Doing Business and Firm Growth

Electricity Connections and Construction Permits

Master Thesis International Economics and Business Supervisor: Prof. R.C. Inklaar

Co-assessor: Prof. R. Ortega Argiles Author: Babette Vugs - S2227436

Abstract

I. Introduction

The understanding of why some countries grow faster than others has been a popular research subject for decades. Regulation plays a large role in answering this question, which makes it is important for governments to gain insight into the effects of their regulation. Economies with less excessive bureaucracy and more efficient regulation create an environment that stimulates innovation, creativity and human capital development, and in turn will improve living standards and prosperity. Economic growth is highly correlated with firm growth. This creates the aim of government policy to facilitate firm development through a stimulating business environment. This positive effect is confirmed by Djankov et al. (2006), who have investigated the overall relation between regulation and economic growth. They conclude that countries with ‘better’ and more business stimulating regulations grow faster. Based on an self developed index they state that doing business indicators positively affect economic growth between 1993 and 2002. Djankov et al. (2006); Hall and Jones (1999); Acemoglu et al. (2001), emphasise the importance of legal and political institutions as a main indicator of economic growth. These institutions can influence economic welfare prospects by creating a stable and stimulating set of government regulations. This legislation can be defined more specific, based on the World Bank doing business indicators. These indicators analyse how difficult it is for enterprises to start, maintain and close a firm within a country, based on perceived obstacles of established business legislation. Existing literature mostly discusses finance- and entry related constraints of doing business (i.e. Fisman and Svensson, 2007; Klapper et al., 2006; Davidsson and Henrekson, 2002; Fonseca et al., 2001). They overall conclude that an increase in the obstacle of getting credit-, taxation- or an increased barrier to entry negatively affects firm activity and in turn firm growth. This article will contribute to this field of research with a focus on two doing business indicators that are limited discussed in economic literature, which are the procedures for obtaining an electricity connection and construction permit. Besides the vital requirement of capital in the expansion process of firms, they also need to receive electric power and the permission to build. An fast and easy application process allows firms to quickly react to evolved opportunities of growth.

3  

intensity is calculated as the industry average electricity consumption (in TJ) per value added and capital-intensity as the amount of capital stock per value added (all expressed in US dollars). The duration of the application process, expressed in days, combined with the intensity ratios results in the tested interaction term. A negative beta of this interaction term states that more dependent firms experience more restriction to firm growth than less dependent firms, when the duration of the application process increases. This results in the following two tested hypotheses. First, whether more electricity-dependent firms will experience a higher restriction to firm growth when the obstacles of getting electricity increase, compared to less dependent firms. And secondly, if more capital-dependent firms will experience a higher restriction to firm growth when the obstacles of obtaining a construction permit becomes higher, compared to less capital intensive firms.

1.1 Short Overview Data, Method and Results

The required firm-level data is collected from the World Bank Enterprise Surveys (WBES). This data, collected with standardized surveys, gives an overview of the economy’s private sector and is highly reliable, while it is gathered by private impartial contractors. Surveys of countries with less than 1000 observations and were examined before the cut-off in 2008, due to a possible bias created by the financial crisis, are excluded from the sample. This results in 13 middle income countries, which still face problems regarding inefficient business regulation and high levels of bureaucracy. Because the selected sample only considers firms that actually have applied, it is important to correct for a possible selection bias with the use of the two-step Heckman model. This model measures if the selected observations, firms that have applied for an electricity connection or construction permit, have the same average characteristics as the total sample. In other words, if the sample can be characterised as a simple random sample. The results of the first step of the Heckman selection model shows a significant selection bias for all firm characteristics (except firm investment in the sample of building permits). In the electricity sample there are three negative lambda’s for firm size, -investment and –productivity, which indicate that a firm that has applied is on average smaller, invests less and is less productive than a random firm in the total sample. Firms in the selected sample of building permits are significantly larger and less productive than random firms in the total sample. These significant selection biases indicate the importance to correct the estimated regressions with lambda.

takes longer for larger firms, which is also supported by the singular variable of duration. The results for construction permits show a negative significant relation for firm productivity and is in turn the only characteristic that confirms the hypothesis. The negative coefficient of the interaction term (-0,37%) indicates a larger restriction in productivity for firms in capital intensive markets (-0,97%) when the duration becomes longer, compared to firms in less capital intensive industries (-0,40%). This shows a possible barrier to entry in more capital intensive industries, while the obtaining of an electricity connection is more crucial for firm growth in these industries. In order to draw general conclusions on firm growth it is important to look at the results of all three firm characteristics instead of only one. Furthermore, firm size as firm characteristic is sensitive to many other factors than included in the model, so does alone not indicate a strong conclusion. The overall conclusion is in turn that there is no strong relationship between firm growth and the duration increase of the application processes for an electricity connection and building permit. Several limitations of this conclusion have to be mentioned. The two largest problems regarding the method used are reverse causality and data limitations. The latter makes it impossible to examine the affect over time and only for a limited number of countries, because this information is not available.

Section II will discuss the theoretical background between regulation and firm growth. Section III will describe the measurement method to examine the hypotheses and section IV examines the required data for these models. The results will be interpreted in section V and section VI concludes and deals with limitations of the method used.

II. Literature Review

There are many articles written about the causes of economic growth and the consequences of government regulation. In order to sketch a theoretical background, several relations discussed in literature are summarized in this section. First, the relation between economic growth and firm growth is discussed in more depth. Secondly, regulation is defined more specific based on indicators of the business environment and the conclusion of this section will discuss the contribution and hypotheses of this article.

2.1 Economic Growth

5  

technological knowledge and specialize in more productive sectors will stimulate the economy to effectively reallocate resources across sectors, which is necessary to achieve long-term economic growth (Nelson and Pack, 1999). This can be seen in the ‘Asian Miracle’, where Korea, Taiwan, Singapore and Hong Kong have developed themselves from poor and underdeveloped economies towards modern developed economies in less than forty years. This development was mainly driven by policy that stimulated foreign learning, innovation and specialization reforms. Specialization in capital intensive sectors stimulates countries to shift labour away from low productive sectors, such as agriculture, towards more productive sectors, such as manufacturing. This reallocation of labour is necessary to increase overall productivity of the country. Another way to look at economic growth, besides the assimilation view, is the accumulation view, which is represented by Solow (1956). He states that long-term growth in income per capita is caused by growth in Total Factor Productivity (TFP), which is the amount of output that cannot be explained by the inputs used for production (Comin, 2006). This can be improved due to efficient accumulation of physical and human capital. Economic growth can thus be created by the improvement of TFP, which has institutions as an important determinant (Isaksson, 2007; Chanda and Dalgaard, 2003), while institutions affect the composition of an economy between agriculture and non-agriculture. Aggregate TFP difference within a country can mostly be explained by the level of efficient resource allocation across industries. If markets can be characterised by imperfect market conditions these welfare improvements, in terms of economic growth, do not occur. Inklaar et al. (2015), Hsieh and Klenow (2009) and Busso et al. (2012), all emphasise the fact that TFP will be negatively affected in the case of resource misallocation. Hskeih and Klenow (2009) state that when firms become more productive and grow, they need more inputs to facilitate this growth. In the case of misallocation, in their article caused by policy distortions, this unavailable growth of inputs causes firms to stay small and to invest less than necessary to sustain this firm growth. In short, economic growth can be stimulated with effective resource allocation across industries and TFP improvements, which should be facilitated by government policy.

