DOES CORRUPTION IMPACT THE PROBABILITY FIRMS APPLY FOR LOANS?

AN EMPIRICAL RESEARCH ON DEVELOPING COUNTRIES IN

EAST ASIA & PACIFIC AND SOUTH ASIA

Current version: December 2018

Jean-Christophe Statnik

jcstatnik@univ-lille2.fr

Thi-Le-Giang Vu 1

thilegiang.vu.etu@univ-lille.fr

Abstract: Corruption is a big problem for development and gets the attention of many scholars. Nevertheless, the impact of corruption on the ability firms apply for loans for developing countries is still a question. Using Enterprise Survey from World Bank and other data from Transparency International and The World Economic Forum, we find corruption increases the ability firms apply for loans in developing countries. Moreover, the effect of corruption depends on the level of country development. In upper-middle income countries, corruption discourages firms from applying loans, while in lower-middle income countries, corruption encourages firms to apply loans. We also find burden of government regulation is a channel to explain for this effect. To get over burdens in regulation, firms find corruption as one way to reach their targets at lower costs.

1 Corresponding author

1. INTRODUCTION

Capital plays a vital role in the development of enterprises. Therefore, access to finance is a

topic getting the attention of not only entrepreneurs but also many researchers in the world.

Most of researches about this topic focus on applicants to find constraints hampering them

from getting loans (Cavalluzzo, Cavalluzzo, & Wolken, 2002; Cole, 1998, 2008; Petersen &

Rajan, 1994; Stiglitz & Weiss, 1981). Although non-applicants account for a large proportion,

researches about them are still less. Studies about this field pay more attention to developed

countries while developing countries receive much less attention. Therefore, we choose to

study about accessing to finance for non-applicants in developing countries to field this gap.

More detail, we concentrate our study to corruption and find how corruption impact the non-

applicants and discourage them from applying loans.

Prior researches found corruption may have both positive and negative impacts on applicants

for bank credit. As a negative effect, high corruption reduces the amount of loans which banks

grant to borrowers (Anaere, 2014; Weill, 2011)

and mitigate firm

Weill, 2015; Weill, 2011). For non-applicants, as our best knowledge, only Galli (2017) find

corruption constrains small and medium enterprises in developed countries from applying for

loans. From the findings in literature, especially the results of Galli (2017), we question about

the effect of corruption on non-applicants in developing countries. Is this effect similar or

different from developed countries?

To exploit this question, we use the Enterprise Surveys (ES) conducted by World Bank from

2012 to 2016 for upper-middle income countries and lower-middle income countries2 in (1)

East Asia & Pacific and (2) South Asia. The survey collected objective data based on the

experience and the perception of enterprises about their operation environment. The data

includes 14 countries Cambodia, China, Indonesia, Lao, Malaysia, Myanmar, Papua New

Guinea, Philippines, Solomon Island, Timor-Leste, Thailand, Vietnam, India and Pakistan. Of

these, China, Malaysia and Thailand are upper-middle income countries and the rest are lower-

middle income countries.

Classification is based on the criteria of World Bank. Before July 01, 2017, upper-middle income countries will have GNI/capita range from USD4036 12475; lower-middle income countries will have GNI/capita range from USD1026 4035 and low-income countries are smaller than USD1026. Since July 01, 2017, new threshold for upper-middle income countries is USD 3956-12235; lower-middle income countries are USD 1006-3955 and low-income countries are less than USD1006.

Our research is important in the following issues: First, in access to finance field, compare to

researches about applicants who applied for finance, study about non-applicants is a new field

of research. The importance of this field is gradually confirmed when recent researches find

the number of discouraged borrowers3 is twice or even three times as the number of borrowers

who are rejected from applying for loans (Freel, Carter, Tagg, & Mason, 2012; Gama, Duarte,

. Besides, most of predecessor researches about

discouraged borrowers are conducted in developed areas or developed countries such as Europe,

United Kingdom, United Stated, Australia and Canada. The percentage of discouraged

borrowers compare to total loan seekers4 in these countries range from 8% to 23%. Cross

country researches, especially researches for developing countries, are rare. As our best

knowledge, there are three studies of Chakravarty and Xiang (2009), Chakravarty and Xiang

(2013) for ten developing countries around the world and Gama et al. (2017) for less developed

countries in Eastern Europe and Central Asia. As our opinion, researches about discouraged

borrowers are especially important for developing countries because in these countries, funds

to finance for development are vital while the percentage of discouraged borrowers are very

high, but studies to understand them are less. Contributing to research of discouraged borrowers

in developing countries, our study focuses on upper-middle income countries and lower-middle

income countries in (1) East Asia & Pacific and (2) South Asia. These two regions are the

fastest growing regions in the world. Therefore, capital for development is must be very large.

However, the incidence of discouraged borrowers in these regions is very high around 67% of

loan seekers. Finding determinants constraining firms from applying for loans in these regions

may have significant meanings for authorities in reducing the rate of discouraged borrowers

and contributes to the literature understandings about non-applicants in developing countries.

Second, there are abundant researches about the threats of corruption to economy and society.

Nevertheless, studies investigating the impact of corruption on discouraged borrowers are rare.

Studies of Chakravarty and Xiang (2009), Martin Brown (2011) and Chakravarty and Xiang

(2013) mention about this effect. However, they find a significant impact of corruption

on discouraged borrowers. Only Galli (2017) concludes that small and medium enterprises in

European are more likely refrain from applying for bank loans if corruption is high. In our

study, we investigate developing countries and we find the effect of corruption on discouraged

borrowers is significant. Moreover, we prove that the impacts of corruption differ according to

Discouraged borrowers are non-applicants who need capital but apply for loans because of fearing to be rejected. 4 Loan seekers are enterprises who need credit. They are both borrowers who applied for credit and discouraged borrowers who need credit but apply for credit.

level of country development. In detail, in higher development countries, higher level of

corruption will discourage firms from applying for loans. However, in lower development

countries, if level of corruption increases, firms are more willing to apply for credit. This is a

new point contributing to the literature about the impact of corruption on discouraged

borrowers. We explain for this result by the burden of regulation. More precisely, the burden

of regulation is a channel for the negative effect of corruption on discouraged borrowers. More

burden of regulation will push the firms into using corruption as one way to get over these

burdensome.

In the next section, we review researches related to discouragement. The third part proposes

hypothesis. The fourth part mentions about data and methodology. The final parts show the

results and conclusion.

2. LITERATURE REVIEW

2.1 Discouraged borrowers and elements affect discouragement

Although empirical researches about discouraged borrowers was conducted since 1990s by

Jappelli (1990) for households in United State, by Mayank Raturi (1999) in Zimbabwe and by

Levenson and Willard (2000) in United State, the first theory was built later by Kon and Storey

(2003). In the research, Kon and Storey (2003) define discouraged borrowers are good

borrowers apply for loans because they think they will be rejected. They take into

account the application costs and those prevent them from applying for loans. Application costs

can be financial costs such as costs to gather information required by banks, costs for time to

travel to banks, time to wait for application process, time to meet bank officers. Application

costs can also be psychic costs such as reluctance to provide information to outsiders, lack of

knowledge and skills to complete application documents, reluctance to enter the bank.

Kon and Storey (2003) explain that asymmetric information is a reason for discouragement. If

banks have clear information about borrowers, their mistakes in deciding for financing will be

reduced and when borrowers know that banks make good decisions, their applications will

increase. Therefore, discouraged borrowers decrease when there is less asymmetric

information. Kon and Storey (2003) conclude three elements impacting the scale of

discouragement are screening error of the banks, application costs and the different between

interest rate of bank and interest rate of money lender. Reducing screening error of the banks

will mitigate the discouragement. While discouragement will increase if application costs are

high and sources of funding are less.

Following the theory of Kon and Storey (2003), empirical researches find the rate of collateral,

relationship lending, entrepreneur characteristics, firm characteristics and distance are factors

impacting on discouraged borrowers. These researches prove that the above elements impact

asymmetric information, hence affect the incidence of discouraged borrowers. Besides,

literature also find perceptions about operating environment and economic conditions are

significant effects on discouragement.

Rate of collateral

Rate of collateral impacts not only applicants who apply for loans but also non-applicants who

may be discouraged from applying. For those non-applicants, they are good enough to apply

nk. Therefore, they are

ateral does increase

the incidence of discouraged borrowers (Kon & Storey, 2003).

Lending relationship

Lending relationship play an important role in reducing asymmetric information. Close

relationship between lenders and borrowers can mitigate credit restriction (Berger & Udell,

Rajan, 1994). Researches about discouragement find two opposite impacts of lending

relationship on discouraged borrowers. Chakravarty and Xiang (2009), Freel et al. (2012) and

Chakravarty and Xiang (2013) find stronger relationship between a lender and a firm can

reduce the likelihood of firm being discouraged. This conclusion is also get in the research of

Chakravarty and Yilmazer (2009) when they find that if firms have relationship with more

credit sources, the probability of applying for loans will increase. However, Han, Fraser, and

Storey (2009) conclude that risky borrowers are more likely be discouraged when they make

longer relationship with banks. In the line with Han et al. (2009), Vincent (2010) also confirms

discouraged borrowers have stronger relationships with financial institution. During the time

building relationship, they understand both financial institution and themselves. Therefore,

they know their ability to success applying for loans and that discourages them from applying

for credit.

Entrepreneur characteristics

Next factor influencing discouraged borrowers is entrepreneur characteristics. Characteristics

of managers have significant impacts on the development and prospect of firms. These

characteristics also impact on the tendency to apply for credit. Entrepreneur characteristics

include gender, educational level, professional experience, credit quality and wealth.

Chakravarty and Xiang (2009) and Gama et al. (2017) suggested that if top manager is male,

that firm will be less self-rationing. Explain for this reason, researches of Treichel and Scott

2006; Watson 2006; Coleman 2000 cited in Freel et al. (2012) find that women are less self-

confidence, more risk averse and want to keep control of their business. They have the feeling

that applying for loans will create more risks. Therefore, they are less likely apply for credit

than men.

About the educational level of owner, most of research found the positive relationship between

educational level and credit availability (Parker and Van Praag 2006, Krasniqi 2010 cited in

Bellier, Sayeh, and Serve (2012)). Chakravarty and Xiang (2009) conduct research for

developing countries and report that if top managers of the firm have high education (university

or higher level), that firm will less discourage themselve from applying for loan. In the line

with Chakravarty and Xiang (2009), Fraser (2014) applies both quantitative and qualitative

methods on SMEs in UK and confirms that owners of discouraged firms are lower levels of

education.

In addition, owners of discouraged firms are also less wealthy and less professional experience

(Fraser, 2014). Fastenbauer and Robson (2014) implement an qualitative research in UK and

Austria and they conclude that experience of owners impact on the probability to apply for loan.

Experience owners understand the development of their firm. They also can evaluate more

accurate the benefit they get if they apply for loans. Hence, they are less self-discouraged.

Recent research of Gama et al. (2017) confirms the effect of owner professional experience on

discouragement.

Firm characteristics

It is widely agreed that firm size impact on discouragement and small firms are less likely apply

for credit (Cole, 2008; Ferrando & Mulier, 2015; Fraser, 2014; Freel et al., 2012; Han et al.,

2009; Mac an Bhaird, Vidal, & Lucey, 2016; Xiang, Worthington, & Higgs, 2015). This is

reasonable because small firms usually have opaque information. Therefore, they are hard to

prove their creditworthiness with lenders. In addition, small firms usually face with high

probability of application rejection (Holton, Lawless, & McCann, 2014) and the feeling that

they will be turned down again discourages them from applying for new loans (Fraser, 2014;

Xiang et al., 2015).

Besides firm size, researchers find other characteristics affecting to discouraged borrowers.

