WPS6622 Policy Research Working Paper 6622 Coagglomeration of Formal and Informal Industry Evidence from India Megha Mukim The World Bank Finance, Economics and Urban Development Department September 2013 Policy Research Working Paper 6622 Abstract A large and growing informal sector is a major explain much of formal-informal coagglomeration. Also, feature of developing countries. This paper analyzes within-industry coagglomeration matters mostly to coagglomeration patterns between formal and informal small- and medium-sized formal firm births. Traditional manufacturing enterprises in India. It studies (a) the measures of agglomeration remain important in causes underlying these patterns and (b) the positive explaining new industrial activity, whether in the formal externalities, if any, on the entry of new firms. The or the informal sector. analysis finds that buyer-supplier and technology linkages This paper is a product of the Finance, Economics and Urban Development Department. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The author may be contacted at mmukim@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 Coagglomeration of Formal and Informal Industry: Evidence from India * Megha Mukim ‡ Keywords: Coagglomeration, Informal Sector, India JEL Classification: R1, R3, O1 * I am grateful to Diana Weinhold, Holger Görg, Henry Overman, Waltraud Schelkle, Maarten Bosker, Harry Garretson and Walker Hanlon for helpful comments and suggestions, to Michaela Fuchs at the European Regional Science Association conference in Sweden, to Elisabet Viladecans- Marsal at the Urban Economic Association conference in Denver, and to Vernon Henderson, Nathanial Baum-Snow, Adam Storeygard and other participants at the Urban Studies seminar at Brown University for useful discussions on an earlier draft. Kristen Behrans and two anonymous referees provided excellent suggestions for the current draft. ‡ Economist, Finance, Economics and Urban Development, World Bank. 1 Introduction The unorganized, or informal, sector 1 is an important means of livelihood to millions of people in developing countries. Because of its very nature – it is unregulated by government – data collection and subsequent analysis lags far behind that for the formal sector. In India, the informal sector 2 often falls outside the scope for planned development efforts, and thus remains in the shadows with regard to productivity, social security and statistics. Within the non-farm sector, informal enterprises account for 43.2% of Net Domestic Product (NDP) and employ 71.6% of the workforce, a fact hardly peculiar to the Indian economy. Informal activity accounts for a majority of employment and makes a significant contribution to NDP in several developing countries in Asia, Africa, Eastern Europe and Latin America - on average the informal share of employment ranges from 24 percent in transition economies, to 50 percent in Latin America and over 70 percent in Sub-Saharan Africa (see Kar and Marjit 2008 and Jutting and Laiglesia 2009 for a range of rates of informal employment). The importance of the informal sector in India shows little sign of abating, while that of the formal sector remains stagnant, and yet little is known about the relationship, whether symbiotic or competitive, between the two sectors. In India, new firm activity is strongly related to existing firm activity – however the literature often assumes that formal and informal sectors function independently of one another. Since systematic data on the transactions of firms in the informal sector are scarce, any knowledge of its relationship with formal firms is often based on qualitative case studies – Pieters et al (2010, 2011) and Bairagya (2010) are exceptions. Given that policy makers are interested in encouraging new economic activity, they need a better understanding of the links between existing formal and informal firms and whether these has any beneficial spillovers for new firms. In this paper, I investigate two main issues. First, to what extent and why informal firms cluster with those in the formal sector, and second, whether the coagglomeration of formal and informal firms has any effect on the rate of new firms setting up shop within a given location. I focus on formal and informal manufacturing firms in India. Data on informal firms 1 A number of countries, including India, often use the terms ‘unorganised sector’ and ‘informal sector’ interchangeably. This paper will stick with the term ‘informal’. 2 I will use ‘sector’ to indicate the division between formal and informal, and ‘industry’ to differentiate between different types of manufacturing sub-divisions, such as furniture, paper, textiles etc. 2 are taken from surveys conducted by the National Sample Survey Organisation (NSSO) in 2000-2001 and 2005-2006. Data on formal firms are taken from the Annual Survey of Industries (ASI) for the two years corresponding to the NSSO surveys. I assume that firms’ entry could be driven by the effects of agglomeration and/or coagglomeration, market access, the size of the market and a set of unobserved attributes, which could be specific to the location or to the industry. I study formal and informal firms across 22 2-digit manufacturing industries, 459 districts and two years (2000-01 and 2005-06). This paper will start with a brief overview of the existing literature – studies that focus on the relationship of the formal and the informal sectors, and those that provide evidence on the factors that drive firm location decisions. I will then describe patterns of coagglomeration between formal and informal manufacturing firms in India and subsequently investigate the factors that could help explain the observed trends. In section 4, I will estimate the effect of coagglomeration and other types of clustering on new firms’ location decisions and demonstrate how formal firms differ from their informal counterparts. Data sources relevant to the analysis will be described early-on in the sections. Section 5 will discuss the results and section 6 will conclude. I provide a flavor of the main results upfront. Coagglomeration between formal and informal manufacturing firms in India is not very high and it has been falling over time, except for industries such as apparel, furniture and metal-making where it has risen over the period of analysis. When formal and informal firms do coagglomerate, it seems to be driven mostly by buyer-supplier linkages and owing to spillovers of equipment and design from formal to informal firms. Buyer-linkages are especially strong between smaller formal firms and their informal counterparts. New informal firms tend to choose locations based strongly on own- industry agglomeration. Input-output linkages are important for both formal and informal firms. Large formal firms tend to avoid industrially diverse districts and informal-formal coagglomeration is almost entirely irrelevant for them. Informal firms are attracted to markets, formal firms to locations with better access to markets elsewhere. 