CORRUPTION, PROVINCIAL INSTITUTIONS AND CAPITAL STRUCTURE: NEW EVIDENCE FROM A TNSITIONAL ECONOMY

Using a unique !rm-provincial level panel dataset "om 2005 to 2011, this study for the !rst time investigates the role played by corruption and provincial institutions in determining a company’s capital structure in Vietnam’s legal environment. Contrasting to the majority of previous studies, the results show that corruption has an insigni!cant in#uence on a company’s bank loans, consistent with institutional theory. However, the role of corruption is di$erent for types of various capital structures a%er controlling for both unobservable characteristics and endogeneity problems. More speci!cally, corruption has signi!cantly positive in#uence on short-term capital structure, but a negative impact on long-term loans. All of these results hold a%er a series of robust tests.


Introduction
eoretically, corruption has been considered as a crucial factor in constructing a state's legal system, resource distribution and rms' behavior (Fan, Titman, & Twite, 2012). Corruption a ects a company's capital structure decision in di erent ways. On the one hand, corruption can lead to a decrease in bank credit. When investors intend to invest in a company, they expect to regain their capital based on criteria speci ed in the contract (Bolton & Dewatripont, 2005;Leland & Pyle, 1977). However, investors su er a higher risk in seriously corrupt countries and in the condition of the loose legal environment. ese higher risks and potential implementation costs make banks reluctant to o er credits or increase the credit standard in a manner that will increase the cost of securing external funding banks (La Porta, Lopez-de-Silanes, Shleifer & Vishny, 1997;Porta, Lopez-de-Silanes, Shleifer & Vishny, 1998;Shleifer & Vishny, 1993). In other words, banks are not motived to grant more loans or o er more credits to rms, or even upgrade the credit standards in the context of higher corruption.
By way of contrast, other perspectives (e.g., Stiglitz & Weiss, 1981) show that there is a positive linkage between corruption and capital structure. Adverse selection caused by asymmetric prior information between bank and debtors can lead to credit rationing. e existence of credit rationing suggests that some debtors choose to pay an interest rate far in excess of the o cial rate. Consequently, they are motivated to bribe bank o cers to obtain credit. When debtors actively bribe bank o cers to increase their chances of receiving credit, the corruption increases the company's bank credit.
In another approach, the role of rms' corruption behavior on the performance and capital structure is explained by institutional theory. is is considered as one of the most popular perspectives in transitional economies (e.g., Hoskisson, Eden, Lau & Wright, 2000;Wright, Filatotchev, Hoskisson & Peng, 2005). is approach shows that paying bribe is simply an entry cost of rms to join an established game and hence it may not a ect the e ciency and rm capital structure (North, 1990). e story can go as follows: When rms pay informal costs, this puts the pressure on neighbouring rms to follow their behaviours. As a result, corruption may have li le impact on their performance.
In the light of above theoretical perspectives, many empirical studies have been conducted from various countries, but the ndings are inconclusive, making it hard to make generalized inferences. For example, Welch (2011) takes advantage of the data from banks and regional corruption indexes, nding that Russia's corruption resulted in blocked bank credit. Similarly, De Carvalho (2008) used corporate data on Brazil and found that corruption prevents corporations from obtaining bank credit. On the other hand, Chen, Liu, and Su (2013) indicate that corruption contributes to companies' receipt of bank credit. Similarly, Fungáčová, Kochanova and Weill (2015) analysed data from 14 transition countries. eir common nding is that there is a positive correlation between corruption level and a company's receipt of bank credit.
Compared with previous studies on capital structure, this study has several di erences. Firstly, this study examines not only rm-level corruption but also the e ects of quality of provincial institution on capital structure. Secondly, most studies focus on the analysis in the US and other developed countries. ere is less empirical evidence on capital structure in developing countries, especially for transitional nations. is question is conducted by studying the context of Vietnam because there is no empirical evidence of the impact of bribe on rm capital structure in Vietnam. Also, despite implementation of the anti-corruption and anti-waste laws and various anti-corruption campaigns, bribes to public o cials remain a major challenge for business environment in Vietnam. Furthermore, this study considers not only the e ects of corruption on capital structure but also on types of capital structure. Finally, in terms of methodology, several empirical challenges arise when considering the linkage between corruption and rm capital structure. Unobservable characteristics and the endogeneity of ex-planatory variables are the main concerns; more importantly, the presence of potential dynamic endogeneity that can be understood as the past rms' leverage a ecting the current rms' leverage. Following Wintoki et al. (2012), I overcome these problems by using the two-step system dynamic panel GMM models.
