FIRM-LEVEL DETERMINANTS OF POLITICAL INFLUENCE

Transcription

1 ECONOMICS & POLITICS DOI: /j x Volume 22 November 2010 No. 3 FIRM-LEVEL DETERMINANTS OF POLITICAL INFLUENCE ALBERTO CHONG* AND MARK GRADSTEIN This paper uses a large cross-country survey of business firms to assess their influence on government policies. When controlling for endogeneity, we find that such an influence is associated with larger firms and to a lesser extent with government ownership, but not with the degree of competition. We also find that firms perception of being politically influential is enhanced with the country s level of institutional quality. 1. INTRODUCTION TWO DIAMETRICALLY opposite influential theories speculate about the motivation and rationale of government intervention in the economy. The public interest theory, put forward in Pigou (1938), states that the government acts to achieve social benefit and to correct market failures. In contrast, the capture theory, originated in Stigler (1971), hypothesizes that the government is an agent of powerful commercial interests; 1 similar arguments come out of the rent-seeking literature (Krueger, 1974). These competing views and some of their implications are discussed in depth in Glaeser and Shleifer (2003). In reality, however, government policies rarely correspond to either of the two extremes. Progressive income taxation, uniform public education, and old age policies, as well as air pollution regulations, are all examples of public interestminded approaches, and they are commonly used across countries. In contrast, monopoly regulation, trade policies, or financial regulations are often viewed as being, to a large extent, influenced by commercial interests, the degree of which may in principle vary significantly across countries. In the light of these considerations, it seems useful to characterize the circumstances of firms influence over government policies. In particular, one issue is the profile of politically influential firms; specifically, what characteristics make firms more likely to exert political influence? Another interesting issue is identifying the extent to which these firms stand to disproportionately gain and the policy aspects that are especially prone to political influence. In particular, to the extent that government policies are found to be responsive to the influence of business firms, this would provide support for Stigler s (1971) view of government intervention. Corresponding author: Alberto Chong, Inter-American Development Bank, 1300 New York Avenue NW, Washington, DC 20577, USA. albertoch@iadb.org 1 Cf. With its power to prohibit or compel, to take or give away money, the state can and does selectively help or hurt a vast number of industries (Stigler, 1971). 233

2 234 CHONG AND GRADSTEIN This paper provides an empirical scrutiny of these issues by studying a range of outcomes, pertaining to policy impact. Existing literature typically uses direct involvement of politicians in the operation of a firm as a proxy for political connections. Arguing that such direct political connections are only one channel through which firms may affect policy making, we focus on the more general issue of the firms influence using information on their own perceptions. 2 This approach captures, in particular, lobbying of politicians, which has been of documented significance in the United States (Kroszner and Stratmann, 1998). This paper is the first to use all inclusive measures of political influence; while self-reported, it captures the overall level of a firm s connection. This aspect is novel to the existing literature, which typically focuses on one dimension of political connections. To this end, we use a large firm-level survey across countries recently conducted by the World Bank, the World Business Environment Survey, which elicited responses by firms managers about their respective policy influence at various levels of policy making and across the globe. These responses serve as our proxies for actual political influence. Consistent with existing literature, we find that larger firms perceive of themselves as exerting more policy influence. Government ownership is also associated with a larger perception of influence, but the relationship is not always statistically significant. Foreign shares ownership is associated with less influence, but again this relationship is not always robust. We also find that institutional weakness is associated with stronger perceived policy influence. Unlike existing literature, however, we do not find that a competitive environment is associated with less political influence. Conceptually, this paper is related to the literature on the motives of public officials, specifically, in the field of regulation as reviewed in Glaeser and Shleifer (2003). Also related is the literature on corruption; see Aidt (2003) for a review. The model of Choi and Thum (2007), which views the interaction between politicians and firms in the context of a mutual exchange of favors, is one way of thinking about the shaping of political influence; in the Appendix, we present a simple extension that is used to motivate some of the variables used. Empirically, this paper is related to the emerging literature that seeks to determine the extent to which politically connected firms are able to generate gains for themselves. This literature typically focuses on financial market outcomes such as access to credit or firm value (see for example Faccio, 2006a; Goldman et al., 2006; Khwaja and Mian, 2005, and references therein). Another focus, closer to this paper s emphasis, has been firms ability to affect legislation (Stratmann, 2002; see also De Figueiredo and Silverman, 2007, for analysis of university lobbying in this regard). Earlier literature, represented by Fisman (2001) and Goldman et al. (2006, 2 Goldman et al. (2006), in the U.S. context, use the amount of political contributions as a measure of political influence.

