- Table 2 (selected entries) — Industry informational opacity (R-squared; Informativeness Index; Opacity Index):
  - Finance, Insurance, And Real Estate: 0.0034; 0.9966; 5.684
  - Wholesale Trade: 0.0039; 0.9961; 5.550
  - Mining: 0.0042; 0.9958; 5.477
  - Transportation, Communications, Electric, Gas: 0.0059; 0.9941; 5.123
  - Retail Trade: 0.0060; 0.9940; 5.111
  - Services: 0.0066; 0.9934; 5.019
  - Agriculture, Forestry, And Fishing: 0.0068; 0.9932; 4.986
  - Construction: 0.0072; 0.9928; 4.931
  - Manufacturing: 0.0085; 0.9915; 4.756
- Opacity index construction: informativeness highly skewed; normalized via ln(1/R2)-ln(R2); opacity index reported as informational opacity; U.S. data only.

### Control variables (firm, industry, country)
- Country-level controls (13 listed):
  1. Economic development: ratio of credit to GDP.
  2. Policy rate (central bank policy rate).
  3. Aggregate economic volatility: standard deviation of annual GDP growth.
  4. Rule of law.
  5. Regulatory quality.
  6. Government effectiveness.
  7. Voice and accountability.
  8. Control of corruption.
  9. Political stability.
  10. Efficiency of debt enforcement.
  11. Creditor rights.
  12. Shareholder protection.
  13. Bankruptcy costs.
- Formal Institutions Index (PCA-derived composite):
  - Formal Institutions Index = Rule of Law*0.41 + Efficiency of Debt Enforcement*0.34 + Control of Corruption*0.40 + Regulatory Quality*0.39 + Government Effectiveness*0.40 + Voice and Accountability*0.36 + Political Stability*0.31.
  - Weights obtained via principal components analysis.
- Industry-level controls:
  - Competition/concentration index: Herfindahl index (sum of squares of market shares in industry; range 0 to 1.0).
  - Industry informational opacity (as above).
- Firm-level controls:
  - Firm size: log of total assets (sum of fixed and current assets).
  - Leverage: total liabilities / total assets.
  - Profitability (ROA): EBIT / total assets.
  - Sales growth: annual logarithmic growth of sales.
  - Dependence on external finance: Rajan-Zingales index.
- Data treatment:
  - Winsorize firm-level controls at the 1% level in both tails; sales growth and leverage also winsorized at top 5% of distribution.
  - Final sample: 50,000 firms across 400 industries in 51 countries.

### Hypotheses (numbered as in text)
- H1: Uncertainty aversion will directly negatively influence firm risk-taking.
- H2: Individualism will directly positively influence firm risk-taking.
- H3: Power distance will directly negatively influence corporate risk-taking.
- H4: Masculinity will directly positively influence firm risk-taking.
- H5: Individualism strengthens the positive relationship between formal institutional development and corporate risk-taking; power distance weakens it.
- H6: Individualism and power distance will indirectly affect corporate risk-taking through bankruptcy protection and creditor/shareholder rights; lower uncertainty aversion strengthens shareholders’-rights → corporate risk-taking link.
- H7: Higher uncertainty aversion reduces risk in informationally opaque industries; higher individualism and masculinity increase risk-taking in informationally opaque industries.

### Empirical model and estimation method
- Hierarchical linear mixed models (HLM) used to exploit multilevel dataset: firms nested in industries nested in countries.
- Sample period: 2000-2012.
- Dependent variables: z-score (natural log of Z), σ(ROA), R&D index.
- Mean-centering: independent variables grand-mean centered; additional variables constructed with suffixes “_indus_ctry_mean” and “_firm_var” to separate within- and between-level variation.
- Basic model structure provided (multilevel decomposition with firm, industry, country controls and cultural variables).

### Key empirical findings (direct effects)
- Direct and robust effects (HLM specifications):
  - Uncertainty Aversion (UAI): negative and significant across models (supports H1).
  - Individualism (IDV): positive and significant across models (supports H2).
  - Power Distance (PDI): negative and significant for z-score and R&D index, less robust for σ(ROA) (supports H3 partially).
  - Masculinity (MAS): not significant in HLM specifications (H4 rejected).
- Economic significance (reported in text using one standard deviation changes):
  - One standard deviation in uncertainty aversion (21.7) associated with:
    - Change in z-score: 0.1085 (0.005*21.7); mean z-score = 0.019; stdev = 1.12.
    - Change in σ(ROA): 4.1447 (0.191*21.7); mean σ(ROA) = 23.367; stdev = 51.18.
    - Change in R&D index: 0.2387 (0.011*21.7); mean R&D index = 0.02; stdev = 0.04.
  - One standard deviation in individualism (28.9) associated with:
    - Change in z-score: 0.2312 (0.008*28.9).
    - Change in σ(ROA): 6.5314 (0.226*28.9).
    - Change in R&D index: 9.9127 (0.343*28.9).
  - One standard deviation in power distance (18.7) associated with:
    - Change in z-score: 0.0187 (0.001*18.7).
    - Change in R&D index: 2.6741 (0.143*18.7).
- Firm-level control patterns:
  - Firm profitability and sales growth: significant and negative across models (more profitable/higher sales firms take less risk).
  - Firm size and leverage: mixed/less robust across specifications.

