Policy Research Working Paper 10048 Firm-Level Input Distortion in Indian States Ritam Chaurey Ruchita Manghnani Viviana M.E. Perego Siddharth Sharma South Asia Region Office of the Chief Economist May 2022 Policy Research Working Paper 10048 Abstract This paper measures trends in factor misallocation in India and by firm size. Overall, the findings show that adjust- between 1999 and 2014, using data from a rich panel of ment costs declined over time for labor and land but with Indian firms. The misallocation of a factor is modeled as an significant heterogeneity with respect to state growth rate adjustment cost, that is, an implicit variable cost incurred and firm size. Using these stylized facts on trends in factor by a firm when using that factor. Trends in the adjustment adjustment costs, as well as in-depth field interviews with cost are estimated using a new adaptation of the firm-level firms in two Indian states, the paper also discusses potential cost-minimization approach. The paper documents these policy developments behind these trends, including a pre- trends for four factors of production (permanent labor, liminary examination of the role of state-level governance contract labor, land, and fixed capital) across Indian states in the implementation of relevant factor market policies. This paper is a product of the Office of the Chief Economist, 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://www.worldbank.org/prwp. The authors may be contacted at ssharma1@worldbank.org. 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 Firm-Level Input Distortion in Indian States Ritam Chaurey 1 Ruchita Manghnani 2 Viviana M.E. Perego 3 Siddharth Sharma 4 Keywords: Misallocation, Firms, Distortions, Land Markets, Labor Markets, India JEL Codes: O12, O43, L20, D22 1 SAIS, Johns Hopkins University 2 The World Bank 3 The World Bank 4 Corresponding Author, The World Bank. Email: ssharma1@worldbank.org. The views expressed are those of the individual authors and do not necessarily reflect those of the World 1. Introduction5 There is growing evidence that factor misallocation has constrained firm performance in India. A large literature suggests that rigid industrial labor regulation reduces employment and productivity in firms. Besides labor, there has been rising interest in understanding the misallocation of land and capital, and their effects on the manufacturing sector. Some recent studies suggest that rigid land use regulation, such as land ceiling laws, have reduced productivity. Research on a related topic – the impact of policies promoting industrial parks and Special Economic Zones – is also growing. Misallocation is important to study because it suggests that aggregate output could be increased by reallocating existing resources towards firms that can use them more productively. By one estimate, simply reallocating resources between existing Indian firms to match U.S. levels of efficiency in resource allocation could increase India’s aggregate productivity by 60%-80% (Hsieh and Klenow, 2009). Factor misallocation could be one of the reasons for the absence of economic convergence between India’s states. The widening economic gap between Indian states has policy makers increasingly concerned.6 It is a puzzle, contradicting a basic paradigm in growth theory – the idea of convergence — and contrasting with the experience of the U.S. and Japan (cf. e.g., Barro and Sala-i-Martin, 1992; Caselli and Coleman, 2001). Since state governments have jurisdiction over many policy areas affecting factor markets, the design and implementation of factor market policies varies across states. It could be that some states remain poorer than others not because of a deficiency of resources, but because on an inefficient allocation of existing resources. Despite the strong influence of state governments on factor market policies, there has been surprisingly little systematic analysis of how factor market misallocation and related policies vary across Indian states. This paper contributes to this policy-relevant research agenda by using Indian firm-level panel data from 1999-2014 to describe trends in factor market distortions across Indian states. We look at ‘permanent’ labor, temporary (‘contract’) labor, land and fixed capital, and establish stylized facts about how factor market distortions have evolved across Indian states.7 Complementing the results of the data analysis with information learned through firms interviews in two states, we suggest areas of focus for future policy analysis. The results of this paper can also be a useful filter for identifying cases of successful reforms or good practices in individual states which could be shared with other states. We adapt the cost-minimization approach (De Loecker and Warzynski, 2012) to measure trends in factor ‘adjustment costs’ over the years. The ‘adjustment cost’ of a factor is conceptualized as an implicit variable cost incurred by a firm in using that factor. This conceptualization is similar to the ‘implicit taxes’ on inputs modeled in Hsieh and Klenow (2009), a seminal study of factor misallocation in India. The potential causes of this implicit cost include market frictions, institutional factors and factor market regulations which make it harder for firms to employ or adjust their inputs. The analysis indicates that at the all-India level, the adjustment costs for permanent labor, contract labor and land have fallen between 1999 and 2014. There was a small increase in the adjustment cost for permanent labor between 2004-2007, but the trend has been downwards since 2009. The decline is most 5 We are grateful to Amit Khandelwal for in-depth advice on how to estimate adjustment costs. We also thank Sam Asher, Leonardo Iacovone, Rinku Murgai and Bob Rijkers for helpful comments and suggestions. We thank Sarur Chaudhary, Ruifan Shi and Sargam Jain for superb research assistance. 6 See, for example, Chapter 10 in Government of India (2017). 7 We distinguish between permanent labor (workers employed directly by the firm) and contract labor (workers employed through labor contractors) as they are subject to different sets of regulation. 2 dramatic in the case of contract labor. In contrast to labor and land, the adjustment cost for fixed capital (excluding land) has not declined over the years. We examine how these trends vary across states and by firm attributes such as size.8 Splitting states into two groups based on their manufacturing growth rates over the study period, we find that the decline in permanent and contract labor adjustment costs has been significantly faster in fast-growing states compared to slow-growing states. The decline in land and capital adjustment costs, in contrast, is not correlated with state-level growth performance. As regards firm attributes, we show that firm size plays a role, particularly for land. The decline in permanent labor adjustment costs has been experienced by firms of all sizes, but at a significantly faster pace among ‘large’ firms (those with more than 50 workers) compared to `small’ firms (those with 10 to 50 workers). Contract labor adjustment costs, in contrast, have fallen faster among smaller firms. Strikingly, land-related adjustment costs have trended in opposite directions for large firms relative to small firms, declining for large firms and increasing for small firms. Firm-level variation in the adjustment cost of an input is associated with a misallocation of that input across firms (Hsieh and Klenow, 2009). For instance, the variance of factor adjustment costs across firms within a state is indicative of within-state misallocation across firms. We find that the variance of permanent and contract labor adjustment costs has been falling faster in fast-growing states compared to slow-growing states, suggesting that the within-state misallocation of labor is falling faster in faster growing states. But for land, the within-state dispersion has been rising for both fast- and slow-growing states. Discussing the potential causes of these trends, we also present a review of the institutional framework concerning labor and land in India, supported by grassroots evidence on the constraints faced by firms in the country gathered through our field interviews with small and medium enterprises in Telangana and Uttar Pradesh. Among the most interesting considerations emerging from the exercise, we document an important role of governance in shaping the effective enforcement of factor market policy, and show that broad state- level governance measures are strongly correlated with firms’ perceptions about business environment, regulatory ease, and transparency. Using an index of governance quality to split states into a “high” and a “low” governance group, however, we puzzlingly find that adjustment costs for certain productive factors fell significantly more in the low-governance group, compared to the high-governance one. This pattern is robust to restricting the sample to neighboring districts located on opposite sides of the borders between low and high governance states, increasing our confidence that it is related to initial differences in state- level institutions, rather than other spatially varying factors. Interestingly, a preliminary examination of available indices of governance quality over time suggests a slow convergence in governance quality across Indian states. The states that did not have robust governance structures earlier seem to have also been the most effective in improving their governance over time, which could in turn have produced a more sustained reduction in firms’ adjustment costs. The rest of this paper is as follows. Section 2 discusses the relevant literature. Section 3 describes our empirical methodology, as well as our main data sources. Section 4 discusses our results on the main trends in factor adjustment costs, and on how they vary by firm size and across fast- and slow-growing states. Drawing on in-depth firm interviews conducted in two states, Section 5 discusses the key factor market regulations which could be relevant to explaining these results. Section 6 conducts a preliminary exploration of the role of governance, examining how the trends in adjustment costs have varied across high- and low- governance states. Section 7 concludes with suggestions for priority areas for future policy analysis, as suggested by the results of this paper. 8The firm-level data set used for the analysis, the Annual Survey of Industries, covers factories with 10 or more workers. We are unable to measure these trends for firms with fewer than 10 workers. 3 2. Literature A growing body of work examines how firm-level misallocation affects aggregate productivity. This literature is reviewed comprehensively in Restuccia and Rogerson (2017). One branch of this literature has focused on quantifying resource misallocation across firms and its impact on aggregate productivity. The misallocation considered is of the ‘static’ variety, in the sense that the distribution of firm-level productivity is taken as a given. Studies of this type on India suggest that the misallocation of resources towards unproductive firms – much of which is related to factor market issues – has significantly constrained productivity growth in India. Given the distribution of firm-level productivity, Hsieh and Klenow (2009) estimate that simply reallocating resources between existing Indian firms to match U.S. levels of efficiency in resource allocation could increase India’s aggregate productivity by 60%-80%. Market distortions can also have an adverse effect on firms’ lifecycle decisions to invest in productivity improvement. This ‘dynamic misallocation’ could be a reason why the distribution of firm-level productivity differs across countries. For example, older plants in India are smaller and much lessproductive than their counterparts in the United States (Hsieh and Klenow, 2014). This could be because larger, more productive firms in India face higher regulatory costs, reducing their incentive to invest in better management and innovative capacity. Firm-level data availability is often a binding constraint on how well firm-level misallocation and productivity can be measured. Rotemberg and White (2019) argue that the measured differences in misallocation between the US and India measured by Hsieh and Klenow (2009) study could be an artifact of differences in firm-level data cleaning procedures across the US and India. This issue is less of a concern in our study, which is focused on within-country differences over time and relies on estimates from a single survey. Further, as explained in our methodology section below, we limit our work to measuring quantities that do not require production function estimation, as that would put additional demands on the data. Another strand of the misallocation literature examines the impact of specific factor market-related polices on resource allocation and productivity. In the case of India, there is extensive work on inflexible labor laws in India and their effects on industrial performance. A large literature has found negative economic impacts of amending the IDA regulations in pro-worker directions, thereby making it harder to fire permanent workers - lower output, employment, investment, and productivity in formal manufacturing (Aghion et al., 2008; Ahsan and Pages, 2009; Besley and Burgess, 2004), lower sensitivity of industrial employment to local demand shocks (Adhvaryu et al., 2013) and lower employment in the retail sector (Amin, 2009). A further consequence of these inflexible labor laws has been the increased use of contract workers, who are not subject to IDA regulations (Chaurey, 2015; Ramaswamy, 2013; Sen et al.., 2013). These workers have temporary contracts and are very often hired indirectly through a contracting agency. Bertrand et al. (2017) show that an increased reliance by firms on contract labor is associated with an increase in their size, a decrease in the average product of labor, an increase in employment variability, anda decrease in the average cost of labor. Amirapu and Gechter (2017) look at the effects of labor regulationsimposed on firms with 10 or more workers and not covered by the IDA and find that they substantially increase the firm’s unit labor costs. Furthermore, they find a positive association between regulatory costs and corruption measures at the state level. 4 Apart from labor market distortions, there has been an increased focus on the misallocation of land and building inputs and its effects on the manufacturing sector.9 Duranton et al. (2015a) show that land and building misallocation in India leads to output misallocation in local areas and leads to lower labor productivity in districts. They estimate that a one standard deviation decrease in the misallocation of land and buildings is associated with a 20-25% increase in output per worker. In particular, the paper also looks at the repeal of the Urban Land Ceiling and Regulation Act (ULCRA, cf. infra) between 1999 and 2003, which led to large reductions in the misallocation of land and buildings in the areas where this strict regulation was in place: the repeal of the ULCRA seems to have resulted in an increase in output per worker of about 3% in treated areas. Furthermore, land misallocation is also an important determinant of financial misallocation for firms because land is often used as a collateral for loans in India (Duranton et al., (2015b)). In this regard, both Special Economic Zones and Industrial Areas have played a significant role in fostering industrial growth. Hyun and Ravi (2017) find that because of the 2005 SEZ Act in India, firms in the formal sector grew, and there was a shift in economic activity from the informal sector to the formal sector. Blakeslee et al. (2018) study the effects of the Industrial Areas (IAs) program that facilitated the establishment of industrial firms in areas that had previously been restricted to agriculture. They find that IAs caused a large increase in the number of firms and employment, and that there were substantial spill- overs to neighboring villages. 