3 FIRM-LEVEL DETERMINANTS OF POLITICAL INFLUENCE 2008), analyzed important events that could have affected firms depending on the extent of their political connections. Much of this research was carried out in the context of a specific country often a developing country or, alternatively, in the U.S. context. For example, Goldman et al. (2008) analyze how the change in party control in the aftermath of the 1994 midterm election in the United States affected the value of procurement contracts of public corporations. More recently, Faccio (2006a, 2006b) and Faccio and Parsley (2006) have studied the outcomes of political connections in a cross-country sample. This literature has provided very useful empirical frameworks and insights, generally concluding that political connections matter, more so in countries with weak institutions EMPIRICAL STRATEGY Our data come from a survey from the World Bank Group that seeks to provide detailed information from enterprises on the condition of the private sector in many countries around the world. In particular, the aim of the survey is to measure the quality of governance and public services including the extent of corruption and to provide better information on constraints to private sector growth from the enterprise perspective. An important advantage of this survey is that it allows international comparisons as the data collection standards follow homogeneous methodologies (World Bank, 2000). The field work was conducted between 1999 and 2000 by private polling of each firm that filled some basic requirements. In particular, the survey was targeted to a representative sample of firms filling criteria such as sector, size, location, and ownership characteristics. 3 The objective was to gather information on a sizeable number of firms in countries around the world, which was accomplished for most of the sample. 4 The sample consists of firm-level survey 3 The particular requirements that had to be filled by the sample selected were as follows. Sector: in each country, the sectoral composition in terms of manufacturing (including agro-processing) vs. services (including commerce) will be determined by relative contribution to GDP, subject to a 15% minimum for each category. Size: at least 15% of the sample shall be in the small and 15% in the large size categories. Ownership: at least 15% of the firms will have foreign control. Exporters: at least 15% of firms will be exporters, meaning that some significant share of their output is exported. Location: at least 15% of firms will be in the category small city or countryside. 4 The countries and number of firms (in parentheses) included in the survey are as follows: Albania (133), Argentina (57), Armenia (106), Azerbaijan (102), Belarus (98), Bolivia (55), Botswana (49), Brazil (80), Bulgaria (99), Cameroon (39), Canada (43), Chile (45), China (47), Colombia (57), Costa Rica (31), Cote d Ivoire (52), Croatia (97), Czech Republic (110), Dominican Republic (58), Ecuador (42), Egypt (11), El Salvador (39), Estonia (120), Ethiopia (35), France (33), Germany (47), Ghana (31), Guatemala (22), Haiti (20), Honduras (23), Hungary (105), Indonesia (39), Italy (48), Kazakhstan (97), Kenya (59), Lithuania (112), Madagascar (48), Malawi (30), Malaysia (22), Mexico (30), Moldova (98), Namibia (47), Nicaragua (17), Nigeria (32), Pakistan (30), Panama (30), Peru (51), Philippines (44), Poland (196), Portugal (16), Romania (100), Russia (498), Senegal (18), Singapore (64), Slovakia (106), Slovenia (100), South Africa (63), Spain (59), Sweden (69), Tanzania (25), Thailand (211), Trinidad and Tobago (50), Tunisia (30), Turkey (119), the United Kingdom (32), the United States (32), Uganda (53), Ukraine (197), Uruguay (31), Venezuela (54), Zambia (42), and Zimbabwe (66).

4 236 CHONG AND GRADSTEIN responses of thousands of firms in some 80 countries, many of them developing andintransition. 5 The survey asked each business to rank the constraints or problems impacting on their operations. This process involved an extensive questionnaire undertaken via a face-to-face interview with either the firm managers or the firm owners of each company. As a result, the survey reports comparative measurements based on firms perceptions about their business environment as shaped by a variety of economic and policy factors. Choi and Thum (2007) and an extension in the Appendix contains a simple modeling approach to firms political influence. These models point to firm size and institutional quality as its important determinants. Consequently, we posit the following empirical relationship: I ic ¼ a þ b 1 Institut:Quality ic þ b 2 Size i þ b 3 X i þ b 4 Z c þ e ic ; ð1þ where I ic represents the level of influence firm i has on the government of country c. We focus on firm size and institutional quality proxy as our main variables of interest, also including additional firm-level explanatory variables, represented by vector X ic, that include the ownership of the firm, the number of competitors in a particular market, the sector where it operates, the logarithm of the sales amount the previous year, the sales growth in the previous three years, whether the firm exports or not, and whether the firm is a holding or has operations in other countries. We also add a set of country dummies, represented by Z c. Finally, e ic is a random error term. This full specification is considered to be our benchmark specification. We select proxies that capture the extent to which a firm may influence the executive, the legislature, and sector ministries, as detailed in Table 1, which lists all the definitions of the variables used in this paper. Also, in Table 2 the summary statistics of all the variables are presented, only considering the sample with the observations that do not have missing data in the variables we need. Still, our usable sample is made up of nearly 4,000 observations. Our dependent variable is based on the responses of the firms to questions on government influence that range from category 1 (never influential) to 5 (very influential). Inspection of the distribution of the responses to these questions yields somewhat similar patterns. For instance, about 7% of the surveyed firms consider themselves frequently or very influential across the above-mentioned channels of influence, while approximately 30% report being just influential or seldom influential, and around 60% consider themselves never influential. The pattern of responses hovers around similar percentages regardless of the specific variable considered. In fact, we use all these variables as alternative measures in order to test for the robustness of our findings. 5 Since not all the firms answered all our questions of interest, the sample size used in this paper is somewhat smaller, as described below.

