Are Corruption and Regulation Less Burdensome in Special Economic Zones?

Abstract

Many developing country governments would like to attract investment and create jobs in manufacturing and high-tech industries. Heavy and unpredictable laws and regulations, frequent demands for bribes, high taxes, poor-quality roads, slow and inefficient ports, and unreliable power, however, deter private investors. Moreover, political opposition and fiscal constraints prevent governments from resolving the numerous issues. Rather than attempting to solve everything everywhere, many governments have tried to fix problems in only small regions. These special economic zones (SEZs) often have lower taxes, more liberal regulation, and better infrastructure. This paper asks whether firms located in African and South Asian SEZs report less regulation and corruption than other firms in the same countries. We find, on average, being located in an SEZ is associated with lower burdens due to corruption and regulation. Firms in the zones are less likely to report paying bribes than firms outside the zones and report spending less time dealing with inspections and regulations. However, this is not true in Africa; firms in African zones report that corruption and regulation are as troublesome as for similar firms outside the zones.

1. Introduction

Developing countries often see private investment as an instrument for creating jobs, boosting exports, and diversifying into manufacturing and high-tech industries. Although some successfully attract private and foreign investors, others struggle. Private firms do not want to risk entering countries with costly and unpredictable regulations, unreliable electricity and roads, high taxes on formal businesses, and corrupt officials demanding bribes. When countries have poor business environments, private firms search for more attractive destinations.

The best way to increase private investment would be to provide all firms with a better investment climate. Solving all problems across entire countries, however, can be challenging. Existing firms with market power often oppose policies that make it easier to start new businesses. Improving roads, ports, and electricity is expensive and time-consuming, especially after years of underinvestment. Cutting taxes on formal businesses damages public finances in countries with narrow tax bases and many informal firms. And social and environmental interest groups often oppose reforms to reduce red tape that they believe might undermine the country’s social and environmental goals. Completing all needed reforms would often be overwhelming and expensive.

Governments, therefore, often adopt a second-best approach. Rather than fixing everything everywhere, they establish special economic zones (SEZs) that offer attractive business environments in limited areas. The zones aspire to attract new investors with low taxes, light regulation, responsive and honest bureaucracies, reduced barriers to imported goods, and dependable power and transportation. Although different zones offer different benefits, temporary or permanent tax incentives are almost universal.

Despite their prominence, SEZs have not always succeeded, especially in sub-Saharan Africa. With a few exceptions, such as Mauritius, Farole (2011a) finds African zones have not increased investment, created jobs, or boosted exports. Farole and Moberg (2014) argue African zones often fail because the zones cannot compete with the most attractive investment destinations. Similarly, Aggarwal (2023) finds that although South Asian SEZs provide good infrastructure and tax incentives, they have not delivered the same improvements in governance and corruption.

This paper explores why some zones, especially in Africa, have failed to entice new foreign investors. It asks whether firms in African and South Asian SEZs report less corruption and more modest regulatory burdens—two critical components of the investment climate—than other firms. Using data from the World Bank’s Enterprise Surveys in Africa and South Asia, we compare how firms inside and outside the zones in these countries answer questions about bribery and red tape. To the best of our knowledge, this is the first paper to do this using a large cross-country sample.1 The larger sample will make it easier to uncover statistically significant differences and will allow us to compare Africa with South Asia.

We find, on average, that being located in an SEZ is associated with lighter regulation and less corruption. SEZ firms, on average, report spending less time dealing with government requirements and report paying fewer bribes than other firms. Although unobserved firm-level differences might affect these associations, they are suggestive.

These differences, however, are not visible everywhere. Although South Asian SEZ firms report lower corruption and lighter regulation than non-zone firms, sub-Saharan African SEZ firms do not. This could explain why many African zones have struggled to attract investment and create employment.

2. Methodology

2.1. Background

Special economic zones are economic zones within countries where legal regimes, taxes, regulations, and other investment climate policies are different from elsewhere in the country (Aggarwal, 2023; Farole, 2011a).2 Farole (2011a, p. 17) defines special economic zones as:

… spatially delimited areas within an economy that function with administrative, regulatory, and often fiscal regimes that are different (typically more liberal) than those of the domestic economy.

Because governments want to use the zones to attract private and foreign investment, boost exports, and create jobs, they adopt policies that will make the zones more attractive to private investors.

The number of special economic zones, and the number of countries with zones, has been growing. Using data from the International Labour Organization’s database, Farole (2011a) reports there were 176 zones in 47 countries in 1986. By 2006, there were 3500 zones in 130 countries. By 2019, there were 5383 zones (United Nations Conference on Trade and Development, 2019).3

As the number of zones has increased, they have become more diverse.4 Different zones have different goals, offer different incentives, and cover different industries. The smallest zones can contain only a single firm, while the largest can cover entire regions or even the entire country.5 Some focus on a single sector such as textiles or high tech industry, while others are large and diversified (Stein, 2008). Some, such as free ports and export processing zones (EPZs), focus on exports while others do not. Many recent SEZs are designed to be attractive sites that will link firms into global value chains (Aggarwal, 2023).

Although different zones offer different benefits, tax incentives are among the most important. In Togo, twelve of seventeen firms in EPZs said tax incentives were the main reason they located in the zones.6 Similarly, Kinyondo et al. (2016) found firms in Tanzanian SEZs reported the same: tax incentives were the most important benefit they received. Finally, based on firm surveys in African SEZs, Farole (2011a) found tariffs and corporate taxes were the fourth and fifth most important criteria—out of eleven—when deciding where to invest.7 Perhaps because of this, most special economic zones offer some tax incentives. In a survey of SEZs in twenty-six sub-Saharan African countries, Newman and Page (2017) found all but one offered some tax incentives to firms in the zones. Despite their popularity, however, Frick et al. (2019) found that tax exemptions were not successful in encouraging zone growth in the poorest countries.