3 2 Related Literature The literature, both theoretical and empirical, on coagglomeration of firms is new 3, but fast- growing – Ellison and Glaeser (1997) coined the term ‘coagglomeration’ to describe the tendency of various industries to agglomerate in close proximity to one another. Helsley and Strange (2012) provide a succinct but thorough review of the theory describing the foundations for coagglomeration. Most of existing literature focuses on coagglomeration dynamics across industries (for instance, whether textile producers coagglomerate with leather producers) and not across sectors within industries (for instance, whether formal and informal textile producers coagglomerate). There are few empirical papers that study patterns of coagglomeration across sectors within a given industry. For instance, Barrios et al (2006) study coagglomeration between domestic and foreign firms in Ireland and find evidence of FDI-related local spillovers. He et al (2012) provide detailed descriptions of coagglomeration between exporters, non-exporters and foreign firms within industries and across locations in China. Howard and Tarp (2012) investigate coagglomeration between low and high-tech clusters in Vietnam and find that technology transfers and skills correlations explain the observed patterns. What explains coagglomeration, and what sorts of benefits might this lead to? Firms could coagglomerate because of greater access to specialized inputs, labor-market pooling, and knowledge and technology spillovers (Ellison and Glaeser 2010). Firms could also be attracted to the same location owing to the presence of natural endowments or public goods. The focus of this paper is on the coagglomeration between formal and informal firms. There are a number of studies that have investigated the degree of complementarity or substitutability between the formal and informal sectors in India. Mitra (2009) finds that the incidence of informality is high in more industrialised Indian states, suggesting that informal activity could be complementary to the formal sector. Some working papers focus on studying the production linkages between the formal and informal sectors in India, and its effect on informal employment and on the size of the informal sector – see Pieters et al (2010), Pieters et al (2011) and Bairagya (2010). Pieters et al (2010, 2011) find evidence of complementarities between the modern informal sector and firms in the formal sector, which are linked through outsourcing. Their analysis is conducted across 2-digit manufacturing 3 Ciccone and Peri (2006), Bacolod et al (2008, 2009), Gabe and Abel (2013) study complementarities across workers in urban labor markets. 4 industries at the level of the state. Since states in India are often the size of small countries, this paper uses districts as the unit of analysis for location. In urban regions, districts often correspond to cities and surrounding areas. Marjit (2003) uses a general equilibrium model and argues that only the capital-intensive segment of the informal sector is complementary to the formal sector. On the other hand, Dutta et al (2011) find that informal activity serves as a substitute to the formal sector when they observe increased informal activity in states in India that suffer from higher levels of corruption (that inhibits formal activity). The informal sector in India largely ignores labor regulations, officially recognized collective bargaining processes, taxes or institutional obligations. There is some research (Marjit and Kar 2009) to show that informal manufacturing and self-employed units accumulate fixed assets and invest and that often they are able to do so in times when their formal counterparts are mired in complex regulations. Informal production can be an important input to the production of intermediate goods, processed exports and import substitutes, supported by supply side contracts with the formal sector, leading to beneficial spillovers for both sectors. For instance, informal carpet weavers in Agra operate alongside larger, more formal carpet designers and exporting firms in the city. Ranis and Steward (1999) illustrate how better formal-informal linkages could lead to growth within the informal sector. House (1984) and Arimah (2001) find that larger informal firms (in Kenya and Nigeria, respectively) often benefit from subcontracting or direct sales linkages with formal firms. Bairagya (2010) finds that in the presence of trade liberalization, production linkages between formal and informal manufacturing firms leads to an increase in the size of the informal sector. Overman and Venables (2005) and Duranton (2008) describe a number of case studies illustrating the generation of agglomeration economies from the interaction between formal and informal enterprises. Firms, whether formal or informal, could also benefit from the more traditional types of agglomeration - i.e. with other similar firms from the same industry, related industry, or within a location with a diverse industry mix. Marshall (1919) theorized that clusters of firms, predominantly in the same sector, could take advantage of localization economies, such as the sharing of sector-specific inputs, skilled labor and knowledge. Firms in a given industry and those in related industries might agglomerate to enjoy the benefits such as inter- industry linkages, buyer-supplier networks, and opportunities for efficient sub-contracting (Venables 1996). An overall large size of the urban agglomeration and its more diverse 5 industry mix is also thought to provide external benefits beyond those realized within a single sector or due to a tight buyer-supplier network (Henderson 2003). The focus of this paper will be to describe the effect of formal-informal coagglomeration versus that of other types of agglomeration on the generation of new activity, i.e. firm births, in the formal and the informal sectors. There are a number of papers that study firm births in the formal and the informal sectors in India – these are referred to as firms’ location decisions (Lall and Mengistae 2005, Lall and Chakravorty 2005), or entrepreneurship (Ghani et al 2011a,b). Mukim and Nunnenkamp (2012) consider the location decisions of foreign firms in India and find that foreign investors tend to increasingly favor locations that already host other foreign investors. Lall and Mengistae (2005) find that formal manufacturing firm location choices in India are strongly influenced by the local business environment and own- industry agglomeration economies. Lall and Chakravorty (2005) find that new manufacturing activity is attracted to locations with existing industrial concentrations, exacerbating spatial inequalities. Lall et al (2003, 2004) study the effect of agglomeration economies on the performance of existing formal manufacturing firms in India. A priori, there is no reason to assume that informal sector activity remains unaffected by agglomeration economies. In the absence of access to formal credit facilities, or alternatively since they are untouched by changes in regulations, the importance of buyer-supplier linkages and informal networks of social interaction could be more important to informal firms than to firms operating in the formal sector. Mukim (2011) studies the location decisions of new informal firms, manufacturing and services, in India and finds that input-output linkages are important in determining spatial entry patterns. Ghani et al (2011a) also study the spatial determinants of new formal and informal firm births, i.e. entrepreneurial activity. They find that input-output agglomeration economies matter for manufacturing firm entry, for both formal and informal firms. Ghani et al (2011b) compare the spatial determinants of entry separately for male and female entrepreneurs in informal manufacturing and services in India. They find that the effect of female-owned firm agglomeration economies (particularly input- output linkages) is a strong predictor of new female entrepreneurial activity. This paper will build upon the results from earlier analysis in two ways. First, it will focus on the coagglomeration of informal and formal activity within manufacturing industries in India. While there are papers that have studied the production linkages between formal and informal 6 firms in India, they have ignored other kinds of linkages and have focused their analysis at the level of states. Second, the paper will contribute to the growing literature on spatial work on entry patterns of formal and informal firms in India, and explicitly account for the effect of formal-informal coagglomeration versus that of traditional measures of agglomeration. 