Store Formats, Market Structure, and Consumers’ Food Shopping Decisions

Store Formats, Market Structure, and Consumers’ Food Shopping Decisions Abstract A growing literature in health and nutrition suggests that healthy foods are less available and more expensive at nontraditional store formats such as supercenters, convenience stores, and drug stores. We use Nielsen Homescan data to investigate the relationship between store format and the healthfulness of consumers’ grocery shopping. Accounting for a rich set of controls, as well as food retail market structure, we simultaneously estimate the healthfulness of consumers’ food purchases and the shares of food expenditure at traditional and nontraditional store formations. We find that healthier food choices are generally associated with higher food expenditure shares at supermarkets and supercenters and lower shares at drug stores and convenience stores. In addition, market concentration has a negative effect on shopping healthfulness. dietary quality, consumer choices, food expenditures, store formats, market structure, household scanner data D12, L11, Q12, I10 Introduction Obesity, overweight status, and diabetes are three examples of prevalent health concerns that can be attributed, at least in part, to people’s food choices. Each has been cited as a significant cause of death and source of healthcare costs in the United States (Bhattacharya and Bundorf 2009). It is widely understood that consumers do not, on average, meet USDA guidelines for healthy eating. Americans eat too few fruits, vegetables, and whole grains while eating too many fats, added sugars, and refined grains (Mancino et al. 2008; Dong and Lin 2009). Economists, social scientists, and nutritionists have devoted much research to the determinants of poor diet choices and possible solutions. One area of research that has received relatively little attention is the impact of store formats and store characteristics on consumers’ food purchasing decisions. Consumers are increasingly shopping at nontraditional store formats, which include supercenters, club stores, mass merchandisers that sell food, drug stores, and convenience stores. Martinez (2007) showed that the share of consumers’ food-at-home expenditures at outlets other than supermarkets increased from 17% to 31% from 1994 to 2005. This study uses Nielsen Homescan consumer purchase data to investigate the relationship between store formats and the healthfulness of consumers’ grocery shopping. Because store-format choice and food-purchase healthfulness are interrelated decisions, we estimate them in a simultaneous system. Taking care to control for probable confounding factors such as the food environment, demographics, and relative prices, we estimate how store format choices, as reflected by format-specific food expenditure shares, affect the healthfulness of households’ food purchases. We generally find that healthier food choices are associated with higher food expenditure shares at supermarkets and supercenters and lower shares at drug stores and convenience stores. In addition, increased retail food industry concentration has a negative effect on shopping healthfulness. Background on Store Choice and Food Shopping Decisions There are a number of reasons to expect that store formats might significantly shape consumers’ food choices. For instance, product assortments, average prices, and promotional strategies differ substantially across store formats (e.g., Blattberg et al. 1995; Seiders et al. 2000; Leibtag et al. 2010; Market Force 2011). A considerable literature in health and nutrition (e.g., Liese et al. 2007) is devoted to the availability and price of foods at different types of stores as they differ according to socioeconomic indicators. A regularity uncovered by a review of these studies is that