The zones do not only offer tax incentives; many also provide more liberal regulatory regimes.8 Governments can allow firms to avoid complying with labor regulations, make hiring foreign managers and specialists easier, and reduce the number of required licenses—especially for importing and exporting (Farole, 2011b, p. 2). Some zones also have specialized agencies that can either provide firms with the licenses they need or can help them get licenses from other departments (Farole & Moberg, 2014; Moberg, 2018). These ‘one-stop shops’ can also help firms handle other laws and regulations—something especially important for foreign investors who are unfamiliar with the country. Firms in African zones ranked regulation as the third most important criteria (of eleven) when deciding where to invest, while firms in non-African zones ranked it fifth (Farole, 2011a).

If these policies reduce the cost of complying with regulation, zone firms should spend less time dealing with regulation than similar firms in other parts of the country. To test this, we must control for other differences between firms that affect the regulatory burden. For example, managers of exporting firms might spend more time dealing with customs than managers of non-exporting firms. If zone firms are more likely to export, it might appear the zones failed to reduce the regulatory burden if we do not control for this. This leads to our first hypothesis:

Hypothesis 1.

All else equal, we would expect firms inside the zones to spend less time dealing with government regulation than firms outside the zones.

Policies that liberalize regulation and cut taxes might also reduce corruption. Managers might be willing to pay small bribes if doing so allows them to get an import license more quickly or avoid installing expensive equipment to comply with environmental regulations. But when approvals are faster and complying with regulations less expensive or time-consuming, managers might be less willing to pay bribes.9 Similarly, when taxes are low, businesses have less reason to bribe tax inspectors. Reforms that lighten the regulatory burden, reduce taxes, and streamline licensing procedures will reduce managers’ reasons to pay bribes and bureaucrats’ ability to ask for them.

Empirical studies support the idea that improving regulation also reduces corruption. Corruption is a smaller problem in countries with lighter regulation (Knack & Keefer, 1995; Langbein & Knack, 2010; Mauro, 1995). Similarly, bribes are more common in countries where registering a business takes longer (Djankov et al., 2002; Svensson, 2005). Finally, firms are more likely to pay bribes when they meet more often with government agencies and spend more time dealing with regulation (Clarke, 2011; Gonzalez et al., 2007).10

Governments can also use direct policies to target corruption in the zones. For example, they can set up watchdogs to monitor corruption, pay officials higher salaries, or have ‘one-stop shops’ that can issue licenses and permits, bypassing corrupt officials in existing agencies.

Although reducing corruption is difficult, reforms that only affect the zones might face less political opposition. Privileged interest groups outside the zones might not oppose—or even know about—reforms that only affect zone firms. Governments can therefore experiment with controversial policies and programs in the zones that would be too politically difficult for the whole country (Auty, 2011; Moberg, 2015; Stein, 2008). For example, corruption watchdogs that focus only on zone officials might not threaten bribetakers outside the zones. Similarly, introducing a new licensing regime in a new SEZ will be less threatening if the government preserves the old regime elsewhere—especially if zone firms focus on exporting or competing with imports.11 Officials working in agencies that process licenses might resist changes that simplify applications or eliminate licenses if they would lose their jobs or their opportunities to take bribes.12 But if the reforms only affect new investors in the SEZs, they might feel less threatened. This leads us to our second hypothesis:

Hypothesis 2.

All else equal, we would expect firms inside the zones to be less likely to pay bribes than firms outside of the zones.

For the reasons outlined above, corruption and regulation should be less costly inside the zones. Newman and Page (2017), however, argue reform has not always succeeded in Africa. Among the SEZs and EPZs they studied, only some reduced regulation or improved institutions.

One problem is that agencies that enforce regulations might not treat SEZ firms differently from other firms. First, overwhelmed agencies might find it difficult to quickly process applications from firms within the zones with their limited resources. Second, some agencies might actively oppose the reforms. They might feel the regulations they enforce are important and thus oppose reforms within the zones that weaken them. They might also oppose the reforms if they feel it affects their agency’s power or undermines their ability to collect bribes. Third, some agencies might lack the resources to effectively manage and enforce multiple regimes. The agencies might therefore fail to cooperate with zone officials.

One way to circumvent existing agencies is to set up independent agencies that directly regulate zone firms or help zone firms navigate the existing bureaucracy. Although ‘one-stop shops’ seem attractive, these agencies often fail. One problem is they often lack the authority to issue licenses directly, instead relying on bureaucrats seconded from their parent agencies (Farole & Moberg, 2014). Because these officials’ appointments remain in their parent agencies, their interests might not change. They, therefore, might fail to process licenses and permits efficiently. In other cases, the one-stop shop can only route the applications to the relevant agencies rather than processing them directly. Without the same incentives as one-stop shop officials, officials in the existing agencies might treat applications from zone firms in the same way as they treat other applications. Furthermore, workers in the one-stop shops might not have the institutional power or influence to force recalcitrant officials to behave any differently. Consequently, the process in the zones is often no faster or less costly than elsewhere—a common problem in Africa.13 Newman and Page (2017) write: “there are few African countries where central SEZ authorities have the decision-making power over regulatory authorities” (p. 24). They note the one-stop shop in Lesotho could not ensure officers from the various ministries worked together. They also note in other countries, including Tanzania, Nigeria, and Kenya, that there was no formal institutional link between the agencies in the SEZs.

Another reason zones might not have lower regulatory burdens is they often make firms comply with extra rules (Moberg, 2015). For example, firms might need to hire enough local workers, export enough, or invest enough in the zones. The new requirements introduce new opportunities for corrupt bureaucrats to ask for bribes and for firms to offer bribes to avoid complying. Based on a survey of twenty-four firms in Tanzanian EPZs, Kinyondo et al. (2016) report officials inspected the average firm thirteen times to ensure they were fulfilling zone requirements, mostly related to exporting or fulfilling policy and technical requirements.

