WPS6160 Policy Research Working Paper 6160 Understanding the Business Environment in South Asia Evidence from Firm-Level Surveys Wendy Carlin Mark Schaffer The World Bank South Asia Region Human Development Department August 2012 Policy Research Working Paper 6160 Abstract This paper examines the relationship between firm macroeconomic conditions, rule of law, etc.—for the performance and growth and the business environment growth of their firm. The analysis finds, in line with this in the countries of the South Asia Region—Afghanistan, approach, that higher-productivity and better-performing Bangladesh, Bhutan, India, Maldives, Nepal, Pakistan, firms in the region, and in particular firms that recently and Sri Lanka—using firm-level data from the World expanded their employment and created jobs, report Bank’s Enterprise Surveys. The analysis uses an approach significantly higher constraints in terms of the supply of in which the responses of firms to questions about the public inputs. The authors discuss the differences across quality of the business environment can be interpreted countries in the importance of various industries, how as shadow prices: estimations by managers of the cost they relate to various firm characteristics, how informal imposed on the firm by inadequacies of an aspect of the and rural sector firms are constrained by public inputs, business environment—public inputs such as regulation, and how firms in the South Asia Region countries physical infrastructure, availability of skilled labor, compare with firms in the rest of the world. This paper is a product of the Human Development Department, South Asia Region. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The author may be contacted at m.e.schaffer@hw.ac.uk. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Understanding the Business Environment in South Asia: Evidence from Firm-Level Surveys * Wendy Carlin 1 Mark Schaffer 2 JEL: D22, O17, O18, O43, O53 Key words: Firm Behavior, Formal and Informal Sectors, Business Environment, Constraints to Firm Growth, South Asia * We are very grateful to Paul Seabright for discussions on our related work which have influenced this paper. Comments and feedback from Reema Nayar, Pradeep Mitra and their colleagues in the Human Development Department of the South Asia Region, and from the audience at a June 2011 seminar in the SAR Chief Economist Seminar Series, are gratefully acknowledged. The usual caveat applies. 1 University College London and CEPR 2 Heriot-Watt University, CEPR and IZA 1. Introduction This paper examines the relationship between firm performance and growth and the business environment in the countries of the South Asia Region (SAR): Afghanistan, Bangladesh, Bhutan, India, Maldives, Nepal, Pakistan and Sri Lanka. The analysis uses data collected for the Enterprise Surveys conducted by the World Bank of formal sector firms in the region, along with surveys of informal sector and rural firms in several countries. The paper serves as an input to the World Bank South Asia Region’s flagship report, “More and Better Jobs in South Asia�. Surveys of firms across countries can be used to provide a rich description of how managers perceive the costs to them of the business environment in which they operate. Interpreting these surveys requires a conceptual framework, and the analysis in this paper uses the approach set out in Carlin et al. (2006, 2010, 2012) and Carlin and Schaffer (2012). In this framework, the responses of firms to questions about the quality of the business environment can be interpreted as shadow prices: estimations by managers of the cost imposed on the firm by inadequacies of an aspect of the business environment – regulation, physical infrastructure, availability of skilled labor, macroeconomic conditions, rule of law, etc. – for the growth of their firm. Expenditures on mitigation of these costs as reported by firms also fit naturally into this framework. When taken to the data, these predictions inform the policy maker whether, for example, it is the case that improvement in a particular element of the business environment is likely to benefit well- or poorly performing firms; whether public input bottlenecks are more pressing in urban or rural areas; and whether there are important differences across industries in the burden of weaknesses in the business environment. Since the focus of the flagship report is on jobs, we examine in particular how different elements of the business environment affect firms that are expanding employment. Our modeling framework allows us to interpret clearly the different business- environment related measures collected in enterprise surveys and to formulate hypotheses as to how they relate to firm efficiency and growth. It predicts in particular that well-performing firms report higher costs of constraints. This insight 2 has the implication that relaxing the constraints on these firms is likely to produce the largest increases in output and that scarce resources available to policy makers would be better directed toward easing such bottlenecks than in focusing on the bottlenecks reported by poorly performing firms. The structure of the paper is as follows. In Section 2, we summarize the analytical framework used in the rest of the paper and the predictions it generates about how a firm’s evaluation of the costs of constraints imposed by its external environment would vary with its characteristics; the model is set out in detail in Appendix 1. Section 3 describes the Enterprise Survey data available for the SAR countries and how the variables in the model map onto the data. In Sections 4-6 we analyze the survey data from formal firms in the region. As expected, firms that had created jobs in the preceding three years reported higher costs of constraints than did other firms. Reported costs – shadow prices – varied in the expected way with a number of other firm characteristics such as size and location. In Sections 7-8, we use country averages of costs of constraints reported by formal sector firms to assess which elements of the external environment are more problematic for firms across the region and in each country, and in Section 9 we compare these country averages for the SAR countries with those reported by formal sector firms in other countries at a similar level of development. Section 10 compares business environment constraints reported by formal sector firms in the region with those reported by rural and informal sector firms. Section 11 concludes. 2. Modeling framework Our modeling builds on the framework in Carlin et al. (2006, 2010, 2012). The framework allows us to interpret clearly the different business-environment related measures collected in the Enterprise Surveys and to formulate hypotheses as to how they relate to firm efficiency and growth. Our modeling framework predicts that well-performing firms report higher costs of constraints. This insight has the implication that relaxing the constraints on these firms is likely to produce the largest increases in output and that scarce resources would be better directed toward easing such bottlenecks than in focusing on the bottlenecks reported by poorly performing 3 firms. However, it is important to remember that the data collected in the business environment surveys provides information only about the constraints facing existing enterprises. It is not a useful source for addressing the constraints facing potential firms. Before discussing our framework in more detail, it is useful to address the question of why the importance of the business environment for economic performance cannot be readily estimated directly using firm-level indicators. Why can’t the importance of the business environment be estimated directly using firm-level indicators? A commonly-employed approach to using firm-level data to estimate the impact of variations in the business environment on firm performance is to estimate a regression in which firm performance is the dependent variable and with a measure of the business environment as reported by the firm used as a regressor along with various covariates. For example, firm-level data can be used to estimate a total factor productivity equation in which output appears on the left-hand side and inputs and other covariates appear on the right-hand-side along with one or more business environment measures. There are three reasons why such an attempt to estimate the effect of variations in the business environment at the level of the firm on productivity (or growth) is likely to be unsuccessful. First, the many dimensions of the business environment are likely to be correlated, which makes it very difficult to tease out their separate effects on performance. Inclusion in the regression of single measures of the business environment reported by the firm is likely to result in endogeneity/omitted variable bias, whereas including many measures and controls will typically lead to very imprecisely estimated coefficients (the “curse of dimensionality� problem). This is the same problem faced by attempts to uncover the institutional determinants of growth in cross-country studies. 3 3 See, e.g., Easterly (2009), Dethier (2008, 2010), or Durlauf et al. (2005) for discussions of the macro literature. 4 Second, even if there was only one candidate dimension of the business environment, for its effect on firm-level productivity to be estimated requires that it vary at the level of the firm. This is clearly not the case for a number of elements of the business environment such as macroeconomic stability and the court system. Third, in cases where there is variation at the level of the firm, its effect on performance can be estimated only if there is a way of isolating the quality of such a firm-level micro-business environment from the firm’s characteristics. A simple example illustrates the problems. It is plausible that a higher productivity firm will attract more attention from rent-seeking bureaucrats; hence, a naïve regression of firm performance on the number of visits would produce a positive estimate of the effect of bureaucratic attention on performance. The major research strategy adopted to get around this problem and uncover the effect of inspections on firm performance separate from the effect of firm performance in attracting inspections has been to use the so-called “cell averages� approach. Instead of using the firm’s own report on the number of visits, the average number reported by firms with similar characteristics (such as firm size, industry and location) is used. The cell averages approach is one of the strategies for addressing the endogeneity bias problem recommended by Dethier et al. (2008, 2010) in their survey paper. However, the cell averages approach runs into an immediate problem. Unobservable characteristics that cause or are correlated with higher productivity of the firm in question will also tend to raise the productivity levels of the other firms in the cell (e.g., a local demand or industry-specific shock will boost capacity utilization and performance). This will tend to raise the prevalence of inspections, expenditure on abatement such as bribes and the seriousness of this element of the business environment reported by the firm. This is an example of Manski’s (1993) “reflection problem�. The econometric challenge in trying to tease apart differences in the institutional environment faced by firms in a single country while avoiding the problem of endogeneity, is too much for the data to bear, which is why a recent careful study of 5 the data for transition economies that tried to do just this, Commander and Svejnar (2009), found largely null results. The three problems with attempting to uncover the relevance of elements of the business environment by estimating directly a production function augmented by business environment indicators can be avoided by taking a different approach. Following Carlin et al. (2006, 2010), we take as our starting point that the business environment is external to the firm and that to an important extent, firms in a country share the same environment. This suggests that firm level information be used in a different way from the augmented production function method. Specifically, we formulate predictions as to how the firm’s response to its business environment in terms of its expenditure on abatement and its evaluation of the costs imposed on it by deficiencies in the business environment varies with its characteristics, including its performance. In short, a firm-level assessment of the business environment is an endogenous variable in the modeling framework and as the dependent variable in estimations. The framework can be also used to address the relative importance of different aspects of the business environment in different countries. Modeling framework: Summary The model is set out in Appendix 1 and we summarize it here. We use a simple single-period firm production function with 4 inputs, L, E, B and G, which are combined to product output Y. L is employment; E is an intermediate input that is a flow of services which results from combining a public input B with G, an input the firm purchases in order to mitigate the effects of the unreliability of the public input. Firms also differ in productivity, which we capture with a productivity parameter A. We index countries by j and firms by i. In our basic model, the public input B that is supplied on identical terms to all firms in a country, so we write it as B j . This captures the notion of a shared “business environment� – B j varies across countries but not across firms within a given country. We then extend the model to cover the case where the public input varies with the firm’s productivity or profitability. 6 For example, if B j is the quality of the electricity supply from the grid, and Gij is the firm’s generator, then B j and Gij combine to create the firm’s electricity input Eij . Spending money on Gij means the firm can generate its own electricity when there is an interruption in supply from the main electricity grid ( B j ). Electricity is then combined with labor to create output. Although the quality of the business environment at country level cannot be directly observed, there are country-level proxies for B j such as the degree of reliability of the electricity supply (e.g., in terms of outages). Corruption is another example. In this case, B j is a measure of the honesty of government bureaucrats in country j, but the public input Bij supplied to firm i depends not only on B j but also on productivity Aij : high productivity firms will attract attention from dishonest government officials looking for bribes. As a consequence, the quality of the public input supplied to high productivity firms will be lower than that supplied to low productivity firms that are ignored by the bribe- seeking officials. Thus if B j is the honesty of the bureaucracy in country j, Bij is the inverse of the number of inspections that a firm with productivity Aij attracts (more inspections means a lower quality public input Bij supplied to the firm), and Gij is bribes. Central to our analysis are the “Subjective Severity� indicators collected in the Enterprise Surveys. These are responses to questions about a feature of the business environment faced by the firm, where the question takes the form, “How much of an obstacle is X to the operation and growth of your business?�, and the respondent rates the severity on a 5-point scale of 0 (“no obstacle�) to 4 (“very severe obstacle�). The key point about these subjective severity indicators is that these are not estimates of the country-wide public input B j , or even of the public input Bij supplied to firm i; they are valuations of the public input. A simple and intuitive interpretation is that the “reported cost� Rij of a public input is the gap between the firm’s profit in the 7 hypothetical situation where the public input provided is of such high quality that it poses a negligible obstacle to the firm’s operations, and the firm’s profit in reality, given the actual quality of public input provided. We show in the Appendix that Rij can be interpreted as the shadow price of shortcomings in the public input B j . The core predictions of the model (see the Appendix) are that better performing firms (faster growing, higher productivity, etc.) ∗ • spend more on mitigation, Gij (e.g., are more likely to have a generator; are more likely to pay bribes); ∗ • report better public input services Eij in cases where the input service is not a function of firm-level productivity (e.g., are more likely to report fewer delays at customs); ∗ • report higher or lower public input services Eij in cases where the input service is a positive function of firm-level productivity (e.g., outcome depends on the offsetting effects of a higher number of inspections and greater expenditure on bribes); ∗ • report higher costs of public input constraints, MRCij . 3. Mapping the framework to the data In this section, we show how the framework is matched with the data in the Enterprise Surveys. We begin by identifying a number of proxy variables for unobserved firm productivity and firm growth. We then summarize the variables that are proxies for * the reported costs or shadow prices of constraints, Rij , mitigation costs Gij , the flow * of public input services Eij and the shared business environment B j . Performance variables We define the following measures of firm performance that are available in the Enterprise Surveys. 8 • Growth of permanent employment: this is a dummy variable that takes the value of 1 if there was an increase in the number of permanent employees over the preceding three years. 4 • Labor productivity: the log of value added per worker, where value added is defined as sales less spending on raw materials and deflated using a PPP-based year-specific exchange rate. • Total factor productivity: this variable is constructed as a simple residual using logs of sales, employment, fixed capital and material, with weights on the latter three variables set to 0.25, 0.10 and 0.65, respectively. • R&D-firm: this is a dummy variable that takes the value of 1 if the firm reports that it does R&D. • Process or product innovation: this is a dummy variable that take the value of 1 if the reports that it introduced a new process or product during the previous three years. • Sales to MNCs: the percentage of sales reported to have been made to multi- national companies. We interpret this as an indicator of external evaluation of the firm’s quality. • Education top level of the manager: this is measured in years of education. • Average education level of the production workforce: this is also measured in years of education. • Training offered: this is a dummy variable that indicates whether the firm has an in-firm training program. • Per cent of production workers trained: this relates to the firm’s in-firm training. The reported costs of public input constraints Measures of Rij , the reported cost or shadow price of public inputs, are available in the SAR data: “How much of an obstacle is XX to the operation and growth of your business?