3. Methodology and data 3.1 Theory: Using a cost-minimization approach to infer factor adjustment costs Our method for estimating the adjustment cost is adapted from the cost-minimization approach used in studies such as De Loecker and Warzynski (2012) and De Loecker and others (2016). One advantage of this method is that it does not require us to specify a particular structure of market demand for output or estimate a specific production function.1 The following is a simplified description of the methodology in De Loecker and Warzynski (2012). 1 2 Consider a firm i’s production function with two variable factors of production, and .2 Cost minimization by the firm gives us the Lagrangian: 1 2 1 1 1 2 2 2 1 ( , , ) = ( ) + ( ) + ( − (. )) (1) Here t is time, is the level of production that the firm is trying to minimize costs for. is the price paid by firm i for factor v. The price can vary by firm, but it cannot be affected by the firm (that is, the firm acts as a price taker in factor markets). For instance, the price can vary by firm location. Here, let ≥ 1 be the adjustment cost for factor v. It is a variable cost that is not explicitly paid (that is, it is not included in the firm’s reported expenditure on the factor). It is similar to the concept of “implicit input tax” in Hseih and Klenow (2009). The first order conditions for this Lagrangian are: 1 This exercise resembles previous work by Ghani et al. (2012), who estimate misallocation at the district level. However, there are two key differences. First, the method in Ghani at al. is based on recovering an index of misallocation which is not based on a specific model of markets and firm behaviour. In contrast, the method we propose recovers firm-level adjustment costs which have a specific interpretation in the context of a model of cost-minimization by firms and are more amenable to further analysis on the causes of misallocation. Second, Ghani et al. do not focus on inter-state differences. 2 The model can be generalized to the case of more than two variable inputs. The key assumption is that the production function is continuous and twice differentiable with respect to at least one input, and that input is adjustable. This assumption restricts the technology so that the firm can adjust its output quantity by changing at least one variable input. 5 = − (2) = (3) where is the firm’s mark-up and = , or the ratio of the expenditure on input v to sales.3 The expression denotes the elasticity of output with respect to input v, which would need to be estimated from a production function. Equation (3) is similar to De Loecker and Warzynski (2012), but with a multiplicative adjustment cost.4 Suppose we have firm-level input and output data, including data on input expenditures and revenue. We could then use equation (3) to infer the product of the price mark-up and the adjustment cost, , but it would require estimating a production function to measure the elasticity of output with respect to input v. However, it is possible to measure the change in input adjustment costs over time without estimating a production function. We make two assumptions: A1. The production function is Cobb-Douglas. This implies time invariant output elasticity. A2. The production function of firm i stays fixed over the study period. These assumptions are common in studies using firm-level panel data to estimate production functions. If we were to estimate the production function, we would have to make additional assumptions to address the endogeneity of input use. Taking ratios of equation (3) for two inputs, we get 2 2 1 1 = 1 2 . (4) The ratio of input expenditures is inversely proportional to the ratio of the input adjustment costs. 1 To simplify, set = 1. In other words, we are redefining the input 2 adjustment cost relative to that on 2 input 1. If > 1, this would mean that, in sum, factor market distortions are taxing input 2 relative to input 1. Note that the ‘implicit input tax’ τ in the two-input model in Hsieh and Klenow (2009) is defined in this relative sense: an input market distortion that favours one factor of production over the other. Taking logs of equation (4) we have: 2 1 1 log ( ) = log 2 + log 2 (5) Since the production function is fixed over time, taking the difference between two consecutive time periods 1 and 2, we have 2 2 log ( ) = log 1 (6) 3 De Loecker and Warzynski (2012) define the output markup as the output price to marginal cost ratio. Given this definition, the optimal cost-minimization expression in Equation 3 is not dependent on the form of (static) price competition among firms. 4 If there are no adjustment costs to either input, then the mark-up computed from either input should be the same. This also implies 1 2 1 2 =1 6 We can thus recover the change in the factor adjustment cost from data on the change in the ratio of input expenditures. This requires no production function estimation. It is a double difference within a firm across inputs over time. Importantly, z can be interpreted as the τ from Hsieh and Klenow (2009), where τ’s are distortions that raise the marginal product of one input relative to the other input. This can be generalized to the case of k variable inputs, in which changes in k-1 relative adjustment costs over time can be estimated. Note that the price mark-up does not appear in equations (4) to (6). This method cannot be used to estimate the price mark-up - or, more generally, any distortion that affects all inputs proportionally (such as the ‘input-neutral implicit tax’ in Hsieh and Klenow, 2009)). As equation (3) shows, estimating the mark-up and the level of the adjustment cost would require production function estimation. 3.2 Empirics: Estimating trends in factor adjustment costs from firm-level panel data The main objective of our descriptive analysis is to apply equation (5) to firm-level data to estimate how factor adjustment costs have changed over time in Indian states. Suppose there are k variable inputs, with being the adjustment cost of input v (relative to input 1) for a firm. Rearranging Equation (5), 1 1 log = log ( ) − log (7) 1 where the ratio of output elasticities, log , is by assumption fixed over time. Now suppose that log ( ) = + . + ∑ . . + (8) This equation specifies that the adjustment cost consists of a firm-level fixed component , annual shocks which depend on a set of initial firm characteristics , and an idiosyncratic firm-specific shock . The initial firm characteristics could include, for example, the state in which the firm is located, firm size, sector and age. We do not observe log ( ), but can infer the ′ and ’s – the changes over time –from the ratio of input 1 expenditure shares log , as described in equation (7). Combining equations (7) and (8), we arrive at our baseline regression specification: 1 log = + . + ∑ . . + (9) 1 where ≝ − log 2 . We are mainly interested in how the adjustment cost varies over time by state. As specified in equation (8), we regress the ratio of firm-level input shares on firm fixed effects and on year dummies interacted with state dummies. The firm fixed effect absorbs the time-invariant output elasticity ratio and all time invariant components of the adjustment cost. Thus, the regression identifies how the adjustment cost has changed relative to the baseline year, and how that change varies by state. We are also interested in how the adjustment cost varies over time for firms of different size and age. This 7 involves regressing the ratio of firm-level input shares on firm fixed effects and on year dummies interacted with initial firm size and age category dummies. 3.3 Estimating the within-state variance in adjustment cost We also examine the variance in the factor adjustment cost across firms within the same state, and how that has changed over time. A higher within-state variance in the adjustment cost for input v implies that there is greater misallocation of that input across firms within the state. In the Hsieh and Klenow (2009) model, the variance of the implicit input tax τ within an economy is positively correlated with the extent of firm- level misallocation (and the loss of aggregate productivity due to firm-level misallocation). To measure the variance of log ( ) in a cross-section of firms, we make the additional assumption that the production function is invariant not only over time, but also across firms within the same 3-digit industry 1 in India. This would imply that the input elasticity ratio log 2 is invariant over time, and across firms within the same 3-digit industry in India. Thus, equation (9) modifies to: 1 log = + + . + ∑ . . + (10) Here, the industry fixed effect absorbs the input elasticity ratio. 1 We implement this in two stages. First, we regress log on 3-digit industry fixed effects and compute the residual from the regression. Per equations (9) and (7), this residual is directly proportional to the level of the input adjustment cost. We then compute the variance of this residual across firms within every state, for every year t. If one is willing to accept the additional assumption about the production function being invariant within the same 3-digit industry group in India, this variance is directly proportional to the variance of log ( ) within the state. 3.4 Choosing the reference input As discussed, with a total of k variable inputs, we can measure changes in the adjustment costs of k-1 inputs relative to a reference input. The choice of reference input matters to the interpretation of the results. For example, if the reference input is fixed capital (excluding land), then any policy change which has the same effect on land and fixed capital will be difficult to discern, as the relative adjustment cost of land will be unaffected by it. Further, a good reference input should be relatively frictionless, to make it less likely that changes in the adjustment cost of the reference input might dominate changes in our measured ratios. We pick raw materials as the reference input on the supposition that it is a relatively frictionless input, and plausibly not affected directly by land and labor market policies. Our results are however generally robust to using expenditure on purchased electricity as the reference input. However, as many firms in India use electricity generators as a backup source of power, we prefer raw material because to avoid imputing a value to self-generated electricity. 3.5 Data The main sources of data for our analysis are summarized in Table 3.1. 8 The bulk of our work is conducted on data from the Annual Survey of Industries (ASI) by the Indian Ministry of Statistics and Programme Implementation (MoSPI). The survey spans over many years and covers the entire manufacturing sector comprising industrial units (“factories”) registered under Sections 2(m)(i) and 2(m)(ii) of the Factories Act, 1948.14 This includes all firms employing 10 or more workers using power, and those with 20 or more workers regardless of the use of power. The ASI data collect detailed information on employment, material inputs, and output that can be conveniently used to recover mark-ups and estimate production functions. The data set also provides information on the location of the factory at the state and district level. The ASI data are available both as detailed, unit-level cross-sections without firm identifiers, and as panel data with firm identifiers.15 Both these data sets have been widely used in the literature, most notably by Allcott et al. (2016), Hsieh and Klenow (2009), and Martin et al. (2017). In our work, we use the panel version for the period 1999-2014 for the sample of the major Indian states.16 For our analysis on adjustment costs, we use four main outcome variables from the ASI – wage bill of permanent labor, wage bill of contract labor, value of land (owned and leased), and value of fixed capital excluding land. In line with our discussion in Section 3.2.3, we further divide all our outcome variables by the expenditure on materials consumed (or on electricity consumed and purchased in our robustness checks), which is our reference input for measuring relative input adjustment costs.17 Descriptive statistics by year on the main ASI variables used in our analysis are reported in Table 3.2. As regards our classification of fast- and slow-growing states, we use value added time series provided by the Institution for Transforming India Policy Commission of the Government of India (NITI Aayog) and compute the growth rate of manufacturing value added from 1999 to 2014. States are classified as fast- growing if they have an above-median rate of growth (Annex Table A3.1). Our main governance classification is sourced from a study carried out by Transparency International India and the Centre for Media Studies (CMS) in 2005 on the experience of the average citizen with 11 public services18 across the country. The measure combines information on aspects such as level of satisfaction in the services of the relevant State Department, or on number of households reporting paying bribes, or using influence to speed up administrative processes. States are then ranked with respect to a weighted average of the scores obtained by the various Departments, and we assign a State to a “low governance” group if its score lies in the bottom 50 percent of the ranking. Alternative governance indicators come from the Indian National Council of Applied Economic Research (NCAER), the Enterprise Survey of the World Bank, the Indian National Crime Records Bureau, and various policy reports, and are described in more detail in Section 6. 14 The main central government’s law on conditions of work. See Section 5 for details. 15 The cross-sectional data sets are available for the years 1983-84, 1984-85, 1989-90, 1993-94, 1994-95, and more recently from 1996-2014, whereas the panel data is available from 1998 onwards. 16 Our final sample includes Andhra Pradesh, Assam, Bihar, Chhattisgarh, Delhi, Goa, Gujarat, Haryana, Himachal Pradesh, Jharkhand, Karnataka, Kerala, Madhya Pradesh, Maharashtra, Odisha, Punjab, Rajasthan, Tamil Nadu, Telangana, Uttar Pradesh, Uttarakhand, West Bengal. 17 To define expenditure on materials consumed, we use the Tabulation Programme provided by the MoSPI. Materials consumed include expenditure on total basic items, non-basic chemicals, packing items, consumable stores, and total imports consumed. 9 4. Trends in factor adjustment costs In this section, we show our results on the trends in adjustment costs for permanent labor, contract labor, land, and fixed capital (all relative to raw material), as produced by running Equation (9) from Section 3.2.1 on our sample of Indian firms between 1999 and 2014. For each input, we show three sets of graphs corresponding to a) all India trends in adjustment costs; b) trends in adjustment costs in fast-growing states relative to slow-growing states; and c) trends in adjustment costs for large firms relative to small firms. The regression tables from which these graphs were generated are shown in the Annex. The all-India graphs plot the α coefficients on the year dummies in our baseline regression (9). The graphs for fast- and slow-growing states and for firm size, instead, plot the β coefficients on the interaction between the year dummies and the covariates’ matrix X. We respectively focus on an indicator for fast-growing states (cf. Section 3.3), and on an indicator for whether firm size is greater than or equal to 50 employees.19 In each of these graphs, the omitted year is 1999, which means that all the year dummy (and respective interaction) coefficients plotted over time should be interpreted as being relative to 1999. For example, a negative coefficient on a year dummy should be interpreted as a reduction in the adjustment cost for that input in that year relative to 1999. 