79454 AUTHOR ACCEPTED MANUSCRIPT FINAL PUBLICATION INFORMATION Competition and Demographics in Large Indian Cities The definitive version of the text was subsequently published in Journal of Development Studies, 47(9), 2011-08-05 Published by Taylor and Francis THE FINAL PUBLISHED VERSION OF THIS ARTICLE IS AVAILABLE ON THE PUBLISHER’S PLATFORM This Author Accepted Manuscript is copyrighted by the World Bank and published by Taylor and Francis. It is posted here by agreement between them. Changes resulting from the publishing process—such as editing, corrections, structural formatting, and other quality control mechanisms—may not be reflected in this version of the text. You may download, copy, and distribute this Author Accepted Manuscript for noncommercial purposes. Your license is limited by the following restrictions: (1) You may use this Author Accepted Manuscript for noncommercial purposes only under a CC BY-NC-ND 3.0 Unported license http://creativecommons.org/licenses/by-nc-nd/3.0/. (2) The integrity of the work and identification of the author, copyright owner, and publisher must be preserved in any copy. (3) You must attribute this Author Accepted Manuscript in the following format: This is an Author Accepted Manuscript of an Article by Amin, Mohammad Competition and Demographics in Large Indian Cities © World Bank, published in the Journal of Development Studies47(9) 2011-08-05 http://creativecommons.org/ licenses/by-nc-nd/3.0/ © 2013 The World Bank Competition and Demographics in Large Indian Cities Mohammad Amin* World Bank August, 2010 Abstract Recent studies suggest that consumer-household attributes may be as important in determining the level of competition in certain markets as firm characteristics and the number of firms. However, evidence on which consumer-household attributes matter for competition is limited, especially for developing countries. Focusing on India’s retail sector, the present paper contributes to this literature by showing that the number of adult non-workers per household in the city, a proxy for shopping time opportunity cost, has a strong effect on competition between retailers. Policy implications of our findings in light of the ongoing dramatic reductions in non-workers in India are discussed. Keywords: Competition, Retailing, India, Demographics JEL: D24, L11, L81 ________________________ *Enterprise Analysis Unit, World Bank, Washington DC, 20433. Email: mamin@worldbank.org. Phone: (202)-473-1915. I would like to thank Penelope J. Brook, Simeon Djankov, Alvaro Gonzalez, Ernesto Lopez-Cordova and Siddharth Sharma for helpful comments. All remaining errors are my own. 1. Introduction Mainstream economics views competition and the competitiveness of markets exclusively in terms of the number of firms and their behavior. Competition policies around the world are largely aimed at preventing collusive agreements between firms and the abuse of dominant position by a single firm. In contrast, a small but growing literature argues that in many industries, consumer attributes and consumer behavior may be equally important in determining the level of competition (Waterson, 2003). It suggests tying of competition policy to consumer attributes to keep markets truly competitive. However, this literature is still in its infancy and empirical evidence on which consumer attributes matter for competition is very limited. The present paper takes one step towards filling this gap in the literature by providing evidence on how an important consumer- household attribute, the number of adult non-workers per household, affects the level of competition in India’s retail sector. Our hypothesis is that shopping time opportunity cost is lower for households that have more adult non-workers. Therefore, such households are likely to search more intensively than others for best prices and deals, increasing the level of competition in the retail market. To test this hypothesis, we exploit variation in the number of adult non- workers per household (henceforth, non-workers) using data on retail stores spread across 41 large cities of India. Consistent with the stated hypothesis, we find a large positive effect of more non-workers on the level of competition. We discuss the policy implications of our findings which we believe are particularly important for India, given that the country is witnessing a rapid decline in the size of its adult non-working population. 1 The motivation for exploring the link between competition and non-workers is two-fold. First, it provides a better understanding of the determinants of competition in retailing. The importance of search cost for market competition has been empirically established but only for the case of developed countries. The present paper makes a first attempt at extending this literature to a developing country. Second, our findings are important for the appropriate design and targeting of competition policy. As economic development spreads to the relatively poorer countries, household structure in these countries is likely to change as it is currently happening in India. Additional policy measures that factor in the structure of households could play an important role in ensuring that retail markets remain truly competitive. Tying of competition policy to specific consumer attributes (different from ours) is also recommended by Giulietti et. al. (2005) based on their finding that pensioner households, low-income households, rural dwellers and consumers with disabilities benefited differently from the recent deregulation of the U.K. energy markets. There is no previous work on the relationship between non-workers or other socio-demographic consumer-household attributes and competition. Goldman et al. (2002) argue that the number of non-working adults in the household is a good surrogate for the household shopping time opportunity cost. However, the authors do not examine the implications of this for the level of competition in the market.1 Other studies focusing on consumer attributes show that competition in the market is significantly correlated with consumers’ past experience with switching suppliers (Giulietti et. al. 2005), commuting distance to shops (Baron et. al. 2004, Claycombe 2000), consumer perceptions on search and switching costs and whether expected benefit from comparison 2 shopping is long-term or short-term (Giulietti et. al.), competition policy and the cost of switching suppliers (Lalive and Schmutzler 2007, Giulietti et. al., Calem and Mester 1995, Knittel 1997). Waterson (2003) provides a useful summary of the related literature. There is some hint in the literature that a household’s income level may affect its search efforts (for best prices and deals) and therefore the level of competition (in the retail markets). Lower income households have higher marginal utility of income (savings from more intensive search) and also lower opportunity cost of time. Hence, they are likely to search more intensively implying a negative relationship between income and the level of competition (Marvel 1976, Hoch et. al. 1995).2 Our results below do reveal such a negative relationship but it disappears when we control for non-workers. This suggests that the income-competition relationship highlighted in the literature may be spuriously driven by the relationship between non-workers and competition. We test the hypothesis of a positive relationship between non-workers and competition using micro data on 1,948 retail stores in 41 large cities of India. These data were collected by the World Bank’s Enterprise surveys in 2006. The choice of India for the present study is important for two reasons. First, the retail and wholesale sector in India is one of the largest in the country, but a formal analysis of the sector is virtually non-existent. The sector is the second largest employer after agriculture accounting for 10 percent of all formal jobs in the country and 14 percent of the national GDP. Hence, our findings have important implications for a large section of the Indian economy. Second, India is