healthful options, measured broadly by fruits, vegetables, and other whole foods, are fewer in number and higher in price at small retailers such a drug stores and convenience stores than they are at supermarkets. In rural Texas, Bustillos et al. (2008) find that the variety of healthful foods, including canned and frozen options, is significantly higher at supermarkets than at convenience stores, dollar stores, or mass merchandisers. For rural South Carolina, Liese et al. (2007) find that healthful food options are significantly less numerous and higher priced at convenience stores and other smaller formats. Studying a nationally representative of Food Stamp participants, Rose and Richards (2004) conclude that access to large supermarkets is directly correlated with the consumption of fruits and vegetables. Zenk et al. (2005), studying the purchase patterns of African American women, find that lower-income shoppers are more likely to shop at nontraditional formats and, as a result of differences in product availability across formats, purchase less fresh produce. A few studies go further and link the food environment to health outcomes. Powell and Bao (2009) find an inverse relationship between body mass index (BMI) and supermarket availability across metropolitan areas in the United States. Chen et al. (2016) show that household members residing in food deserts, areas defined in part by poor access to supermarkets, are more likely to be obese. A closely related stream of literature, which fits within in this category of studies on food choices by store format, examines food expenditures by income group. Goodman (1968), MacDonald and Nelson (1991), Kaufman et al. (1997), and Chung and Myers (1999) use a variety of empirical approaches to investigate whether poor households pay more for food, per unit, than do higher-income households. At the heart of each study is the notion that low-income households often lack access to large supermarkets and therefore shop at smaller stores with fewer choices and higher prices. Another relevant research topic, grounded in economics and marketing, investigates the determinants of store choice, particularly among formats. The study most similar in spirit and approach to our own is Fox et al. (2004), who find that much of the variation in household food expenditures, about 31%, can be explained by format choice, which in turn is determined more by promotional activities than by prices or location. Using a survey of representative households, Carpenter and Moore (2006) find that product assortment is more important for frequent shoppers at both supermarkets and club stores than for supercenter shoppers, whereas frequent supercenter shoppers value price competitiveness more than shoppers of alternative formats. Finally, a small group of research papers links market structure to food choice. For example, several studies identify a positive connection between food prices and market concentration (e.g., Lamm 1981; Yu and Connor 2002). However, supermarkets also compete along a variety of other dimensions, including variety (Richards and Hamilton 2006) and quality (Matsa 2011). Therefore, market concentration may shape consumers’ food choices through a variety of mechanisms, under the presumption that competition across firms decreases with concentration. In addition, the particular impacts of nontraditional store formats such as supercenters is a relatively new area of study. For example, Courtemanche and Carden (2011) attribute an increase in BMI and obesity in American cities to the opening of Wal-Mart Supercenters, while Volpe et al. (2013) find that increased supercenter market share within metropolitan areas is associated with decreased dietary quality among grocery purchases. Alternative Store Formats and Consumers’ Evolving Store Choices Table 1 