### Key empirical findings (indirect effects and institutional/industry links)
- Formal Institutions Index: positive and significant across models — better formal institutions → more firm risk-taking.
- Shareholders' rights: positive and significant in several specifications (e.g., 17.687** in stdev(ROA) HLM column).
- Creditor rights: associated with lower corporate risk-taking and lower σ(ROA) (e.g., -3.525**).
- Industry effects:
  - Industry competition (Herfindahl): positive association with firm risk-taking.
  - Industry informational opacity: negative association with firm risk-taking (firms in more opaque industries tend to take less risk on average).

### Moderating / interaction effects (Table 7 highlights)
- H5 confirmed in part:
  - IDV x Formal Institutions: interaction positive and significant (e.g., 0.006** for inverse z-score, 0.155* for σ(ROA), 0.259* for R&D index) — higher individualism strengthens the positive relationship between formal institutions and corporate risk-taking.
  - PDI x Formal Institutions: interaction not significant, suggesting power distance weakens but does not significantly interact in models.
- H6: no consistent pattern; mixed interaction results (some significant coefficients, but authors reject hypothesis H6 overall).
  - Specific interactions: IDV x Creditor Rights = -0.009*** (inverse z-score specification), IDV x Shareholders' Rights = 0.037*** (inverse z-score specification).
  - PDI x 1/Eff. of Debt Enforcement = -0.118* (inverse z-score specification).
- H7 partly confirmed and notable surprising result:
  - UAI x Industry Opacity: negative and significant across models (e.g., -0.001* inverse z-score; -0.006** σ(ROA); -0.058* R&D) — high UAI in opaque industries → even less risk-taking.
  - IDV x Industry Opacity: highly positive and significant across models (e.g., 1.185* inverse z-score; 27.323** σ(ROA); 94.36* R&D) — in highly opaque industries, firms in highly individualistic societies take even more risk. Authors note this is surprising and may reflect greater competition in opaque industries in individualistic societies.
  - PDI and MAS interactions with industry opacity: not significant.

### Foreign vs. domestic firms (identification test)
- Sample of foreign firms is limited (only 500 foreign firms).
- Key finding: for foreign subsidiaries, the parent (origin) country’s cultural values explain their risk-taking behavior more than host country cultural values.
  - Example coefficients (foreign firms, origin country effects):
    - Individualism (origin country) significant and positive for σ(ROA) and R&D models (e.g., 0.554*** σ(ROA) column).
    - Power Distance (origin country) negative and significant across models (e.g., -0.031*** inverse z-score; -0.778** σ(ROA); -0.231*** R&D).
- Interpretation: firms retain cultural behavior from country of origin; domestic regulatory implications may differ for domestic vs. foreign firms.
- Caveats: small foreign-firm sample; analysis does not account for length of foreign presence or cultural composition of management.

### Conclusions and policy-relevant implications
- Main conclusions:
  - National cultural values exert both direct and indirect effects on corporate risk-taking, even after controlling for institutional, economic, and industrial environments.
  - Firms in societies that are highly uncertainty-averse, low in individualism, and high in power distance tend to take less risk.
  - Culture interacts with formal institutions and industry characteristics to produce heterogeneous outcomes (e.g., individualism amplifies the effect of strong formal institutions on risk-taking).
  - Foreign subsidiaries tend to reflect the culture of the parent country more than the host country, suggesting regulatory outcomes may vary by firm cultural origin.
- Suggested considerations for further research (authors’ proposed directions):
  - Examine cultural composition of firms’ boards and managers to refine firm-level cultural measures.
  - Decompose formal versus informal institutions in more depth and clarify whether cultural dimensions reflect “as is” practices or “as should be” values.
  - Investigate regional and organizational heterogeneity within countries.
  - Better decompose risk-taking into managerial decisions versus institutional/regulatory constraints and analyze cross-industry regulatory effects.

*Source: _wp12210 - Appendix A details the survey questions used by Hofstede (2001) to calculate the national cultural values.*

### REFERENCES

### _wp12210 - REFERENCES

### Appendix A: Measuring National Culture
- Construction of indices:
  - Uncertainty aversion index: country mean scores for three questions:
    - Rule orientation: agreement with “Company rules should not be broken, even when it is in the company‟s best interest.”
    - Employment stability: employees‟ statement that they intend to remain employed at IBM for a certain number of years.
    - Stress: mean answer to “How often do you feel nervous or tense at work?”
  - Individualism index: country mean scores for three items:
    - Importance attached to having enough personal/family time.
    - Importance attached to having good physical working conditions and a secure employment.
    - Importance attached to having a challenging, adventurous job.
  - Power distance index: country mean scores for three items:
    - Perceptions of subordinates‟ fear of disagreeing with superiors;
    - Subordinates‟ perception that their boss takes decisions in an autocratic/paternalistic way;
    - Subordinates‟ preference for an autocratic, a persuasive/paternalistic, or a democratic/consultative style of decision-making.
  - Masculinity index: country mean scores for six items:
    - Job content and learning: how important is to have a job on which there is a great deal of day-to-day learning.
    - Rewards: how important it is to have opportunity for high earnings/career advancement.
    - Interpersonal relations: how important is to work with people who are team-workers.
    - Security: how important is to have job security of not being transferred to a less desirable job.
    - Comfort: how important is to have good physical working conditions.
    - Company: how important is to work in a company that is successful and modern.

