تاثیر بزرگراه های اصلاح شده در شرکت های هند

India's Golden Quadrilateral Program, a major highway project, aimed at improving the quality and width of existing highways connecting the four largest cities in India. It affected the quality of highways available to firms in cities that lay along the routes of the four upgraded highways, while leaving the quality of highways available to firms in other cities unaffected. This feature of the project allows for a difference-in-difference estimation strategy, where status on and off the improved highways, and distance from them, are used as treatment variables. This strategy is implemented using data from the 2002 and 2005 rounds of the World Bank Enterprise Surveys for India. Firms in cities affected by the Golden Quadrilateral highway project reduced their average stock of input inventories by between 6 and 12 days’ worth of production. Firms in cities where road quality did not improve showed no significant changes. The reduction in stocks of input inventories also varied inversely with the distance between the city in which a firm was located and the nearest city on an improved highway. Firms on the Golden Quadrilateral were also more likely to have switched the supplier who provided them with their primary input, suggesting that they saw reason to re-optimize their choice of supplier after the arrival of better highways. Consistent with these findings, firms on the improved highways reported decreased transportation obstacles to production, while firms in control cities reported no such change.

This paper studies the effects of an ambitious program of highway improvements in India on firms in that country. It is motivated by – and contributes to – a large literature on the economic effects of investments in large-scale infrastructure, a key issue in development. Highways, the quintessential example of such infrastructure investments, are studied in papers such as Michaels, 2008, Chandra and Thompson, 2000 and Fernald, 1998. Such infrastructure investments are often posited as being essential for higher economic growth. Recently, for example, the World Bank has argued that Africa suffers from an “extensive infrastructure deficit” (Foster and Briceño-Garmendia 2009). Its simulations suggest that bridging this gap could have large growth dividends. For example, if “all African countries were to catch up with Mauritius in infrastructure, per capita economic growth in the region could increase by 2.2 percentage points” (ibid). Significant budgetary resources are allocated to (or sought for) highway construction in many countries, based in part on such hypothesized causal link between infrastructure development and economic growth. Yet, empirical evidence on this issue is “extremely controversial, and consists of studies that are divided on both the magnitude and direction of the net effect of infrastructure spending on economic growth” (Chandra and Thompson 2000). This is because estimating the economic effects of infrastructure must contend with several issues that complicate the analysis. In what follows, I briefly lay out the nature of these problems, before discussing how I attempt to deal with them in this paper through a combination of the identification strategy, data, and outcome variables used. The first key issue that arises in trying to estimate the economic effects of infrastructure is that of the endogenous placement of new infrastructure, which makes it difficult to clearly quantify causal effects. This problem can be summed up as follows: do areas with (more, better) infrastructure show better economic outcomes because of the infrastructure, or is better infrastructure attracted by (the potential for) better economic outcomes? For example, we may observe that areas with better roads grow faster. However, if we suspect that roads are likely to have been built or improved in areas that had the highest potential for economic growth, simple correlations between road quality and growth are likely to overstate the impact of roads on growth (see, for example, the discussion in Qian et al., 2009). Only truly random placement of a major infrastructure project can hope to deal with the endogenous placement of infrastructure in a completely satisfactory fashion. Particularly in the case of large-scale projects like inter-state highways or big dams, such random placement is difficult to conceive of. But in its absence, a series of recent papers cognizant of the