5 FIRM-LEVEL DETERMINANTS OF POLITICAL INFLUENCE 237 TABLE 1 VARIABLE DESCRIPTION Variable Firm characteristics Foreign shares ownership State shares ownership Sector Number of competitors log(sales in the last year) Firm growth If it exports If the firm has holdings or operations in other countries Size Legal organization Influence on the government Influence on the executive Influence on the legislature Description If the firm has shares of foreign ownership, the variable takes the value of 1, and otherwise, 0. Source: World Bank (2000) If the firm has shares of state ownership, the variable takes the value of 1, and otherwise, 0. Source: World Bank (2000) The sector of the firm could belong to the following sectors: (1) manufacturing, (2) service, (3) agriculture, (4) construction, and (5) others (commerce, mining and quarrying, electricity, gas, and water). The last one (others) is the omitted category in all the regressions of the document. Source: World Bank (2000) Number of competitors in the same line of business. Source: World Bank (2000) Logarithm of the amount of sales last year. Source: World Bank (2000) Percentage of growth of sales in the last three years. Answer to the question: Please estimate the growth of your company s sales over the past three years. Source: World Bank (2000) Dummy variable that takes the value of 1 if the firm exports and 0 if it does not. Source: World Bank (2000) Dummy variable that takes the value of 1 if the firm has holdings or operations in other countries, and 0 if it does not. Source: World Bank (2000) The survey defines the size of the firms according to the number of employees. In that way, we have (1) small (50 employees or fewer), (2) medium (between 51 and 500 employees), and (3) large firms (more than 500 employees). The variable is assumed as continuous in all the regressions. Source: World Bank (2000) Four dummies that answer the question: What is the legal organization of this company: (1) single individual proprietorship, (2) corporation, privately held, (3) corporation listed on stock exchange, and (4) cooperative, partnership, and others. The omitted dummy in the regressions is the last category. Source: World Bank (2000) The question in the survey was: When a new law, rule, regulation, or decree is being discussed that could have a substantial impact on your business, how much influence does your firm typically have at the national level of the executive on the content of that law, rule, regulation, or decree? Would you say very influential, frequently influential, influential, seldom influential, or never influential? The variable we use ranges from (1) never influential to (5) very influential. Source: World Bank (2000) The question in the survey was: When a new law, rule, regulation, or decree is being discussed that could have a substantial impact on your business, how much influence does

6 238 CHONG AND GRADSTEIN TABLE 1 Continued Variable Influence on ministry Institutional quality proxies Strong institutions: taxes and regulations Good functioning of the judiciary Strong institutions: tax regulation and administration Strong institutions: high taxes Description your firm typically have at the national level of the legislature on the content of that law, rule, regulation, or decree? Would you say very influential, frequently influential, influential, seldom influential, or never influential? The variable we use ranges from (1) never influential to (5) very influential. Source: World Bank (2000) The question in the survey was: When a new law, rule, regulation, or decree is being discussed that could have a substantial impact on your business, how much influence does your firm typically have at the national level of the ministry on the content of that law, rule, regulation, or decree? Would you say very influential, frequently influential, influential, seldom influential, or never influential? The variable we use ranges from (1) never influential to (5) very influential. Source: World Bank (2000) Answer to the question: Please judge on a four-point scale how problematic is the taxes and regulations factor for the operation and growth of your business: (1) no obstacle; (2) minor obstacle; (3) moderate obstacle; and (4) major obstacle. Source: World Bank (2000) Answer to the question: Please judge on a four-point scale how problematic is the tax administration regulations factor for the operation and growth of your business: (1) no obstacle; (2) minor obstacle; (3) moderate obstacle; and (4) major obstacle. The variable scale was changed to: (1) major obstacle (that implies bad functioning) to (4) no obstacle (good functioning). Source: World Bank (2000) Answer to the question (in the regulation section of the survey): Please judge on a four-point scale how problematic is the regulatory area of tax administration regulations factor for the operation and growth of your business: (1) no obstacle; (2) minor obstacle; (3) moderate obstacle; and (4) major obstacle. Source: World Bank (2000) Answer to the question (in the regulation section of the survey): Please judge on a four-point scale how problematic is the regulatory area of high taxes factor for the operation and growth of your business: (1) no obstacle; (2) minor obstacle; (3) moderate obstacle; and (4) major obstacle. Source: World Bank (2000) Country-level control variables Average years of schooling Average schooling years in the total population over age 25. Source: World Bank (2007) log(gdp) Logarithm of the average per capita GDP for the period , expressed in constant 2000 US dollars. Source: World Bank (2007)