- Figure A1. National Culture by Income Group (tabulated values preserved):
  - Uncertainty Aversion | Individualism | Power Distance | Masculinity
  - High Income: 64.00 57 48 51
  - Middle Income: 68.00 27 71 50
  - Low income: 70.00 25 67 42
  - Correlation with income: -0.99* 0.98* -0.92* 0.77*

- Visuals referenced:
  - Figure A2. Heat Map of Uncertainty Aversion Index
  - Figure A3. Heat Map of Individualism Index
  - Figure A4. Heat Map of Power Distance Index
  - Figure A4. Heat Map of Masculinity Index

### Appendix B: Measuring Industry Informational Opacity
- Stepwise methodology:
  - Step 1: Firm f‟s stock price returns as weekly logarithmic growth:
    - r_f,t = ln (price_t/price_t-1)         (1)
  - Step 2: Regress firm returns on industry-specific and market-specific returns:
    - r_i,f,t = α_f,0 + β_f,i * r_i,t + φ_f,m * r_m,t + Ɛ_i,f,t      (2)
    - Industry returns: market capital-weighted portfolio of all firms in industry i excluding firm f (weighting by firm f‟ share of market capital in industry i).
    - Market returns: market capital-weighted portfolio including all NYSE, AMEX and NASDAQ listed stocks (weighting by firm f‟s share of market capital in market m).
    - Exclusion of firm f prevents spurious correlations in industries with a limited number of firms.
  - Step 3: Stock price informativeness measure:
    - Stock price informativeness = 1- R^2      (3)
    - Interpretation: the more informative stock prices are, the lower the R^2 will be.
  - Step 4: Logistic transformation to construct industry opacity index:
    - Industry opacity = ln (1- R^2/ R^2) = ln (Stock price informativeness/ R^2) (4)
    - Purpose: make the index less skewed and less leptokurtic than 1- R^2.
    - Interpretation: higher industry opacity implies less informative stock prices in that industry.

- Dataset summary:
  - The table shows the industry opacity index for 140 industries.

- Selected entries from Table B. Informational Opacity Index by Industry (values preserved exactly as in source):
  - SIC Code 616 | Industry: Mortgage Bankers & Loan Correspondents | R-squared 0.0010 | Informativeness index 0.9990 | Opacity Index 6.903
  - SIC Code 614 | Industry: Personal Credit Institutions | R-squared 0.0014 | Informativeness index 0.9986 | Opacity Index 6.573
  - SIC Code 651 | Industry: Real Estate Operators (No Developers) & Lessors | R-squared 0.0015 | Informativeness index 0.9985 | Opacity Index 6.511

- Additional notes on Table B:
  - The table (extended across pages) lists industry-level R-squared, Informativeness index, and Opacity Index for each SIC-coded industry, with values ranging from very low R-squared (e.g., 0.0010) and high opacity (e.g., 6.903) through much higher R-squared and correspondingly lower opacity (e.g., 0.47750 R-squared, 0.5225 Informativeness, 0.090 Opacity Index).

*Italic: Content derived from the PDF chapter/section "_wp12210 - REFERENCES" (appendices A and B as provided).*

### Appendix C: Regression Results

### Appendix C: Regression Results

### Notes and estimation framework
- Dependent variables: Risk-Taking measured as (a) Inverse of the z-score; (b) Standard Deviation of Return on Assets; (c) Research & Development Index (R&D expenditures to market capitalization).
- Estimation period: 2000-2012; sample includes firms from all industries; models nest firms across two levels: industry and country.
- Models: Hierarchical linear mixed model (HLM) estimations reported alongside OLS for comparison. OLS regressions employ robust standard errors clustering at the country level with industry fixed effects (not reported).
- Significance notation: ***, **, and * denote significance at the 1, 5, and 10 percent level.
- Table-level notes: Some regressions exclude U.S. firms (noted in specific tables).