4.1 Permanent labor We first look at the trends in adjustment costs for permanent labor in Figure 4.1. At the all-India level in panel 4.1a, adjustment costs for permanent labor have gone down over the time relative to 1999, our starting year. Although there was a slight increase in the trend between 2004-07, the adjustment cost for permanent labor has declined again post-2009. The corresponding regression results are shown in column 1, Annex Table A4.1. In Figure 4.1b, we plot the differences in adjustment cost trends for permanent labor between fast- and slow-growing states. The trends in adjustment costs for permanent labor in fast-growing states are lower than the corresponding trends in slow-growing states, and the differences in the coefficients are statistically significant over the years. In other words, the adjustment costs on permanent labor have been falling faster over time for firms in fast-growing states compared to slow-growing states. The corresponding 18 The services considered in the study are electricity, government hospitals, income tax services, judiciary (lower courts), land administration, municipal services, police, public distribution systems, rural financial institutions, public schools up to 12th grade, water supply. 19 The cutoff is chosen to reflect the size threshold after which various provisions of the Industrial Disputes Act (IDA) become binding (cf. Section 5). Results are robust to using a higher cutoff of 100 employees (after which many IDA regulations become even stricter). 10 regression results are shown in Annex Table A4.2, column 1. Finally, in Figure 4.1c (regression results in Annex Table A4.3, column 1), we find that the trends in adjustment costs on permanent labor are lower over time for larger firms than for smaller firms, and these differences are statistically significant for most years after 2003. This suggests that adjustment costs on permanent workers have been falling faster for larger firms relative to small firms. The declining trend across India may be related to the relaxed implementation of the provisions of regulations pertaining to permanent workers (see Section 5), thereby making it easier for firms to hire permanent workers over time. The difference in the declining trend of fast- and slow-growing states in turn provides suggestive evidence that the implementation of labor laws may have varied across states, with fast-growing states making it relatively easier to hire permanent workers than slow growing states. Finally, considering that labor regulations are stricter for larger firms (see Section 5.1), our finding that the decline in adjustment costs was more pronounced for these establishments may be also suggestive of a relaxed implementation of labor laws. Figure 4.1: Adjustment cost trends – Permanent labor a: All India b: Fast-growing states relative to c: Large firms relative to small slow-growing states firms Data source: ASI 1999 – 2014. Note: Coefficients on year dummies in Panel a (baseline year = 1999). Coefficients on interaction of year dummies with a dummy for fast growing states in Panel b. Coefficients on interaction of year dummies with a dummy for large firm in Panel c. 4.2 Contract labor In Figure 4.2, we look at the trends in the adjustment costs for firms for hiring contract labor. In Figure 4.2a, we find a secular decline in the adjustment costs from hiring contract workers faced by firms all over India. Furthermore, in comparison to slow-growing states, firms in fast-growing states experienced a larger decline in contract labor-related adjustment costs (Figure 4.2b). The differences between firms across fast- and slow-growing states are also statistically significant (see Annex Table A4.2, column 2). Small firms saw larger reductions in these adjustment costs compared to large firms (Figure 4.2c). These differential effects are statistically significant after 2009 (see Annex Table A4.3, column 2). 11 Figure 4.2: Adjustment cost trends – Contract labor a: All India b: Fast-growing states relative to c: Large firms relative to small slow-growing states firms Data source: ASI 1999 – 2014. Note: Coefficients on year dummies in Panel a (baseline year = 1999). Coefficients on interaction of year dummies with a dummy for fast growing states in Panel b. Coefficients on interaction of year dummies with a dummy for large firm in Panel c. 4.3 Land We next look at trends in adjustment costs for land faced by firms over time in Figure 4.3. In Figure 4.3a, we find that adjustment costs for land have substantially declined over time, especially since 2006. In Figure 4.3b, we compare the land-related adjustment costs for firms across fast- and slow-growing states. We do not find statistically significantly different effects across these groups after 2005, indicating that adjustment costs for land have gone down for firms in both fast- and slow-growing states. Finally, we compare adjustment cost trends across large and small firms in Figure 4.3c. The relative trend for large firms is negative, and sizably so compared to the average trend for land. Indeed, the estimates imply that land- related adjustment costs have trended in opposite directions for large firms relative to small firms. For large firms, adjustment costs have gone down over time, whereas for small firms these adjustment costs have risen over the years. This suggests that over the years, it has become increasingly easier for large firms to access industrial land; on the contrary, for small firms accessing land remains a major constraint, which has even been steadily increasing over time. These differential trends across firm sizes are statistically significant (see Annex Table A4.3 column 3). 4.4 Fixed capital We finally turn our attention to trends in adjustment costs related to fixed capital in Figure 4.4. In Figure 4.4a, we find evidence of an increasing trend in fixed capital-related adjustment costs for firms across India. Although adjustment costs were lower than in 1999 until 2009, they rose after 2010. In Figure 4.4b, we compare these costs for firms in slow- and fast-growing states. There does not seem to be a clear pattern in the differences between firms in the two sets of states. In Figure 4.4c, we find that fixed capital-related adjustment costs have trended upwards relatively slowly among larger firms (also see Annex Table A4.5, column 4). 12 Figure 4.3: Adjustment cost trends – Land a: All India b: Fast-growing states relative to c: Large firms relative to small slow-growing states firms Data source: ASI 1999 – 2014. Note: Coefficients on year dummies in Panel a (baseline year = 1999). Coefficients on interaction of year dummies with a dummy for fast growing states in Panel b. Coefficients on interaction of year dummies with a dummy for large firm in Panel c. Figure 4.4: Adjustment cost trends – Fixed capital (excluding land) a: All India b: Fast-growing states relative c: Large firms relative to small to slow-growing states firms Data source: ASI 1999 – 2014. Note: Coefficients on year dummies in Panel a (baseline year = 1999). Coefficients on interaction of year dummies with a dummy for fast growing states in Panel b. Coefficients on interaction of year dummies with a dummy for large firm in Panel c. 4.5 Within-state variance in adjustment costs 13 Dispersion in factor adjustment costs across firms within a state would be indicative of factor misallocation across firms within the state (Hsieh and Klenow, 2009). For example, a state-level distortion that implicitly makes it cheaper for larger firms to access land as compared to small firms would in turn lead to higher dispersion in the adjustment costs across firms in that state. In this subsection, we therefore analyze the trends in the dispersion of adjustment costs across firms within fast- and slow-growing states. Formally, as discussed in Section 3.2.2., we first infer the levels of the adjustment costs at the firm level by using the residuals from regression of input expenditure shares on industry fixed effects, as in Equation (10) above. Then, we calculate the dispersion of these adjustment costs across firms for each state-year cell.20 Finally, we average these variances over fast- and slow-growing states, over time, and for each of our four productive inputs. Figure 4.5 shows the results of the analysis. For permanent labor, the within- state variance has declined more for fast-growing states compared to slow-growing states. Similarly, for contract labor, after 2005, the within-state dispersion in adjustment costs has declined more for fast-growing states. This means that the wedge between large and small firms in terms of labor-related adjustment costs has narrowed more, over time in fast- than in slow-growing states. Interestingly, for land, the within-state dispersion has trended upwards over time for both groups of states: in both sets of states, therefore, there seems to have been a divergence in the land-related adjustment costs faced by different firms. Figure 4.5: Within-state variance in adjustment costs Data source: ASI 1999 – 2014. 4.6 Alternative interpretations and robustness checks A first, obvious concern with our findings is that the trends we observe in adjustment costs for firms might be due to organic technological changes rather than to misallocation over time: for instance, the use of new machineries may require installation, adaptation, training of the workforce, which would all affect relative input costs even absent any change in factor market distortions. To the extent that natural technological 20 Figures A4.1-A4.4 (see Annex) show these plots for each state for the four inputs. 14 changes happen at the industry level over time, they should be adequately captured by the inclusion of industry-year fixed effects in our baseline regressions. In our robustness exercises, therefore, we control for a set of industry-year interaction dummies, and obtain very similar trends as in our main specifications, regardless of whether sectors are identified at the two- or three-digit level. 21 A potential confounder to the interpretation of our analysis could be related to selection bias as firms enter and exit the market in response to unrelated changes in the business environment. For example, distortions may result spuriously high both in states that are growing rapidly and have many entrants who have not yet reached productive efficiency, and in less well-off states that have a high share of struggling and exiting firms – but this would obviously have very different policy implications in the two cases. To deal with this concern, we first exploit information in the ASI data on the reasons why an establishment in the sample frame could not be included in the survey, in order to get some information on firm closures.22 Annex Table A5.1 shows that in our sample period there was indeed quite a high turnover of firms, but that the difference between the firm exit rates of high- and low-growth states does not seem to be substantial, nor to follow a defined pattern over time. In Annex Table A5.2, we also document firm entry,23 which turns out to be equally limited in our sample, as well as evenly distributed across states with different rates of growth. As a further robustness check, we also replicate our analysis excluding new entrants from the sample and find very similar results to our baseline specification. Finally, one could be concerned that land as an input is quite different in scope from capital or labor, as firms do not often make year to year changes in their land holdings, instead undertaking lumpier land adjustments over longer periods of time. In that case, our results might be picking up noise or measurement error, rather than true adjustment costs. To avoid this source of bias, we thus replace our year-on-year specification for land-related adjustment costs with long differences (3- and 5-years): even though only around 20 percent of our original sample have sufficient information on land usage to allow for long differences estimation, we reassuringly find comparable results to our main specification. 5. Relevant factor market policies in Indian states Could differential change in land and labor regulation across India states explain the trends in their adjustment costs? While finding specific causal mechanisms for these patterns is beyond the scope of this paper, the institutional framework surrounding factor markets within and across Indian states is critical to understanding the implications of our results. This section presents a short survey of the most salient aspects of land and labor regulation in Indian states, complemented by in-depth firm interviews in two states (Telangana and Uttar Pradesh)24 which were conducted during 2017-18 to gather grassroots information on broad issues of governance, policy implementation, and law enforcement in these areas. 21 Results for this section are available from the authors upon request. 22 For this exercise, we consider as closure the following categories: closed, non-operational, non-existent within three years, closure but in existence, non-existent now, non-existent for more than three years, deregistration. 23 We define as new entrants those establishments in their first year of operations, that is, those created in the same year as the survey. 24 We conducted semi-structured interviews and focus groups with more than fifty MSMEs, Industry Associations, consulting firms and informed academics. The two states were selected so that they respectively belonged to our “fast -” and “slow-growth” state groups. The fact that Telangana is a recently formed state (2014) made it also a particularly interesting case-study for our analysis on regulations and quality of institutions. 