endogeneity issue have attempted to tackle it creatively by using plausible instruments or a natural experiment to aid identification. For example, Duflo and Pande (2007) use gradient to instrument for the placement of hydroelectric projects in India to obtain plausibly causal estimates of the economic effects of large hydroelectric projects. In the case of highway construction, both Chandra and Thompson, 2000 and Michaels, 2008 use a feature of the US Interstate Highway construction program that allows them to treat it as a natural experiment that affected counties through which the new interstate highways passed differently from those it bypassed. The idea derives from the nature of the highway-building exercise. When a highway is built to connect cities A and B, it must pass through areas that lie in between the two, thus contributing to improved infrastructure in places that happen to lie in between the (possibly endogenously chosen) points that the highway is built to connect. If the precise route of the highway was not manipulated to include some intermediate areas –whether counties, districts, or cities – and exclude others based on factors correlated with the outcomes of interest, then the highway construction can be treated as exogenous to the areas that the highway runs through. This paper approaches the issue of endogeneity using an identification strategy similar in spirit to that used by Chandra and Thompson (2000) and developed further in Michaels (2008). The nature of the project studied (India's Golden Quadrilateral, or “GQ” highway scheme) and panel data on a sample of firms representative of the country's entire non-agricultural private economy enable me to tease out plausibly causal estimates of the impact of improved highways on firms in India. I do so using a differences-in-differences approach, where changes in relevant outcomes for firms affected by the Golden Quadrilateral are compared to changes in outcomes for other firms. The Golden Quadrilateral program sought to improve the quality and width of 5,846 km of existing highways connecting the four largest cities in India (Delhi, Mumbai, Kolkata, and Chennai). This dramatically improved the quality of highway transportation available to firms in cities that happened to lie along the routes of the four existing highways that were upgraded. However, firms in cities not on these highways did not directly benefit from comparable increases in the quality of highways available to them (see Fig. 1 for a depiction of the location of cities in the data relative to the upgraded highways). The position of the city where a firm is located relative to the highways that were upgraded thus created variation in the extent to which a firm should have seen the quality of the highways available to it improve as a result of the Golden Quadrilateral project. Full-size image (87 K) Fig. 1. Location of Cities in Data Relative to Golden Quadrilateral Highways. Figure options I use data from two waves of the World Bank's Enterprise Surveys for India, which collected information about a random sample of firms in the formal sector, stratified by sector of activity, firm size, and geographical location so as to generate a sample1 of firms ”representative of the whole non-agricultural private economy”. The panel structure of the data allows me to compare the responses of a representative sample of 1,091 firms in 37 Indian cities in the year 2002, when the project had just begun, to the responses of the same firms in the year 2005, when it was approximately two-thirds complete, providing plausibly causal estimates of the effects of the project. Table 1a and b summarize relevant features of the cities, industries, and firms in the panel. Table 1a. Characteristic of Cities in Sample, By Location On and Off Golden Quadrilateral. I II III IV V VI VII City State Nodal City or Suburb? Nearest City on GQ Driving distance in km from City in IV No. of Firms Population in Millions, 2001 Panel A: Cities On The Golden Quadrilateral Ahmedabad Gujarat No Itself 0 59 4.53 Bangalore Karnataka No Itself 0 43 5.71 Kolkata West Bengal Yes Itself 0 67 13.21 Chennai Tamil Nadu No Itself 0 35 6.56 Delhi Delhi Yes Itself 0 50 12.90 Faridabad Harvana Yes Itself 0 23 1.06 Ghaziabad Uttar Pradesh Yes Itself 0 22 0.97 Guntur Andhra Pradesh No Itself 0 17 0.51 Gurgaon HalTana Yes Itself 0 12 0.23 Hosur Karnataka No Itself 0 21 0.08 Hubli-Dharwad Karnataka No Itself 0 16 0.79 Kanpur Uttar