Interestingly, contrary to the many ndings of previous studies, I nd that corruption does not a ect rms' capital structure a er controlling for heterogeneity, simultaneity and dynamic endogeneity. is nding supports the viewpoints of institutional theory and re ects the fact that corruption is widespread in Vietnam. Accordingly, engagement in corruption is considered as an entry fee and not related with rms' capital structure. However, paying bribe has negative impacts on rms' short-term capital structure and positive linkages with long-term rms' capital structure. e rest of this paper is structured as follows. e next section presents the background of the study. Data and methodology are presented in Section 3, and Section 4 displays empirical results. e last section contains conclusion and the summary of ndings.
2. e background of the study  stocks provides 13.7 percent of total nancial market, bonds and insurance companies only account for 1.8 percent and 9.3 percent, respectively. Also, a er nearly 30 years of renovation (Doi Moi), the Vietnamese economy has gained many achievements, transforming into one of the most dynamic markets in South East Asia. Besides the rapid growth and development, according to Nguyen and Van Dijk (2012), corruption in Viet Nam is more widespread than before. In spite of the anti-corruption activities implemented by the local government, the Vietnamese ranking was very low at 112 out of 168 countries in terms of corruption level in 2015 according to Transparency International (TI).
In addition, for Vietnam, big gaps between formal institutions laws and the enforcement capacity of the local authorities have been documented. Furthermore, the institutional quality across provinces developed unevenly -several provinces lag behind, others witness a signi cant improvement in economic governance and business investment (Malesky, 2007). is situation motivates us to consider whether corruption has an e ect on the capital structure of rms, and if so, how.

Data source
is study uses two data sources. First, data are extracted from the surveys by the Danish International Development Agency with the assistance of the Institute of Labour Science and Social A airs, the Central Institute for Economic Management and the University of Copenhagen. e results of these surveys are based on questionnaires every two years from 2005 to 2011, and this study employs the data in years 2005, 2007, 2009 and 2011. ese sources provide the information about over ten thousand private manufacturing enterprises in ten provinces in the Southern, Central and Northern regions of Vietnam. rough the surveys, many useful indicators such as the rm size, age and export, the gures about capital structure, i.e. the proportion of total debt to total asset, the short-term capital structure, the long-term capital structure and, especially, forms of bribery are recorded. As a result, the availability of data allows this study to consider the impact of corruption on rm capital structure.
Another data set is taken from the surveys of the Vietnam aggregated Provincial Competitiveness Index (PCI), which were implemented by the Vietnam Competitiveness Initiative in collaboration with the Vietnam Chamber of Commerce and Industry in the period 2005-2011 to assess the institutional quality of provinces or local governments. e survey o ers nine institutional sub-indices across the years of the period. ese indices include: First, entry costs including (i) time for a rm registration and land acquisition, (ii) time required for rms to complete all the necessary licenses needed to begin a business as well as the level of di culty to have such licenses/permits. Second, access to the acquired land and the security of business premises a er land acquisition.
ird, transparency and access to information, that is whether enterprises have access to appropriate planning and legal documents for doing their business, training and labour, as well as whether new laws are provided to enterprises su ciently and predictably implemented. Fourth, cost of time to handle regulatory compliance measure, e.g. bureaucratic compliance or decisions to implement local regulations. Fi h, informal payments measuring an enterprise's perception about the corruption from local o cials. Sixth, distortion o ering privileges to state owned enterprises, e.g. incentives, policy, and allocation of capital and credit sources toward state-owned enterprises. Seventh, services for private sector development, provinces private sector business growth promotion programs and the development of industrial zones and parks. Eighth, employment and worker training -whether/how provincial authorities promote vocational training and skills development for local rms. Ninth, legal institutions measuring the trust from rms on provincial courts and contract enforcement.
Combining two data sets together, I created a unique province -rm level panel dataset with 2684, 2483, 2515, 2449 observations in 2005, 2007, 2009 and 2011, respectively. More speci cally for the dataset, Table 2 provides the de nitions and statistical description of main variables in the model.