summarizes the Food Institute’s (2010) descriptions of each of the store formats used in our study: supermarkets, drug stores, mass merchandisers, supercenters, club stores, convenience stores, and a category of other stores that consists mostly of dollar stores and military commissaries. While there are no universally accepted definitions and classifications of food retail store formats, throughout this study we use the store-format names provided by Nielsen, who collects the Homescan data. The format descriptions are drawn from a consensus of Progressive Grocer, the Food Marketing Institute, Nielsen, and the US Census for retail. Table 1 Descriptive Statistics for Food Retailing Store Formats, 2009 Format Description Number Stores (US) Avg. Store Area (sq. feet) Avg. No. SKUsa Avg. Weekly Sales Δ HH Penetration 2001–‘09 Supermarket Food retailer with greater than 9,000 sq. feet of selling space and $2 million annually in food sales. 40,205 52,100 45,500 $221,345 −2% Mass Merchandiser Department stores selling primarily general merchandise as well as a limited selection of foods, including perishables 3,683 66,400 95,000 $228,639 −18% Supercenter Large stores that combine department stores with full supermarkets 3,366 184,100 100,000 $903,311 19% Drug Store Stores featuring prescription-based pharmacies and generating at least 20% of revenues from food and general merchandise 24,660 12,300 5,400 $46,229 −6% Club Store Large stores selling food and general merchandise in bulk to consumers with paid memberships 1,304 130,500 5,100 $1,108,329 0% Convenience Store Similar to drug stores but without pharmacies. May or may not feature gas stations 150,704 2,600 4,200 $19,538 −7% Othersd A collection of smaller formats including dollar stores, direct-to-consumer outlets, and hospitals. 22,407 7,600 5,500 $20,228 5% Format Description Number Stores (US) Avg. Store Area (sq. feet) Avg. No. SKUsa Avg. Weekly Sales Δ HH Penetration 2001–‘09 Supermarket Food retailer with greater than 9,000 sq. feet of selling space and $2 million annually in food sales. 40,205 52,100 45,500 $221,345 −2% Mass Merchandiser Department stores selling primarily general merchandise as well as a limited selection of foods, including perishables 3,683 66,400 95,000 $228,639 −18% Supercenter Large stores that combine department stores with full supermarkets 3,366 184,100 100,000 $903,311 19% Drug Store Stores featuring prescription-based pharmacies and generating at least 20% of revenues from food and general merchandise 24,660 12,300 5,400 $46,229 −6% Club Store Large stores selling food and general merchandise in bulk to consumers with paid memberships 1,304 130,500 5,100 $1,108,329 0% Convenience Store Similar to drug stores but without pharmacies. May or may not feature gas stations 150,704 2,600 4,200 $19,538 −7% Othersd A collection of smaller formats including dollar stores, direct-to-consumer outlets, and hospitals. 22,407 7,600 5,500 $20,228 5% Source: The Food Institute Industry Review (2010). a The number of SKUs and average weekly sales include nonfood items and products not available at supermarkets. b In the 2010 Review this is a category only includes dollar stores when measuring penetration. Otherwise it reports a weighted average of dollar stores and military commissaries based on their expenditure shares in the Homescan data. Table 1 Descriptive Statistics for Food Retailing Store Formats, 2009 Format Description Number Stores (US) Avg. Store Area (sq. feet) Avg. No. SKUsa Avg. Weekly Sales Δ HH Penetration 2001–‘09 Supermarket Food retailer with greater than 9,000 sq. feet of selling space and $2 million annually in food sales. 