7 FIRM-LEVEL DETERMINANTS OF POLITICAL INFLUENCE TABLE 1 Continued 239 Variable Description Variables compared to the political influence proxies Number of steps to start a Natural logarithm of the number of different procedures that business a start-up business has to comply with to obtain legal status, i.e. to start operating as a legal entity. Source: World Bank (2008) Number of days to start a business Investor protection index Central or national government is very helpful for your business Local or regional government is very helpful for your business Natural logarithm of the number of days required to obtain legal status to operate a firm in Source: World Bank (2008) The strength of investor protection index is the average of the extent of disclosure index, the extent of director liability index, and the ease of shareholder suits index. The index ranges from 0 to 10, with higher values indicating more investor protection. Source: World Bank (2008) Answer to the question: All in all, for doing business I perceive the central or national government as (1) very unhelpful, (5) very helpful. Source: World Bank (2000) Answer to the question: All in all, for doing business I perceive the local or regional government as (1) very unhelpful, (5) very helpful. Source: World Bank (2000) Developing new rules, Answer to the following question: Please evaluate the regulations or policies is usually statement: The process of developing new rules, regulations or such that businesses are policies is usually such that businesses are informed in informed in advance of changes advance of changes affecting them. Answers range from (1) affecting them never to (6) always. Source: World Bank (2000) Changes in economic and financial policies, which materially affect the firm, are completely predictable Changes in rules, law, and regulations, which materially affect the firm, are completely predictable For important changes in laws or policies, the government takes into account concerns voiced by the firm or its business association Information of the laws and regulations affecting the firm is easy to obtain Changes in economic and financial policies are (1) completely unpredictable, (5) completely predictable. Source: World Bank (2000) Changes in rules, law, and regulations, which materially affect the firm, are (1) completely unpredictable, (5) completely predictable. Source: World Bank (2000) Answer to the following question: In case of important changes in laws or policies affecting my business operation the government takes into account concerns voiced either by me or by my business association. Answers range from (1) never to (6) always. Source: World Bank (2000) Agreement with the statement: In general, information on the laws and regulations affecting my firm is easy to obtain. Source: World Bank (2000) In Table 3, we provide an additional set of variables that show that the responses to our variables of interest are also consistent with other measures typically linked with influence and potentially specific favorable legislation to firms. In fact, when using variables related to the acknowledgment of help

8 240 CHONG AND GRADSTEIN TABLE 2 SUMMARY STATISTICS Variable No. of observations Mean SD Minimum Maximum Influence on the government Influence on the executive 3, Influence on legislature 3, Influence on ministry 3, Firm characteristics Size 3, Foreign shares ownership 3, State shares ownership 3, Number of competitors 3, Sector: manufacturing 3, Sector: service 3, Sector: agriculture 3, Sector: construction 3, log(sales in the last year) 3, Firm growth 3, If the firm exports 3, If the firm has holdings or 3, operations in other countries Legal org.: single proprietorship 3, Legal org.: corporation, privately held 3, Legal org.: corporation listed on a stock exchange 3, Institutional quality (proxies) Strong institutions: taxes regulation 3, Good functioning of judiciary 3, Strong institutions: tax regulations 3, and admin. Strong institutions: high taxes 3, Country-level variables Average years of schooling log(gdp) at different levels of government, the development of new rules, regulations, or policies, and, in general, laws and policies that favor businesses, we find that firms acknowledge that they are informed in advance of changes affecting them, that the government takes into account concerns voiced by the firm or its business association, and that government at all levels is perceived as being helpful to firm development. Furthermore, we also consider other country-level external measures, namely the number of steps and days to start a business in a country, and the investor protection index, all obtained from the well-known Doing Business database from the World Bank.

9 FIRM-LEVEL DETERMINANTS OF POLITICAL INFLUENCE 241 TABLE 3 FIRMS POLITICAL INFLUENCE PROXIES Extent of influence firms have on the Executive Legislature Ministry Number of steps to start a business Number of days to start a business Investor protection index Central or national government is very helpful for your business Local or regional government is very helpful for your business Developing new rules, regulations, or policies is usually such that businesses are informed in advance of changes affecting them Changes in economic and financial policies, which materially affect the firm, are completely predictable Changes in rules, law, and regulations, which materially affect the firm, are completely predictable For important changes in laws or policies, the government takes into account concerns voiced by the firm or its business association Information of the laws and regulations affecting the firm is easy to obtain Notes: Pairwise correlation coefficients, with the corresponding p-value of significance level below. All variables are defined in Table 1. We find that all the corresponding pairwise correlation coefficients yield positive and statistically significant coefficients. The two main explanatory variables are the size of the firm and the quality of institutions. The former is measured in terms of number of employees and classified into three categories, namely (i) small firms, which are ones with 50 employees or fewer; (ii) medium firms, which are ones that have between 51 and 500 employees; and (iii) large firms, defined as those firms with more than 500 employees. On the other hand, institutional quality is captured by the perception of how problematic taxes and regulations are as a factor for the operation and growth of the firm. The answers to this question range from 1 (no obstacle) to 4 (major obstacle). Thus, the more problematic taxes regulation is a factor for the firm, the stronger the institution of taxation. We use an ordered probit technique to analyze the key determinants of the likelihood of having influence. This approach allows us to take into