� The model predicts that these will be positively related to measures of productivity and growth: they will be increasing in Aij and positively correlated with 4 If permanent employment data is not available for a firm for 3 years earlier, a shorter period is used. 9 measures that are also correlated with Aij . The public inputs for which we have data are: • Electricity • Telecoms • Transport • Customs • Unfair competition • Access to land • Crime/theft/disorder • Tax administration • Business licensing • Political instability • Government policy instability • Corruption • Operation of the courts and legal system • Macroeconomic instability • Labor regulation • Skilled labor shortages There is another question that is best interpreted along with the reported costs measures. Firms are asked about the adequacy of their access to water. When they report “insufficient water�, this can be interpreted as meaning that “insufficient water� is an obstacle to production. Had they answered “sufficient water� then this would be equivalent to the answer above of “not an obstacle�. We therefore consider water together with the other public input constraints. • Sufficient supply of water Mitigation outlays * Measures of mitigation costs Gij will also be increasing in Aij . The following indicators of mitigation outlays are included in the Enterprise Surveys. • Bribes (=1 if the firm paid bribes, =0 if not) • Managerial time dealing with regulatory issues (%) • Generator (=1 if the firm has its own generator, =0 if not) The flow of public input services * Measures of the flow of services Eij that are available in the Enterprise Surveys are: 10 • Average number of days for exporter to get goods through customs. This is an inverse measure of the speed with which goods are processed through customs. • Total number of inspections by officials per year. This is also an inverse measure; a higher number of inspections reduces the flow of services from the relevant public input. Firm-level estimates of the shared business environment There is only one indicator of B j in the Enterprise Surveys, namely the firm’s experience of interruptions to the power supply. We use this in dummy variable form: • Power cuts (=1 if more than one per month, =0 otherwise) How can the answers to the “Access to Finance� and “Tax Rates� questions be interpreted? There are a number of reasons why “finance� and “tax rates� cannot be interpreted in the same way as the public input constraints (Carlin et al. 2006). If finance had the character of a public good like telecoms or customs regulation then we would expect that a high score would mean that better access to finance would boost output. However, because of the agency problems characteristic of the firm, better access to finance may result in more funding of pet projects that do not raise output (on average) but lead to higher default rates. Where the finance system is working well, we would expect that access to finance is a constraint on at least some managers. This is not the case of a well-functioning electricity system or customs administration. Finally, if financial institutions are functioning well, the perception of its availability as a constraint should be inversely related to the quality of investment projects the firm has available to fund, so that high scores may indicate poor quality projects rather than the potential for increased output. 11 The interpretation of responses to managers as to the importance of tax rates for the operation of their business faces similar difficulties as the interpretation of the importance of access to finance, and partly for similar reasons. Just as managers do not take account of the social benefits of a financial system that constrains access to finance by aiming to screen out poor projects, they also do not take account of the social benefits arising from government spending funded by the taxes they pay. We suggest that the way to interpret the responses of managers on tax rates is that they point to the costs imposed on firms if public inputs are supplied at the cost of higher taxes than necessary. Tax rates are very highly ranked as a constraint by managers in virtually all countries (irrespective of their level of development) but it does not follow that it is a priority everywhere to cut taxation. A high ranking is probably better interpreted as indicating that policies to reduce tax rates while holding other aspects of public infrastructure provision constant (for instance, by improving administrative efficiency) would improve firm performance. For the public input constraints, we use the survey answers to get at the private cost to the firm of inadequate or unreliable public inputs or burdensome regulation. It is not possible to use micro data of this kind to uncover whether tax rates are a key constraint to firms – firms are unlikely to take into account the public inputs that are paid for by taxation when responding to the question. 4. Are job-creating firms in the South Asia Region also high-performance firms? The analysis in this paper relates to formal-sector firms in 8 South Asia Region (SAR) countries: Afghanistan, Bangladesh, Bhutan, India, the Maldives, Nepal, Pakistan and Sri Lanka. The data were collected in a series of surveys over the 10-year period 2000-2010. Tables 1 and 2 present the number of firms and the median permanent employment by country and year. Firms in the three most-populous countries – India, Bangladesh and Pakistan – account for 80% of the total sample. Most of the firms surveyed are SMEs; median permanent employment is 19 persons. This is true of the individual surveys as well; the exceptions are the surveys in Bangladesh and Sri Lanka in 2002 and 2003, where permanent employment in the median firm exceeded 12 100 persons. All the estimations reported below, both regressions and correlations, use country fixed effects. This includes pairwise correlations, i.e., these are pairwise partial correlations with country effects partialed out. When investigating the role of external constraints on firm performance, our modeling framework highlights that it is the firms that are performing well that are predicted to report higher costs of constraints. This insight has the implication that relaxing the constraints on these firms is likely to produce the largest increases in output and that scarce resources would be better directed toward easing such bottlenecks than in focusing on the bottlenecks reported by poorly performing firms. Before introducing the results on business environment constraints, it is important to understand the correlations among the indicators of firm performance. Country means for these performance measures are given in Table 3. Differences across countries will be discussed in a separate piece. Note that these are unconditional means, and the differences by country are influenced by differences in sample composition (e.g., manufacturing vs. services). Note also that some measures, notably R&D activity and sales to MNCs, are not available for some countries. Is it the case that permanent employment growth is concentrated in firms that are also successful when measured by the other performance criteria? Broadly speaking the answer is yes. The green cells in Table 4 show a positive and significant correlation between the growth of employment variable and whether the firm does R&D, its introduction of a new process or product, the education level of the top manager and of the workforce, as well as measures of in-firm training. These are unconditional correlations (more precisely, they are conditioned on country fixed effects only). It is also important to know whether the performance correlations with the growth of employment remain once other firm characteristics are taken into account. As a first step, Table 5 shows that larger, more established, more internationally engaged firms, and those located in large cities are also the ones where jobs are most likely to have been created. 13 Definitions of the control variables for firm characteristics: • Firm size: measured by the log of the firm’s average number of employees over the previous three years. • New firm: this is a dummy variable equal to one if the firm is 4 years old or less. • Services: this is a dummy variable that takes the value of one for firms in the services sector; the benchmark is manufacturing. • Foreign: this is a dummy variable that takes the value of one if the firm has a foreign owner with a stake of at least 10%. • Exporter: this is a dummy variable that the value of one if the firm exports at least 10% of its sales, either directly or through intermediaries. • Importer: this is a dummy variable equal to one if the firm imports directly. 5 • Small city or rural: this is a dummy variable that takes the value of one if the firm is located in a small city or rural area. Table 6 reports the means for these controls by country and for the sample as a whole. Once we partial out these firm characteristics and re-do the correlations of performance measures, Table 7 shows that the patterns shown in Table 4 remain broadly unchanged. The employment growth variable is significantly correlated with most of the other performance measures – and each of those continues to be correlated with most of the other measures. Note that employment growth is not positively correlated with the level of labor productivity or total factor productivity. This is explained by the fact that there will be a spurious negative correlation between a firm’s employment growth and its level of productivity when it hires additional labor in order to expand output. When the control variables are included, the correlation between employment growth and labor productivity is significantly negative. The reason is that part of the positive correlation between the controls and both job growth and labor productivity is 5 Since the importer status variable is missing for some countries, we include a “missing importer� dummy variable that allows us to include many firms for which this variable is missing. 14 removed once the controls are included. Hence, the spurious negative correlation between employment growth and labor productivity is then much stronger. The outcome of this initial descriptive analysis is that job creation is associated with specific firm characteristics and with a wide range of other indicators of firm performance. Job growth takes place in larger, more established firms that are internationally engaged – and these firms are also more likely than others to be innovative, and to have a more educated manager and workforce. On the basis of this, we move to examine how the reported costs of constraints vary with firm performance. 5. How does firm performance affect the evaluation of constraints imposed by the external environment? Our modeling framework predicts that better-performing firms report higher constraints from the external environment. Do we find this pattern in the data? In the analysis that follows, we look first at how the evaluation of business environment constraints varies with firm performance. In order to do this, we regress the firm’s evaluation of the seriousness of each element of the external environment on the performance measure and on the standard set of controls introduced above (firm size, firm age, industry, ownership, exporter status, importer status, and location, plus country fixed effects). The results we report use heteroskedatic-robust standard errors. Means by country of the firm’s evaluation of constraints, and of mitigation outlays, the flow of services of public inputs, and of the shared business environment, are reported in Tables 8 and 9. The virtually universal problems with electricity supply in Afghanistan, Bangladesh, Nepal and Pakistan are highlighted by the results reported in column 8 of Table 9: almost all firms report at least one power cut per month (and most report at least one per week). Note that although almost all firms in these countries face a poor quality business environment in this respect, their valuations of this aspect of the environment (the “subjective severity� question on electricity) vary. Over one-third of firms in each of these three countries report that 15 electricity is at most a “moderate� obstacle (3 on the 0-4 scale). This provides a good illustration of how our framework works: all firms in these three countries report that the quality of this particular public input ( B j ) is poor, but they differ in their estimates of how costly this is for them (Rij). Public input constraints and firm performance We begin with the public input constraints (Table 10). The coloring of the cells in the table provides an immediate visual impression that firms performing well across each of the performance dimensions shown in the top row of the table report higher costs of constraints than do less well performing firms (green cells). In the lower part of the table, a number of indicators of mitigation expenditure are shown. The predominance of green cells there shows that expenditure on mitigation (e.g., bribes, generators) is also positively correlated with firm performance. Benchmark performance measure: job-creation Turning to firms that have recently expanded permanent employment (first column in Table 10), it is apparent that such firms report higher costs of constraints across virtually the whole range of external constraints both in terms of physical infrastructure and in the regulatory and broader policy environment. The only exceptions among the sixteen constraints (including insufficient water supply) are those imposed by “unfair competition� and political instability. Job-creating firms also report a higher tendency to pay bribes, are the target of more inspections by officials, report a higher frequency of power outages and are more likely to have a generator. Alternative performance measures Using the static efficiency measures of performance, higher productivity firms complain more about most institutional aspects of the business environment (see Table 10). Since higher productivity firms are not necessarily engaged in expansion, it is unsurprising that physical infrastructure and access to land and skilled labor are reported as less onerous than is the case for job-creating firms. Of the physical infrastructure elements, higher productivity firms appear somewhat more troubled by poor telecoms than are less productive firms. 16 Firms that have undertaken R&D and those that have innovated (introduced a new process or product) do not report higher costs of electricity or telecoms constraints than do other firms. R&D-firms report fewer problems with transport than other firms but higher constraints across the board otherwise. The complaints of innovating firms are a little more narrowly focused than for R&D-firms. In both cases, however, higher costs of anti-competitive behavior, customs regulation, corruption, access to land, an inadequately educated labor force, policy uncertainty and macroeconomic instability are reported. Firms with more highly educated managers also identify a broad range of elements of the external institutional environment as more costly than do firms with less well- educated managers. Like higher productivity firms, those that do R&D and have highly educated managers are more likely to pay bribes, have a generator and attract more inspections. They also report more management time spent with officials. The constraints reported by firms with a higher share of educated production workers are rather different. In particular, such firms report lower costs associated with tax administration and corruption than do other firms and there is no greater tendency to pay bribes, spend management time with officials or be inspected than is the case for other firms. It is notable that firms with a more educated labor force do not report access to skilled labor as more costly than other firms. However, firms that are engaged in training do report access to skilled labor as more costly. Firms that sell to MNCs are prone to report higher costs of access to suitably qualified labor and of the courts than firms that are not involved in an MNC relationship. Otherwise, reported constraints are lower or no different from non-MNC related firms. To summarize, across a wide range of performance indicators, reported costs of public input constraints are significantly higher in better-performing firms. Such firms are also likely to be engaged in more activity to mitigate the effects of poor physical and institutional infrastructure. 