Pradesh No Itself 0 44 2.72 Mumbai Maharashtra Yes Itself 0 32 16.43 NOIDA Uttar Pradesh Yes Itself 0 13 0.31 Pune Maharashtra No Itself 0 24 3.76 Surat Gujarat No Itself 0 35 2.81 Thane Maharashtra Yes Itself 0 15 1.26 Vadodara Gujarat No Itself 0 66 1.49 Vijayawada Andhra Praclesh No Itself 0 45 1.04 Panel B: Cities Off The Golden Quadrilateral Bhopal Madhya Pradesh No Agra 541 25 1.46 Calicut Kerala No Bangalore 355 8 0.88 Chandigarh Punjab/Han-ana No Delhi 238 16 0.81 Cochin Kerala No Bangalore 533 16 1.36 Coimbatore Tamil Nadu No Bangalore 340 46 1.46 Gwalior Madhya Pradesh No Agra 118 26 0.87 Hyderabad Anclhra Pradesh No Vijavawada 267 56 5.74 Indore Madhva Pradesh No Ahmedabad 415 41 1.52 Jalandhar Punjab No Delhi 364 22 0.72 Lucknow Uttar Pradesh No Kanpur 77 13 2.25 Lucthiana Punjab No Delhi 305 33 1.40 Madurai Tamil Nadu No Chennai 444 22 1.20 Mangalore Karnataka No Bangalore 347 10 0.54 Mysore Karnataka No Bangalore 139 35 0.79 Nagpur Maharashtra No Surat 747 29 2.13 Nashik Maharashtra No Mumbai 185 18 1.15 Palakkad Kerala No Bangalore 392 15 0.20 Panipat Harvana No Delhi 85 21 0.26 Table options This paper's second contribution is an attempt to empirically document some of the microeconomic channels through which infrastructure affects economic outcomes. Theory suggests that firm-level variables (such as inventories, input costs, or capacity utilization) may respond to improvements in transport infrastructure (see, for example, Shirley and Clifford, 2004). But while a number of papers identify macroeconomic effects of infrastructure on growth, employment, or price convergence (see, for example, Demurger, 2001, Chandra and Thompson, 2000 and Donaldson, 2010), the microeconomic channels identified by the theoretical literature have received relatively little empirical attention beyond a small number of instructive case studies (see, for example, Gulyani, 2001 and Holl, 2004) and a recent paper by Duranton et al. (2011) which explores the effect of highways on the volume, value and composition of trade. However, the Enterprise Survey data contain firms' responses to questions that allow me to directly measure how the choices that firms make about inventories and input suppliers are affected by the quality of highway infrastructure. I use two “treatment variables” to capture the degree to which a given firm is affected by the highway program. The first is a binary variable, which takes the value 1 if the firm is located in one of the 19 cities in the data on an upgraded highway and 0 if it is located in one of the other 18 cities. However, firms in “off-project” cities, whose cargo could also use the improved highways for at least part of their journey, were also affected by the project, albeit indirectly and to a smaller extent. This motivates a second, continuous, treatment variable: the driving distance of the city a firm is located in from the nearest city on the Golden Quadrilateral. I also use two samples. The first (the “full sample”) includes all the cities in the sample, while for the other I exclude the four metropolitan cities (Delhi, Mumbai, Kolkata, and Chennai) and their contiguous suburbs (Gurgaon, Faridabad, Ghaziabad and NOIDA in the case of Delhi, and Thane in the case of Mumbai). This “restricted sample” thus includes firms off the Golden Quadrilateral and those in non-nodal Golden Quadrilateral cities. The rationale for excluding the nodal cities from the analysis is that their status as 'on-Golden Quadrilateral' cities was a matter of design rather than fortuitousness, since these cities formed the nodes of the new system of highways. The choice of sample makes no difference to the sign and significance of the results I find, though the magnitudes of the effects naturally vary by sample. In the discussion, I emphasize the restricted sample because the identification is cleanest in this case. Further, as is clear for instance from Table 2a, the effects found are driven almost entirely by the cities in the restricted sample. Table 1b. Distribution of Firms Across Treatment and Control Cities and Industries. Non-Nodal Nodal GQ Full Sample Non-GQ GQ cities cities No. of Firms 1091 452 370 269 Industrial Distribution (Per Cent of Firms in San pie in Each Industry) Textiles 139 11.06 13.51 14.50 Auto components 136 10.62 13.78 13.75 Garments 122 13.94 4.32 15.99 Drugs and phaima 117 7.52 16.22 8.55 Food processing 94 13.94 5.95 3.35 Electiical appliances 93 8.63 7.57 9.67 Machine tools 72 