Methodology
Using a dynamic panel modelling approach to solve the dynamic nature of economic processes is becoming increasingly important in recent years (Flannery & Hankins, 2013). is dynamic process means that the current rm performance and other rmspeci c characteristics are driven by past performance. To address the "dynamic endogeneity", empirical analyses using rm performance as a dependent variable must be investigated in a dynamic framework in which lagged dependent variable(s) are used as explanatory variable(s) (Wintoki et al., 2012). Wooldridge (2009) noted that including lagged dependent variable(s) as explanatory variables in empirical models allows empiricists to account for unobserved historical factors which have potential impacts on current rm performance, thus mitigating omi ed variable bias. In addition, corruption is also a dynamic process, and hence the lag of corruption is also entered as an independent variable in the model. Furthermore, corruption can be di erent at sizes, age and industries. Consequently, a series of interactions between corruption with size, age and industries are controlled for and the model is speci ed as below: (1) Where Where: Y it is the outcome variable (as measured by a rm's capital structure) of rm i in year t; a s is the estimated coe cient on lagged dependent variables; Corruption is commonly understood as the abuse of power by public o cials for private gains (Svensson, 2005). According to Rand and Tarp (2012), bribe is measured as a dummy variable based on the question whether rms paid informal or communication fees in this study. e bribe payment or communications fees are mainly used for several purposes. For example, they are used to get connected with public services, to get licenses and permits, to gain government contract, to deal with procedures with banks.
Z is a set of rm-related explanatory variables ( rm size, rm age, and export) included in the model as guided by previous studies (e.g., Alves & Ferreira, 2011;Fungáčová et al., 2015). I also account for potential e ects arising from di erences across industries by including dummy variables for industry classi cation in the models. μ i represents time-invariant unobserved rm characteristics; time-speci c e ects are denoted by ω t , and ε it represents the classical error term.
Previous studies on rm performance (Vu, Tran, Nguyen, & Lim, 2016;Wintoki et al. (2012) suggest that the past information be captured adequately by two lags of the dependent variable. To examine this issue, I used a model speci cation in which the current capital structure is a dependent variable being regressed on two lags of past rm performance, and other covariates as in equation (1). Using this formulation, an insigni cant impact of Y it-2 on current rm nancial performance was con rmed. erefore, this suggests that using one-year lagged dependent variable as an explanatory variable in a rst-order autoregressive [AR(1)] structure is enough to address the potential dynamic endogeneity. e results are similar for other lagged values of other independent variables. is is in accordance with a study by Zhou, Fa , and Alpert (2014), which argues that an AR(1) structure appears to be unavoidable when almost all panel datasets used in corporate nance research are short. e AR(1) panel model speci cation is given in detail as follows: Prior studies also indicate that not controlling for institutional quality factors may bias the e ect of corruption on rm capital structure. For example, De Jong, Kabir, and Nguyen (2008) noted that it is not only rms' a ributes that have a direct impact on their capital structure but also factors such as the institutional quality of a country or a company's business practices and so forth that will produce in uences on the choice of capital structure. Corruption can "grease or sand the wheel" if the institutional environment is good or bad (Méon & Weill, 2010). us, indexes of institutional quality at provincial level (P m,jt ) are controlled for in the model. Also, lagged values of indices of institutional quality are entered in the model to account for unobserved historical factors which have potential impacts on current rm capital structure: With respect to estimation approach, given the presence of the AR(1) structure in equation (2), the pooled OLS (OLS) and the OLS with xed-e ects (FE) methods are likely to provide inconsistent estimates (Flannery & Hankins, 2013;Wintoki, et al., 2012). us, studies o en use traditional IV approach to obtain consistent estimates.
Unfortunately, it tends to be infeasible to nd a set of external instrumental variables when almost all independent variables are o en considered not exogenous. To rectify this issue, the current study uses the two-step system generalised method of moments estimator (System GMM) developed by Blundell and Bond (1998). is estimator is superior to the OLS or FE as it controls for time-invariant unobserved heterogeneity across rms, simultaneity, and dynamic endogeneity (Blundell & Bond, 1998;Wintoki et al., 2012).

Empirical results and discussions
is section provides the results of the empirical analyses for the role of corruption on rm capital structure by using the dynamic two-step GMM approaches.    Following Schultz et al., (2010) and Wintoki et al., (2014), rm age and year dummies are considered to be exogenous.
First, regarding the main variable of interest, columns 1, 2 and 3 of Table 3 show that corruption impacts insigni cantly on the rm capital structure. e nding supports the perspectives of institutional theory, and this may be explained by the fact that corruption is very popular in Vietnam and hence it is considered as an entry payment for every rm which must pay to participate in the market or compete with neighbors for survival. When a rm pays a bribe, neighbor rms also pay bribe, thus, there are no statistically signi cant di erences in the e ect of corruption on rms and their neighbors.
is nding is partly consistent with Vu et al. (2016), who also indicate that there is an insigni cant relation between corruption and rm performance.