Cole (2008) stated in the results that discouraged firms are younger than other firms in the

sample. These firms also have worse credit quality and more likely locate in urban areas. In the

line with Cole (2008), Ferrando and Mulier (2015) studies 9 countries in Europe and confirms

that discouraged borrowers are younger and riskier than applying firms. In addition, if trade

credit is more available for firms, they will be likely to discourage themselves. Trade credit is

another alternative source of fund with lower cost. Hence if trade credit is available, firms have

tendency to use more trade credit than applying for loans. In contrast, Biais and Gollier (1997)

and Voordeckers and Steijvers (2006) proved that trade credit can work as a signaling

instrument to reduce asymmetric information. When firms use trade credit, their suppliers will

get private information from them and the private information can convey to banks. Based on

this argument, Gama et al. (2017) find trade credit has negative impact on discouragement.

Mention other firms characteristics, research of Han et al. (2009) in US small business find

firms using external finance such as business credit cards or business credit lines seem less to

be discouraged because they have signal that they are creditworthiness. About the ownership

structure, Martin Brown (2011) and Xiang et al. (2015) state that foreign ownership increases

discouragement because these firms reply more on internal funds from parent firms and less

apply for external finance sources. Martin Brown (2011) also find government-owned firms

are less seeking for credit. Besides, exporting firms are more likely apply for credit because

they need capital to finance for their operation.

Firm perception about the operating environment

Firm perception about the operating environment has a significant impact on the probability

firms apply for bank loans. These perceptions can be firm perception about access to finance

or level of corruption in the economy. Mac an Bhaird et al. (2016) find firms will discourage

themselves from applying for loans if they feel access to finance is a severe problem. Impact

of corruption on discouragement will be discussed in detail in the following part.

Distance between lenders and borrowers

Closer distance from a bank office to a firm can make it easier for that firm to get financing

(Agarwal & Hauswald, 2010) because bank can collect soft information and get benefit from

private information. Kon and Storey (2003) also explain the distance between lenders and

borrowers can reduce application costs for bank and helps lenders make better decisions.

Hence, the incidence of discouraged borrowers decreases. Confirm theory of Kon and Storey

(2003), Gama et al. (2017)

lenders and they find firms located in the big city are less likely become discouraged borrowers.

Economic conditions and legal environment

Chakravarty and Xiang (2009); Chakravarty and Xiang (2013) use the average GDP growth

rate during research period to proxy for country characteristic. They conclude that there is a

negative correlate between growth rate and discouragement. Higher growth rate will reduce the

discouragement. Gama et al. (2017) find quite similar result that firms are less probability of

being discouraged in countries which GDP per capita is high. Because in these countries, firms

face with lower financing obstacle -Kunt, Laeven, & Maksimovic, 2006).

About the legal environment, -Kunt and Maksimovic (1998) conclude that the

incidence of firms financing their growth by external sources is higher in countries which legal

systems is efficient. This result implies that in inefficiency legal system countries, firms will

use more internal sources to finance for their growth and limit their applications for external

sources. Besides, - r-Kant (2011) find bank loan maturity is 5. Firms are more likely get long-term

In addition,

in countries which have better institutional environments, banks are more willing to provide

long-term debts to SMEs. Findings of - -Kant (2011) are

important to authorities because their policies to improve legal and institutional environment

have direct impact on In short, findings from studies of

-Kunt and Maksimovic (1998); - -Kant (2011) imply

that the efficiency of legal environment has a positive effect on the ability firms apply for loans.

In the line with previous studies, Gama et al. (2017) also expect the strength of the legal

protections of creditors and borrowers and the number of discouraged borrowers have negative

correlation. However, the results they find is opposite with their expectation. Increasing legal

protections will increase the number of discouraged borrowers. To explain for this result, they

find in their research the strength of legal protection has positive relation with collateral.

Stronger legal protections increase the collateral requirements and hence increase the incidence

of discouraged borrowers. Mac an Bhaird et al. (2016) also find greater regulatory quality will

increase the number of discouraged borrowers and they

- and -Kant (2011) use economic freedom index as a proxy for institutional environment. The index includes 10 institutional factors: trade policy, fiscal burden of government, government intervention in the economy, monetary policy, capital flows and foreign investment, banking and finance, wages and prices, property rights, regulation, and black market activity. It is measured by the Heritage Economic Freedom Index.

2.2 Corruption and discouragement

Corruption and its measures

Corruption is a complicated issue and it still exists arguments about corruption definitions.

However, there is a consensus in the literature that corruption is the misuse of public office for

private gains (Nye, 1967; Philp, 2016; Treisman, 2000). Based on the measurement of

corruption from internal, external or hybrid. Asiedu and Freeman (2009) classify three ways to

capture corruption. The first way measures firms perception/experiences about corruption in

the environment in which they operate. This way captures internal information. It reflects the

perception/experiences about level of corruption in a country. However, the accurateness

depends on the way to collect samples and answers of firms because corruption is a sensitive

issue. In addition, each country has its own culture, economic and political environment so firm

perception about corruption in each country is different. The same answer severe corruption

mean that two countries have the same level of corruption. Hence,

comparing level of corruption across countries may not be easily. The second way is the

assessment of analysts who are outside a country. This way overcomes the shortcomings of the

first method that there is the same criterion to assess countries, so the result can compare across

countries. Nevertheless, the assessment is usually based on media reports, so it may not

accurately reflect the country level of corruption. The third way is a combination of many

sources about corruption into an index, for example Corruption Perceptions Index. Therefore,

this measurement mitigates the shortcomings of both above methods.

Using a different way of classification, Gutmann, Padovano, and Voigt (2015) analyze two

proxies for corruption that are perception versus experience with bribery. They conclude

that two proxies may not highly correlated because respondent characteristics and country

.

Corruption and discouragement

Theories and empirical researches have proved vigorous evidences that corruption harms

economic development. Nevertheless, at another side, literature also find that in case of

cumbersome regulations, corruption can improve the efficiency and help growth (Huntington,

1968; Leff, 1964; Leys, 1965). This efficiency is shown by corruption can help things to be

done faster (Leys, 1965; Lui, 1985) or like competitive bidding, corruption can allocate

contracts to lowest cost firms (P. J. Beck & Maher, 1986; Lien, 1986).

In access to finance field, corruption is also found having both positive and negative impacts.

In the positive side, find higher level of corruption can increase bank

debt ratio. Chen et al. (2013)

decisions to grant loans are not affected by bribery, but bribery enables better economic

performance firms to receive larger amount of loans. In the negative side, Anaere (2014) shows

that the banks less likely grant loan to firms and households in private sector if the level of

corruption in that country is high and Weill (2011) concludes corruption constrains bank credit

in both developed and developing countries. However, most of researches in access to finance

field focus on applicants, while researches about impacts of corruption on discouraged

borrowers are rare.

Researches about discouraged borrowers of Chakravarty and Xiang (2009), Martin Brown

(2011) and Chakravarty and Xiang (2013) added corruption as an independent variable.

Nevertheless, they find no statistical evidence that corruption impact on discouragement.

Recently, Galli (2017) conducts the research on micro, small, and medium-sized enterprises in

Europe. They find that high level of corruption impedes firms from applying for bank loans

and behaviors of small firms are more effected by corruption than larger firms. In addition, the

percentage of small firms refraining from applying for credit is higher in countries which have

high level of corruption. To proxy for corruption, Galli (2017) use perception corruption at

country level such as Corruption Perceptions Index launched by Transparency International

and Control of Corruption from Worldwide Governance Indicators published by World Bank.

3. HYPOTHESIS DEVELOPMENT

The study about the impact of corruption on discouraged borrowers of Galli (2017) concludes

that high corruption increase the probability of discouragement in small firms and this research

investigates developed countries where corruption is under the control and low. Our research

is conducted in developing countries where corruption is quite high. The average of Corruption

Perceptions Index in the study of Galli (2017) is 68, while in our data is 366. This value

confirms the pervasion of corruption in developing countries. Axelrod (1986) stated that a

behavior will become a norm when it spreads in a population. In the context of corruption,

when corruption is pervasive in the society, it will become a normal behavior. The feelings of

guilt and the loss of reputation will decrease as the number of persons violating the rules

increase (Andvig, 1991). In addition, the ability to be detected and punished is more difficult

Corruption Perceptions Index is scaled from 0 to 100 with 0 is very severe perception of level of corruption and 100 is the lowest perception of corruption

because of the compromise of the community for this action. In that case, Della Porta (2017)

confirms that

Conducting the research for developing countries and based on these above arguments, we

expect corruption may reduce the discouraged borrowers in developing countries, because

corruption can reduce application cost and increase the chance firms receive the loans.

Hypothesis 1: Corruption has a negative impact to discourage borrowers.

Besides concluding that corruption has a positive impact on discouraged borrowers, further

research about the impact of corruption, Galli (2017) divide the data into two parts based on

the level of corruption and author finds firms is more self-discouraged in high corruption

countries than in low corruption countries. Moreover, according to Treisman (2000) and

Pellegrini (2011), there is a correlation between the level of country development and

corruption. More develop countries have a lower level of corruption. The above results suggest

two ideas which lead to the second hypothesis: (1) it may exist different impacts of corruption

on discouraged borrowers in different groups of countries depending on their level of

development and (2) the same result Galli (2017) got on developed countries, we expect the

effect of corruption on discouraged borrowers may get positive sign for countries which are

higher level of development.

Hypothesis 2: The effect of corruption on discouraged borrowers depends on the level of

country development and in higher development countries, corruption hamper firms from

applying loans.

4. DATA AND METHODOLOGY

4.1 Data

We use The Enterprises Surveys (ES) conducted by World Bank. This is a project of World

Bank to gather data based on experiences and perception of the operating

environment. The project has collected data from about 155,000 firms in 148 countries.

Standard methodology to collect information has been applied in 139 countries to better

compare different countries and different time. The survey use stratified random sample

method to collect data. Follow this method, population is divided into homogeneous sub groups

and simple random samples are selected from each group with sample sizes proportional to

strata sizes. This method ensures representation of samples across the population. Firm size,

business sector and geographic region are strata of the survey.

The ES is face to face interview with business owners and top managers. Private contractors

have been hired by World Bank to conduct the surveys. These contractors Government

agencies or organizations associated with Government because of sensitive questions related

to Government. To collect the data, at first interviewers use a screener questionnaire over the

phone to check if enterprises are eligible and make appointment. After that, interview is

conducted by face to face with the owner or managers of each enterprise.

The target regions of our research are (1) East Asia and Pacific and (2) South Asia. We choose

these areas because they are two fastest growing regions in the world so capital for development

is vital. Discouragement will be an important issue if it happens in these regions. The research

period is from 2012 to 2016. During this 5-year period, the economy have much

fluctuation, so there is less bias between countries surveyed in 2012 and countries surveyed in

2016. In addition, there is just a slightly difference between the survey version in 2012-2013

and the version in 2014-2016. Therefore, we can collect as much common information as we

need to build the research model. Based on these criteria, we get 17 countries classified by

upper-middle income countries, lower-middle income countries and low-income countries.

Rioja and Valev (2004) find financial development and growth have a positive relationship in

middle region and high region (countries with middle and high levels of financial

development). This effect is even stronger in middle region and insignificant in low region.

The low or insignificant impact of financial development on growth in low income countries

is also found in other research of Deidda and Fattouh (2002), Rousseau and Wachtel (2002).

Based on the above researches, we could expect that discouragement phenomenon is an

important constraint for the growth of middle-income countries. Therefore, our target focuses

on middle income countries which include upper-middle income countries and lower-middle

income countries.