In summary, firms might find regulation and corruption less costly inside SEZs than elsewhere in the country. First, governments often adopt reforms—simplified regulatory regimes and one-stop shops—that could reduce the regulatory burden in the zones. Lighter and improved regulation—and lower taxes—might also reduce firms’ incentives to pay bribes and officials’ ability to demand them. Second, political opposition might be lower if the government implements reform only in the zones, especially when the SEZ firms are new entrants focused on exporting. If these reforms succeed, firms might find regulation and corruption to be lesser constraints in the SEZs. But reforms sometimes fail, and, therefore, the zones might not have less regulation or corruption.

2.2. Data

This paper analyzes data from all World Bank Enterprise Surveys (WBESs) that enquire whether the firm is in an SEZ.14 The core survey—the standardized questions asked worldwide—does not ask about SEZs.15 Some surveys, however, do; World Bank regional and country teams can insert questions into specific surveys.16 Although regional teams in sub-Saharan Africa and South Asia chose to add the SEZ question, the teams in Latin America, Europe and Central Asia, East Asia, and the Middle East and North Africa did not.17 We can, therefore, analyze data from only 50 low- and middle-income countries in South Asia and sub-Saharan Africa (see Appendix A for a list).

The WBESs cover manufacturing, retail trade, and other services. Because government agencies supply the lists for the sampling frames, the samples exclude informal or unregistered firms. The surveys also exclude microenterprises with fewer than five workers. Because the surveys focus on the private sector, they omit fully—but not partly—government-owned firms. Table 1 presents the main variables’ sample means.

Table 1. Sample statistics.

	Means
Firm is in an SEZ	38%
Ave. percent of time spent dealing with regulation	6.3
% of firms reporting paying bribes	26%
Age of firm (years)	17.6
Number of workers	105.7
% of firms exporting	19%
% of firms foreign-owned	5%
% of firms partly government-owned	0%

Note: Based only on observations in the regression in column 3 of Table 2.

Table 1. Sample statistics.

	Means
Firm is in an SEZ	38%
Ave. percent of time spent dealing with regulation	6.3
% of firms reporting paying bribes	26%
Age of firm (years)	17.6
Number of workers	105.7
% of firms exporting	19%
% of firms foreign-owned	5%
% of firms partly government-owned	0%

Note: Based only on observations in the regression in column 3 of Table 2.

Table 2. Differences in bribes and regulation between firms inside and outside SEZs.

	(1)	(2)	(3)	(4)
	Firm Pays Bribes (Dummy)		% of Time Spent Dealing with Regulation
Firm is in a special economic zone
Firm is in a special economic zone	−0.130 ***		−1.381 ***
[dummy]	(−3.67)		(−2.83)
Firm is in an export processing zone		−0.312 ***		−6.261 ***
[dummy]		(−4.27)		(−4.01)
Firm is an in industrial zone		−0.255 ***		−2.385 ***
[dummy]		(−5.52)		(−2.82)
Other firm characteristics
Age of firm	−0.014	0.032	0.845 ***	1.344 ***
[years, natural log]	(−0.82)	(1.37)	(3.44)	(3.24)
Number of workers	0.325 ***	0.336 ***	3.589 ***	3.809 ***
[natural log]	(6.50)	(5.45)	(5.44)	(3.64)
Number of workers squared	−0.038 ***	−0.042 ***	−0.350 ***	−0.375 ***
[natural log]	(−6.17)	(−5.36)	(−4.43)	(−3.00)
Firm is an exporter	0.139 ***	0.139 ***	6.159 ***	9.177 ***
[dummy]	(3.59)	(2.91)	(10.48)	(9.96)
Firm is foreign-owned	−0.148 **	−0.059	0.849	2.363 *
[dummy]	(−2.56)	(−0.68)	(1.36)	(1.91)
Firm is partly government-owned	−0.273	0.002	−0.488	−1.524
[dummy]	(−1.41)	(0.01)	(−0.19)	(−0.29)
Sector dummies	Yes	Yes	Yes	Yes
Country-year dummies	Yes	Yes	Yes	Yes
Observations	20,491	13,579	23,428	15,818
Number of country-years	48	24	50	26
Pseudo R-squared	0.205	0.104	0.0224	0.0236
H0: EPZ and other zones coefficients equal (chi-squared [1])		0.56		5.60 **
(p-value)		0.45		0.02

Source: Author’s calculations based upon data from the World Bank Enterprise Survey data. Note: t-statistics in parentheses. All regressions include country-year and sector dummies (ISIC 4-figure). The bribe variable is a dummy, and so the model is a probit model. The regulation variable is censored at 0 and 100 percent, and so the model is a tobit model. Standard errors are clustered at the survey strata level. ***, **, * Statistically significant at 1%, 5%, and 10% significance levels.

Table 2. Differences in bribes and regulation between firms inside and outside SEZs.

	(1)	(2)	(3)	(4)
	Firm Pays Bribes (Dummy)		% of Time Spent Dealing with Regulation
Firm is in a special economic zone
Firm is in a special economic zone	−0.130 ***		−1.381 ***
[dummy]	(−3.67)		(−2.83)
Firm is in an export processing zone		−0.312 ***		−6.261 ***
[dummy]		(−4.27)		(−4.01)
Firm is an in industrial zone		−0.255 ***		−2.385 ***
[dummy]		(−5.52)		(−2.82)
Other firm characteristics
Age of firm	−0.014	0.032	0.845 ***	1.344 ***
[years, natural log]	(−0.82)	(1.37)	(3.44)	(3.24)
Number of workers	0.325 ***	0.336 ***	3.589 ***	3.809 ***
[natural log]	(6.50)	(5.45)	(5.44)	(3.64)
Number of workers squared	−0.038 ***	−0.042 ***	−0.350 ***	−0.375 ***
[natural log]	(−6.17)	(−5.36)	(−4.43)	(−3.00)
Firm is an exporter	0.139 ***	0.139 ***	6.159 ***	9.177 ***
[dummy]	(3.59)	(2.91)	(10.48)	(9.96)
Firm is foreign-owned	−0.148 **	−0.059	0.849	2.363 *
[dummy]	(−2.56)	(−0.68)	(1.36)	(1.91)
Firm is partly government-owned	−0.273	0.002	−0.488	−1.524
[dummy]	(−1.41)	(0.01)	(−0.19)	(−0.29)
Sector dummies	Yes	Yes	Yes	Yes
Country-year dummies	Yes	Yes	Yes	Yes
Observations	20,491	13,579	23,428	15,818
Number of country-years	48	24	50	26
Pseudo R-squared	0.205	0.104	0.0224	0.0236
H0: EPZ and other zones coefficients equal (chi-squared [1])		0.56		5.60 **
(p-value)		0.45		0.02