3 Formal-Informal Coagglomeration 3.1 Descriptive Statistics There is a continuum along which informal and formal manufacturing firms could exist. Informal firms are those that are not registered under the Factories Act of 1948, which requires all firms engaged in manufacturing to register if they employ 10 workers or more and use power, or if they employ 20 workers or more. All public sector enterprises are assumed to be in the formal sector. Small formal firms (for instance, those employing 10-20 workers and using electricity) might be very similar to their larger informal counterparts. However, the data seems to suggest that formal firms are a very different animal. In 2005-06, the average number of employees per formal firm was 141 while the average informal firm had 3 employees. Informal firms are predominantly very small-scale enterprises – between 90 to 93 percent of the establishments in the sample employ less than 5 workers, and between 68 and 70 percent employ no hired labor at all. Within the formal sector, small firms (i.e. less than 5 workers) comprise 5 percent of the sample, medium (i.e. employ more than 5, but less than 100 workers) comprise 71 percent and large (i.e. employ more than 100 workers) comprise 23 percent. Informal firms tend to belong mostly to industries such as apparel, food, textiles, wood and furniture. This includes cottage and household industries, khadi and village industries, handlooms, handicrafts, coir, sericulture etc, set up all over the country in rural, semi-urban and urban environments. Formal firms are fairly well distributed across different manufacturing industries, such as non-metallic products, chemicals, basic metals and machinery and equipment. Data on informal firms and employment is drawn from the Fifty-Sixth Round (July 2000 - June 2001) and the Sixty-Second Round (July 2005 - June 2006) of the National Sample Survey Organisation. The surveys cover rural and urban areas, and the sample size varies 7 from 100 units in remote areas to over 10,000 units in major cities. Data on formal firms is drawn from the Annual Survey of Industries (ASI), which conducts repeated cross-sectional annual surveys that cover all firms registered under the Factories Act of 1948. ASI survey data is available for each year starting in 1980 – however, for the purposes of this paper, data corresponding to the 2000-01 and 2005-06 surveys is used. This paper focuses only on manufacturing industries that cover National Industrial 2-digit Classification (NIC) 15 to 37. The surveys cover around 558 districts in 2000-01 and 576 districts in 2005-06 – an increase of 18 districts. Since the total landmass of the country remains unchanged, the new districts are formed out of existing districts. For example, mean district population fell from 1.88 million in 2000-01 to 1.71 million in 2005-06. For the purpose of analysis between the two time periods, I combine the new districts, essentially keeping the 2000-01 district boundaries intact. For instance, Delhi is one district in 2000-01, but is divided into 9 districts in 2005-06 – I use the combined data from the 9 new districts so as to match this to the data from the original district boundaries 4. In short, the analysis includes the original 558 districts, of which data is available for a total of 459 districts. After the 2005-06 district boundaries are combined in line with the 2000-01 boundaries, the mean district population for 2005-06 rises to 2.11 million. Do formal and informal firms cluster in the same locations? I calculate the Theil Index for formal and informal firms and list the top 20 districts – see Table 1. Only two districts (Mumbai and Ludhiana) are common to the top five districts across both sectors and an additional two (Delhi and Ahmadabad) common to the top ten, indicating that there could be limited concordance across formal and informal clustering. To measure coagglomeration between the formal and informal sector directly, I follow Ellison and Glaeser (1997, 2010) in principle but modify their Coagglomeration Index somewhat. I compute coagglomeration measures formal and informal sectors within each 2-digit manufacturing industry, across each of the two years for which I have data at my disposal. In other words, the index for the coagglomeration of sectors i and j is: 4 More difficult examples of combination include the new Dantewara district (in the state of Chhattisgarh) which was carved out of Bastar district in the same state – information that was gleaned from searches on Google. 8 ∑ =1( − )( − ) = , 1 − ∑ 2 =1 Where m indexes districts, i indexes the formal sector and j the informal sector. is the share of formal sector employment contained in district m, is the share of informal sector employment contained in district m and measures the aggregate size of district m, which is measured here as the mean employment share in the district across the two sectors (formal and informal). The coagglomeration index is based on measures of formal and informal sector employment. Informal employment within industries across districts averaged 962,000, while formal employment averaged 2,386. In addition, on average, informal employment across industries and districts grew by 950% between 2000-01 and 2005-06, while formal employment grew by a mere 13%. Table 1: Contributions to the Theil Index District Informal District Formal Mumbai 255.43 Bangalore Urban 549.39 Ludhiana 146.34 Mumbai city 433.85 South Tripura 100.84 Coimbatore 334.99 Kolkata 80.03 Vellore 246.75 Delhi 52.53 Ludhiana 180.36 Ahmadabad 47.11 Pune 163.48 Jaipur 44.08 Thane 143.64 South 24 Parganas 43.08 Pudukkottai 129.05 Coimbatore 42.63 Delhi 117.97 West Tripura 42.19 Surat 110.00 Surat 39.93 Ahmadabad 96.51 Thane 39.70 Chennai 93.01 North 24 Parganas 39.52 Thiruvallur 84.87 Haora 37.08 Guntur 83.19 Murshidabad 36.44 Kollam 81.97 Srinagar 34.17 Nizamabad 78.29 Hyderabad 34.00 Gurgaon 76.82 Varanasi 32.53 Gautam Buddha Nagar 75.07 Virudhunagar 31.18 Daman 68.29 Vellore 29.69 Rangareddi 68.27 Table 2 lists the coagglomeration measures for formal-informal sectors by industry for 2000- 01 and 2005-06. Negative values of the index arise when formal and informal sectors are agglomerated in different areas. For example, when the coagglomeration index equals - 9 0.0259 for Manufacture of Coke in 2000-01, it implies that informal coke manufactures tend to agglomerate in locations different from formal coke manufacturers. Industries like wood, food, apparel and fur, non-metallic mineral products tend to rank higher on the coagglomeration index from one year to the next. Averaged across industries, coagglomeration dropped from 0.0102 in 2000-01 to 0.0042 in 2005-06. Similar to the Ellison-Glaeser agglomeration index, the no-coagglomeration benchmark is when the value of the index is zero. In general, if the coagglomeration index is greater than 0.05, the sectors are considered to be highly concentrated. In 2000-01, food products, tobacco, textiles and non-metallic minerals seemed reasonably coagglomerated with values averaging above the 0.03 mark, although these dropped to 0.01 in 2005-06. And indeed, except for three industries – apparel, fabricated metal products and furniture, the measure of coagglomeration fell in 2004-05, suggesting that formal and informal firms were less likely to coagglomerate. Table 2: Formal-Informal Coagglomeration NIC Industry Name 2000-01 2005-06 15 Food products and beverages 0.0318 0.0173 16 Tobacco products 0.0300 -0.0001 17 Textiles 0.0314 0.0152 18 Apparel and Fur 0.0118 0.0172 19 Leather, luggage, and footwear 0.0181 0.0095 20 Wood, except furniture 0.0311 0.0174 21 Paper and paper products 0.0144 0.0066 22 Publishing, printing, recorded media -0.0002 0.0142 23 Coke -0.0259 -0.0064 24 Chemicals and chemical products 0.0214 0.0097 25 Rubber and plastic products -0.0052 0.0068 26 Other non-metallic mineral products 0.0361 0.0166 27 Basic metals 0.0086 0.0064 28 Fabricated metal products 0.0138 0.0169 29 Machinery and equipment 0.0147 0.0119 30 Office 0.0252 -0.0540 31 Electrical machinery and apparatus -0.0040 0.0113 32 Radio -0.0219 -0.0100 33 Medical, watches etc 0.0102 -0.0131 34 Vehicles 0.0037 -0.0012 35 Other transport 0.0125 0.0025 36 Furniture 0.0139 0.0158 Notes: Ellison-Glaeser (2007) coagglomeration index for formal and informal sectors within every 2- digit manufacturing industry. 