Finally, we got 14 countries in both regions. They are Cambodia, China, Indonesia, Laos,

Malaysia, Myanmar, Papua New Guinea, Philippines, Solomon Island, Timor-Leste, Thailand,

Vietnam, India and Pakistan. Of these, China, Malaysia and Thailand are upper-middle income

countries and the rests are lower-middle income countries. To ensure that the results are

reliable, we take only responses rated by interviewers are truthful and somewhat truthful. If

using only truthful responses, we will lose almost half of observations. Therefore, truthful

responses are used for the robustness check to confirm our main findings.

In addition, we also get data from Transparency International and World Economic Forum for

country-level variables. From Transparency International, we extract Corruption Perceptions

Index. This index measures level of corruption in public sector through perceptions of business

people and country experts. The index is averaged from different sources in a country. Selected

sources must meet certain criteria to ensure the reliability, the comparableness and the

development of the index in the future. At least three sources are needed to calculate Corruption

Perceptions Index for one country. This index is scaled from 0 to 100 with 0 is very severe

perception of level of corruption and 100 is the lowest perception of corruption.

Besides, we use data from Global Competitive Index (GCI) issued by World Economic Forum.

The GCI includes the set of institutions, policies, and factors collected from internationally

recognized agencies to determine the level of productivity of a country. From the GCI, we

extract burden of government regulation This indicator measures the level of

burden for business to comply with governmental administrative requirements (e.g. permits,

regulations, reporting). The indicator is collected from answers of leading business executives

and ranks from 1 to 7 where 1 is extremely burdensome and 7 is not burdensome at all.

4.2 Methodology

In order to test our first hypothesis dealing with the impact of corruption on discouraged

borrowers we use the following equation.

DISCOURAGED BORROWERSi,k = 1*CORRUPTIONi,k + 2*LENDING RELATIONSHIPi,k +

3*ENTREPRENEUR CHARACTERISTICSi,k + 4*FIRM CHARACTERISTICSi,k + 5*PERCEPTION

OPERATING ENVIRONMENTi,k + 6*DISTANCEi,k + 1*COUNTRYk 2*INDUSTRYn 3*YEARt i,k

(1)

Dependent variable

Discouraged borrowers is observed by the answers of firms to the question What was the main

reason why this establishment did not apply for any line of credit or loan? . If answer

for the question is No need for a loan - establishment had sufficient capital , they are not

discouraged borrowers. The firms will be considered as discouraged borrowers if their reasons

for not applying are application procedures were complex, interest rates were not favorable,

collateral requirements were too high, size of loan and maturity were insufficient, or they did

think the application would be approved.

Dependent variable got value 1 if firm i in country k is considered discourage borrowers based

on its answers and got value 0 if firm applies for loans. This measure of discouragement is

applied by many previous researches (Chakravarty & Xiang, 2009, 2013; Galli, 2017; Gama et

al., 2017).

Independent variables

CORRUPTION

We use some proxies for corruption. The first proxy is at firm level and based on the survey.

Firms are required to choose the degree in which corruption is an obstacle to their operation.

The levels of corruption range from 0 to 4

We define corruption as an

corruption is major obstacle or very severe obstacle and 0 for no obstacle, minor obstacle or

moderate obstacle. In the Enterprise Survey, perception about corruption is the most general

question capturing the level of corruption in a country and this question also got the highest

rate of answers. Therefore, choosing it as a main variable, we get the biggest number of

observations. In addition, perception corruption is also widely used in researches about

corruption (Mo, 2001; Podobnik, Shao, Njavro, Ivanov, & Stanley, 2008; Treisman, 2000;

Weill, 2011).

Nevertheless, this proxy may have problems when compare across countries. The tolerant

towards corruption can be various in different countries (Cameron, Chaudhuri, Erkal, &

Gangadharan, 2005), so two countries having the same value for perception corruption may

have different levels of corruption. To overcome this disadvantage, we use another proxy in

the survey at firm level to robustness for the main result. This variable is calculated by the

number of times firms make gifts or informal payments to public officials divided by the

number of times they apply or use public services. Doing this way, we measure experience

about corruption instead of perception corruption, so this variable look corruption at another

edge. Besides, we also use the Corruption Perceptions Index issued by Transparency

International as one proxy for corruption at country level.

Other variables in the model are control variable. We control for lending relationship,

entrepreneur characteristics, firm characteristics, distance, firm perception about the operating

environment and economic conditions.

LENDING RELATIONSHIP: One way to measure lender borrower relationship is scope .

If a borrower buys most of it financial products such as checking account, saving

the asymmetric information reduces (Steijvers & Voordeckers, 2009). Norden and Weber

(2010) finds that information on credit line usage and account activity will be used to manage

the relationship between banks and their customers. Therefore, we use financial products which

banks offer borrowers such as checking or saving account, overdraft facility as proxies for

lending relationship. When firms buy these services, the banks can get more precise

information about the firms by observing the changes in checking or saving accounts (Petersen

& Rajan, 1994). This increases the understanding of the banks with the firms. Financial services

are dummy variables got value 1 if firms either have a checking, saving account with banks or

an overdraft facility. Negative impacts of these variables on the incidence of discourage

borrowers is expected.

ENTREPRENEUR CHARACTERISTICS are measured by gender and experience of top

managers.

Gender is a dummy variable received value 1 if top managers are female and 0 for other. If top

managers are female, the probability of discouraged borrowers is expected increase.

Experience of top managers is measured by number of years working in the sector. Experience

of top managers may mitigate discouragement.

FIRM CHARACTERISTICS include age of firm, size of firm, ownership and other

characteristics.

Firm age: as long as firms operate in the market, possible for credit institutions to get

sufficient information from older firms (Berger & Udell, 1995; Kysucky & Norden, 2015;

Petersen & Rajan, 1994). Therefore, asymmetric information reduces, and older firms are

expected less discouragement.

Firm size is measured by the number of employees. There are three sizes: small (firms have 5

to 19 employees), medium (firms have 20 to 99 employees) and large (firms have 100

employees or more). Larger firm is expected less discouragement.

Foreign owned firm is a dummy variable received value 1 if the percentage of firm owned by

foreign individuals, companies, organizations is greater than 50% and 0 for others. As Martin

Brown (2011) and Xiang et al. (2015), we expect foreign ownership increases self-rationing

because these firms reply more on internal funds from parent firms and less apply for external

finance sources.

Direct export is captured by a percentage of sale. Martin Brown (2011) finds exporting firms

are more likely apply for credit because they need capital to finance for their operation. We

expect direct export has negative impact on the probability of discouraged borrowers.

Trade credit is measured by a proportion of total annual purchases of material inputs bought

on credit. Biais and Gollier (1997) and Voordeckers and Steijvers (2006) proved that trade

credit can work as a signaling instrument to reduce asymmetric information so we expect that

trade credit has negative impact on discouragement.

Innovation is a dummy variable received value 1 if firms spend on formal research and

development activities. Although, R&D is highly related to information opacity, this activity

also requires financing. Therefore, a negative impact between innovation and discouragement

is expected (Gama et al., 2017).

Internationally-recognized quality certification is a dummy variable received value 1 if firms

got internationally-recognized quality certification. This variable is expected to have a negative

effect on discouragement because quality certificate can reduce asymmetric information.

Annual financial statements checked and certified by an external auditor is a dummy variable

received value 1 if annual financial statements were checked and certified by an external

auditor. This variable is also expected to have a negative relationship to the probability of

discouraged borrowers.

Credit line is a dummy variable got value 1 if firms have a line of credit or loan from a financial

institution. Petersen and Rajan (1994) measure information transparency of firm by credit line

so we expect firms having credit line will be less self-rationing.

DISTANCE BETWEEN LENDERS AND BORROWERS.

Two proxies for distance are main business city and city with population over one million.

Financial institutions are usually located in business and crowed cities. Hence, firms located in

these areas may close to their lenders. Therefore, lenders would have better information about

them and can reduce screening errors, so discouragement can be reduced.

FIRMS PERCEPTION ABOUT THE OPERATING ENVIRONMENT

Access to finance is a dummy variable got value 1 if firm feels access to finance is a major

obstacle or very severe obstacle to their operation. We expect that firms which feel access to

finance is a problem to their operation are less likely apply for credit.

Besides the above control variables, we also control for country, industry and year as fixed

effects. Model cluster standard errors by country.

For the second hypothesis testing the impact of corruption on discouraged borrowers at

different level of country development, we add in the equation (1) an interaction term between

the corruption variable and GDP per capita.

DISCOURAGED BORROWERSi,k 1*CORRUPTIONi,k*GDPk,t 2*LENDING RELATIONSHIPi,k

3*ENTREPRENEUR CHARACTERISTICSi,k 4*FIRM CHARACTERISTICSi,k 5*PERCEPTION

OPERATING ENVIRONMENTi,k + 6*DISTANCEi,k 1*COUNTRYk 2*INDUSTRYn 3*YEARt i,k

(2)

In addition, to confirm the result, we divide the sample into two groups based on the GDP per

capita. The first group has GDP per capita higher than 5800USD, and this group is also called

upper-middle income countries based on the classification of World Bank. The second group

has GDP per capita lower than 3400USD and belongs to lower-middle income countries group

of World Bank. Using equation (1), we retest the impact of corruption on discouraged

borrowers in two groups.

5. RESULTS

5.1 Univariate analysis

- Discouraged borrowers in East Asia & Pacific and South Asia

In our sample very high (the average is 85.1%).

Reasons for not applying is stated in the Table A2. About 53% of non-applying firms reveal

that they have sufficient capital, so they s and they are not

application procedures were complex, interest rate were not favorable, collateral were too high,

size of loan and maturity were insufficient, or they fear to be rejected. These firms are called

discouraged borrowers.

The percentage of discouraged borrowers accounts for 67% of loan seekers (Table A1). This

rate is extremely high in East Asia & Pacific and South Asia. Compare to Eastern and Western

Europe in the research of Martin Brown (2011), the incidence of discouraged borrowers is 22.6%

or Eastern Europe and Central Asia in the research of Gama et al. (2017), the percentage of

discouraged borrowers is 34%. These numbers confirm the discouraged borrowers in

developing countries is an important issue and should be got the attention.

- Discouraged borrowers versus applicants

Statistics in table A4 show that discouraged borrowers and applicants are significant

differences. Corruption perception is more severe for discouraged borrowers than applicants.

However, applicants pay more gifts/unformal payments to officials than discouraged borrowers.

Applicants also seem have more relationship with banks when they usually use financial

services from banks. They have more women being top managers and their top managers have

longer years of experience. About firm characteristics, in the line with literature review,

applicants are older and bigger. Their export as percentage of sales is higher. They more invest

in innovation and use more trade credit. I applicants got internationally-

recognized quality certifications more than discouraged borrowers and their financial

statements are checked frequently by an external auditor. In addition, applicants have more line

of credit than discouraged borrowers, and they usually located in crowded cities. However,

statistic result reveals that applicants are owned by foreign individuals, companies or

organizations more than discouraged borrowers.

5.2 Findings from the main model

Regarding our equation (1), the column 1 in Table B3 shows that corruption has a negative

impact on discouraged borrowers as expected. The more level of corruption which firms feel,

the more they apply for loans. This result opposites with the finding of Galli (2017). However,

an important difference between our research and Galli (2017) is that we focus on developing

countries while Galli (2017) conducts the research on developed countries. In developing

countries, corruption is more prevalent (Olken & Pande, 2012). Hence, corruption behaviors

are perceived by firms as less serious as it would be. Officials who take part in corruption

activities less feel loss of their reputation when their actions are detected (Andvig, 1991). In

that case, they will more likely break the rule for their own benefit (Akerlof, 1980) and firms

may use corruption as a way to achieve their target

that they will pay higher Della Porta (2017). Therefore, our

result may reasonable that high level of corruption can make firms more likely apply for loans

because corruption increases their probability receiving the loans and reduces the application

costs.