currently witnessing a remarkable decline in the number of adult non-workers per household. For example, between 1991 and 2001 and for the fourteen major states of India, the decline averaged 7.2 percent with a high of 18.6 percent in the state of 3 Haryana, 12.7 percent in Kerala and 12.6 percent in Punjab (Figure 1).3 The downward trend is likely to continue, being part of a larger socio-economic-demographic churning in the country characterized by greater participation of men and women in the labor market, increasing proportion of the young in the working age group and lower fertility rates. Findings in the present paper help understand the likely effect of these socio- economic-demographic changes on the retail sector and also to identify the sorts of cities most in need of corrective policy measures aimed at reducing consumer search costs. We contrast the effect on competition of non-workers and the number of children per household in the city (henceforth, children) as an informal test of the mechanism through which non-workers affect the level of competition. The test is based on two assumptions. First, more children imply higher shopping time opportunity cost for households and therefore less intensive search and less competition in the retail markets. This is opposite to the predicted effect of more non-workers on competition. Second, aspects of overall development and income levels of cities are likely to be correlated with children and non-workers in the same direction (evidence provided below). Hence, if the relationship between non-workers and competition is spuriously driven by differences in overall development or income levels across cities then controlling for children should weaken this relationship, and non-workers and children should affect competition in the same direction. However, if our search cost based interpretation for the relationship between non-workers and competition is indeed true then children and non-workers should have opposite effects on competition and controlling for children should not weaken the estimated effect of non-workers on competition. Results presented below strongly confirm both these predictions. 4 The rest of the paper is as follows. In section 2 we describe the data and our main variables. In section 3 we present our main empirical results. Robustness checks are discussed in section 4. In section 5 we take a more detailed look at traditional vs. non- traditional retail stores and the income-competition relationship. A summary of the main findings and scope for future work are provided in the concluding section. 2. Data and Main Variables As mentioned above, we use store level data collected by the World Bank in 2006 (Enterprise surveys), complemented with external data sources discussed below. The Enterprise survey data consist of a stratified random sample of 1,948 retail stores (cross- section) operating in the formal sector and located in 16 major states and 41 cities of India.4 The National Industrial Classification (NIC-1998, Industry Division 52) classifies retailers into those operating through established stores and the rest who usually operate from home. All stores in our sample belong to the former category. A formal definition of all the variables used in the regressions is provided in Table 1. Summary statistics of the main variables and the correlation between them are provided in Tables 2 and 3, respectively.5 2.1 Dependent variable Our dependent variable attempts to capture the degree of competition faced by the retailers in the sample. Existing studies suggest a number of measures for this purpose including concentration ratio or Herfindahl index, price cost margin (Lerner index), a count of the number of competitors faced by a given firm, and specific incidents of 5 privatization and liberalization that lead to more competition in the market. However, these measures are not free of conceptual and data related problems. For example, concentration ratios or the number of competitors do not capture the contestability of markets. Hence, it is entirely plausible that competition in the market may be greater with just two firms (Bertrand competition) than with more firms (Cournot competition). Increased competition from specific incidents of liberalization is irrelevant for our purpose because such incidents are not an outcome of our variable of interest, non- workers. Hence, there is no reason to expect any systematic relationship between non- workers and increased competition from policy shocks. The Lerner index does not suffer from these problems but data limitations do not allow us to use this index.6 While the difficulty in obtaining reliable estimates of marginal cost required for computing the Lerner index is not specific to our paper, it is aggravated by our focus on the retail sector in a developing country. In particular, quality of retailing service, location related benefit to consumers and the marginal cost of maintaining inventories are extremely hard to measure (Baily and Solow, 2001). In this paper, we take a novel approach towards measuring competition. To this end, we use an experienced-based measure as reported by the retailers in the survey on how important competition is for the prices of the main products sold by their stores. Specifically, the Enterprise survey asked store managers how important the pressure from domestic competitors is over prices of the store’s main products.7 The response was recorded on a 1-4 scale defined as not at all important (1), slightly important (2), fairly important (3) and important (4). We use these self-reports or responses of the stores on the 1-4 scale as our dependent variable, Competition. We note that the dependent variable 6 is defined at the store level and higher values of the variable imply greater competition as experienced by the stores. The mean value of Competition equals 2.2 and the standard deviation is 1.12. Looking at the full distribution, about 35 percent of the stores reported competition as not all important (Competition equal to 1), 27 percent as slightly important, 19 percent as fairly important and the rest 19 percent as important (Competition equal to 4). The approach of using firms’ experience-based measures and experts’ perception to construct economic measures is becoming increasingly popular in the literature. For example, Safavian and Sharma (2007) use firms’ self-reports (similar to the one above) on the efficiency of courts to estimate the differential effect of creditor laws on credit supply across countries with good and bad quality of courts. One concern with such self- reports is that the reference point or the definition of what is, for example, an “important” and “fairly important” effect of competition on prices could vary across stores depending on their characteristics such as size, age and location. This noise in the dependent variable could lead to an omitted variable bias problem with our main results if systematic variations in the reference point mentioned above are also correlated with our main explanatory variables. In the sections that follow, we outline a number of reasons why the potential omitted variable bias referred to above is unlikely to be much of a problem for our main results. For example, we show that the correlation between our main explanatory variable, non-workers, and various store characteristics is extremely small, reducing the possibility of any large omitted variable bias. Our main results easily survive controls for various store characteristics. The same holds for various city characteristics. 