### Table C1 — Explaining Corporate Risk-Taking Through the Inverse of the Z-Score (selected coefficients)
- Culture variables (HLM and OLS consistently):
  - Uncertainty Aversion:
    - Basic OLS: -0.001*** [0.000]
    - Basic HLM: -0.002*** [0.000]
    - Medium OLS: -0.004*** [0.000]
    - Medium HLM: -0.003*** [0.001]
    - Complex OLS: -0.005*** [0.001]
    - Complex HLM: -0.005* [0.003]
    - Very complex OLS: -0.005*** [0.001]
    - Very complex HLM: -0.005*** [0.003]
  - Individualism:
    - Ranges from 0.002*** to 0.008*** across specifications (example: Basic OLS 0.005*** [0.000]; Basic HLM 0.007** [0.003]).
  - Power Distance:
    - Negative and significant in most specifications (example: Basic OLS -0.009*** [0.001]; Medium OLS -0.006*** [0.001]).
  - Masculinity:
    - Generally small and often insignificant (e.g., Basic OLS 0.000 [0.000]; complex specifications show some positive 0.001 values with larger SEs).
- Country, industry, and firm controls (selected):
  - Financial Development: ranges from -0.001 [0.001] to 0.015** [0.007] depending on specification.
  - Central Bank Policy Rate: consistently negative and often significant (example: Basic OLS -0.010*** [0.001]; Basic HLM -0.026* [0.014]).
  - Shareholders' Rights Index: Basic OLS 0.359*** [0.053]; Basic HLM 0.147 [0.291] (variation across specs).
  - Industry Opacity: -0.035*** [0.008] (reported in more complex specs).
  - Firm Size: consistently negative and highly significant (e.g., Basic OLS -0.118*** [0.003]).
  - Firm Profitability: consistently negative and highly significant (e.g., Basic OLS -0.112*** [0.001]).
- Sample and fit:
  - Observations vary by specification: 131446; 104886; 104886; 94997.
  - Number of firms examples: 26777; 26777; 22022; 20220; 22002; 20220; 20055; 20055.
  - R-square overall examples: 0.192; 0.200; 0.202; 0.207.

### Table C2 — Explaining Corporate Risk-Taking Through the Standard Deviation of ROA (selected coefficients)
- Culture variables (strong magnitudes relative to SEs):
  - Uncertainty Aversion:
    - Basic OLS -0.400*** [0.011]; Basic HLM -0.161** [0.075]; Medium OLS -0.512*** [0.013]; Medium HLM -0.200** [0.084]; Complex OLS -0.577*** [0.012]; Complex HLM -0.166* [0.086]; Very complex OLS -0.568*** [0.013]; Very complex HLM -0.191** [0.086].
  - Individualism:
    - Basic OLS 0.453*** [0.011]; Basic HLM 0.215*** [0.075]; across specs up to 0.726*** [0.015] and 0.226** [0.103].
  - Power Distance:
    - Negative in some specs (Basic OLS -0.104*** [0.014]) and large negative in Complex OLS -0.453*** [0.018].
  - Masculinity:
    - Positive and significant in several specs (e.g., Basic OLS 0.155*** [0.010]).
- Country/institutional variables:
  - Shareholders' Rights Index: very large positive coefficients (e.g., Basic OLS 45.436*** [1.158]; Complex OLS 54.309*** [1.501]).
  - Creditor Rights: large negative coefficients (e.g., Basic OLS -17.255*** [0.313]; Complex OLS -18.086*** [0.333]).
  - Aggregate Real Econ Volatility: 575.432*** [32.444] in Complex OLS; other specs report large SEs or non-significant values.
- Firm-level controls:
  - Firm Size: positive and significant for σ(ROA) (e.g., Basic OLS 2.228*** [0.063]; Basic HLM 2.943*** [0.095]).
  - Firm Leverage: mixed; some negative and significant HLM estimates (e.g., Basic HLM -0.042*** [0.003]).
  - Firm Profitability: positive in HLMs (e.g., Basic HLM 0.718*** [0.006]).
- Sample and fit:
  - Observations examples: 227993; 145469; 145469; 132571.
  - Number of firms examples: 38398; 38398; 26683; 26683; 24197.
  - Rsquare overall examples: 0.314; 0.470; 0.480; 0.504.

### Table C3 — Explaining Corporate Risk-Taking Through R&D Expenditures (selected coefficients)
- Culture variables:
  - Uncertainty Aversion:
    - Basic OLS -0.035*** [0.012]; Basic HLM -0.015*** [0.0078]; Medium OLS -0.101*** [0.028]; Medium HLM -0.080** [0.037]; Complex OLS -0.050** [0.036]; Complex HLM -0.074** [0.036]; Very complex OLS -0.050** [0.027]; Very complex HLM -0.011** [0.006].
  - Individualism:
    - Positive across specs, varying significance (e.g., Complex OLS 0.142*** [0.046]; Very complex HLM 0.343** [0.160]).
  - Power Distance:
    - Strongly negative and significant across specifications (e.g., Basic OLS -0.126*** [0.031]; Very complex HLM -0.143*** [0.058]).
  - Masculinity:
    - Mixed; some specifications show significance (Very complex OLS 0.077*** [0.015]).
- Country/institutional and macro variables:
  - Financial Development: positive and significant in several OLS specs (e.g., Basic OLS 0.237*** [0.047]).
  - Shareholders' Rights Index: positive and significant in several OLS/HLM (e.g., Basic OLS 4.703*** [1.360]; Basic HLM 12.283* [7.200]).
  - Aggregate Real Econ Volatility: 249.115* [136.537] in some specifications.
- Industry and firm controls:
  - Industry Opacity: negative and significant in some specs (e.g., Medium OLS -2.081** [1.007]).
  - Industry Competition: positive and significant in OLS (e.g., Basic OLS 6.475*** [1.881]).
  - Firm Profitability: consistently negative and significant (e.g., Basic OLS -0.167*** [0.060]; Basic HLM -0.150*** [0.029]).
- Sample and fit:
  - Observations examples: 229700; 204118; 151029; 132755.
  - Number of firms examples: 39898; 39898; 35713; 59713; 28046; 24328.
  - R-square overall examples: 0.050; 0.300; 0.290; 0.060.