40,205 52,100 45,500 $221,345 −2% Mass Merchandiser Department stores selling primarily general merchandise as well as a limited selection of foods, including perishables 3,683 66,400 95,000 $228,639 −18% Supercenter Large stores that combine department stores with full supermarkets 3,366 184,100 100,000 $903,311 19% Drug Store Stores featuring prescription-based pharmacies and generating at least 20% of revenues from food and general merchandise 24,660 12,300 5,400 $46,229 −6% Club Store Large stores selling food and general merchandise in bulk to consumers with paid memberships 1,304 130,500 5,100 $1,108,329 0% Convenience Store Similar to drug stores but without pharmacies. May or may not feature gas stations 150,704 2,600 4,200 $19,538 −7% Othersd A collection of smaller formats including dollar stores, direct-to-consumer outlets, and hospitals. 22,407 7,600 5,500 $20,228 5% Format Description Number Stores (US) Avg. Store Area (sq. feet) Avg. No. SKUsa Avg. Weekly Sales Δ HH Penetration 2001–‘09 Supermarket Food retailer with greater than 9,000 sq. feet of selling space and $2 million annually in food sales. 40,205 52,100 45,500 $221,345 −2% Mass Merchandiser Department stores selling primarily general merchandise as well as a limited selection of foods, including perishables 3,683 66,400 95,000 $228,639 −18% Supercenter Large stores that combine department stores with full supermarkets 3,366 184,100 100,000 $903,311 19% Drug Store Stores featuring prescription-based pharmacies and generating at least 20% of revenues from food and general merchandise 24,660 12,300 5,400 $46,229 −6% Club Store Large stores selling food and general merchandise in bulk to consumers with paid memberships 1,304 130,500 5,100 $1,108,329 0% Convenience Store Similar to drug stores but without pharmacies. May or may not feature gas stations 150,704 2,600 4,200 $19,538 −7% Othersd A collection of smaller formats including dollar stores, direct-to-consumer outlets, and hospitals. 22,407 7,600 5,500 $20,228 5% Source: The Food Institute Industry Review (2010). a The number of SKUs and average weekly sales include nonfood items and products not available at supermarkets. b In the 2010 Review this is a category only includes dollar stores when measuring penetration. Otherwise it reports a weighted average of dollar stores and military commissaries based on their expenditure shares in the Homescan data. In addition to the format descriptions, table 1 also includes format-specific descriptive statistics as well as estimates of the changes in household penetration, or the share of US households shopping at least some of the time at each of the formats, for the years 2000 through 2009. This statistic supports the narrative that Americans are increasingly shopping among nontraditional formats for their groceries. The total penetration for supermarkets fell slightly over this time, while it increased substantially for supercenters. Much of the increase in supercenter penetration is explained by the large-scale conversion of mass merchandisers to supercenters. The final store-format category, “others,” is an amalgamation of smaller formats defined by Nielsen. As of 2010, the breakdown of this category in Homescan is as follows: Forty percent of the receipts in this category come from dollar stores. Military commissaries and specialized department stores such as sporting goods and electronic stores account for 15% each. Online food shopping accounts for another 10%, and the remaining 20% is attributed to a collection of outlets such hospitals, vending machines, and fast food restaurants where consumers are able to buy snacks and drinks that are suitable for consumption at home. This category saw an increase in household penetration of 5%, the timing of which coincides with the rapid expansion of dollar stores during the recession of 2007–2009. Table 2 reports how consumers’ food spending has changed since 1999 based on data from the Census of Retail Trade, as compiled and organized by the USDA