17 Access to finance is different We conclude this section by examining the results for access to finance. As explained above, a firm’s evaluations of access to finance will behave differently from the public good constraints, because when asked about the obstacle to their activities posed by difficulties with access to finance, the firm’s own circumstances will directly affect the terms on which finance is available and hence its answer to the question. In particular, the firm’s answer to a question about how much of an obstacle to its operation and growth is posed by access to finance will be influenced by its investment plans and its internal financial resources (retentions). One element of the firm’s ability to access external finance will be the “objective� state of the financial system in the economy (or region). However, the terms on which external finance is available and hence the firm’s answer as to its ease of access will be heavily conditioned by its need for finance relative to internal sources, and the collateral it has available. A characteristic of a well-functioning financial system is precisely its ability to direct finance according to firm (or project) -specific quality. Looking at Table 10, we can see that irrespective of the performance measure, for public input constraints, the typical pattern is for better performing firms to report higher costs of constraints (green cells). This is not the case for access to finance. More efficient firms (as indicated by higher labor productivity and TFP) report lower problems with access to finance than do other firms (red cells). This is also the case for firms that sell to multinational companies. This highlights the likely causal chain from good firm performance to the availability of internal finance and easier access to external finance based on sound prior performance and the associated availability of collateral. High productivity firms may also be closer to their optimal capital stock and hence have a more limited need for additional external finance. Consistent with this interpretation is the finding that firms doing R&D report higher financial constraints than firms that do not do R&D. This suggests that informational asymmetries associated with innovation are likely to make access to external finance difficult – even if R&D is a signal of a potentially dynamic firm. 18 6. How do the costs of constraints vary with firm characteristics? In this section, we report the results of our baseline regression, where each business constraint is regressed on the job creation variable and a set of controls. These are reported in Table 11. The coefficient on the indicator for expanding employment is identical to that in Table 10 and was discussed above. Here we discuss the how the reported cost of the public input constraints are related to other characteristics of the firm. In our model, the size of the firm is endogenous, i.e., the level of employment chosen is the profit maximizing one given the constraint of the supply of public input services, the firm’s level of productivity, and prices. The implication is that if we could observe the level of productivity, there would be no separate scale effect on reported costs of constraints. However, since productivity is not measured perfectly, we would expect that variables that are correlated with it to also attract a positive coefficient in the reported constraints regressions. • In particular, firm size is positively correlated with productivity and we would therefore expect larger firms to report higher costs and more mitigation. As column 1 in Table 11 shows, this prediction is borne out by the data. Apart from electricity and access to land, larger firms report higher costs of constraints, and in most cases the effect is significant (green cells). Larger firms report significantly lower costs associated with electricity and with access to land. Larger firms report fewer outages are more likely to have a generator. Nevertheless, the negative and significant coefficient on electricity is an unusual finding. The sign is positive and significant in a large sample of firms from across the world; including in the sub-sample of countries with levels of per capita GDP similar to those in the SAR sample • Firms in services are typically less capital-intensive, less unionized, more dependent on communications, and less engaged in trade than are manufacturing firms. They would be predicted to be less burdened by the electricity network, and by labor regulations and customs administration, and to report a higher cost of poor telecommunications. This is the case in the data. Services firms also report a significantly lower burden in relation to anti- 19 competitive behavior and corruption and a higher burden of the courts, political instability, access to land and business licensing. • Given their access to their parent firm’s internal capital market it would be predicted that foreign-owned firms report fewer problems with access to finance than domestically owned ones. This is indeed the case among the surveyed firms. They are also less prone to bribe and encounter fewer days of customs delay. The only dimensions on which foreign-owned firms report more costly constraints than do domestic firms is in relation to political instability and government policy uncertainty. • Exporters and importers. There are a number of interesting differences in the constraints reported by these firms. In terms of physical infrastructure, exporters report electricity as a costly constraint whereas importers report telecoms and transport as well as electricity as more costly than do non- importing firms. It is importers rather than exporters that are particularly bothered by customs administration. In line with expectations, exporters report fewer problems with anti-competitive behavior than do non-exporters; importers report the opposite. Macroeconomic instability is especially problematic for both types of internationally engaged firms, which is likely to reflect sensitivity to exchange rate movements and uncertainty. • Location. A direct extension of the modeling framework is the prediction that in an economy characterized by uneven development between less dynamic rural and more dynamic urban locations, firms in the more urban locations would report higher costs of constraints. Under the assumption that the supply of public goods is uniform across the country, this prediction simply follows from the greater demands on public inputs in the faster-growing locations, which in low income countries are typically the urban ones. To the extent more rural locations have objectively worse levels of public input supply, this would tend to make it less likely that we would observe the predicted dualism result of higher reported constraints in more urban locations. However, the predominance of red cells in the final column of Table 11 confirms the dualism prediction: firms in small cities or rural locations report lower costs of constraints across the board (including water); the only exception is telecoms, which is a low-ranked constraint everywhere and where there is no significant 20 difference by location. Firms in more rural locations are more likely to have a generator and report more power outages than do firms in more urban locations, but large-city firms report a higher cost of constraint in relation to electricity. • Established versus new firms. The modeling framework does not have any particular predictions as to how firm age should relate to the reported cost of constraints. However, as noted above, it is established firms that tend to be expanding employment. In terms of reported constraints, there is no difference between new and established firms across most institutional elements. New firms report higher costs of constraints for telecoms and transport, and lower costs for corruption and crime. 7. Are there country differences in the way constraints vary? Before looking at the way constraints vary across countries, we check for country differences in the correlations between the job creation variable and firm characteristics (Table 12). The procedure is the same as that for the pooled sample: for each country sample, we report the correlation and partial correlation between the job creation variable and the characteristic in question. The “corr� and “pcorr� columns report these correlations. For comparison purposes, the correlations for the pooled sample in Table 5 are repeated in Table 12 in the rows labeled “All�. We also report whether these correlations are significantly different from those for the rest of the pooled sample; these are reported in the “Diff?� columns. These latter tests are obtained by a pooled estimation in which a dummy variable for the country of interest is interacted with the other variables. As noted above, in the sample as a whole, firms that expanded employment tended to be larger, more established ones located in urban areas that were internationally engaged via ownership, exporting or importing. In the India sample, we see that the rural-urban pattern is less pronounced than that in the sample as a whole: firms in large cities report higher employment growth than in small cities but the difference disappears once the other controls are added. The pattern we saw above for the 21 sample at a whole – employment growth in large cities – is driven by the firms in Bangladesh, Bhutan, Pakistan and Sri Lanka. The positive correlation between employment growth and the age of the firm in the sample as a whole is driven by firms in the sample from Afghanistan, Bangladesh and Pakistan. Elsewhere there is no strong age of firm effect. Once other characteristics are accounted for, the positive association between employment growth and international engagement is strongest in India. In the sample as a whole, there is no strong association between job-creating firms and either sector or ownership. However, there are some country variations. In Bhutan, Nepal, Pakistan and Sri Lanka, it is services rather than manufacturing firms that are more likely to be job-creating. Reported constraints – do countries differ? Our investigation of differences across countries in reported constraints uses a similar approach to that above. We estimate the same regression as with the pooled sample, but separately for each country. Significance tests are reported based on these country-by-country estimations. The results are reported in Table 13; for comparison purposes, the results for the pooled sample in Table 11 are repeated here in the rows labeled “All�. We also report in the “Different?� columns whether regression coefficients are different in the country of interest vs. the rest of the sample, again using the simple procedure of a pooled regression in which the country dummy is interacted with the other regressors. As we have seen earlier, in the sample as a whole in the benchmark regression (job creation plus controls), firms that expanded employment reported higher costs of constraints virtually across the board. When we look at the individual country samples, we find this pattern very clearly for India and Pakistan (note the green and the “greater� cells). The countries that look different are Afghanistan and Bangladesh. In both the latter countries, job-creating firms tend to report lower rather than higher costs of constraints (red and “smaller� cells). It is also growing firms that 22 are less likely to pay bribes and spend less manager time dealing with officials. In Bangladesh, growing firms report lower constraints from anti-competitive behavior by other firms and labor regulation. There are also some interesting cross-country differences in reported constraints for the firm characteristics, which may be relevant for policy. The most striking difference among the SAR countries concerns the location characteristic. Earlier we found that in contrast to the rest of the region, job-creation was uncorrelated with location in India once we controlled for other firm characteristics. Elsewhere, employment creation was more prevalent in big city locations. Turning to the constraints regression analysis, we find once again that India does not reflect the dualism pattern found elsewhere in low income countries. The typical pattern is that it is firms in urban (large city) locations that report higher costs of constraints than do firms in rural areas including small cities. Indeed for electricity, Indian firms in more rural locations report higher constraints than do firms in urban areas. In India, large-city firms report higher constraints only in relation to transport, courts, labor regulation, and customs administration. However, for the majority of constraints, there is no significant large city – small city gap in the reported cost of constraints in India. The standard dualism pattern for low-income countries (found in the large multi- region dataset) is characteristic of most SAR countries apart from India. It is especially pronounced in Bangladesh and Pakistan (red cells). This suggests that in terms of the reported constraints on growth, India is a more integrated economy between rural, small city and large city areas than are the other countries in this sample. The positive firm size effects across a range of constraints are weaker in India than elsewhere. The difference between services and manufacturing firms is also different in India, where it is services firms that complain more about electricity. Unlike the case elsewhere, they are also more likely to have a generator and have more frequent power outages than do manufacturing firms. In Pakistan, services firms complain much less about electricity than do manufacturing firms and are more likely to have a 23 generator and fewer outages. However, services firms in Pakistan report more problems in relation to the availability of educated labor than do manufacturing firms. The other key difference is in relation to exporting firms: in the sample as a whole (and elsewhere in the world) exporting firms tend to complain more about physical infrastructure and institutions than do non-exporters. In India these exporter effects are even stronger. In Bangladesh by contrast, the results are the opposite: exporters report lower costs of constraints relative to non-exporters for electricity and most aspects of institutional infrastructure. Thus, in relation to exporters, it is Bangladesh that stand out from the regional and multi-region sample: for example, exporters complain less about customs administration than do non-exporters. In Bangladesh, also in contrast to experience elsewhere, exporters complain less about macroeconomic instability than do non-exporters. Although the constraints reported by Bangladeshi importers are very much in line with the sample overall, Pakistani importers stand out from the sample. Importing firms there generally report lower constraints than non-importers and than importing firms elsewhere in the region. These differences point to the need for more detailed research to uncover what lies behind them. Are there important industry-specific effects? An important question for policy-makers is whether there are systematic differences across industries in the extent to which the quality of public inputs imposes costs on firms. We focus on six industries (garments, food and beverages, chemicals, electronics, machinery and textiles) and use the samples from the three large countries (Bangladesh, India and Pakistan) where each of the six industries is well-represented. As a first step, we look at whether job creation is concentrated in particular industries. We do this by using the pooled sample for the six industries in the three countries and, for each industry separately, regress the job creation variable on the dummy variable for the industry of interest plus country fixed effects. The results are in Table 14 in the “corr� column. They show that job creation is higher in garments and textiles (except in India) and lower in food and beverages, and electronics (except in Pakistan). It is lower in the machinery sector in Pakistan. 24 However, from a policy perspective, what matters is whether these industry differences persist once we control for the other important firm characteristics, i.e. size, age, etc. It is very striking that once the standard set of firm-level controls are introduced, the differences across industries in job creation mainly disappear. This is shown in the column “pcorr� in Table 14, where for the pooled sample, only the electronics industry dummy retains its significance. There are few significant country differences. Strikingly in the case of Bangladeshi garments, its role as especially job- creating in the initial regression without the controls switches to the opposite once we control for the firm characteristics – it is then less likely to be job-creating than other industries. Finally, when we look at how reported constraints vary by industry either in the pooled sample or by country, there are very few significant industry effects. The implication of these results for the policy-maker is that there is no basis for discriminating across industries when identifying priorities for improving the business environment, once the key characteristics we have analyzed here – international engagement, size, location, etc. – have been accounted for. Summary We summarize the core results of this section by returning to the predictions of the modeling framework. Using our baseline performance indicator of job creation, the model predicts that firms that expanded employment will: • Report higher costs of public input constraints Rij , i.e., higher shadow prices. Such firms report higher costs of constraints in fourteen of the sixteen dimensions (including adequacy of water supplies). ∗ • Spend more on mitigation, Gij . The results reported above confirm that such firms are more likely to have a generator and to pay bribes. Contrary to expectation, they are not more likely to spend management time dealing with officials. ∗ • Report better public input services Eij (in cases where the input service is not a function of firm-level productivity). We find that such firms do not report 25 shorter delays at customs (although higher productivity firms do report shorter delays). These results provide a strong case for a policy focus on the constraints identified by firms as most costly. In the next piece, we examine how the relative importance of different public input constraints reported by firms varies across countries in the sample. 8. Which elements of the business environment matter most for firms and how do they vary across countries? In this section, we use the country averages of the costs of public inputs reported by firms, Rij , to assess which elements of the external environment are more problematic for firms across the region and in each country. We also compare the country average evaluations with those in other countries at a similar level of development. The data that we use in this section pools the surveys of the SAR countries used above with data from the surveys available from the World Bank’s Enterprise Surveys portal. We consider only formal sector firms in the analysis. Altogether, the surveys cover almost 120,000 firms from over 230 different surveys in 126 countries over the period 2000-2010. Of these, about 16,000 firms were from the 8 SAR countries. We use these firm-level data to calculate country average evaluations, i.e., country means. In order to correct for differences in survey samples when comparing reported constraints across countries, we construct the “conditional country means� for each constraint for a standard firm. The controls included in the estimations are the same as those used above: size (log employment), and dummy variables for whether the firm is newly established, expanding employment, has substantial foreign ownership, is a significant exporter or importer, and is located in a small city. 6 Log employment is centered on ln(30). The estimations are identical to the country regressions reported earlier; the country conditional means are, in fact, the estimated intercepts 6 Also as previously, we include a dummy variable in case the importer dummy is missing, in order to increase the sample size and number of surveys included. 