Interestingly, the second hypothesis is also confirmed when the effect of corruption on

discouraged borrowers depends on the level of country development. The column 2 in Table

B3 reveals that when GDP per capita equal to 0, the effect of corruption on discouragement is

negative. However, this negative impact is reducing when GDP per capital increases, and it

turns to a positive impact when GDP per capital are high enough. These results are also stable

when we split the sample into two groups: upper-middle income countries and lower-middle

income countries. The coefficient of the variable orrupt is positive and highly significant in

upper-middle income countries (column 3 Table B3). On the contrary, it is negative and highly

significant in lower-middle income countries (column 4 Table B3).

Further investigating, we calculate the marginal effect of the interaction variable on

discouragement. We find that when GDP per capita of a country is less than about USD2200,

the effect of corruption on discouragement is negative (Table B4). That means increasing level

of corruption will reduce the incidence of discouraged borrowers and this result is significant

at 1% level. The negative effect is still significant at 10% level until the GDP per capita is

around USD2600. In contrast, in higher development countries, the impact of corruption on

discouragement is positive. At GDP per capita higher than about USD6800, the more

corruption is, the more firms discourage themselves from applying for loans and this result is

significant at 10% level (Table B4).

In summary, for developing countries in our study, corruption grease the wheel and increase

the ability firms apply for loans. However, the impact of corruption on discouraged borrowers

depends on the level of country development. In higher development countries, increasing

corruption will increase the probability firms discourage themselves from applying loans. In

contrast, in lower development countries, higher corruption may encourage firms to apply more

for loans.

For other control variables in the model, our findings are confirmed by literature. Firms using

more financial services will less discouragement. This result is also found by Gama et al. (2017).

Like Fastenbauer and Robson (2014), we find experienced top managers are likely more apply

for loans. Besides, firm characteristics proving the creditworthiness such as firm size (size);

firms getting internationally-recognized quality certification (certificate); firms having

financial statements checked and certified by an external auditor (ExterAudit) or firms having

a line of credit from a financial institution (creditline) have significant negative impacts on

discouraged borrowers. These are reasonable because creditworthy firms are less opaque

information and their chances to receive the loans are also higher. Therefore, they less

discourage themselves. These results get consensus by Martin Brown (2011), Chakravarty and

Xiang (2013) et Gama et al. (2017). In the same line with Martin Brown (2011) and Xiang et

al. (2015), we also find foreign ownership (Ffirm) increase discouragement. Finally, firm

characteristic variables related to expense such as export firms (diexport); innovation

(innovation), are negative impacts on discouraged borrowers. These firms need more capital to

finance for their operation, so they are more apply for loans (Gama et al., 2017; Martin Brown,

2011).

5.3 Explanation for the impact of corruption on discouraged borrowers

There is a consensus in literature that corruption is highly correlated with regulation. Poor

enforcement of legal institutions leads to higher corruption (Tonoyan, Strohmeyer, Habib, &

Perlitz, 2010). In case of poor enforcement of law, bureaucrats may create a lot of red tape to

make difficulties for enterprises and force them to pay bribes (Banerjee, 1997). Enterprises

involve in corruption because they want to overcome cumbersome regulations (Shleifer &

Vishny, 1993). Follow Wei and Kaufmann (1999),

benchmark to explain for the impact of corruption on discouraged borrowers. More precisely

we test the following equation:

DISCOURAGED BORROWERSi,k 1*CORRUPTIONi,k*BURDENk,t 2*LENDING

RELATIONSHIPi,k 3*ENTREPRENEUR CHARACTERISTICSi,k 4*FIRM CHARACTERISTICSi,k +

5*PERCEPTION OPERATING ENVIRONMENTi,k + 6*DISTANCEi,k 1*COUNTRYk 2*INDUSTRYn

3*YEARt i,k (3)

The interaction between corruption variable and burden of government regulation reveals that

corruption has a negative impact on discouraged borrowers when government regulation is

extremely burdensome (Table B5). Nevertheless, when government regulation is less

burdensome, the negative effect reduces and may turn to positive effect. These results are very

significant at 1% level. In detail, marginal effects in Table B6 state that if the value of

BURDEN variable is lower than 3.5, the higher corruption reduces discouragement.

However, when this benchmark is higher than 3.9, the higher corruption increases the

probability of discouragement.

The descriptive statistic shows that upper-middle income countries have BURDEN variable

higher than 4, except Thailand (average 3.5) and lower-middle income countries have

BURDEN variable lower than 3.5, except Laos (3.9) and Indonesia (3.8). The above findings,

one more time, confirm the negative effect of corruption on discouraged borrowers in lower-

middle income countries and the positive effect in upper-middle income countries. More

important, these effects are explained by burden of government regulation. In lower-middle

income countries where regulation is less efficient, to overcome burdensome regulations, firms

find corruption as a solution. This way may save their time and give them more chance to reach

their target with lower costs. In addition, the ability to be caught and punished is low for both

of giving and receiving corruption because of the pervasion of corruption in the social.

(Leys, 1965; Lui, 1985). In contrast, in upper-

middle income countries, the enforcement of regulations is more efficient, high corruption

discourages firms from applying for loans. This result is supported by

and Galli (2017).

5.4 Robustness check

- Applying other measurements for corruption

To robustness check for our model, we apply other ways to measure corruption. First, we

capture corruption by the number of times firms give gifts or pay informal payments to officials

against the number of times they apply or use public services. The higher this rate is, the more

corruption firms suffer. Second, we use Corruption Perceptions Index (CPI) published by

Transparency International. This index measures perception corruption in public sector through

experts, business people and uses a scale of 0 to 100. In contrast with the above measurements,

higher value of CPI means lower level of corruption and CPI capture perception corruption at

country level.

Columns (1) and (4) in Table B7 reveal that for the full data, corruption have negative effects

on discouragement. These results confirm our main finding that in developing countries,

corruption has a negative impact on discouraged borrowers. Testing the effect of corruption on

different groups of country, we get robustness that higher level of corruption reduces the

number of discouraged borrowers in lower-middle income countries (Columns (3), (6) in

Table7). For upper-middle income countries, only Corruption Perceptions Index confirm that

corruption hampers firms from applying for loans (Column (5) Table B7). In summarize,

corruption may reduce the application costs, so it alleviates discouraged borrowers.

Nevertheless, this impact is more significant in countries which their level of development is

low. For higher development countries, the effect slightly turns to positive and firms less apply

for loans if they feel corruption is high.

- Applying another measurement for discouraged borrowers

In further, we measure dependent variable using a stricter method. Kon and Storey (2003)

d

they will be rejected. Applying this definition, we capture discouraged borrowers by: firms

have a line of credit from a financial institution and for loans not because they

have sufficient capital, but because application procedures were complex, interest rates were

not favorable, collateral requirements were too high, size of loan and maturity were insufficient,

would be approved.

When a firm has a line of credit from a financial institution, it could prove its credit worthiness

with that financial institution. Petersen and Rajan (1994) also measured information

transparency of firm by credit line. Therefore, we classify firms having line of credits as good

borrowers and use the measurement of discouraged borrowers as the above discussion. This

way of measure captures the full definition of Kon and Storey (2003). Dependent variable got

value 1 if firms are considered discourage borrowers and got value 0 if firms apply for loans.

Result in column (1) Table B8 reveals that corruption have a negative effect on discouraged

borrowers. Although this result is insignificant, it is in the line with the main result. Hypothesis

2 is confirmed when the results of other columns in Table B8 show the effect of corruption on

discouraged borrowers depends on the level of countries development. In countries which GDP

per capita are lower than 2600USD, high corruption decreases the level of discouragement,

while in countries which GDP per capita are higher than 4000USD, high corruption increases

the probability of discouragement (Table B9). The negative effect in lower-middle income

countries and the positive impact in upper-middle income countries are still robustness at 1%

level of significant when we split the sample. These results confirm our main findings.

- Robustness check by using only truthful responses in the survey

To ensure the reliability of the results, we use only truthful responses in the survey to re-test

our main findings. We lost 2847 observations in the model. However, the results we get are the

same with the main results and all hypothesis are confirmed. In developing countries,

corruption has negative impact on discouraged borrowers and the impact of corruption depends

on the level of country development. All results are shown in Table B10.

- Robustness check by excluding China and India from the sample

China belongs to group of upper-middle income countries and India belongs to group of lower-

middle income countries. They are two big countries comparing to the rests and the number of

observations in these two countries dominate the sample. Therefore, to avoid the bias, we retest

our work by excluding China and India out of the sample. We again find high corruption

prevents applying for loans in upper-middle income countries but encourages applying for

loans in lower-middle income countries. Our findings are robustness when we interact

corruption with GDP per capita and when we split the sample (Table B11). These results

.

6. CONCLUSIONS

Using Enterprise Survey conducted by World Bank in the period from 2012 to 2016 for upper-

middle income countries and lower-middle income countries in East Asia & Pacific and South

Asia, we find that in developing countries, firms less discourage themselves when they feel

high corruption. Moreover, the effect of corruption on discouragement depending on the level

. High corruption can encourage firms in lower-middle income

countries while discourage firms in upper-middle income countries from applying for loans.

For upper-middle income countries, our finding is in the same line with research of Galli (2017).

For lower-middle income countries, although the negative effect of corruption on

in the literature. In this case, corruption can work as a lubricant to grease the wheel and mitigate

obstacles in accessing to finance . We explain this

result by the fact that burden of government regulation is one of explanations for the different

impacts of corruption on discouragement.

Our findings are robust when we measure corruption by different ways both at firm level and

country level. These results still hold when we use stricter way to capture discouraged

borrowers and we also prove that our findings

dominate the research sample.

Finally, although we find higher corruption can encourage firms in lower-middle income

countries to apply for loans, good effect for the economy.

Borrowers are less hesitance to apply for loans because corruption can reduce their application

costs and they feel that they have more chance to receive the loans. They can be both good and

bad borrowers. In the robustness check, our results state that in lower-middle income countries,

higher corruption can encourage good discouraged borrowers to apply for loans (Table B8).

(Table B3). Therefore, higher corruption more benefit for good borrowers or bad

borrowers is still a question for the future researches.

REFERENCES

Agarwal, S., & Hauswald, R. (2010). Distance and private information in lending. The Review of Financial Studies, 23(7), 2757-2788.

Akerlof, G. A. (1980). A theory of social custom, of which unemployment may be one consequence. The Quarterly Journal of Economics, 94(4), 749-775.

Anaere, C. (2014). Effect of corruption on bank lending; Evidence from sub-Saharan Africa. J. Int. Eco. Bank, 2, 15-27.

Andvig, J. C. (1991). The economics of corruption: A survey. Studi economici, 43(1), 57-94.

Asiedu, E., & Freeman, J. (2009). The effect of corruption on investment growth: Evidence from firms in Latin America, Africa, and transition countries. Review of Development Economics, 13(2), 200-214.

Axelrod, R. (1986). An evolutionary approach to norms. American political science review, 80(4), 1095-1111.

Banerjee, A. V. (1997). A theory of misgovernance. The Quarterly Journal of Economics, 112(4), 1289-1332.

Beck, P. J., & Maher, M. W. (1986). A comparison of bribery and bidding in thin markets. Economics Letters, 20(1), 1-5.

Beck, T., -Kunt, A., Laeven, L., & Maksimovic, V. (2006). The determinants of financing obstacles. Journal of international money and finance, 25(6), 932-952.

Bellier, A., Sayeh, W., & Serve, S. (2012). What lies behind credit rationing? A survey of the literature.

Berger, A. N., & Udell, G. F. (1995). Relationship lending and lines of credit in small firm finance. Journal of business, 351-381.

Biais, B., & Gollier, C. (1997). Trade credit and credit rationing. The Review of Financial Studies, 10(4), 903-937.