Economics Research Service (ERS). The table also includes the most comparable numbers as compiled for Progressive Grocer (PG) in Major (2013). There are obvious and important differences in the way the two sources chronicled in table 2 categorize and define retail food formats. However, both data sources corroborate the trend away from supermarkets and toward supercenters and club stores. The ERS numbers show the supermarket share of the US grocery dollar falling from 72.7% in 1999 to 63.8% in 2011. Meanwhile, the share for supercenters and club stores combined rose from 6% to 16%. Martinez (2007), also using US Census data, confirms that this trend began in earnest in 1994. The PG numbers show that the supermarket share falls from 72.3% in 1997 to 58.9% in 2011, while the share for warehouse clubs and supercenters grows from 9% to 23%. Table 2 Shares of Consumers’ Food Expenditures, by Store Format and Data Source Economic Research Service Year Supermarkets Convenience Stores Other Grocery Specialty Food Stores Warehouse Clubs and Supercenters Mass Merchandise Other Stores Home Deliveries, Mail Orders Direct to Consumera 1999 72.7 2.5 1.1 2.0 5.8 1.8 4.7 3.6 5.8 2000 70.9 2.5 1.4 2.0 7.2 1.7 4.9 3.4 6.0 2001 70.1 2.6 1.2 2.0 8.9 1.7 4.8 3.1 5.5 2002 67.4 2.6 1.3 2.0 11.7 1.6 4.8 3.0 5.6 2003 66.8 2.6 1.3 2.1 12.6 1.4 4.9 3.0 5.7 2004 66.3 2.7 0.8 2.3 13.4 1.2 4.9 2.8 5.6 2005 65.8 2.6 0.7 2.3 14.1 1.0 4.9 2.8 5.8 2006 65.5 2.6 0.8 2.3 14.7 0.9 4.9 2.8 5.5 2007 65.2 2.6 0.6 2.4 15.1 0.7 4.8 2.7 5.9 2008 64.9 2.5 0.7 2.4 15.6 0.7 4.8 2.6 5.9 2009 65.0 2.5 0.6 2.3 15.9 0.7 4.7 2.4 5.9 2010 64.4 2.5 0.9 2.3 16.1 0.6 4.8 2.4 5.9 2011 63.8 2.4 1.4 2.3 16.0 0.6 4.9 2.7 5.9 Progressive Grocer Supermarkets Pharmacy and Drug Stores Discount Deparment Stores Warehouse Clubs and Supercenters Mass Merchandise 1997 72.3 1.2 2.3 9.1 N/A 2002 66.0 1.2 2.2 15.6 0.7 2007 60.2 1.5 2.1 22.0 1.4 2011 58.9 1.5 1.7 23.2 1.5 Economic Research Service Year Supermarkets Convenience Stores Other Grocery Specialty Food Stores Warehouse Clubs and Supercenters Mass Merchandise Other Stores Home Deliveries, Mail Orders Direct to Consumera 1999 72.7 2.5 1.1 2.0 5.8 1.8 4.7 3.6 5.8 2000 70.9 2.5 1.4 2.0 7.2 1.7 4.9 3.4 6.0 2001 70.1 2.6 1.2 2.0 8.9 1.7 4.8 3.1 5.5 2002 67.4 2.6 1.3 2.0 11.7 1.6 4.8 3.0 5.6 2003 66.8 2.6 1.3 2.1 12.6 1.4 4.9 3.0 5.7 2004 66.3 2.7 0.8 2.3 13.4 1.2 4.9 2.8 5.6 2005 65.8 2.6 0.7 2.3 14.1 1.0 4.9 2.8 5.8 2006 65.5 2.6 0.8 2.3 14.7 0.9 4.9 2.8 5.5 2007 65.2 2.6 0.6 2.4 15.1 0.7 4.8 2.7 5.9 2008 64.9 2.5 0.7 2.4 15.6 0.7 4.8 2.6 5.9 2009 65.0 2.5 0.6 2.3 15.9 0.7 4.7 2.4 5.9 2010 64.4 2.5 0.9 2.3 16.1 0.6 4.8 2.4 5.9 2011 63.8 2.4 1.4 2.3 16.0 0.6 4.9 2.7 5.9 Progressive Grocer Supermarkets Pharmacy and Drug Stores Discount Deparment Stores Warehouse Clubs and Supercenters Mass Merchandise 1997 72.3 1.2 2.3 9.1 N/A 2002 66.0 1.2 2.2 15.6 0.7 2007 60.2 1.5 2.1 22.0 1.4 2011 58.9 1.5 1.7 23.2 1.5 Source: The Economic Research Service estimates are calculated using data from the US Census Bureau and the Bureau of Labor Statistics. They are available at the Food Expenditures Topic Page, http://www.ers.usda.gov/data-products/food-expenditures.aspx. The Progressive Grocer estimates are drawn from Major (2013) and are available here: http://www.progressivegrocer.com/top-stories/headlines/trending-topics/id39214/supermarkets-emerge-as-grocery-share-leaders/. a Via farmers, processors, and wholesalers. Table 2 Shares of Consumers’ Food Expenditures, by Store Format and Data Source Economic Research Service Year Supermarkets Convenience Stores Other Grocery Specialty Food Stores Warehouse Clubs and Supercenters Mass Merchandise Other Stores Home Deliveries, Mail Orders Direct to Consumera 1999 72.7 2.5 1.1 2.0 5.8 1.8 4.7 3.6 5.8 2000 70.9 2.5 1.4 2.0 7.2 1.7 4.9 3.4 6.0 2001 70.1 2.6 1.2 2.0 8.9 