15 5.1 Labor The system of labor legislation and rules in India builds on a complex architecture. Labor regulations vary across government authorities (central and states), establishment sizes, and types of laborers. The Indian Labor Bureau (2004), for instance, estimates that in 2003 a typical enterprise might have to deal with about 25 central government and between 10 to 15 pieces of state regulations. The practical costs of complying with these regulations in turn depend heavily on various characteristics of the firm, such as size and composition of the workforce. The main law governing conditions of work is the Central Government’s Factories Act 1948. The Act contains specific requirements in terms of health and safety of factory workers, including hours of work, overtime and annual leave, and female and child labor conditions. The Act applies to all manufacturing premises that use electricity and employ 10 or more workers, and to non-powered premises with 20 or more employees. State governments can however lower the threshold number of workers, except in the case of family-based businesses (Mitchell et al., 2014). The regulation of smaller manufacturing units, shops, and other types of small workplaces is enacted through State-level Shops and Commercial Establishments Acts. The Minimum Wages Act 1948 mandates that minimum wages be set for certain types of employment or industries. Depending on the specific case, the competent authority may be the Central government or the State, which means that rates of pay vary not only from industry to industry, but from state to state, and from region to region25 (Mitchell et al., 2014). Variations in wage regulation by establishment size can be found in the Payment of Bonus Act 1965, which mandates payment of an annual bonus to all employees with wages below a specified limit in establishments employing at least 20 persons, and in the Equal Remuneration Act 1976, which prescribes equal pay for equal work between male and female workers in establishments with 10 or more employees. Security and welfare obligations also depend to an important extent on firm size. For example, employers and employees contribute to the Employees’ State Insurance Scheme (ESI), which provides insurance cover for employees in the case of death, sickness, workplace injury and disablement, and maternity, but the scheme is limited to enterprises with 20 or more workers. Similarly, compulsory contributions to the Employees’ Provident Fund (EPF) are mandated only for establishments above the 20-workers threshold and operating in certain industrial sectors. The Payment of Gratuity Act 1972 adopts a lower threshold, and provides for a gratuity at the point of superannuation, retirement, or resignation, to be paid to employees with continuous service of at least five years, and by establishments employing at least 10 workers. The most frequently cited source of distortions in the Indian labor law (Bhattacharjea, 2006, 2017), however, is the Industrial Disputes Act 1947 (IDA). The IDA covers a vast array of issues, from resolution of industrial disputes, to hiring and firing workers, to closure of establishments in the formal sector (Chaurey, 2015). The rules on lay-off, retrenchment and closure in particular have been identified as especially distortive (Chaurey, 2015; Dougherty et al., 2014; Ramaswamy, 2012): Chapter V-A of the Act requires notice and compensation to firms with 50 or more workers, and Chapter V-B requires notice, compensation, and permission from government authority if the firm has more than 100 employees.26 Similarly, an establishment that were to close down would be required to inform the government with a sixty- or ninety- day prior notice, depending on whether it employed more than 50 or 100 workers (Chaurey, 2015). Since its inception in 1947 and its amendment in 1982, the IDA has been further amended multiple times by State 25 For instance, rural areas tend to have lower minimum rates than urban areas, to account for the higher cost of living in the latter. 26 Originally, Chapter V-B applied to firms employing 300 or more workers. The threshold was decreased with the Industrial Disputes (Amendment) Act of 1982, with effect from 1984 (Bhattacharjea, 2017). 16 governments, with swinging levels of support for employers or workers. Besley and Burgess (2004) have documented how these amendments have in practice introduced significant variation in the rigidity of labor regimes across Indian States. Chaurey (2015) and Ramaswamy (2012) have also shown how the IDA regulations might affect the composition of a firm’s labor force, providing strong incentives to employ contract workers (i.e. workers hired through registered labor contractors). As IDA is not applicable to non-permanent workers, their lay- off or termination does not require notice, compensation or permission. Moreover, the size thresholds in the Act are defined based on the number of permanent workers in a given factory, and hence contract labor use is a way to stay below the legal cut-off size for IDA’s provisions. Although an Amendment Act to the IDA in 1982 declared the continuing employment of workers on casual or temporary contracts to be an “unfair labor practice” (Mitchell et al., 2014), and despite some restrictions contained in the Contract Labor (Regulation and Abolition) Act of 1970,27 the use of contract labor has in fact been shown to have increased substantially during the 1990s (Ahsan et al., 2008). The Contract Labor (Regulation and Abolition) Act is applicable to establishments employing at least 20 contractors on any day over the previous 12 months, and entitles contract workers to minimum wages, workplace health and safety provisions, and social security cover such as Employee Provident Fund benefits. This is the primary responsibility of the contractor, but needs to be verified by the employer (the firm).28 This shared responsibility may have introduced ambiguities in the enforcement of the Act. Another regulatory grey area is whether contract workers can be used for the ‘core’ activities of an establishment. In general, the Act does not prohibit this, but some states governments have made amendments to prohibit the use of contract labor for core activities (Rajeev, 2009). 5.2 Land According to some authors (Ghani et al., 2012; Lall and Chakravorty, 2005), land is probably the factor that most severely affects the efficient location choices of manufacturing establishments in India, and as such it might be among the most important drivers of misallocation of industrial production in the country. Despite a long history of land reform ever since Indian independence in fact (cf. Besley and Burgess, 2000), high spatial inequality can be observed in the allocation of agricultural land (Chakravorty, 2012),29 and, at the same time, various policies restrict both the transfer of farmland and the supply of land in urban areas. The transfer of land and the change of land usage are under the jurisdiction of State governments. For example, the transfer of farmland to a non-agriculturist30 is strictly prohibited in Gujarat, Himachal Pradesh, Karnataka, and Maharashtra, with further restrictions for agriculturists from different States. In general, nevertheless, buyers do need to apply for non-agricultural clearance (NAC) from local/State governments to convert farmland to other uses (Morris and Pandey, 2007). Another barrier to efficient land transactions is the requirement in the Indian Stamp Act 1899 that non- judicial stamp duties be paid on the sale and purchase of real property. As each State enacts its own stamp 27 At the Central government level, the Act prohibits the employment of contract Labor in certain categories of work in certain sectors (for example in mines, on railways, and port facilities), and variously regulates work conditions of contract Laborers (cf. Mitchell et al., 2014). 28 Among other provisions, the Act mandates that if the contractor fails to pay contract wor kers’ wages, the principal employer is liable to pay the workers (ILO, 2008). 29 The 2005-2006 Agricultural Census of India documented a national average size of agricultural landholdings of around 3 acres, the lowest average ever recorded. Chakravorty (2012) notes however how the national figure masks significant cross-state variation in plot sizes: while some States like Bihar and Kerala have fragmented, very small holdings, others such as Gujarat, Haryana, Madhya Pradesh, Punjab, or Rajasthan, have much larger, more consolidated ones. 30 A non-agriculturist is defined as an individual not involved in the cultivation of crops and lacking family ties to agriculture. 17 duties, cross-state variation in duty rates is substantial; however, with rates as high as 12-14 percent in many states, and frequent instances of separate stamp incidence for land acquisition and development, thesetaxes tend to be at least five times higher in India than in most high-income countries (Morris and Pandey,2007). Duranton et al. (2015a) report how these prohibitive costs discourage land transactions, thereby reducing land supply on the market, and lead to widespread under-reporting of registration, which in turn stifles investment as land cannot be used as collateral when applying for loans. For urban land, the Urban Land (Ceiling and Regulation) Act 1976 (ULCRA) used to impose ceiling limits for holdings of vacant land, prohibit transfers of land and buildings, and restrict building construction in large urban agglomerations. In 1999, a largely unanticipated Repeal Act gave rights to State governments to repeal it, which happened in a staggered way between 2003 and 2008. Duranton et al. (2015a, 2016) show how the repeal of the Act was associated with a decline in land and building misallocation in the country, and the more so among early-adopter States.31 Despite the ULCRA repeal though, an array of Central and State policies still restricts urban land supply and lead to higher property prices (cf. Chakravorty, 2013 and Duranton et al., 2015a). Chakravorty (2013) for instance cites earmarking for (often inefficient) public use,32 as well as lack of titles in slum areas, which makes the land non-marketable in practice, as examples in this direction. Other harmful measures include the enforcement of low building height and low floor space indices33 (rarely higher than 1.5, compared to a range between 5 and 15 in other Asian cities), urban land ceilings in certain cities like Kolkata, and rent control measures favoring tenants over owners, which in practice transfer ownership rights to tenants, but not the actual title and ability to sell the property. After the Government of India designed a Model Rent Legislation in 1992, many States did actually start formulating new rent acts addressing these dysfunctional rent issues, but the process seems still far from complete (cf. Duranton et al., 2015a for details). A more recent development in land administration is represented by the Digital India Land Records Modernization Programme (DILRMP), launched by the Government of India in 2008. The main objective of the program is to develop a modern, comprehensive, and transparent land records management systemin the country, improving the quality of the land-titling system and minimizing the scope of land and property disputes. Among the major components of the program are the computerization of all land records, the digitization of maps and integration of textual and spatial data, and the survey and re-survey ofall settlement records, including the creation of original cadastral records where necessary. Land records are supposed to be handled through a single-window system, which automates processes following registration and guarantees for the correctness of land titles. The degree of implementation of the program across States varies quite substantially by the specific DILRMP component. For example, while as of today the computerization of the records of rights is complete or almost complete in most of the biggerStates, the full digitization of cadastral maps seems to be far less common, and to display a significantly higher cross-State variation.34 5.2.1 Industrial zones 31 These included Delhi, Gujarat, Haryana, Karnataka, Madhya Pradesh, Orissa, Punjab, Rajasthan, and Uttar Pradesh. 32 Examples include land used for defense (cantonments, army barracks), sick industries, unused airports and rail facilities. 33 The floor space index (FSI) is a measure of the total surface that can be built upon a plot of land. It is computed as the ratio between the floor space covered in all floors of a building to the area of the plot on which the building stands. 34 State-level progress on the major program components can be tracked on the dedicated DILRMP web portal. See http://nlrmp.nic.in/nlrmpmap/nlrmpmap.html#. The IDFC Institute, in collaboration with the World Bank, has also recently embarked on the creation of a land governance index to assess State-level performance in modernizing land records. The index will comprise sub-indicators on (i) transparency of land records; (ii) transparency of land registration; (iii) reliability of land records; (iv) reliability of land registration; and (v) reliability of spatial records. 18 A subject that deserves specific consideration when addressing land issues in manufacturing is the treatment of industrial estates. “Industrial Estate” (IE) is a general label that applies to a number of place -based policies, introduced to promote industrialization. Although the basic idea of industrial parks is always to set aside land for industrial development, in order for firms to benefit from scale economies, shared infrastructure, and clustering, the various types of IEs differ in terms of their economic objectives, the incentives offered, and activities promoted (Blakeslee et al., 2017). For example, States like Karnataka have introduced Industrial Areas (IAs) programs, whereby the government acquires agricultural land, develops it with basic utilities such as power system, recycling, and infrastructure, and then provides it to large non-agricultural firms at market rates. As no financial incentives are offered to firms to locate their operations in the areas, the main advantage for participating establishments is that land development and conversion to non-agricultural use are performed ex ante by the government – so that IAs programs function as a de facto reform of local land-use restrictions (Blakeslee et al., 2017). A far more complex case is represented by Special Economic Zones (SEZs), which were introduced in 2005 with the Special Economic Zones Act.35 These zones, which evolved from existing Export Processing Zones (EPZs), are industrial estates that produce mainly export-oriented goods and are envisioned as comprehensive industrial townships with social facilities for employees of participating enterprises, like housing blocks, schools, and hospitals. Administrative procedures within SEZs are facilitated by the “single-window” mechanism, which enables faster clearances and resolution of bureaucratic red-tape through to revision of applications by a single regulatory body, the Board of Approval, that brings together the Central and State governments (Hyun and Ravi, 2017). SEZ developers and firms, moreover, enjoy strong financial incentives, such as subsidies, tax exemptions, customs privileges, and more flexible labor regulations especially in terms of hiring and firing practices. Besides Central-level incentives, States also waive a number of other levies (e.g. water and electricity duties) and make compliance with certain regulations (e.g. for environmental assessments) more agile for firms located within the SEZ.36 More recently, the Indian government has launched a new zonal development model, the National Investment and Manufacturing Zones (NIMZs), in the ambit of its National Manufacturing Policy 2011. Although similar in spirit (large integrated industrial townships with necessary social infrastructure and facilities, land use on the basis of zoning, “single-window” mechanism), the NIMZs differ from SEZs under many respects. First, while SEZs mostly target export-oriented production, NIMZs focus substantially on domestic technology and manufacturing: as such, while the service sector tends to be predominant in the former, for the latter there is a requirement that 30 percent of the area be reserved for manufacturing. As a result, NIMZs are also way bigger in size: where SEZs are allowed to cover a surface ranging between a minimum of 10-1,000 hectares depending on the sector (IT zones can be small) and a maximum of 5,000 hectares, the minimum size for NIMZs is 5,000 hectares. Finally, incentives are lower in NIMZs, which do not offer complete tax holidays and enforce stricter labor regulations (for example, sub-contracting of labor is not allowed, and lay-off and retrenchment policies are not as flexible as in SEZs). On the administrative side, NIMZs are established as Public-Private Partnerships (PPPs) through Special Purpose Vehicles (SPV), with the State government acquiring the land, the Central government funding the infrastructure, and a 35 Since 2005, more than 300 SEZs began operation across the nation, 80 percent of which are located in Andhra Pradesh, Gujarat, Karnataka, Kerala, Maharashtra, Tamil Nadu, and Uttar Pradesh. Since developers of SEZs can be both public and private entities, 70 percent of existing zones are either private or joint sector initiatives (Hyun and Ravi, 2017). 