10 3.2 Why do firms coagglomerate? Theories of agglomeration predict that firms cluster to be in close proximity to goods, people and ideas. Or, they could agglomerate owing to the presence of natural advantages in a location. These explanations could apply equally to firms across the formal and the informal sector. I describe below the different variables I construct to measure the possible explanations for formal-informal coagglomeration. Buyer-Supplier Linkages (proximity to goods): Firms from a given sector could locate close to a firm from another sector if they bought or sold goods to and from one another 5. For example, Arimah (2001) provides evidence of linkages between the formal and informal sector in Nigeria in the form of sub-contracting and the flow of consumer goods and raw materials. Carr et al (2000) illustrate that formal enterprises often provide materials and inputs to formal enterprises, which then transform them to send back to formal enterprises. In such cases, proximity would lower costs for both formal and informal firms. To assess the importance of downstream linkages (i.e. supplier linkages) I use data from the NSSO on informal firms that are sub-contracting, i.e. working on a contract with firms in the private sector, co-operatives or with a contractor. In other words, →, refers to the proportion of informal firms that are supplying to the formal sector within a given industry, district and year. To assess the importance of upstream (i.e. buyer) linkages, I use data from the ASI on formal firms that subcontract their inputs (excluding labor) from outside the firm. Thus, →, is the proportion of total input expenses that are being subcontracted 6 within a given industry. Labor-market linkages (or proximity to people): Agglomeration could occur because workers are able to move across firms and industries, or in this case, between firms in the informal and the formal sector. Rosenthal and Strange (2001) provide an overview of the importance of labor market pooling in explaining the spatial concentration of firms. To assess labor movement across the informal and formal sectors, I use data from the ASI on labor employed 5 It could also be argued that buyer-supplier linkages and coagglomeration are endogenous – in other words, firms may use the outputs of (or sell to) particular sectors simply because these sectors tend to coagglomerate (for other reasons). 6 This is a proxy for subcontracting by formal firms to informal firms – but it is equally possible that formal firms could subcontract inputs to other formal firms. 11 through a contract. In other words, →, is the proportion of employees that are being subcontracted from outside the firm. Technology Spillovers (proximity to ideas): Firms could also co-locate if this helped the transfer of technology within a cluster. Porter (1990) points out that knowledge sharing often takes place through the processes of buying and supplying. To assess the importance of technological linkages I use data from the NSSO on the sub-set of informal firms that are sub-contracting and who receive either equipment or design from the contractor. I construct a variable ℎ→, that refers to the proportion of informal firms within a given industry that are receiving a technology transfer from the contractor. Table 3 summarizes the mean values of each of the explanatory variables across districts and how these vary across different manufacturing industries from 2000-01 to 2005-06. In 2005- 06, on average, formal firms sub-contracted 27 percent of their total employment – industries like tobacco, coke and non-metallic mineral products tended to subcontract more, while publishing, office equipment and textiles were on the lower side. The average proportion of inputs subcontracted was also high (37 percent in 2005-06), with textiles, apparel and basic metals in the lead, and coke, chemicals and non-metallic minerals lagging behind. Interestingly, a simple pair-wise correlation reveals that labor-market linkages seem to be negatively correlated with buyer linkages, suggesting that firms that subcontract labor are less likely to subcontract inputs. The proportion of informal firms subcontracting, i.e., supplier linkages as defined above, is relatively low at an average of 8 percent in 2005-06. More sophisticated manufacturing products, such as vehicles, radio, publishing and rubber and plastics seem to subcontract more, while informal firms in industries like food and tobacco subcontract less. The proportion of informal firms supplied with equipment and design through subcontracting is the lowest, averaging around 2 percent for 2005-06. Informal firms in industries such as other transport, medical/watches, publishing and machinery and equipment seem to have better technological linkages through their contractors, while radio, coke, tobacco and office equipment have the least linkages. 12 Table 3: Descriptive Statistics Labor market Technology Linkages Buyer Linkages Seller Linkages Linkages NIC Industry Name 2000 2005 ∆ 2000 2005 ∆ 2000 2005 ∆ 2000 2005 ∆ 15 Food products, beverages 0.19 0.28 47% 0.30 0.37 25% 0.02 0.02 5% 0.01 0.01 -33% 16 Tobacco products 0.38 0.52 37% 0.41 0.38 -9% 0.03 0.02 -41% 0.01 0.00 -58% 17 Textiles 0.12 0.17 45% 0.20 0.52 155% 0.09 0.09 -3% 0.01 0.02 33% 18 Apparel, Fur 0.19 0.29 50% 0.44 0.46 5% 0.08 0.06 -32% 0.01 0.02 27% 19 Leather, luggage, footwear 0.21 0.26 20% 0.31 0.35 11% 0.06 0.06 2% 0.01 0.01 34% 20 Wood 0.21 0.28 29% 0.21 0.28 32% 0.08 0.11 29% 0.01 0.02 80% 21 Paper, paper products 0.21 0.23 13% 0.09 0.41 333% 0.05 0.07 30% 0.01 0.01 81% 22 Publishing, printing etc 0.08 0.15 77% 0.24 0.34 46% 0.21 0.19 -7% 0.03 0.03 -2% 23 Coke 0.25 0.41 61% 0.11 0.19 73% 0.03 0.01 -58% 0.03 0.00 -100% 24 Chemicals, chemical products 0.19 0.28 45% 0.18 0.27 48% 0.03 0.01 -57% 0.02 0.01 -58% 25 Rubber, plastic products 0.14 0.25 76% 0.21 0.40 90% 0.06 0.12 83% 0.01 0.02 28% 26 Non-metallic mineral products 0.33 0.44 33% 0.09 0.27 183% 0.02 0.05 147% 0.01 0.01 183% 27 Basic metals 0.25 0.30 22% 0.26 0.43 66% 0.10 0.10 0% 0.04 0.02 -45% 28 Fabricated metal products 0.20 0.34 70% 0.28 0.42 50% 0.11 0.10 -15% 0.02 0.02 -14% 29 Machinery and equipment 0.11 0.19 80% 0.30 0.41 36% 0.10 0.08 -18% 0.02 0.03 26% 30 Office 0.13 0.16 24% 0.32 0.31 -3% 0.17 0.07 -60% 0.00 0.00 31 Electrical machinery etc 0.15 0.25 71% 0.21 0.36 71% 0.05 0.06 26% 0.01 0.02 57% 32 Radio 0.11 0.18 70% 0.23 0.35 55% 0.07 0.12 77% 0.00 0.00 33 Medical, watches etc 0.07 0.16 113% 0.32 0.32 2% 0.06 0.09 56% 0.01 0.04 615% 34 Vehicles 0.15 0.26 67% 0.31 0.40 28% 0.15 0.11 -31% 0.01 0.01 17% 35 Other transport 0.15 0.27 79% 0.30 0.41 36% 0.14 0.11 -17% 0.05 0.05 -13% 36 Furniture 0.26 0.29 11% 0.28 0.35 26% 0.13 0.13 7% 0.02 0.02 60% To understand the extent to which each of these linkages might be driving the coagglomeration of formal and informal firms, I run the following regression: , = → , + →, + →, + ℎ→, + + + where the dependent variable is coagglomeration between formal (a) and informal (b) firms in any given industry k in year t. The fact that some locations might be attractive to both informal and formal firms owing to certain types of natural advantages is captured by a district-specific dummy, and I also include a year-specific dummy. Since the dependent variable varies only by industry and year, I do not include industry dummies. Table 4 summarizes the results. Coagglomeration between the formal and informal sectors seems to be driven largely by buyer-supplier linkages and technology transfers. A unit increase in technological linkages, i.e. the proportion of informal firms that receive equipment or designs from the contractor, leads to a 0.0044 unit increase in coagglomeration 13 – see model (3). Given that the average value of the coagglomeration index in 2005-06 was 0.0042, this is a sizeable effect. Surprisingly, the effect of labor-market linkages was either negative or insignificant - if stringent labor regulations in the formal manufacturing sector led firms to subcontract labor from the informal market, then that should’ve given formal firms more reason to coagglomerate with their informal counterparts. This might indicate that formal firms are more likely to subcontract informal laborers that do not operate or work in informal