1.7 4.8 3.1 5.5 2002 67.4 2.6 1.3 2.0 11.7 1.6 4.8 3.0 5.6 2003 66.8 2.6 1.3 2.1 12.6 1.4 4.9 3.0 5.7 2004 66.3 2.7 0.8 2.3 13.4 1.2 4.9 2.8 5.6 2005 65.8 2.6 0.7 2.3 14.1 1.0 4.9 2.8 5.8 2006 65.5 2.6 0.8 2.3 14.7 0.9 4.9 2.8 5.5 2007 65.2 2.6 0.6 2.4 15.1 0.7 4.8 2.7 5.9 2008 64.9 2.5 0.7 2.4 15.6 0.7 4.8 2.6 5.9 2009 65.0 2.5 0.6 2.3 15.9 0.7 4.7 2.4 5.9 2010 64.4 2.5 0.9 2.3 16.1 0.6 4.8 2.4 5.9 2011 63.8 2.4 1.4 2.3 16.0 0.6 4.9 2.7 5.9 Progressive Grocer Supermarkets Pharmacy and Drug Stores Discount Deparment Stores Warehouse Clubs and Supercenters Mass Merchandise 1997 72.3 1.2 2.3 9.1 N/A 2002 66.0 1.2 2.2 15.6 0.7 2007 60.2 1.5 2.1 22.0 1.4 2011 58.9 1.5 1.7 23.2 1.5 Economic Research Service Year Supermarkets Convenience Stores Other Grocery Specialty Food Stores Warehouse Clubs and Supercenters Mass Merchandise Other Stores Home Deliveries, Mail Orders Direct to Consumera 1999 72.7 2.5 1.1 2.0 5.8 1.8 4.7 3.6 5.8 2000 70.9 2.5 1.4 2.0 7.2 1.7 4.9 3.4 6.0 2001 70.1 2.6 1.2 2.0 8.9 1.7 4.8 3.1 5.5 2002 67.4 2.6 1.3 2.0 11.7 1.6 4.8 3.0 5.6 2003 66.8 2.6 1.3 2.1 12.6 1.4 4.9 3.0 5.7 2004 66.3 2.7 0.8 2.3 13.4 1.2 4.9 2.8 5.6 2005 65.8 2.6 0.7 2.3 14.1 1.0 4.9 2.8 5.8 2006 65.5 2.6 0.8 2.3 14.7 0.9 4.9 2.8 5.5 2007 65.2 2.6 0.6 2.4 15.1 0.7 4.8 2.7 5.9 2008 64.9 2.5 0.7 2.4 15.6 0.7 4.8 2.6 5.9 2009 65.0 2.5 0.6 2.3 15.9 0.7 4.7 2.4 5.9 2010 64.4 2.5 0.9 2.3 16.1 0.6 4.8 2.4 5.9 2011 63.8 2.4 1.4 2.3 16.0 0.6 4.9 2.7 5.9 Progressive Grocer Supermarkets Pharmacy and Drug Stores Discount Deparment Stores Warehouse Clubs and Supercenters Mass Merchandise 1997 72.3 1.2 2.3 9.1 N/A 2002 66.0 1.2 2.2 15.6 0.7 2007 60.2 1.5 2.1 22.0 1.4 2011 58.9 1.5 1.7 23.2 1.5 Source: The Economic Research Service estimates are calculated using data from the US Census Bureau and the Bureau of Labor Statistics. They are available at the Food Expenditures Topic Page, http://www.ers.usda.gov/data-products/food-expenditures.aspx. The Progressive Grocer estimates are drawn from Major (2013) and are available here: http://www.progressivegrocer.com/top-stories/headlines/trending-topics/id39214/supermarkets-emerge-as-grocery-share-leaders/. a Via farmers, processors, and wholesalers. Methodological Approach To investigate the healthfulness of households’ food purchases, we adopt the methodology developed by Volpe and Okrent (2012) and assign a healthfulness score, hereafter called the USDAScore, to the quarterly shopping baskets of each household. This score relies on the classification of all grocery products, undertaken by the USDA Center for Nutrition Policy and Promotion (CNPP) and reported by Carlson et al. (2007), into a series of broader food categories. These categories are comprehensive of all products in the Nielsen Homescan data and are nonoverlapping. The Carlson et al. (2007) report provides spending recommendations, allotting food dollars across these CNPP categories, to assist consumers in making grocery purchases that abide by the Dietary Guidelines for Americans (DGA). Moreover, these recommendations are distinct according to age and gender. With the rich demographics of the Homescan data, we are able to construct household-specific USDA spending recommendations based on household composition. The USDAScore is based on the differences between the category-specific observed expenditure shares and the USDA recommended expenditure shares. More specifically, the USDAScore for household i in quarter q is given by USDAScoreicq=∑c((expshareicq−USDAexpshareic)2)−1 where c denotes the CNPP food categories. The complete details behind the construction of this score, and information on the CNPP food categories (including how food items in Homescan are mapped to these categories), are available in Volpe and Okrent (2012). The USDAScore is unitless and should be thought of as an index, where higher