36 Up-to-date information on Central and State-level facilities and incentives can be easily retrieved on the dedicated SEZ web page of the Indian Ministry of Commerce and Industry, Department of Commerce: http://sezindia.nic.in/cms/facilities-and-incentives.php. 19 private player developing the zone. Contrary to SEZs, finally, State governments have in general a leading role in the development of zones, as well as in setting up the SPVs that manage them.37 5.3 Implementation and governance Despite the non-negligible variation in State regulations on both labor and land documented in Sections 5.1 and 5.2, many commentators have argued that implementation and enforcement of policy tends to be far more uneven – so that governance might ultimately be what really matters in determining divergence in state outcomes. As regards labor for example, our interviews with firms suggested that enforcement varies in many ways. In terms of regulation avoidance, Mitchell et al. (2014) and the OECD (2007) cite the example of “voluntary” retirement schemes (VRS), which sidestep the rules on retrenchment. More simply, the ILO (2008) also reports that businesses often have the capacity to arbitrage between states with varying degrees of enforcement, and just relocate to areas where it is more fluid. These observations are in line with recent empirical evidence from Hsieh and Olken (2014), who find no discontinuities in the distribution of firm size at the IDA thresholds of 50 and 100 workers. On the other hand, instead, Amirapu and Gechter (2017) argue that the distortionary effect of labor regulations depends critically on the quality of governance through the extent and type of corruption present in regulatory enforcement: firms do face a disincentive to expand beyond the 10-worker threshold of the Factories Act, but this is due more to corrupt practices by inspectors once a factory is registered than to the cost of regulatory compliance.38 Warnecke and De Ruyter (2012) also report that public officials have considerable discretion in implementing government regulations, and that corruption is widespread in courts too. There are strong elements of discretion in contract labor law as well, for example in the definition of ‘core’ activities and other activities for which use of contract labor may be prohibited in certain states, and the precise responsibilities of the labor contractor and the employer (the firm) in ensuring the contract is enforced. A qualitative survey suggests that contract labor law is poorly enforced, with workers complaining about collusive arrangements between contractors, firms and inspectors (Rajeev, 2009). Governance seems to matter equally crucially for land-related issues (Chakravorty, 2012). Despite the DILRMP reforms, for example, a report by the World Bank (2015) on six Indian States 39 highlights that rural land records are generally not updated regularly (in Bihar, for instance, most maps seem not to have been updated since 1922) and lack spatial reference, while urban land records are lacking for many properties. Even when the legal and institutional framework is relatively robust, the World Bank assessment points to significant challenges in terms of implementation and enforcement: the frequent sharing of responsibility for land management by a range of different departments (up to four in certain states) and agencies causes horizontal overlap and lack of coordination in the first place, and registration of land transactions seems to be often plagued by errors and fraud. Duranton et al. (2015a), for example, note how the process of evaluation for land that is publicly purchased offers wide opportunities for corruption, as valuation is done by the public agent acquiring the property without specific guidelines. Morris and Pandey (2007) report widespread poor protection of land property rights, as well as a high degree of discretion in granting the non-agricultural clearance for the conversion of use of farmland. 37 Detailed information and comments on NIMZ guidelines and benefits can be respectively found in DIPP (2011) and ASA (2015). 38 Amirapu and Gechter (2017) even cite evidence that inspectors, while threatening to overreport violations, tend to demand bribes proportional to the number of workers employed in an establishment. 39 The States surveyed were Andhra Pradesh, Bihar, Jharkhand, Karnataka, Odisha, and West Bengal. 20 This framework was confirmed during our fieldwork, when virtually all the interviewed firms (both in Uttar Pradesh and Telangana) mentioned implementation and governance issues among the main challenges they face in their regular operations. Many firms reported being subject to numerous inspections throughout the year, with inspectors having significant discretionary powers. 40 On inspections, an independent exercise has recently been led by the Department of Industrial Policy and Promotion (DIPP) of the Ministry of Commerce and Industry of the Government of India. Every year since 2014, the Department has outlined in a Business Reform Action Plan (BRAP) the specific actions required in all States in the areas of procedural and inspection reform, ranking States on the extent of reform implementation. A comprehensive implementation of the program would allow enterprises to maintain unified rather than multiple registers and be subject to joint regulatory compliance inspections for all laws, instead of separate inspections for each law. Moreover, the same official would not be allowed to visit any firm in two consecutive inspections. Finally, as regards environmental clearances, States are supposed to identify establishments that need to be inspected based on a computerized risk assessment: establishments falling under the low- and medium-risk categories would be allowed to submit respectively a self- and third- party certification, instead of being subject to inspections. Although these policy changes are potentially transformative, interviewed firms consistently expressed the view that the impact of the reforms is not yet being fully felt on the ground. Many firms were not fully aware of the changes. Implementation issues, too, seem to be watering down the spirit of the reform programs: for instance, very few firms are currently categorized as low-risk, so that inspections are still prevalent. Similarly, many interviewees claimed that inspectors often do not follow the reformed procedures in practice, for example visiting the same firm more than once or setting up multiple inspections rather than joint ones. Even beyond inspection costs, however, red tape, inconsistencies, and delays in the effective implementation of regulations and policies emerged during our interviews as a remarkable hurdle to firms –especially so for smaller enterprises. For example, while large firms mentioned that they tended to hire specific “consultants” to take care of paperwork and compliance with red tape, this would not be a realistic possibility for smaller enterprises. Similarly, delayed payments by government agencies, both for public procurement contracts and for the award of grants and subsidies, appeared to be a pervasive issue significantly disrupting firms’ operations. The government, both at the State and Central level, has consequently been quite active in trying to improve governance and in seeking ways to provide for a more effective service delivery to Indian citizens. One example in this direction has been a concerted effort to move towards the digitization of services and widespread use of e-governance, with many States developing their own e-governance projects to provide specific electronic services to citizens since the late 1990s.41 More recently, the National e-Governance Plan (NeGP) was endorsed by the Central government in 2006 as an ambitious initiative to develop e- government services, networking infrastructure, State data centers, and village-level centers for the delivery of core services to citizens (Bussell, 2012). The provision of NeGP services can be under the purview of 40 An MSME typically receives inspections by the Labor Department (Provident Fund, Employee State Insurance, Minimum Wage), Factories Act, Fire Department, Pollution Control Board, National Ground Tribunal, Electricity Department, Income Tax, GST. 41 These State programs targeted services such as ration cards, income certificates, land records, building licenses, and income tax payments. Examples include eSeva (Andhra Pradesh), Civic Centres (Gujarat), Sugam (Himachal Pradesh), Nemmadi (Karnataka), Friends (Kerala), e-Mitra (Rajasthan), RASI (Tamil Nadu), e-Suvidha (Uttar Pradesh). Bussell (2012) provides a detailed description of many of these programs. 21 either the Central (as regards, for example, income tax, pensions, postal services) or the State government (for instance for education, health, and crucially land records).42 Among the integrated NeGP projects, the Single Window Business Registration Act 2013 provided for the establishment of a single-window registry as a one-stop shop for the provision of online government-to- business (G2B) services to investors and business communities in the country. The resulting online platform (eBiz, launched in 2015) can now be used by anyone planning to start operations or having an existing business in India, and provides online submission and processing of composite forms and one-time payments.43 In addition, States continue to have several other e-governance programs, as well as dedicated single-window portals.44 Firms interviewed during our fieldwork almost unanimously expressed satisfaction in how these initiatives are in general reducing the complexity in obtaining the information and services required to start and run a business. In Uttar Pradesh, interviewees also appreciated how the introduction of electronic platforms is starting to bear some fruit in terms of reduced corruption. Nevertheless, both in Telangana and Uttar Pradesh enterprises noted how in practice many applications for various licenses and clearances still require the relevant documentation to be handed in manually into public offices, despite the existence of State single- window portals. In Telangana, the same applies to the single window for land acquisition under the land record modernization program. Once more, this seems to weigh differently on bigger and smaller firms,as the latter do not have a large enough workforce to assign employees to these tasks without significantly compromising operations. 6. Does governance matter? In line with our observations from Section 5, where we documented how institutional quality and the practical implementation of laws and regulations seem to be significant determinants of firm performance, we now turn our attention to governance. We assign states to two groups based on their level of governance (see Table 6.1), and we are interested in assessing whether the pattern of factor misallocation over time differs between the two sub-samples. As discussed in Section 3.3 on data sources above, we distinguish between “high-” and “low-governance” states based on the governance rankings provided by Transparency International India and the Centre for Media Studies (2005). In Tables 6.2 and 6.3, we confirm that the measure correlates well with firm perceptions on both business obstacles and corruption as captured by the latest wave of the Enterprise Survey of the World Bank.45 When we group States according to our “quality of governance” classification, we observe in Table 6.2 that firms in States with lower governance are far more likely to report major or severe obstacles to operations along 42 Details can be found on the web portal of the Ministry of Electronics and Information Technology of the Government of India, cf. http://meity.gov.in/content/mission-mode-projects. 43 eBiz has been developed as a PPP between Infosys Technologies Limited (Infosys) and the Department of Industrial Policy and Promotion (DIPP) of the Ministry of Commerce and Industry of the Government of India. Businesses can apply for 20 central government services (plus a number of state government services in Andhra Pradesh, Odisha and Delhi), to obtain licenses, approvals, clearances, no objection certificates, permits, and for filing of returns. The services available on the portal include those from Ministry of Corporate Affairs (MCA), Reserve Bank of India (RBI), Department of Industrial Policy and Promotion (DIPP), Central Board of Direct Taxes (CBDT), Directorate General of Foreign Trade (DGFT), and Employees’ Provident Fund Organisation (EPFO). 44 An updated list is kept on the web portal of the national level initiative India Development Gateway (InDG), see http://vikaspedia.in/e-governance/national-e-governance-plan/copy_of_e-governance-in-state-and-services. 45 Although a wave of the Enterprise Survey for the year 2005 exists, we focus on the latest wave (year 2014) due to its improved sample frame, which is representative at the State level. This was not the case for 2005 edition of the survey. 22 all the dimensions considered. The difference between the two groups is particularly striking for access to land, for which there is an almost six-fold increase in the likelihood of encountering difficulties when moving from column 1 to Column 2, and for business licensing, where the percentage of firms lamenting problems grows by a factor of around 3.5. Similarly, in Table 6.3, firms in States with weaker governance tend to report a higher incidence of harassment and less-than-impartial behaviors on the part of public officials. It is interesting to notice, for example, that while the likelihood to be visited by a tax inspector, as well as the number of visits per year, does not differ substantially between the two groups of States (if anything, a firm is more likely to receive a visit in a high-governance State), the likelihood that an informal gift to the inspector is expected during one such visit more than triples in size in the low governance group, where this seems to be the case for a quarter of the respondents. In Figure 6.1, we further show that, as governance quality deteriorates, various indicators of economic efficiency and investment climate produced by the Indian National Council of Applied Economic Research (NCAER, 2017) worsen too.46 For example, land issues become more problematic in general, in particular with less land available for industrial purposes as a share of total surface area of the state, lower digitization of land registration processes, and a higher share of entrepreneurial projects that are stalled because of land- related problems vis-à-vis other reasons. Labor issues display a similar pattern, and, interestingly, the share of contract workers in the total labor force is also higher in worse-governance States. As regards other areas of interest, financial depth (measured as the ratio of outstanding commercial bank credit and gross State GDP) decreases with institutional quality, as does (albeit mildly) the completion rate of legal cases by courts; finally, higher shares of entrepreneurial projects are delayed by environmental clearances. 