### Table C4 — Explaining Corporate Risk-Taking (With Decomposed Institutional Variables) — selected coefficients
- Core cultural effects remain across decomposed institutional specifications:
  - Uncertainty Aversion: -0.005*** [0.001]; other specs show larger negative magnitudes for σ(ROA) and R&D (e.g., -0.568*** [0.013]; -0.191** [0.086]; -0.011** [0.006]).
  - Individualism: positive and significant across measures (e.g., 0.003*** [0.001] up to 0.726*** [0.015]).
  - Power Distance: negative and significant (e.g., -0.006*** [0.001]; Complex OLS -0.415*** [0.019]).
  - Masculinity: small positive values in σ(ROA) specs (0.194*** [0.015]) and mixed elsewhere.
- Decomposed institutional measures (examples):
  - Rule of Law: 0.042** [0.020] in one inverse-z specification; 14.979*** [0.408] in σ(ROA) OLS; other decomposed institutional coefficients show large magnitudes and wide SEs.
  - Control of Corruption: -0.079*** [0.012] in one inverse-z specification; large negative coefficients in σ(ROA) OLS (-17.531*** [0.228]) with large SEs in other specs.
  - Efficiency of Debt Enforcement: 0.004*** [0.001] in inverse-z OLS; other specs show negative σ(ROA) effects (e.g., -0.243*** [0.014]).
- Industry/opactiy and competition:
  - Industry Opacity: negative and significant for inverse z-score (-0.035*** [0.008]) and σ(ROA) (-0.433* [0.224]) and R&D (-2.081** [1.007]).
  - Industry Competition: positive and significant in many specifications (e.g., 0.091*** [0.034]; 8.005*** [1.377]).
- Firm controls and sample:
  - Firm Size: negative for inverse-z (-0.121*** [0.004]) and positive for σ(ROA) in some specs (2.331*** [0.089]); negative for R&D in complex specs.
  - Observations examples: 94997; 132571; 132755.
  - Number of firms examples: 20055; 20055; 24197; 24197; 43282; 43282.
  - R-square overall examples: 0.207; 0.504; 0.060.

### Table C5 — Accentuating/Moderating Effects (interaction tests; selected coefficients)
- Interactions and moderators reported across three dependent variables (Inverse of the z-score; σ(Return on assets); Research & Development) with H5/H6/H7 hypotheses grouping:
- Key interaction patterns (selected coefficients and significance):
  - Uncertainty Aversion (main effects across models):
    - Inverse z-score: -0.005*** [0.003]; -0.007** [0.003]; -0.004* [0.003] across columns (1)(2)(3).
    - σ(ROA): -0.297*** [0.096]; -0.339** [0.170]; -0.360*** [0.097].
    - R&D: -0.010** [0.004]; -0.089** [0.49]; -0.038** [0.184].
  - Individualism (main and interactions):
    - Inverse z-score: 0.013* [0.006]; 0.004** [0.002]; 0.015*** [0.004].
    - σ(ROA): 0.538*** [0.183]; 0.381* [0.199]; 0.444*** [0.098].
    - R&D: 0.113* [0.060]; 0.267 [0.437]; 0.297** [0.132].
    - Interaction Individualism X Formal Institutions: 0.006** [0.003]; 0.155* [0.083]; 0.259* [0.141].
    - Individualism X Creditors' Rights: -0.009*** [0.002] (significant in an inverse-z specification).
    - Individualism X Shareholders' Rights: 0.037*** [0.010] (inverse-z).
  - Power Distance:
    - Main effects mixed; interactions include Power Distance X 1/Eff. of Debt Enforcement: -0.118* [0.068].
    - Power Distance X Industry Opacity often small or insignificant.
  - Industry Opacity and Competition interactions:
    - Industry Opacity: consistently negative and significant in inverse-z and σ(ROA) models (e.g., inverse-z -0.037*** [0.007]; σ(ROA) -0.048*** [0.013]).
    - Industry Competition: positive and significant coefficients in many interaction specifications (e.g., inverse-z 0.206* [0.124]; σ(ROA) 14.284*** [4.844]; R&D 12.808 [7.566]).
  - Firm controls (consistent effects across interactions):
    - Firm Size: inverse-z negative and significant (e.g., -0.082*** [0.004]); σ(ROA) positive and significant (2.789*** [0.116]); R&D negative in several interaction specs (-1.235* [0.748]).
    - Firm Leverage: inverse-z small negative and significant (-0.003*** [0.000]); σ(ROA) negative and significant (-0.094*** [0.007]).
    - Firm Profitability: inverse-z -0.029*** [0.000]; σ(ROA) 0.728*** [0.007]; R&D -0.175*** [0.046].
    - Firm Sales Growth: inverse-z -0.290*** [0.022]; R&D large negative effects in some specs (e.g., -7.570** [3.046]).
    - Firm Dependence on External Finance: inverse-z -0.003*** [0.000]; σ(ROA) 0.026*** [0.006]; R&D -0.049 [0.041] (non-significant in some specs).
- Sample size and scope for interaction regressions:
  - Number of firms: 20055 repeated across columns (interactions reported on the 50,000 firms, from 50 countries baseline; specific interaction tables use the sample size 20055).
- Notes:
  - Interactions tested include combinations such as Individualism X Formal Institutions, Power Distance X Cost of Bankruptcy, Uncertainty Aversion X Industry Opacity, and many others; reported coefficients show heterogeneous moderating effects with varying magnitudes and significance levels.
  - Robust standard errors in parentheses.