2.1.1 Firm Growth

be expressed in employment and/or sales (Coad and Holzl, 2010). Firm productivity is used to estimate relative firm growth and represents more efficient use of inputs. The third measure is firm investment, which indicates the degree in which firms take advantage of evolving growth opportunities (Lan et al. 1995). These three firm characteristics are closely related and can therefore not be seen separately. Firm size and productivity, in both absolute terms and TFP, are positively related (Leung et al., 2008). Furthermore, investment in positive NPV projects in the long-run result in an increase productivity and in turn in firm size (Lang et al. 1996). This can also be argued the other way around; more established and productive firms are often larger and have therefore the ability to invest more on a yearly basis.

The firm characteristics are affected by a country’s policy-uncertainty. This is mainly translated into a decrease of firm investment in unstable business environments, which in turn decreases opportunities for growth (Rodrik, 1990). Firms located in less developed and more indebt countries first have to wait if policy adjustments sustain before investing in the enforcement of new policy. Rodrik (1990) shows that even small amounts of policy uncertainty (i.e. exchange rate changes or trade policy changes) can act as a major constraint of investment, because the risk of a decreasing firm performance is too high for uncertain policy reforms. Given the fact that investments are often necessary for reforms to succeed, a self-fulfilling prophecy is created; when many companies don’t expect the policy reform to be successful and restrain investment, the reform can end up being dropped. Kang et al. (2013) build upon these ideas and state that when economic policy and the corresponding costs of regulation become less stable, this translates into more uncertain costs of doing business. In this case, again, firms become more reserved about investment and tend to wait until more information becomes available about the regulation adjustments. They show a significant negative effect between higher policy uncertainty and firm’s investment for both short- and longer term. Furthermore, they find that the effect of new-based policy shocks is considerable larger than effects caused by federal expenditure policy shocks. They also check their results for different effects between firms from different sizes. Smaller firms experience a significant negative effect after the news of policy uncertainty, while larger firms no longer show a significant negative effect and are more able to cope with these shocks. This is supported by Beck et al. (2006) who state that a stable legal environment and well established property rights protection positively affects firm growth in terms of size. This overall shows the importance of a stable and stimulating business environment in order for firms to effectively grow.

7  

economic environment is stable. The second step to economic growth are structural improvements regarding efficient resource allocation across industries and TFP improvements, which government policy should stimulate. The ultimate goal of firm growth can be measured by three firm characteristics; firm size, -productivity and –investment. These characteristics are closely related with each other and combined give a good overview of established firm growth.

2.2 Regulation

The former part of this section emphasises the important role of governments to develop and improve their business environment in order to effectively stimulate economic- and in turn firm growth. So the effectiveness and quality of regulatory institutions determine how well markets function and in turn how the economy is influenced (Jalilian, 2006). This indicates the importance of institutions, which is also emphasised by North (1990), who state that economic performance of countries is highly dependent on the evolvement of institutions. He argues that institutions affect the cost functions of the economy, while they influence the costs associated with transaction and transformation (production). Higher costs and more inefficient levels of bureaucracy restrain economic development, while this decreases the ability of firms to quickly adapt and respond to risen opportunities of growth. It is important to make the broad definition of ‘government regulation’ more specific to be able to examine the effect of specific regulations on firm growth.

2.2.1 Regulation and firm characteristics

Schiffer and Weder (2001) have investigated whether there are differences in terms of perceived obstacles in doing business between different firm sizes. Their main finding is that smaller firms indeed face more obstacles in doing business than larger firms. In other words, they find a negative relationship between firm size and perceived obstacles. This difference is large when comparing large and small businesses, but the difference between medium and large enterprises is very small. Hampel Milagrosa et al. (2015) try to counter argue the conclusion of Schiffer and Weder (2001) and argue that networks and enterprise characteristics are vital in determining whether firms grow. Although this seems more important in some countries, most firms still indicate the business climate as the major constraint to SME development. Especially the costs of the procedures (money and time) regarding licensing, taxations and inspections were mentioned. Most small firms could not pay the costs for these procedures in advance, especially because they do not know whether their requests will be approved. On a broader industry level, USAID (2008) concludes that economies consisting of many small firms (less than 20 employees), experience more economic growth due to a more stimulating business environment than economies with larger firms. In short, smaller firms experience more restriction to firm growth due to business regulation. This also indicates that these firms will benefit the most from legal development towards more stimulating regulation.

2.2.2 Business Regulation

9  

Henrekson (2002), and Fonseca et al. (2001) confirm this conclusion and state that institutional barriers negatively affect firm growth, while higher start-up costs decrease the incentive for entrepreneurs to start for themselves. Consequences due to strict labour regulation is emphasised by Almeida and Carneiro (2008). They state that stricter enforcements of labour regulation constrain firm growth and leads to higher unemployment. They state that this negative influence is being caused by an increase of monitoring costs and mandatory payments for health- and safety regulations. This creates an incentive for firms to stay small, while in this case they often are less visible and can evade regulation reforms. Their conclusions also show that the enforcement of regulation leads to less informal employees and higher unemployment rates. In short, the business environment can be analysed with the use of business indicators. Financial obstacles of doing business are negatively related to firm growth as is shown in economic literature.

2.2.3 Electricity Connection and Construction Permits

Two less investigated indicators are respectively concentrated on the ease of getting an electricity connection and construction permit. Getting reliable and affordable electricity is an essential factor for a successful business environment, while it is a part the infrastructure network that is essential to attract and develop local and foreign businesses.

firm growth based on micro-level data have not yet been investigated. This article tries to fill this gap in existing literature.

The second indicator that is not explicitly covered in existing literature focuses on the ability to easily obtain a construction permit, which gives firms permission to build a new construction or expand their pre-existing structures. Rankings from the World Bank show that in Singapore it is the easiest to get a building permit, with an average 10 procedures that take on average 26 days. Afghanistan ranks on the last place with on average 11 procedures with an average total duration of 353 days. Examination of these procedures indicate problems with bureaucratic and inefficient steps in the application process. When it is easier and cheaper to get a construction permit, fewer firms will illegally expand their firms, which will result in greater public safety (World Bank, 2015). This is caused by the obligated step to submit a building plan and to hold technical consultation with several government authorities. Furthermore, when the permit is easily obtained this motivates firms to expand their business, which stimulates firm’s growth in absolute terms. This also affects the productivity level of the firm (Leung et al., 2008). Restriction to firm growth directly caused by the obstacle to obtain a construction permit is not yet investigated in existing literature, which makes it contributing to examine this relation based on micro-level data.

The simplified relation between duration of the application processes and firm growth has little economic relevance due to reverse causality difficulties. This problem is (partly) tackled by the inclusion of an interaction term between the application duration and industry dependency, on respectively an electricity connection and building permit. Rajan and Zingales (1996) use the same approach when estimating the relation between financial dependence and growth. They examine whether firms that are more dependent on external finance experience relatively higher growth in countries with more developed financial markets. This interaction makes it possible to estimate whether these more developed markets contribute to firm growth. This interaction term makes it possible to draw a more precise and relevant conclusion across industries. Firms that are more dependent on an electricity connection are more electricity-intensive, while firms in these industries cannot operate without these connections. Capital-intensive firms will be more dependent on construction permits, as firm expansion is more important in these industries.