enterprises. Since I do not include industry-specific dummies, it might also be more or less feasible to subcontract to the informal sector in some industries than others, perhaps because they are regulated differently. Table 4: Explaining Formal-Informal Sector Coagglomeration All Firms SMEs Only Variable (1) (2) (3) (4) Labor-market linkages -0.0031*** 0.0009** -0.0005 -0.0008 [0.000] [0.000] [0.001] [0.000] Buyer Linkages -0.0007 0.0017*** 0.0014*** 0.0044*** [0.000] [0.000] [0.000] [0.001] Supplier Linkages 0.0038*** 0.0039*** 0.0041*** 0.0003 [0.001] [0.001] [0.001] [0.000] Technological Linkages 0.0049*** 0.0038** 0.0044*** 0.0036 [0.002] [0.002] [0.002] [0.002] Constant 0.0029*** 0.0038*** 0.0033* 0.0030 [0.000] [0.000] [0.002] [0.002] Year Fixed Effects ×    District Fixed Effects × ×   Observations 2,328 2,328 2,328 1,763 R-squared 0.029 0.184 0.284 0.324 Notes: Dependent variable is the coagglomeration index for formal-informal clustering for each manufacturing industry. Standard errors in square brackets clustered at the industry level *** p<0.01, ** p<0.05, * p<0.1 I also calculate separately the coagglomeration index and the explanatory variables for the sub-set of small and medium enterprises (SMEs) in the formal sector, i.e. firms with less than 100 employees, and then re-run the regression – see model (4). I find that whilst the effect of buyer-linkages, i.e. the proportion of inputs being subcontracted by formal SMEs, is now thrice as strong, that of other variables is now irrelevant. This indicates that SMEs in the 14 formal sector coagglomerate with informal firms primarily for the purpose of production linkages. 4 New Firms, Agglomeration and Coagglomeration Why should we care about coagglomeration between the formal and informal sectors? In this section, I measure the effect of coagglomeration, relative to other sorts of agglomeration, on firm births in the formal and the informal sectors. Since I am studying the decisions of new establishments, i.e. firm births, the existing environment is taken as exogenous. Both sources of data on formal and informal firms, the ASI and the NSSO respectively, provide information on firm births for the two years of interest - 2000-01 and 2005-06. The NSSO asks informal firms in the survey if the enterprise had been in operation for less than 3 years – firms that reply in the affirmative are coded as firm births. ASI provides information on the year of initial production of the formal firm, which allows me to identify which firms were recent entrants to the market in the two years of interest. In line with the NSSO survey, I take new entrants to the market for the given year and the two preceding years. I create a repeated cross-sectional data set, pooled across the two years – 2000-01 and 2005-06 for the dependent variable, i.e. the count of new firms per capita. Table 5: Characteristics of the data Data Std. Source Year Variable # Mean Dev. Min. Max. count 5915 0.72 2.8 0 136 2000-01 count>0 1511 2.8 4.9 1 136 ASI count 6192 0.4 2.08 0 89 2005-06 count>0 961 2.6 4.7 1 89 count 5915 228 1109 0 46476 2000-01 count>0 2656 508 1612 1 46476 NSSO count 6192 324 1730 0 57801 2005-06 count>0 2748 731 2540 1 57801 Table 5 summarizes the data on firm births. Clearly, not every 2-digit-industry and district pair experience positive values of firm births and new firm entrants in the formal sector are rarer and fewer than those in the informal sector. I assume that these zero values are genuine, in that they indicate no new firm entrants in the given year, industry and district. 15 The large number of zero firm births in the data bring with them computational problems. Rosenthal and Strange (2003) use tobit estimations to deal with the problem of zero births. The bias resulting from noisy estimates of fixed effects in nonlinear models is not a problem when the number of observations per fixed effect is large – in this case, the data lacks enough variation when district-level dummies are introduced since there are 459 districts in the data. Thus, I run tobit regressions with year and industry dummies only. I also use a probit model that predicts the discrete zero-positive outcome for firm births defined at the level of districts, industries and the two years. I use the dprobit command in Stata that calculates marginal probability effects, in this case, at the sample mean values of the regressors. And, finally, I estimate a linear OLS model, which drops observations with zero births, but the results of which remain unbiased with the use of district, industry and year dummies. My results, presented in Section 5, continue to be robust. I am mainly interested in parceling out the effect of the traditional agglomeration variables, i.e. within-industry clustering, input-output or inter-industry clustering, industrial diversity and market access, from the effect of formal-informal coagglomeration on firm births. I regress the count of new firms per capita within a location as a function of factors common to the location (i.e. industrial diversity, market access and wealth) and those common to particular industries (i.e. intra and inter-industry clustering and formal-informal coagglomeration). My main estimating equation is as follows: =∝ +−1 + −1 + −1 + + −1 + −1 + + + + I run separate regressions for firm births in the formal and the informal sectors. The dependent variable in the model is the count of new firms in 2-digit industry k in district j and at time t , whilst all the explanatory variables, with the exception of coagglomeration, in the model are defined at time t − 1. The section below describes each of the explanatory variables. 16 4.1 Specification of Variables The economic geography variables in this model are represented by market access (−1 ), intra-industry or localization economies (−1 ), inter-industry linkages (−1 ), the Hirschman-Herfindahl Index for industrial diversity (−1) and coagglomeration between the formal and informal sectors ( ). , , and are the district, industry and year dummies and are used wherever possible. To reduce clutter the time sub-script has been excluded in the descriptions of the variables below. Localization economies are measured as the proportion of sector k ' s employment in district j as a share of all of sector k ' s total employment in the country. The higher this value, the higher the expectation of intra-industry concentration benefits in the district. E j ,k σ jk = Ek To evaluate the strength of buyer-supplier linkages for each industry, a summation of district industry employment weighted by the industry’s input and output coefficients from the column and row vectors from the national input-output account are used: n τ jk = ∑ ( wki + wko )e jk k =1 where, τ jk is the strength of the buyer-supplier linkages, wki is industry k’s national input coefficient column vector, wko is industry k’s national output coefficient row vector, and e jk is total employment for industry k in district j. The measure examines local level inter- industry linkages based on national input-output accounts. I also compute buyer and supplier linkages separately to use in my regressions, but I find that these variables are highly correlated (correlation coefficient=0.83) and that this leads to serious multicollinearity problems. Thus, I use an input-output variable that captures the strength of both buyer and supplier linkages. The Hirschman-Herfindahl Index is used to measure the degree of economic diversity. Unlike measures of specialization, which focus on one industry, the diversity index considers the 17 industry mix of the entire regional economy. The largest value for U j is one when the entire regional economy is dominated by a single industry. Thus a higher value signifies lower level of economic diversity.  