26 reported in the first column of Table 13. The intercepts can therefore be interpreted as estimates of the constraint level in a given country for a benchmark manufacturing firm with 30 employees that is domestically owned, with no foreign ownership or import/export activity, and that is located in a large or capital city. These intercepts are the “country conditional means� that we analyze below. 7 For those SAR countries for which we have surveys in multiple years, we obtain year-specific conditional means from the means by year of the residuals from the firm-level regression. The implication of the modeling framework set out above is that job-creating firms are likely to benefit most from the success of policy-makers in relaxing the most pressing external environment constraints. For the reasons discussed earlier, we focus on business environment constraints with a public good character and therefore do not include the “Tax Rates� or “Access to Finance� constraints in the main analysis. The latter measures are considered briefly in Appendix 2. The ranking of constraints by country The conditional country means are reported for each of the SAR countries and for the region as a whole in Table 15. The countries are listed according to GDP per capita. Because the conditional means are sometimes based on surveys from more than one year, GDP per capita used in the table and in the diagrams is a weighted average of GDP per capita, where the weights are the survey sample size in the relevant sample years. For the region as a whole, political instability, electricity and corruption are the three top-ranked constraints. However, there is considerable variation across the countries. The differences between “finance� and the public good constraints discussed above mean it is not possible to interpret the country conditional mean for finance as comparable with the other constraints. As an example, think of the following experiment. If the economy is affected by a positive productivity shock, we would expect the average firm to report a higher cost of a public good constraint but lower cost of access to finance. Since the responses to the finance question are non- 7 The results for the unconditional country means, i.e., when using simple means across firms in a given survey, are very similar. 27 comparable with the answers to the other constraint questions, it is not possible to create a ranking of constraints that includes finance. Finance is therefore not reported in Table 15. For each country, the country mean score for each constraint is reported and the constraints are ordered from the most to the least costly. • In every country except Bhutan and the Maldives, electricity is one of the two highest ranked constraints. • In five of the eight countries, corruption is among the four highest ranked constraints. • Political instability is in the top three except in Bhutan– but was not included in the survey instrument for India, Sri Lanka and the Maldives. Nepal Political instability is the top-ranked constraint in Nepal. This is followed by electricity, transport and corruption. Only in Bhutan is transport also one of the top- ranked constraints. Least problematic are access to land, business licensing and the courts. Nepal was surveyed in 2000 and 2009. Electricity and transport were included in the two surveys and reported constraints went up in both cases. Bangladesh In Bangladesh, the ranking is electricity, political instability, corruption, and then access to land. Least problematic to firms are transport, the courts, and labour regulation. Surveys were conducted in Bangladesh in 2003 and 2007. Most aspects of the business environment were included in each survey. There were few substantial changes over time: although concern with customs, transport, tax administration and anti-competitive practices went down. 28 Afghanistan Political instability is the top-ranked constraint in Afghanistan, followed by electricity, corruption, and access to land. Problems with business licensing, access to skilled labor and labor regulation come at the bottom. Surveys were conducted in Afghanistan for the years 2006, 2009 and 2010. Only a handful of the set of constraints were included in all three surveys. All three surveys were relatively small, however, each with only about 100-200 firms used in the estimations. No clear trends emerge. We note that the prevalence of power outages more than once a month fell sharply between 2009 and 2010 in the unconditional mean estimate but once sample composition is controlled for, there is no difference between the two years. This highlights the importance of controlling for the influence of sample composition. Pakistan The top-ranked constraint in Pakistan is tax administration, followed by a group comprising electricity, political instability, government policy uncertainty, the courts and corruption. The only other SAR country where the courts are highly ranked as a problem is the Maldives. At the bottom of the ranking in Pakistan are transport, access to skilled labor and telecoms. Pakistan was surveyed in 2002, 2007 and 2010. As elsewhere in the SAR, only a handful of constraints were included in the 2010 survey. The reported cost of electricity as a constraint went up over time. By contrast, concern with corruption had a hump-shaped pattern, with the peak in 2007. India Consistent with their importance elsewhere in the region, electricity and corruption are ranked top in India. The next most costly elements of the business environment are tax administration and labor regulation. Access to land is a problem elsewhere in the region but in India, it is one of the least problematic aspects of the business environment, along with courts and telecoms. 29 For India, there are surveys from the years 2002, 2003, 2005 and 2010. Only a handful of constraints were included in the 2010 survey. However, for those that were included, a sharp increase in problems is reported. In particular, electricity emerged clearly as the most costly constraint for firms in 2010. It is interesting to note that the likelihood of power outages remained fairly constant between 2005 and 2010. This is consistent with higher reported costs from electricity as an obstacle to production if firms were improving their performance / growing faster and hence suffering larger losses from an unchanged level of unreliability of the electricity grid. Labor regulation is also reported to be markedly more problematic in 2010 than in 2005. Sri Lanka The top five ranked constraints in Sri Lanka are electricity, government policy uncertainty, macroeconomic instability, anti-competitive practices, and labor regulations. Firms were not asked for their evaluation of the extent to which policy uncertainty and macro instability affected their business in Afghanistan, Bhutan and Nepal. Among the countries where these aspects were included, Sri Lanka is the only one where they are highly ranked as problems. Firms in Sri Lanka rated business licensing, customs, and access to land as least problematic. (A question about the obstacles posed by the courts was not included in the surveys in Sri Lanka.) Sri Lanka was surveyed in 2004 and 2010 but few aspects of the business environment were included in 2010; there were no marked changes. Maldives Given its physical environment, it is not surprising that the top-ranked constraint in the Maldives is access to land. This is followed by corruption, the courts, and access to skilled labor. The least problematic aspects of the business environment in the Maldives are telecoms, macro instability, and business licensing. The Maldives was only surveyed once, and the survey size was small (about 140 firms were used in the calculation of most of the conditional means), so these results should also be treated cautiously, despite their plausibility. 30 Bhutan The business environment in Bhutan (which was only surveyed once, in 2009) impinges on firms quite differently from elsewhere in the region. The top-ranked constraints are transport, access to skilled labor regulation, and tax administration. Of least concern are the courts, telecoms and political instability. The contrast between Bhutan and Nepal in the importance of electricity as an obstacle is particularly striking; across the SAR countries, the constraint is lowest in Bhutan and second highest in Nepal. This is not explained simply by the abundance of hydropower in Bhutan (about 4/5ths of total generation), since Nepal has the second-most abundant hydropower in the region (over 1/4th of total generation). The size of the Bhutan sample was relatively small, however – about 250 firms were used to calculate the country conditional means – so the above results should be treated cautiously. 9. How do the SAR countries compare with those outside the region? It is useful to compare the constraints reported by firms in the SAR countries with those outside the region but at similar levels of GDP per capita. The non-SAR country conditional means are constructed using the same set of controls as for the SAR countries, but in a single regression with country-survey fixed effects. The fixed effects correspond to the conditional means for individual country surveys. For compatibility with the results reported above, we continue to use a single conditional mean for each SAR country, i.e., we do not disaggregate by survey. The SAR and non-SAR conditional means are reported for the main constraints of interest in Figures 1-9 below (results for the other constraints are reported in Appendix 2). The log of GDP per capita in PPPs is on the horizontal axis and the reported cost of the constraint on the vertical. Heteroskedasticity-robust confidence intervals of 5% are shown individually for the SAR countries; these are simply the confidence intervals for the estimated constants in Table 13. The figures also show a regression line based on a quadratic in log GDP per capita for the country-survey conditional means for the rest of the world. A 95% confidence interval based on the 31 standard error of the predicted mean is also displayed; this confidence interval is robust to heteroskedasticity and within-country correlation. 8 Electricity, outages and generators There is a clear downward-sloping gradient in the electricity chart (Fig 1): firms in richer countries report lower obstacles to production from electricity. The high level of constraint reported in Nepal, Afghanistan and Bangladesh are partly accounted for by these being very poor countries. However, even for such poor countries, the problems look serious. This is particularly striking for Nepal, which as noted above is well-endowed with hydro power. Outages are also very high for these three countries (Fig 1). The constraints imposed on firms by unreliable electricity in Pakistan and India are, respectively, at the upper and lower ends of the confidence interval for countries at their levels of GDP per capita. Sri Lanka looks unusual in that it reports costs of constraints on the high side for its level of development but low outages. However, this may be partly accounted for by its higher use of generators. In general, use of generators is higher in the SAR countries than is typical elsewhere. Fig. 1 highlights the fact that the low reported constraints in the Maldives and Bhutan (although imprecisely estimated because of small sample sizes) are partly related to their higher income per head than the other countries in the region. But the chart also indicates that these two countries are at the low end of problems with electricity (including outages) compared with other countries at their level of development. Corruption and bribes The relationship between the reported cost of the corruption constraint and GDP per capita is characterized by low levels and low variation in reported constraints in rich countries and much more variation in poor and middle-income ones (Fig. 2). Of the richer SAR countries, Sri Lanka and Bhutan have lower levels of reported constraints than is typical at that level of GDP per capita. This appears true of India and 8 A non-overlap of a SAR country confidence interval and the non-SAR country regression line confidence interval provides a conservative test of the difference between the estimated means. That is, non-overlap suggests a statistically significant difference, whereas the estimated means could still be significantly different even if the confidence intervals overlap. 32 (possibly) Nepal as well. Bangladesh has higher levels than typical, whereas Afghanistan and. Pakistan look unexceptional. The data on the prevalence of bribes line up with those on corruption, with a high prevalence of firms making bribe payments in Bangladesh even for such a poor country and a low prevalence in Nepal. The low cost of constraint reported in the richer SAR countries of Sri Lanka and Bhutan is reflected in a low prevalence of bribes for countries at this level of GDP per capita. Political instability Among the SAR countries in which the role of political instability was included in the survey, there is a very clear downward income-constraint locus. The richer country Bhutan registers a very low cost of this constraint (in the cross-country comparison), which contrasts with the higher than typical costs reported in Nepal, Afghanistan and Bangladesh. Pakistan is also on the high side for countries at its level of GDP per capita. Access to Land There is an interesting degree of variation in the evaluation of problems related to access to land among the SAR countries (Figure 4). Across countries in the world, the constraint slopes downward as income per capita rises. Among the SAR countries, four have predicted values above the regression line and four, below. There are richer and poorer countries in each group. In the “high� group are Bangladesh, Afghanistan, Pakistan and the Maldives. The exceptional problems of the Maldives are highlighted in the chart. Afghanistan and Bangladesh (and Pakistan in the 2007 survey) also look unusual relative to other poor countries in the extent of access to land problems. The contrast with Nepal is very clear. In India and Sri Lanka access to land appears to be less problematic as compared with countries at a similar level of development. Water Problems with the availability of water (Figure 5) tend to be high in the SAR countries as compared with countries at similar levels of development (the exception is the Maldives). The problems in Pakistan are especially notable. 33 Labor regulation The chart (Figure 6) suggests that this element of the business environment becomes more costly to firms as GDP per capita rises. Most of the SAR countries report elevated levels of this constraint as compared with other countries at their level of development. This is especially in the case for Nepal, India and Sri Lanka. Courts Problems with the courts look unusually serious in Pakistan (Figure 7). By contrast, this dimension of the business environment appears less problematic than might be expected at their level of GDP per capita in Nepal, India and Bhutan. The contrast between Pakistan and India along this dimension of the business environment is very striking. Inadequately educated labor Firms in the SAR countries do not identify problems posed by inadequate access to skilled labor as especially serious – as compared with countries at their level of development (Figure 8). The only exception is the case of Bhutan. Indeed for Nepal, Afghanistan, India and Pakistan, the costs of this element of the external environment are rated below that of their comparators. Telecoms The chart (Figure 9) highlights the fact that for most countries in the world, telecoms is a low-ranked constraint. This is also true of firms in the South Asia region. Problems appear more serious in Bangladesh, Afghanistan and Sri Lanka than is typical of countries at their levels of development. 10. A comparison of business environment constraints between formal sector firms and rural and informal sector firms Surveys have been conducted of rural firms in Bangladesh, Pakistan & Sri Lanka and of informal sector firms in Afghanistan, India and Pakistan. However, the sample for India is much larger (over 2,000 firms as compared with about 200 and 400 firms for the other two countries with informal sector surveys). We therefore confine the 34 comparison of informal firms to India. We compare the reported constraints on firm growth for rural and informal sector firms with those of formal sector firms, which were used in the previous analysis in the paper. The informal sector survey of India includes only manufacturing firms. For the formal/rural comparison, we pool the rural sector firms for the three countries. For both the formal/rural and formal/informal comparisons, the comparison group are the firms in the Enterprise Surveys for the relevant set of countries, which we call “formal�. Only manufacturing firms in the Enterprise Surveys are included for the comparison with informal firms. We do two exercises for the formal/rural and formal/informal comparisons. The first is to compare the levels of the reported constraints in the two groups of firms (Tables 16 and 18. The second is to examine whether there are any differences between formal and rural or informal sector firms in how constraints are ranked (Tables 17 and 19). In the left hand panel of Table 16 (specification 1), we show the unconditional mean evaluations of each obstacle for formal sector firms and informal or rural sector ones. To facilitate comparison between the formal and informal/rural sectors, we show the results for small formal sector firms (dropping from the samples those firms with more than 20 employees). The column headed “Diff?