6.1 The dynamics of governance: Is there convergence across states? Given our focus on time trends, we would be ideally interested in providing at least some evidence on the evolution of governance patterns across Indian States over time. Since we do not have data on the governance score by Transparency International India and CMS for years other than 2005, we try to do so by exploring a range of additional sources. First, we use an indicator of State government effectiveness in the delivery of core public services as developed by Mundle et al. (2016), who focus on infrastructure, social services, fiscal performance, justice, law and order, and quality of the legislature. Then, we turn to another work by the Centre for Media Studies (2017) and borrow their statistics on the proportion of households who were forced to pay bribes in order to obtain public services. Finally, we consult the yearly reports by the Indian National Crime Records Bureau, and retrieve the number of corruption records to be investigated by the Anti-Corruption and Vigilance Departments of each State (both as a stock and as the number of new cases reported each year). The exercise delivers two stylized facts: (i) governance is sticky; and (ii) there nonetheless seems to be convergence in governance. In other words, it seems that states that did not have robust governance capabilities might be effectively catching up with the rest of the country, despite still displaying worse outcomes overall. 46The NCAER study assesses the factors creating investment opportunities and driving investment decisions in Indian states, including the availability of factors of production, the efficiency in the use of these productive factors, States’ growth pr ospects, and firms’ perceptions of investment opportunities as created by the socio-political and economic climate. NCAER started producing their State Investment Potential reports only in 2016, and their first edition did not feature information on land issues. For this reason, we only focus on the 2017 wave of the study. 23 These suggestive results are illustrated in Figure 6.2, where we plot our various alternative measures of governance for Indian States at two different points in time, to compare earlier values to more recent ones. Stickiness would be implied by a positive correlation between the two set of values, that is, by the baseline values on the x-axis being good predictors of the more recent values on the y-dimension. Convergence would in turn result in the linear relationship between the past and present values cutting the 45-degree line (in grey in the figure) from above: a lower value today is associated with a larger change tomorrow. As it is apparent from Figure 6.2, this seems to be the case for all the measures considered. Figure 6.1 Does governance matter? Data source: Transparency International India and the Centre for Media Studies (2005) and NCAER (2017). 24 Figure 6.2 The dynamics of governance Data sources: Mundle et al. (2016); Centre for Media Studies (2017); National Crime Records Bureau (2015) 6.2 Trends in adjustment costs by governance Keeping in mind the suggestive evidence presented above on the correlation with governance with various measures of institutional quality and economic efficiency, as well as on the quality of its dynamic evolution, we now look at how governance shapes the trends in adjustment costs for the four inputs (permanent labor, contract labor, land, and fixed capital) for firms over time. Since our governance indicator is measured at the state level, we look at the trends separately for firms in states with high governance quality and those with low governance quality. Furthermore, we also look at these trends for large and small firms across states with differing governance qualities. In our main figures, we show results estimated for all firms in these states. To make sure that our estimated trends only capture governance quality, not other state level measures, we repeat the analysis by restricting the sample to districts along the border in states with high versus low governance quality, with similar results (see Annex Tables A6.1 and A6.2). Surprisingly, the adjustment costs of permanent labor and land (Figures 6a and 6c, respectively) have fallen significantly more in the low-governance group, compared to the high-governance one. This pattern is robust to restricting the sample to districts along the border in states with high versus low governance 25 quality (Annex Figures A6a and A6c). Moreover, this differential looks significantly larger among larger firms vis-à-vis the smaller ones (Annex Table A6.2). Note that this does not imply that the levels of adjustment costs are lower in low-governance states compared to high-governance states: similar to our analysis in Section 4, all the reported coefficients are in fact relative to the omitted year 1999. In other words, the figures only show that the adjustment costs have fallen more in low-governance states relative to their own corresponding levels in 1999 as compared to the same trends in high governance states. Nevertheless, this pattern appears at least counter-intuitive at first, as one would reasonably expect to see better rates of cost decline in better-governed states. The conundrum, however, makes more sense once the dynamic path of governance is taken into account: if low-governance states improved their governance effectiveness over time at a faster pace, it may well be that this affected the speed of adjustment of factor misallocations more than the absolute differences in governance quality per se. Figure 6: Adjustment cost trends in low relative to high governance quality states a. Permanent Labor b. Contract Labor c. Land d. Fixed Capital Data source: ASI 1999 – 2014. Note: Coefficients on interactions of year dummies with a dummy for low governance states. 26 7. Conclusion This paper has documented trends in factor market distortions in Indian states for four factors of production over a 15-year time span. We have shown an overall decline in factor misallocation for land and labor, with important heterogeneity across both state- and firm-level attributes. In particular, three policy-relevant findings can be readily derived from our exercise: (i) the speed of adjustment cost reduction and the growth of the manufacturing sector tend to be correlated; (ii) firm size matters; and (iii) governance matters too. The fact that states diverge not just in terms of manufacturing growth, but also in terms of factor market efficiency, is to be viewed with some concern. Even though our current analysis is silent about the direction of causality between the two dimensions, our evidence hints at the existence of a self-reinforcing pattern in which misallocation leads to low growth which worsens misallocation, and so on. As regards firm-level attributes, size seems to represent a significant advantage for firms when it comes to input markets. In particular, our results for India are suggestive of increasing land misallocation towards larger firms. This observation raises the question why the process of acquiring land or adjusting land use may be harder for small firms. It is possible that policies on industrial parks (for example on how land is allocated or transferred to new users), the processing of land conversion requests, or the easing of urban land ceiling laws are tilted in practice towards larger firms, possibly because of their greater bargaining power. Indeed, this is what emerged to some extent during our field consultations with private sector players. The finding that the decline in permanent labor adjustment costs too has been more pronounced for larger establishments is further suggestive of a business environment favorable to these firms. On paper, most labor laws in India are stricter for larger firms, with the most stringent provisions applying onlybeyond a certain size cut-off. As the laws did not change significantly during the study period, it could be that changes in their implementation are creating a de facto bias against SMEs. Relatedly, our findings paint a prominent role for institutional quality, as upgrading governance seems to have paid off in terms of better factor market efficiency. In our review of Indian policies, we have provided examples of programs aimed at improving institutional efficiency and streamlining administrative processes. Our preliminary results on the dynamics of governance would suggest that these reforms have been more effectively implemented by states that were initially worse-off, and that this in turn has had a magnifying effect on the secular reduction in factor misallocation observed for the entire country. For as appealing as this interpretation may look however, there remains the potentially problematic fact that states with lower absolute levels of governance seem to have had a more favorable evolution of factor adjustment costs over time. This in turn raises the intriguing question of how exactly institutional quality, policy implementation, and factor market efficiency interact in practice. Input misallocation has been recognized as one of the driving forces behind the welfare and growth differentials that persist across countries in the world. In this paper, we have provided stylized facts and suggestions on how regulation, law enforcement, and overall governance style could have affected the path of distortions in the allocation of resources across firms in a big developing country. Considering the severe consequences of factor misallocation on aggregate productivity growth, a more rigorous understanding of these dynamics would be a critical contribution to the development literature. 27 References Adhvaryu, A., Chari, A., Sharma, S. (2013). Firing costs and flexibility: evidence from firms' employment responses to shocks in India. Review of Economics and Statistics 95(3): 725-740. Aghion, P., Burgess, R., Redding, S., and Zilibotti, F. (2008). he unequal effects of liberalization: Evidence from dismantling the license Raj in India. American Economic Review, 98(4):1397–1412. Ahsan, A., Pages, C., and Roy, T. (2008). Legislation, enforcement and adjudication in Indian labor markets: Origins, consequences and the way forward. In Mazumdar, D. and Sarkar, S., editors, Globalization, Labor Markets and Inequality in India. International Development Research Centre, Ottawa. Allcott, H., Collard-Wexler, A., and O'Connell, S.D. (2016). How do electricity shortages affect industry? Evidence from India. American Economic Review, 106(3), 587-624. Amin, M. (2009). Labor regulation and employment in India's retail stores. Journal of Comparative Economics, 37(1): 47–61. Amirapu, A. and Gechter, M. (2017). Labor regulations and the cost of corruption: Evidence from the Indian firm size distribution. Boston University IED Working Paper No. 266. ASA (2015). A brief report on NMIZ in India. ASA & Associates LLP. Barro, R.J. and Sala-i-Martin, X. 1992. Regional Growth and Migration: A Japan-U.S. Comparison. NBER Working Papers No. 4038. National Bureau of Economic Research. Bertrand, M., Hsieh, C., and Tsivanidis, N. (2017). Contract Labor and Firm Growth in India. Working paper. Besley, T. and Burgess, R. (2000). Land reform, poverty reduction, and growth: evidence from India. The Quarterly Journal of Economics, 115(2):389–430. Besley, T. and Burgess, R. (2004). Can regulation hinder economic performance? Evidence from India. The Quarterly Journal of Economics, 119(1):91–134. Bhattacharjea, A. (2006). Labor market regulation and industrial performance in India: A critical review of the empirical evidence. The Indian Journal of Labor Economics, 49(2):211–232. Bhattacharjea, A. (2017). Threshold effects of Indian labor laws. In Kapila, U., editor, India’s Economy: Pre-Liberalization to GST. Academic Foundation, New Delhi. Blakeslee, D., Chaurey, R., and Malik, S. (2017). Structural transformation and spillovers from Industrial Areas. Mimeo, available at SSRN: https://ssrn.com/abstract=3049617. Bussell, J. (2012). E-governance and corruption in the States: Can technology serve the Aam Aadmi? Economic & Political Weekly, XLVII(25):211–232. Caselli, F. and Coleman, W.J. II, 2001. The U.S. Structural Transformation and Regional Convergence: A Reinterpretation. Journal of Political Economy, 109(3):584-616. Centre for Media Studies (2017). CMS - India corruption study. perception and experience with public services and snapshot view for 2005-17. Technical report, Centre for Media Studies, New Delhi. 28 Chakravorty, S. (2012). Can regulation hinder economic performance? Evidence from India. Eurasian Geography and Economics, 53(1):21–43. Chakravorty, S. (2013). A new price regime: Land markets in urban and rural India. Economic & Political Weekly, XLVIII(17):45–54. Chaurey, R. (2015). Labor regulations and contract labor use: Evidence from Indian firms. Journal of Development Economics, 114:224–232. Chaurey, R. (2017). Location-based tax incentives: Evidence from India. Journal of Public Economics, 156:101–120. De Loecker, Jan, and Frederic Warzynski. 2012. Markups and Firm-Level Export Status. American Economic Review, 102(6): 2437-71. DIPP (2011). National Manufacturing Policy - Guidelines for establishment of National Investment & Manufacturing Zones (NIMZs). Department of Industrial Policy and Promotion, India. Dougherty, S., Frisancho, V., and Krishna, K. (2014). State-level labor reforms and firm-level productivity in India. Technical Report 10(1), National Council of Applied Economic Research. Duranton, G., Ghani, E., Grover Goswami, A., and Kerr, W. (2015a). The misallocation of land and other factors of production in India. The World Bank Group, Policy Research Working Paper No. 7221. Duranton, Gilles; Ghani, Ejaz; Grover, Arti Goswami; Kerr, William Robert. (2015b). Effects of land misallocation on capital allocations in India. Policy Research working paper; no. WPS 7451. Washington, D.C.: World Bank Group. Duranton, G., Ghani, E., Grover Goswami, A., and Kerr, W. (2016). A detailed anatomy of factor misallocation in India. The World Bank Group, Policy Research Working Paper No. 7547. Ghani, E., Grover Goswami, A., and Kerr, W. (2012). Is India’s manufacturing sector moving away from cities? NBER Working Paper No. 17992. Government of India, 2017. Economic Survey of India 2016-17. Hasan, Rana, Devashish Mitra and Asha Sundaram, (2013). What explains the high capital intensity of Indian manufacturing? Indian Growth and Development Review, 6(2): 212-241. Hsieh, C. and Klenow, P.J. (2009). Misallocation and Manufacturing TFP in China and India, TheQuarterly Journal of Economics, 124(4): 1403–1448. Hsieh, Chang-Tai, and Enrico Moretti. 2019. "Housing Constraints and Spatial Misallocation." American Economic Journal: Macroeconomics, 11 (2): 1-39. Hsieh, C. and Olken, B. (2014). The missing “missing middle”. NBER Working Paper No. 19966. Hyun, Y. and Ravi, S. (2017). The effect of place-based development policies: Evidence from Indian SEZs. Job Market Paper. ILO (2008). Proceedings of the national tripartite meeting on the employment relationship in India. International Labor Organization, Geneva. 