*Source: Appendix C: Regression Results (tables C1–C5), estimation details, coefficients, standard errors, observations, and notes as reported in the source PDF.*

### Appendix D: Data Sources and Definitions

### Appendix D: Data Sources and Definitions

### Country-specific data
- Uncertainty aversion — Uncertainty aversion indicator - UAI. Higher values reflect higher national uncertainty aversion.  
  - Source: Hofstede (2001)
- Individualism — Individualism indicator - IDV. Higher values reflect higher national individualism.  
  - Source: Hofstede (2001)
- Power distance — Power distance - PDI. Higher values reflect higher national power distance.  
  - Source: Hofstede (2001)
- Masculinity — Masculinity - MAS. Higher values reflect higher national masculinity.  
  - Source: Hofstede (2001)
- Religion  
  - Source: Gene Shackman, 2009, "The Global Social Change Research Project", Available at: http://gsociology.icaap.org
- Legal origin — Legal origin variable separating countries into: French, German, Scandinavian, Socialist, and British legal origin.  
  - Source: Shleifer et al. (1999)
- Rule of law — Captures perceptions of the extent to which agents have confidence in and abide by the rules of society, and in particular the quality of contract enforcement, property rights, the police, and the courts, as well as the likelihood of crime and violence.  
  - Source: Worldwide Governance Indicators, World Bank
- Control of corruption — Captures perceptions of the extent to which public power is exercised for private gain, including both petty and grand forms of corruption, as well as "capture" of the state by elites and private interests.  
  - Source: Worldwide Governance Indicators, World Bank
- Corruption — Alternative measure of the corruption level.  
  - Source: Transparency International
- Regulatory quality — Captures perceptions of the ability of the government to formulate and implement sound policies and regulations that permit and promote private sector development.  
  - Source: Worldwide Governance Indicators, World Bank
- Government effectiveness — Captures perceptions of the quality of public services, the quality of the civil service and the degree of its independence from political pressures, the quality of policy formulation and implementation, and the credibility of the government's commitment to such policies.  
  - Source: Worldwide Governance Indicators, World Bank
- Voice and accountability — Reflects perceptions of the extent to which a country's citizens are able to participate in selecting their government, as well as freedom of expression, freedom of association, and a free media.  
  - Source: Worldwide Governance Indicators, World Bank
- Political stability — Captures perceptions of the likelihood that the government will be destabilized or overthrown by unconstitutional or violent means, including politically-motivated violence and terrorism.  
  - Source: Worldwide Governance Indicators, World Bank
- Formal institution index — Index of quality of formal institutions, calculated by the formula:  
  - Formal institutions = [Rule of law*0.4248 + Government effectiveness*0.4198 + Control of corruption*0.4187 + Regulatory quality*0.4093 + Political stability*0.3954 + Accountability*0.3797]  
  - Source: Li and Zahra (2012); Author's calculations.
- Creditor rights — An index aggregating creditor rights, following La Porta et al. (1998). The index ranges from 0 (weak creditor rights) to 4 (strong creditor rights).  
  - Source: La Porta et al. (1998)
- Property rights — Index of property rights in 2004.  
  - Source: The Heritage Foundation, http://www.heritage.org/Index/
- Legal protection of minority shareholders (shareholder rights) — An index measuring the legal protection of minority shareholders against expropriation by corporate insiders. Named the anti-self-dealing index in Djankov et al. (2006).  
  - Source: Djankov et al. (2006)
- Disclosure requirements index — Variable "disclose", from La Porta et al. (1998).  
  - Source: La Porta et al. (1998)
- Efficiency of debt enforcement — Variable "Case A Efficiency".  
  - Source: Djankov et al. (2006)
- Enforceability of contracts — Index measuring enforceability of contracts.  
  - Source: Djankov et al. (2003)
- Bankruptcy costs — Assessment of the efficiency of bankruptcy law. Scale from 0 to 6, where higher scores indicate higher compliance.  
  - Source: World Economic Forum Global Competitiveness Report (2005).
- Propensity to file for bankruptcy — Index calculated as the inverse of bankruptcy costs.  
  - Source: Author's calculations.
- Economic development — GDP per capita.  
  - Source: International Finance Statistics