7 Additionally, we explore the relationship between our competition variable and factors at the city level such as the number of children per household, extent of power outages faced by an individual retailer vs. the rest of the retailers in the city, retailer density and the income level of consumers. We show that all these relationships make intuitive sense, an unlikely result if the measurement error or the level of noise in our competition variable were too severe. We believe that our results for the comparison between the effects of children vs. non-workers on competition is particularly revealing as far as the omitted variable bias problem is concerned. This issue is discussed in greater detail below. Another concern with the dependent variable as defined above could be that it relates to price-competition alone and therefore may not adequately capture the broader competitive environment. For example, pricing restrictions for certain products (by law) may blunt price-competition, but stores may still compete with each other for the precious few buyers by providing a greater range of products and better quality of service. While this problem cannot be ruled out completely, we provide some evidence which suggests that it is unlikely to be severe. Specifically, in another survey question, stores were asked how important the influence of domestic competitors is in their decision to introduce new product lines. Responses were recorded on the same 1-4 scale as above. The correlation between the response of stores on this question and the one on price-competition above equals 0.738. The high correlation is reassuring in that it suggests that our measure of competition captures the broader competitive environment rather than the narrow specifics of price setting. 8 2.2 Explanatory variables 2.2.1 Non-workers Our main explanatory variable is the number of adult non-workers per household in the city. The variable equals total number of adult non-workers in the city divided by the total number of households in the city (Non workers). We use (lagged) 1991 values of the variable taken from Census of India (1991).8 Shopping for many products such as grocery items is typically done for the entire household which motivates our preference for non- workers per household over, for example, non-workers as a proportion of city population.9 We follow the Census definition of an adult and a household. The former is defined as all agents above seven years of age, and the latter as a set of individuals living in a common house and sharing a common kitchen.10 For the cities in our sample, Non workers varies between 2.01 (Noida) and 3.88 (Patna) with a mean value of 2.9 and a standard deviation of 0.38. 2.2.2 Motivation for other controls Reverse causality from the level of competition to non-workers is unlikely given that the latter is lagged by 15 years. A relatively more serious problem with our results could be a bias due to omitted variables. Broadly, there are four potential sources of such a bias. First, job opportunities are likely to be fewer in the relatively poorer cities. Hence, we predict a negative correlation between income levels and non-workers. As suggested in the literature, higher income households tend to search less intensively for best prices since they have lower marginal utility of income (cost savings from intensive search). This implies a negative relationship between income level and competition across cities. 9 The predicted negative relationship between income and non-workers and between income and competition suggests that failure to control for income levels across cities could bias the estimated effect of non-workers on competition in the upward direction. Second, aspects of less development such as frequent power outages, poorer roads and inadequate access to credit are likely to have a negative effect on the competitiveness of markets. Also, less developed cities have few job opportunities and therefore more non-workers. The structure of correlations here suggests a downward bias in the estimated effect of non-workers on competition from the failure to adequately control for the stated aspects of overall development. Third, firm characteristics may vary systematically with both, the level of competition and non-workers. Although it is difficult to sign the direction of the implied omitted variable bias, below we provide evidence that suggests that this source of bias is very small in magnitude and therefore virtually irrelevant for the present study. Last, as discussed above, self-reports of retailers on the importance of competition for the prices of their main products could differ across stores due to differences in the reference point or the definition of what is “important” and “fairly important”. While we do not suspect any direct effect of non-workers on the reference point, however, it is possible that factors such as firm and city characteristics that lead to differences in reference point may also be spuriously correlated with non-workers. We provide detailed evidence below to show that this form of omitted variable bias is unlikely to be severe for our main results. 2.2.3 Number of children per household 10 To address the potential omitted variable bias problem discussed above, we begin by contrasting the effect on competition of non-workers and the number of children per household in the city (Children). We use (lagged) 1991 values of Children taken from Census of India (1991). The motivation is that Non-workers and Children are likely to be correlated with overall development in the same direction (negatively). Table 3 clearly confirms that this is indeed the case. Hence, if the relationship between competition and non-workers is primarily driven by differences in the level of overall development across cities then we can expect both, Non-workers and Children to affect competition in the same direction. In contrast, search theory predicts that children and non-workers should have opposite effects on competition. More children imply higher shopping time opportunity cost and therefore less intensive search and lower competition. The prediction for non-workers is just the opposite. The contrasting predicted effects of non- workers and children on the dependent variable despite the fact that the two are positively correlated provides a useful test for the search cost based interpretation of the relationship between competition, non-workers and children. 2.2.4 Other controls Our second defense against the omitted variable bias problem is to show that our results are robust to a number of controls for city and store characteristics. Based on the discussion above, in our main specification we control for a proxy measure of household income which is the mean per capita expenditure of the district population (Expenditure). The data source for the variable is National Sample Survey Organization (NSSO, 50th round, 1991-92). The NSSO routinely conducts surveys of household expenditure levels 11 that are available at the district level.11 Reliable estimates of GDP at the city or district level are not available for India. Next, we control for two proxy measures of overall development of cities that include total adult population of cities (Population) and a dummy variable, Metro, that equals 1 if a store is located in a metropolitan city (Bangalore, Chennai, Hyderabad, Kolkata, Mumbai and New Delhi) and 0 otherwise. As with Non-workers and Expenditure, we use lagged (1991) values of Population taken from Census of India (1991).12 Larger cities (in terms of population) and the metropolitan cities are known to be more developed and richer and also the main beneficiaries of the ongoing retail boom in the country. In the robustness section, we include a number of additional controls in our specification such as literacy rate, sex ratio, number of workers per household, store-size, duration of power outage faced by stores, number of retailers in the city, etc. 3. Estimation All estimation results discussed below are obtained using an ordered logit specification with Huber-White robust standard errors clustered on the city.13 Without much loss of generality, we report the marginal effects for the highest value of the dependent variable (Competition equal to 4). That is, the regression tables discussed below show the impact, evaluated at the margin, of a unit increase in the various explanatory variables on the probability that Competition takes its highest value of 4 versus a lower value.14 All marginal effects are evaluated at the mean value of the