10 242 CHONG AND GRADSTEIN consideration the ordinal nature of our dependent variables. Thus, the empirical specification follows the following methodology: ProbðI i ¼ 1Þ ¼Fðm 1 b 0 WÞ; ProbðI i ¼ 2Þ ¼Fðm 2 b 0 WÞ Fðm 1 b 0 WÞ;.. ProbðI i ¼ nþ ¼1 Fðm n 1 b 0 WÞ; where I i is a random variable indicating firm influence and W is the full set of explanatory variables. After some manipulation in order to remove indeterminacy, the log-likelihood can be derived by defining, for each individual, d ij ¼ 1 if alternative j is chosen by individual i, and 0 if not, for the n possible outcomes. For each i, one and only one of the d ij is 1 (Schmidt and Strauss, 1975). In particular, the log-likelihood is ln L ¼ Xn i¼1 X 4 fd i1 ln Fðm 1 b 0 xþþ j¼2 ð2þ d ij ln Fðm j b 0 xþ Fðm j 1 b 0 xþ þ d i5 ln½1 Fðm 4 b 0 xþšg: ð3þ Once the likelihood is formed, the estimation of the unknown parameters m and b can be undertaken. The estimated ancillary parameters, m, along with the estimated b, maximize the log-likelihood function (3). The impact of a change in an explanatory variable on the estimated probabilities of the lowest and the highest of the ordered classification is direct. Note that an estimated b value does not estimate the change in the probability of a given outcome. Manipulating (2), it can be shown that that the marginal effects of the attributes on the corresponding probabilities i ¼ 1Þ ¼ 1 b 0 WÞb k ; i ¼ 2Þ 2 b 0 WÞ fðm 1 b 0 WÞŠb k ; i ¼ nþ ¼ 4 b 0 WÞb k : ð4þ k Finally, a fit measure may be obtained through the following restricted log-likelihood: ln L 0 ¼ Xn n j ln t j ¼ Xn t j ln p j ; ð5þ t j¼1 j¼1

11 FIRM-LEVEL DETERMINANTS OF POLITICAL INFLUENCE where p j is the sample proportion of observations corresponding to the particular quintile. All our empirical results include fixed effects and robust standard errors that have been clustered at the country level. Finally, in order to test for the robustness of our results, we use other proxies related to the perception of the institutional quality of the country, in particular, the functioning of the judiciary as well as two other firm constraints related to taxes. We also include other basic controls related to the characteristics of the firm, as described in the lines above DETERMINANTS OF INFLUENCE The basic findings are presented in Table 4. This table shows results for three different specifications for each of the influence proxies, namely, the extent of influence that firms have on the executive, the legislature, and ministries. The first specification includes our main explanatory variables, namely size of the firm and institutional quality as measured by how taxes and regulations affect the firm. The second specification includes additional firm characteristics, in particular, state and foreign ownership shares, sector dummies, sales, and the rate of growth of the firm. The third specification adds variables related to foreign operations, namely, whether the firm exports and whether the firm has holdings or operations in other countries. The latter is the most complete specification. The cost of the addition of these variables is the sharp reduction in the number of observations, though. However, the findings among the different specifications presented do not vary much. We use the most complete specification as our benchmark in order to calculate corresponding marginal effects for each of our dependent variables. These calculations are presented in Table 5. Unlike in simple probits, presenting the full set of marginal effects using ordered probits becomes quite impractical as one would have to show the marginal coefficients for each of the categories of the dependent variable (see equation (4)). Thus, for economy, we focus on the marginal effects at mean levels, which, in the case of our proxies of influence, are 2, as shown in Table 2. 6 We observe that our results are very similar across regressions, which is not very surprising, given the relatively high correlations among the various channels of influence as noted above. Our two main explicative variables are statistically significant at conventional levels in all regressions. The lower the quality of the institutions, the higher the probability that the firms will have more influence over the government. With respect to firm size we find that larger firms are linked to a higher probability of having influence, as expected. These findings are shown in Table 4. In terms of the marginal effects, we find that if the index of institutional quality increases by one unit, the 6 Full set of marginal coefficients are available on request. Results do not change.

12 244 CHONG AND GRADSTEIN TABLE 4 DETERMINANTS OF POLITICAL INFLUENCE (ORDERED PROBIT REGRESSIONS, COEFFICIENTS REPORTED) Dependent variable: extent of influence firms have on the Executive Legislature Ministry (1) (2) (3) (4) (5) (6) (7) (8) (9) Strong institutions: taxes regulation (0.023) (0.024) (0.025) (0.023) (0.025) (0.025) (0.022) (0.026) (0.027) Size (0.032) (0.036) (0.035) (0.032) (0.039) (0.039) (0.034) (0.038) (0.037) Foreign shares ownership (0.060) (0.062) (0.057) (0.065) (0.055) (0.061) State shares ownership (0.073) (0.074) (0.065) (0.065) (0.054) (0.055) Number of competitors (0.039) (0.038) (0.034) (0.034) (0.038) (0.038) Sector: manufacturing (0.274) (0.299) (0.294) (0.319) (0.195) (0.212) Sector: service (0.271) (0.296) (0.311) (0.338) (0.205) (0.222) Sector: agriculture (0.286) (0.312) (0.319) (0.347) (0.212) (0.227) Sector: construction (0.282) (0.308) (0.315) (0.340) (0.216) (0.232) log(sales in the last year) (0.012) (0.012) (0.010) (0.010) (0.011) (0.011) Firm growth (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) If the firm exports If the firm has holdings or operations in other countries (0.043) (0.044) (0.040) (0.059) (0.056) (0.056) Observations 6,033 3,902 3,834 6,042 3,915 3,846 6,030 3,900 3,829 Number of countries Log pseudo-likelihood 6,106 3,847 3,782 5,951 3,787 3,710 6,116 3,848 3,760 Pseudo-R w 2 25,954 12,639 34, ,566 49,621 46,091 80,614 23,123 24,563 Notes: Robust standard errors are in parentheses. All regressions include country dummies and standard errors adjusted for clustering by country. Significant at 10%. Significant at 5%. Significant at 1%.