� shows whether there is a statistically significant difference between the mean for the small formal and the informal/rural sector firms. The right hand panel of Table 16 (specification 2) includes a size control for firm characteristics interacted with an informal or rural dummy (plus country-survey fixed effects). This allows the two groups of firms being compared (formal vs. informal and formal vs. rural) to have their own coefficients on the size control. We can interpret the constant or the constant plus the dummy as means conditioned on size. The benchmark firm is the same as in the rest of the analysis except that it is centered at average employment of 5 persons, which is more appropriate given the size distribution of informal and rural firms. In the comparison of small formal vs. rural 35 firms in Bangladesh, India and Sri Lanka, median employment in small formal firms is 10 (compared to 35 employees for all formal firms in these three countries), and in rural firms is 1.5. In the comparison of small formal vs. informal manufacturing firms in India, median employment in small formal firms is again 10 (compared to 20 employees for all formal manufacturing firms in the India surveys), and in informal firms is 4. The remaining size disparity between small formal firms, on the one hand, and rural or informal firms, on the other, is why we condition on size in specification 2 as described above. Levels comparisons between rural and formal sector firms The results for the comparison of rural with formal sector firms are clear and echo the results reported above in Section 7 for the large city / small city comparison using the Enterprise Survey data (Table 16). Rural firms almost always report that the elements of the business environment represent less serious obstacles to their firm’s production and growth than is the case for managers of formal sector firms. This is highlighted by the predominance of red shaded cells in both specifications. (Note that the number of observations for “Competition� and “Crime, theft and disorder� is very small and these results should be treated with caution.) The results in the lower part of the table are interesting because they accord with what one would expect of rural firms: they have much lower engagement with officials (bribes, inspections, management time) and report lower costs of corruption and the range of institutional constraints. Rural firms report similar levels of power outages and they are more likely to have a generator than formal sector firms. They rate the problems associated with access to electricity as less problematic than do formal sector firms, which suggests that they make less use of electricity in their production process. Comparison of rankings of constraints by rural and formal sector firms Table 17 presents the results in a slightly different way in order to highlight how the different firm types rank the constraints. Using the unconditional means in the first column, we see the results familiar from above, namely that electricity, political instability, and corruption are the highest ranked constraints for formal sector firms. 36 In the second column, we see that little changes if we restrict attention to the small formal sector firms. The constraints most important to rural firms are rather different. Three of the top four constraints are common to the two groups of firms: electricity, macroeconomic instability and political instability. The most striking difference is that for rural firms, corruption is a much lower ranked constraint: this is consistent with the results reported above that rural firms are less engaged in bribes and spend much less time with officials. Rural firms rank transport more highly as a constraint than do formal sector firms. Thus even though rural firms rate transport as less problematic for their business than do formal sector firms, they rank it more highly as a concern. The former result is consistent with the model used in the paper: formal sector firms value the losses associated with unreliable transport more highly than do rural firms. However, rural firms rank transport more highly, which likely reflects the fact that the transport infrastructure is considerably poorer in rural areas and that among the various business environment constraints, this one is especially burdensome. The higher ranking of anti-competitive practices and crime, theft and disorder by rural than formal sector firms is based on a very small sample and probably not much can be inferred safely from this. Levels comparisons between informal and formal sector firms (Manufacturing & India only) It is important to keep in mind when reviewing these results that they are drawn from a single country (India) and relate only to manufacturing sector firms. In contrast to the results for the rural firms, informal sector firms report higher constraints in 12 of the 17 cases (Table 18). They report lower constraints in two (electricity and customs) and the same level for corruption. Informal sector firms report a lower engagement in bribes and spend less management time with officials – as would be expected. However, they report the same number of 37 inspections as do formal sector firms. They report more frequent power cuts but are no more likely to have a generator than formal sector firms. Comparison of rankings of constraints by informal and formal sector firms In terms of rankings (Table 19), we note first that political instability was absent from all of the India surveys. Electricity, corruption and tax administration were the three highest ranked constraints among formal sector firms (“all� and “small�). The ranking of constraints by informal sector firms was quite different. Electricity was in the middle of the ranking – so although informal firms reported more power cuts and there was the same prevalence of generators, their lower ranking of electricity suggests they were typically much less dependent on it in their production process than were formal sector firms. The top-ranked constraints for informal sector firms were transport, business licensing, inadequately educated labor force and access to land. Access to land is ranked much lower by formal sector firms – as are business licensing and transport, and to a lesser extent access to educated labor. It seems that informal firms answering the question about the obstacles to the growth of their firm are reflecting on the core factors of production that limit them attaining scale and perhaps making it worthwhile attaining “formal� status. 11. Conclusion We have used firm-level data from surveys of firms in the South Asia Region to assess the constraints presented to firms by shortcomings in the provision of public inputs. As predicted by the modeling framework, better performing firms, and notably those that are job-creators, are more constrained by these shortcomings than are other firms, and are more likely to be engaged in activity to mitigate these shortcomings. We also find that characteristics associated with high productivity, such as export activity and firm size, are also associated with higher reported constraints. We also find significant variation across countries in terms of which firm characteristics are associated with constraints originating with which public inputs, 38 and in terms of which public input constraints firms rank as most important in their countries. 39 Table 1: Sample sizes by country and year Country 1999 2001 2002 2003 2004 2005 2006 2008 Total Afghanistan 325 640 965 Bangladesh 1,085 1,504 2,589 Bhutan 250 250 India 886 1,825 260 2,286 5,257 Maldives 145 145 Nepal 222 368 590 Pakistan 964 931 1,895 Sri Lanka 425 425 Total 1,108 2,789 1,345 425 2,431 1,256 1,504 1,258 12,116 Table 2: Median employment by country and year Country 1999 2001 2002 2003 2004 2005 2006 2008 Total Afghanistan 15 10 11 Bangladesh 153 45 80 Bhutan 15 15 India 40 18 28 17 20 Maldives 28 28 Nepal 82 15 25 Pakistan 26 12 20 Sri Lanka 113 113 Total 48 20 108 113 17 13 45 11 24 Table 3: Means of measures of firm productivity and growth Country (1) (2) (3) (4) (5) (6) (7) (8) (9) Afghanistan 0.23 9.37 3.07 0.79 na na 10.81 7.4 0.13 Bangladesh 0.41 8.27 2.84 0.32 0.45 5.83 11.97 7.06 0.2 Bhutan 0.64 9.61 na na na na na na 0.26 India 0.38 9.45 3.11 0.28 0.53 32.7 13.38 10.72 0.18 Maldives 0.61 9.63 3.56 na 0.44 43.92 11.55 9.33 0.69 Nepal 0.43 8.61 2.76 na 0.21 na 12.47 7.63 0.1 Pakistan 0.37 9.55 3.27 0.47 0.12 35.46 12.65 6.87 0.11 Sri Lanka 0.51 8.78 3.07 0.1 0.21 28.95 11.44 9.05 0.4 Total 0.39 9.16 3.07 0.31 0.44 31.7 12.63 8.92 0.18 Notes: (1) Expanding employment; (2) log VA/L; (3) TFP; (4) R&D; (5) Sales to MNCs; (6) Education of the top manager; (7) education of the workforce; (8) training programme in place; (9) trained workers. 40 Table 4: Correlations of expanding permanent employment with other indicators of firm productivity and growth expand_p log_lprod tfp rd newPP sales_MNCedu_topmaedu_prodWftrain pct_prod_workers_train expand_p . Pos Pos Pos Pos Pos Pos Pos Expanding permanent employment N 7,824 6,662 9,117 9,633 2,352 11,007 9,587 11,468 6,885 log_lprod . Pos Pos Pos Pos Pos Pos Pos Pos Log VA/L N 7,824 6,463 6,031 6,405 2,112 7,282 6,503 7,415 4,351 tfp Pos . Pos Pos Pos Pos Pos Pos Pos TFP N 6,662 6,463 5,577 5,547 1,860 6,444 5,730 6,357 4,110 rd Pos Pos Pos . Pos Pos Pos Pos Pos Pos Firm does R&D N 9,117 6,031 5,577 7,959 2,279 8,835 8,228 8,973 6,085 newPP Pos Pos Pos Pos . Pos Pos Pos Pos Introduced new product/process N 9,633 6,405 5,547 7,959 2,278 9,164 8,366 9,155 5,926 sales_MNC Pos Pos Pos Pos Pos . Pos Pos Pos Percent of sales to MNCs N 2,352 2,112 1,860 2,279 2,278 2,332 2,177 2,320 367 edu_topma Pos Pos Pos Pos Pos Pos . Pos Pos Pos Education level of top manager in years N 11,007 7,282 6,444 8,835 9,164 2,332 9,070 10,515 6,548 edu_prodW Pos Pos Pos Pos Pos Pos . Pos Pos Average education level of the production workforce in years N 9,587 6,503 5,730 8,228 8,366 2,177 9,070 9,554 6,135 ftrain Pos Pos Pos Pos Pos Pos Pos Pos . Pos Training offered N 11,468 7,415 6,357 8,973 9,155 2,320 10,515 9,554 6,565 pct_prod_w Pos Pos Pos Pos Pos Pos Pos . % of production workers that were trained N 6,885 4,351 4,110 6,085 5,926 367 6,548 6,135 6,565 Notes: Partial correlations; country fixed effects are partialled out. “Pos� and green indicates a positive correlation significant at the 5% level. 41 Table 5: Correlations of expanding permanent employment with firm characteristics Characteristic Country corr N pcorr N Size (log average L) All 0.049** 11,953 0.035** 11,893 New firm All -0.056** 12,139 -0.039** 11,893 Services All -0.009 12,201 0.045* 11,893 Foreign All 0.050 12,201 -0.012 11,893 Exporter All 0.143** 12,201 0.087** 11,893 Importer All 0.121** 12,201 0.053** 11,893 Small city or rural All -0.027** 12,201 -0.003 11,893 corr = Controls are country fixed effects. pcorr = Controls are country fixed effects plus other firm characteristics Notes: Green indicates a positive correlation; red a negative correlation. *=significant at the 5% level; **=significant at the 1% level. Table 6: Means of firm controls Country (1) (2) (3) (4) (5) (6) (7) Afghanistan 2.59 0.51 0.77 0.05 0.06 0.34 0.06 Bangladesh 4.38 0.13 0.08 0.02 0.34 0.40 0.18 Bhutan 2.85 0.24 0.65 0.06 0.18 0.65 0.71 India 3.38 0.15 0.01 0.02 0.16 0.14 0.31 Maldives 3.39 0.29 0.66 0.00 0.09 0.66 1.00 Nepal 3.35 0.21 0.19 0.05 0.18 0.09 0.52 Pakistan 3.16 0.11 0.05 0.02 0.16 0.13 0.21 Sri Lanka 4.80 0.08 0.00 0.19 0.68 0.42 0.76 Total 3.54 0.17 0.12 0.03 0.21 0.23 0.29 Notes: (1) Log employment; (2) New firm; (3) Services firm; (4) Foreign ownership >10%; (5) Exporting >10% of sales; (6) Direct importer; (7) Small city or rural location. 42 Table 7: Partial correlations of expanding permanent employment with other indicators of firm productivity and growth expand_p log_lprod tfp rd newPP sales_MNCedu_topmaedu_prodWftrain pct_prod_workers_train expand_p . Neg Pos Pos Pos Pos Pos Expanding permanent employment N 7,791 6,625 8,903 9,375 2,290 10,731 9,509 11,299 6,859 log_lprod Neg . Pos Pos Pos Pos Pos Pos Pos Pos Log VA/L N 7,791 6,437 6,006 6,373 2,095 7,252 6,476 7,383 4,339 tfp Pos . Pos Pos Pos Pos Pos TFP N 6,625 6,437 5,545 5,511 1,835 6,410 5,701 6,321 4,098 rd Pos Pos Pos . Pos Pos Pos Pos Pos Firm does R&D N 8,903 6,006 5,545 7,751 2,228 8,642 8,171 8,870 6,065 newPP Pos Pos Pos Pos . Pos Pos Pos Pos Introduced new product/process N 9,375 6,373 5,511 7,751 2,216 8,930 8,291 9,033 5,902 sales_MNC Pos Pos Pos Pos Pos . Pos Pos Pos Percent of sales to MNCs N 2,290 2,095 1,835 2,228 2,216 2,271 2,140 2,274 362 edu_topma Pos Pos Pos Pos Pos Pos . Pos Pos Education level of top manager in years N 10,731 7,252 6,410 8,642 8,930 2,271 9,000 10,378 6,527 edu_prodW Pos Pos Pos . Pos Average education level of the production workforce in years N 9,509 6,476 5,701 8,171 8,291 2,140 9,000 9,477 6,114 ftrain Pos Pos Pos Pos Pos Pos Pos . Pos Training offered N 11,299 7,383 6,321 8,870 9,033 2,274 10,378 9,477 6,540 pct_prod_w Pos Pos Pos Pos . % of production workers that were trained N 6,859 4,339 4,098 6,065 5,902 362 6,527 6,114 6,540 Notes: Controls are: country fixed effects; log employment (size); new firm; services; foreign; exporter; importer; missing importer indicator; small city. “Pos� and green indicates a positive correlation significant at the 5% level. “Neg� and red indicates a negative correlation significant at the 5% level. 43 Table 8: Means of evaluations of external constraints (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) (17) Afghanistan 2.69 1.58 1.41 1.37 1.43 2.12 1.32 1.17 2.95 2.25 1.09 1.98 0.38 0.99 na na 0.24 Bangladesh 2.93 1.37 1.26 1.56 1.33 1.96 1.99 1.32 2.80 2.46 1.21 1.91 0.83 1.52 1.84 1.83 0.22 Bhutan 0.84 0.53 1.30 1.09 0.58 0.86 1.29 1.20 0.20 0.74 0.50 1.67 1.31 1.35 na na 0.36 India 1.75 0.52 0.85 1.01 0.87 0.71 1.55 0.89 na 1.70 0.44 1.16 1.13 1.02 1.03 0.92 0.21 Maldives 1.13 0.81 1.25 1.12 1.07 2.27 0.59 0.47 na 1.87 1.69 2.59 1.23 1.93 1.52 0.87 0.01 Nepal 2.74 0.57 1.58 0.98 0.97 0.29 0.90 0.33 3.41 1.32 0.11 1.11 1.02 0.54 na na 0.21 Pakistan 2.33 0.75 0.94 1.02 1.28 1.45 1.84 1.05 2.15 2.21 2.00 1.64 0.88 0.95 1.96 2.12 0.45 Sri Lanka 2.04 0.95 1.16 0.88 1.07 0.50 0.80 0.50 na 1.04 na 1.14 1.37 1.32 1.66 1.60 0.35 Total 2.24 0.80 1.08 1.17 1.10 1.23 1.61 1.00 2.55 1.95 0.91 1.51 0.97 1.13 1.43 1.44 0.26 Notes: (1) Electricity; (2) Telecoms; (3) Transport; (4) Customs; (5) Competition; (6) Access to land; (7) Tax admin.; (8) Bus. licensing; (9) Political instability; (10) Corruption; (11) Courts; (12) Access to financing; (13) Labour regulations; (14) Inadequately trained labour; (15) Government policy uncertainty; (16) Macro instability; (17) Insufficient water supply. 44 Table 9: Means of measures of mitigation, public input services, and the shared business environment Country (1) (2) (3) (4) (5) (6) Afghanistan 0.49 12.25 0.75 n.a. 11.88 1.00 Bangladesh 0.86 4.01 0.62 16.59 8.72 0.96 Bhutan 0.05 17.17 0.23 n.a. 2.10 0.24 India 0.45 6.48 0.59 8.91 8.44 0.69 Maldives 0.38 6.33 0.41 2.98 4.67 0.04 Nepal 0.08 8.50 0.48 n.a. 10.08 0.99 Pakistan 0.46 2.62 0.36 9.52 10.01 0.56 Sri Lanka 0.16 3.84 0.75 16.16 7.80 n.a. Total 0.51 5.97 0.56 11.27 8.59 0.75 Notes: (1) Bribes; (2) Management time; (3) Firm owns a generator; (4) Number of inspections; (5) Days delayed in customs; (6) Power cuts more than once per month. 45 Table 10: Public input constraints and firm productivity/growth Sales to Man. educ L educ Expand L N Log VA/L N TFP level N R&D (1/0) N Innovate N MNCs N (years) N (years) N Electricity 0.081** 10,913 0.011 6,921 -0.003 5,825 -0.052 7,951 0.004 8,397 0.001 2,285 0.005 9,809 -0.014** 8,893 Telecoms 0.060** 9,192 0.020* 5,364 0.040* 4,328 0.022 6,746 0.044 6,945 0.000 2,266 0.015** 8,348 0.010* 7,535 Transport 0.140** 10,809 0.017 6,885 0.008 5,807 -0.061* 7,921 0.104** 8,343 -0.002** 2,284 -0.002 9,712 0.011** 8,852 Customs 0.112** 9,916 0.045** 6,408 0.001 5,453 0.107** 7,643 0.116** 7,652 -0.001 2,239 0.025** 8,932 0.026** 8,344 Competit 0.040 10,398 0.029* 6,493 0.045 5,452 0.087* 7,706 0.232** 7,901 -0.001 2,274 0.010* 9,351 0.021** 8,613 AccessLand 0.049* 10,590 -0.002 6,832 -0.021 5,777 0.212** 7,699 0.032 8,101 -0.002** 2,282 -0.016** 9,530 -0.005 8,611 TaxRates 0.114** 10,556 0.050** 6,649 0.055* 5,600 0.158** 7,893 -0.001 8,078 0.000 2,263 0.025** 9,510 -0.006 8,792 TaxAdministration 0.170** 10,530 0.048** 6,643 0.062* 5,594 0.164** 7,891 0.020 8,057 0.000 2,261 0.038** 9,487 -0.009* 8,772 BusLicensing 0.078** 10,437 0.022* 6,650 0.001 5,729 0.103** 7,660 0.039 7,973 0.000 2,197 0.013** 9,381 0.006 8,481 PoliticalInstability 0.040 3,607 0.030 2,650 0.001 1,958 0.107 1,388 0.052 2,378 n.a. - 0.012 2,839 -0.002 2,324 Corruption 0.216** 10,588 0.023 6,677 0.061* 5,615 0.183** 7,920 0.070* 8,143 -0.001 2,272 0.035** 9,537 -0.012* 8,835 Courts 0.098** 5,747 0.012 4,469 0.047 3,582 0.108* 3,479 0.019 4,418 0.001* 1,975 0.028** 4,921 0.016* 4,170 AccessFinancing 0.012 10,783 -0.029** 6,831 -0.063** 5,779 0.132** 7,939 0.020 8,300 -0.002** 2,278 -0.005 9,707 -0.002 8,810 LaborReg 0.054* 10,321 0.017 6,666 -0.012 5,617 0.066* 7,687 0.063* 7,877 -0.001 2,269 0.018** 9,307 0.000 8,590 InadEducLabor 0.131** 10,815 0.014 6,873 -0.003 5,804 0.081** 7,923 0.109** 8,324 0.002* 2,286 0.022** 9,734 0.001 8,827 GovPolicyUnc 0.114** 8,149 0.034** 5,492 0.016 5,084 0.194** 7,642 0.199** 7,186 0.000 2,255 0.020** 7,954 0.006 7,648 MacroInstability 0.109** 8,847 0.018 6,010 -0.011 5,303 0.192** 7,529 0.147** 7,885 0.000 2,245 0.019** 8,514 0.002 8,391 bribe 0.035** 9,711 0.014** 5,958 0.027** 5,051 0.059** 7,562 0.066** 7,403 0.000 2,233 0.010** 8,739 0.001 8,144 mng_time -0.252 11,534 0.019 7,528 -0.102 6,408 0.772** 8,830 -1.363** 9,074 0.014* 2,251 0.130** 10,437 -0.064* 9,442 generator 0.020* 10,779 0.044** 7,341 0.036** 6,280 0.076** 8,645 0.091** 8,980 0.002** 2,234 0.019** 9,864 0.010** 9,422 insufficient_water 0.035** 9,518 0.012** 6,769 0.018* 5,859 0.015 7,445 0.070** 7,878 0.000** 1,867 0.003 8,838 0.010** 8,216 num_insp 2.078** 7,995 0.392* 5,752 0.016 5,390 2.651** 7,581 -0.642 7,047 -0.008 1,835 0.216** 7,841 0.043 7,324 days_customs -0.495 2,926 -0.529* 1,948 -0.264 1,798 -0.798 2,657 -0.845 2,559 -0.007 381 -0.319* 2,757 0.053 2,728 m_power_out 0.052** 9,797 -0.001 6,434 0.000 5,369 -0.035** 7,397 0.049** 7,868 0.000 1,979 0.000 8,798 -0.010** 8,350 Notes: Controls are: country fixed effects; log employment (size); new firm; services; foreign; exporter; importer; missing importer indicator; small city. “Pos� and green indicates a positive coefficient; “Neg� and red indicates a negative coefficient. *=significant at the 5% level; **=significant at the 1% level. Standard errors are robust to heteroskedasticity. 