explanatory variables. Regression results for our benchmark or the main specification are provided in Table 4. Without any additional controls, a unit increase in Non workers raises the 12 probability of a store facing the highest level of competition by 8.9 percentage points, significant at close to the 1 percent level (column 1). The estimate implies that moving from the lowest (city of Noida) to the 25th percentile value (city of Bhubaneswar) of Non workers raises the probability of a store facing the highest level of competition by 5.2 percentage points, a large effect given that less than 19 percent of the stores in the sample face the highest level of competition. Alternatively, a one standard deviation increase in Non workers raises the stated probability by 3.4 percentage points. As the later results reveal, these estimates of the positive effect of more non-workers on competition are on the conservative side. In columns 2 and 3 (of Table 4) we report the independent effects of children and expenditure on competition. As expected, both these effects are negative but neither of them is significant at the 10 percent level. Earlier we had argued that if our search based interpretation for the relationship between competition and non-workers is indeed correct then non-workers and children should have opposite effects on competition even though they are positively correlated with one another. This also implies that failure to control for either of them would bias the estimated coefficient of the other towards zero. Both these predictions are confirmed in column 4 where we control for Non workers and Children simultaneously. The estimated marginal effect of Non workers rises sharply from 0.089 (column 1) above to 0.133 (column 4) and is significant at less than the 1 percent level. The estimated marginal effect of Children also increases (in absolute value) from -0.076 (column 2) to -0.227 (column 4) and is significant at less than the 10 percent level (p-value of 0.085). 13 The contrasting effects of non-workers and children on competition along predicted lines discussed above serve to increase our confidence against the omitted variable bias problem and in favor of the search cost based interpretation of the non- workers and competition relationship. This is further corroborated when we control for other aspects of overall development through population and the metropolitan city dummy. Controlling for these variables causes the estimated marginal effect of non- workers (on the probability that the store faces the highest level of competition) to further increase from 0.133 above to 0.170 (column 5), significant at less than the 1 percent level. Similarly, the estimated marginal effect of children also increases in absolute value from -0.227 above to -0.346 and is now significant at less than the 5 percent level (column 5). Individually, none of the measures of overall development in the specification above show any statistically significant impact on the level of competition. City population and the metropolitan city dummy show a weak negative association with the level of competition, while the expenditure measure shows a positive association (column 5). We discuss these results in detail in section 5.2. 4. Robustness It is entirely plausible that the various proxy measures of overall development mentioned above may not capture all possible aspects of economic development that are correlated with both non-workers and the level of competition. If this were true, the results discussed above could still suffer from the omitted variable bias problem. Similarly, the observed positive relationship between non-workers and competition above may also be 14 spuriously caused by differences in store characteristics and in the structure of the retailing activity across the sampled cities. Below, we provide a number of robustness checks to raise our confidence against such problems. Broadly, these checks include additional controls for the level of overall development of cities and the structure of retail stores. Results from these robustness checks are provided in Table 5. We begin by controlling for two additional popularly used measures of overall development at the city level. These measures include adult literacy rate (Literacy) and the ratio of females to males (Sex ratio). Both these variables are taken from Census of India (1991 values).15 We would like to mention that some care is needed in interpreting what sex ratio proxies for as far as the level of competition in retailing is concerned. In addition to higher income levels, higher values of sex ratio are typically associated with greater gender parity (favoring women). It is possible that with greater gender parity, women may be less involved in traditional roles including household shopping, implying a negative relationship between sex ratio and the level of competition. Our empirical results discussed below do not reject this hypothesis. In the Enterprise Survey, stores singled out inadequate power supply and poor access to finance as the two biggest obstacles to doing business. Differences in power supply and access to finance across cities could have a large effect on the structure of the retail sector. We control for these potential differences from affecting our results using the proportion of stores in the city that have a checking or savings account (Checking) and the proportion of stores in the city that have either a line of credit or overdraft facility (Financial Access).16 For power supply, we control for two variables: the duration of power outage on average per day in 2005-06 as reported by the stores (Outage – own 15 store) and the duration of power outages per day on average reported by all other stores in the city (Outage). The motivation for the two power outage variables is that power outages in one’s own store are likely to deflect customers to the neighboring stores implying greater competitive pressure for the store in question. By the same logic, the opposite holds for power outages in the other (neighboring) stores. Data source for all these variables is Enterprise Surveys. It is natural to expect that the density of retail shops in a given city may affect the level of competition. Similarly, more stringent business regulations may lower the threat of new entry, lowering the level of competition. We control for these factors using two proxy measures that include total employment in retail and distribution as a proportion of total adult population defined at the city level (Retailer density), and a measure of business regulations following Sevensson and Fisman (2007) defined as the percentage of store’s senior management’s time spent in dealing with business regulations and averaged at the city level (Regulation). Data source for retailer density is Census of India (1991) and is the closest available proxy measure of the density of retail shops at the city or state level. Data source for Regulation is Enterprise Surveys. Our final robustness check consists of controlling for basic store characteristics including (log of) floor area of the shop (Size)17, age of the store (Age), a dummy variable of the 1 if a store is part of a larger chain and 0 otherwise (Chain), percentage of the firm (store) held by the largest shareholder (Ownership concentration) and store-type fixed effects.18 The survey classifies all stores into traditional stores (selling grocery items), consumer durable stores (selling consumer durables) and modern format stores (large stores part of a shopping complex). Store-type fixed effects are dummy variables that 16 capture factors common to all stores within each of these categories (a formal definition of these fixed effects is provided in Table 1). Regression results provided in columns 1-4 of Table 5 clearly show that the various robustness controls mentioned above do not change our main results much from above. Specifically, the estimated marginal effect of non-workers on the probability of a store facing the highest level of competition declines somewhat from -0.170 (column 5, Table 4) to -0.144 (column 4, Table 5) due to the controls listed above and remains significant at less than the 1 percent level. As above, the corresponding marginal effect of children remains negative and significant at less than the 5 percent level, decreasing slightly in magnitude from -0.346 (column 5, Table 4) to -0.302 (column 4, Table 5). As predicted, the two power outage variables have contrasting effects on the dependent variable and they are both significant at less than the 1 percent level. The estimated effect of retailer density is also along predicted lines (positive), significant at close to the 5 percent level and quite large in magnitude. For example, an increase in retailer density from its 25th to the 75th percentile value increases the probability of a store facing the highest level of competition by a large 6.3 percentage points. More burdensome regulation has a negative effect on the probability of a store facing the highest level of competition as we might expect, although this effect is statistically significant only between 10-15 percent levels. For the remaining variables, higher values of sex ratio and larger size of stores is associated with significantly lower probability of a store facing the highest level of competition, although the size-competition relationship becomes insignificant once of we control for the store-type fixed effect (column 3 vs. column 4, Table 5).19 For store-type 17 fixed effects, we find that, relative to consumer durable stores, the probability that a traditional store faces the highest level of competition is lower by 6.2 percentage points and this difference is significant at less than 1 percent level (not shown). We do not find any significant difference between consumer durable and modern format stores in the level of competition they face. The estimated marginal effect of age and being part of a larger chain is weak and statistically insignificant, while the same for more concentrated ownership is negative and significant at less than the 10 percent level. 4.1 Other robustness checks We performed a number of additional robustness checks and found that our main results discussed above remained intact. We provide a brief outline of these checks. About 30 percent of the stores in our sample are located in the metropolitan cities. Since the structure of retailing could be different across metropolitan and the remaining cities, we dropped all the metropolitan cities from the sample to check if our results hold for the majority of cities in the country. Regression results for the sub-sample of non- metropolitan cities are provided in column 5 of Table 5 and these are qualitatively similar to the ones discussed above. Next, we controlled for additional store characteristics which include years of store manager’s experience in retailing, initial level of employment when the store first started operations, number of days of inventory maintained by the store, percentage of a store’s annual sales (in 2005-06) that were never paid for, losses due to theft (as percentage of sales in 2005-06) averaged at the city level, a measure of crime at the city level derived from store’s perception of crime as an obstacle to their business, number of 18 workers per household (1991 values, at the city level) and measures of vehicle availability at the city level which include the proportion of households that have a four- wheeler (car, jeep, van, etc) and the proportion of households that have a two-wheeler (scooter, motor cycle, moped, etc).20 The relationship between non-workers and competition discussed above easily survived all these controls. Summarizing, our empirical results show a large positive effect of more non- workers on the level of competition. According to our most conservative estimate (column 1, Table 4), a move from the lowest to the 25th percentile value of non-workers increases the proportion of retail stores in the city that face the highest level of competition from other retailers by 5.1 percentage points, a large effect given that only 18.6 percent of the stores in the full sample face the highest level of competition. The finding is robust to a number of controls for city and store characteristics such as income levels, density of retailers, infrastructure availability, store-size, age and the type of products sold by the stores. 5. Extensions 5.1 Traditional vs. non-Traditional stores It is well known that the bulk of the retailing activity in India is done by the small and traditional stores selling grocery items. Traditional stores may also involve more frequent visits or repeat purchases, suggesting that the non-workers and competition relationship discussed above may be stronger for traditional than the non-traditional stores.21 Hence, it is important to check if our results discussed above hold for traditional stores and how they vary across traditional and non-traditional stores. 19 To this end, we use classification of stores into traditional stores (64.3 percent of the sample) and non-traditional stores is as provided by the Enterprise survey (see, Table 1).22 Regression results for traditional stores and the rest are provided separately in Table 6. These results show that the relationship between competition and non-workers holds for traditional as well as non-traditional stores. The relationship is slightly stronger for the traditional stores, although not significantly so.23 5.2 Income-competition relationship In this section we focus on why our results fail to show any significant relationship between expenditure and competition. This is important because our expenditure variable is a proxy measure for per capita income of cities in the sample. Our results for non- workers discussed above could be biased upwards if differences in income levels across cities are not adequately controlled for by the expenditure and other related variables. There are three possible reasons for the weak expenditure-competition relationship that we find in our regressions. First, the correlation between expenditure and population is 0.548 and 0.466 between expenditure and the dummy for the metropolitan cities. The potential multicollinearity problem is evident here. However, this is at most a contributory factor since expenditure continues to show a weak effect on the dependent variable even if when we do not control for population and the metropolitan dummy (shown below). Second, cities with higher levels of expenditure in our sample have fewer children per household, less burdensome regulations and fewer power outages.24 Note that all these covariates of expenditure tend to increase the level of competition in cities with higher expenditure levels, diluting an otherwise negative expenditure-competition 20 relationship. However, this reason also does not explain why the expenditure-competition relationship is weak since we controlled for all these covariates above. This brings us to our last point. That is, expenditure or income levels do not have a direct effect on competition and they are at best proxy measures for non-workers. Regression results provided in Table 7 are consistent with the reasons discussed in the previous paragraph. We exclude population and the dummy for the metropolitan city from all the specifications in the Table to avoid the multicollinearity problem mentioned above. The best possible scenario for the expenditure-competition relationship is provided in columns 1-3, where we do not control for non-workers. In all these specifications, higher expenditure is associated with a lower probability of a store facing the highest level of competition and this relationship is significant at less than the 5 percent level. However, controlling for non-workers clearly destroys this strong and negative relationship (columns 4-6). We conclude that existing studies showing a negative income-competition nexus could be spuriously picking up the effect of non- workers on competition. More work is needed to ascertain or reject this claim. 