13 FIRM-LEVEL DETERMINANTS OF POLITICAL INFLUENCE 245 TABLE 5 DETERMINANTS OF POLITICAL INFLUENCE (ORDERED PROBITS, MARGINAL EFFECTS SHOWN: CHANGE FROM CATEGORY 2 TO 3, Prob(Y ¼ 3 X)) Dependent variable: extent of influence firms have on the Executive Legislature Ministry (1) (2) (3) Strong institutions: taxes regulation (0.003) (0.003) (0.003) Size (0.005) (0.005) (0.005) Foreign shares ownership (0.007) (0.007) (0.007) State shares ownership (0.010) (0.009) (0.009) Number of competitors (0.005) (0.004) (0.005) Sector: manufacturing (0.038) (0.038) (0.028) Sector: service (0.035) (0.038) (0.028) Sector: agriculture (0.040) (0.045) (0.033) Sector: construction (0.040) (0.044) (0.033) log(sales in the last year) (0.001) (0.001) (0.001) Firm growth (0.000) (0.000) (0.000) If the firm exports (0.005) (0.005) (0.005) If the firm has holdings or operations in other countries (0.008) (0.007) (0.008) Notes: The marginal effects shown in this table come from the regressions (3), (6), and (9) of Table 4. Robust z-statistics are in parentheses. Standard errors clustered at the country level. Significant at 10%. Significant at 5%. Significant at 1%. probability that the firm increases its influence from two to three units is reduced by around 0.5% for each of the three influence measures used. On the other hand, in the case of firm size if the increase goes from the second size category to the third size category, the probability of influence increases by about 3%. In short, when measuring the marginal effects at the mean, size appears to have a larger bearing on influence rather than institutional quality. 7 7 The addition of country-level variables as controls instead of country dummies tends to increase the statistical significance of our main explanatory variables. However, the added country-level variables tend to yield little statistical significance.

14 246 CHONG AND GRADSTEIN Interestingly, we do not find that foreign ownership of domestic companies is linked to a higher probability of perceiving them as influential at any level of the government as the corresponding coefficients in our regressions in Table 4 are not statistically significant at conventional levels. On the contrary, firms that share ownership with the government are found to be substantially more influential than the ones only privately owned, as coefficients are statistically significant and positive. The corresponding marginal effects show that if the firm has state-owned ownership, the probability that the firm is just influential increases by 4%. Additionally, sector dummies also appear to be important, especially for firms that belong to the service sector. The corresponding coefficients of the service dummies are consistently positive and statistically significant at conventional levels in all the regressions. Along the same lines, higher rates of growth in firms are linked to a higher probability of having an influence on the government, at all levels of government. However, the corresponding economic linkages appear to be limited as the marginal effect is small. In fact, the latter is lower than 0.1% even though the coefficient is still statistically significant at conventional levels. Finally, as shown in Tables 4 and 5, both firms that export and firms that hold operations in other countries are positive and statistically significant. Furthermore, firms that export have around more than 1% increased probability of having influence, regardless of the type of public forum considered. Interestingly, firms that have holdings or operations in other countries have an increased probability of influence of around 3%. Firm size, firm growth, and our proxy of institutional quality may be endogenous to the extent of the political influence of firms. One can plausibly argue that firms with more influence are the ones that grow more in size and sales. At the same time, if there are firms that have more influence on the government, it is possible that the institutional quality could be affected, by the reduction of efficiency or strength in taxes and regulation. In order to deal with this potential problem, we apply an instrumental variables approach. The instruments used are the legal organization of the firm, the country s average years of schooling, and the country s gross domestic product (GDP). The first variable is a firm-level variable that asks whether the firm has a single proprietorship, whether it is a corporation privately held, whether it is a corporation listed on the stock exchange, and whether it is a cooperative, partnership, or others. It is reasonable to expect that, on average, corporations listed on stock exchanges will be larger and less fragile than the ones with individual owners. This is also associated with the fact that the simple correlation between the firm organization dummies and firm size is about 0.20 in all cases and is statistically significant. On the other hand, the correlation of this variable with the dependent variables related to the extent of influence firms have is around 0.08 or less and it is not statistically significant at standard levels.