Looking more closely, I explore the e ects of corruption on di erent types of rms' capital structure. Interestingly, the e ect of corruption is di erent for various types of capital structure. e results from columns 4, 5 and 6 of Table 3 show that paying bribe helps rms gain more external credit. is stems from the fact that although the rapid development of Vietnam's capital market has o ered more options for corporate funding, capital market remains relatively underdeveloped and banks remain the major capital provider for corporations and occupy a monopolistic position in the credit market ( see Table 1). Although the state Bank decreases the banks' monopoly pro t by controlling interest rates, bank o cers have various methods of evading those controls and taking rent from corporations by virtue of their monopolistic position, obtaining illegal pro ts in the process. To increase the likelihood of receiving credit, corporations and rms are willing to bribe such bank o cers. In the meantime, regulators are strict about banks' control of their non-performing loan rate, which means that banks must control the risk of not regaining their capital. us, banks would prefer to grant more short-term credit, especially in regions in which corruption is very serious. As a result, corruption impacts positively on short-term debt of rms. e ndings support some perspectives from previous studies (e.g., Diamond, 1991;Jiang & Li, 2005). Such studies also reveal that short-term credits are favorable for banks to obtain timely and constant information about debtors, thus placing corporations under the banks' close supervision and control in regions with high corruption level.
While bribery helps to boost short-term bank debts, it hinders long-term bank debts. As shown by columns 7, 8 and 9 of Table 3, rms paying bribe have lower access to long-term bank credit than those without paying bribe. is can be interpreted in the way that banks are more hesitant to o er long-term loans if they are in a very corrupt environment. Long-term bank loans are less prevalent and more strictly controlled inside banks compared with short-term bank loans. e choice to grant such loans may depend more on the legal framework through the protection of creditors and the enforcement of loan contracts. Banks may not receive the capital back or they may have to add extra expenses to ensure their business safety in countries with a poor law-enforcement system. us, corruption has a negative in uence on rms' ability to obtain long-term bank credit in this context. In terms of rm-level characteristics, as expected, rm size has a positive impact on rm capital structure performance. For example, column 2 of Table 3 shows 1 percent increase in sizes of rm coupled with 0.03 percent increase in accessing external nance, with other things constant. e results are consistent with most ndings in the literature. is nding provides the same perspectives with trade-o theory which reported the positive relationship between company size and level of debt nancing. Accord-ing to Deesomsak, Paudyal, and Pesce o (2004), large companies have small volatility of cash ow and more transparency in nancial activities that help them access credit more easily. In addition, export activities have an insigni cant impact on rm capital structure, while rm age as measured by the number of years in business witnesses a negative association with rms' capital structure. e estimated results in the columns of Table 3 reveal that the role of provincial quality of institution on rm capital structure is di erent depending on the model speci cation. For example, time cost has a negative and signi cant impact on rm capital structure. In addition, while actions supporting private sectors (Private act) do not help rms to access external nance because of complicated administrative procedure, transparency in nance and information helps rms access bank's credit. Surprisingly, labour training has no impact on rm capital structure, and this may be explained by the outdated education in Vietnam where theory is heavily focused on, while practical skills are hardly dealt with, quality of instructors is at worrying low levels, and teaching equipment is not updated.

Conclusion
e paper aims to estimate the e ect of corruption on Vietnamese SMEs' capital structure from 2005 to 2011. Unlike previous studies, this study considers for the rst time the impact of corruption at both rm and provincial levels on rms' capital structure in Vietnam. Some main interesting ndings are presented below.
Some provincial institutional factors such as time costs for private sector are negatively associated with rms' capital structure; while transparency in nance and information improves the rms' access to bank's credit. is implies that simpli cation of procedure to save time for enterprises, and increasing actions to make information and procedures clear are necessary to improve access to external nancial sources. e nding of a negative linkage between corruption and the improvement in rms' long-term nancial access implies that measures against corruption are necessary for the development of SMEs. Anti-corruption campaign has been implemented for many years, but its results are limited. is suggests that political e orts and willingness from central government, particularly from the Communist Party, is crucial for the success of anti-corruption. A transparent legal framework and e ective enforcement should be encouraged and e ectively implemented to control corruption in Vietnam.
ere is some limitation in the current study. e study used data from manufacturing SMEs, so its ndings might not represent whole enterprise. Especially, the ndings might not be true for large enterprises which own di erent resources and business behaviours including markets and negotiating powers. is suggests that further research on larger rms and other sectors beyond manufacturing should be done to make a general conclusion about the relationship between corruption and rms' capital structure in Vietnam. Finally, future research should be conducted in other transitional economies using the same methodology as in the current study to examine whether a negative association between corruption and rm capital structure is found to be consistent beyond Vietnam.