6. Conclusion The present paper contributes to a small but growing literature that links the level of competition in product markets to consumer-household attributes that shape consumer shopping behavior. This literature also suggests tying of competition policy to consumer- household attributes to ensure that markets remain truly competitive. This is in sharp contrast to the traditional view of competition as an outcome of firm behavior and 21 competition policies exclusively focused on preventing collusive agreements between firms and the abuse of dominant position by a single firm. Our results show an economically large effect on competition of an important consumer-household attribute, the number of adult non-workers per household. The suggested underlying mechanism is that more non-workers lower a household’s shopping time opportunity cost, leading to a more intensive search for best prices and deals and therefore greater competition in retailing. We contrasted the effects of non-workers and children on competition as an informal test of the stated mechanism. Policy implications of our findings are important, especially because India is currently witnessing a rapid decline in the number of non-workers. Cities that currently have fewer non-workers and those witnessing largest declines in non-workers are most in need of pro-competitive reforms aimed at reducing consumer search costs. For example, one policy reform could be facilitating the spread of internet and internet based shopping. The present paper suggests exciting opportunities for future research. A natural first step would be look at other consumer-household attributes that may be important for the level of competition in retailing. For example, our results show that the gender composition of cities is highly correlated with the level of competition. More work is needed to understand why this is so. We hope that the present paper will motivate future research for a better understanding of the competitiveness of retailing and other markets in developing countries. 22 Notes 1 Goldman et al. estimate how the number of adult non-workers in the household affects household’s choice of shopping at wet markets relative to superstores and conventional supermarkets. 2 The caveat to this is that poorer households consume less and this could lower their propensity to search for best prices if there is a fixed cost of searching. 3 These figures are computed from Census of India (1991, 2001). For more details, see Figure 1. 4 The sampling frame for the survey was the list of retail stores regularly interviewed by AC Nielson for inventory verification on behalf of distributors of branded goods. This list covers stores in 41 cities across India for three major industry segments: Fast Moving Consumer Goods (FMCG) stores (traditional stores), consumer durable stores and the modern format stores. A definition of these industry segments is provided in Table 1. The sample was stratified according to segment-specific criteria. FMCG stores were stratified based on turnover, number of salesmen, number of FMCG product and the presence of cooling equipment. Consumer durable and modern format stores were stratified based on turnover. The sample size was determined so as to minimize the standard error in the sample variables, given the available resources for each surveying stratum. Once the sample size was determined, the sample was allocated to strata using Neymann’s allocation rule. More information about the survey and methodology is available at www.enterprisesurveys.org. 5 For the regressions reported below, we leave out the city of Kozhikode from the sample (1.8% of the sample) because it is a gross outlier in most specifications. Including the city in the sample does not have much impact on the sign or the significance level of estimated coefficient of non- workers (or other important explanatory variables) although the magnitude is lower. The remaining cities included in the survey include (in alphabetical order): Ahmedabad, Bangalore, Bhopal, Bhubaneswar, Chandigarh, Chennai, Coimbatore, Cuttack, Delhi, Dhanbad, Faridabad, Ghaziabad, Greater Mumbai, Guntur, Gurgaon, Gwalior, Hubli-Dharwad, Hyderabad, Indore, Jaipur, Jamshedpur, Kanpur, Kochi, Kolkata, Kota, Lucknow, Ludhiana, Madurai, Mangalore, Mysore, Nagpur, Nashik, Noida, Patna, Pune, Surat, Vadodara, Vijayawada and Vishakhapatnam. 6 Information on the prices of the stores’ product(s) is not available in the Enterprise survey. The same holds for various elements of cost such as purchase of (intermediate) goods, inventory maintenance, etc. 7 There is virtually no competition from foreign retailers in India. 8 Data on non-workers is available every ten years with the latest year being 2001. Our results do not change much if we use 2001 values of the variable. The correlation coefficient between 1991 and 2001 values of non-workers per household is 0.875. 9 For the cities in our sample, the correlation coefficient between Non workers as defined above and non-workers as a proportion of total city population equals 0.768. The high correlation between the two variables makes it difficult to infer whether the distribution of non-workers across households within cities (as opposed to the distribution of non-workers across cities) matters for the level of competition. However, in some specifications where our results are particularly strong such as the sample of traditional stores (selling grocery items), the effect of Non workers on competition remains significant even if we control for the proportion of non- workers in total city population. 10 Data on non-workers by other age groups are not available. 11 Districts are bigger than the cities. 12 Our results do not change much if we use year 2001 values of Expenditure and Population. The correlation between 1991 and 2001 values of Expenditure equals 0.808 and 0.901 for Population. 13 Our main results are slightly stronger if we do not adjust standard errors for clustering. 14 Marginal effects for the remaining values of the dependent variable are available on request from the author. These results are consistent with the broad findings discussed in the paper. For example, an increase in the non-workers lowers the probability of the dependent variable taking 23 its least value (Competition equal to 1). This effect is statistically significant at less than the 5 percent level, with our without the various controls discussed above. 15 We use 1991 values because data for non-workers and children are also for the same year. However, 1991 and 2001 values of these variables are almost perfectly correlated (correlation of over .90) and our results do not change much if we use 2001 values instead. 16 The survey also provides information on whether a store has a checking/savings account and a line of credit. Our results do not change much if we control for these measures also (discussed towards the end of the section). 17 Our results do not change much if we use labor employment or annual store sales as measures of store-size. 18 Our motivation in controlling for ownership concentration is that it could affect store efficiency which may in turn determine the competitive pressure faced by a store. Excluding the variable form the regressions does not change any of our results significantly. 19 Size varies sharply across store-types. The average value of Size for traditional stores equals 4.78 (traditional stores), 5.34 (consumer durable stores) and 7.03 (modern format stores). 20 Data on two-wheelers and four-wheelers was first collected in the 2001 Census and it is available at the district level. We use these data for the urban part of the district population. 21 There is some evidence to suggest that the elements of search cost (non-workers, in our case) may be more important for the level of competition for products that require more frequent purchases or trips to the retail store (see, for example, Sorensen 2000). However, given the limited nature of research on this issue, more work is needed to ascertain or reject its validity. 22 See, variables “Traditional store” and “Modern Format store” in Table 1. 