11  

III. Methodology

3.1 Model Description

In order to effectively picture firm growth, the dependent variable is in threefold; firm size, -productivity and -investment. The combination of these three indicators is important, while they all have their shortcomings and are subject to many other factors than business regulation. Firm size can be calculated with the use of employment and sales. Employment is highly dependent on the type of products made, while labour intensity differs among industries. The beneficial factor of this measure is that it is not expressed in monetary terms and therefore does not need to be corrected for currency differences or inflation, which makes cross-country differences easy. The second way to measure firm size is based on sales. The downside of this measure is that it does not incorporate the actual value added by the firm, which gives a distorted view when the firm buys expensive inputs and does not add much to it before selling. The second measure of firm growth is productivity, which combines the two measures of size, as it expresses contributed sales per employee. This results in a more reliable measurement of firm growth. The third, and last measure of firm growth is based on firm investment, which can be calculated as investment per value of sales. This ratio is made so make sure that investment does not estimate the same as firm size.

These firm characteristics are measured for firm i, in manufacturing industry j, for country c in one point in time. As stated in the contribution to existing literature, it is important to distinguish between industries when examining the influence of doing business obstacles instead of only looking at the overall economy. A distinction between corporations that are more (or less) dependent on an electricity connection or construction permit makes it possible give a more explicit policy recommendation regarding firm growth. The interaction also (partly) tackles the problem of reverse causality. The interaction term requires measurement of industry intensities, expressed in ratios for electricity- and capital dependency. When these ratios are combined with obstacles data, expressed in the amount of days it takes to obtain an electricity connection or construction permit, regression coefficients for the interaction effects can be calculated. When beta of the interaction term is negative, this supports the hypothesis that more intensive sectors experience a larger restriction to firm growth than firms in less intensive industries. In order to limit biases in the model, it is important to correct for industry- and country-specific characteristics. Results can be affected by political- and regulation differences between countries, because the enforcement degree of regulation enforcement and corruption differ across countries (Schiffer and Weder, 2001). Country- and industry dummy variables are created, because it is impossible to incorporate all possible control variables.

potentially felt too discouraged to even start the application process or firms for which the application was denied. This creates a sample-selection bias (𝜆), because these firms also experience restrictions to firm growth due to business obstacles of electricity and construction permits. In order to correct for this selection bias, the Heckman selection model is used. This model is discussed more in depth in the next section.

The three dependent variables are explained in a model that includes the interaction term of application duration and industry intensity and these two variables separately. Lastly, indicator variables for industries and countries are included and lambda to correct for the possible selection bias. This results in the following two models:

Electricity Connection 1      𝐹𝑖𝑟𝑚  𝐺𝑟𝑜𝑤𝑡ℎ_! = 𝛼!+ 𝛽! 𝐸𝑙𝑒𝑐𝑡𝑟𝑖𝑐𝑖𝑡𝑦  𝐼𝑛𝑡𝑒𝑛𝑠𝑖𝑡𝑦!  ×  𝐸𝑙𝑒𝑐𝑡𝑟𝑖𝑐𝑖𝑡𝑦  𝑂𝑏𝑠𝑡𝑎𝑐𝑙𝑒! + 𝛽_!  𝐸𝑙𝑒𝑐𝑡𝑟𝑖𝑐𝑖𝑡𝑦  𝐼𝑛𝑡𝑒𝑛𝑠𝑖𝑡𝑦! + 𝛽!  𝐸𝑙𝑒𝑐𝑡𝑟𝑖𝑐𝑖𝑡𝑦  𝑂𝑏𝑠𝑡𝑎𝑐𝑙𝑒!+ 𝛽!𝐷!  𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦   +   𝛽!𝐷!  𝐶𝑜𝑢𝑛𝑡𝑟𝑦 + 𝜆 + 𝜀! Construction Permits 2      𝐹𝑖𝑟𝑚  𝐺𝑟𝑜𝑤𝑡ℎ! = 𝛼!+ 𝛽! 𝐶𝑎𝑝𝑖𝑡𝑎𝑙  𝐼𝑛𝑡𝑒𝑛𝑠𝑖𝑡𝑦!  ×  𝐵𝑢𝑖𝑙𝑑𝑖𝑛𝑔  𝑃𝑒𝑟𝑚𝑖𝑡  𝑂𝑏𝑠𝑡𝑎𝑐𝑙𝑒! + 𝛽!𝐶𝑎𝑝𝑖𝑡𝑎𝑙  𝐼𝑛𝑡𝑒𝑛𝑠𝑖𝑡𝑦! + 𝛽_!𝐵𝑢𝑖𝑙𝑑𝑖𝑛𝑔  𝑃𝑒𝑟𝑚𝑖𝑡  𝑂𝑏𝑠𝑡𝑎𝑐𝑙𝑒_!+ 𝛽_!𝐷_!𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦 +   𝛽_!𝐷_!  𝐶𝑜𝑢𝑛𝑡𝑟𝑦 + 𝜆 + 𝜀_! 3.2 Measure of Dependency

13  

favourable than an intensity solely based on the United States or Western Europe. The World Input-Output Database (WIOD) is used to construct electricity consumption- and capital stock data for OECD countries. The specifics of the calculations of these intensities are discussed below.

3.2.1 Electricity

The benchmark for electricity intensity is based on the year 2009, because this is the most recent year with all essential information available in the Environmental Accounts of the WIOD. Furthermore, the sample is based on country-surveys from 2010 onwards. The intensity is based on all 2009 OECD members (table 1 and 2 Appendix). Data was not available for Iceland, New Zealand, Norway and Switzerland, which evidently drop out. Also Mexico and Turkey drop out, because they are part of the country sample (the sample selection is discussed in section IV). The most significant and most influential economies (i.e. the US and Western Europe) are included in the benchmark.

Based on gross electrical use (in TJ) it becomes possible to estimate which industries demand most electricity. These total industry consumption figures are biased by industry size, while large industries reasonable demand more power than smaller industries. Electricity intensity is for this reason calculated as a ratio between input, electricity demand, and output, value added. Gross value added industry data (at current basic prices) is available in the Socio Economic Accounts of the WIOD. This paper will focus on manufacturing industries, identified through ISIC industry codes. The manufacturing sector contains ISIC codes 15 to 33. These codes make it possible to correctly link industry electricity consumption and value added. All data is expressed in monetary terms of the local currency. After correcting for the average 2009 exchange rate, the ratios can be averaged to construct the industry average electricity consumption per US dollar of value added. Japan is excluded from the average benchmark, because this extreme outlier distorted the average electricity intensity. The calculation and average of industry intensities are summarized in table 1 (and table 13 Appendix).

Table 1: Industry Electricity Intensity Ratio (base year 2009) Measured as electricity consumption (gross energy use in TJ) divided by the industry total value added (adjusted for US dollar based on the average exchange rates). Table displays an average (equally weighting) of intensities of OECD countries. Excluding Iceland, New Zealand, Norway, Switzerland, Mexico, Turkey and Japan. Industries are ranked by electricity intensity. Data: World Input Output Database; Environmental Accounts and Socio Economic Accounts, available for download online.

Electricity Intensity Ratio (Electricity Consumption TJ per US dollar Value Added)

Industry ISIC Electricity Intensity

Basic Metals and Fabricated Metal 27t28 0,373

Other Non-Metallic Mineral 26 0,408

Coke, Refined Petroleum and Nuclear Fuel 23 0,414

Chemicals and Chemical Products 24 0,540

Wood and Products of Wood and Cork 20 0,623

Food, Beverages and Tobacco 15t16 0,824

Pulp, Paper, Paper , Printing and Publishing 21t22 0,912

Textiles and Textile Products 17t18 0,953

Rubber and Plastics 25 1,197

Leather, Leather and Footwear 19 1,995

Electrical and Optical Equipment 30t33 2,552

Table 1 shows that Machinery (ISIC 29) and Electrical and Optical Equipment (ISIC 30t33) are the most electricity intensive sectors, while Basic Metals and Fabricated Metal (ISIC 27t28) are the least intensive sectors. The ratios indicate that in the least intensive industries, one US dollar value added only requires 0,373 TJ of electricity consumption, while in the most intensive industry this requires 4,841 TJ of electricity.