E 2 U j = ∑E  jk  k  j  Market access, or the potential for opportunities for interactions with neighboring districts is defined as: MA j = ∑ Sm m dbj −m where MA j is the accessibility indicator estimated for location j, Sm is a size indicator at destination m (district population), d jm is a measure of distance between origin j and destination m, and b describes how increasing distance (orthodomic) reduces the expected level of interaction. In the original model proposed by Hanson (1959), b is an exponent describing the effect of the travel time between the zones. In this specification, the exponent value is set to 1.Transport is allowed to occur along a 500-kilometre radius. The size of the district j is not included in the computation of market access – only that of neighboring districts is taken into account. I also include the percentage of high-income households ( WE j ) within a district as an indicator of the measure of consumer expenditures within a district. The variable is constructed using household consumption data and refers to those households that belong to the highest monthly per-capita consumption expenditure group. The actual MPCE category differs depending on the year of the survey, the type of district (rural or urban) and the population of the district. Coagglomeration ( ) is defined in Section 3, and measures coagglomeration between the formal and informal sectors within each 2-digit manufacturing industry. It is also standard to control for location-level infrastructure such as transport, electricity, banks etc. However, as physical infrastructure changes relatively slowly over time (in this case over a 5-year period), the effects should be more or less soaked up by district dummies. 18 While I run separate regressions for births in the formal sector and those in the informal sector, note that the explanatory variables are common across the two sets of regressions since these have been constructed taking total employment, i.e. across the formal and informal sector, into account. The simple descriptive statistics of the explanatory variables is presented in Table 6. Table 6: Descriptive Statistics Expected Mean Variable Varies by Sign 1999-2000 2004-2005 Localisation + District × Industry × Year 0.0040 0.0045 Input + District × Industry × Year 8,773 8,531 Urbanisation - District × Year 0.359 0.382 Coagglomeration* + Industry × Year 0.012 0.005 Market Access + District × Year 870,570 904,737 Wealth** + District × Year 5.98 5.58 Population + District × Year 2,155,122 1,983,314 Notes: *Refers to 2000-01, and 2005-06. **Refers to percentage of the population. There are a total of 579 districts and 22 2-digit manufacturing industries in the sample. Data on the explanatory variables is drawn from the Employment and Unemployment Surveys - Round 55.10 (July 1999 – June 2000) and Round 61.10 (July 2004 – June 2005). This data, which is disaggregated by industry and district, allows me to construct the agglomeration variables. Data on wealth and population within the district is also drawn from the household surveys. 4.2 Endogeneity Concerns Although the dependent variable is the count per capita of firm births and the existing industrial structure is taken as given, the effect of unobservable factors could bias the coefficients. The underlying assumption within the model is that if a particular location offers some inherent features that improve the profitability of certain economic industries, this will increase firm births. Such inherent features may be related to natural endowments or regulatory specificities, but they could also have to do with essentially un-measurable factors such as local business cultures. It is important to distinguish whether firms are attracted by a common unobservable, whether they derive benefits from being located in close proximity to one another, or whether it is some combination of the two. 19 I include an arsenal of fixed effects to control for omitted variables bias in my estimations, specifically, industry, district and year fixed effects. District fixed effects successfully control for any time-invariant characteristics of the district, such as the presence of natural resources, climate, proximity to the coast – in short, all features of the natural geography of the district. Industry fixed effects control for any time-invariant characteristics particular to the industry. Year fixed effects control for any unobservable time-specific shocks. However, there could be industry-district conditions that could also affect the coefficients. I lack enough variation in the data to be able to include district × industry dummies, and so am unable to control satisfactorily for certain types of omitted variables bias. 5 Results and Discussion To deal with the zero firm births in the data, I use a Tobit model, wherein the response variable is the count of firm births per capita – this includes zero firm births. However, since the number of observations per district fixed effect is not large enough for computation, I only include industry and year dummies. I then use a Probit model, where the dependent variable is a discrete variable representing the zero-positive outcomes for firm births. I use the full range of district, industry and year fixed effects and report marginal effects. And, finally, I estimate a linear OLS model, where the dependent variable is the log of non-zero counts of new firms per capita, and I include the full range of fixed effects. Buyer-supplier linkages, the Herfindahl index, market access are in logs, while localization, coagglomeration and wealth are not. I run separate regressions for informal and formal firm births – see Tables 7 and 8. My results are reasonably robust across the different models. The tobit coefficients describe the linear effect on the uncensored latent variable and not the observed outcome. In other words, if localization were to increase by one unit, the expected count of new firms per capita would increase by 0.00214 while holding all other variables in the model constant – see Table 7. The probit model reports marginal effects, i.e. a change in the probability of a firm birth for a given change in the explanatory variable. In other words, if localization were to increase by one unit, the probability of a firm birth would increase by 1.508. The OLS coefficients can be interpreted as follows – if localization were to increase by one unit, informal firm births per capita would increase by 84.1 percent. 20 I find that on average, localization i.e. intra-industry clustering seems to positively affect the probability of informal firms entrants, and it also results in new firm births per capita. In the same vein, buyer-supplier linkages also have a strong and positive effect on new informal firm entry and count per capita. Ghani et al (2011a) also note that the strength of buyer and supplier linkages has a positive effect on new employment in informal manufacturing in India. Coagglomeration between formal and informal firms seems to have no discernible effect, although it seems to negatively affect the probability of entry of new informal firms. Table 7: Informal Firm Births Tobit Probit OLS Variable (1) (2) (3) Localization 0.00214 1.508* 0.841*** [0.00134] [0.873] [2.792] Buyer-Supplier 9.51e-05*** 0.0360*** 0.161*** [9.69e-06] [0.00661] [0.0297] Herfindahl Index -0.000159* -0.0366 -0.0699 [2.51e-05] [0.0523] [0.295] Coagglomeration -0.00172 -0.911* -0.837 [0.00256] [2.063] [8.362] Market Access 3.65e-06 0.0132 0.419 [2.32e-05] [0.0482] [0.0116] Wealth 1.25e-05* 0.00687** 0.0101 [2.27e-06] [0.00335] [0.0124] Year Fixed Effects    Industry Fixed Effects    District Fixed Effects ×   Observations 5,829 5,584 3,367 R-squared 0.0266 0.287 0.458 The dependent variables are as follows: for model (1) it is the count of new informal firms per capita; for model (2) it is a dummy variable indicating zero and positive counts; for model (3) it is the log of count of new informal firms per capita. Marginal effects are reported for the Probit model. Numbers in brackets are standard errors, clustered at the district level; McFadden’s Pseudo R-squared values are presented for models (1) and (2). *** p<0.01, ** p<0.05, * p<0.1 On the other hand, localization economies seem to have no effect on the probability or the count per capita of formal firm births – see Table 8. Instead, a decrease in industrial diversity, given by a rise