23 To formally check for the difference in the effect of non-workers on competition across traditional stores vs. the rest, we interacted non-workers with the dummy for traditional stores. However, the estimated coefficient of this interaction term was statistically insignificant at the 10 percent level or less. This finding was robust to control for other interaction between the traditional dummy and children, expenditure, etc. 24 For the cities in our sample, the correlation between Expenditure and Children, Outage and Regulation equals, respectively, -0.290, -0.065 and -0.143. 24 References [1] Ausubel, Lawrence M. (1991), “The Failure of Competition in Credit Market,” American Economic Review, 81(1), pp. 50-81. [2] Baily, Martin N. and Robert Solow (2001), “International Productivity Comparis ons Built from the Firm Level,” Journal of Economic Perspectives, 15(3), pp. 151-172. [3] Baron, Jon M., Beck A. Taylor and J. R. Umbeck (2004), “Number of Sellers, Average Prices and Price Dispersion,” International Journal of Industrial Organization, 22, pp. 1041-1066. [4] Calem, P.S. and L.J. Mester (1995), “Consumer Behavior and the Stickiness of Credit -Card Interest Rates,” American Economic Review, 85(5), pp. 1327-1336. [5] Claycombe, Richard J. (2000), “The Effects of Market Structure on Prices of Clothing and Household Furnishings,” International Journal of Industrial Organization, 28, pp. 827-841. [6] Giulietti, Monica, Catherine W. Price and Michael Waterson (2005), “Consumer Choice and Competition Policy: A Study of UK Entergy Markets,” Economic Journal, 115, pp. 949-68. [7] Goldman, Arieh, S. Ramaswami and Robert E. Krider (2002), “Barriers to the Advancement of Modern Food Retail Formats: Theory and Measurement,” Journal of Retailing, 78, pp. 281-95. [8] Hoch, S., B. Kim, A. Montgomery and P. Rossi (1995), “Determinants of Store-Level Price Elasticity,” Journal of Marketing Research, 32(1), pp. 17-29. [9] Knittel, C. (1997), “Interstate Long Distance Rates: Search Costs, Switching Costs and Market Power,” Review of Industrial Organization, 12(4), pp. 519-536. [10] Lalive, Rafael and Armin Schmutzler (2007), “Exploring the Effects of Competition for Railway Markets,” International Journal of Industrial Organization, forthcoming. [11] Marvel, Howard P. (1976), “The Economics of Information and Retail Gasoline Price Behavior: An Empirical Analysis,” Journal of Political Economy, 84, pp. 1003-59. [12] Pierre, Gaelle and Stefano Scarpetta (2006), “Employment Protection: Do Firm’s Perceptions Match with Legislation?” Economics Letters, 90, pp. 328-334. [13] Prendergast, C. (2002), “Consumers and Agency Problems,” Economic Journal, 112 (478), pp. C34-51. [14] Safavian, Mehnaz and Siddharth Sharma (2007), “When Do Creditor Rights Work?” Journal of Comparative Economics, 35 (3), 484-508. [15] Sorenson, Alan T. (2000), “Equilibrium Price Dispersion in Retail Markets for Prescription Drugs,” Journal of Political Economy, 108 (4), pp. 833-850. [16] Svensson, Jakob and Ray Fisman (2007), “Are Corruption and Taxation Really Harmful to Growth? Firm Level Evidence,” Journal of Development Economics, 83 (1), pp. 63-75. [17] Waterson, M. (2003), “The Role of Consumers in Competition and Competition Policy,” International Journal of Industrial Organization, 21(2), pp. 129-50. 25 Percentage change over 1991-2001 in the number of adult non-workers per household 10.00 5.00 0.00 Utta Mad Mah Raj And Tam G uja O ri s Kar We Pun Ker Al l s Bi h Har asth nata st al a ar yan r Pr hra sa jab tate h ya a ra il N rat Be n -5.00 s htr an a ade adu Pra ka s (a P ra gal des sh a vera d es h h ge) -10.00 -15.00 -20.00 Figure 1 Source: Census of India, 1991 and 2001. Percentage changes in the figure above equal the number of adult non-workers per household in 2001 minus the same in 1991 and expressed as a percentage of the 1991 values of the variable. 26 Table 1: Description of Main Variables Variable Description “Last fiscal year” below means fiscal year 2005-06. Competition Response of retail stores to the following question asked in the survey: For this store, how important are each of the following influences over prices of its main products? a. Pressure/Influence from domestic competitors Not at all important (1), Slightly important (2), Fairly important (3) and Important (4). Source: World Bank Enterprise Surveys (www.enterprisesurveys.org) Non workers Total adult non-workers in the city divided by total number of households in the city (1991 values). Source: Census of India (1991). Children Total number of children (below 7 years) divided by the total number of households in the city in 1991. Source: Census of India (1991) Expenditure Per capita expenditure in the urban part of the district (in thousand Indian Rupees) where the cities in our sample are located. Source: National Sample Survey Organization (50th Round, 1991-92), Government of India. Population (in millions) Total (adult) population of the city in 1991. Source: Census of India (1991). Metro A dummy variable equal to 1 for a store located in a metropolitan city (Delhi, Mumbai, Kolkata, Chennai, Bangalore and Hyderabad) and 0 otherwise. Source: World Bank Enterprise Surveys (www.enterprisesurveys.org) Table 1: Description of Other Variables Variable Description Literacy Number of adult literates in the city per 1000 adult population of the city (1991 values). Source: Census of India (1991). Retailer density Total employment in retail and distribution in the city divided by adult city population (1991 values). Source: Census of India (1991). Sex ratio Ratio of females to males in the city in 1991. Source: Census of India (1991) Size Total selling area of the store measured in square feet (log values). Source: World Bank Enterprise Surveys (www.enterprisesurveys.org) Traditional store A dummy variable equal to 1 if a store is a 27 “traditional store” and 0 otherwise. Traditional stores belong to the Fast Moving Consumer Goods (FMCG) section of retailing and include grocers, general stores, chemists, food stores, cosmetic stores, etc. Source: World Bank Enterprise Surveys (www.enterprisesurveys.org) Modern Format store A dummy variable equal to 1 is a store is a “modern format” store and 0 otherwise. Modern format stores are large stores typically part of a bigger shopping complex. Source: World Bank Enterprise Surveys (www.enterprisesurveys.org) Store-type fixed effects Two dummy variables indicating whether a store is a traditional store or a modern format store (as defined above). The omitted category is that of consumer durable stores. Source: World Bank Enterprise Surveys (www.enterprisesurveys.org) Outage-own store Hours of power failure faced by the store on a typical day during the last fiscal year. Source: World Bank Enterprise Surveys (www.enterprisesurveys.org) Outage Hours of power outage on a typical day faced on an average by all other stores in the city during the last fiscal year. Source: World Bank Enterprise Surveys (www.enterprisesurveys.org) Financial Access Proportion of stores in the city that have a line of credit or overdraft facility. Source: World Bank Enterprise Surveys (www.enterprisesurveys.org) Checking Proportion of stores in the city that have a checking or savings account. Source: World Bank Enterprise Surveys (www.enterprisesurveys.org) Regulation Percentage of store’s senior management’s time spent in dealing with business regulations during the last fiscal year (average values at the city level). Source: World Bank Enterprise Surveys (www.enterprisesurveys.org) Chain A dummy variable equal to 1 if a store is part of a larger chain and 0 otherwise. Source: World Bank Enterprise Surveys (www.enterprisesurveys.org) Ownership concentration Percentage of the firm (store) held by the largest shareholder. Source: World Bank Enterprise Surveys (www.enterprisesurveys.org) Age Age of the store. It equals 2006 minus the year store was established. Source: World Bank Enterprise Surveys (www.enterprisesurveys.org) Workers Total number of workers in the city divided by the total number of households in the city (1991 values). Source: Census of India (1991). 