3.2.2 Construction Permits

The second hypothesis requires an intensity that characterises firms that are increasingly dependent on construction permits. Firms that want to expand their production capacity through company extension are often more capital intensive, while these extensions require large amounts of capital (Kumar et al, 1999). The intensity of capital is measured by the ratio of real fixed capital stock to value added for each industry. In other words, the amount of capital required to produce one dollar of value added. Data of capital stock is again based on the Socio Economic Accounts of the WIOD. Most characteristics of the capital intensity benchmark are the same as for the electricity intensity benchmark. The only adjustment is the base year of the benchmark, 2007, while capital stock information is not available for a large part of OECD countries in 2009. This, however, should not differ much within such a small time span, while the capital intensity of industries develops very slowly (Wolff, 1991). The industry capital intensity ratios are displayed in figure 2 (and figure 14 Appendix). Coke, Refined Petroleum and Nuclear Fuel (ISIC 23) is the most capital-intensive sector within most countries (2,647 US$ of capital stock per US$ value added), while Machinery (ISIC 29) is the least intensive sector within manufacturing (1,091 US$ of capital stock per US$ value added).

Capital Intensity Ratio (Capital Stock per US dollar Value Added)

Industry ISIC Electricity intensity

Machinery, Nec 29 1,091

Basic Metals and Fabricated Metal 27t28 1,258

Electrical and Optical Equipment 30t33 1,288

Rubber and Plastics 25 1,468

Wood and Products of Wood and Cork 20 1,600

Textiles and Textile Products 17t18 1,676

Pulp, Paper, Paper , Printing and Publishing 21t22 1,760

Other Non-Metallic Mineral 26 1,787

Food, Beverages and Tobacco 15t16 1,790

Chemicals and Chemical Products 24 1,830

Leather, Leather and Footwear 19 2,211

Coke, Refined Petroleum and Nuclear Fuel 23 2,647

15  

IV. Data

4.1 Data Source

Firm level data is collected from the World Bank Enterprise Surveys (WBES), which contains a broad range of economic data for 130.000 firms in 135 countries. This standardized survey gives an economy’s private sector overview of formal firms (it does not include informal businesses). It captures topics such as crime, finance, infrastructure, performance and workforce. This data is collected over the period between 2002 and 2014, gathered by different units of the World Bank and the timing of the surveys are country specific. The raw material and datasets are publicly available for researchers. The surveys contain country specific questions, however, the aggregated country database only includes general questions. This standardization makes the survey an ultimate device to examine cross-country comparisons concerning the indicators of the business environment. While many questions are related to the government, and in turn contain sensitive information, private contractors conduct the surveys on behalf of the World Bank. This results in a reliable set of firm data related to the business environment, suitable for micro-level studies. Generally the top manager or owner of the company answers the questions. The World Bank states in their methodology that ‘typically 1.200-1.800 interviews are conducted in larger economies, 360 interviews are conducted in medium-sized economies, and for smaller economies, 150 interviews took place’ (World Bank, 2010). This random selection of firms results in an overall picture of the private sector of a country. Cross-country comparisons to estimate the indictors’ influence are justified, while information on firm characteristics are documented at the same moment as the perspective of the firm’s business environment. Also the comparison of surveys from different base years does not create a threat when the range of base years is minimal. The total database contains of surveys from a variety of countries and base years. For some countries there are more surveys available, but this data is to small to consider panel data evaluation for the hypothesis stated above. For this reason a cross-country dataset is constructed.

4.2 Sample Selection

Mexico (2010), Nigeria (2014), Pakistan (2013), Peru (2010), Russia (2012), Turkey (2013) and the Ukraine (2013). These surveys combined result in a total sample of 32.376 firms. A cross-country comparison is possible while the range of base years, 2010-2014, is small enough to not bias the estimated results. The survey includes all industries, but this thesis focuses on firms within the manufacturing industries, this exclusion is based on the ISIC code within the survey. This leaves a total of 19.503 firms within the manufacturing sector. As discussed in the methodology section, only firms that applied for an electricity connection or construction permit within the last two years are included in the examined subsamples (based on #C3 and #G2 WBES: this number is not corrected for the amount of firms that actually estimated the duration of the application process). This results in 2.102 firms that applied for an electricity connection within the last two years and 1.865 firms that applied for a construction permit (table 3). This large drop out of firms emphasises the importance of the self-selection estimation measured by the Heckman two-step model discussed later in this section.

Table 3: Total sample description based on selected 13 WBES countries. The second column displays the amount of firms in the manufacturing sector (based on ISIC classification). The third and fourth column displays the amount of firms that respectively applied for an electricity connection and building permit.

Overall, firms are well scattered across countries and industries, which causes a good foundation to estimate the models. Most firms within this sample are located in India (6.490 firms) and the least in the Ukraine (718 firms). Furthermore, most firms operate within the food and beverages industry (2.758 firms) and just 10 firms operate within a part of the electrical and optical equipment industry. For this reason, the industries ISIC groups, identified in the Standard Industry List (OECD, 2015), are taken into account in the identification of the industry intensities. Most electricity applications in the last two years are done in Nigeria (262 firms) and most building permits in the emerging economy of India (367 firms).

Data requirements for the independent and dependent variables are discussed and summarized below. The last part of this section discusses the Heckman selection model in more detail.

Total Sample Description

Country Survey Year Manufacturing Firms Electricity Application Building Permit Application

17  

The first model requires data of the Doing Business indicator ‘getting electricity’. This indicator is part of the WBES. If firms have applied for an electricity connection in the past 2 years (#C2 WBES) they have to estimate the number of days it took to actually receive an electrical connection service (#C3 WBES). The model estimates if firm characteristics are negatively affected if it takes more days to get connected to the network and if more intensive firms experience relatively more restriction to firm growth if the duration of application process increases. The second model looks at building permits. This indicator is implemented as the number of days it takes to obtain a construction related permit (#G3 WBES) if they applied a permit in the last two years (#G2 WBES). Lastly, firms that did apply for an electricity connection or construction permit but could not estimate the duration of these processes, got denied, or were still in process of the application, are excluded from the examined subsample. This distribution is displayed in table 4.

Table 4: Distribution of respectively the duration of electricity connections and construction permits Column 1 shows the number of firms within the manufacturing industry that have applied and estimated the succeeded application process. Column 2, 3, 4 and 5 are expressed in number of days. The row ‘total’ does not expresses the sum of rows, but the characteristics of the total subsample.

The first subsample, the left part of table 4, displays the distribution of firms that have applied and estimated the duration of an electricity connection. In total, of the 2.102 firms that have applied for a connection in the last two years, 1.750 firms have also estimated the amount of days it took to get connected to the network and have actually obtained a connection. Most of these firms are located within Nigeria (239 firms) and Mexico (230 firms) and operate in the food- and textile industry (331 firms). On average it took 45,5 days to actually obtain an electricity connection, varying between 1 and 850 days. This wide spread in duration of the process can be explained due to possible enforcement differences of the government between firms and due to extreme outliers. The distribution also shows large differences across countries, while in Turkey and the Ukraine it takes on average 7,8 days and in Russia 109,4 days.