in the Herfindahl Index (higher levels=lower diversity) seems to positively affect both, the probability and the count per capita of firm births. Again, Ghani et al (2011a) find a similar result for new employment in formal manufacturing. In addition, buyer- supplier linkages across industries are an important and positive predictor of formal firm births. The effect of informal-formal coagglomeration increases the probability of births, but its effect on the count per capita is unclear. Accessibility i.e. being located close to larger, more populated districts is not a predictor of firm births, but it does positively affect the count per capita of firm births. Table 8: Formal Firm Births Tobit Probit OLS Variable (1) (2) (3) Localization 7.41e-07 0.900 0.766 [1.27e-05] [0.750] [2.361] Buyer-Supplier 1.04e-06*** 0.0665*** 0.133*** [1.11e-07] [0.00695] [0.0292] Herfindahl Index 2.37e-06*** 0.0723 0.441** [2.60e-07] [0.0530] [0.223] Coagglomeration 9.55e-05** 0.202* 1.673 [2.95e-05] [1.790] [4.001] Market Access 5.16e-07** 0.00260 0.0219** [2.45e-07] [0.0411] [0.0165] Wealth 3.64e-08 0.00188 -0.0120 [2.38e-08] [0.00302] [0.0121] Year Fixed Effects    Industry Fixed Effects    District Fixed Effects ×   Observations 5,829 4,223 1,303 R-squared 0.0344 0.343 0.636 The dependent variables are as follows: for model (1) it is the count of new informal firms per capita; for model (2) it is a dummy variable indicating zero and positive counts; for model (3) it is the log of count of new informal firms per capita. Marginal effects are reported for the Probit model. Numbers in brackets are standard errors, clustered at the district level; Pseudo R-squared values are presented for models (1) and (2). *** p<0.01, ** p<0.05, * p<0.1 It seems that linkages to buyers and suppliers in related industries are a significant and consistent predictor of informal and formal firm births. Informal firm entry is affected by intra-industry clustering, while formal firm entry seems to be attracted to regions with less 22 industrial diversity, but higher informal-formal coagglomeration within a given industry. The percentage of wealthy households positively affects informal firm activity, while access to larger markets positively affects formal firm activity. This suggests that informal firm births might have more to do with population patterns within a location, while formal firm births are drawn to locations with better access to other, larger, locations. 5.1 Differentiating by Size As a robustness check, I carry out the same exercise as above but differentiate between firms of different sizes. Perhaps large firms in the informal sector could be similar to smaller firms within the formal sector – and perhaps it is not formality, but size, that is the important delineating parameter. I divide firms in the informal and the formal sector by the number of employees. For informal firms, I divide the sample of enterprises into those that are small (i.e. employ less than 5 workers) and large (i.e. they employ more than 5 workers). Almost 90 percent of the firms in the sample, thus defined, are small-scale enterprises. For formal firms, I divide the sample of enterprises into those that are small (i.e. employ less than 5 workers), medium (i.e. employ more than 5, but less than 100 workers) and large (i.e. employ more than 100 workers). Small formal firms comprise 5 percent of the sample, medium comprise 71 percent and large comprise 24 percent. If formal and informal firms are different primarily as a function of their size, then we should expect to see small informal firms behave like to small formal firms, and large informal firms perhaps behave like to medium formal firms. Since I am mainly interested in studying the factors that drive up the count of new firms per capita, and because I would like to control for unobservables at the level of districts, as well as industries and across time, I retain the simple OLS specification. The dependent variable is the log of the count of new firm births per capita. A few interesting observations emerge – see Table 9. Small firms in the informal sector do behave somewhat like their small counterparts in the formal sector – firm births across both types of firms are positively affected by intra-industry clustering, a result that was not observed when firm births in the formal sector were taken as a whole. Large (i.e. more than 5 employees) informal firms seem to shy away from wealthier locations, same as medium-sized 23 formal firms (i.e. more than 5, but less than 100 employees). However, this is where the similarities end. Buyer-supplier linkages across industries are significant for all kinds of firms, except for small formal firms. Industrial diversity, or lack thereof, positively affects formal firm births irrespective of firm size. Formal-informal coagglomeration positively affects small and medium formal firm births – recall that the effect of this variable was insignificant for the full sample of formal firms. It is possible that formal-informal clustering within a given industry might encourage new formal firms if these firms were more likely to buy from or sell to existing coagglomerations of formal-informal firms. Indeed, Holl (2008) finds that industry agglomeration in Spain tends to facilitate subcontracting between large and small manufacturing firms. Table 9: OLS Regressions by Size Informal Formal Variable Small Large Small Medium Large Localization 0.702** 0.545 0.61*** -0.413 -0.884 [3.267] [3.251] [3.405] [1.113] [1.862] Buyer-Supplier 0.0724** 0.145*** -0.0435 0.104*** 0.194*** [0.0355] [0.0460] [0.0330] [0.0119] [0.0261] Herfindahl Index 0.236 -0.563 1.158*** 0.565*** 0.432** [0.299] [0.505] [0.228] [0.0989] [0.189] Coagglomeration -8.813 -10.06 0.3*** 0.24** 1.047 [6.823] [9.243] [29.23] [4.813] [7.451] Market Access -0.148 -1.271 0.288*** 0.00522 0.469** [0.240] [1.611] [0.0876] [0.0736] [0.192] Wealth 0.0153 -0.0330* 0.0105 -0.0184*** 0.00752 [0.0144] [0.0196] [0.0195] [0.00514] [0.0112] Year Fixed Effects      Industry Fixed Effects      District Fixed Effects      Observations 1,805 1,152 293 2,776 935 R-squared 0.400 0.466 0.901 0.673 0.737 The dependent variable is the log of count of new firms per capita. Numbers in brackets are standard errors, clustered at the district level *** p<0.01, ** p<0.05, * p<0.1 Broken down by firm size, the results for input and output economies buttress those observed for the general sample – the effect of these is positive and significant for firm births in both the formal and the informal sectors. The results suggest that larger enterprises are affected 24 more strongly by the clustering of buyers and suppliers in related industries. This is in contrast to a study on Italian firms carried out by Lafourcade and Mion (2003) who find that small firms (i.e. less than 20 employees) are more sensitive to input-output linkages than large firms. 