Source: Author’s calculations based upon data from the World Bank Enterprise Survey data. Note: t-statistics in parentheses. All regressions include country-year and sector dummies (ISIC 4-figure). The bribe variable is a dummy, and so the model is a probit model. The regulation variable is censored at 0 and 100 percent, and so the model is a tobit model. Standard errors are clustered at the survey strata level. ***, **, * Statistically significant at 1%, 5%, and 10% significance levels.

The dependent variables measure whether firms say they pay bribes and how much time managers spend dealing with government regulations. We focus on objective rather than subjective questions for two reasons. First, it is easier to quantify the difference between firms in SEZs and the rest of the country when using objective questions. Second, things other than corruption and regulation might affect answers to subjective questions about them.18

The question on corruption reads:

We’ve heard that establishments are sometimes required to make gifts or informal payments to public officials to “get things done” with regard to customs, taxes, licenses, regulations, services etc. On average, what percent of total annual sales, or estimated total annual value, do establishments like this one pay in informal payments or gifts to public officials for this purpose?

(World Bank, 2007)

We use this question to make a dummy with a value ‘one’ if the manager reported firms must pay bribes. We focus on whether the manager says firms must pay bribes rather than on how much firms need to pay because earlier studies have found there are problems with the amounts they report. Managers can answer the question in local currency or as a percentage of sales. Because the managers also report sales, we can calculate bribes either as a percentage of sales or in local currency for all firms. Although it should not matter how the manager answers the question, it does (Clarke, 2011, 2026; Malomo, 2013). Firms that report bribes as a percentage of sales report paying between four and fifteen times more than firms that report bribes in monetary terms. Furthermore, this difference is not because of either observable or unobservable differences between firms that report bribes as a percentage of sales and firms that report bribes in local currency (Clarke, 2011, 2026). This suggests either managers who report bribes as a percentage of sales overreport them or managers who report bribes in local currency underreport them. Because we cannot compare the answers of managers who report in different ways, we focus on whether the firm paid a bribe.

Another important observation is the question asked what the manager thinks other firms do rather than what the firm does. The reason the survey does this is that it allows managers to report bribes without admitting they have done anything illegal. Although there are valid questions about how firms respond to indirect questions, we will assume managers answer them thinking about their own firm.19 We can justify this in three ways. First, managers might recognize the survey asks the question indirectly to protect them and that the interviewer really wants to know what they do—not what they think their competitors do (Johnson et al., 2002). Second, even if they do not recognize this, managers who pay bribes might believe others also do the same. This might be reasonable—people believe others act and think like they do even when others act and think differently (Ross et al., 1977).20 Third, it is convenient to do this for expositional reasons. It is less clumsy to write “firms in SEZs are less likely to report paying bribes” than to write “firms in SEZs believe firms like theirs are less likely to pay bribes.”

Zone firms were less likely to pay bribes than firms outside the zones. Whereas only 20 percent of firms in the SEZs reported paying bribes, about 30 percent of other firms did.

The other dependent variable looks at how much time senior managers spend dealing with regulations, inspections, and other legal requirements. The question reads:

In a typical week over the last year, what percentage of total senior management’s time was spent on dealing with requirements imposed by government regulations? [By senior management I mean managers, directors, and officers above direct supervisors of production or sales workers. Some examples of government regulations are taxes, customs, labor regulations, licensing and registration, including dealings with officials and completing forms]

(World Bank, 2007)

Managers could spend all their time or no time dealing with regulation. Because of this, the amount is censored below at zero percent and above at 100 percent. In practice, few managers reported spending all their time dealing with government regulations (less than one percent), while many reported spending no time dealing with them (38 percent). Because of the censoring, we estimate the model as a two-sided tobit model. This implicitly assumes the error term is normally distributed.

Firms inside and outside the zones spend similar time dealing with government regulations. On average, managers of zone firms said they spent 6.2 percent of their time dealing with regulations—only slightly lower than other firms’ managers (6.3 percent).

3. Results

This section presents the econometric models we will use in this paper.

3.1. Econometric Model

The first model asks whether firms in special economic zones are less likely to report paying bribes than other firms (Hypothesis 2). The second asks whether firms in the zones report spending less time dealing with government regulation than other firms (Hypothesis 1).

To see whether SEZ firms are less likely to pay bribes, we assume the firm’s propensity to pay depends on whether it is in a zone and on other firm characteristics:

$$\mathrm{Propensity}\mathrm{to}\mathrm{pay}{\mathrm{bribes}}_{ijkt}=\alpha +\beta {SEZ}_{ijkt}+\gamma X_{ijkt}+{\lambda }_{jt}+{\varsigma }_k+{\epsilon }_{ij}$$

(1)

We do not observe the manager’s propensity to pay bribes and thus cannot estimate Equation (1) directly. Instead, we only see whether they said firms like theirs pay bribes. As discussed in the previous section, we assume managers who pay bribes will be more likely to answer ‘yes’ when asked whether firms like theirs pay bribes. We assume the error term, ɛ_ij, has a normal distribution and thus estimate the model, using maximum likelihood estimation, as a probit model. In the models that use instrumental variables, we estimate the models using maximum likelihood estimation and that take into account that the bribe variable and the zone variables are dummies.