The second subsample, on the right part of table 4, displays the same distribution for the Duration of Electricity Connection Duration of Construction Permit

application of construction permits. In total 1.444 firms have applied for a construction permit and have estimated the duration of this completed process. Most of these firms were located in India (23%) and within the food and textiles industry. In Russia it took on average, again, the longest to get a permit, with an average of 208 days and Ukraine is on average the fastest with an average of 2,11 days. Two firms located in Russia and China reported the longest duration of 730 days. The overall average shows that it takes 66 days before the construction permit is allocated, with a standard deviation of 105,5 days.

In both models it can be seen that the standard deviation values are relatively high compared to the averages. This indicates the possibility that the variables are not normally distributed, which is confirmed by normality tests. For this reason, the obstacles expressed in number of days, are transformed to logs. This is also done for the intensity ratios calculated above. This results in a approximately normal distribution of the independent variables, which is necessary for the use of the Heckman selection model and the estimated regressions.

4.4 Dependent variables

19  

(3) 𝐹𝑖𝑟𝑚  𝑆𝑖𝑧𝑒!=

(  𝑃𝑒𝑟𝑚𝑎𝑛𝑒𝑛𝑡  𝐹𝑇  𝐸𝑚𝑝𝑙𝑜𝑦𝑚𝑒𝑛𝑡_!+ 𝑇𝑒𝑚𝑝𝑜𝑟𝑎𝑟𝑦  𝐹𝑇  𝐸𝑚𝑝𝑙𝑜𝑦𝑚𝑒𝑛𝑡_!)

(4)  𝐹𝑖𝑟𝑚  𝐼𝑛𝑣𝑒𝑠𝑡𝑚𝑒𝑛𝑡!=

!""#$%  !"#$%&'()*$  !"  !"#$%&'()_!!!""#$%  !"#!"#$%&'!  !"  !"#$  !"#  !"#$%#&'(_! !!"#$_!

(5) 𝐹𝑖𝑟𝑚  𝑃𝑟𝑜𝑑𝑢𝑐𝑡𝑖𝑣𝑖𝑡𝑦!=

𝑆𝑎𝑙𝑒𝑠!

(  𝑃𝑒𝑟𝑚𝑎𝑛𝑒𝑛𝑡  𝐹𝑇  𝐸𝑚𝑝𝑙𝑜𝑦𝑚𝑒𝑛𝑡!+ 𝑇𝑒𝑚𝑝𝑜𝑟𝑎𝑟𝑦  𝐹𝑇  𝐸𝑚𝑝𝑙𝑜𝑦𝑚𝑒𝑛𝑡!)

Firm characteristics are not available for all firms that have applied, because some firms have responded in the survey that they did not know the answer (reported as -9), did not want to answer (-8), or did not want to apply (-7). These missing observations should not numerical be included in the distribution and regressions. For this reason the regressions are done for firms for which that specific characteristic is available and observations containing the value of -9, -8 and -7 are dropped from the samples. This results in 6 smaller subsamples containing data of firms that have applied, succeeded and estimated the application processes and for which the necessary firm characteristics data is available. This is displayed in table 5.

4.5 Sample Description

Table 5: Firm characteristics of firms in respectively the electricity- and building permit sample The first column (obs.) displays the amount of firms for which all necessary information was available. The first row for each firm characteristic summarizes the resulting average distribution based on formulae (3)(4) and (5) stated in this article. The input for these formulas, for each firm characteristic, are listed under the first row of each characteristic. Firm size is expressed in number of workers. Firm investment and productivity are given in US dollars.

Also in the second model concentrating on construction permits, not for all firms are all firm characteristics available. Table 5 shows that there is relatively less information available about firms that applied for a construction permit than for an electricity connection. Based on 1.375 firms that have applied the average firm employs 349 fulltime workers in their corporation. Firms invest on average 4,9 million in land and buildings and 0,7 million US dollars in equipment, which results in an investment turnover of 0,95 per dollar of generated sales. This high investment rate can be explained by the fact that these firms have applied for a building permit, therefore many firms have a relatively high amount invested in construction and land, which evidently results in a higher average investment ratio. Firm productivity of an average firm is equal to $80.067.

When comparing these two samples it should be noticed that on average firms that apply for an electricity connection are smaller (206 employees) than firms that apply for a construction permit (349 employees). This can be empirically explained, because young, and often smaller, firms have to be connected to electricity, while established, and often larger, firms already have an electricity connection and are focussing on company expansion. The expenditures on investments are larger in the construction permit sample, because a construction permit requires a considerable amount of investment in land and buildings, which evidently increases the average overall investments. That the second sample reflects more established firms is also shown in firm sales, which is higher in the sample of building permits (37,6 mln US$) than in electricity (17,7 mln US$). These comparisons empirically show that these three firm characteristics are closely related and that the sample selection influences the average firm characteristics. This emphasises the importance of the Heckman model.

Finally, while the firm characteristics in both samples show a non-normal distribution, (i.e. high standard deviations and large spreads within each country) the dependent variables are also transposed into logs, which results in approximately normal distribution.

Firm Characteristics

Electricity Sample Construction Permits Sample

Variable Obs. Mean Std. Dev. Min. Max. Obs. Mean Std. Dev. Min. Max.

Firm Size 1670 205,964 807,036 0 21.000 1.375 348,730 1.118,524 3 21.000

FT Permanent 1670 188,028 784,254 0 21.000 1.375 317,031 1.075,042 3 21.000

FT Temporary 1670 17,935 95,850 0 3.000 1.375 31,699 133,824 0 3.000

Firm Investment 846 0,468 5,019 0 127,5 769 0,950 16,349 0 445

Equipment 846 2.126.138 37.700.000 0 1.090.000.000 769 4.910.871 46.600.000 0 1.090.000.000

Land & Buildings 846 309.246 2.037.411 0 39.600.000 769 761.917 4.828.022 0 94.800.000

Sales 846 22.500.000 87.200.000 350 1.600.000.000 769 59.500.000 212.000.000 105 3.060.000.000

Firm Productivity 1513 98.733 1.610.371 0 62.500.000 1.285 80.067 248.809 2 5.382.041

Sales 1513 17.700.000 89.800.000 5 1.600.000.000 1.285 37.630.000 166.000.000 105 3.060.000.000

FT Permanent 1513 185,378 783,641 2 21.000 1285 296,766 988,256 3 21.000

21  

When the two samples displayed in table 5 are split according to a threshold measure, based on the duration of the obstacle (expressed in days), it becomes possible to predict which signs the regression coefficients will show. Four dummy variables are created for each sample, which are displayed in table 6.

Dummies Threshold for Sample Examination

Dummies Electricity Obstacle Construction Permits Obstacle

D1 ≤10 days ≤10 days

D2 10<days≤30 10<days≤30 D3 30<days≤90 30<days≤90 D4 >90 days >90 days

Table 6: Dummies Threshold for Sample Examination. Based on duration of the application process.

Table 7 shows several different patterns. The hypothesis stated in section II would expect a negative relation between the increase in duration (D) and firm characteristics. However, the results show no clear linear relation between these two. Firm size peaks in respectively D4 and D2, firm investment in D1 and D4 and firm productivity in D2 and D3. Overall, firms in group D4 are not extremely smaller, have not invested relatively less and experience lower productivity than firms in group D1. This indicates that there is a large possibility that the hypotheses will not be supported.

Table 7: Firm Characteristics per Dummy Group for respectively the electricity- and construction permit sample. The first column (obs.) shows the number of firms for which all necessary information was available. The first row of each firm characteristic shows the distribution

of all firms for which information was available, while the rest of the rows summarize the distributions per classified group. Firm size is displayed in number of employees, while the rest is in US dollars. Calculations of the firm characteristics are based on formulae (3)(4) and (5) stated in this article.