28 Table 2: Summary statistics of the main variables Variable Mean Standard Range deviation (minimum, maximum) Competition 2.217 1.115 (1, 4) Non workers 2.85 0.381 (2.01, 3.88) Children 0.766 0.144 (0.491, 1.04) Expenditure 0.423 0.115 (0.292, 0.765) (in thousand Indian Rupees) Population (in millions) 1.31 1.68 (0.103, 8.58) Retailer density 0.082 0.017 (0.040, 0.114) Size (square feet, log 5.13 1.17 (2.08, 11.41) values) Sex ratio 0.877 0.065 (0.752, 1.011) 29 Table 3: Correlation coefficients for Non workers and Children with various indicators of overall development of the cities Non workers Children Non workers 1 0.488 Children 0.488 1 Expenditure -0.290 -0.126 (per capita expenditure in 1991- district level values) Population -0.096 -0.160 (adult city population in millions in 1991) Retailer density -0.020 -0.257 (total employment in the retail sector in the city as a proportion of adult population in the city, 1991 values) Sex ratio -0.036 -0.321 (ratio of females to males in the city, 1991 values) Size -0.027 -0.139 (Log of floor area of the retail stores in the sample measured in square feet– city level average) Power outage 0.132 0.368 (hours of power outage in a typical day faced by stores during 2005-06) Financial Access -0.047 -0.135 (proportion of stores in the city that have a line or credit or overdraft facility) Checking -0.007 -0.138 (proportion of stores in the city that have a checking or savings account) Regulation -0.037 -0.151 (amount of time spent by senior management of stores in dealing with business regulation during 2005-06 and averaged at the city level) Literacy rate -0.013 -0.496 (proportion of adults in the city that are literate, 1991 values taken from Census of India.) Workers 0.028 0.026 (total number of workers divided by total number of households in the city, 1991 values taken from Census of India.) 1. All variables are defined at the city level. Detailed description of the variables along with data sources is provided in Table 1. 2. Main point of the table is to show that Non workers and Children are correlated in the same direction with various proxy measures of overall development of the cities. 30 Table 4: Marginal effects for the highest level of competition Dependent variable: Competition (1) (2) (3) (4) (5) Non workers 0.089** 0.133*** 0.170*** (0.012) (0.000) (0.001) Children -0.076 -0.227* -0.346** (0.575) (0.085) (0.022) Expenditure -0.109 0.335* (0.365) (0.093) Population -0.007 (0.524) Metro -.093 (0.108) Predicted 0.183 0.186 0.185 0.181 0.179 probability Sample Size 1866 1866 1866 1866 1866 p-values in brackets; all standard errors clustered on the city; significance level is denoted by *** (1% or less), ** (5% or less) and * (10% or less). Estimates shown in the table are marginal effects for the highest level of competition ( Competition equal to 4) obtained using Ordered Logit estimation method. 31 Table 5: Marginal effects from ordered logit regressions: Robustness Dependent variable: Competition (1) (2) (3) (4) (5) Non workers 0.165*** 0.165*** 0.147*** 0.144*** 0.149*** (0.001) (0.000) (0.000) (0.000) (0.002) Children -0.429*** -0.370*** -0.288** -0.302** -0.404** (0.002) (0.012) (0.023) (0.015) (0.035) Expenditure 0.312 0.327 0.224 0.220 0.016 (0.104) (0.123) (0.230) (0.202) (0.951) Population -0.011 -0.017 -0.023 -0.022** -0.058 (0.317) (0.208) (0.037) (0.040) (0.142) Metro -0.089 -0.077 -0.094 -0.091 (0.130) (0.216) (0.120) (0.120) Literacy -2.29 -2.94 -2.67 -2.74 -2.48 (0.416) (0.274) (0.352) (0.328) (0.477) Sex ratio -0.498** -0.648*** -0.653*** -0.640*** -0.688** (0.040) (0.002) (0.006) (0.005) (0.027) Outage -0.021*** -0.020*** -0.020*** -0.021** (0.003) (0.005) (0.005) (0.013) Outage–own store 0.009*** 0.008*** 0.008*** 0.010*** (0.000) (0.000) (0.000) (0.000) Financial Access 0.094 0.048 0.056 0.183 (0.607) (0.777) (0.731) (0.383) Checking 0.006 0.101 0.101 0.110 (0.975) (0.439) (0.413) (0.486) Retailer density 2.88* 2.74* 2.92** (0.051) (0.053) (0.046) Regulation -0.006 -0.006 -0.009* (0.138) (.123) (.063) Size 0.015** 0.010 0.0001 (0.036) (0.259) (0.994) Store type fixed effects Yes Yes Age 0.0003 -0.0003 (0.613) (0.752) Chain -0.027 -0.004 (0.357) (0.926) Ownership concentration -0.001* -0.001* (0.066) (0.068) Predicted probability 0.175 0.171 0.168 0.166 0.184 Sample Size 1866 1859 1850 1850 1279 p-values in brackets; all standard errors are clustered on the city; significance level is denoted by *** (1% or less), ** (5% or less) and * (10% or less).The sample in column 5 excludes the metropolitan cities of Chennai, Delhi, Mumbai, Kolkata, Bangalore and Hyderabad. Sample size in columns 1-4 varies due to missing observations. 32 33 Table 6: Traditional vs. Non-Traditional stores (Marginal effects) Dependent variable: Competition (1) (2) (3) (4) (5) (6) Traditional stores Non-Traditional stores Non workers 0.090*** 0.177*** 0.149*** 0.074 0.135** 0.124** (0.010) (0.001) (0.000) (0.134) (0.039) (0.036) Children -0.354** -0.293** -0.299* -0.261 (0.024) (0.016) (0.091) (0.199) Expenditure 0.406** 0.320* 0.206 0.038 (0.046) (0.070) (0.366) (0.852) Population -0.009 -0.025** -0.005 -0.017 (0.429) (0.031) (0.720) (0.290) Metro -0.079 -0.075 -0.118* -0.128 (0.166) (0.185) (0.096) (0.146) Literacy -1.18 -6.04* (0.734) (0.071) Sex ratio -0.809*** -0.217 (0.001) (0.510) Outage -0.021*** -0.013 (0.001) (0.294) Outage– 0.011*** 0.0002 own store (0.000) (0.969) Financial 0.012 0.139 Access (0.935) (0.622) Checking 0.095 0.179 (0.437) (0.265) Retailer 2.09* 4.36* density (0.060) (.053) Regulation -0.004 -0.010* (0.296) (0.097) Size 0.016 -0.004 (0.127) (0.776) Age 0.0001 0.0007 (0.936) (0.442) Chain -0.030 (0.322) Ownership 0.0002 -0.002*** concentration (0.867) (0.001) Modern Format -0.037 stores (0.368) Predicted 0.165 0.160 0.148 0.215 0.211 0.196 probability Sample Size 1195 1195 1184 671 671 665 1. p-values in brackets; all standard errors are clustered on the city; significance level is denoted by *** (1% or less), ** (5% or less) and * (10% or less). Sample size varies due to missing observations. 2. There are no chain stores among traditional stores. Hence, Chain is excluded in column 3. 3. Main point of the table is to show that the significant effect of Expenditure (columns 1-3) disappears when we control for Non workers (columns 4-6). 34 Table 7: Marginal effects from ordered logit regressions (Robustness) Dependent variable: Competition (1) (2) (3) (4) (5) (6) Non workers 0.113*** 0.119*** 0.124*** (0.010) (0.009) (0.006) Children -0.077 -0.017 -0.034 -0.189* -0.202* -0.228** (0.475) (0.883) (0.770) (0.062) (0.078) (0.045) Expenditure -0.238** -0.396** -0.388** -0.116 -0.214 0.202 (0.046) (0.031) (0.026) (0.402) (0.338) (0.342) Literacy 0.548 -0.330 -1.87 -2.21 (0.841) (0.905) (0.507) (0.433) Sex ratio -0.664*** -0.663*** -0.644*** -0.586*** -0.488** -0.502** (0.004) (0.006) (0.005) (0.004) (0.047) (0.034) Outage -0.020*** -0.018*** -0.017*** -0.019*** -0.016** -0.015** (0.000) (0.006) (0.005) (0.001) (0.018) (0.021) Outage– 0.009*** 0.009*** 0.008*** 0.009*** 0.009*** 0.008*** own store (0.000) (0.000) (0.001) (0.000) (0.000) (0.000) Financial 0.073 -0.073 0.180 0.182 Access (0.692) (0.679) (0.368) (0.341) Checking 0.057 0.052 0.057 0.048 (0.712) (0.723) (0.720) (0.745) Retailer 2.29 2.17 1.58 1.45 density (0.115) (0.133) (0.218) (0.244) Regulation -0.004 -0.004 -0.006 -0.006 (0.335) (0.307) (0.202) (0.174) Size 0.008 0.011 (0.384) (0.225) Store type fixed effects Yes Yes Age 0.001 0.001 (0.294) (0.420) Chain -0.033 -0.034 (0.244) (0.210) Ownership -0.001 -0.001 concentration (0.154) (0.145) Predicted 0.179 0.177 0.174 0.176 0.175 0.171 probability Sample Size 1859 1859 1849 1859 1859 1849 1. p-values in brackets; all standard errors are clustered on the city; significance level is denoted by *** (1% or less), ** (5% or less) and * (10% or less). Sample size varies due to missing observations. 2. Main point of the table is to show that the significant effect of Expenditure (columns 1-3) disappears when we control for Non workers (columns 4-6). 35