6.64 9.19 2.97 Electiouics 61 3.76 4.86 9.67 Metals 60 8.41 2.43 4.83 Plastics 51 5.09 5.68 2.60 Other chemicals 44 1.99 6.76 3.72 Leather 37 2.65 2.70 5.58 Rubber 20 2.21 1.35 1.86 Paper 14 0.66 2.16 1.12 Agro processing 8 0.44 1.08 0.74 Paints 6 0.66 0.81 0.00 Marine processing 6 0.88 0.54 0.00 Sugar 4 0.00 0.54 0.4 Cosmetics 3 0.44 0.27 0.00 Wood 3 0.44 0.00 0.37 Mineral processing 1 0.00 0.27 0.00 Table options Table 2a. Days to Inventory Held. 2002 2005 Change, 2002-5 Non-GQ cities Days of Inventory Held 23.13 23.27 0.14 (S.E) (29.22) (19.71) Non-Nodal GQ cities Days of Inventory Held 38.10 27.54 -10.56*** (S.E.) (52.36 (31.33) Nodal GQ cities('and their suburbs) Days of Inventory Held 26.78 24.31 -2.47 (S.E.) (35.68) (21.35) All GQ cities Days of Inventory Held 33.44 26.18 -7.26* (S.E) (46.54) (27.59) Table options The key findings relate to firms’ holdings of inventories of their principal input, and are as follows. A comparison of means shows that non-nodal firms on the Golden Quadrilateral highways reduced their average input inventory (measured in terms of the number of days of production the inventory held was sufficient for) by 10.5 days more than firms situated on other highways. In the full sample, the corresponding effect was 7 days’ worth of production. The regression estimates are both statistically and economically significant, and suggest that being on the Golden Quadrilateral reduces inventory holdings by the equivalent of at least 6 days' worth of production (with the highest estimate from the restricted sample showing an effect just under twice as large). The effects are most pronounced for industries such as pharmaceuticals, food processing and electronics (see Table 2c). Being an additional kilometer further from the Golden Quadrilateral causes a firm to hold between 0.13 and 0.22 days' more worth of inventories. Thus, inventory management became significantly leaner for treatment firms relative to control firms after the improved highways were put into place. Table 2b. Length of Relationship with Supplier 2002 2005 Change. 2002-5 Non-GQ cities Years with Main Input Supplier 4.15 4.54 0.39** (S.E.) (1.39) (1.01) Non-Nodal GQ cities Years with Main Input Supplier 4.68 4.83 0.15 (S.E.) (0.84) (0.61) Nodal GQ cities(and their suburbs) Years with Main Input Supplier 4.65 4.75 0.1 (S.E) (0.80) 0.73) All GQ cities Years with Main Input Supplier 4.67 4.79 0.12 (S.E) (0.82) (0.67) *denotes significance at 90% level, ** at 95% and *** at 99%. Table options Table 2c. Inventory Holding by Industry. Inventory Holdings Industry City Category 2002 2005 Change 2002-05 Garments Non-GQ 126 20.8 22.2 1.4 Non-nodal GQ 26 41.7 29.7 -12 Textiles Non-GQ 96 33.6 27.41 -6.19** Non-nodal GQ 100 39.9 29.8 -10.1*** Drugs & Pharmaceuticals Non-GQ 66 24.6 24.7 0.1 Non-nodal GQ 120 53.8 37.2 -16.6*** Electronics Non-GQ 32 35.2 33.9 -1.3 Non-nodal GQ 36 42.2 20.2 -22*** Electrical Appliances Non-GQ 76 21.4 19.9 -1.5 Non-nodal GQ 54 38.3 29.9 -8.4** Machine Tools Non-GQ 56 30.9 28.5 -2.4 Non-nodal GQ 68 38.9 34.3 -4.6 Auto Components Non-GQ 94 18 23.1 5.1 Non-nodal GQ 100 24.7 20.8 -3.9 Leather Goods Non-GQ 24 21 14.41 -6.6*** Non-nodal GQ 20 33.5 24.6 -8.9** Food Processing Non-GQ 124 16.4 18.8 2.4 Non-nodal GQ 38 51.7 25.6 -26.1** Plastics Non-GQ 46 19.8 23.1 3.3 Non-nodal GQ 42 21.5 18.9 -2.6 Metals & Metal Peoducts Non-GQ 68 17.8 19.3 1.5 Non-nodal GQ 18 22.4 15.3 -7.1* Chemicals Non-GQ 18 17.6 15.2 -2.4 Non-nodal GQ 48 36.5 20.9 -15.6* Notes: * denotes significance at 90%, ** at 95* and *** at 99%. The other industries in the data-set had too few observations (randing from I to 20 firms in the entire sample) to make any meaningful comparison of means (See Table 1b). Table options I also find that firms in cities that gained better highway access were more likely to have switched the supplier who provided them with their primary input than firms in cities where road quality did not undergo a comparable improvement. Since any supplier available to a firm before the change continues to be available after the change, the fact that more firms on the improved highways change their primary input supplier suggests that a better supplier becomes feasible after the highway construction. This is a revealed preference argument: if highway quality was not constraining firms' input supplier choices, the choice set would still expand when the highways were upgraded, but firms would not necessarily change suppliers. That they do