46 Table 11: Public input constraints, firm growth and firm characteristics Country Indicator Size (log L) New firm Expanding L Services Foreign Exporter Importer Missing importer Small city Obs All Electricity -0.053** 0.004 0.081** -0.164** -0.089 0.079* 0.023 0.261** 0.043 10,913 All Telecoms 0.041** 0.092** 0.060** 0.142** -0.094 -0.055 0.146** 0.002 -0.014 9,192 All Transport 0.018 0.070* 0.140** -0.057 0.052 -0.030 0.270** -0.072 -0.101** 10,809 All Customs 0.101** 0.077* 0.112** -0.016 0.028 0.025 0.294** -0.265** -0.217** 9,916 All Competit 0.017 0.063 0.040 -0.111* 0.047 -0.172** 0.126** -0.013 -0.115** 10,398 All AccessLand -0.045** 0.080* 0.049* 0.062 0.000 0.072* 0.041 0.186** -0.048 10,590 All TaxRates 0.058** -0.094** 0.114** -0.061 -0.184* -0.079* 0.066 -0.079 -0.190** 10,556 All TaxAdministration 0.082** -0.059 0.170** 0.003 -0.143 -0.051 0.058 -0.322** -0.178** 10,530 All BusLicensing 0.024* -0.031 0.078** 0.052 0.023 0.095** 0.099** 0.110* -0.104** 10,437 All PoliticalInstability 0.011 -0.008 0.040 0.120* 0.221* 0.323** 0.067 0.028 -0.226** 3,607 All Corruption 0.038** -0.088* 0.216** -0.032 -0.050 0.038 0.163** 0.202** -0.154** 10,588 All Courts 0.049** -0.032 0.098** 0.129* 0.116 0.174** 0.031 -0.061 -0.121** 5,747 All AccessFinancing 0.000 0.081* 0.012 -0.317** -0.308** -0.111** 0.016 -0.257** -0.145** 10,783 All LaborReg 0.084** -0.032 0.054* -0.094* -0.123 0.083** 0.023 -0.263** -0.130** 10,321 All InadEducLabor 0.045** 0.016 0.131** 0.085 -0.107 -0.010 0.029 -0.153** -0.062* 10,815 All GovPolicyUnc 0.051** -0.079 0.114** -0.020 0.181 -0.047 0.145** 0.272** -0.189** 8,149 All MacroInstability 0.050** -0.046 0.109** -0.013 0.083 0.086* 0.204** 0.405** -0.110** 8,847 All bribe 0.010* -0.001 0.035** -0.018 -0.056* 0.029* 0.040** -0.157** -0.008 9,711 All mng_time 0.327** 0.446 -0.252 0.518 -0.176 1.221** 0.070 -1.034** -0.248 11,534 All generator 0.131** -0.012 0.020* 0.114** -0.039 0.111** 0.071** 0.010 0.093** 10,779 All insufficient_water 0.011** 0.015 0.035** 0.085** 0.002 -0.008 0.038** -0.180** -0.036** 9,518 All num_insp 2.568** 0.099 2.078** 2.322 1.496 1.977 2.848** 2.193 1.438 7,995 All days_customs 0.593** -0.492 -0.495 0.406 -2.921** 2.261** -1.465* -1.614 1.416* 2,926 All m_power_out -0.031** 0.000 0.052** -0.045** -0.045 0.015 0.027* 0.282** 0.076** 9,797 Notes: “Pos� and green indicates a positive coefficient; “Neg� and red indicates a negative coefficient. *=significant at the 5% level; **=significant at the 1% level. Standard errors are robust to heteroskedasticity. 47 Table 12: Country-specific correlations of expanding permanent employment with firm characteristics Characteristic Country corr Diff? N pcorr Diff? N Size (log average L) All 0.049** n.a. 11,953 0.035** n.a. 11,893 New firm All -0.056** n.a. 12,139 -0.039** n.a. 11,893 Services All -0.009 n.a. 12,201 0.045* n.a. 11,893 Foreign All 0.050 n.a. 12,201 -0.012 n.a. 11,893 Exporter All 0.143** n.a. 12,201 0.087** n.a. 11,893 Importer All 0.121** n.a. 12,201 0.053** n.a. 11,893 Small city or rural All -0.027** n.a. 12,201 -0.003 n.a. 11,893 Size (log average L) Afghanistan -0.005 Smaller 941 0.006 Smaller 938 New firm Afghanistan -0.187** Smaller 982 -0.184** Smaller 938 Services Afghanistan 0.074* Greater 985 0.108** Greater 938 Foreign Afghanistan -0.067 Smaller 985 -0.031 938 Exporter Afghanistan -0.062 Smaller 985 -0.104 Smaller 938 Importer Afghanistan -0.052 Smaller 985 -0.037 Smaller 938 Small city or rural Afghanistan 0.234** Greater 985 0.214** Greater 938 Size (log average L) Bangladesh 0.079** Greater 2,589 0.045** 2,581 New firm Bangladesh -0.086** 2,598 -0.079** 2,581 Services Bangladesh -0.246** Smaller 2,606 -0.045 Smaller 2,581 Foreign Bangladesh 0.237** Greater 2,606 0.105 Greater 2,581 Exporter Bangladesh 0.188** Greater 2,606 0.047 2,581 Importer Bangladesh 0.204** Greater 2,606 0.054* 2,581 Small city or rural Bangladesh -0.250** Smaller 2,606 -0.125** Smaller 2,581 Size (log average L) Bhutan 0.027 250 0.015 250 New firm Bhutan -0.004 250 0.031 250 Services Bhutan 0.192** Greater 250 0.169* 250 Foreign Bhutan -0.038 250 -0.066 250 Exporter Bhutan -0.125 Smaller 250 -0.028 250 Importer Bhutan 0.140* 250 0.114 250 Small city or rural Bhutan -0.180** Smaller 250 -0.140* Smaller 250 Size (log average L) India 0.038** Smaller 5,122 0.028** 5,089 New firm India -0.017 Greater 5,234 0.003 Greater 5,089 Services India -0.003 5,269 -0.048 5,089 Foreign India -0.031 Smaller 5,269 -0.116* Smaller 5,089 Exporter India 0.196** Greater 5,269 0.154** Greater 5,089 Importer India 0.110** 5,269 0.043 5,089 Small city or rural India 0.078** Greater 5,269 0.090** Greater 5,089 Size (log average L) Maldives 0.029 145 0.033 145 New firm Maldives 0.150 Greater 148 0.172* Greater 145 Services Maldives -0.079 148 -0.131 Smaller 145 Foreign Maldives 0.000 n.a. 148 0.000 n.a. 145 Exporter Maldives 0.001 148 -0.155 145 Importer Maldives 0.071 148 0.000 n.a. 145 Small city or rural Maldives 0.000 n.a. 148 0.000 n.a. 145 Size (log average L) Nepal 0.036* 590 0.045** 587 New firm Nepal -0.049 588 -0.032 587 Services Nepal 0.025 591 0.093 587 Foreign Nepal 0.199* 591 0.172 Greater 587 Exporter Nepal -0.002 Smaller 591 -0.028 Smaller 587 Importer Nepal -0.047 Smaller 591 -0.022 587 Small city or rural Nepal -0.033 591 -0.031 587 Size (log average L) Pakistan 0.052** 1,891 0.032** 1,881 New firm Pakistan -0.039 1,890 -0.050 1,881 Services Pakistan 0.170** Greater 1,900 0.212** Greater 1,881 Foreign Pakistan 0.185* 1,900 0.067 1,881 Exporter Pakistan 0.058 Smaller 1,900 0.043 1,881 Importer Pakistan 0.133** 1,900 0.049 1,881 Small city or rural Pakistan -0.118** Smaller 1,900 -0.112** Smaller 1,881 Size (log average L) Sri Lanka 0.041** 425 0.032 422 New firm Sri Lanka 0.026 449 0.027 422 Services Sri Lanka 0.000 n.a. 452 0.000 n.a. 422 Foreign Sri Lanka -0.038 452 -0.112 422 Exporter Sri Lanka 0.080 452 0.018 422 Importer Sri Lanka 0.118* 452 0.108 422 Small city or rural Sri Lanka -0.134* Smaller 452 -0.073 422 Notes: corr=Controls are country fixed effects (“All� only). pcorr=Controls are country fixed effects (“All� only) plus other firm characteristics. Green indicates a positive correlation; red a negative correlation. *=significant at the 5% level; **=significant at the 1% level. 48 Table 13: Country-specific public input constraints, firm growth and firm characteristics Notes: “Pos� and green indicates a positive coefficient; “Neg� and red indicates a negative coefficient. *=significant at the 5% level; **=significant at the 1% level. Standard errors are robust to heteroskedasticity. Country Indicator Size (log L) Different? New firm Different? Expanding L Different? Services Different? Foreign Different? Exporter Different? Importer Different? All Electricity -0.053** n.a. 0.004 n.a. 0.081** n.a. -0.164** n.a. -0.089 n.a. 0.079* n.a. 0.023 n.a. All Telecoms 0.041** n.a. 0.092** n.a. 0.060** n.a. 0.142** n.a. -0.094 n.a. -0.055 n.a. 0.146** n.a. All Transport 0.018 n.a. 0.070* n.a. 0.140** n.a. -0.057 n.a. 0.052 n.a. -0.030 n.a. 0.270** n.a. All Customs 0.101** n.a. 0.077* n.a. 0.112** n.a. -0.016 n.a. 0.028 n.a. 0.025 n.a. 0.294** n.a. All Competit 0.017 n.a. 0.063 n.a. 0.040 n.a. -0.111* n.a. 0.047 n.a. -0.172** n.a. 0.126** n.a. All AccessLand -0.045** n.a. 0.080* n.a. 0.049* n.a. 0.062 n.a. 0.000 n.a. 0.072* n.a. 0.041 n.a. All TaxRates 0.058** n.a. -0.094** n.a. 0.114** n.a. -0.061 n.a. -0.184* n.a. -0.079* n.a. 0.066 n.a. All TaxAdministration 0.082** n.a. -0.059 n.a. 0.170** n.a. 0.003 n.a. -0.143 n.a. -0.051 n.a. 0.058 n.a. All BusLicensing 0.024* n.a. -0.031 n.a. 0.078** n.a. 0.052 n.a. 0.023 n.a. 0.095** n.a. 0.099** n.a. All PoliticalInstability 0.011 n.a. -0.008 n.a. 0.040 n.a. 0.120* n.a. 0.221* n.a. 0.323** n.a. 0.067 n.a. All Corruption 0.038** n.a. -0.088* n.a. 0.216** n.a. -0.032 n.a. -0.050 n.a. 0.038 n.a. 0.163** n.a. All Courts 0.049** n.a. -0.032 n.a. 0.098** n.a. 0.129* n.a. 0.116 n.a. 0.174** n.a. 0.031 n.a. All AccessFinancing 0.000 n.a. 0.081* n.a. 0.012 n.a. -0.317** n.a. -0.308** n.a. -0.111** n.a. 0.016 n.a. All LaborReg 0.084** n.a. -0.032 n.a. 0.054* n.a. -0.094* n.a. -0.123 n.a. 0.083** n.a. 0.023 n.a. All InadEducLabor 0.045** n.a. 0.016 n.a. 0.131** n.a. 0.085 n.a. -0.107 n.a. -0.010 n.a. 0.029 n.a. All GovPolicyUnc 0.051** n.a. -0.079 n.a. 0.114** n.a. -0.020 n.a. 0.181 n.a. -0.047 n.a. 0.145** n.a. All MacroInstability 0.050** n.a. -0.046 n.a. 0.109** n.a. -0.013 n.a. 0.083 n.a. 0.086* n.a. 0.204** n.a. All bribe 0.010* n.a. -0.001 n.a. 0.035** n.a. -0.018 n.a. -0.056* n.a. 0.029* n.a. 0.040** n.a. All mng_time 0.327** n.a. 0.446 n.a. -0.252 n.a. 0.518 n.a. -0.176 n.a. 1.221** n.a. 0.070 n.a. All generator 0.131** n.a. -0.012 n.a. 0.020* n.a. 0.114** n.a. -0.039 n.a. 0.111** n.a. 0.071** n.a. All insufficient_water 0.011** n.a. 0.015 n.a. 0.035** n.a. 0.085** n.a. 0.002 n.a. -0.008 n.a. 0.038** n.a. All num_insp 2.568** n.a. 0.099 n.a. 2.078** n.a. 2.322 n.a. 1.496 n.a. 1.977 n.a. 2.848** n.a. All days_customs 0.593** n.a. -0.492 n.a. -0.495 n.a. 0.406 n.a. -2.921** n.a. 2.261** n.a. -1.465* n.a. All m_power_out -0.031** n.a. 0.000 n.a. 0.052** n.a. -0.045** n.a. -0.045 n.a. 0.015 n.a. 0.027* n.a. 49 Table 13 (continued): Country-specific public input constraints, firm growth and firm characteristics Country Indicator Size (log L) Different? New firm Different? Expanding L Different? Services Different? Foreign Different? Exporter Different? Importer Different? Afghanistan Electricity 0.023 -0.105 -0.231* Smaller -0.358** Smaller 0.386 Greater -0.206 0.134 Afghanistan Telecoms 0.015 0.138 -0.160 Smaller 0.147 -0.246 0.452* Greater 0.252* Afghanistan Transport 0.088 -0.059 0.209 0.166 Greater -0.129 0.414* Greater 0.320** Afghanistan Customs 0.146** 0.067 -0.163 Smaller -0.131 0.105 0.371* 0.822** Greater Afghanistan Competit 0.124* Greater 0.234* -0.203 Smaller -0.164 -0.302 0.332 Greater 0.128 Afghanistan AccessLand 0.051 Greater 0.191 -0.049 -0.005 -0.012 0.227 -0.137 Afghanistan TaxRates 0.056 0.035 -0.130 Smaller -0.014 -0.591** Smaller -0.079 0.380** Greater Afghanistan TaxAdministration 0.074 0.224* Greater -0.018 Smaller 0.069 -0.383 0.094 0.231* Greater Afghanistan BusLicensing -0.118** Smaller -0.051 0.173 0.234* Greater -0.277 0.160 0.360** Greater Afghanistan PoliticalInstability 0.158** Greater -0.030 -0.209* Smaller 0.068 0.381 0.146 0.124 Afghanistan Corruption 0.048 -0.025 -0.023 Smaller -0.060 -0.661** Smaller 0.361* 0.402** Greater Afghanistan Courts 0.040 -0.064 0.204 -0.080 Smaller -0.177 0.152 0.128 Afghanistan AccessFinancing 0.022 0.130 -0.177 -0.225 -0.635** 0.173 0.128 Afghanistan LaborReg 0.030 Smaller 0.020 0.078 0.001 0.056 0.072 0.105 Afghanistan InadEducLabor -0.058 Smaller 0.007 -0.025 0.243* 0.274 0.031 -0.106 Afghanistan GovPolicyUnc n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. Afghanistan MacroInstability n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. Afghanistan bribe 0.050** Greater 0.045 -0.189** Smaller -0.011 -0.098 0.101 0.274** Greater Afghanistan mng_time 2.341** Greater 3.420** Greater -3.444** Smaller 3.806* Greater 0.692 3.378 0.356 Afghanistan generator 0.025 Smaller -0.027 0.005 -0.134** Smaller 0.080 -0.092 Smaller 0.003 Afghanistan insufficient_water -0.022 -0.037 -0.076 0.183** Greater 0.072 0.068 0.068 Afghanistan num_insp n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. Afghanistan days_customs 4.173* -6.235* Smaller -1.905 3.943 -18.180* Smaller -1.817 3.969 Afghanistan m_power_out -0.004 Greater 0.006 0.005 Smaller -0.006 Greater 0.003 0.002 0.005 50 Table 13 (continued): Country-specific public input constraints, firm growth and firm characteristics Country Indicator Size (log L) Different? New firm Different? Expanding L Different? Services Different? Foreign Different? Exporter Different? Importer Different? Bangladesh Electricity -0.064** -0.062 0.018 0.005 -0.029 0.059 -0.056 Bangladesh Telecoms 0.086** 0.142 0.017 -0.294* Smaller 0.149 -0.219* Smaller 0.074 Bangladesh Transport 0.021 0.118 0.057 -0.404** Smaller 0.031 -0.097 0.240** Bangladesh Customs 0.117** 0.127 0.050 -0.264** Smaller -0.485** Smaller -0.298** Smaller 0.376** Bangladesh Competit -0.037 Smaller 0.097 -0.128* Smaller -0.275** Smaller -0.183 -0.395** Smaller 0.133 Bangladesh AccessLand -0.054* 0.118 -0.011 0.474** Greater 0.203 0.079 -0.076 Smaller Bangladesh TaxRates 0.088** -0.205** -0.083 Smaller 0.098 0.034 -0.245** Smaller 0.067 Bangladesh TaxAdministration 0.128** Greater -0.142 0.041 Smaller 0.191 -0.198 -0.234** Smaller 0.158* Greater Bangladesh BusLicensing 0.009 0.084 Greater -0.001 Smaller 0.110 -0.011 0.156** -0.015 Smaller Bangladesh PoliticalInstability -0.042 Smaller -0.125 -0.024 0.164 0.430** 0.188** Smaller 0.164* Bangladesh Corruption 0.066** -0.190* 0.100 Smaller -0.102 0.285 Greater -0.040 0.101 Bangladesh Courts 0.020 -0.112 0.080 0.333** Greater 0.473* 0.187* -0.042 Bangladesh AccessFinancing -0.071** Smaller 0.143* -0.112* Smaller -0.450** Smaller -0.542** -0.192** 0.056 Bangladesh LaborReg 0.019 Smaller 0.042 -0.172** Smaller -0.218** Smaller -0.185 0.178** 0.099 Bangladesh InadEducLabor 0.003 Smaller 0.135* Greater 0.143** -0.150 Smaller 0.174 0.073 Greater 0.016 Bangladesh GovPolicyUnc 0.035 -0.142 0.045 0.118 0.238 -0.107 0.135* Bangladesh MacroInstability 0.083** -0.197* Smaller 0.075 0.123 0.173 -0.164** Smaller 0.133* Bangladesh bribe 0.046** Greater -0.070** Smaller -0.007 Smaller 0.045 Greater -0.080 -0.029 Smaller -0.002 Smaller Bangladesh mng_time 0.235** -0.217 0.837** Greater -1.301** Smaller -0.184 0.468 Smaller 1.042** Greater Bangladesh generator 0.173** Greater -0.019 0.007 n.a. n.a. -0.110* 0.026 Smaller 0.029 Bangladesh insufficient_water -0.006 Smaller 0.000 0.015 -0.086** Smaller 0.006 0.065** Greater 0.042 Bangladesh num_insp 1.836** -2.416 3.243** 1.056 10.461 1.672 5.527** Bangladesh days_customs 0.544 1.595 Greater -2.623** Smaller 10.019** Greater -0.676 0.488 Smaller -0.874 Bangladesh m_power_out -0.012** Greater -0.035* Smaller 0.023** Smaller 0.004 Greater -0.015 -0.031** Smaller -0.036** Smaller 51 Table 13 (continued): Country-specific public input constraints, firm growth and firm characteristics Country Indicator Size (log L) Different? New firm Different? Expanding L Different? Services Different? Foreign Different? Exporter Different? Importer Different? Bhutan Electricity 0.117* Greater 0.042 0.204 -0.062 -0.426** Smaller -0.189 0.006 Bhutan Telecoms 0.131* 0.226 0.077 0.026 -0.200 -0.023 -0.041 Bhutan Transport 0.263** Greater 0.299 0.167 -0.312 -0.121 -0.089 -0.277 Smaller Bhutan Customs 0.135* 0.384** Greater 0.081 0.060 -0.006 -0.073 -0.045 Bhutan Competit -0.034 0.051 0.070 -0.307 -0.075 -0.221 0.038 Bhutan AccessLand 0.095 0.048 0.237 0.258 -0.388 -0.141 0.416* Greater Bhutan TaxRates -0.073 -0.105 0.052 -0.084 -0.419 -0.156 0.399* Greater Bhutan TaxAdministration 0.049 0.018 0.159 -0.147 -0.901** Smaller -0.620** Smaller 0.110 Bhutan BusLicensing 0.089 0.125 0.117 -0.078 1.635** Greater -0.032 -0.182 Bhutan PoliticalInstability 0.014 0.103 0.032 -0.068 0.099 -0.061 Smaller -0.176 Smaller Bhutan Corruption 0.042 0.167 0.088 -0.230 -0.433** Smaller -0.423* Smaller 0.022 Bhutan Courts -0.041 0.055 -0.142 Smaller -0.096 -0.458** Smaller 0.007 0.204 Bhutan AccessFinancing -0.060 0.160 -0.047 -0.089 0.374 -0.118 -0.188 Bhutan LaborReg 0.047 -0.122 0.140 -0.289 -0.807** Smaller -0.254 0.250 Bhutan InadEducLabor 0.269** Greater -0.091 0.162 -0.256 -0.627** Smaller -0.240 0.245 Bhutan GovPolicyUnc n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. Bhutan MacroInstability n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. Bhutan bribe -0.010 0.031 -0.014 0.036 -0.068** -0.003 0.022 Bhutan mng_time 2.459 -1.885 -1.504 0.724 -10.584** Smaller -2.766 -0.885 Bhutan generator 0.112** -0.002 -0.004 0.067 0.064 0.038 -0.038 Bhutan insufficient_water 0.044 0.109 -0.007 0.120 -0.018 -0.133 -0.190* Smaller Bhutan num_insp n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. Bhutan days_customs 0.626 -1.326* 0.294 0.161 n.a. n.a. 0.491 -3.236 Bhutan m_power_out 0.062* Greater 0.126 0.032 -0.079 -0.055 -0.271** Smaller 0.009 52 Table 13 (continued): Country-specific public input constraints, firm growth and firm characteristics Country Indicator Size (log L) Different? New firm Different? Expanding L Different? Services Different? Foreign Different? Exporter Different? Importer Different? India Electricity -0.079** 0.099 Greater 0.165** Greater 0.415* Greater -0.438** Smaller 0.253** Greater -0.185** Smaller India Telecoms -0.008 Smaller 0.023 0.069* 0.118 0.037 -0.075* 0.187** India Transport 0.011 0.056 0.169** 0.186 0.096 0.014 0.125* Smaller India Customs 0.066** Smaller 0.044 0.168** Greater 0.065 0.128 0.286** Greater 0.102 Smaller India Competit 0.003 -0.021 0.040 0.394 0.305 -0.009 Greater 0.278** Greater India AccessLand -0.061** 0.057 0.020 -0.128 0.079 0.068 0.139** Greater India TaxRates 0.007 Smaller -0.067 0.219** Greater -0.098 -0.080 0.170** Greater -0.070 Smaller India TaxAdministration 0.015 Smaller -0.092 0.279** Greater 0.087 -0.140 0.182** Greater -0.074 Smaller India BusLicensing 0.045** Greater 0.005 0.076* 0.156 -0.019 0.121* 0.190** India PoliticalInstability n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. India Corruption 0.007 Smaller -0.089 0.271** Greater 0.375 0.024 0.214** Greater 0.075 India Courts 0.075** -0.107* 0.081 n.a. n.a. 0.418* 0.130* 0.035 India AccessFinancing 0.006 0.028 0.040 0.262 Greater -0.168 -0.053 0.092 Greater India LaborReg 0.064** -0.016 0.150** Greater 0.224 -0.058 0.185** Greater -0.001 India InadEducLabor 0.032* 0.021 0.106** 0.572** Greater -0.185 -0.066 -0.007 India GovPolicyUnc 0.039* -0.060 0.103* 0.068 0.325* 0.105 Greater 0.173** India MacroInstability 0.032* 0.098 Greater 0.054 0.292 0.097 0.268** Greater 0.298** Greater India bribe -0.017** Smaller -0.018 0.080** Greater -0.348** Smaller -0.054 0.111** Greater 0.037 India mng_time -0.071 Smaller -0.172 -0.036 -3.966** Smaller 0.545 2.671** Greater -1.728** Smaller India generator 0.108** Smaller -0.019 -0.001 0.272** Greater 0.012 0.147** Greater 0.108** Greater India insufficient_water -0.002 Smaller -0.017 Smaller 0.018 0.069 -0.066 -0.058** Smaller 0.105** Greater India num_insp 2.274** 0.721 2.466* -0.827 -0.735 3.294* 3.887* India days_customs 0.054 Smaller -1.230 0.771 Greater -6.322** Smaller -1.395 5.544** Greater -0.711 Greater India m_power_out -0.051** Smaller 0.044* Greater 0.088** Greater 0.126** Greater -0.152* Smaller 0.094** Greater -0.018 Smaller 53 Table 13 (continued): Country-specific public input constraints, firm growth and firm characteristics Country Indicator Size (log L) Different? New firm Different? Expanding L Different? Services Different? Foreign Different? Exporter Different? Importer Different? Maldives Electricity -0.265** Smaller -0.251 0.117 0.112 n.a. n.a. -0.060 