4.6 Heckman selection

As stated before, there is a large possibility that the sample is affected by a selection bias, while only firms that actually have applied for an electricity connection or construction permit are taken into account in the estimated model. For this reason, it is important to test if this bias is

Firm Characteristics according to Dummy Groups Firm Characteristics of Electricity Sample

Firm Characteristics of Construction Permits Sample

Variable Obs. Mean Std. Dev. Min. Max. Obs. Mean Std. Dev. Min. Max.

significant and, if this is the case, adjust the regressions for this bias. This can be done in a simple and empirical way by splitting the total sample between firms that did and did not apply (figure 8).

Table 8: Firm Characteristics for firms that did or did not apply for respectively an electricity- or construction permit. The first column (obs.) shows the number of firms for which all necessary information was available. The first row of each firm characteristic shows the distribution of all firms for which information was available, while the rest of the rows summarize the distributions of firms that did or did not apply. Firm size is displayed in number of employees, while the rest is in US dollars. Calculations of the firm characteristics are based on formulae (3)(4) and (5) stated in this article.

The distributions in table 8 firstly show that firms, in both samples, that have applied for respectively an electricity connection or construction permit are on average larger than firms that have not applied. In the electricity sample there is also a large difference in the relative amount invested, which is much higher for firms that did not apply (2,126) compared to firms that have applied (0,432). This is also the case for firm productivity in this sample (104,6 mln. versus 78,9 mln. US dollars). These large differences are not present in the sample of construction permits. These results suggest that there is a selection bias in firm size for both selected samples and in firm investment and –productivity in the electricity sample, while firms that are selected (have applied) are already larger, and in the electricity sample invest less and are more productive, than an average firm in the total sample. This indicates the importance precisely estimate this bias with the use of the Heckman model.

The Heckman correction, which is a two-step statistical procedure, corrects for non-random selection in the sample. In the first stage the probability of selection is estimated based on a probit model founded on economic theory. The probit model looks like:

𝑃𝑟𝑜𝑏 𝐷 = 1 𝑍 = 𝜙 𝑍𝛾 .

D indicates if the firm has applied for respectively an electricity connection or construction permit (D=1) or not (D=0) based on the total sample of 19.503 firms. Z is the vector of variables that possibly explain why certain firms apply and others do not and 𝛾 represents the unknown parameters. Lastly, 𝜙 is the distribution function. This regression estimates the probability of application for each firm in the sample. Four variables (Z) that are similar for both models are taken into account to estimate this probability for application. The first and second variables are respectively the amount of years that the enterprise has experience within the country (survey year minus funded year in country #B5) and the

Electricity Sample (Applied vs Not Applied)

Construction Permits Sample (Applied vs Not Applied) Obs. Mean Std.dev. Min. Max. Obs. Mean Std.dev. Min. Max.

23  

sector experience of the top manager (#A7). Dierickx and Cool (1989) state that companies build a comparative advantage within a host country when they gather experience and knowledge over time. This country-specific knowledge drives performance improvements (Barkema and Vermeulen, 1997; Schiffer and Weder, 2001), while more knowledge about institutions can help firms to deal with bureaucratic procedures of the application process. This country- and sector knowledge advantage will help and encourage firms to start the application process. The third variable incorporates the firm’s perception of the government corruption level. The WBES asks firms if the perceive the court system as fair, impartial and uncorrupted based on a Likert scale (#H7a WBES). This scale was classified as strongly disagree (1), tend to disagree (2), tend to agree (3) and strongly agree (4), higher values indicate that the firm perceived the government as more fair and uncorrupted. Firms that think of the government as very corrupt will less likely begin the application procedure, while they might be to discouraged from the start.The last variable incorporates if the firm is part of a larger firm (#A7). This can stimulate firms to begin the application process, while it is financially and legally backed-up by the larger head corporation.

This results in the following model for both obstacles:

(6) 𝐴𝑝𝑝𝑙𝑖𝑐𝑎𝑡𝑖𝑜𝑛_!

= 𝛼!+ 𝛽!  𝐶𝑜𝑢𝑛𝑡𝑟𝑦  𝐸𝑥𝑝𝑒𝑟𝑖𝑒𝑛𝑐𝑒!+ 𝛽!  𝑆𝑒𝑐𝑡𝑜𝑟  𝐸𝑥𝑝𝑒𝑟𝑖𝑒𝑛𝑐𝑒!+ 𝛽!  𝐶𝑜𝑟𝑟𝑢𝑝𝑡𝑖𝑜𝑛  ! + 𝛽!  𝐿𝑎𝑟𝑔𝑒𝑟  𝐸𝑛𝑡𝑒𝑟𝑝𝑟𝑖𝑠𝑒!  

The second stage of the Heckman model corrects the original model for the possible self-selection found in step one expressed in 𝜆. If lambda is significantly different from zero, this indicates a selection bias. In this case, the regression includes 𝜆 to correct for this bias, while the observations in the regression have all applied (D=1). In order to use the Heckman model it is important to note that the variables that possibly explain the decision to apply (Z) should not all statistically correlate with the dependent variable. In other words, at least one variable that significantly explains the selection of the firm (D) should not also statistically significant explain the variation in the dependent variable. These two regressions have to be compared to judge if 𝜆 is justified in the overall regression model.

V. Results

5. 1 Heckman Selection Model

In order to use the Heckman model there should at least be one variable that explains whether firms apply or not, which may not significantly explains the dependent variable. These two regressions are compared in table 9. The left regression explains the dependent variable with the selected variables explained in section III and the right regression displays the first step of the Heckman two-step selection model and in turn explains if firms apply for an electricity connection.

Electricity Connection

Firm Size Firm Investment Firm Productivity

Firm Size Application Firm Inv. Application Firm Prod. Application Country Experience 0,000** 0,000 0,000 -0,001* 0,000*** 0,000 Sector Experience 0,010*** 0,009*** -0,003 0,005*** 0,016*** 0,009*** Part of larger firm -1,074*** -0,162 0,130 -0,197*** -0,387*** -0,15*** Honest Government 0,033*** -0,009*** -0,014 0,009 0,008 -0,013** Constant 5,452*** -1,160*** -3,532*** -0,579*** 10,117*** -1,173*** Observations 14.969 14.696 3.367 3.485 13.329 13.061 R-squared 0,107 0,001 0,023 Adj. R-squared 0,106 0,000 0,023 Lambda -0,148 -0,579 1,453 P Heckman 0,897 0,000 0,000

Table 9: Electricity obstacle; examination for the use of the Heckman model. The first column for each firm characteristic displays the regression with the firm characteristic as the dependent variable. The second column displays the first step of the Heckman model with dependent variable whether firms have applied for an electricity connection. Note: *** p<0,01; ** p<0,05; * p<0,1.

25  

would result in a sample with on average more younger firms, which translates in less fulltime employees and less investment in absolute terms due to less internal funds.

The same procedure is done for the obstacle construction permits (Table 10).