15 FIRM-LEVEL DETERMINANTS OF POLITICAL INFLUENCE The second instrument used is a country-level variable that reflects the level of education in each economy. Economies with higher levels of education are expected to have higher growth rates through higher firm productivity. The correlation index between education and influence shows that this correlation is around 0.6 and it is statistically significant with firm growth, but it is less than 0.2 and it is not statistically significantly correlated with political influence. That is, education per se appears not to be correlated with political influence. The third instrument used is the logarithm of GDP (average ). Countries with a higher GDP tend to have better managed institutions, as extensively shown in the literature (Kaufmann et al., 2005). The domestic product, however, is not a priori highly correlated with potential influence in the formal organization of the political system. In fact, we find that the correlation between our potential instrument and our explanatory variable is around 0.76 and is statistically significant at 5%, but such a variable yields a correlation index of less than 5% with the dependent variables on influence of political system, which, additionally, are not statistically significant. We find that our main results hold when using instrumental variables, as shown in Table 6. In the nine regressions presented in this table, which correspond to the specifications of Table 4, we find that the size of the firm yields a positive and statistically significant coefficient. Furthermore, we find that the proxy for institutional quality yields a negative and statistically significant link to political influence. When calculating marginal effects, we find that the effect of institutional quality is economically larger than the one of size. The former yields an increased probability of around 9% 10% while the corresponding marginal effects of the latter change much less in relation to the uncorrected results, as shown in Table 4. In relation to our other controls, we find that the coefficient of foreign ownership remains negative but becomes statistically significant at conventional levels, which is consistent with related previous evidence (O Neil, 1994). However, government ownership yields weak statistical significance in the case of influence on the executive, and no statistical link with the other dependent variables considered. Also, sector dummies become statistically significant. It appears that firms in manufacturing, service, agriculture, or construction exert about 5% 13% more political influence than firms in other sectors. In order to test for the robustness of our findings above, we also test our benchmark specifications with other proxies for institutional quality, in particular (i) the perception on the functioning of the judiciary; (ii) a measure on high taxes; and (iii) a measure on taxes regulation and administration. (The definitions of these variables are provided in Table 1.) The results are presented in Table 8. With respect to our main variables of interest, size and institutional quality, our results are very similar to before both in terms of size and statistical significance. In particular, when using these proxies we 247

16 248 CHONG AND GRADSTEIN TABLE 6 DETERMINANTS OF POLITICAL INFLUENCE (IV ORDERED PROBITS, COEFFICIENTS REPORTED) Dependent variable: extent of influence firms have on the Executive Legislature Ministry (1) (2) (3) (4) (5) (6) (7) (8) (9) Strong institutions: taxes regulation (0.205) (0.648) (0.156) (0.235) (0.661) (0.137) (0.243) (0.606) (0.165) Size (0.064) (0.151) (0.151) (0.072) (0.146) (0.146) (0.071) (0.151) (0.151) Foreign shares ownership (0.090) (0.072) (0.101) (0.077) (0.088) (0.069) State shares ownership (0.153) (0.094) (0.174) (0.100) (0.152) (0.112) Number of competitors (0.078) (0.047) (0.074) (0.044) (0.068) (0.046) Sector: manufacturing (0.284) (0.210) (0.235) (0.093) (0.228) (0.130) Sector: service (0.255) (0.193) (0.212) (0.090) (0.206) (0.129) Sector: agriculture (0.281) (0.253) (0.203) (0.145) (0.195) (0.154) Sector: construction (0.334) (0.213) (0.316) (0.137) (0.275) (0.163) log(sales in the last year) (0.014) (0.015) (0.012) (0.013) (0.013) (0.016) Firm growth (0.007) (0.002) (0.008) (0.002) (0.007) (0.002) If the firm exports If the firm has holdings or operations in other countries (0.058) (0.061) (0.051) (0.075) (0.074) (0.070) Observations 3,569 3,569 3,569 3,532 3,532 3,532 3,491 3,491 3,491 Number of countries Log pseudo-likelihood 3,536 3,524 3,524 3,390 3,377 3,377 3,420 3,405 3,405 Pseudo-R w 2 56,277 10,741 30,750 11,054 8,927 27,002 12,520 19,501 26,445 Notes: Robust standard errors are in parentheses. All regressions include country dummies and standard errors adjusted for clustering by country. In the first-stage regressions, the instrumented variables are firm size, firm growth, and the institutional quality proxies, and were estimated using a lineal probability (for firm size) and OLS models. The instruments used for each variable were three dummy variables that indicate the legal organization of the firm, the country s average years of schooling, and the country s GDP, respectively. Significant at 10%. Significant at 5%. Significant at 1%.