n.a. n.a. Maldives Telecoms -0.114 Smaller -0.073 0.058 0.536** n.a. n.a. -0.026 -0.146 Maldives Transport -0.052 0.215 0.151 0.128 n.a. n.a. -0.516 0.070 Maldives Customs -0.110 Smaller -0.190 -0.034 0.776** Greater n.a. n.a. -0.031 -0.114 Maldives Competit 0.012 0.150 -0.061 0.281 n.a. n.a. 0.177 0.244 Maldives AccessLand -0.148 0.019 0.139 -0.117 n.a. n.a. 0.098 n.a. n.a. Maldives TaxRates -0.045 -0.403 0.418 -0.161 n.a. n.a. 1.059 -0.162 Maldives TaxAdministration -0.064 Smaller -0.176 0.116 0.238 n.a. n.a. 0.388 -0.691** Smaller Maldives BusLicensing -0.059 -0.176 0.221 0.283 n.a. n.a. 0.129 n.a. n.a. Maldives PoliticalInstability n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. Maldives Corruption -0.140 0.298 0.518 -0.103 n.a. n.a. -0.523 n.a. n.a. Maldives Courts 0.050 0.436 0.396 -0.148 n.a. n.a. -0.359 n.a. n.a. Maldives AccessFinancing 0.023 0.130 0.291 -0.037 n.a. n.a. 0.078 -0.141 Maldives LaborReg 0.126 -0.161 -0.143 0.288 n.a. n.a. 0.671 n.a. n.a. Maldives InadEducLabor 0.057 0.027 0.185 -0.023 n.a. n.a. -0.881 0.587* Greater Maldives GovPolicyUnc -0.056 0.381 -0.342 0.584* Greater n.a. n.a. 0.785 0.072 Maldives MacroInstability -0.084 0.098 0.024 0.559* Greater n.a. n.a. 0.472 n.a. n.a. Maldives bribe -0.014 0.075 0.025 -0.010 n.a. n.a. -0.133 -0.259** Smaller Maldives mng_time -0.324 -2.868 4.068 8.164** Greater n.a. n.a. 3.933* 1.333 Maldives generator 0.175** -0.055 0.034 -0.014 n.a. n.a. 0.097 n.a. n.a. Maldives insufficient_water 0.004 -0.021 0.022 0.013 Smaller n.a. n.a. 0.003 -0.218** Smaller Maldives num_insp 0.106 Smaller -1.385 1.310 4.836** n.a. n.a. 1.220 n.a. n.a. Maldives days_customs -2.607 Smaller -1.902 -1.508 8.430** Greater n.a. n.a. 3.599 2.703 Maldives m_power_out 0.017 Greater -0.058* Smaller 0.000 -0.021 n.a. n.a. -0.043 n.a. n.a. 54 Table 13 (continued): Country-specific public input constraints, firm growth and firm characteristics Country Indicator Size (log L) Different? New firm Different? Expanding L Different? Services Different? Foreign Different? Exporter Different? Importer Different? Nepal Electricity -0.031 -0.002 -0.120 0.185 Greater -0.046 -0.465** Smaller -0.113 Nepal Telecoms 0.103 -0.023 -0.194 Smaller 0.376* -0.223 0.031 n.a. n.a. Nepal Transport -0.138** Smaller -0.217 Smaller 0.033 0.544** Greater -0.227 -0.442** Smaller -0.063 Nepal Customs 0.159** 0.028 -0.014 0.619** Greater 0.156 -0.097 0.101 Nepal Competit -0.003 -0.121 -0.038 0.195 Greater 0.534 -0.256 -0.642* Smaller Nepal AccessLand -0.038 0.139 0.022 0.189 0.095 -0.164** Smaller 0.044 Nepal TaxRates 0.147* -0.276* 0.099 0.250 Greater -0.049 0.598** Greater 0.269 Nepal TaxAdministration 0.166** -0.065 0.223* 0.018 0.207 0.563* Greater 0.231 Nepal BusLicensing 0.096** Greater 0.152 0.078 -0.063 0.258 0.202 0.242 Nepal PoliticalInstability 0.083 0.070 0.059 0.116 -0.300 Smaller 0.292 0.109 Nepal Corruption 0.315** Greater -0.057 0.115 0.249 -0.081 -0.283 -0.059 Nepal Courts 0.076* 0.023 0.050 0.030 0.321 -0.094 Smaller 0.253* Nepal AccessFinancing 0.092* Greater 0.082 -0.055 0.065 Greater -0.256 0.117 -0.138 Nepal LaborReg 0.320** Greater -0.079 0.067 -0.207 -0.020 0.204 -0.056 Nepal InadEducLabor -0.009 -0.142 0.023 0.123 0.082 -0.463** Smaller -0.009 Nepal GovPolicyUnc n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. Nepal MacroInstability n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. Nepal bribe 0.016 -0.070** Smaller 0.009 -0.071* 0.048 -0.049 0.026 Nepal mng_time 0.347 -0.928 -3.747* Smaller -1.190 -2.474 -5.083* Smaller 7.270 Nepal generator 0.106** 0.105 -0.051 n.a. n.a. -0.154 0.158** 0.283** Greater Nepal insufficient_water 0.005 -0.034 0.085 n.a. n.a. -0.117 -0.057 0.010 Nepal num_insp n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. Nepal days_customs 1.623 -3.521 -1.907 -6.383 46.191 -13.746* Smaller -6.007 Nepal m_power_out 0.002 Greater 0.008 0.005 Smaller 0.010 Greater 0.011 Greater -0.015 -0.003 Smaller 55 Table 13 (continued): Country-specific public input constraints, firm growth and firm characteristics Country Indicator Size (log L) Different? New firm Different? Expanding L Different? Services Different? Foreign Different? Exporter Different? Importer Different? Pakistan Electricity -0.019 -0.237* Smaller -0.004 -0.643** Smaller -0.143 0.083 0.322** Greater Pakistan Telecoms 0.063** 0.160 0.181** Greater 0.605** Greater 0.280 Greater 0.029 0.207* Pakistan Transport 0.016 0.056 0.077 -0.164 0.415* -0.110 0.136 Pakistan Customs 0.152** Greater -0.058 0.093 0.365** Greater 0.274 -0.117 -0.071 Smaller Pakistan Competit 0.040 0.007 0.263** Greater 0.029 0.152 -0.225* -0.023 Pakistan AccessLand -0.060* -0.028 0.091 -0.420** Smaller -0.087 0.142 0.039 Pakistan TaxRates 0.207** Greater -0.130 0.168* -0.456** Smaller -0.471 -0.339** Smaller -0.260* Smaller Pakistan TaxAdministration 0.175** Greater -0.325** Smaller 0.104 -0.048 -0.174 -0.485** Smaller -0.342** Smaller Pakistan BusLicensing 0.058* -0.229* Smaller 0.149* -0.270* Smaller 0.164 -0.156 Smaller -0.143 Smaller Pakistan PoliticalInstability 0.054 -0.013 0.279** Greater 0.489** Greater 0.251 0.631** Greater -0.376* Smaller Pakistan Corruption 0.053 -0.138 0.375** Greater -0.379* Smaller -0.223 -0.110 0.019 Pakistan Courts 0.080 0.432* Greater 0.154 0.554** Greater -0.137 0.624** Greater -0.313 Pakistan AccessFinancing -0.005 -0.169 Smaller 0.057 -0.182 -0.317 -0.414** Smaller -0.501** Smaller Pakistan LaborReg 0.139** Greater -0.144 0.094 0.254* Greater -0.237 -0.227** Smaller -0.048 Pakistan InadEducLabor 0.080** -0.071 0.155** 0.393** Greater -0.357* 0.105 0.138 Pakistan GovPolicyUnc 0.129** -0.105 0.234* -0.529* -0.136 -0.370** Smaller -0.085 Pakistan MacroInstability 0.042 -0.280* Smaller 0.231** Greater -0.250 -0.053 0.018 -0.039 Smaller Pakistan bribe 0.042** Greater 0.003 0.050* 0.092 -0.258** Smaller -0.075* Smaller -0.103* Smaller Pakistan mng_time 0.875** Greater 0.793 -0.353 0.865 3.323 -0.158 Smaller 0.676 Pakistan generator 0.167** Greater -0.068* 0.048* 0.302** Greater -0.093 0.101** -0.017 Smaller Pakistan insufficient_water 0.009 -0.003 0.052* 0.817** Greater -0.089 -0.032 -0.089* Smaller Pakistan num_insp 5.045** Greater 1.959 -3.156 Smaller 13.842 12.870 -3.146 Smaller -2.293 Smaller Pakistan days_customs -0.483 Smaller -1.896 -3.766** Smaller -1.874 -3.506 -5.749** Smaller -0.768 Pakistan m_power_out -0.011 -0.089* Smaller 0.013 Smaller -0.335** Smaller 0.054 0.001 0.157** Greater 56 Table 13 (continued): Country-specific public input constraints, firm growth and firm characteristics Country Indicator Size (log L) Different? New firm Different? Expanding L Different? Services Different? Foreign Different? Exporter Different? Importer Different? SriLanka Electricity 0.052 Greater 0.413 0.220 n.a. n.a. 0.019 -0.093 -0.099 SriLanka Telecoms 0.076 0.232 -0.104 n.a. n.a. -0.224 -0.076 -0.423** Smaller SriLanka Transport 0.107* 0.271 0.217 n.a. n.a. 0.087 -0.124 0.577** SriLanka Customs 0.070 -0.237 0.032 n.a. n.a. 0.009 -0.125 0.820** Greater SriLanka Competit 0.037 -0.235 0.129 n.a. n.a. -0.002 -0.470** 0.094 SriLanka AccessLand 0.011 0.297 0.298** Greater n.a. n.a. -0.207 0.132 0.057 SriLanka TaxRates 0.065 -0.276 0.142 n.a. n.a. -0.179 -0.634** Smaller 0.397* Greater SriLanka TaxAdministration 0.008 -0.444** Smaller 0.030 n.a. n.a. 0.038 -0.059 0.401** Greater SriLanka BusLicensing 0.021 -0.121 -0.020 n.a. n.a. -0.038 -0.056 0.084 SriLanka PoliticalInstability n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. SriLanka Corruption 0.165** Greater -0.408* 0.036 n.a. n.a. -0.074 -0.241 0.505** Greater SriLanka Courts n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. SriLanka AccessFinancing -0.041 -0.001 0.161 n.a. n.a. -0.222 -0.032 0.211 SriLanka LaborReg 0.222** Greater -0.342 -0.079 n.a. n.a. -0.161 -0.253 Smaller 0.027 SriLanka InadEducLabor 0.197** Greater -0.360 0.079 n.a. n.a. -0.262 -0.166 -0.142 SriLanka GovPolicyUnc 0.068 -0.311 -0.039 n.a. n.a. 0.018 -0.373* 0.361* SriLanka MacroInstability 0.115* -0.235 -0.022 n.a. n.a. -0.049 0.122 0.302 SriLanka bribe -0.003 0.025 0.031 n.a. n.a. 0.010 0.001 0.136** Greater SriLanka mng_time 0.264* -0.033 -0.443 n.a. n.a. -0.178 0.256 Smaller -0.616 SriLanka generator 0.125** -0.074 0.025 n.a. n.a. -0.066 0.128** 0.061 SriLanka insufficient_water 0.006 -0.025 0.037 n.a. n.a. 0.084 -0.063 0.197** Greater SriLanka num_insp 3.650** -5.238* Smaller -2.303 Smaller n.a. n.a. -3.022 -0.517 -6.970** Smaller SriLanka days_customs 4.121** Greater -1.707 -1.301 n.a. n.a. -7.004** Smaller 5.113 -16.293** Smaller SriLanka m_power_out n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. 57 Table 14: Industry- and country-specific correlations of expanding permanent employment with firm characteristics Industry Country corr Diff? N pcorr Diff? N Garments All 0.057** n.a. 7,809 0.011 n.a. 7,718 Garments Bangladesh 0.077** 1,866 -0.075* Smaller 1,843 Garments India 0.039 4,590 0.009 4,530 Garments Pakistan 0.068* 1,353 0.039 1,345 Food & beverages All -0.078** n.a. 7,809 -0.032 n.a. 7,718 Food & beverages Bangladesh -0.139** Smaller 1,866 -0.018 1,843 Food & beverages India -0.055* 4,590 -0.054* 4,530 Food & beverages Pakistan -0.001 Greater 1,353 0.074 Greater 1,345 Chemicals All 0.003 n.a. 7,809 -0.003 n.a. 7,718 Chemicals Bangladesh 0.062 1,866 0.074* Greater 1,843 Chemicals India -0.011 4,590 -0.012 4,530 Chemicals Pakistan -0.016 1,353 0.030 1,345 Electronics All -0.049** n.a. 7,809 -0.028 n.a. 7,718 Electronics Bangladesh -0.108** 1,866 -0.057 1,843 Electronics India -0.042* 4,590 -0.016 4,530 Electronics Pakistan -0.026 1,353 -0.028 1,345 Machinery All -0.004 n.a. 7,809 0.016 n.a. 7,718 Machinery Bangladesh -0.058 1,866 0.046 1,843 Machinery India 0.010 4,590 0.026 4,530 Machinery Pakistan -0.046 1,353 -0.056 Smaller 1,345 Textiles All 0.029* n.a. 7,809 0.019 n.a. 7,718 Textiles Bangladesh 0.054 1,866 0.035 1,843 Textiles India 0.032 4,590 0.016 4,530 Textiles Pakistan -0.012 1,353 -0.017 1,345 Notes: corr=Controls are country fixed effects. pcorr=Controls are country fixed effects plus other firm characteristics. Green indicates a positive correlation; red a negative correlation. *=significant at the 5% level; **=significant at the 1% level. 58 Table 15: Rankings of constraints by country TOTAL TOTAL AFG AFG BGD BGD NPL NPL PAK PAK IND IND PoliticalInstability 2.58 PoliticalInstability 3.16 Electricity 3.11 PoliticalInstability 3.20 TaxAdministration 2.29 Electricity 1.61 Electricity 2.17 Electricity 2.89 PoliticalInstability 2.86 Electricity 3.11 Electricity 2.15 Corruption 1.58 Corruption 1.83 Corruption 2.13 Corruption 2.39 Transport 2.21 PoliticalInstability 2.12 TaxAdministration 1.44 TaxAdministration 1.66 AccessLand 2.08 AccessLand 2.13 Corruption 1.63 GovPolicyUnc 2.09 LaborReg 1.12 GovPolicyUnc 1.37 CrimeTheftDisorder 1.69 TaxAdministration 1.99 LaborReg 1.46 Courts 2.06 InadEducLabor 0.98 MacroInstability 1.32 Competit 1.55 GovPolicyUnc 1.89 Competit 1.29 Corruption 2.04 GovPolicyUnc 0.97 Competit 1.16 Telecoms 1.32 MacroInstability 1.83 Customs 1.03 MacroInstability 1.97 Customs 0.93 CrimeTheftDisorder 1.15 TaxAdministration 1.24 Competit 1.59 TaxAdministration 1.03 Competit 1.36 CrimeTheftDisorder 0.93 AccessLand 1.15 Customs 1.20 Customs 1.50 InadEducLabor 0.83 CrimeTheftDisorder 1.36 BusLicensing 0.83 Customs 1.13 Transport 1.18 Telecoms 1.48 CrimeTheftDisorder 0.81 AccessLand 1.35 MacroInstability 0.83 InadEducLabor 1.07 Courts 1.17 InadEducLabor 1.47 Telecoms 0.48 Customs 1.27 Competit 0.82 LaborReg 1.05 BusLicensing 0.76 CrimeTheftDisorder 1.32 AccessLand 0.38 LaborReg 1.12 Transport 0.81 Transport 1.04 InadEducLabor 0.74 BusLicensing 1.29 BusLicensing 0.26 BusLicensing 1.08 AccessLand 0.68 BusLicensing 0.94 LaborReg 0.33 Transport 1.24 Courts 0.10 Transport 0.92 Telecoms 0.47 Courts 0.87 MacroInstability - Courts 1.15 MacroInstability - InadEducLabor 0.91 Courts 0.47 Telecoms 0.71 GovPolicyUnc - LaborReg 0.87 GovPolicyUnc - Telecoms 0.67 PoliticalInstability - 59 Table 16 Comparison of reported constraints by formal, small formal and rural firms Unconditional means Survey fixed effects and log(L) (centered on L=5) All formal Small formal Rural Diff? Small formal Rural Diff? Obs (SF) Obs (R) Electricity 2.575 2.563 1.470 Smaller 2.446 1.828 Smaller 2,234 4,888 Telecoms 0.969 0.694 0.318 Smaller 0.602 0.316 Smaller 1,355 4,881 Transport 1.081 0.874 0.696 Smaller 0.864 0.725 Smaller 2,166 4,750 AccessLand 1.606 1.701 0.100 Smaller 1.613 0.066 Smaller 1,747 4,157 InadEducLabor 1.254 0.986 0.183 Smaller 0.929 0.322 Smaller 1,755 4,210 MacroInstability 1.978 1.939 1.619 Smaller 1.795 1.785 1,716 2,428 GovPolicyUnc 1.868 1.622 0.535 Smaller 1.608 0.395 Smaller 955 1,846 PoliticalInstability 2.457 2.290 0.681 Smaller 2.206 0.884 Smaller 1,316 4,029 AccessFinancing 1.677 1.655 1.086 Smaller 1.687 1.013 Smaller 2,143 4,827 Competit 1.308 1.243 1.394 1.200 1.430 1,766 66 TaxAdministration 1.854 1.569 0.099 Smaller 1.328 0.137 Smaller 1,756 4,197 TaxRates 1.820 1.669 0.146 Smaller 1.431 0.205 Smaller 1,762 4,195 LaborReg 0.904 0.677 0.042 Smaller 0.631 0.056 Smaller 2,143 4,202 Customs 1.296 0.811 0.020 Smaller 0.660 0.029 Smaller 1,520 4,091 BusLicensing 1.140 1.035 0.177 Smaller 0.933 0.173 Smaller 1,668 1,763 Courts 1.585 1.553 0.163 Smaller 1.543 0.195 Smaller 1,000 4,231 Corruption 2.137 1.978 0.146 Smaller 1.888 0.159 Smaller 2,211 4,229 CrimeTheftDisorder 1.405 1.271 0.791 Smaller 1.199 0.819 Smaller 1,799 67 bribe 0.591 0.464 0.288 Smaller 0.404 0.212 Smaller 1,691 52 bribe1 0.269 0.205 0.077 Smaller 0.181 0.062 Smaller 1,394 52 mng_time 4.467 2.884 0.118 Smaller 2.436 0.286 Smaller 2,170 5,398 generator 0.515 0.146 0.062 Smaller 0.056 0.091 Greater 1,631 5,416 insufficient_water 0.343 0.302 0.014 Smaller 0.321 0.020 Smaller 1,664 4,031 num_insp 12.652 5.809 0.451 Smaller 5.168 1.608 Smaller 1,283 2,970 days_customs 8.242 5.204 5.204 n.a. 7.396 7.396 n.a. 142 0 m_power_out 0.751 0.727 0.669 Smaller 0.723 0.702 2,014 2,671 Small numbers of observations are noted in red. Differences are heteroskedasticity-robust tests with a 5% significance level. 60 Table 17 Ranking of constraints: formal, small formal and rural firms Unconditional means Survey fixed effects and log(L) (centered on L=5) Formal Small formal Rural Small formal Rural Electricity Electricity MacroInstability Electricity Electricity PoliticalInstability PoliticalInstability Electricity PoliticalInstability MacroInstability Corruption Corruption Competit Corruption Competit MacroInstability MacroInstability CrimeTheftDisorder MacroInstability PoliticalInstability GovPolicyUnc AccessLand Transport AccessLand CrimeTheftDisorder TaxAdministration GovPolicyUnc PoliticalInstability GovPolicyUnc Transport AccessLand TaxAdministration GovPolicyUnc Courts GovPolicyUnc Courts Courts Telecoms TaxAdministration InadEducLabor CrimeTheftDisorder CrimeTheftDisorder InadEducLabor Competit Telecoms Competit Competit BusLicensing CrimeTheftDisorder Courts Customs BusLicensing Courts BusLicensing BusLicensing InadEducLabor InadEducLabor Corruption InadEducLabor Corruption BusLicensing Transport AccessLand Transport TaxAdministration Transport Customs TaxAdministration Customs AccessLand Telecoms Telecoms LaborReg LaborReg LaborReg LaborReg LaborReg Customs Telecoms Customs Small numbers of observations are noted in red. 61 Table 18 Comparison of reported constraints by formal, small formal and informal sector firms (manufacturing & India only) Unconditional means Survey fixed effects and log(L) (centered on L=5) All formal Small formal Informal Diff? Small formal Informal Diff? Obs (SF) Obs (I) Electricity 1.790 1.818 1.572 Smaller 1.817 1.574 Smaller 2,247 1,951 Telecoms 0.515 0.497 0.920 Greater 0.522 0.998 Greater 2,144 1,951 Transport 0.905 0.826 1.836 Greater 0.838 1.840 Greater 2,244 1,951 AccessLand 0.713 0.752 1.672 Greater 0.901 1.686 Greater 2,149 1,951 InadEducLabor 1.013 0.974 1.714 Greater 0.860 1.720 Greater 2,151 1,951 MacroInstability 0.913 0.832 1.149 Greater 0.677 1.160 Greater 2,104 402 GovPolicyUnc 1.033 0.954 1.192 Greater 0.809 1.184 Greater 2,114 402 AccessFinancing 1.203 1.168 1.533 Greater 1.113 1.567 Greater 2,238 1,951 Competit 0.863 0.824 1.376 Greater 0.767 1.382 Greater 2,127 1,951 TaxAdministration 1.551 1.486 1.455 1.432 1.414 2,149 1,951 TaxRates 1.682 1.639 1.531 Smaller 1.586 1.492 2,152 1,951 LaborReg 1.169 1.061 1.576 Greater 0.879 1.588 Greater 2,234 1,951 Customs 1.011 0.878 0.590 Smaller 0.783 0.595 Smaller 2,021 402 BusLicensing 0.891 0.802 1.826 Greater 0.710 1.818 Greater 2,032 1,549 Courts 0.441 0.336 1.591 Greater 0.272 1.585 Greater 1,129 1,951 Corruption 1.681 1.637 1.451 Smaller 1.507 1.431 2,241 1,951 CrimeTheftDisorder 0.960 0.993 1.582 Greater 0.994 1.561 Greater 2,148 1,951 bribe 0.444 0.438 0.237 Smaller 0.312 0.266 Smaller 2,230 1,936 bribe1 0.175 0.173 0.084 Smaller 0.158 0.096 Smaller 1,966 1,623 mng_time 6.624 6.436 1.624 Smaller 5.685 1.704 Smaller 2,538 1,942 generator 0.588 0.417 0.197 Smaller 0.246 0.226 2,410 1,951 insufficient_water 0.213 0.207 1.000 n.a. 0.205 1.000 Greater -999 5 num_insp 8.743 5.832 5.063 4.140 5.125 2,309 128 days_customs 8.470 7.136 7.136 n.a. 5.928 5.928 n.a. 295 0 m_power_out 0.686 0.719 0.942 Greater 0.671 0.943 Greater 2,179 1,320 Small numbers of observations are noted in red. Differences are heteroskedasticity-robust tests with a 5% significance level. 