6 Conclusion This paper seeks answers to the following questions: What is the extent of coagglomeration between formal and informal firms in India and what determines these trends; and to what extent does formal-informal coagglomeration, relative to traditional measures of agglomeration, have an effect on firm births within a location? The paper has two important findings: (1) informal-formal coagglomeration is explained by linkages between buyers and suppliers, and, (2) formal-informal coagglomeration has a positive effect on the entry of formal SMEs in India. The empirical analysis finds that buyer- supplier linkages between formal and informal firms within the same industry and technological spillovers are the most significant factors that explain coagglomeration. The effect of coagglomeration is not overwhelming in the face of traditional forms of clustering – such as intra-industry clustering (that matters most for new informal activity), or inter- industry clustering (that matters strongly for new formal and informal activity). However, coagglomeration does have a positive effect on one subset of firms – it positively affects births of SMEs in the formal sector. This could be a factor of labor market regulations in the formal sector, or simply because these firms find it cost-effective for other reasons to outsource production or labor. The research also finds that formal and informal firms in India behave differently with regard to their response to agglomeration, market access and the extent of consumption. Informal firms are attracted to firms in similar industries, formal firms to districts with low levels of industrial diversity. Informal firms tend to start shop in more populated districts, while formal firms do so in districts that are closer to other larger districts. The importance of this research is underscored by two inter-related factors. First, as Duranton (2008) points out, much more evidence is needed on informal manufacturing activity and its relationship to production externalities in order to design policies that would encourage 25 economic activity in developing countries. This paper fills in part of the gap in the literature by providing an explanation for factors that encourage informal activity and to what extent this is related to linkages with formal firms. Second, this paper illustrates how formal- informal interactions can lead to positive spillovers for the formal SME sector. These findings have important implications for policy. The existence of positive production externalities indicate that policies that try to limit the development of informal activities and attempts to shift the balance to formal activities are fraught with danger and that it might be efficient to allow these clusters to grow. If governments care about the performance of small and medium sized-enterprises, they need to pay attention to the performance of informal firms, firms that provide crucial inputs to the formal sector. Processes of labor and product outsourcing have created new possibilities for informal actors to be integrated in national and international value chains – studies have shown how integration of informal Guatemalan handicraft producers (Dunn and Villeda 2005) or Tanzanian furniture producers (Murphy 2007) could facilitate skills upgrading, job creation and technical innovation for both formal and informal actors. Policy makers should think about supporting certain types of informal firms, not just those that are at the cusp of formality, but also those that are inextricably linked to the formal sector. There are several directions for further research. A valuable contribution would be to extend the analysis by studying the performance of new entrants over time and in the presence of agglomeration economies, since ultimately policy-makers would be interested in encouraging industrial activity that helps to sustain growth (whether in terms of productivity or jobs). Another interesting question would be the effect of formal-informal linkages on the productivity of the latter, and whether supplying to better or larger firms helps smaller or less formal firms move up the productivity ladder. And lastly, the government has used a number of tools targeting those within the informal sector. For instance, special health insurance schemes have been set up, social security measures were put into place and funds have been made available for technical, marketing and credit facilities to households, workers and firms in the informal sector. Understanding the impact of these interventions on the performance of not just targeted firms, but also downstream formal firms, would provide handy evidence on the effectiveness of existing policies. 26 References Arimah, B.C. (2001) ‘Nature and determinants of linkages between informal and formal sector enterprises in Nigeria’, African Development Review, 13(1): 114-144. Bacolod, M., Blum, B. and Strange, W.C. (2008) ‘Skills and the city’, Journal of Urban Economics 65:127-135. Bacolod, M., Blum, B. and Strange, W.C. (2009) ‘Urban interactions: soft skills vs. specialization’, Journal of Economic Geography 9:227-262. Barrios S., Bertinelli., and Strobl E. (2006) ‘Coagglomeration and growth’, Regional Science and Urban Economics 36:467–481. Carr, M., Chen, M.A. and Tate J. (2000) ‘Globalisation and Home-Based Workers’, Feminist Economics 6(3):123–142. Dunn, E. and Villeda. L. 0(2005) ‘Weaving Micro and Small enterprises into Global Value Chains’, USAID Advanced Microenterprise Advancement Program. Duranton, G. (2008) ‘Cities: Engines of Growth and Prosperity for Developing Countries?’, Commission on Growth and Development Working Paper 12, World Bank. Duranton, G. and Overman, H.G. (2005) ‘Testing for localization using micro-geographic data’, Review of Economic Studies, 72(4): 1077-1006. Dutta, N., Kar, S. and Roy, S. (2011) ‘Informal sector and corruption: An empirical investigation for India’, Discussion Paper No. 5579, IZA, Bonn, Germany. Ellison, G. and Glaeser, G.L. (1997) ‘Geographic Concentration of U.S. Manufacturing Industries: A Dartboard Approach’, Journal of Political Economy, 105 (5): 889–927. Ellison, G., Glaeser, E.L. and Kerr, W.R. (2010) ‘What causes industry agglomeration? Evidence from coagglomeration patterns’, American Economic Review 100: 1195-1213. Gabe, T.M. and Abel, J.R. (2013) ‘Shared knowledge and coagglomeration of occupations’, federal Reserve Bank of New York Staff Report No. 612. Ghani, E., Kerr, W. and O’Connell, S.D. (2011a) ‘Spatial determinants of entrepreneurship in India’, NBER Working Paper 17514. Ghani, E., Kerr, W. and O’Connell, S.D. (2011b) ‘Local industrial structures and female entrepreneurship in India’, NBER Working Paper 17596. Henderson, J.V. (2003) ‘The urbanization process and economic growth: the so-what question’, Journal of Economic Growth, 8: 47-71. Holl, A. (2008) ‘Production subcontracting and location’, Regional Science and Urban Economics 38: 299-309. 27 Jütting J and Laiglesia JR De (2009) ‘Is Informal Normal? Towards More and Better Jobs in Developing Countries’, Paris: OECD Development Centre. Lafourcade, M. and Mion, G. (2003) ‘Concentration, spatial clustering and the size of plants: disentangling the sources of co-location externalities’, CORE Discussion Paper No. 2003/91. Lall, S.V., Koo, J. and Chakravorty, S. (2003) ‘Diversity Matters: The Economic Geography of Industrial Location in India’, Policy Research Working Paper 3072, World Bank Lall, S.V. and Chakravorty, S. (2005) ‘Industrial Location and spatial inequality: Theory and Evidence from India’, Review of Development Economics, 9 (1): 47-68. Lall, S.V., Shalizi, Z. and Deichmann, U. (2004) ‘Agglomeration economies and productivity in Indian industry’, Journal of Development Economics, 73 (2): 643-673. Lall, S.V. and Mengistae, T. (2005) ‘The impact of business environment and economic geography on plant-level productivity: an analysis of Indian industry’, Policy Research Working Paper No. 3664, World Bank Marjit, S. and Kar, S. (2009) ‘A contemporary perspective on the informal labour market: theory, policy and the Indian experience’, Economic and Political Weekly, Vol. XLIV, No. 14. Marshall, A. (1919) Industry and Trade. London: Macmillan Mitra, A. (2009) ‘Impact of trade on service sector employment in India’, Institute of Economic Growth, Working Paper Series No. E/298. Mukim, M (2011) ‘Industry and the urge to cluster: A study of the informal sector in India’, SERC Discussion Paper No. 72. Mukim, M. and Nunnenkamp, P. (2012) ‘The location choices of foreign investors: a district- level analysis in India’, The World Economy 35(7): 886-918. Murphy, J.T. (2007) ‘The Challenge of Upgrading in African Industries: Socio-Spatial Factors and the Urban Environment in Mwanza, Tanzania’, World Development 35 (10): 1754-1778. Porter, Michael E. 1990. The Competitive Advantage of Nations. New York: The Free Press. Rosenthal, S. and Strange, W. (2001) ‘The determinants of agglomeration’, Journal of Urban Economics 50: 191– 229. Venables, A.J. (1996) ‘Equilibrium locations of vertically linked industries’, International Economic Review, 49: 341-359. Wu, F.L. (1999) ‘Intrametropolitan FDI firm location in Guangzhou, China - A Poisson and negative binomial analysis’, Annals of Regional Science 33 (4): 535-555. 28