suggests that their old choice now appears sub-optimal. The data therefore suggest that the benefit from switching in terms of lower prices, more convenient location, etc. for some firms exceeds any switching costs, which, given the benefits of long-term relationships with suppliers, might be substantial. The statistical significance of these subsidiary results on firms' supplier choices is less robust to changes in the sample and specification. Their signs and magnitudes are, however, stable. Finally, the surveys also provide direct evidence about firms' perceptions of the degree to which transportation constitutes an obstacle to their productive activities. Such data are difficult to interpret because of the inherent subjectivity of criteria like “severity”. This is particularly problematic in a cross-sectional setting: one person may have very different notions of what constitutes a “major obstacle” to another's understanding of the same issue. Panel data obviates this problem to a certain extent, since it means that we are comparing the same firms’ responses before the project to their subsequent replies. I find that firms in cities affected by the highway project became 7.6 percentage points, or about 60 per cent, less likely to report that transportation constituted a 'major' or ' very severe' obstacle to production, while there was no significant change in the responses of firms in off-Golden Quadrilateral cities. I do not emphasize these results because of the well-known problems with perceptions-based data. But even if little can be said about this final finding in isolation, what firms say they experiences does reinforce the picture painted using other, more “objective” data they also report. These results, particularly those on inventories, support the idea that the Golden Quadrilateral project eased the extent to which transportation infrastructure constrains firms, as well as giving us a sense of the channels through which such constraints operated. These results thus link this paper with recent literature on the channels through which infrastructure reduces the costs of economic activity. However, it is worth emphasizing at the outset that the nature of the data used here preclude any attempts to measure the effects of the project on the number of firms, product prices (as, for example, in Donaldson, 2010), etc. In that sense, the effects measured here only provide a partial, and necessarily incomplete, glimpse of the changes firms experienced as a result of access to better highways. The plan of the rest of the paper is as follows. Section 2 provides background information on the specific context of this paper, the Indian highway network and the Golden Quadrilateral project. Section 3 explains the identification strategy used in the paper, which relies on using the Golden Quadrilateral Project as a positive shock to the quality of highway infrastructure available to cities it passed through versus those it bypassed that is unrelated to (firms in) the cities' (potential) economic performance. Section 4 motivates the focus on inventory behavior and supplier relationships, and discusses the data and variables used in this paper. Section 5 presents the key results and their interpretation. Section 6 concludes.

Firms in cities that lay along one of the four national highways connecting the four largest cities in India that the Indian government upgraded as part of its Golden Quadrilateral report holding about 10.5 days’ worth of production less of input inventories in 2005, when much of the project had been implemented, than in 2002, when work had just begun, while firms which lay in cities off the Golden Quadrilateral highways report no such change. Similar results are obtained when distance from the upgraded highways is used as a measure of treatment. Such firms also show a greater propensity to change suppliers between the two years, suggesting that a larger fraction of them found their existing arrangements sub-optimal than did firms not on the new highways. Firms on the upgraded highways also became much less likely to report that transportation was a major obstacle to obstacle to production in 2005 relative to their responses to the same question in 2002. Seen together, these pieces of evidence suggest that improved highways facilitated productive choices that firms may have wanted to make even earlier, but were constrained from being able to make by the quality of highways available to them.