We estimate the models as unweighted regressions. We do this for two main reasons. First, unweighted maximum likelihood estimators are often efficient with exogenous stratification (Wooldridge, 2001). Second, the weights tend to increase the importance of large countries when using Enterprise Survey data.21 This is a problem in our models because the cross-country sample becomes excessively dominated by a single country, India, when using weights.22

We use clustered standard errors to allow for the error terms, ${\epsilon }_{ij}$, to be correlated for firms in the same sampling strata. The sampling strata are based on firm size, sector of activity, and geographical location within countries (World Bank, 2022). We also use clustered standard errors when testing other hypotheses. This tends to increase the size of the standard errors, and reduce t-statistics, relative to using unclustered standard errors.

The dummy showing whether firm $i$ in country $j$ and industry $k$ at time $t$ operates in an SEZ (${SEZ}_{ijkt})$ interests us most. We code the dummy as 1 for SEZ firms and 0 otherwise. The post-2009 surveys asked whether the zone was an EPZ or a different industrial zone. For these surveys, we can therefore include two dummies indicating the type of zone. Because the early surveys did not collect information on zone type, including the two dummies reduces sample size. If the firms in the zones report less corruption, the dummy’s coefficient will be negative.

As well as the SEZ dummy, the regression includes several controls ($X_{ijkt})$. These include three dummies representing whether the firm has foreign owners, whether it exports, and whether the government partly owns it.23 Second, the regression also controls for the firm’s age and size.24 Third, the model also includes 34 industry dummies (${\varsigma }_k)$ at the four-figure ISIC 3.1 level to control for differences in regulation across industries. If firms in some industries meet with government officials more often, they might face more frequent demands for bribes. Finally, it includes country-year dummies ${(\lambda }_{jt})$ to control for differences between countries and over time that might affect the likelihood firms pay bribes.25 For example, firms might be more likely to pay bribes when a country has worse institutions, less effective courts, or where civil service pay is low. Because the regressions include country-year dummies, we can interpret the results as comparing SEZ firms with other firms in the same country in the same year.

To see whether firms in SEZs report a lower burden of regulation than firms outside them, we also run the following regression:

$$\mathrm{Percent}\mathrm{of}\mathrm{time}\mathrm{spent}\mathrm{dealing}\mathrm{with}{\mathrm{regulations}}_{ijkt}=\alpha +\beta {SEZ}_{ijkt}+\gamma X_{ijkt}+{\lambda }_{jt}+{\varsigma }_k+{\epsilon }_{ij}$$

(2)

The dependent variable measures how much time senior managers spend dealing with regulations, inspections, and other legal requirements. The variable is greater than or equal to 0 percent—some managers spend no time dealing with government requirements—and less than or equal to 100 percent, although few managers spend all their time dealing with regulation. We therefore estimate the model, using maximum likelihood estimation, as a two-sided tobit model, which assumes the error, ${\epsilon }_{ij}$, has a normal distribution. In the models that use instrumental variables, we estimate the models using maximum likelihood estimation and that take into account that the regulation variable is truncated and the zone variables are dummies. As before, we allow the standard errors to be correlated within strata.

The variable that most interests us is the SEZ dummy. If zone firms spend less time dealing with regulations, the dummy’s coefficient will be negative. If zone firms spend more time dealing with regulation—perhaps because SEZ firms must file extra paperwork related to export or labor requirements—it will be positive.

The regression also includes the firm-level controls and country-year and industry dummies included in the previous regression. Because the country-year dummies control for differences between countries for specific years, we can interpret the SEZ dummy as showing the difference in the time that SEZ and non-SEZ firms report spending on regulation.

3.2. Main Results

Table 2 presents results from the regressions for whether the firm paid a bribe and for the percent of time management spent dealing with regulation.

Table 3. Marginal differences between firms inside and outside SEZs.

	Probability That Firm Has Paid Bribe	% of Time Dealing with Regulations
Firms in EPZs	12.7%	6.6
Firms in other industrial zones	13.8%	8.1
Firms outside of zones	19.5%	9.1
Difference between EPZ and non-SEZ firms	−6.8%	−2.5
Difference between firms in industrial zones and non-SEZ firms	−5.8%	−1.0
Firms in SEZs	23.8%	7.7
Firms outside of zones	27.1%	8.4
Difference between SEZ and non-SEZ firms	−3.3%	−0.7

Source: Author’s calculations based upon data from the World Bank Enterprise Survey data. Note: Levels are calculated by calculating the probability that the firm reports paying a bribe or the amount of time it reports dealing with regulation for each observation, assuming each firm is in an EPZ, then assuming each firm is in another industrial zone, and then assuming that each firm is in neither type of zone. The probabilities for each firm are then averaged over all observations.

shows the marginal differences between firms located in special economic zones and other firms.

Likelihood of paying a bribe. Consistent with the first hypothesis, the SEZ dummy’s coefficient is negative and statistically significant (see column 1). This suggests an association between location in an SEZ and corruption: firms in SEZs are less likely to report paying bribes than other firms after controlling for observable firm and country characteristics.

When we separate the zones into export processing and other industrial zones, the coefficients on both dummies are negative and significant (see column 2). This suggests firms in both EPZs and other industrial zones are less likely to report paying bribes than other firms. The coefficient is larger for EPZ firms than for firms in other industrial zones. The difference between the two, however, is not statistically significant (chi-squared [1] = 0.56, p-value = 0.45).