Cameron, L., Chaudhuri, A., Erkal, N., & Gangadharan, L. (2005). Do attitudes towards corruption differ across cultures? Experimental evidence from Australia, India, Indonesia and Singapore.

Cavalluzzo, K. S., Cavalluzzo, L. C., & Wolken, J. D. (2002). Competition, small business financing, and discrimination: Evidence from a new survey. The Journal of Business, 75(4), 641-679.

Chakravarty, S., & Xiang, M. (2009). What discourages small businesses from asking for loans? The international evidence on borrower discouragement. The international evidence on borrower discouragement.

Chakravarty, S., & Xiang, M. (2013). The international evidence on discouraged small businesses. Journal of Empirical Finance, 20, 63-82.

Chakravarty, S., & Yilmazer, T. (2009). A multistage model of loans and the role of relationships. Financial Management, 38(4), 781-816.

Chen, Y., Liu, M., & Su, J. (2013). Greasing the wheels of bank lending: Evidence from private firms in China. (37). (7)

Cole, R. A. (1998). The importance of relationships to the availability of credit. Journal of Banking & Finance, 22(6), 959-977.

Cole, R. A. (2008). Who needs credit and who gets credit? Evidence from the surveys of small business finances.

Cotugno, M., S., & Sampagnaro, G. (2013). Relationship lending, hierarchical distance and credit tightening: Evidence from the financial crisis. Journal of Banking & Finance, 37(5), 1372-1385.

Deidda, L., & Fattouh, B. (2002). Non-linearity between finance and growth. Economics Letters, 74(3), 339-345.

Della Porta, D. (2017). Corrupt exchanges: Actors, resources, and mechanisms of political corruption: Routledge.

-Kunt, A., & Maksimovic, V. (1998). Law, finance, and firm growth. The journal of finance, 53(6), 2107-2137.

Fastenbauer, J., & Robson, P. (2014). Discouraged Borrowers in the UK and Austria. Paper presented at the Proceedings of Global Business and Finance Research Conference.

Ferrando, A., & Mulier, K. (2015). The real effects of credit constraints: evidence from discouraged borrowers in the euro area.

Fraser, S. (2014). Back to borrowing? perspectives on the of Enterprise Research Centre, White Paper(8).

Freel, M., Carter, S., Tagg, S., & Mason, C. (2012). The latent demand for bank debt: characterizing Small Business Economics, 38(4), 399-418.

Z., Kochanova, A., & Weill, L. (2015). Does money buy credit? Firm-level evidence on bribery and bank debt. World development, 68, 308-322.

Galli, E. (2017). Does Corruption Affect Access to Bank Credit for Micro and Small Businesses? Evidence from European MSMEs.

Gama, A. P. M., Duarte, F. D., & J. P. (2017). Why discouraged borrowers exist? An empirical (re) examination from less developed countries. Emerging Markets Review, 33, 19-41.

Gutmann, J., Padovano, F., & Voigt, S. (2015). Perception vs. experience: explaining differences in corruption measures using microdata.

Han, L., Fraser, S., & Storey, D. J. (2009). Are good or bad borrowers discouraged from applying for loans? Evidence from US small business credit markets. Journal of Banking & Finance, 33(2), 415-424.

- G., & -Kant, J. (2011). SME financing in Europe: Cross-country determinants of bank loan maturity. International Small Business Journal, 29(5), 489-507.

Holton, S., Lawless, M., & McCann, F. (2014). Firm credit in the euro area: a tale of three crises. Applied economics, 46(2), 190-211.

Huntington, S. P. (1968). Political order in changing societies: Yale University Press.

Jappelli, T. (1990). Who is credit constrained in the US economy? The Quarterly Journal of Economics, 105(1), 219-234.

Jiangli, W., Unal, H., & Yom, C. (2008). Relationship lending, accounting disclosure, and credit availability during the Asian financial crisis. Journal of money, credit and banking, 40(1), 25-55.

Kon, Y., & Storey, D. J. (2003). A theory of discouraged borrowers. Small Business Economics, 21(1), 37-49.

Kysucky, V., & Norden, L. (2015). The benefits of relationship lending in a cross-country context: A meta-analysis. Management Science, 62(1), 90-110.

Leff, N. H. (1964). Economic development through bureaucratic corruption. American behavioral scientist, 8(3), 8-14.

Levenson, A. R., & Willard, K. L. (2000). Do firms get the financing they want? Measuring credit rationing experienced by small businesses in the US. Small Business Economics, 14(2), 83-94.

Leys, C. (1965). What is the Problem about Corruption? The Journal of Modern African Studies, 3(2), 215-230.

Lien, D.-H. D. (1986). A note on competitive bribery games. Economics Letters, 22(4), 337-341.

Lui, F. T. (1985). An equilibrium queuing model of bribery. Journal of political economy, 93(4), 760-781.

Mac an Bhaird, C., Vidal, J. S., & Lucey, B. (2016). Discouraged borrowers: Evidence for Eurozone SMEs. Journal of International Financial Markets, Institutions and Money, 44, 46-55.

Martin Brown, S. O., Alexander Popov, Pinar (2011). Who needs credit and who gets credit in Eastern Europe? Economic Policy, 26(65), 93-130.

Mayank Raturi, A. V. S. (1999). Explaining Ethnic Differentials in Credit Market Outcomes in Zimbabwe. Economic Development and Cultural Change, 47(3), 585-604.

Mo, P. H. (2001). Corruption and economic growth. Journal of comparative economics, 29(1), 66-79.

Norden, L., & Weber, M. (2010). Credit line usage, checking account activity, and default risk of bank borrowers. Review of Financial Studies, 23(10), 3665-3699.

Nye, J. S. (1967). Corruption and political development: A cost-benefit analysis. American political science review, 61(2), 417-427.

Olken, B. A., & Pande, R. (2012). Corruption in developing countries. Annu. Rev. Econ., 4(1), 479-509.

Pellegrini, L. (2011). Causes of corruption: a survey of cross-country analyses and extended results Corruption, development and the environment (pp. 29-51): Springer.

Petersen, M. A., & Rajan, R. G. (1994). The benefits of lending relationships: Evidence from small business data. The journal of finance, 49(1), 3-37.

Philp, M. (2016). Corruption and Measurement Measuring corruption (pp. 61-72): Routledge.

Podobnik, B., Shao, J., Njavro, D., Ivanov, P. C., & Stanley, H. E. (2008). Influence of corruption on economic growth rate and foreign investment. The European Physical Journal B, 63(4), 547-550.

Rioja, F., & Valev, N. (2004). Does one size fit all?: a reexamination of the finance and growth relationship. Journal of Development economics, 74(2), 429-447.

Rousseau, P. L., & Wachtel, P. (2002). Inflation thresholds and the finance growth nexus. Journal of international money and finance, 21(6), 777-793.

Shleifer, A., & Vishny, R. W. (1993). Corruption. The Quarterly Journal of Economics, 108(3), 599-617.

Steijvers, T., & Voordeckers, W. (2009). Collateral and credit rationing: a review of recent empirical studies as a guide for future research. Journal of Economic Surveys, 23(5), 924-946.

Stiglitz, J. E., & Weiss, A. (1981). Credit rationing in markets with imperfect information. The American economic review, 71(3), 393-410.

Tonoyan, V., Strohmeyer, R., Habib, M., & Perlitz, M. (2010). Corruption and entrepreneurship: How formal and informal institutions shape small firm behavior in transition and mature market economies. Entrepreneurship theory and practice, 34(5), 803-832.

Treisman, D. (2000). The causes of corruption: a cross-national study. Journal of public economics, 76(3), 399-457.

Vincent, C. (2010). An Interpretation of Discouraged Borrowers based on relationship Lending.

Voordeckers, W., & Steijvers, T. (2006). Business collateral and personal commitments in SME lending. Journal of Banking & Finance, 30(11), 3067-3086.

Wei, S.-J., & Kaufmann, D. (1999). Does grease money speed up the wheels of commerce? : The World Bank.

Weill, L. (2011). Does corruption hamper bank lending? Macro and micro evidence. Empirical Economics, 41(1), 25-42.

Xiang, D., Worthington, A. C., & Higgs, H. (2015). Discouraged finance seekers: An analysis of Australian small and medium-sized enterprises. International Small Business Journal, 33(7), 689-707.

APPENDIX

Appendix A Descriptive Statistic

Table A1 Discourage borrowers in East Asia & Pacific and South Asia

Country Total

(1) = (2)+(3)

Don't apply

(2)

Apply

(3)

Discouraged borrowers

(4)

Loan seekers

(5) = (3)+(4)

DBs/Loan seeker

(6) = (4) / (5)

China 2,556 1,913 643 771 1,414 54.5%

Malaysia 911 673 238 200 438 45.7%

Thailand 767 706 61 349 410 85.1%

Cambodia 316 276 40 99 139 71.2%

Indonesia 1,255 1,092 163 546 709 77.0%

Laos 358 306 52 100 152 65.8%

Myanma 579 502 77 156 233 67.0%

Papua New Guinea 65 49 16 11 27 40.7%

Philippines 986 804 182 66 248 26.6%

Solomon 146 118 28 18 46 39.1%

Timor Leste 109 92 17 35 52 67.3%

Vietnam 963 542 421 91 512 17.8%

India 8,777 8,010 767 3,050 3,817 79.9%

Pakistan 1,087 999 88 315 403 78.2%

Total 18,875 16,082 2,793 5,807 8,600 67.5%

Table A2 Main reasons for not applying loans Countries Main reasons for not applying loans/line of credit

No need a loan

sufficient capital

Complex application procedure

Not favorable interest

rates

High collateral

Insufficient size/

maturity

Did not think it

would be approved

Other Total

China 57.7% 11.2% 10.0% 8.6% 6.7% 4.3% 1.6% 100.0%

Malaysia 54.1% 5.9% 14.8% 6.1% 2.0% 1.5% 15.6% 100.0%

Thailand 49.1% 14.4% 18.3% 5.8% 4.7% 6.6% 1.1% 100.0%

Cambodia 61.6% 17.3% 10.7% 5.2% 2.2% 1.1% 1.8% 100.0%

Indonesia 41.9% 5.2% 25.6% 9.6% 6.0% 4.2% 7.6% 100.0%

Lao 54.3% 13.6% 8.3% 5.0% 1.0% 5.3% 12.6% 100.0%

Myanma 64.7% 17.9% 5.0% 2.8% 2.6% 3.0% 4.0% 100.0%

Papua New Guine 69.4% 6.1% 8.2% 2.0% 0.0% 6.1% 8.2% 100.0%

Philippines 88.5% 1.8% 3.8% 1.8% 0.4% 0.6% 3.1% 100.0%

Solomon Island 82.2% 5.9% 1.7% 6.8% 0.8% 2.5% 100.0%

Timor Leste 60.9% 14.1% 9.8% 7.6% 3.3% 3.3% 1.1% 100.0%

Vietnam 80.1% 5.6% 4.6% 4.1% 1.1% 1.5% 3.0% 100.0%

India 45.6% 7.7% 11.9% 9.9% 6.1% 3.1% 15.7% 100.0%

Pakistan 62.2% 5.9% 16.8% 6.5% 1.2% 3.0% 4.4% 100.0%

All sample 53.1% 8.3% 12.3% 8.1% 4.8% 3.2% 10.2% 100.0%

Tab

le A

3 M

ain

vari

able

s at

fir

m le

vel a

nd o

ther

var

iabl

es a

t cou

ntry

leve

l

Table A4 Summary statistics of variables

Variable Obs Mean Std. Dev. Min Max

Discouraged borrowers Applicants

Mean (SD)

Mean (SD)