scores indicate greater adherence to the DGA.1 Our general goal is to investigate how store format decisions and other factors affect the household-specific USDAScore. Despite the rich set of demographics afforded by the Homescan data, a single-equation approach modeling USDAScore as a function of format expenditure shares and controls could be subject to potential endogeneity. Shoppers may select a store format based on their food preferences, particularly their demand for healthful food. For example, if supermarkets offer the healthiest food, then it would stand to reason that people seeking a healthy lifestyle will choose these stores. With these complexities in mind, we use a system of equations approach to account for the potential endogeneity of format shares. In this setting, we simultaneously model two consumer choices: what store format to shop at and how healthy the shopping basket will be. Drawing on the theoretical and applied literature on store choice and food choice, we specify other covariates that may affect simultaneously a household’s observed store-format shares and USDAScore. Our empirical framework relies on eight reduced-form equations, seven that model store-format expenditure share as a function of prices, the food environment, and household demographics, and one that models the USDAScore as a function of the format expenditure shares, prices, market structure, and household demographics. These eight equations are represented by Fimtj=fPRmtj,Smt,Dimt+υimt,forj=1to7,and (1) USDAScoreimt=fFimt,Phmt,Mmt,Dimt+ωimt, (2) where Fimtj is the format share for household i, in market m and quarter t, for store format j. Fimtj is a function of demographics (D), including proxies for time constraints, the local food environment (S), and retail-format prices (PR). USDAScoreimtj is a function of prices for healthy foods (Ph), market structure (M), store format shares (F), and household demographics (D). Estimating the simultaneous equation system given by equations (1) and (2) presents a number of empirical concerns. First, we drop one of the format share equations to avoid a linear dependency. A second concern involves specifying S, D, and M. Here, we draw on related literature and describe how these vectors are specified in the next section. A third concern is that the store format shares, Fimtj, are bounded by zero, thus making the equations given by (1) left censored.2 This censoring makes estimating the system computationally intensive. To overcome this problem, we rely on the general concepts found in Shonkwiler and Yen (1999), who employ a two-step procedure to estimate a left-censored system. To apply their procedure to our system, we construct and estimate individual probit equations corresponding to the equations in (1), recover the Inverse Mills ratios, then estimate uncensored linear equations for (1) and (2) as a system, with the Inverse Mills ratios included as extra terms in (1). The last concern involves prices, which pose several empirical difficulties. One difficulty, discussed in the next section, involves the question of how to specify PR and Ph. However, a second and especially troublesome difficulty, potential endogeneity, is discussed here. If underlying determinants of prices are unobserved by the researcher but correlated with the error terms, then the coefficient estimates in the simultaneous system can be biased. A general strategy to overcome this endogeneity issue is to use instruments to estimate PR and Ph. We follow this strategy and use Hausman-style instruments composed of prices from outside markets that have uncorrelated demand or supply shocks. Thus PR and Ph are replaced by their predicted values from instrumental variable estimations. Data and Variable Specification The Homescan data set consists of the self-scanned purchases of a sample