The differences between reported bribes for firms inside and outside of zones are large. After controlling for size, age, export status, ownership, sector, and country, the average likelihood that a firm inside an EPZ reported paying a bribe is 12.7 percent. For firms in other industrial zones, it is 13.8 percent, and for firms outside the zones, it is 19.5 percent.26 The six-percentage point difference means non-zone firms were close to 50 percent more likely to report paying bribes.

The difference is smaller—although still significant—in the regression with a single SEZ dummy. The likelihood the average SEZ firm reported paying a bribe is 23.8 percent, while the average likelihood for firms outside the zones is 27.1 percent—a 3.3 percentage point difference. The weaker results for the single dummy could be due to sample differences—the earliest surveys did not ask about the type of zone. We explore this further in the robustness checks.

Although these results would be consistent with the idea that locating in an SEZ lowers the burden that corruption imposes on firms, this is not the only possible explanation for the difference. The differences could be due to unobserved differences between firms that locate inside and outside of zones. Although the regressions control for some firm level differences—age, size, export status, sector, country, and ownership—other unobserved firm-level characteristics could also be important.

Time spent dealing with regulation. Consistent with the second hypothesis, managers of SEZ firms report spending less time dealing with regulations than managers of firms outside the zones after controlling for observable differences. The SEZ dummy’s coefficient and the EPZ and other zone dummies’ coefficients are negative and significant. In contrast to bribes, however, firms in other industrial zones report spending significantly more time dealing with regulations than EPZ firms do (chi-squared [1] = 5.6, p-value = 0.02).

Although the difference between SEZ firms and non-zone firms is significant, it is smaller than the difference for paying bribes—especially for firms in other industrial zones. After controlling for observable differences, managers in zones spend, on average, about 7.7 percent of their time dealing with regulations, while managers at firms outside of the zones report spending 8.4 percent of their time. The difference between firms in EPZs and other firms is larger. The manager of the average EPZ firm spends 6.6 percent of their time dealing with regulations, while managers of firms in other zones and firms outside zones spend 8.1 percent and 9.1 percent of their time doing the same.

3.3. Results for Subgroups

In this sub-section, we look at differences between zone and non-zone firms in different groups of countries. We run separate regressions by region, then by income, and then by how corrupt the country is.

Table 4. Results for subgroups and additional robustness checks.

Column	(1)	(2)	(3)	(4)	(5)	(6)
Dependent Variable	Corruption			Regulation
Coefficient on:	SEZ	EPZ	Other	SEZ	EPZ	Other
All	−0.130 ***	−0.312 ***	−0.255 ***	−1.381 ***	−6.261 ***	−2.385 ***
By Region
Africa	0.037	0.008	−0.170 **	1.503 **	0.576	3.454 **
South Asia	−0.267 ***	−0.554 ***	−0.315 ***	−3.364 ***	−9.366 ***	−4.006 ***
By Income Level
Low income	0.023	−0.223 **	−0.153	0.350	−2.303	0.835
Middle income	−0.219 ***	−0.340 ***	−0.268 ***	−3.259 ***	−8.254 ***	−3.429 ***
By income and region
Low-inc. Africa	0.096	0.023	0.107	0.151	−2.931	−0.670
Low-inc. Asia	−0.067	−0.419 ***	−0.287 ***	0.734	−0.366	2.172
Middle-inc. Africa	−0.049	−0.033	−0.254 ***	3.121 **	3.943	4.871 **
Middle-inc. Asia	−0.344 ***	−0.594 ***	−0.314 ***	−6.264 ***	−13.569 ***	−5.507 ***
By corruption
High corruption	−0.072	−0.239 ***	−0.224 ***	1.106 *	−0.969	2.274 *
Low corruption	−0.176 ***	−0.422 ***	−0.277 ***	−3.513 ***	−10.994 ***	−4.708 ***
2SLS
Endogenous zone	−0.610 ***	−0.541 ***	−0.560 ***	−5.963 ***	−9.850 ***	−6.889 ***
Alt. corruption measure
Alt. measure	−0.003	−0.206 ***	−0.046
Sample
Small sample	−0.265 ***	−0.312 ***	−0.255 ***	−3.056 ***	−6.261 ***	−2.385 ***

Source: Author’s calculations based upon data from the World Bank Enterprise Survey data. Note: Table reports coefficients from regressions like those in Table 2. The coefficient in column (1) corresponds to the regression in column (1) of Table 2, the coefficients in columns (2) and (3) correspond to column (2) of Table 2, the coefficient in column (4) corresponds to the regression in column (3) of Table 2, and the coefficients in columns (5) and (6) correspond to the regression in column (4) of Table 2. All regressions include control variables from Table 2, including country-year and sector dummies (ISIC 4-figure). The bribe variable is a dummy, and so the model is a probit model. The regulation variable is censored at 0 and 100 percent, and so the model is a tobit model. Standard errors are clustered at the survey strata level. ***, **, * Statistically significant at 1%, 5%, and 10% significance levels.

summarizes the results from these additional regressions, focusing on the coefficients on the zone dummies.

Results from Africa. We first run separate models for Africa and Asia (see Table 4). The sample contains more African than Asian countries (39 compared with 9), but fewer African observations (8318 compared with 12,154).

The results are weaker for Africa than for the whole sample. In Africa, SEZ firms are no less likely to report paying bribes than non-zone firms. When the model includes two dummies—one for EPZs and one for other industrial zones—the EPZ dummy’s coefficient is insignificant. The other dummy’s coefficient, however, is negative and significant.

The results for regulation are also weaker. Firms in African SEZs reported spending more, not less, time dealing with regulations than firms outside the zones. Senior managers in firms in the zones spend, on average, about 11.6 percent of their time dealing with government regulation, while managers at firms outside the zones spend, on average, about 10.7 percent. When we split the zones into EPZs and other zones, firms in other zones, but not firms in EPZs, report spending more time dealing with regulation. In summary, firms in African SEZs do not report spending less time dealing with regulations than firms outside the zones and firms in EPZs are no less likely to report paying bribes.