DBs 8600 0.68 0.47 0 1 1

(0) 0

(0)

corrupt 19388 0.28 0.45 0 1 0.29 (0.45)

0.22*** (0.42)

corrupt1 11270 0.16 0.34 0 1 0.17 (0.35)

0.19** (0.36)

CPI 20212 36.26 4.76 21 50 36.62 (4.28)

36.58 (5.43)

relation1 20093 0.87 0.33 0 1 0.86 (0.35)

0.86 (0.34)

relation2 19577 0.42 0.49 0 1 0.4 (0.49)

0.47*** (0.50)

gender 20170 0.17 0.37 0 1 0.15 (0.36)

0.19*** (0.39)

experience 19738 15.11 9.56 1 70 14.53 (9.11)

16.71*** (9.52)

age 19981 18.35 13.07 0 162 17.6 (12.32)

18.03 (12.60)

size 20140 1.91 0.77 1 3 1.82 (0.76)

2.13*** (0.76)

Ffirm 20226 0.03 0.18 0 1 0.02 (0.12)

0.03*** (0.18)

diexport 20168 7.99 23.02 0 100 6.17 (20.44)

11.59*** (26.39)

tradecr 18920 45.11 35.23 0 100 41.97 (34.13)

49.44*** (33.46)

innovation 20099 0.25 0.44 0 1 0.23 (0.42)

0.33*** (0.47)

certificate 19914 0.37 0.48 0 1 0.34 (0.47)

0.43*** (0.50)

ExterAudit 19713 0.65 0.48 0 1 0.60 (0.49)

0.66*** (0.47)

creditline 19262 0.29 0.45 0 1 0.26 (0.44)

0.76*** (0.43)

access 20037 0.12 0.32 0 1 0.16 (0.36)

0.15 (0.36)

city1m 20308 0.50 0.50 0 1 0.49 (0.50)

0.60*** (0.49)

maincity 20307 0.70 0.46 0 1 0.69 (0.46)

0.63*** (0.48)

GDP 20308 2983.40 2302.54 1196 9649 2905.47 (2220.71)

3701.76*** (2630.67)

BURDEN 19391 3.58 0.48 3 5 3.63 (0.41)

3.69*** (0.57)

Significant levels 1% (***), 5% (**), 10% (*) for the difference in the means between discouraged borrowers and applicants

Appendix B Econometric models

Table B1 Variables definitions

Symbol Explain Sources

DEPENDENT VARIABLE DBs Dummy =1 for discouraged borrowers and 0 for applicants ES

DBs1 Dummy=1 for discouraged borrowers who have credit lines and 0 for applicants ES

INDEPENDENT VARIABLES

Corruption

corrupt Dummy =1 if corruption is major obstacle and very severe obstacle and 0 for otherwise ES

corrupt1 Number of times paying gifts (informal payments) divided by number of times applying public services ES

CPI Corruption perception index range from 0-100 with 100 is clear of corruption TI7

Lending relationship

relation1 Dummy =1 if firms have a checking or savings account and 0 for otherwise ES

relation2 Dummy =1 if firms have a overdraft facilitiy and 0 for otherwise ES

Entrepreneur characteristics

gender Dummy=1 if top managers are female and 0 for otherwise ES

experience Number of years of experience top managers work in the sector ES

Firm characteristics

age Firm age measured by number of year establishment operates ES

size Firm size measured by logarith number of employees for firms having more than 1 employee ES

Ffirm Dummy =1 if percentage of firm owned by foreigners is greater than 50% and 0 for otherwise ES

diexport Direct export as a percentage of sale ES

tradecr Proportion of total annual purchases of material inputs purchased on credit ES

innovation Dummy =1 if firms spent on formal research and development activities and 0 for otherwise ES

certificate Dummy =1 if firm got internationally-recognized quality certification ES

ExterAudit Dummy =1 if firms' annual financial statements are checked and certified by an external auditor ES

creditline Dummy = 1 if firms have a line of credit or loan from a financial institution ES

Firms perception about the operating environment

access Dummy =1 if access to finance is major obstacle and very severe obstacle and 0 for otherwise ES

Distance between lenders and borrowers

city1m Dummy = 1 if firm located in city with population over 1 million ES

maincity Dummy = 1 if firm located in main business city ES

Economic conditions GDP GDP per capita World Bank

BURDEN Burden of government regulation WEF8

7 Transparency International 8 The World Economic Forum

Tab

le B

2 -

Cor

rela

tion

Table B3 Main models (Equation 1 and 2)D d apply for loans because of complex procedures, high interest rate, high collateral, insufficient size of loan & maturity, fearing to be rejected and got value 0 if firms applied for loans. Column (1) tests the effect of corruption on discouraged borrowers for full data; Column (2) interact corruption and GDP per capita; Columns (3) & (4) split the sample into 2 groups (Upper-middle income countries and lower-middle income countries).

Regression of columns (1), (3) and (4) use Equation 1: Discouraged borrowersi,k 1*corrupti,k 2*relation1i,k + 3*relation2i,k + 4*genderi,k 5*experiencei,k 6*agei,k 7*sizei,k 8*Ffirmi,k 9*diexporti,k 10*tradecri,k 11*innovationi,k 12*certificatei,k + 13*ExterAuditi,k 14*creditlinei,k 15*accessi,k 16*city1mi,k 17*maincity + 1*COUNTRYk 2*INDUSTRYn 3*YEARt i,k

Regression of column (2) uses Equation 2: Discouraged borrowersi,k 1*corrupti,k*GDPk,t 2*relation1i,k + 3*relation2i,k 4*genderi,k 5*experiencei,k 6*agei,k 7*sizei,k 8*Ffirmi,k 9*diexporti,k 10*tradecri,k 11*innovationi,k 12*certificatei,k 13*ExterAuditi,k 14*creditlinei,k 15*accessi,k 16*city1mi,k 17*maincity 2*INDUSTRYn 3*YEARt i,k

(1) (2) (3) (4) VARIABLES Full data Interact GDP Upper Lower GDP 0.00003 (0.00008) corrupt -0.270** -0.66555*** 2.217*** -0.337*** (0.118) (0.19299) (0.177) (0.050) corrupt*GDP 0.00019** (0.00008) relation1 0.157 0.51565** -0.031 0.089 (0.150) (0.20377) (0.115) (0.175) relation2 -0.190** -0.05926 -0.125 -0.111** (0.094) (0.13490) (0.216) (0.045) gender -0.018 0.05300 -0.014 0.070 (0.121) (0.15194) (0.218) (0.125) experience -0.017*** -0.01938** -0.004 -0.019*** (0.005) (0.00806) (0.034) (0.004) age -0.002 0.00450 -0.025 -0.002 (0.003) (0.00725) (0.020) (0.004) size -0.239*** -0.21666** -0.123 -0.228** (0.087) (0.09339) (0.099) (0.101) Ffirm 0.393** 0.25885 0.575*** 0.620*** (0.192) (0.16729) (0.191) (0.224) diexport -0.002** -0.00351*** -0.004* -0.003** (0.001) (0.00134) (0.002) (0.001) tradecr -0.006 -0.00546 0.003 -0.009*** (0.004) (0.00378) (0.004) (0.002) innovation -0.132*** -0.33363*** -0.049 -0.116*** (0.035) (0.11183) (0.097) (0.043) certificate -0.175*** -0.17400*** -0.093 -0.175*** (0.055) (0.06360) (0.171) (0.049) ExterAudit -0.412* -0.49437* -1.021*** -0.253 (0.243) (0.26598) (0.381) (0.177) creditline -1.908*** -2.03057*** -3.278*** -1.431*** (0.563) (0.55916) (0.620) (0.346) access 0.136 0.15930 -0.061 0.209 (0.196) (0.20247) (0.567) (0.216) city1m 0.001 -0.09851 0.499 -0.032 (0.148) (0.24564) (1.034) (0.110) maincity -0.052 0.25003 0.115 -0.165 (0.162) (0.30218) (0.451) (0.129) Constant 2.774*** 2.18681*** 2.495*** 2.287*** (0.322) (0.64248) (0.835) (0.246) Observations 7,370 7,370 1,848 5,522 Correctly classified 81.94% 80.73% 85.50% 81.06% Country FE YES YES YES Industry FE YES YES YES YES Year FE YES YES YES YES Cluster country YES YES YES YES Method LOGIT LOGIT LOGIT LOGIT

Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

Table B4 Marginal effect of corruption on the probability of discouragement at different level of GDP per capita

margins, dydx(corrupt) at (GDP=(1000 (200) 9000)) Average marginal effects Number of obs = 7370 Model VCE : Robust Expression : Pr(DBs), predict() dy/dx w.r.t. : 1.corrupt

1.corrupt _at GDP dy/dx

Delta-method Std. Err. z P>z [95% Conf. Interval]

1000 -0.0749 0.0198 -3.78 0.000 -0.1138 -0.0360

1200 -0.0686 0.0183 -3.76 0.000 -0.1044 -0.0328

1400 -0.0624 0.0169 -3.70 0.000 -0.0955 -0.0293

1600 -0.0562 0.0157 -3.58 0.000 -0.0870 -0.0254

1800 -0.0501 0.0148 -3.38 0.001 -0.0791 -0.0211

2000 -0.0440 0.0142 -3.10 0.002 -0.0719 -0.0162

2200 -0.0380 0.0139 -2.73 0.006 -0.0653 -0.0107

2400 -0.0320 0.0140 -2.28 0.023 -0.0595 -0.0045

2600 -0.0261 0.0145 -1.80 0.071 -0.0545 0.0023

2800 -0.0202 0.0152 -1.33 0.184 -0.0501 0.0096

6600 0.0787 0.0481 1.64 0.102 -0.0155 0.1729

6800 0.0832 0.0500 1.66 0.096 -0.0147 0.1811

7000 0.0876 0.0519 1.69 0.091 -0.0141 0.1892

7200 0.0919 0.0538 1.71 0.087 -0.0135 0.1972

7400 0.0961 0.0556 1.73 0.084 -0.0129 0.2052

7600 0.1003 0.0575 1.74 0.081 -0.0124 0.2130

7800 0.1044 0.0594 1.76 0.079 -0.0120 0.2208

8000 0.1084 0.0613 1.77 0.077 -0.0117 0.2284

8200 0.1123 0.0631 1.78 0.075 -0.0114 0.2360

8400 0.1161 0.0650 1.79 0.074 -0.0112 0.2434

8600 0.1199 0.0668 1.79 0.073 -0.0111 0.2508

8800 0.1235 0.0686 1.80 0.072 -0.0110 0.2581

9000 0.1271 0.0705 1.80 0.071 -0.0110 0.2652

Note: dy/dx for factor levels is the discrete change from the base level.

Graph B1 Average marginal effects of corruption on the probability of discouragement at different level of GDP per capita

0 2000 4000 6000 8000 10000GDP per capita in $

Average Marginal Effects of 1.corrupt with 95% CIs

Table B5 Interaction between corruption and burden of government regulation (Equation 3)

D apply for loans because of complex procedures, high interest rate, high collateral, insufficient size of loan & maturity, fearing to be rejected and got value 0 if firms applied for loans.