29 Indian Labor Bureau (2004). India labor year book. Technical report, Ministry of Labor and Employment, Government of India, New Delhi. Lall, S. and Chakravorty, S. (2005). The evolution of labor law in India: An overview and commentary on regulatory objectives and development. Review of Development Economics, 9(1):47–68. Martin, L.A., Nataraj, S. and Harrison, A.E. (2017). In with the Big, Out with the Small: Removing Small- Scale Reservations in India. American Economic Review, 107(2): 354-86. Mitchell, R., Mahy, P., and Gahan, P. (2014). The evolution of labor law in India: An overview and commentary on regulatory objectives and development. Asian Journal of Law and Society, 1(2):413–453. Mundle, S., Chowdhury, S., and Sikdar, S. (2016). Governance performance of Indian States 2001-02 and 2011-12. National Institute of Public Finance and Policy, New Delhi, Working Paper No. 164. National Crime Records Bureau (2000). Crime in India. Technical report, Ministry of Home Affairs, Government of India, New Delhi. National Crime Records Bureau (2015). Crime in India. Technical report, Ministry of Home Affairs, Government of India, New Delhi. NCAER (2017). The NCAER State investment potential index, second edition. Technical report, National Council of Applied Economic Research, New Delhi. Rajeev, Meenakshi, 2009. Contract Labor Act in India: A Pragmatic View. Indira Gandhi Institute of Development Research (IGIDR) Proceedings/Project Reports Series. PP-062-33. Ramaswamy, K. (2013). Size-dependent Labor Regulations and Threshold Effects: The Case of Contract- worker Intensity in Indian Manufacturing. Indira Gandhi Institute of Development Research, Mumbai Working Papers 2013–012. Indira Gandhi Institute of Development Research, Mumbai, India. Restuccia, D. and Rogerson, R. (2017). The Causes and Costs of Misallocation. NBER Working Paper no. 23422. Rotemberg, M. and White, T.K. (2019). Measuring Cross-Country Differences in Misallocation. Working Paper. Transparency International India and the Centre for Media Studies (2005). India corruption study. Technical report, Transparency International India, New Delhi. 30 TABLES AND FIGURES Tables Table 3.1 Data Sources Data sources Variables Source Years Firm expenditure on productive factors ASI 1999-2014 Firm size ASI 1999-2014 Growth rate of manufacturing sector NITI Aayog 1999, 2014 State population Indian Population Census 1991 Governance score Transparency International India and CMS (2005) 2005 Investment climate indicators NCAER 2017 Firms' perceptions on obstacles to operations Enterprise Survey 2014 Firms' perceptions on corruption Enterprise Survey 2014 State government effectiveness Mundle et al. (2016) 2001, 2011 % households paying bribes CMS (2017) 2005, 2017 Registered corruption cases National Crime Records Bureau (2015) 2000, 2015 31 Table 3.2: Summary Statistics of key ASI variables Fixed Capital - Permanent Labor Contract Labor Land Firm Age Firm Size Year Excluding Land Standard Standard Standard Standard Standard Standard Mean Deviation Mean Deviation Mean Deviation Mean Deviation Mean Deviation Mean Deviation 1999 2.104 2.411 3.300 2.583 0.844 2.351 0.938 2.315 0.0275 0.1636 0.5189 0.4997 2000 2.175 2.252 3.366 2.408 0.858 2.291 0.845 2.225 0.0290 0.1677 0.5141 0.4998 2001 2.216 2.239 3.363 2.411 0.938 2.300 0.793 2.141 0.0512 0.2205 0.4822 0.4997 2002 2.246 2.247 3.266 2.384 0.902 2.303 0.750 2.102 0.0885 0.2841 0.5227 0.4995 2003 2.297 2.247 3.236 2.349 0.928 2.268 0.799 2.068 0.1284 0.3346 0.5386 0.4985 2004 2.340 2.299 3.161 2.400 0.786 2.247 0.870 2.155 0.1660 0.3721 0.4492 0.4974 2005 2.359 2.375 3.174 2.458 0.787 2.248 0.898 2.192 0.2133 0.4096 0.4126 0.4923 2006 2.424 2.357 3.016 2.537 0.667 2.267 0.870 2.146 0.2906 0.4540 0.4073 0.4913 2007 2.546 2.321 2.931 2.611 0.663 2.286 0.846 2.107 0.3561 0.4789 0.4043 0.4908 2008 2.622 2.305 3.096 2.509 0.569 2.321 0.854 2.077 0.3933 0.4885 0.4444 0.4969 2009 2.618 2.311 3.257 2.347 0.557 2.322 0.902 2.098 0.3777 0.4848 0.4484 0.4973 2010 2.558 2.310 3.034 2.444 0.476 2.314 0.820 2.131 0.4107 0.4920 0.4258 0.4945 2011 2.619 2.269 3.038 2.413 0.502 2.291 0.909 2.033 0.4469 0.4972 0.4470 0.4972 2012 2.637 2.261 3.024 2.378 0.453 2.312 0.903 2.028 0.4771 0.4995 0.4558 0.4981 2013 2.647 2.162 3.160 2.325 0.392 2.425 0.916 2.145 0.4917 0.4999 0.4320 0.4954 2014 2.630 2.093 3.170 2.240 0.389 2.435 0.924 2.148 0.5221 0.4995 0.4301 0.4951 Total 2.469 2.280 3.132 2.420 0.639 2.323 0.868 2.125 0.3165 0.4651 0.4502 0.4975 Av. # Obs. 31694 10361 24491 33454 36699 39286 Note: The dependent variables are input-share ratio in logs. The dependent variables are binary variables: For firm age, the value takes 1 when a firm’s initial year of operation is after 2000. For firm size, the value takes 1 when a firm employs more than 50 employees. 32 Table 5.1: Factor Laws Table 6.1: Governance classification Data source: Transparency International India and the Centre for Media Studies (2005). 33 Table 6.2: Major or severe obstacles to operations (percentages) Data source: Enterprise Survey, year 2014. Table 6.3: Corruption perceptions, 2014 Data source: Enterprise Survey, year 2014. Number of non-missing observations in parentheses. 34 Annex Tables Table A3.1: Growth rate of manufacturing value added, Indian States, 1999-2014 State Growth rate Fast-growing Andhra Pradesh 0.655636 0 Assam 0.349316 0 Bihar 0.488523 0 Chattisgarh 0.463236 0 Delhi 0.553769 0 Jharkhand -0.28426 0 Kerala 0.641139 0 Madhya Pradesh 0.486132 0 Odisha 0.30178 0 Uttar Pradesh 0.689997 0 West Bengal 1.040271 0 Goa 1.168414 1 Gujarat 1.800211 1 Haryana 1.329213 1 Himachal Pradesh 1.982666 1 Karnataka 1.224 1 Maharashtra 1.374972 1 Punjab 1.438939 1 Rajasthan 1.564837 1 Tamil Nadu 1.452238 1 Uttarakhand 12.16795 1 35 Table A4.1: Adjustment Cost Trends – All India (1) (2) (3) (4) VARIA Permanent Contract Land Fixed Capital BLES Labor Labor (Excluding Land) 2000 -0.0649*** -0.110*** 0.0292* -0.0550*** (0.0136) (0.0294) (0.0161) (0.0157) 2001 -0.0997*** -0.103*** 0.0689*** -0.0710*** (0.0129) (0.0275) (0.0153) (0.0149) 2002 -0.156*** -0.171*** 0.00746 -0.143*** (0.0127) (0.0270) (0.0151) (0.0147) 2003 -0.154*** -0.224*** 0.00918 -0.125*** (0.0127) (0.0268) (0.0150) (0.0147) 2004 -0.121*** -0.207*** -0.0225 -0.0932*** (0.0123) (0.0262) (0.0146) (0.0142) 2005 -0.0764*** -0.188*** -0.0142 -0.0333** (0.0126) (0.0266) (0.0150) (0.0146) 2006 -0.0694*** -0.238*** -0.0597*** -0.0548*** (0.0125) (0.0264) (0.0148) (0.0145) 2007 -0.0283** -0.238*** -0.0529*** -0.0270* (0.0126) (0.0262) (0.0149) (0.0145) 2008 -0.0311** -0.249*** -0.101*** -0.00253 (0.0127) (0.0263) (0.0151) (0.0147) 2009 -0.0382*** -0.257*** -0.145*** 0.0296** (0.0128) (0.0264) (0.0151) (0.0148) 2010 -0.104*** -0.340*** -0.179*** -0.00616 (0.0127) (0.0262) (0.0149) (0.0146) 2011 -0.0762*** -0.354*** -0.124*** 0.101*** (0.0127) (0.0261) (0.0149) (0.0146) 2012 -0.104*** -0.383*** -0.128*** 0.128*** (0.0127) (0.0261) (0.0149) (0.0146) 2013 -0.145*** -0.396*** -0.134*** 0.143*** (0.0126) (0.0262) (0.0148) (0.0145) 2014 -0.199*** -0.440*** -0.0900*** 0.161*** (0.0126) (0.0263) (0.0148) (0.0145) Obs. 507,102 165,778 391,848 535,262 R- 0.867 0.899 0.845 0.788 squared Firm FE YES YES YES YES 36 Table A4.2: Adjustment Cost Trends – Fast vs. Slow Growing State Groups (1) (2) (3) (4) VARIABLES Permanent Contract Land Fixed Labor Labor Capital (Excluding Land) 2000 -0.0482** -0.0304 0.0650*** -0.0912*** (0.0209) (0.0480) (0.0252) (0.0243) 2001 -0.0671*** -0.0672 0.111*** -0.0996*** (0.0199) (0.0452) (0.0238) (0.0230) 2002 -0.125*** -0.129*** 0.0364 -0.155*** (0.0195) (0.0442) (0.0234) (0.0226) 2003 -0.109*** -0.158*** 0.0664*** -0.148*** (0.0194) (0.0438) (0.0234) (0.0225) 2004 -0.0723*** -0.156*** 0.0242 -0.118*** (0.0188) (0.0429) (0.0226) (0.0218) 2005 -0.0296 -0.105** 0.0429* -0.0568** (0.0195) (0.0436) (0.0234) (0.0226) 2006 -0.0193 -0.160*** -0.0343 -0.0987*** (0.0193) (0.0432) (0.0232) (0.0224) 2007 0.0300 -0.164*** -0.0459** -0.0762*** (0.0193) (0.0429) (0.0232) (0.0224) 2008 0.0538*** -0.137*** -0.0823*** -0.0152 (0.0196) (0.0431) (0.0234) (0.0227) 2009 0.0603*** -0.123*** -0.118*** 0.0511** (0.0198) (0.0433) (0.0237) (0.0230) 2010 0.0103 -0.172*** -0.173*** 0.0273 (0.0197) (0.0430) (0.0234) (0.0227) 2011 0.0198 -0.192*** -0.130*** 0.0818*** (0.0197) (0.0429) (0.0234) (0.0227) 2012 -0.00425 -0.221*** -0.143*** 0.110*** (0.0198) (0.0429) (0.0235) (0.0228) 2013 -0.0510*** -0.219*** -0.144*** 0.111*** (0.0195) (0.0432) (0.0232) (0.0225) 37 2014 -0.100*** -0.206*** -0.0714*** 0.156*** (0.0196) (0.0434) (0.0232) (0.0226) Fast Growing 2000 -0.0322 -0.127** -0.0610* 0.0620* (0.0275) (0.0607) (0.0328) (0.0319) Fast Growing 2001 -0.0592** -0.0571 -0.0723** 0.0495 (0.0261) (0.0569) (0.0310) (0.0303) Fast Growing 2002 -0.0549** -0.0668 -0.0497 0.0213 (0.0258) (0.0557) (0.0306) (0.0298) Fast Growing 2003 -0.0803*** -0.105* -0.0973*** 0.0393 (0.0257) (0.0553) (0.0305) (0.0297) Fast Growing 2004 -0.0853*** -0.0802 -0.0797*** 0.0425 (0.0248) (0.0541) (0.0296) (0.0287) Fast Growing 2005 -0.0848*** -0.133** -0.0951*** 0.0406 (0.0256) (0.0550) (0.0304) (0.0296) Fast Growing 2006 -0.0906*** -0.124** -0.0431 0.0740** (0.0254) (0.0545) (0.0302) (0.0293) Fast Growing 2007 -0.104*** -0.118** -0.0130 0.0835*** (0.0255) (0.0542) (0.0302) (0.0294) Fast Growing 2008 -0.148*** -0.180*** -0.0320 0.0228 (0.0258) (0.0544) (0.0306) (0.0298) Fast Growing 2009 -0.170*** -0.217*** -0.0460 -0.0323 (0.0260) (0.0546) (0.0307) (0.0300) Fast Growing 2010 -0.196*** -0.270*** -0.0125 -0.0511* (0.0257) (0.0543) (0.0304) (0.0297) Fast Growing 2011 -0.166*** -0.260*** 0.00790 0.0333 (0.0257) (0.0541) (0.0304) (0.0297) Fast Growing 2012 -0.171*** -0.260*** 0.0212 0.0309 (0.0258) (0.0541) (0.0304) (0.0298) Fast Growing 2013 -0.163*** -0.279*** 0.0142 0.0536* (0.0255) (0.0543) (0.0301) (0.0294) Fast Growing 2014 -0.171*** -0.365*** -0.0318 0.0104 (0.0256) (0.0545) (0.0302) (0.0295) Observations 507,102 165,778 391,848 535,262 R-squared 0.867 0.899 0.845 0.788 Firm FE YES YES YES YES 38 Table A4.3: Adjustment Cost Trends – Large vs. Small Firms (1) (2) (3) (4) VARIABLES Permanent Contract Land Fixed Labor Labor Capital (Excluding Land) 2000 -0.0625*** -0.0482 0.109*** 0.00680 (0.0241) (0.0913) (0.0286) (0.0276) 2001 -0.0795*** -0.140* 0.226*** 0.0138 (0.0229) (0.0823) (0.0270) (0.0262) 2002 -0.131*** -0.223*** 0.223*** -0.0263 (0.0230) (0.0818) (0.0271) (0.0263) 2003 -0.112*** -0.324*** 0.275*** 0.00970 (0.0231) (0.0814) (0.0272) (0.0264) 2004 -0.103*** -0.297*** 0.277*** 0.0378 (0.0219) (0.0785) (0.0257) (0.0250) 2005 -0.0694*** -0.248*** 0.301*** 0.110*** (0.0223) (0.0791) (0.0262) (0.0254) 2006 -0.0579*** -0.344*** 0.287*** 0.110*** (0.0222) (0.0789) (0.0261) (0.0254) 2007 -0.00726 -0.346*** 0.328*** 0.122*** (0.0226) (0.0787) (0.0265) (0.0257) 2008 0.00359 -0.294*** 0.292*** 0.156*** (0.0231) (0.0793) (0.0271) (0.0263) 2009 0.0162 -0.325*** 0.274*** 0.191*** (0.0230) (0.0799) (0.0269) (0.0263) 2010 -0.0825*** -0.491*** 0.245*** 0.130*** (0.0227) (0.0792) (0.0265) (0.0259) 2011 -0.0790*** -0.565*** 0.320*** 0.165*** (0.0229) (0.0790) (0.0267) (0.0260) 2012 -0.110*** -0.640*** 0.322*** 0.162*** (0.0230) (0.0793) (0.0268) (0.0262) 2013 -0.143*** -0.647*** 0.342*** 0.135*** 39 (0.0227) (0.0798) (0.0263) (0.0258) 2014 -0.207*** -0.690*** 0.356*** 0.124*** (0.0228) (0.0800) (0.0264) (0.0259) Large Firm 2000 -0.00147 -0.0645 -0.0717** -0.0508 (0.0307) (0.0972) (0.0353) (0.0351) Large Firm 2001 -0.0144 0.0633 -0.151*** -0.0554* (0.0292) (0.0879) (0.0334) (0.0334) Large Firm 2002 -0.0459 0.0806 -0.232*** -0.106*** (0.0290) (0.0871) (0.0332) (0.0331) Large Firm 2003 -0.0615** 0.137 -0.291*** -0.112*** (0.0289) (0.0866) (0.0331) (0.0330) Large Firm 2004 -0.0462* 0.126 -0.332*** -0.116*** (0.0278) (0.0838) (0.0317) (0.0317) Large Firm 2005 -0.0262 0.104 -0.350*** -0.124*** (0.0284) (0.0844) (0.0324) (0.0324) Large Firm 2006 -0.0252 0.162* -0.375*** -0.129*** (0.0282) (0.0840) (0.0321) (0.0321) Large Firm 2007 -0.0600** 0.178** -0.419*** -0.123*** (0.0283) (0.0837) (0.0323) (0.0322) Large Firm 2008 -0.0905*** 0.108 -0.455*** -0.165*** (0.0287) (0.0841) (0.0327) (0.0326) Large Firm 2009 -0.120*** 0.113 -0.525*** -0.199*** (0.0287) (0.0847) (0.0326) (0.0327) Large Firm 2010 -0.0744*** 0.182** -0.550*** -0.157*** (0.0284) (0.0840) (0.0323) (0.0324) Large Firm 2011 -0.0366 0.255*** -0.588*** -0.0842*** (0.0284) (0.0838) (0.0323) (0.0323) Large Firm 2012 -0.0600** 0.280*** -0.613*** -0.0897*** (0.0285) (0.0840) (0.0323) (0.0324) Large Firm 2013 -0.0482* 0.240*** -0.647*** -0.0639** (0.0282) (0.0844) (0.0320) (0.0321) Large Firm 2014 -0.0512* 0.249*** -0.646*** -0.0402 (0.0283) (0.0846) (0.0320) (0.0322) Observations 466,141 155,258 358,267 493,267 R-squared 0.868 0.900 0.846 0.789 40 Firm FE YES YES YES YES Table A5.1 Firm closures, 1999-2014 Table A5.2 Firm entry, 1999-2014 Difference Difference Year All India Slow growing High growing Year All India Slow growing High growing slow-high* slow-high* 1999 0.062 0.068 0.057 0.011 1999 0.008 0.008 0.007 0.001 2000 0.250 0.272 0.235 0.037 2000 0.006 0.007 0.006 0.001 2001 0.230 0.234 0.228 0.006 2001 0.007 0.009 0.005 0.004 2002 0.186 0.181 0.189 -0.008 2002 0.006 0.008 0.005 0.003 2003 0.163 0.160 0.165 -0.005 2003 0.005 0.006 0.004 0.002 2004 0.178 0.174 0.181 -0.007 2004 0.007 0.008 0.006 0.002 2005 0.172 0.164 0.177 -0.013 2005 0.009 0.011 0.007 0.004 2006 0.206 0.192 0.214 -0.022 2006 0.012 0.016 0.010 0.006 2007 0.194 0.205 0.187 0.018 2007 0.014 0.014 0.013 0.001 2008 0.225 0.220 0.229 -0.009 2008 0.010 0.012 0.009 0.003 2009 0.186 0.202 0.177 0.025 2009 0.008 0.010 0.008 0.002 2010 0.207 0.210 0.205 0.005 2010 0.008 0.006 0.008 -0.002 2011 0.140 0.101 0.161 -0.060 2011 0.002 0.003 0.001 0.002 2012 0.140 0.106 0.159 -0.053 2012 0.004 0.005 0.003 0.002 2013 0.176 0.191 0.165 0.026 2013 0.004 0.005 0.003 0.002 2014 0.133 0.145 0.125 0.020 2014 0.003 0.005 0.002 0.003 * in bold when significant at 5 percent level * in bold when significant at 5 percent level Data source: ASI 1999 – 2014. Table A6.1: Adjustment Cost Trends – High Governance Quality vs. Low Governance Quality (1) (2) (3) (4) (5) (6) (7) (8) All Sample Only Bordering Districts Fixed Fixed Permanent Contract Capital Permanent Contract Capital VARIABLES Land Land Labor Labor (Excluding Labor Labor (Excluding Land) Land) 2000 -0.0308 -0.157*** 0.0913*** -0.0665*** 0.0738 -0.136 0.150*** -0.0694 (0.0199) (0.0400) (0.0235) (0.0231) (0.0452) (0.0915) (0.0466) (0.0527) 2001 -0.0635*** -0.144*** 0.0730*** -0.0850*** 0.0145 -0.154* 0.182*** -0.0578 (0.0190) (0.0375) (0.0223) (0.0220) (0.0427) (0.0849) (0.0439) (0.0498) 2002 -0.125*** -0.214*** 0.0182 -0.186*** -0.0738* -0.277*** 0.112*** -0.179*** (0.0188) (0.0367) (0.0221) (0.0217) (0.0416) (0.0826) (0.0428) (0.0485) 2003 -0.100*** -0.268*** 0.0273 -0.157*** -0.0279 -0.243*** 0.128*** -0.156*** (0.0187) (0.0364) (0.0220) (0.0217) (0.0410) (0.0815) (0.0422) (0.0478) 2004 -0.0673*** -0.243*** -0.0191 -0.123*** 0.0398 -0.168** 0.0565 -0.0598 (0.0181) (0.0357) (0.0213) (0.0209) (0.0396) (0.0805) (0.0409) (0.0462) 2005 -0.0161 -0.224*** -0.0114 -0.0674*** 0.0451 -0.257*** 0.0648 -0.0824* (0.0185) (0.0361) (0.0218) (0.0214) (0.0406) (0.0810) (0.0418) (0.0472) 41 2006 0.00366 -0.284*** -0.0514** -0.0742*** 0.0519 -0.336*** 0.0322 -0.0715 (0.0183) (0.0358) (0.0216) (0.0212) (0.0402) (0.0800) (0.0413) (0.0467) 2007 0.0645*** -0.278*** -0.0149 -0.0247 0.133*** -0.327*** 0.0674 -0.0186 (0.0184) (0.0357) (0.0217) (0.0212) (0.0403) (0.0792) (0.0415) (0.0467) 2008 0.0892*** -0.261*** -0.0431** 0.0202 0.189*** -0.231*** 0.045 0.0927** (0.0186) (0.0358) (0.0219) (0.0215) (0.0407) (0.0793) (0.0418) (0.0472) 2009 0.0826*** -0.260*** -0.0958*** 0.0631*** 0.185*** -0.259*** 0.0179 0.191*** (0.0187) (0.0360) (0.0220) (0.0216) (0.0409) (0.0799) (0.0419) (0.0476) 2010 0.0117 -0.357*** -0.105*** 0.0269 0.0925** -0.365*** 0.00419 0.164*** (0.0185) (0.0358) (0.0217) (0.0214) (0.0410) (0.0801) (0.0421) (0.0477) 2011 0.0173 -0.381*** -0.0760*** 0.0879*** 0.0691* -0.446*** 0.0559 0.175*** (0.0185) (0.0357) (0.0217) (0.0213) (0.0412) (0.0799) (0.0421) (0.0477) 2012 0.00015 -0.442*** -0.0527** 0.126*** 0.108*** -0.468*** 0.0613 0.206*** (0.0185) (0.0357) (0.0217) (0.0214) (0.0416) (0.0805) (0.0425) (0.0482) 2013 -0.0582*** -0.465*** -0.0235 0.120*** 0.0675 -0.512*** 0.0990** 0.237*** (0.0185) (0.0360) (0.0217) (0.0214) (0.0420) (0.0829) (0.0430) (0.0488) 2014 -0.135*** -0.500*** -0.0122 0.134*** -0.0067 -0.458*** 0.129*** 0.313*** (0.0186) (0.0361) (0.0217) (0.0214) (0.0422) (0.0829) (0.0430) (0.0489) Low Governance -0.0651** 0.097 -0.110*** 0.021 -0.173** 0.0841 -0.121* 0.0664 Quality 2000 (0.0275) (0.0600) (0.0327) (0.0318) (0.0674) (0.1570) (0.0699) (0.0784) Low Governance -0.0640** 0.0891 0.00062 0.0279 -0.152** 0.0557 -0.033 0.0668 Quality 2001 (0.0262) (0.0562) (0.0310) (0.0303) (0.0633) (0.1440) (0.0657) (0.0736) Low Governance -0.0562** 0.0759 -0.0183 0.0787*** -0.155** 0.0351 -0.0773 0.0693 Quality 2002 (0.0259) (0.0551) (0.0306) (0.0299) (0.0614) (0.1390) (0.0639) (0.0715) Low Governance -0.101*** 0.0753 -0.0305 0.0554* -0.188*** -0.0288 -0.111* 0.086 Quality 2003 (0.0258) (0.0548) (0.0305) (0.0298) (0.0606) (0.1380) (0.0629) (0.0704) Low Governance -0.0979*** 0.0559 0.00218 0.0499* -0.204*** -0.138 -0.0036 0.0349 Quality 2004 (0.0249) (0.0535) (0.0295) (0.0288) (0.0588) (0.1340) (0.0612) (0.0683) Low Governance -0.113*** 0.0599 0.00357 0.0603** -0.112* 0.0492 -0.0342 0.151** Quality 2005 42 (0.0256) (0.0544) (0.0303) (0.0295) (0.0601) (0.1360) (0.0625) (0.0696) Low Governance -0.137*** 0.0908* -0.0083 0.0338 -0.123** -0.0042 0.0179 0.11 Quality 2006 (0.0254) (0.0539) (0.0301) (0.0293) (0.0596) (0.1350) (0.0620) (0.0690) Low Governance -0.174*** 0.074 -0.0609** -0.0021 -0.167*** -0.0288 -0.0979 0.0474 Quality 2007 (0.0255) (0.0537) (0.0302) (0.0294) (0.0603) (0.1350) (0.0628) (0.0697) Low Governance -0.221*** 0.01 -0.0999*** -0.0425 -0.260*** -0.111 -0.130** -0.0437 Quality 2008 (0.0259) (0.0539) (0.0306) (0.0298) (0.0609) (0.1350) (0.0634) (0.0703) Low Governance -0.230*** -0.0029 -0.0970*** -0.0747** -0.331*** -0.0438 -0.128** -0.174** Quality 2009 (0.0259) (0.0540) (0.0305) (0.0299) (0.0606) (0.1360) (0.0628) (0.0702) Low Governance -0.217*** 0.0276 -0.152*** -0.0502* -0.253*** 0.0145 -0.177*** -0.121* Quality 2010 (0.0257) (0.0537) (0.0303) (0.0296) (0.0612) (0.1350) (0.0633) (0.0709) Low Governance -0.174*** 0.0627 -0.104*** 0.00597 -0.156** 0.103 -0.0969 0.0349 Quality 2011 (0.0257) (0.0535) (0.0302) (0.0296) (0.0615) (0.1360) (0.0635) (0.0711) Low Governance -0.197*** 0.133** -0.168*** -0.0233 -0.301*** 0.15 -0.228*** -0.0972 Quality 2012 (0.0258) (0.0534) (0.0303) (0.0297) (0.0619) (0.1360) (0.0638) (0.0716) Low Governance -0.162*** 0.148*** -0.228*** 0.0163 -0.253*** 0.198 -0.251*** -0.0235 Quality 2013 (0.0256) (0.0535) (0.0299) (0.0294) (0.0621) (0.1370) (0.0640) (0.0717) Low Governance -0.119*** 0.134** -0.167*** 0.0209 -0.186*** 0.0458 -0.224*** -0.0841 Quality 2014 (0.0257) (0.0536) (0.0300) (0.0295) (0.0626) (0.1370) (0.0642) (0.0722) Observations 493,570 159,762 380,535 521,012 103,264 33,024 86,816 108,556 R-squared 0.867 0.9 0.844 0.788 0.867 0.897 0.835 0.754 Firm FE YES YES YES YES YES YES YES YES 43 Table A6.2: Adjustment Cost Trends – Triple Interaction of Governance Quality and Firm Size (1) (2) (3) (4) (5) (6) (7) (8) All Sample Only Bordering Districts Fixed Capital Fixed Capital Permanent Contract Permanent Contract VARIABLES Land (Excluding Land (Excluding Labor Labor Labor Labor Land) Land) 2000 -0.0568* -0.0178 0.183*** -0.0663* 0.00427 0.0129 0.256*** -0.0513 (0.0323) (0.1160) (0.0408) (0.0374) (0.0927) (0.4270) (0.0936) (0.1070) 2001 -0.0740** -0.222** 0.205*** -0.0658* 0.0276 -0.404 0.381*** 0.113 (0.0305) (0.1030) (0.0384) (0.0354) (0.0875) (0.3700) (0.0884) (0.1010) 2002 -0.0962*** -0.310*** 0.221*** -0.119*** 0.118 -0.835** 0.343*** 0.108 (0.0308) (0.1020) (0.0386) (0.0356) (0.0878) (0.3850) (0.0882) (0.1010) 2003 -0.0842*** -0.425*** 0.294*** -0.0960*** 0.104 -0.494 0.399*** 0.0813 (0.0308) (0.1010) (0.0386) (0.0356) (0.0858) (0.3800) (0.0866) (0.0994) 2004 -0.0659** -0.357*** 0.299*** -0.0456 0.04 -0.516 0.355*** 0.166* (0.0289) (0.0979) (0.0364) (0.0335) (0.0799) (0.3680) (0.0803) (0.0925) 2005 -0.0184 -0.310*** 0.311*** 0.0191 0.00535 -0.5 0.390*** 0.154* (0.0293) (0.0980) (0.0368) (0.0338) (0.0802) (0.3730) (0.0804) (0.0929) 2006 0.00229 -0.435*** 0.300*** 0.00971 0.00682 -0.542 0.313*** 0.183** (0.0291) (0.0973) (0.0364) (0.0336) (0.0802) (0.3680) (0.0796) (0.0924) 2007 0.0635** -0.437*** 0.368*** 0.0382 0.0153 -0.837** 0.399*** 0.141 (0.0292) (0.0971) (0.0366) (0.0337) (0.0813) (0.3670) (0.0813) (0.0934) 2008 0.0946*** -0.359*** 0.364*** 0.102*** 0.0484 -0.750** 0.421*** 0.282*** (0.0298) (0.0976) (0.0373) (0.0343) (0.0828) (0.3700) (0.0832) (0.0954) 2009 0.0811*** -0.367*** 0.363*** 0.156*** 0.121 -0.625* 0.394*** 0.425*** (0.0299) (0.0991) (0.0375) (0.0346) (0.0824) (0.3690) (0.0823) (0.0949) 2010 -0.0058 -0.522*** 0.385*** 0.107*** 0.0501 -0.657* 0.394*** 0.436*** (0.0294) (0.0981) (0.0367) (0.0339) (0.0827) (0.3690) (0.0826) (0.0951) 2011 -0.022 -0.604*** 0.445*** 0.131*** 0.0595 -0.879** 0.489*** 0.445*** (0.0296) (0.0979) (0.0369) (0.0341) (0.0828) (0.3690) (0.0828) (0.0953) 2012 -0.0259 -0.632*** 0.473*** 0.189*** 0.0965 -1.060*** 0.466*** 0.366*** (0.0297) (0.0981) (0.0371) (0.0343) (0.0840) (0.3680) (0.0836) (0.0963) 2013 -0.0952*** -0.574*** 0.524*** 0.173*** -0.0528 -1.355*** 0.406*** 0.365*** (0.0296) (0.0997) (0.0369) (0.0342) (0.0859) (0.3790) (0.0860) (0.0985) 2014 -0.171*** -0.608*** 0.533*** 0.183*** -0.141 -0.633* 0.460*** 0.553*** (0.0297) (0.0998) (0.0369) (0.0342) (0.0863) (0.3800) (0.0856) (0.0989) 44 Low Governance 0.00615 0.0541 -0.167*** 0.0975* -0.0171 0.06 -0.251* 0.240* Quality 2000 (0.0457) (0.1830) (0.0576) (0.0527) (0.1240) (0.5460) (0.1300) (0.1420) Low Governance -0.0229 0.195 -0.0249 0.0532 -0.0292 0.437 -0.157 0.0478 Quality 2001 (0.0433) (0.1660) (0.0541) (0.0499) (0.1160) (0.4780) (0.1210) (0.1320) Low Governance -0.0475 0.155 -0.0605 0.0707 -0.240** 0.878* -0.195 -0.0979 Quality 2002 (0.0435) (0.1650) (0.0544) (0.0501) (0.1160) (0.4870) (0.1200) (0.1320) Low Governance -0.0616 0.197 -0.135** 0.0629 -0.189* 0.545 -0.240** 0.00213 Quality 2003 (0.0435) (0.1650) (0.0544) (0.0501) (0.1130) (0.4850) (0.1180) (0.1300) Low Governance -0.0357 0.133 -0.113** 0.0541 -0.119 0.308 -0.117 -0.0964 Quality 2004 (0.0410) (0.1590) (0.0513) (0.0472) (0.1060) (0.4650) (0.1110) (0.1220) Low Governance -0.0780* 0.0992 -0.0834 0.0599 -0.0739 0.385 -0.133 0.0042 Quality 2005 (0.0416) (0.1600) (0.0520) (0.0479) (0.1070) (0.4720) (0.1110) (0.1230) Low Governance -0.116*** 0.159 -0.129** 0.0286 -0.103 0.283 -0.0819 -0.0613 Quality 2006 (0.0413) (0.1600) (0.0516) (0.0475) (0.1060) (0.4680) (0.1100) (0.1220) Low Governance -0.120*** 0.147 -0.166*** 0.0444 -0.0561 0.599 -0.141 0.0239 Quality 2007 (0.0415) (0.1590) (0.0519) (0.0478) (0.1080) (0.4670) (0.1120) (0.1230) Low Governance -0.146*** 0.0627 -0.186*** 0.0202 -0.121 0.566 -0.190* -0.0921 Quality 2008 (0.0423) (0.1600) (0.0529) (0.0486) (0.1100) (0.4700) (0.1140) (0.1250) Low Governance -0.112*** 0.113 -0.189*** 0.0227 -0.239** 0.346 -0.159 -0.167 Quality 2009 (0.0425) (0.1610) (0.0530) (0.0489) (0.1090) (0.4700) (0.1130) (0.1250) Low Governance -0.129*** 0.164 -0.261*** 0.0293 -0.155 0.353 -0.279** -0.125 Quality 2010 (0.0420) (0.1600) (0.0522) (0.0483) (0.1100) (0.4690) (0.1140) (0.1260) Low Governance -0.0804* 0.234 -0.220*** 0.0817* -0.147 0.503 -0.167 -0.0772 Quality 2011 (0.0421) (0.1600) (0.0524) (0.0484) (0.1110) (0.4710) (0.1150) (0.1260) Low Governance -0.106** 0.22 -0.244*** 0.0365 -0.251** 0.678 -0.264** -0.0998 Quality 2012 (0.0424) (0.1600) (0.0526) (0.0486) (0.1120) (0.4690) (0.1160) (0.1280) Low Governance -0.0664 0.137 -0.284*** 0.0638 -0.116 1.079** -0.116 0.00053 Quality 2013 (0.0417) (0.1600) (0.0516) (0.0479) (0.1130) (0.4770) (0.1160) (0.1280) 45 Low Governance -0.0269 0.118 -0.218*** 0.0537 -0.072 0.264 -0.123 -0.274** Quality 2014 (0.0418) (0.1610) (0.0517) (0.0480) (0.1130) (0.4780) (0.1160) (0.1290) Large Firm 2000 0.0426 -0.155 -0.131*** 0.00198 0.201* -0.148 -0.121 0.0491 (0.0410) (0.1230) (0.0498) (0.0475) (0.1160) (0.4470) (0.1180) (0.1350) Large Firm 2001 0.0174 0.0867 -0.178*** -0.0254 0.126 0.202 -0.240** -0.0731 (0.0390) (0.1100) (0.0471) (0.0451) (0.1100) (0.3910) (0.1110) (0.1270) Large Firm 2002 -0.0462 0.106 -0.291*** -0.105** -0.157 0.55 -0.239** -0.348*** (0.0389) (0.1100) (0.0470) (0.0450) (0.1080) (0.4040) (0.1090) (0.1250) Large Firm 2003 -0.0254 0.179* -0.392*** -0.0964** -0.0313 0.269 -0.326*** -0.257** (0.0388) (0.1090) (0.0470) (0.0449) (0.1060) (0.3990) (0.1070) (0.1230) Large Firm 2004 -0.0041 0.127 -0.460*** -0.125*** 0.172* 0.319 -0.308*** -0.246** (0.0371) (0.1050) (0.0449) (0.0430) (0.1010) (0.3870) (0.1020) (0.1170) Large Firm 2005 0.00205 0.0902 -0.471*** -0.148*** 0.171* 0.304 -0.337*** -0.222* (0.0379) (0.1060) (0.0457) (0.0438) (0.1030) (0.3930) (0.1040) (0.1200) Large Firm 2006 9.3E-05 0.175* -0.512*** -0.142*** 0.196* 0.235 -0.355*** -0.290** (0.0376) (0.1050) (0.0453) (0.0434) (0.1030) (0.3870) (0.1030) (0.1190) Large Firm 2007 -0.0005 0.186* -0.565*** -0.101** 0.330*** 0.5 -0.436*** -0.149 (0.0377) (0.1050) (0.0454) (0.0436) (0.1040) (0.3860) (0.1040) (0.1200) Large Firm 2008 -0.0111 0.102 -0.612*** -0.135*** 0.326*** 0.545 -0.496*** -0.214* (0.0383) (0.1050) (0.0461) (0.0441) (0.1050) (0.3890) (0.1060) (0.1220) Large Firm 2009 0.00302 0.119 -0.702*** -0.154*** 0.336*** 0.365 -0.511*** -0.154 (0.0384) (0.1060) (0.0462) (0.0444) (0.1050) (0.3880) (0.1050) (0.1210) Large Firm 2010 0.0334 0.193* -0.766*** -0.133*** 0.281*** 0.32 -0.604*** -0.256** (0.0379) (0.1050) (0.0455) (0.0438) (0.1050) (0.3890) (0.1060) (0.1220) Large Firm 2011 0.0732* 0.268** -0.814*** -0.0626 0.349*** 0.512 -0.713*** -0.212* (0.0380) (0.1050) (0.0456) (0.0438) (0.1060) (0.3880) (0.1060) (0.1220) Large Firm 2012 0.0482 0.225** -0.816*** -0.0979** 0.360*** 0.576 -0.696*** -0.121 (0.0381) (0.1050) (0.0457) (0.0439) (0.1070) (0.3870) (0.1070) (0.1230) Large Firm 2013 0.0678* 0.123 -0.855*** -0.0807* 0.521*** 0.737* -0.488*** -0.0509 (0.0381) (0.1070) (0.0455) (0.0439) (0.1120) (0.4000) (0.1120) (0.1290) Large Firm 2014 0.0658* 0.121 -0.852*** -0.0728* 0.637*** 0.237 -0.530*** -0.0103 (0.0381) (0.1070) (0.0456) (0.0440) (0.1130) (0.4010) (0.1120) (0.1290) Low Gov Qual X -0.110* 0.0473 0.0857 -0.116* -0.326** 0.127 0.158 -0.294* Large Firm 2000 (0.0573) (0.1930) (0.0698) (0.0662) (0.1520) (0.5730) (0.1580) (0.1750) 46 Low Gov Qual X -0.0603 -0.124 0.0431 -0.0341 -0.302** -0.242 0.105 -0.135 Large Firm 2001 (0.0544) (0.1770) (0.0659) (0.0629) (0.1420) (0.5040) (0.1470) (0.1640) Low Gov Qual X -0.0078 -0.0908 0.0716 0.0214 0.0583 -0.678 0.128 0.238 Large Firm 2002 (0.0541) (0.1750) (0.0657) (0.0624) (0.1400) (0.5120) (0.1450) (0.1610) Low Gov Qual X -0.0569 -0.144 0.169*** -0.0036 -0.112 -0.523 0.158 0.0891 Large Firm 2003 (0.0540) (0.1750) (0.0656) (0.0623) (0.1380) (0.5090) (0.1420) (0.1590) Low Gov Qual X -0.0936* -0.0898 0.192*** 0.00865 -0.311** -0.287 0.0934 0.114 Large Firm 2004 (0.0518) (0.1690) (0.0627) (0.0597) (0.1310) (0.4900) (0.1360) (0.1510) Low Gov Qual X -0.0438 -0.0394 0.148** 0.0217 -0.16 -0.372 0.0949 0.133 Large Firm 2005 (0.0530) (0.1710) (0.0641) (0.0611) (0.1340) (0.4980) (0.1380) (0.1540) Low Gov Qual X -0.015 -0.0806 0.205*** 0.0342 -0.141 -0.179 0.116 0.213 Large Firm 2006 (0.0526) (0.1700) (0.0635) (0.0606) (0.1330) (0.4920) (0.1370) (0.1530) Low Gov Qual X -0.0747 -0.0876 0.182*** -0.0629 -0.312** -0.469 0.0997 0.0445 Large Firm 2007 (0.0528) (0.1690) (0.0638) (0.0608) (0.1350) (0.4920) (0.1390) (0.1550) Low Gov Qual X -0.116** -0.0554 0.153** -0.0927 -0.326** -0.737 0.112 0.056 Large Firm 2008 (0.0536) (0.1700) (0.0648) (0.0617) (0.1360) (0.4940) (0.1410) (0.1570) Low Gov Qual X -0.191*** -0.136 0.160** -0.153** -0.361*** -0.411 0.0464 -0.104 Large Firm 2009 (0.0537) (0.1710) (0.0648) (0.0619) (0.1360) (0.4940) (0.1400) (0.1560) Low Gov Qual X -0.143*** -0.165 0.200*** -0.124** -0.319** -0.387 0.237* -0.013 Large Firm 2010 (0.0533) (0.1700) (0.0641) (0.0613) (0.1370) (0.4940) (0.1410) (0.1570) Low Gov Qual X -0.155*** -0.213 0.210*** -0.121** -0.372*** -0.497 0.201 0.0287 Large Firm 2011 (0.0533) (0.1690) (0.0641) (0.0613) (0.1370) (0.4950) (0.1410) (0.1580) Low Gov Qual X -0.148*** -0.107 0.142** -0.0907 -0.382*** -0.525 0.183 -0.0056 Large Firm 2012 (0.0535) (0.1700) (0.0643) (0.0615) (0.1390) (0.4940) (0.1420) (0.1590) Low Gov Qual X -0.157*** 0.0147 0.0969 -0.0711 -0.580*** -0.891* -0.151 -0.123 Large Firm 2013 (0.0529) (0.1700) (0.0634) (0.0608) (0.1420) (0.5030) (0.1450) (0.1630) Low Gov Qual X -0.151*** 0.0212 0.087 -0.0417 -0.667*** -0.326 -0.121 -0.0272 Large Firm 2014 (0.0531) (0.1710) (0.0636) (0.0610) (0.1430) (0.5040) (0.1450) (0.1640) 47 Observations 493,570 159,762 379,869 520,837 97,651 27,845 80,336 100,494 R-squared 0.867 0.9 0.845 0.788 0.866 0.87 0.83 0.754 Firm FE YES YES YES YES YES YES YES YES 48 Figures Figure A4.1: Within-State Variance in Adjustment Costs - Permanent Labor 2 3 5 6 7 2.2 2.4 2.6 2.8 3 4 4 5 2.5 4 3 3 2 3 1.5 2 2 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 YR YR YR YR YR 8 9 10 18 19 3.5 5 4 3 4 3.5 2.5 3 4 3 2.5 2 2 3 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 YR YR YR YR YR 20 21 22 23 24 4.5 3.5 3 4 4 4 3.5 2.5 3 3.5 3 2.5 3 2 2 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 YR YR YR YR YR 27 28 29 30 32 3.23.43.6 6.5 10 4 6 6 3.5 8 5.5 5 3 6 5 3 4 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 YR YR YR YR YR 33 8 7 6 2000 2005 2010 2015 YR Figure A4.2: Within-State Variance in Adjustment Costs - Contract Labor 2 3 5 6 7 3.5 6 6 6 2 3 4 5 3 4 4 4 2.5 2 2 2 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 YR YR YR YR YR 8 9 10 18 19 4.5 5 5 5 6 4 5 4 4 4 3.5 3 4 3 3 3 3 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 YR YR YR YR YR 20 21 22 23 24 4.5 3.5 6 5 5 3 4 5 4 4 2.5 3.5 2 4 3 3 3 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 YR YR YR YR YR 27 28 29 30 32 10 4.5 6 6 6 5.5 4 6 8 4 5 4 3.5 5 4 2 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 YR YR YR YR YR 33 8 6 4 2000 2005 2010 2015 YR 49 Figure A4.3: Within-State Variance in Adjustment Costs - Land 2 3 5 6 7 4.5 4 5 4 5 3.5 4 3.5 4 4 3.5 3 3 3 2.5 3 3 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 YR YR YR YR YR 8 9 10 18 19 5 5 9 9 6 8 4.5 4.5 5.5 8 6 7 5 4 4 7 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 YR YR YR YR YR 20 21 22 23 24 4.5 5.5 6 7 6 5 6 5 4 5 4.5 4 5 3 4 3.5 4 4 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 YR YR YR YR YR 27 28 29 30 32 6 6 6 8 7 5 5 5 6 6 4 4 4 5 4 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 YR YR YR YR YR 33 5.5 5 4.5 2000 2005 2010 2015 YR Figure A4.4: Within-State Variance in Adjustment Costs – Fixed Capital (Excluding Land) 2 3 5 6 7 3.5 4 6 3 4 3.5 3 3 2.5 4 2.5 3 2 2.5 2 2 2 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 YR YR YR YR YR 8 9 10 18 19 4.5 4 4 4 3 3.5 4 3.5 2.5 3.5 3 3 2.5 2 3 2 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 YR YR YR YR YR 20 21 22 23 24 3.5 5 4 4 4 3.5 3.5 3.5 3 4 3 2.5 2.5 3 3 3 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 YR YR YR YR YR 27 28 29 30 32 4.5 10 12 9 5 7 8 6 4 4 6 7 6 8 4 5 2 3 3.5 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 2000 2005 2010 2015 YR YR YR YR YR 50 33 7 5 6 4 2000 2005 2010 20 Figure A6: Adjustment Cost Trends in Low relative to High Governance Quality States, NeighboringDistricts Only a. Permanent Labor b. Contract Labor c. Land d. Fixed Capital 51