Results for South Asia. In contrast to Africa, firms in South Asian SEZs report spending less time dealing with regulations and are less likely to report paying bribes than firms outside the zones (see Table 4). The differences are particularly large when comparing EPZ firms with non-EPZ firms.27 The likelihood the average EPZ firm reported paying a bribe is 7 percent, compared with 11 percent for other zone firms, and 17 percent for firms outside the zones. Similarly, the managers of the average SEZ firm spent 3.7 percent of their time dealing with government regulations. By comparison, the managers of the average firm in other zones spent 5.4 percent of their time, and the managers of the average firm outside of the zones spent 7.0 percent. In summary, and in contrast to Africa, firms in South Asian SEZs—and especially in South Asian EPZs—reported spending less time dealing with regulation and were less likely to pay bribes than South Asian firms outside the zones.

Results by income. The results suggest the differences between firms inside and outside SEZs are larger in South Asia than in Africa. The South Asian sample, however, differs from the African sample in several ways. One way is that most African firms are located in low-income countries (29 of 41 surveys), whereas the South Asian sample is more mixed (5 of 9 surveys are from low-income countries).28 We, therefore, rerun the regressions by income class and by income class and region.

SEZ firms in middle-income economies were less likely to report paying bribes and spent less time dealing with government regulations than similar firms outside the zones (see Table 4). The differences between firms inside and outside the zones are large. The average likelihood that an SEZ firm reports paying a bribe is 12.8 percent. The average likelihood for non-zone firms is 17.5 percent. Similarly, managers of SEZ firms spend 7.3 percent of their time dealing with regulations, compared with 8.7 percent for non-zone firms. The differences are also significant when we compare firms in EPZs and other industrial zones with non-zone firms separately.

The results for low-income countries are more mixed. When we do not distinguish between EPZs and other zones, SEZ firms and other firms were equally likely to pay bribes and spent similar time dealing with regulations. The only significant difference was between EPZ firms and other firms.

Breakdown by income class and region. As a next exercise, we divide the sample into four groups: low-income African countries, low-income Asian countries, middle-income African countries, and middle-income Asian countries. Doing this makes the samples smaller and might therefore make it harder to find robust results, especially when we include the two dummies.

For middle-income Asian countries, we find a large difference between firms inside and outside SEZs. Firms inside SEZs were much less likely to report paying bribes and spent much less time dealing with government regulations than other firms. The estimated likelihood the average SEZ firm would pay a bribe was 8.5 percent, compared with 14.5 percent for non-SEZ firms. Similarly, the average SEZ firm’s manager spent 5.3 percent of their time dealing with regulation, compared with 7.7 percent for non-zone firms. Results are similar in the smaller sample with two dummies.

For low-income countries in Africa, the differences were all statistically insignificant. This was true for SEZ and non-SEZ firms and for EPZ, other zone, and non-SEZ firms. For middle-income African countries and low-income Asian countries, the results are more mixed with only some significant differences between zone and non-zone firms. Moreover, firms in SEZs report spending more time dealing with regulation than non-zone firms in middle-income African countries.

In summary, the results are strongest for middle-income countries in South Asia. Firms in SEZs spent less time dealing with regulation and were less likely to pay bribes than firms outside the zones. Some evidence suggests firms in low-income countries in Asia and middle-income countries in Africa were also less likely to pay bribes—but the differences were not always significant. Finally, firms in zones in low-income countries in Africa were no less likely to pay bribes and spent no less time dealing with regulation than firms outside the zones.

High- and low-corruption countries. Rather than splitting the sample into middle- and low-income countries, we next split it by how corrupt the country is. Ranking countries based on the share of firms that paid bribes might be problematic. If the zones affect the burden of corruption and regulation, splitting the sample based on survey data might introduce endogeneity. We, therefore, use the Worldwide Governance Indicators to split the sample (Kaufmann et al., 2009). Highly corrupt countries rank below the sample median on control of corruption, while less corrupt countries rank above it.29

Middle-income countries are less corrupt, on average, than low-income countries. Corruption was high in 9 of 21 middle-income countries and 24 of 37 low-income countries. The partial overlap between income and corruption could make it difficult to know whether dividing the sample by income or corruption is more important.

When we run separate analyses for more and less corrupt countries, the results suggest firms in SEZs were less likely to report paying bribes. The coefficients on the zone dummies are always negative and mostly significant. They are, however, larger in the less corrupt countries.

The zone dummies’ coefficients are also negative and significant in the less corrupt countries in the regulation regressions. This suggests firms inside the zones report paying lower bribes than firms outside the zones in less corrupt countries. In contrast, two of the coefficients are positive and statistically significant for more corrupt countries. This suggests, if anything, that zone firms report spending more time dealing with regulation than non-zone firms.

3.4. Additional Robustness Checks

Instrumental variable estimation. As discussed earlier, endogeneity is a concern in these regressions. Omitted firm-level variables might affect firms’ decisions to locate in the SEZs and affect whether they pay bribes. Although finding good instruments is difficult, a common way of dealing with this problem in firm-level analyses is to use cell averages as instruments (Amin, 2025).30 We use the strata, which are based on size, sector, and region, as cells to calculate the likelihood that firms in that strata are in SEZs.31 In this case, the argument is that if other firms that are in the same region and sector and about the same size are located in SEZs, then it is also more likely that the firm will also be located in an SEZ. That is, if other similar firms are located in an SEZ, it signals the availability of a suitable SEZ in that region.

When we do this, the coefficients remain statistically significant and negative in both regressions, suggesting that, on average, firms in zones report that bribes are less common and that regulation is more burdensome after using the cell average as an instrument (see Table 4). The point estimates of the coefficients are larger in absolute value in these regressions.