Construction Permit

Firm Size Firm Investment Firm Productivity

Firm Size Application Firm Inv. Application Firm Prod. Application Country Experience 0,000** 0,000* 0,000 0,000 0,000*** 0,000 Sector Experience 0,010*** 0,010*** -0,003 0,012*** 0,016*** 0,009*** Part of larger firm -1,074*** -0,275*** 0,13 -0,409** -0,387*** -0,252 Honest Government 0,033*** 0,012* -0,014 0,043*** 0,008 0,011* Constant 5,452*** -1,088*** -3,532*** -0,479*** 10,117*** -1,101 Observations 14.969 14.643 3.367 3.420 13.329 13.061 R-squared 0,1066 0,001 0,023 Adj. R-squared 0,106 0,000 0,023 Lambda 2,005 0,322 -1,453 P-value Heckman 0,000 0,288 0,000

Table 10: Construction Permit obstacle; examination for the use of the Heckman model. The first column for each firm characteristic displays the regression with the firm characteristic as the dependent variable. The second column displays the first step of the Heckman model with dependent variable whether firms have applied for a construction permit. Note: *** p<0,01; ** p<0,05; * p<0,1.

If we allow a significance level of 10%, the same conclusion emerges for the sample of construction permits as was found for electricity connections. Two sample specific variables are therefore included in the dependent variable firm size; whether the land is owned by the firm itself and how much of the land is being used. The expectation is that when the land is owned by the firm itself or if there is space left to build on, there will sooner be an application to build on the firms’ land (table 16 Appendix). Again, this is not an ideal set of explanatory variables, but it makes it possible to compare and interpret the founded results in the adjusted regression model. The selection bias is significant for firm size and –productivity at a 1% significance level. The signs of lambda for firm size- and productivity, respectively 2,189 and -1,453, show in turn slightly different results compared to the sample of electricity connections. It can be argued that well-established and therefore larger firms apply for construction permits, while the firm needs sufficient money to pay for the applied expansion. This can in turn be the reason why firms that have applied are on average less productive than a random firm in the sample.

construction permit. This can be explained by two sides of corruption. On the one hand corruption demotivates firms when they are hindered by the bureaucratic procedures and informal payments. But some firms can be positively affected by corruption, because informal pays can speed up the processes needed for permits. In other words, illegal payments can create an incentive and tool to tackle bureaucratic delays (Ayadm and Hayalogu, 2014).

In the regression results discussed below, the adjusted regressions for the original and the new selection of variables are included. The second selection makes it possible to interpret the corrected regression for the selection bias.

5.2 Electricity Model

The results for the electricity model are displayed in table 11 (and table 17-19 Appendix). In interpreting the results it is important to emphasise that the three characteristics should be examined combined instead of separately, as explained in the literature review. The stated hypothesis would expect a negative sign for the coefficient of the interaction term between the duration of the application process and industry intensity displaying the dependency on electricity. The results show only a significant relation between the interaction term and firm size, which is positive for both selections of explanatory variables for the first step of the selection model. This positive coefficient can indicate a problem regarding reverse causality, even though the interaction should prevent this from happening, it shows that larger firms in more electricity intensive industries have on average a larger duration for the application process of obtaining an electricity connection. The specifications of the procedures in the application process are documented by the World Bank in country-specific analyses. This shows that firms have to submit a document in which they validate their company and define a power supply plan and estimate. This plan will be more complex for larger firms and in turn will take more time to judge for the government institution. Furthermore, Rajan and Zingales (1999) state that larger firms face more and costlier regulations than smaller firms. Based on an example of labour regulation, it can be seen that larger firms have to face stricter regulation and a higher level of enforcement by the government. This can also explain why the application process for larger firms is longer than for smaller firms.

27  

selection bias, is equal to 0,066%, which indicates a smaller affect across industries than after the inclusion of lambda. The significant explanatory variables displayed in table 9 are included as control variables in this regression (3b). The simplified influence of duration shows a positive significant relation for firm size and –productivity. This measure is however very sensitive for reverse causality and emphasises that a longer duration is related to larger (0,108%) and more productive (0,063%) firms.

Table  11:  Regression  results  Electricity  Connection.  Dependent  and  independent  variables  are  expressed  in  logs.  Duration  is   expressed  in  number  of  days  and  intensity  displays  an  industry  specific  electricity  intensity  ratio.  The  regressions  are  characterized  as:   (1)  𝑙𝑛 𝑓𝑖𝑟𝑚  𝑔𝑟𝑜𝑤𝑡ℎ! = 𝛼!+ 𝛽!  𝑙𝑛  𝐸𝑙𝑒𝑐𝑡𝑟𝑖𝑐𝑖𝑡𝑦  𝐼𝑛𝑡𝑒𝑛𝑠𝑖𝑡𝑦!+ 𝛽!ln 𝐸𝑙𝑒𝑐𝑡𝑟𝑖𝑐𝑖𝑡𝑦  𝑂𝑏𝑠𝑡𝑎𝑐𝑙𝑒!+  𝛽!𝐷!  𝐶𝑜𝑢𝑛𝑡𝑟𝑦 + 𝜀!  (2)  

𝑙𝑛 𝑓𝑖𝑟𝑚  𝑔𝑟𝑜𝑤𝑡ℎ! = 𝛼!+  𝛽!  𝑙𝑛  𝐸𝑙𝑒𝑐𝑡𝑟𝑖𝑐𝑖𝑡𝑦  𝐼𝑛𝑡𝑒𝑛𝑠𝑖𝑡𝑦!+ 𝛽!  ln   𝐸𝑙𝑒𝑐𝑡𝑟𝑖𝑐𝑖𝑡𝑦  𝑂𝑏𝑠𝑡𝑎𝑐𝑙𝑒!+ 𝛽!  (ln   𝐸𝑙𝑒𝑐𝑡𝑟𝑖𝑐𝑖𝑡𝑦  𝐼𝑛𝑡𝑒𝑛𝑠𝑖𝑡𝑦!×  ln   𝐸𝑙𝑒𝑐𝑡𝑟𝑖𝑐𝑖𝑡𝑦  𝑂𝑏𝑠𝑡𝑎𝑐𝑙𝑒!) +

 𝛽!𝐷!  𝐶𝑜𝑢𝑛𝑡𝑟𝑦 + 𝛽!𝐷!  𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦 + 𝜀!  and  (3)  𝑙𝑛 𝑓𝑖𝑟𝑚  𝑔𝑟𝑜𝑤𝑡ℎ! = 𝛼!+  𝛽!  𝑙𝑛  𝐸𝑙𝑒𝑐𝑡𝑟𝑖𝑐𝑖𝑡𝑦  𝐼𝑛𝑡𝑒𝑛𝑠𝑖𝑡𝑦!+ 𝛽!  ln   𝐸𝑙𝑒𝑐𝑡𝑟𝑖𝑐𝑖𝑡𝑦  𝑂𝑏𝑠𝑡𝑎𝑐𝑙𝑒!  

+𝛽_!  (𝑙𝑛  𝐸𝑙𝑒𝑐𝑡𝑟𝑖𝑐𝑖𝑡𝑦  𝐼𝑛𝑡𝑒𝑛𝑠𝑖𝑡𝑦!×  𝑙𝑛  𝐸𝑙𝑒𝑐𝑡𝑟𝑖𝑐𝑖𝑡𝑦  𝑂𝑏𝑠𝑡𝑎𝑐𝑙𝑒!) +  𝛽!𝐷!  𝐶𝑜𝑢𝑛𝑡𝑟𝑦 + 𝛽!𝐷!  𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦 + 𝜆 + 𝑐𝑜𝑛𝑡𝑟𝑜𝑙  𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠  𝐻𝑒𝑐𝑘𝑚𝑎𝑛  𝑚𝑜𝑑𝑒𝑙 + 𝜀!.  For  firm  

size  (3a)  estimates  the  selection  model  as  discussed  in  formulae  (6)  in  the  article,  while  (3b)  estimates  the  additional  selection  model.  Note:  ***  p<0,01;   **  p<0,05;  *  p<0,1.  