### Industry-level data
- SIC Classification — 2, 3, and 4 letter standard industrial classification.  
  - Source: US Department of Labor
- Stock price informativeness — 1-R2 of regressing firm‟s returns on industry-specific returns and market-specific returns simultaneously.  
  - Source: Author's calculations.
- Industry opacity — ln (1- R2/ R2)  
  - Source: Author's calculations.
- Industry diversification — Index that measures the diversity of industrial activities.  
  - Source: Author's calculations.
- Industry concentration — Herfindahl index which measures the amount of competition among firms within an industry.  
  - Source: Author's calculations.
- Industry competition — The inverse of industry concentration.  
  - Source: Author's calculations.
- Concentration in highly opaque industries — Herfindahl index which measures the amount of competition among firms within opaque industries (industries whose informational opacity is higher than 3.5).  
  - Source: Author's calculations.
- Concentration in less opaque industries — Herfindahl index which measures the amount of competition among firms in less opaque industries (industries whose informational opacity is lower than 3.5).  
  - Source: Author's calculations.

### Firm-level data
- All firm level data — All firm level data CVU Database constructed by the Macro-Finance Research Unit at the IMF (original data from Datastream, Worldscope, Bankscope)
- Z-score — (Return on Assets + Capital Asset Ratio) / Standard deviation of Return on Assets  
  - Source: Author's calculations.
- z-score — ln(Z-score).  
  - Source: Author's calculations.
- σ(ROA) — Standard deviation of return on assets.  
  - Source: Author's calculations.
- R&D expenditures — Expenditures on research and development divided by total firm market share.  
  - Source: CVU
- Firm size — Log of total assets, where total assets is the sum of fixes and current assets.  
  - Source: CVU
- Firm leverage — Leverage, which is defined as the ratio of total liabilities (current and noncurrent liabilities) to total assets.  
  - Source: CVU
- Profitability — Profitability, measured as the firm‟s return on assets (ROA) and calculated as the ratio of earnings (EBIT) to total assets.  
  - Source: CVU
- Sales growth — Sales growth, calculated as the annual logarithmic growth of sales.  
  - Source: CVU
- Dependence on external finance — Dependence on external finance, proxied by the Rajan-Zingales index (Rajan and Zingales, 1998).  
  - Source: CVU