Although it is difficult to formally test for instrument strength in probit and tobit regressions, the cell averages are highly significant in the first-stage regressions. In the regressions including the SEZ dummy, the t-stats on the cell averages exceed 42. In the regression including the two separate dummies, the t-stats on the relevant cell average exceed 20.32

Hypothesis tests favor the results for the IV regressions over the results that assume exogeneity. For the regressions with the single SEZ dummy, we reject the null hypothesis that the variable is exogenous at a one-percent level in both the bribe and regulation regressions. For the regression with both the EPZ and other zone dummies, we reject the null hypothesis that the other zone dummy is exogenous at a 1 percent level in both regressions. For the EPZ dummy, we are unable to reject the null hypothesis at a 10 percent level in either regression.

Alternative measure of corruption. The main question on corruption asks managers whether they think “establishments like this one” pay bribes. The reason for the ambiguous phrasing was to allow managers to answer honestly without incriminating themselves.33 Because some managers might answer based on their own experience while others answer based on what they think other firms do, this approach might introduce measurement error in the dependent variable, leading to inconsistent parameter estimates and attenuation bias.34

The Enterprise Survey includes additional objective measures of corruption. Firms are asked about a series of transactions and asked whether government officials asked for bribes during these transactions.35 We construct a dummy that is set to one if the firm reports being asked for a bribe in any of the transactions.36

In contrast to the main question on corruption, these questions are asked directly about the firm’s experience. After being asked about the transaction, the respondent is asked whether an informal gift was expected or requested. For example, the question for construction permits is:

In reference to that application for a construction-related permit, was an informal gift or payment expected or requested?

(World Bank, 2007)

Using direct questions should reduce the ambiguity due to the other question’s indirect phrasing. Direct questions, however, might introduce their own bias by making respodents less willing to answer questions honestly.37

The results are weaker than for the indirect bribe measure that we use throughout most of the analysis. In particular, in the regression with the single SEZ dummy, the coefficient remains negative but becomes statistically insignificant (see Table 4). In the regression with the two dummies, the coefficient remains negative and statistically significant for firms in EPZs. In contrast, it remains negative but statistically insignificant for firms in other types of zones.

Restricting sample to smaller sample. As discussed earlier, the results for the smaller sample suggest the differences between firms and non-zone firms are larger than the results for the full sample. The larger differences, which are visible for both EPZ firms and firms in other industrial zones, could occur because the two samples include different countries. If differences between firms in SEZs and other firms are larger in some countries than others—and the earlier results suggest they are—we might see different results in the small and large samples. To see if this is the case, we rerun the regressions for the regression with a single dummy only on the smaller sample.

When we do this, the coefficient on the single SEZ dummy becomes larger than in the full sample in both the bribe and regulation regressions (see Table 4). In both cases, the single dummy’s coefficient is close to, but slightly larger than, the coefficient for other industrial zones. This suggests the stronger results for the separate dummies are due to sample differences. One possible reason is South Asian firms dominate the small sample; they make up three-quarters of the small sample, but only 60 percent of the full sample. The stronger results in the small sample might therefore reflect that differences in reported corruption and regulation are larger in South Asia than in Africa.

4. Discussion and Conclusions

Many developing countries have set up special economic zones to attract investment, create jobs, and increase exports. By offering appealing bundles of light regulation, low taxes, and high-quality infrastructure, governments hope to entice private and foreign firms to invest in the zones. Despite these efforts, some zones, especially in sub-Saharan Africa, have failed to attract much private investment (Farole, 2011a). The Foreign Investment Advisory Service (2008, p. 1) notes: “successes in East Asia and Latin America have been difficult to replicate, particularly in Africa, and many zones have failed.” Consistent with this, Otchia and Wiryawan (2025) find that industrial growth accelerated in Asian and Latin American countries after they introduced SEZ policies, but that growth did not accelerate in Africa.

This paper’s results suggest one reason for the zones’ limited success. Although we find an association between SEZ location and lighter regulation on average, we do not observe it everywhere. Firms in the South Asian zones are more likely to report lower corruption and regulation; SEZ firms reported paying fewer bribes and spending less time dealing with government bureaucracy than other South Asian firms. In contrast, firms in African SEZs spent more, not less, time dealing with bureaucracy and paid similar bribes to other African firms.

Although causation is difficult to assess—the association could be due to unobserved firm-level difference between zone and non-zone firms—these results would be consistent with the idea that the zones have not reduced these burdens. If the zones fail to reduce red tape and corruption, firms will only locate there when they receive generous tax breaks or subsidies. If so, this is disappointing. Successful zones might convince government officials and voters to demand similar national reforms (Hartwell, 2018).38 Stein (2008, p. 9), for example, writes: “The zone allows an experimental forum to develop habits that will lead to efficiencies that can be emulated elsewhere in the country while at the same time building up trust with foreign investors.” But only successful zones will inspire more comprehensive reforms.

This paper could be extended in several ways. One would be to collect similar data for East Asia and Latin America, where SEZs have attracted more investment than Africa and South Asia. We could then compare results for East Asia and Latin America with the results for Africa and South Asia.

It would also be useful to have more recent data for South Asia and Africa. As noted above, questions on SEZs were dropped from the African and South Asian Enterprise Surveys in the early-mid 2010s. It is possible that differences between zone and non-zone firms have become smaller—or larger—as these zones have become more entrenched.39

More recent data might also make it easier to assess causality. As noted earlier, the observed differences between zone and non-zone firms’ responses might be due to unobserved firm-level differences. If firm-level panel data were available, we could use more robust estimation techniques.40 The establishment of new zones or expansion of existing zones might mean that it would be possible to better establish causation using difference-in-difference estimation.

Knowing what benefits the zones offer would also be helpful. Firms in export processing zones reported less corruption and regulation than firms in other industrial zones. These differences might be due to EPZs and other zones providing different benefits. Unfortunately, the survey does not report what incentives the firms received. If we knew what benefits different zones offer, we could better understand how zones can attract foreign and private investment.