As stated in the beginning of this section, the economic relevance increases if the same trend can be found across the different firm characteristics instead of only one. Especially firm size is dependent on various different factors, so a significant interaction term in only this characteristic does not support a general conclusion.

In short, there is no clear relation between the duration of the application process to obtain an electricity connection and firm growth. The regressions show no significant negative coefficients for the interaction term in the selection-bias-adjusted regressions across firm characteristics. Based on the simplified variable of duration this conclusion can not be made due to causality problems.

5.3 Construction permits

The same regressions are estimated for the obstacle of construction permits (table 12 and table 20-23 Appendix). The interaction term of firm productivity shows a negative coefficient (3), which is

Electricity Connection

Firm Size Firm Investment Firm Productivity Variables (1) (2) (3a) (3b) (1) (2) (3) (1) (2) (3) Constant 3,745*** 3,051*** 4,984*** 4,834** -3,090*** -6,322** 14,234*** 7,893*** 10,990*** 9,210*** Ln Duration 0,120*** 0,133*** 0,125*** 0,108*** 0,092* 0,102** 0,034 0,074*** 0,069*** 0,063** Ln Intensity 0,004 0,081 -0,067 0,185 0,971 1,998 -1,054 0,000 -0,468 -0,438 Duration#Intensity 0,066** 0,088*** 0,14** 0,026 -0,046 -0,031 -0,039 Country Experience 0,000 0,000 0,000 Sector Experience 0,009 0,015*** 0,013***

Part of larger firm -0,989*** -0,775*** -0,546***

in line with the stated hypothesis. This indicates that firms in more capital intensive industries experience on average a larger restriction to firm productivity when the duration of the application process increases, compared to lower capital intensive industries. The coefficient of the interaction term is equal to -0,368%, translated into a decline in firm productivity for capital intensive industries of -0,974%, compared to low capital intensive industries decline of -0,401%, when duration increases with 1%. This can indicate a barrier of entry for firms in capital intensive markets. Building permits are important in this industry, but the process to obtain these permits lowers firm productivity. In other words, in order to be able to cope with these productivity decreases it is vital for firms to have an accomplished business, which is often not the case for new entrants on the market. This creates an essential minimum level of productivity for entrants in order to survive on the market, which in turn might demotivate potential new entrants.

Table  12:  Regression  results  Construction  Permits.  Dependent  and  independent  variables  are  expressed  in  logs.  Duration  is   expressed  in  number  of  days  and  intensity  displays  an  industry  specific  capital  intensity  ratio.  The  regressions  are  characterized  as:   (1)  𝑙𝑛 𝑓𝑖𝑟𝑚  𝑔𝑟𝑜𝑤𝑡ℎ! = 𝛼!+ 𝛽!  𝑙𝑛  𝐶𝑎𝑝𝑖𝑡𝑎𝑙  𝐼𝑛𝑡𝑒𝑛𝑠𝑖𝑡𝑦!+ 𝛽!ln 𝐵𝑢𝑖𝑙𝑑𝑖𝑛𝑔  𝑃𝑒𝑟𝑚𝑖𝑡  𝑂𝑏𝑠𝑡𝑎𝑐𝑙𝑒!+  𝛽!𝐷!  𝐶𝑜𝑢𝑛𝑡𝑟𝑦 + 𝜀!  (2)  𝑙𝑛 𝑓𝑖𝑟𝑚  𝑔𝑟𝑜𝑤𝑡ℎ! = 𝛼!+

 𝛽_!  𝑙𝑛  𝐶𝑎𝑝𝑖𝑡𝑎𝑙  𝐼𝑛𝑡𝑒𝑛𝑠𝑖𝑡𝑦!+ 𝛽!ln 𝐵𝑢𝑖𝑙𝑑𝑖𝑛𝑔 𝑃𝑒𝑟𝑚𝑖𝑡  𝑂𝑏𝑠𝑡𝑎𝑐𝑙𝑒!+ 𝛽!  (ln   𝐶𝑎𝑝𝑖𝑡𝑎𝑙  𝐼𝑛𝑡𝑒𝑛𝑠𝑖𝑡𝑦!× ln 𝐵𝑢𝑖𝑙𝑑𝑖𝑛𝑔 𝑃𝑒𝑟𝑚𝑖𝑡  𝑂𝑏𝑠𝑡𝑎𝑐𝑙𝑒!) +  𝛽!𝐷!  𝐶𝑜𝑢𝑛𝑡𝑟𝑦 +

𝛽!𝐷!  𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦 + 𝜀!  and  (3)  𝑙𝑛 𝑓𝑖𝑟𝑚  𝑔𝑟𝑜𝑤𝑡ℎ! = 𝛼!+  𝛽!  𝑙𝑛  𝐶𝑎𝑝𝑖𝑡𝑎𝑙  𝐼𝑛𝑡𝑒𝑛𝑠𝑖𝑡𝑦!+ 𝛽!ln 𝐵𝑢𝑖𝑙𝑑𝑖𝑛𝑔 𝑃𝑒𝑟𝑚𝑖𝑡  𝑂𝑏𝑠𝑡𝑎𝑐𝑙𝑒!  

+𝛽_!  (𝑙𝑛  𝐶𝑝𝑖𝑎𝑡𝑎𝑙  𝐼𝑛𝑡𝑒𝑛𝑠𝑖𝑡𝑦!×  𝑙𝑛  𝐵𝑢𝑖𝑙𝑑𝑖𝑛𝑔  𝑃𝑒𝑟𝑚𝑖𝑡  𝑂𝑏𝑠𝑡𝑎𝑐𝑙𝑒!) +  𝛽!𝐷!  𝐶𝑜𝑢𝑛𝑡𝑟𝑦 + 𝛽!𝐷!  𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦 + 𝜆 + 𝑐𝑜𝑛𝑡𝑟𝑜𝑙  𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠  𝐻𝑒𝑐𝑘𝑚𝑎𝑛  𝑚𝑜𝑑𝑒𝑙 + 𝜀!.  For  firm  

size  (3a)  estimates  the  selection  model  as  discussed  in  formulae  (6)  in  the  article,  while  (3b)  estimates  the  additional  selection  model.  Note:  ***  p<0,01;   **  p<0,05;  *  p<0,1.  

This trend of growth restriction does not hold across firm size and –investment. The interaction term for firm size is significant when estimated for the original selected variables in equation (6) in section IV. However, these coefficients cannot be interpreted with certainty while one of the Heckman model requirements is not met. The adjusted variable selection displayed in (3b) is no longer significant. In other words, the relation between duration and firm size is not robust against the inclusion of the appropriate lambda. Also the simplified duration variable is only significant in the

selection-bias-Construction Permits

Firm Size Firm Investment Firm Productivity Variables (1) (2) (3a) (3b) (1) (2) (3) (1) (2) (3) Constant 4,395*** 4,678*** 2,636*** -0,036 -2,108*** -3,301*** -3,553*** 8,359*** 10,856*** 14,28*** Ln Duration 0,090*** -0,075 -0,076 0,171 0,066 0,198 0,156 0,041 -0,094 -0,141* Ln Intensity 0,853*** -3,409 0,033 1,364 -0,581 0,772 -1,412 -0,076 -1,459 -1,946 Duration#Intensity 0,377*** 0,345* -0,036 -0,290 -0,206 -0,283* -0,368** Country Experience 0,000 0,000 Sector Experience 0,028*** 0,025*** -0,006

Part of larger firm -1,505*** -1,028*** 0,227**

Honest Government 0,109***

Land Owner 0,011***

Percentage Land Used -0,011