### Appendix E: Summary Statistics Tables

### Table E1. Summary Statistics (selected entries)
- Firm — 471820 (unique 50702)
- Year — 471820 (unique 13); Mean: 2005; Std. Dev.: 5; Min: 2000; Max: 2012
- Country — 471820 (unique 51)
- z-score — Obs: 209161; Mean: 0.01; Std. Dev.: 1.12; Min: -11; Max: 12
- Return on Assets (ROA) — Obs: 310279; Mean: -7.42; Std. Dev.: 62.75; Min: -999; Max: 992
- Capital-Asset Ratio (CAR) — Obs: 274928; Mean: -1.43; Std. Dev.: 89.61; Min: -26513; Max: 43
- Standard deviation of ROA — Obs: 413159; Mean: 18.35; Std. Dev.: 44.32; Min: 0; Max: 1081
- Research & Development Index — Obs: 324142; Mean: 0.02; Std. Dev.: 0.07; Min: 0.00; Max: 1.00
- Uncertainty aversion — Obs: 471820; Mean: 56.20; Std. Dev.: 21.70; Min: 8; Max: 112
- Individualism — Obs: 471820; Mean: 58.82; Std. Dev.: 28.99; Min: 12; Max: 91
- Power distance — Obs: 471820; Mean: 53.18; Std. Dev.: 18.78; Min: 11; Max: 104
- Masculinity — Obs: 471820; Mean: 59.46; Std. Dev.: 17.18; Min: 5; Max: 95
- Long-term orientation — Obs: 381871; Mean: 53.24; Std. Dev.: 29.94; Min: 16; Max: 118
- Financial development alternative — Obs: 371479; Mean: 9.67; Std. Dev.: 8.66; Min: 0; Max: 53
- Stock market development — Obs: 463118; Mean: 104.53; Std. Dev.: 63.41; Min: 1; Max: 561
- Equity development — Obs: 471820; Mean: 1.09; Std. Dev.: 1.30; Min: 0; Max: 21
- Central Bank Rate — Obs: 444704; Mean: 3.68; Std. Dev.: 4.54; Min: 0; Max: 60
- Aggregate economic volatility — Obs: 471820; Mean: 0.03; Std. Dev.: 0.02; Min: 0.02; Max: 0.16
- Aggregate real economic volatility — Obs: 471820; Mean: 0.02; Std. Dev.: 0.01; Min: 0.01; Max: 0.08
- Rule of law — Obs: 471820; Mean: 1.06; Std. Dev.: 0.75; Min: -1.64; Max: 2.01
- Formal institutions index — Obs: 471820; Mean: 2.24; Std. Dev.: 1.77; Min: -3.18; Max: 4.87
- Religion — Obs: 471820; Mean: 2.01; Std. Dev.: 0.97; Min: 1.00; Max: 5.00
- Shareholder rights — Obs: 448254; Mean: 0.63; Std. Dev.: 0.20; Min: 0.09; Max: 1.00
- Creditor rights — Obs: 451908; Mean: 1.98; Std. Dev.: 1.04; Min: 0.00; Max: 4.00
- Creditor rights alternative — Obs: 421074; Mean: 1.98; Std. Dev.: 1.07; Min: 0.00; Max: 4.00
- Cost of bankruptcy — Obs: 421074; Mean: 9.49; Std. Dev.: 7.19; Min: 1; Max: 38
- Control of corruption — Obs: 471820; Mean: 1.06; Std. Dev.: 0.92; Min: -1.31; Max: 2.59
- Corruption index — Obs: 471820; Mean: 1.14; Std. Dev.: 0.93; Min: -1.08; Max: 2.39
- Industry informational opacity — Obs: 471820; Mean: 4.65; Std. Dev.: 1.03; Min: 0; Max: 8
- Industry competition — Obs: 471820; Mean: 0.10; Std. Dev.: 0.14; Min: 0; Max: 1
- FIRE industry — Obs: 87248; Mean: 1; Std. Dev.: 0.00; Min: 1; Max: 1
- Non-FIRE industry — Obs: 384572; Mean: 1; Std. Dev.: 0.00; Min: 1; Max: 1
- Firm size — Obs: 325540; Mean: 13.53; Std. Dev.: 3.51; Min: 0; Max: 27
- Firm leverage — Obs: 471820; Mean: 2.32082; Std. Dev.: 110.04; Min: 0; Max: 28646
- Firm current leverage — Obs: 274467; Mean: 5.37; Std. Dev.: 230.34; Min: 0; Max: 56792
- Firm profitability — Obs: 310279; Mean: -5.82; Std. Dev.: 40.06; Min: -274; Max: 38
- Firm sales growth — Obs: 262297; Mean: 0.07; Std. Dev.: 0.60; Min: -12; Max: 11
- Firm dependence on external finance — Obs: 220259; Mean: 3.99; Std. Dev.: 46.02; Min: -151; Max: 326

### Table E2. List of Countries Analyzed (as listed)
- Argentina, Australia, Austria, Belgium, Brazil, Chile, China, Colombia, Czech Republic, Denmark, Egypt, Finland, France, Germany, Greece, Hong Kong, Hungary, India, Indonesia, Ireland, Israel, Italy, Japan, Kenya, Kuwait, Lebanon, Malaysia, Mexico, Netherlands, New Zealand, Nigeria, Norway, Pakistan, Peru, Philippines, Poland, Portugal, Saudi Arabia, Singapore, South Africa, South Korea, Spain, Sweden, Switzerland, Taiwan, Thailand, Turkey, United Arab Emirates, United Kingdom, United States*, Venezuela
- Note: * United States excluded from regressions in order to avoid econometric biases.

### Table E3. Correlation Matrix Between Risk-Taking Measures (selected entries)
- z-score with Return on Assets (ROA): 0.16*
- z-score with Capital-Asset Ratio (CAR): 0.02*
- z-score with σ(ROA): -0.25*
- z-score with R&D Index: -0.07*
- Return on Assets (ROA) with Capital-Asset Ratio (CAR): 0.07*
- Return on Assets (ROA) with σ(ROA): -0.55*
- Return on Assets (ROA) with R&D Index: -0.21*
- Capital-Asset Ratio (CAR) with σ(ROA): -0.06*
- σ(ROA) with R&D Index: 0.18*
- Notes: The stars (*) represent significance at the 1% level. The z-score is negatively proxying for risk-taking. The standard deviation of return on assets and the research and development index are positively proxying for firm risk-taking.

### Table E4. Principle Component Analysis of Governance Indicators
- PCA1 / PCF1 coefficients:
  - Rule of Law — 0.41 / 0.17
  - Regulatory Quality — 0.39 / 0.16
  - Government Effectiveness — 0.40 / 0.16
  - Voice and Accountability — 0.36 / 0.15
  - Control of Corruption — 0.40 / 0.16
  - Political Stability — 0.31 / 0.13
  - Efficiency of Debt Enforcement — 0.34 / 0.14
- Notes: PCA components are obtained through principle component analysis. PCF components are obtained through principle factor analysis. We use PCA1 coefficients, but alternatively PCF1 could also be used since the correlation between PCA1 and PCF1 is 100%.

*Source: Appendix D and Appendix E, _wp12210 - Appendix D: Data Sources and Definitions*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2012/_wp12210.pdf_