Regression of the model : Discouraged borrowersi,k 1*corrupti,k*BURDENk,t 2*relation1i,k + 3*relation2i,k 4*genderi,k + 5*experiencei,k 6*agei,k 7*sizei,k 8*Ffirmi,k 9*diexporti,k 10*tradecri,k 11*innovationi,k 12*certificatei,k 13*ExterAuditi,k

14*creditlinei,k + 15*accessi,k 16*city1mi,k 17*maincity 2*INDUSTRYn 3*YEARt i,k

VARIABLES Interact BURDEN BURDEN 0.304 (0.573) corrupt -4.842*** (1.084) corrupt*BURDEN 1.325*** (0.312) relation1 0.645*** (0.176) relation2 -0.078 (0.137) gender 0.072 (0.170) experience -0.016** (0.006) age 0.004 (0.007) size -0.218** (0.091) Ffirm 0.299 (0.232) diexport -0.003** (0.001) tradecr -0.005 (0.004) innovation -0.354*** (0.111) certificate -0.170*** (0.066) ExterAudit -0.490* (0.263) creditline -2.009*** (0.580) access 0.135 (0.211) city1m -0.038 (0.215) maincity 0.207 (0.278) Constant 0.917 (2.448) Observations 7,063 Correctly classified 80.29% Industry FE YES Year FE YES Cluster country YES Method LOGIT

Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

Table B6 Marginal effect of corruption on the probability of discouragement at different level of burden of government regulation

margins, dydx(corrupt) at (BURDEN=(2.7 (.1) 5.0))

Average marginal effects Number of obs = 7063 Model VCE : Robust

Expression : Pr(DBs), predict() dy/dx w.r.t. : 1.corrupt

1.corrupt

_at BURDEN dy/dx Delta-method

Std. Err. z P>z [95% Conf. Interval]

2.7 -0.2148 0.0327 -6.58 0.000 -0.2788 -0.1507

2.8 -0.1912 0.0287 -6.66 0.000 -0.2474 -0.1349

2.9 -0.1675 0.0251 -6.67 0.000 -0.2168 -0.1183

3.0 -0.1439 0.0219 -6.56 0.000 -0.1870 -0.1009

3.1 -0.1205 0.0193 -6.24 0.000 -0.1584 -0.0827

3.2 -0.0975 0.0173 -5.64 0.000 -0.1313 -0.0636

3.3 -0.0748 0.0160 -4.69 0.000 -0.1061 -0.0435

3.4 -0.0528 0.0154 -3.43 0.001 -0.0829 -0.0226

3.5 -0.0314 0.0155 -2.03 0.043 -0.0617 -0.0010

3.6 -0.0107 0.0161 -0.67 0.505 -0.0423 0.0208

3.7 0.0090 0.0171 0.53 0.598 -0.0245 0.0426

3.8 0.0279 0.0185 1.51 0.131 -0.0083 0.0641

3.9 0.0458 0.0201 2.28 0.023 0.0064 0.0852

4.0 0.0627 0.0220 2.84 0.004 0.0195 0.1059

4.1 0.0785 0.0243 3.23 0.001 0.0308 0.1261

4.2 0.0932 0.0270 3.45 0.001 0.0402 0.1461

4.3 0.1068 0.0302 3.54 0.000 0.0476 0.1659

4.4 0.1192 0.0339 3.52 0.000 0.0529 0.1856

4.5 0.1306 0.0381 3.43 0.001 0.0559 0.2052

4.6 0.1408 0.0429 3.28 0.001 0.0568 0.2248

4.7 0.1499 0.0482 3.11 0.002 0.0555 0.2443

4.8 0.1579 0.0540 2.93 0.003 0.0521 0.2637

4.9 0.1649 0.0602 2.74 0.006 0.0469 0.2829

5.0 0.1709 0.0668 2.56 0.011 0.0399 0.3019

Note: dy/dx for factor levels is the discrete change from the base level.

Table B7 Robustness check: replace other proxies for corruption (Equation1)D apply for loans because of complex procedures, high interest rate, high collateral, insufficient size of loan & maturity, fearing to be rejected and got value 0 if firms applied for loans.

Columns (1), (2), (3) test the impact of number of times firms pay gifts (corrupt1) on discouraged borrowers. Column (1) conducts the test on full data, column (2) for group of upper-middle income countries and column (3) for group of lower-middle income countries.

Regression of columns (1), (2) and (3) : Discouraged borrowersi,k 1*corrupt1i,k 2*relation1i,k 3*relation2i,k 4*genderi,k + 5*experiencei,k 6*agei,k 7*sizei,k 8*Ffirmi,k 9*diexporti,k 10*tradecri,k 11*innovationi,k 12*certificatei,k 13*ExterAuditi,k

14*creditlinei,k 15*accessi,k 16*city1mi,k 17*maincity + 1*COUNTRYk 2*INDUSTRYn 3*YEARt i,k

Columns (4), (5), (6) test the impact of Corruption Perception Index (CPI) on discouraged borrowers. Colume (4) conducts the test on full data, column (5) for group of upper-middle income countries and column (6) for group of lower-middle income countries.

Regression of columns (4), (5) and (6) : Discouraged borrowersi,k 1*CPIk,t 2*relation1i,k + 3*relation2i,k 4*genderi,k + 5*experiencei,k 6*agei,k 7*sizei,k 8*Ffirmi,k 9*diexporti,k 10*tradecri,k 11*innovationi,k 12*certificatei,k 13*ExterAuditi,k

14*creditlinei,k 15*accessi,k 16*city1mi,k 17*maincity 2*INDUSTRYn 3*YEARt i,k

(1) (2) (3) (4) (5) (6) VARIABLES Full data Upper Lower CPI-Full data CPI-Upper CPI-Lower corrupt1 -0.201** 0.045 -0.258*** (0.078) (0.139) (0.070) CPI 0.036 -0.121*** 0.124** (0.037) (0.035) (0.059) relation1 -0.228 0.473* -0.420** 0.398** 0.103 0.093 (0.205) (0.251) (0.208) (0.161) (0.136) (0.178) relation2 -0.212*** -0.041 -0.153*** -0.050 -0.134 0.053 (0.059) (0.077) (0.048) (0.128) (0.168) (0.102) gender 0.086 0.059 0.132 0.039 -0.156 0.063 (0.096) (0.279) (0.131) (0.168) (0.212) (0.089) experience -0.019*** -0.002 -0.022*** -0.018** -0.003 -0.027*** (0.004) (0.026) (0.004) (0.007) (0.030) (0.005) age -0.001 -0.029 0.001 0.004 -0.027 0.005 (0.004) (0.021) (0.005) (0.007) (0.020) (0.007) size -0.315*** -0.119 -0.334*** -0.220** -0.136** -0.229** (0.052) (0.084) (0.065) (0.098) (0.069) (0.112) Ffirm 0.337 -0.090 0.635** 0.192 0.522*** 0.337 (0.266) (0.093) (0.314) (0.182) (0.192) (0.294) diexport -0.003** -0.009** -0.004** -0.004*** -0.004 -0.004*** (0.001) (0.004) (0.001) (0.001) (0.003) (0.001) tradecr -0.006 0.006 -0.010*** -0.006 0.004 -0.011*** (0.004) (0.004) (0.002) (0.004) (0.004) (0.002) innovation -0.045 -0.146*** 0.017 -0.339*** -0.023 -0.301*** (0.053) (0.050) (0.039) (0.112) (0.182) (0.108) certificate -0.234*** 0.005 -0.267*** -0.181*** -0.123 -0.141** (0.079) (0.191) (0.067) (0.065) (0.141) (0.060) ExterAudit -0.438 -1.208*** -0.102 -0.438* -0.849* -0.312 (0.315) (0.457) (0.146) (0.265) (0.463) (0.281) creditline -1.795*** -3.079*** -1.357*** -2.091*** -3.205*** -1.624*** (0.611) (0.617) (0.443) (0.568) (0.648) (0.441) access 0.008 0.233 0.025 0.166 0.201 0.251 (0.257) (0.710) (0.290) (0.234) (0.700) (0.251) city1m 0.183* 0.400 0.134 -0.047 0.210 -0.135 (0.103) (0.840) (0.084) (0.224) (1.185) (0.163) maincity -0.137 -0.152 -0.080 0.239 0.245 0.241 (0.134) (0.550) (0.122) (0.285) (0.550) (0.315) Constant 3.087*** 2.048*** 2.502*** 1.031 7.149*** -1.195 (0.433) (0.342) (0.458) (1.464) (2.091) (2.263) Observations 4,729 1,128 3,601 7,657 1,937 5,720 Correctly classified 80.69% 82.36% 79.76% 80.72% 85.23% 79.98% Country FE YES YES YES Industry FE YES YES YES YES YES YES Year FE YES YES YES YES YES YES Cluster country YES YES YES YES YES YES Method LOGIT LOGIT LOGIT LOGIT LOGIT LOGIT

Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

Table B8 Robustness check: changing the way to measure discouraged borrowers (Equation 1 and 2) have credit line and they need credit, but they

of complex procedures, high interest rate, high collateral, insufficient size of loan & maturity, fearing to be rejected. Column (1) tests the effect of corruption on discouraged borrowers for full data; Column (2) interact corruption and GDP per capita; Columns (3) & (4) split the sample into 2 groups (Upper-middle income countries and lower-middle income countries).

Regression of columns (1), (3) and (4) use Equation 1 : Discouraged borrowersi,k 1*corrupti,k 2*relation1i,k + 3*relation2i,k + 4*genderi,k 5*experiencei,k 6*agei,k 7*sizei,k 8*Ffirmi,k 9*diexporti,k 10*tradecri,k 11*innovationi,k 12*certificatei,k + 13*ExterAuditi,k 14*creditlinei,k 15*accessi,k 16*city1mi,k 17*maincity + 1*COUNTRYk 2*INDUSTRYn 3*YEARt i,k

Regression of column (2) uses Equation 2: Discouraged borrowersi,k 1*corrupti,k*GDPk,t 2*relation1i,k + 3*relation2i,k 4*genderi,k 5*experiencei,k 6*agei,k 7*sizei,k 8*Ffirmi,k 9*diexporti,k 10*tradecri,k 11*innovationi,k 12*certificatei,k 13*ExterAuditi,k 14*creditlinei,k 15*accessi,k 16*city1mi,k 17*maincity 2*INDUSTRYn 3*YEARt i,k

(1) (2) (3) (4) VARIABLES Strict Strict-Interract GDP Strict-Upper Strict-Lower GDP 0.00001 (0.00006) corrupt -0.419 -0.98928*** 2.019*** -0.579*** (0.265) (0.22171) (0.149) (0.155) corrupt*GDP 0.00030*** (0.00005) relation1 -0.108 0.45220 0.019 -0.253 (0.176) (0.29939) (0.180) (0.196) relation2 -0.010 0.17559 0.107 -0.067 (0.125) (0.16797) (0.117) (0.132) gender -0.051 -0.12815 -0.219 0.099 (0.221) (0.18534) (0.210) (0.270) experience -0.031*** -0.04068*** 0.001 -0.035*** (0.005) (0.00591) (0.028) (0.002) age -0.003 0.00633 -0.032 -0.000 (0.004) (0.00789) (0.025) (0.006) size -0.125 -0.15527 -0.022 -0.129 (0.137) (0.12932) (0.220) (0.151) Ffirm 0.537* 0.64286*** 0.898*** 0.440 (0.286) (0.21869) (0.105) (0.489) diexport -0.006*** -0.00761*** -0.002 -0.006*** (0.002) (0.00142) (0.001) (0.002) tradecr -0.008*** -0.00880*** -0.003 -0.010*** (0.003) (0.00302) (0.003) (0.003) innovation -0.103** -0.31436*** -0.199 -0.094 (0.046) (0.10290) (0.205) (0.062) certificate -0.316*** -0.29040** -0.068 -0.398*** (0.107) (0.11834) (0.166) (0.089) ExterAudit -0.528 -0.56706 -1.069* -0.401*** (0.333) (0.36326) (0.628) (0.119) access 0.333 0.45922* 0.213 0.349 (0.225) (0.25539) (0.260) (0.289) city1m -0.166* -0.46418 -0.303 -0.117 (0.093) (0.29262) (1.001) (0.071) maincity -0.222 0.36086 0.138 -0.310 (0.224) (0.44476) (0.530) (0.200) Constant 0.855 0.25070 -0.104 2.205** (0.920) (0.84723) (0.940) (0.879) Observations 3,177 3,177 782 2,395 Correctly classified 77.02% 74.41% 80.18% 76.08% Country FE YES YES YES Industry FE YES YES YES YES Year FE YES YES YES YES Cluster country YES YES YES YES Method LOGIT LOGIT LOGIT LOGIT

Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

Table B9 Marginal effect of corruption on the probability of discouragement at different level of GDP per capita. Discouraged borrowers is measured by strict method.

margins, dydx(corrupt) at (GDP=(1000 (200) 9000))

Average marginal effects Number of obs = 3177 Model VCE : Robust Expression : Pr(DBs2), predict() dy/dx w.r.t. : 1.corrupt

1.corrupt _at GDP dy/dx

Delta-method Std. Err. z P>z [95% Conf. Interval]

1000 -0.1151 0.0293 -3.93 0.000 -0.1725 -0.0577

1200 -0.1055 0.0276 -3.82 0.000 -0.1595 -0.0514

1400 -0.0957 0.0260 -3.68 0.000 -0.1466 -0.0448

1600 -0.0859 0.0245 -3.51 0.000 -0.1339 -0.0379

1800 -0.0760 0.0231 -3.29 0.001 -0.1212 -0.0307

2000 -0.0660 0.0218 -3.02 0.002 -0.1089 -0.0232

2200 -0.0561 0.0208 -2.70 0.007 -0.0967 -0.0154

2400 -0.0460 0.0199 -2.31 0.021 -0.0850 -0.0071

2600 -0.0359 0.0192 -1.87 0.061 -0.0736 0.0017

2800 -0.0258 0.0188 -1.37 0.169 -0.0626 0.0110

4000 0.0352 0.0211 1.67 0.096 -0.0062 0.0766

4200 0.0454 0.0222 2.05 0.041 0.0020 0.0888

4400 0.0556 0.0233 2.39 0.017 0.0099 0.1012

4600 0.0657 0.0245 2.68 0.007 0.0177 0.1137

4800 0.0758 0.0258 2.94 0.003 0.0253 0.1263

5000 0.0859 0.0271 3.17 0.002 0.0328 0.1390

5200 0.0960 0.0284 3.37 0.001 0.0402 0.1517

5400 0.1060 0.0298 3.56 0.000 0.0476 0.1644

5600 0.1159 0.0312 3.72 0.000 0.0549 0.1770

5800 0.1258 0.0325 3.87 0.000 0.0621 0.1896

6000 0.1357 0.0339 4.01 0.000 0.0693 0.2021

6200 0.1455 0.0352 4.13 0.000 0.0765 0.2145

6400 0.1552 0.0365 4.25 0.000 0.0837 0.2268

6600 0.1649 0.0378 4.36 0.000 0.0908 0.2389

6800 0.1744 0.0390 4.47 0.000 0.0979 0.2509

7000 0.1839 0.0402 4.57 0.000 0.1050 0.2628

7200 0.1933 0.0414 4.67 0.000 0.1121 0.2745

7400 0.2026 0.0426 4.76 0.000 0.1191 0.2860

7600 0.2117 0.0437 4.85 0.000 0.1262 0.2973

7800 0.2208 0.0447 4.94 0.000 0.1331 0.3085

8000 0.2298 0.0458 5.02 0.000 0.1401 0.3195

8200 0.2386 0.0468 5.10 0.000 0.1469 0.3303

8400 0.2473 0.0477 5.18 0.000 0.1538 0.3409

8600 0.2559 0.0487 5.26 0.000 0.1605 0.3513

8800 0.2643 0.0496 5.33 0.000 0.1672 0.3615

9000 0.2726 0.0504 5.40 0.000 0.1738 0.3715 Note: dy/dx for factor levels is the discrete change from the base level.

Table B10 Robustness check: using only truthful answers and re-test the main model (Equation 1 and 2)

D interest rate, high collateral, insufficient size of loan & maturity, fearing to be rejected and got value 0 if firms applied for loans. Column (1) tests the effect of corruption on discouraged borrowers for full data; Column (2) interact corruption and GDP per capita; Columns (3) & (4) split the sample into 2 groups (Upper-middle income countries and lower-middle income countries). Data includes only truthful answers. Regression of columns (1), (3) and (4) use Equation 1: Discouraged borrowersi,k 1*corrupti,k 2*relation1i,k + 3*relation2i,k +

4*genderi,k 5*experiencei,k 6*agei,k 7*sizei,k 8*Ffirmi,k 9*diexporti,k 10*tradecri,k 11*innovationi,k 12*certificatei,k + 13*ExterAuditi,k 14*creditlinei,k 15*accessi,k 16*city1mi,k 17*maincity + 1*COUNTRYk 2*INDUSTRYn 3*YEARt i,k

Regression of column (2) uses Equation 2 : Discouraged borrowersi,k 1*corrupti,k*GDPk,t 2*relation1i,k + 3*relation2i,k 4*genderi,k 5*experiencei,k 6*agei,k 7*sizei,k 8*Ffirmi,k 9*diexporti,k 10*tradecri,k 11*innovationi,k 12*certificatei,k 13*ExterAuditi,k 14*creditlinei,k 15*accessi,k 16*city1mi,k 17*maincity 2*INDUSTRYn 3*YEARt i,k

(1) (2) (3) (4) VARIABLES Full data Interact GDP Upper Lower GDP 0.00014* (0.00007) corrupt -0.512*** -0.95441*** 1.570* -0.527*** (0.139) (0.24256) (0.858) (0.047) corruptGDP 0.00023* (0.00012) relation1 -0.476*** 0.10904 0.009 -0.646*** (0.139) (0.20695) (0.416) (0.147) relation2 -0.175 -0.02242 -0.073 -0.099 (0.172) (0.20633) (0.249) (0.119) gender -0.060 -0.01266 -0.373*** 0.070 (0.161) (0.14114) (0.144) (0.168) experience -0.019** -0.02273** -0.029 -0.017*** (0.008) (0.00933) (0.041) (0.004) age -0.004 0.00181 -0.023 -0.004 (0.003) (0.00594) (0.016) (0.004) size -0.232*** -0.23436*** -0.127 -0.208** (0.077) (0.07994) (0.097) (0.095) Ffirm 0.077 -0.00585 0.250* 0.408 (0.176) (0.13921) (0.151) (0.257) diexport -0.003*** -0.00383*** -0.004 -0.003** (0.001) (0.00131) (0.003) (0.001) tradecr -0.005 -0.00510 0.006*** -0.009*** (0.004) (0.00389) (0.002) (0.002) innovation -0.036 -0.19598 -0.343*** 0.096 (0.117) (0.16300) (0.086) (0.065) certificate -0.335*** -0.32744*** -0.190*** -0.415*** (0.073) (0.07888) (0.049) (0.062) ExterAudit -0.348*** -0.39160** -0.434** -0.497*** (0.133) (0.16940) (0.214) (0.100) creditline -1.809*** -1.92255*** -3.428*** -1.222*** (0.691) (0.68086) (0.489) (0.412) access -0.033 -0.06280 -0.265 0.060 (0.194) (0.18247) (0.257) (0.194) city1m 0.167 -0.03415 0.610 0.125 (0.169) (0.24200) (1.820) (0.129) maincity 0.050 0.39781 0.281 -0.116 (0.239) (0.38706) (0.570) (0.180) Constant 3.061*** 1.62536** 2.207 2.855*** (0.533) (0.65359) (1.461) (0.540) Observations 4,523 4,523 1,278 3,245 Correctly classified Country FE YES YES YES Industry FE YES YES YES YES Year FE YES YES YES YES Cluster country YES YES YES YES Method LOGIT LOGIT LOGIT LOGIT

Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

Table B11 Robustness check: exclude China and India from the sample (Equation 1 and 2) apply for loans because of complex procedures, high interest rate,

high collateral, insufficient size of loan & maturity, fearing to be rejected and got value 0 if firms applied for loans. Column (1) tests the impact of corruption on discouraged borrowers for the data excluding China and India; Column (2) interacts corruption and GDP per capita; Column (3) & (4) split the sample into 2 groups (Upper-middle income countries excluding China and lower-middle income countries excluding India).

Regression of columns (1), (3) and (4) use Equation 1: Discouraged borrowersi,k 1*corrupti,k 2*relation1i,k + 3*relation2i,k + 4*genderi,k 5*experiencei,k 6*agei,k 7*sizei,k 8*Ffirmi,k 9*diexporti,k 10*tradecri,k 11*innovationi,k 12*certificatei,k + 13*ExterAuditi,k 14*creditlinei,k 15*accessi,k 16*city1mi,k 17*maincity + 1*COUNTRYk 2*INDUSTRYn 3*YEARt i,k

Regression of column (2) use Equation 2 : Discouraged borrowersi,k 1*corrupti,k*GDPk,t 2*relation1i,k + 3*relation2i,k 4*genderi,k 5*experiencei,k 6*agei,k 7*sizei,k 8*Ffirmi,k 9*diexporti,k 10*tradecri,k 11*innovationi,k 12*certificatei,k 13*ExterAuditi,k 14*creditlinei,k 15*accessi,k 16*city1mi,k 17*maincity 2*INDUSTRYn 3*YEARt i,k

(1) (2) (3) (4) VARIABLES Exclude Chi&Ind Exclude Chi&Ind Upper-Exclude Chi Lower-Exclude Ind GDP 0.00004 (0.00008) corrupt -0.051 -0.91541*** 2.037*** -0.410* (0.353) (0.29824) (0.152) (0.220) corruptGDP 0.00027*** (0.00006) relation1 -0.033 0.21874 -0.496 -0.073 (0.116) (0.16580) (0.446) (0.146) relation2 -0.026 0.02541 0.117 -0.146 (0.174) (0.16491) (0.522) (0.169) gender -0.016 0.18216 0.148 -0.054 (0.162) (0.20044) (0.179) (0.180) experience -0.017 -0.01784 0.094*** -0.033*** (0.015) (0.01723) (0.005) (0.010) age 0.004 0.02045* -0.067*** 0.017** (0.011) (0.01221) (0.006) (0.007) size -0.077 -0.04085 -0.238* -0.034 (0.154) (0.14020) (0.137) (0.189) Ffirm 0.544** 0.50977** 1.252 0.634*** (0.265) (0.21605) (1.317) (0.242) diexport -0.001 -0.00200 0.003 -0.001 (0.002) (0.00245) (0.006) (0.003) tradecr -0.003* -0.00357** -0.008*** -0.002 (0.002) (0.00144) (0.001) (0.002) innovation -0.046 -0.65345*** 0.565*** -0.117 (0.181) (0.21899) (0.190) (0.191) certificate -0.337** -0.34008* -0.453*** -0.217 (0.162) (0.19134) (0.124) (0.260) ExterAudit -0.471* -0.68716*** -0.285 -0.647** (0.252) (0.23135) (0.245) (0.299) creditline -2.493*** -2.63835*** -2.756** -2.516*** (0.359) (0.33339) (1.221) (0.356) access 0.683*** 0.72412*** 1.126** 0.688*** (0.111) (0.12038) (0.444) (0.112) city1m 0.186 0.28839 0.095 0.293** (0.276) (0.30504) (0.766) (0.120) maincity 0.353* 0.68513*** 0.888*** 0.172 (0.200) (0.25985) (0.003) (0.190) Constant 3.481*** 2.74352*** 2.133*** 2.964*** (0.526) (0.49801) (0.538) (0.463) Observations 2,543 2,543 597 1,946 Correctly classified 81.71% 80.30% 85.76% 81.55% Country FE YES YES YES Industry FE YES YES YES YES Year FE YES YES YES YES Cluster country YES YES YES YES Method LOGIT LOGIT LOGIT LOGIT

Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1