17 FIRM-LEVEL DETERMINANTS OF POLITICAL INFLUENCE 249 TABLE 7 DETERMINANTS OF POLITICAL INFLUENCE (ORDERED PROBITS WITH IV, MARGINAL EFFECTS SHOWN: CHANGE FROM CATEGORY 2 TO 3, Prob(Y ¼ 3 X )) Dependent variable: extent of influence firms have on the: Executive Legislature Ministry (1) (2) (3) Strong institutions: taxes regulation (0.018) (0.016) (0.021) Size (0.018) (0.017) (0.019) Foreign shares ownership (0.008) (0.008) (0.008) State shares ownership (0.012) (0.012) (0.014) Number of competitors (0.006) (0.005) (0.006) Sector: manufacturing (0.025) (0.011) (0.015) Sector: service (0.022) (0.010) (0.014) Sector: agriculture (0.033) (0.018) (0.016) Sector: construction (0.027) (0.016) (0.014) log(sales in the last year) (0.002) (0.001) (0.002) Firm growth (0.000) (0.000) (0.000) If the firm exports (0.007) (0.007) (0.006) If the firm has holding or operations in other countries (0.009) (0.008) (0.009) Notes: The marginal effects shown in this table come from the regressions (3), (6), and (9) of Table 6. Robust z-statistics are in parentheses. Standard errors clustered at the country level. Significant at 10%. Significant at 5%. Significant at 1%. find that while the effect of institutions is less than or equal to 1%, the one of firm size ranges from 3% to 4%. In Table 9, we apply instrumental variables to these robustness proxies and further confirm that firm size has a positive and statistically significant effect at the three levels of influence. Also, again, in the case of the institutional quality proxies, all the coefficients yield a negative and statistically significant coefficient. We find again that larger firms have more probabilities of being politically connected about 3% and 5% more and that the extent of their influence is an increasing function of institutional weakness, with a marginal effect that could be as high as 20%.

18 250 CHONG AND GRADSTEIN TABLE 8 ROBUSTNESS: DETERMINANTS OF POLITICAL INFLUENCE (ORDERED PROBITS) Dependent variable: extent of influence firms have on the Executive Legislature Ministry (1) (2) (3) Good functioning of judiciary Coefficients Good functioning of judiciary (0.022) (0.023) (0.024) Size (0.040) (0.043) (0.041) Marginal effect (change from category 2 to 3, Prob(Y ¼ 3 X)) Good functioning of judiciary (0.003) (0.003) (0.003) Size (0.005) (0.005) (0.006) Observations 3,420 3,430 3,419 Number of countries Pseudo-R Strong institutions: high taxes Coefficients Strong institutions: high taxes (0.023) (0.024) (0.023) Size (0.040) (0.042) (0.042) Marginal effect (change from category 2 to 3, Prob(Y ¼ 3 X)) Strong institutions: high taxes (0.003) (0.003) (0.003) Size (0.005) (0.005) (0.006) Observations 3,873 3,885 3,871 Number of countries Pseudo-R Strong institutions: tax regulations and administration Coefficients Strong institutions: tax regulations and administration (0.023) (0.021) (0.021) Size (0.039) (0.041) (0.041) Marginal effect (change from category 2 to 3, Prob(Y ¼ 3 X)) Strong institutions: tax regulations and administration (0.003) (0.002) (0.003) Size (0.005) (0.005) (0.006) Observations 3,886 3,899 3,884 Number of countries Pseudo-R Notes: Robust standard errors are in parentheses. The table reports the coefficients and marginal effects for our variables of interest based on the benchmark specification in Table 4. Significant at 10%. Significant at 5%. Significant at 1%.

19 FIRM-LEVEL DETERMINANTS OF POLITICAL INFLUENCE 251 TABLE 9 ROBUSTNESS: DETERMINANTS OF POLITICAL INFLUENCE (IV ORDERED PROBITS) Dependent variable: extent of influence firms have on the Executive Legislature Ministry (1) (2) (3) Good functioning of judiciary Coefficients Good functioning of judiciary (0.127) (0.112) (0.129) Size (0.142) (0.154) (0.152) Marginal effect (change from category 2 to 3, Prob(Y ¼ 3 X)) Good functioning of judiciary (0.015) (0.013) (0.016) Size (0.017) (0.018) (0.020) Observations 3,359 3,324 3,287 Number of countries Pseudo-R Strong institutions: high taxes Coefficients Strong institutions: high taxes (0.207) (0.187) (0.062) Size (0.155) (0.149) (0.154) Marginal effect (change from category 2 to 3, Prob(Y ¼ 3 X)) Strong institutions: high taxes (0.024) (0.021) (0.009) Size (0.018) (0.017) (0.019) Observations 3,542 3,505 3,465 Number of countries Pseudo-R Strong institutions: tax regulations and administration Coefficients Strong institutions: tax regulations and administration (0.031) (0.031) (0.032) Size (0.153) (0.147) (0.153) Marginal effect (change from category 2 to 3, Prob(Y ¼ 3 X)) Strong institutions: tax regulations and administration (0.004) (0.004) (0.004) Size (0.018) (0.017) (0.019) Observations 3,553 3,516 3,475 Number of countries Pseudo-R Notes: Robust standard errors are in parentheses. The table reports the coefficients and marginal effects for our variables of interest based on the benchmark specification in Table 6. In the first-stage regressions, the instrumented variables are firm size, firm growth, and the institutional quality proxies, and were estimated using a lineal probability (for firm size) and OLS models. The instruments used for each variable were three dummy variables that indicate the legal organization of the firm, the country s average years of schooling, and the country s GDP, respectively. Significant at 10%. Significant at 5%. Significant at 1%.