62 Table 19 Ranking of constraints: formal, small formal and informal sector firms (manufacturing & India only) Unconditional means Survey fixed effects and log(L) (centered on L=5) Formal Small formal Informal Small formal Informal Electricity Electricity Transport Electricity Transport Corruption Corruption BusLicensing Corruption BusLicensing TaxAdministration TaxAdministration InadEducLabor TaxAdministration InadEducLabor LaborReg LaborReg AccessLand CrimeTheftDisorder AccessLand GovPolicyUnc CrimeTheftDisorder Courts AccessLand LaborReg InadEducLabor InadEducLabor CrimeTheftDisorder LaborReg Courts Customs GovPolicyUnc LaborReg InadEducLabor Electricity CrimeTheftDisorder Customs Electricity Transport CrimeTheftDisorder MacroInstability MacroInstability TaxAdministration GovPolicyUnc Corruption Transport Transport Corruption Customs TaxAdministration BusLicensing Competit Competit Competit Competit Competit BusLicensing GovPolicyUnc BusLicensing GovPolicyUnc AccessLand AccessLand MacroInstability MacroInstability MacroInstability Telecoms Telecoms Telecoms Telecoms Telecoms Courts Courts Customs Courts Customs 63 Figure 1. Electricity, Outages and Generators Electricity Constraint level, conditional means; quadratic fit; 5% robust CIs 4 BGD NPL 3 AFG PAK LKA 2 IND MDV 1 BTN 0 6 7 8 9 10 11 Log(GDP p.c.) at PPP Power Cuts >1 per mon. Constraint level, conditional means; quadratic fit; 5% robust CIs 1.5 AFG 1 BGD NPL IND .5 PAK BTN MDV LKA 0 6 7 8 9 10 11 Log(GDP p.c.) at PPP 64 Generator (y/n) 1 Constraint level, conditional means; quadratic fit; 5% robust CIs AFG .8 .6 LKA IND MDV .4 BGD PAK BTN .2 NPL 0 6 7 8 9 10 Log(GDP p.c.) at PPP 65 Figure 2. Corruption and bribes Corruption Constraint level, conditional means; quadratic fit; 5% robust CIs 4 3 BGD AFG PAK 2 MDV NPL IND LKA BTN 1 0 6 7 8 9 10 11 Log(GDP p.c.) at PPP Bribes (>0) Constraint level, conditional means; quadratic fit; 5% robust CIs 1 BGD .8 .6 PAK AFG MDV .4 IND LKA .2 NPL BTN 0 6 7 8 9 10 11 Log(GDP p.c.) at PPP 66 Figure 3. Political instability Polit. Instab. Constraint level, conditional means; quadratic fit; 5% robust CIs 4 NPL AFG 3 BGD PAK 2 1 BTN 0 6 7 8 9 10 Log(GDP p.c.) at PPP Figure 4. Access to Land Access to Land Constraint level, conditional means; quadratic fit; 5% robust CIs 4 3 BGD MDV AFG 2 PAK 1 IND NPL LKA BTN 0 6 7 8 9 10 11 Log(GDP p.c.) at PPP 67 Figure 5. Insufficient water Insuff. Water Constraint level, conditional means; quadratic fit; 5% robust CIs .8 PAK .6 BTN .4 LKA .2 BGD IND AFG NPL MDV 0 6 7 8 9 10 11 Log(GDP p.c.) at PPP Figure 6. Labour Regulation Labour Regulation Constraint level, conditional means; quadratic fit; 5% robust CIs 4 3 2 LKA NPL BTN MDV IND PAK 1 BGD AFG 0 6 7 8 9 10 11 Log(GDP p.c.) at PPP 68 Figure 7. Courts Courts Constraint level, conditional means; quadratic fit; 5% robust CIs 4 3 PAK 2 MDV AFG BGD 1 BTN IND NPL 0 6 7 8 9 10 11 Log(GDP p.c.) at PPP Figure 8. Inadequately educated labour Inad. Educ. Labor Constraint level, conditional means; quadratic fit; 5% robust CIs 4 3 2 BTN BGD MDV LKA 1 IND PAK NPL AFG 0 6 7 8 9 10 11 Log(GDP p.c.) at PPP 69 Figure 9. Telecoms Telecoms Constraint level, conditional means; quadratic fit; 5% robust CIs 4 3 2 BGD AFG LKA 1 PAK BTN NPL IND MDV 0 6 7 8 9 10 11 Log(GDP p.c.) at PPP 70 REFERENCES Wendy Carlin and Mark E. Schaffer (2012), “The Business Environment in the Transition�, forthcoming in Paul G. Hare and Gerard Turley (eds.), Handbook of the Economics and Political Economy of Transition, Routledge 2013. Wendy Carlin, Mark E. Schaffer and Paul Seabright (2010), “A Framework for Cross- Country Comparisons of Public Infrastructure Constraints on Firm Growth�, CEPR Discussion Paper 7662 http://ideas.repec.org/p/cpr/ceprdp/7662.html Wendy Carlin, Mark E. Schaffer and Paul Seabright (2006), “Where Are the Real Bottlenecks? Evidence from 20,000 Firms in 60 Countries About the Shadow Costs of Constraints to Firm Performance�, CEPR Discussion Paper 5719 http://ideas.repec.org/p/cpr/ceprdp/5719.html Wendy Carlin, Mark E. Schaffer and Paul Seabright (2010), “Institutions matter for growth – but which ones and how much?�, http://www.voxeu.org/index.php?q=node/4621 Wendy Carlin, Mark E. Schaffer and Paul Seabright (2012), “Soviet Power Plus Electrification: What is the Long-Run Legacy of Communism?�, CEPR Discussion Paper 9003. http://econpapers.repec.org/paper/cprceprdp/9003.htm Simon Commander and Jan Svejnar (2007), “Do Institutions, Ownership, Exporting and Competition Explain Firm Performance: Evidence from 26 Transition Countries�, Institute for the Study of Labor (IZA) Discussion Paper 2637, Bonn Jean-Jacques Dethier, Maximilian Hirn and Stephan Straub (2010), “Explaining Enterprise Performance in Developing Countries with Business Climate Survey Data� World Bank Research Observer. Jean-Jacques Dethier, Maximilian Hirn and Stephan Straub (2008), “Explaining Enterprise Performance in Developing Countries with Business Climate Survey Data�, World Bank Policy Research Discussion Paper 4792. Steven Durlauf, Paul Johnson and Jonathan Temple (2005), “Growth Econometrics� in Philippe Aghion and Steven N. Durlauf, eds, Handbook of Economic Growth, North-Holland. William Easterly (2009), “Why there’s no ‘GrowthGate’: Frustration vs. Chicanery in Explaining Growth�, Aidwatch, 10 December. http://aidwatchers.com/2009/12/why-there’s-no-“growthgate�-frustration-vs- chicanery-in-explaining-growth/ Mary Hallward-Driemeier and Reyes Aterido (2009), “Comparing Apples with…Apples How to Make (More) Sense of Subjective Rankings of Constraints to Business�, World Bank Policy Research Paper 5054. 71 Mary Hallward-Driemeier, Gita Khun-Jush, Lant Pritchett (2010), “Deals versus Rules: Policy Implementation Uncertainty and Why Firms Hate It�, World Bank Policy Research Working Paper No. 5321. C. Manski (1993), “Identification of Endogenous Social Effects: The Reflection Problem,� Review of Economic Studies, Vol. 60, No. 3, 1993, pp. 531-542 Mitra, P., Selowsky, M and J. Zalduendo (2010). Turmoil at Twenty: Recession, Recovery, and Reform in Central and Eastern Europe and the Former Soviet Union. World Bank: Washington, DC. 72 Appendix 1: Modelling framework Gij is a private input chosen by the firm. B j and Gij are substitutes in the production of intermediate input Eij , created via an intermediate input production function E ( B j , Gij ) . Gij is a mitigation cost or input that substitutes for deficiencies in the public input B j . Intermediate input Eij is combined with labour input Lij via a final output production function F and firms-specific technology level Aij, to generate output Yij = Aij F ( Lij , Eij )) : Eij = E ( B j , Gij ) (1) Yij = Aij F ( Lij , E ( B j , Gij )) (2) The firm’s problem is choose Lij and Gij to maximize profits for given technology Aij, public input B j , and relative prices of labour and mitigation, denoted as wj and pj, respectively (we normalize the output price to 1; all firms in country j face the same prices), and the intermediate input and final output production technology. 9 It is useful to write these as maximum-value or indirect objective functions. Denoting profit-maximizing quantities with a superscript *, we have the input demand functions for labour and mitigation, the supply function for output, and the profit function for the firm’s maximized profit, all written as functions of the exogenous variables Aij, B j , wj and pj. L∗ = L * ( Aij , B j , w j , p j ) ij (3) ∗ Gij = G * ( Aij , B j , w j , p j ) (4) ∗ Eij = E ( B j , Gij ) = E * ( Aij , B j , w j , p j ) * (5) ∗ Yij∗ = Y * ( Aij , B j , w j , p j ) = Aij F ( L∗ , E ( B j , Gij )) ij (6) π ij = π * ( Aij , B j , w j , p j ) = Y * ( Aij , B j , w j , p j ) − w j L∗ − p j Gij ∗ ij ∗ (7) 9 In addition to the assumption of weak separability that we have already made, F(L,B,G)=F(L,E(B,G), and the usual assumptions about the production functions E and F, we also assume that E is strictly quasi-concave and homothetic. 73 So far we have assumed that the public input B j is supplied identically to all firms in a country. An example of a public input of this kind is macroeconomic stability. A more realistic assumption would allow for Bij to vary across firms. This could result from regional variation in the quality of the public input within a country, or simply because of random variation in the reliability of the public input, e.g., some firms were luckier than others with respect to the number power outages they faced. An important issue relates to possible differences in infrastructure quality across locations. We find in the paper that firms in large cities report higher constraints across most dimensions of the business environment than do firms in more rural locations. However, this is not because the supply of public inputs, Bij is lower in cities; in fact, our prior is that if infrastructure quality varies between rural areas and cities, it is higher in the latter. Thus, when we find that firms in large cities are more constrained, this is in spite of having, if anything, better public inputs. In some cases, however, the public input supplied to the firm will vary with the firm’s profitability or productivity. (Since maximized profit is a function of productivity Aij – see above – we simplify and consider productivity Aij as a proxy for profitability.) In this case, we have ∂Bij ∂Bij Bij = B( B j , Aij ) >0 <0 (8) ∂B j ∂Aij Eij = E ( Bij , Gij ) = E ( B( B j , Aij ), Gij ) (9) An example: B j is the honesty of the bureaucracy in country j, Bij is the inverse of the number of inspections that a firm with productivity Aij attracts (more inspections means a lower quality public input Bij supplied to the firm), and Gij is bribes. 74 We now consider how the firm’s optimal choices of inputs and output, and the firm’s valuations of the public input, vary with the quality of the public input B j , and with the productivity of the firm Aij . In the model above, supply and profits are, not surprisingly, increasing in the quality of the public input B j : ∂Yij* ∂π ij * > 0, >0 (10) ∂B j ∂B j Many such country-level measures are available and have been used in country-level studies. Firm-level surveys do collect some information about the quality of the business environment B j . However, these measures are best interpreted as estimates by an individual firm of the quality of the shared environment, in the same way that a firm’s answers on a price survey provide information about the market price for a specific product. An example of such a measure from the Enterprise Surveys would be a firm’s report of the number of electricity supply interruptions it faced. ∗ Information on mitigation costs Gij is also collected from firms. Mitigation expenditures are endogenously chosen by the firm. These expenditures will be decreasing in the quality of the public input B j and increasing in the productivity of the firm, Aij : ∗ ∗ ∂Gij ∂Gij ≤0, ≥0 (11) ∂B j ∂Aij The second expression is of interest to us in the empirical analysis and has a twofold intuitive justification. In the benchmark case where the public input supplied to all firms is identical and independent of firm productivity, i.e. Bij = B j , higher productivity firms spend more on mitigation because the payoff is bigger than it is to low productivity firms. In the case where the quality of the public input varies 75 inversely with firm productivity, as in the example of higher productivity firms attracting more attention from rent-seeking officials, i.e. Bij = B( B j , Aij ) , the effect is reinforced. More profitable firms have an even lower quality public input, and hence the payoff to spending on mitigation is even bigger. The above implies that firm productivity, and proxies for productivity and growth, should be associated with higher mitigation outlays. Moreover, the partial derivative ∗ ∂Gij can vary systematically across countries, and in particular it will be decreasing ∂Aij in the quality of the public input B j : ∗ ∂ 2 Gij ≤0 (12) ∂Aij ∂B j i.e., countries with a lower quality public input, B j should see stronger correlations between mitigation outlays and firm-level productivity. If the quality or quantity of the public input B j is sufficiently high, the marginal cost of additional expenditure on mitigation will be greater than the marginal benefit to the ∗ firm, in which case optimal mitigation Gij is zero. Examples would be expenditure on a new generator when the quality of electricity supply is so high that the cost of the generator cannot be justified or expenditure on bribes when public officials are already so honest that there is no point bribing them. In these circumstances, there would be no correlation between mitigation costs and firm-level productivity. ∗ Firms also provide information about the flow of services Eij obtained from the combination of the public input and mitigation expenditures. An example is the speed with which goods clear customs, which is an endogenous result of the quality of the customs bureaucracy ( B j ) and of the optimal mitigation costs such as management ∗ time and bribes aimed at getting the firm’s goods through customs ( Gij ). In the benchmark case where the public input supplied to all firms is identical and 76 ∗ independent of firm productivity, the flow of intermediate inputs, Eij , is increasing in the productivity of the firm; this follows from the property that mitigation outlays are also increasing in the productivity of the firm: ∗ ∗ ∂Eij ∂E ( B j , Gij ( Aij )) * ∂Eij ∂Gij = = ≥0 (13) ∂Aij ∂Aij ∂Gij ∂Aij A simple and intuitive interpretation of the “Subjective Severity� indicators collected in the Enterprise Surveys is that they represent the “reported cost� Rij of a public input is the gap between the firm’s profit in the hypothetical situation where the public input provided is of such high quality that it poses a negligible obstacle to the firm’s operations, and the firm’s profit in reality, given the actual quality of public input provided. If we denote the level of public input provided in an ideal, high-quality business environment as B j , we have Rij = π * ( Aij , B j , w j ) − π * ( Aij , B j , w j ) (14) The marginal analogue of the reported cost Rij for small changes in the public input, or “marginal reported cost�, is therefore simply the derivative of the profit function: ∂π ij ∗ Rij ≈ ≡ λij (15) ∂B j We can think of the profit function π ij as resulting from a constrained maximization ∗ by the firm, where the public input B j is supplied to the firm at a level or quality that means the firm would prefer a higher quality or more of it. By the envelope theorem for constrained maximization, the derivative of the profit function π ij with respect to ∗ a constrained or fixed input is simply the shadow price of the input. Thus we follow 77 Carlin et al. (2006) and interpret the responses to “Subjective Severity� questions as the shadow price λij of shortcomings in the public input B j . 10 The shadow price of B j is decreasing in B j : ∂λij ∂ 2π ij ∗ ≡ <0 (16) ∂B j ∂B j2 The shadow price of a constraint is also increasing in the productivity of the firm: ∂λij ∂ 2π ij ∗ ≡ >0 (17) ∂Aij ∂B j ∂Aij i.e., a higher productivity firm will report higher costs of a poor public input than a lower productivity firm – even though they share the same business environment. Lastly, we are interested in firm growth as well firm productivity. The simplest extension to the model that accommodates this is to extend the model to include a quasi-fixed input such as capital or workers with permanent contracts. Now, in ∗ addition to the optimizing choice of variable inputs L∗ and Gij , the firm also chooses ij ∗ ∗ an optimal level of investment I ij in the quasi-fixed input. I ij will be increasing in the firm-specific parameters that capture future profitability such as Aij . Hence, we ∗ expect direct measures of I ij , or proxy measures for the parameters that drive the ∗ ∗ ∗ cross-firm variation in I ij , to be correlated with Gij , Eij and MRC * in the same way as Aij is above. 10 Carlin et al. (2010) interpret the responses as “reported costs� (RC) in a slightly different framework to the one adopted here, namely an O-ring production function in which the quality of the public input is measured by the probability that it fails and output is zero. This allows a response of 0 to be interpreted naturally as a zero probability of failure, which in turn implies the firm’s evaluation of the quality of the public input is that it is so high that additional improvements would not benefit the firm. The difference in formal frameworks is immaterial to the analysis here. 78 Appendix 2: Charts for Access to Finance, and Tax Rates, and the remaining elements of the external environment Although we cannot compare the level of the finance obstacle with the public good elements of the business environment, we can look at the cross-country patterns. The results for Nepal and India stand out. In both cases, firms report access to finance as a less serious obstacle than is the case in other countries at comparable levels of development. Access to Finance Constraint level, conditional means; quadratic fit; 5% robust CIs 4 3 BGD MDV AFG 2 PAK BTN LKA NPL IND 1 0 6 7 8 9 10 11 Log(GDP p.c.) at PPP As discussed in the paper, it is difficult to interpret managers’ answer to the question about tax rates. The cross-country chart reflects the finding that firms in countries at all levels of development complain a lot about tax rates. The SAR countries – with the exception of Pakistan and the Maldives, which are typical of countries at their level of development – tend to report lower obstacles from the tax rate than do firms in other countries. 79 Tax Rates 4 3 Constraint level, conditional means; quadratic fit; 5% robust CIs PAK 2 LKA BGD AFG IND BTN MDV 1 NPL 0 6 7 8 9 10 11 Log(GDP p.c.) at PPP Other elements of the business environment Business Lic. Constraint level, conditional means; quadratic fit; 5% robust CIs 4 3 2 BGD BTN PAK LKA 1 AFG IND NPL MDV 0 6 7 8 9 10 11 Log(GDP p.c.) at PPP 80 Competition 4 3 2 Constraint level, conditional means; quadratic fit; 5% robust CIs BGD AFG LKA NPL PAK 1 BTN IND MDV 0 6 7 8 9 10 11 Log(GDP p.c.) at PPP Crime/Theft/Disorder Constraint level, conditional means; quadratic fit; 5% robust CIs 4 3 2 AFG BGD PAK MDV LKA 1 IND NPL BTN 0 6 7 8 9 10 11 Log(GDP p.c.) at PPP 81 Customs 4 3 2 Constraint level, conditional means; quadratic fit; 5% robust CIs BGD AFG PAK BTN NPL 1 IND LKA MDV 0 6 7 8 9 10 11 Log(GDP p.c.) at PPP 82 Gov. Policy Uncert. 4 3 Constraint level, conditional means; quadratic fit; 5% robust CIs PAK 2 LKA BGD MDV 1 IND 0 6 7 8 9 10 11 Log(GDP p.c.) at PPP Macro Instability Constraint level, conditional means; quadratic fit; 5% robust CIs 4 3 2 PAK BGD LKA 1 IND MDV 0 6 7 8 9 10 11 Log(GDP p.c.) at PPP 83 Tax Admin. 4 3 Constraint level, conditional means; quadratic fit; 5% robust CIs PAK 2 BGD IND AFG LKA BTN NPL 1 MDV 0 6 7 8 9 10 11 Log(GDP p.c.) at PPP Transport Constraint level, conditional means; quadratic fit; 5% robust CIs 4 3 NPL 2 BTN BGD AFG LKA 1 PAK MDV IND 0 6 7 8 9 10 11 Log(GDP p.c.) at PPP 84 Number of Inspections 40 30 20 Constraint level, conditional means; quadratic fit; 5% robust CIs LKA 10 BGD PAK IND AFG MDV 0 -10 6 7 8 9 10 Log(GDP p.c.) at PPP Mng. Time on Regulation (%) Constraint level, conditional means; quadratic fit; 5% robust CIs 40 30 BTN 20 10 AFG NPL IND LKA BGD PAK 0 MDV -10 6 7 8 9 10 11 Log(GDP p.c.) at PPP 85 Days to Clear Customs 60 40 Constraint level, conditional means; quadratic fit; 5% robust CIs NPL 20 PAK AFG BGD IND LKA BTN 0 MDV 6 7 8 9 10 11 Log(GDP p.c.) at PPP Bribes (>1% sales) Constraint level, conditional means; quadratic fit; 5% robust CIs 1 .8 .6 AFG .4 BGD PAK .2 MDV IND NPL BTN LKA 0 6 7 8 9 10 11 Log(GDP p.c.) at PPP 86