of households who are asked to record their complete purchases of each shopping trip by scanning UPCs. For the years 2004–2010 the database contains over 50,000 households (most of which are not in the sample the entire period) that span nearly the entire continental United States. The Homescan data include detailed information on product names and characteristics as well as household demographics, thus allowing for an in-depth analysis of quality of food purchases while controlling for extensive confounding variables. While the Homescan data are self-reported, robustness checks and calibration efforts have demonstrated the data to contain errors in line with those found in major government-collected data sets (Einav et al. 2008, 2010). We structure the data into cumulative shopping baskets by household and quarter (1998Q1–2010Q4). The final data set consists of more than 1.1 million quarterly baskets from over 90,000 unique households. A key aspect to every basket is that expenditures are broken down by store format, a treatment that allows for the precise measurement of the contribution each store format has on shopping baskets. Table 3 provides USDAScore and expenditure shares for selected food categories, by store format, and with the exception of club stores, it shows that foods purchased at nontraditional formats are less healthful. It is important to keep in mind throughout that all shares presented at table 3 are averages across formats. That is, if we treat all expenditures at supermarkets as a single shopping basket, it has an average USDAScore of 7.4 and fruits constitute 6.3% of the expenditures. This average does not necessarily reflect the purchasing habits of households in general or those that shop entirely or primarily at supermarkets, as consumers are able to shop among formats to complete their food shopping within any given time period, and in practice often do so (Fox et al. 2004; Gauri et al. 2008).3 Supermarket expenditures have the highest USDAScore save for club stores, which score slightly higher at 7.5. Foods purchased at mass merchandisers and drug stores are the least healthful, with each format scoring 3.6. Convenience stores, the subject of much of the literature that motivates this study, score 4.3. Other stores score 6.4 and supercenter expenditures are slightly less healthful on average than those made at supermarkets, scoring 6.9. Table 3 Average USDAScore and Shares of Total Expenditures Attributed to Selected Food Categories, by Store Format Supermarkets Mass. Merchandise Supercenters Drug Stores Club Stores Convenience Stores Other USDAScore 7.43 3.55 6.94 3.55 7.50 4.27 6.41 Fruits 6.32 2.12 4.70 2.65 9.24 1.79 3.39 Vegetables 7.39 1.23 4.93 0.70 4.98 0.70 3.56 Whole Grains 2.01 2.09 2.02 1.19 1.53 0.39 1.71 Refined Grains 8.64 5.98 7.65 2.27 4.95 5.02 7.94 Low-fat Dairy 4.60 1.51 3.37 2.25 2.26 10.36 1.24 Regular Dairy 8.42 2.78 6.53 4.44 6.70 15.84 3.03 Low-fat Meats 1.10 0.29 1.42 0.04 0.94 0.07 0.21 Regular Meats 5.24 1.04 4.89 0.72 4.64 0.063 1.91 Poultry 0.66 0.15 0.61 0.02 2.03 0.03 0.23 Fats 2.29 0.43 1.87 0.22 1.58 0.39 0.49 Sugar Drinks 5.52 8.99 6.31 7.95 3.73 17.16 8.53 Sweet Packaged 12.37 34.34 17.95 45.71 11.24 18.16 32.93 Savory Packaged 23.22 22.05 24.02 12.43 26.23 14.01 19.91 Supermarkets Mass. Merchandise Supercenters Drug Stores Club Stores Convenience Stores Other USDAScore 7.43 3.55 6.94 3.55 7.50 4.27 6.41 Fruits 6.32 2.12 4.70 2.65 9.24 1.79 3.39 Vegetables 7.39 1.23 4.93 0.70 4.98 0.70 3.56 Whole Grains 2.01 2.09 2.02 1.19 1.53 0.39 1.71 Refined Grains 8.64 5.98 7.65 2.27 4.95 5.02 7.94 Low-fat Dairy 4.60 1.51 3.37 2.25 2.26 10.36 1.24 Regular Dairy 8.42 2.78 6.53 4.44 6.70 15.84 3.03 Low-fat Meats 1.10 0.29 1.42 0.04 0.94 0.07 0.21 Regular Meats 5.24 1.04 4.89 0.72 4.64 0.063 1.91 Poultry 0.66 0.15 0.61 0.02 2.03 0.03 0.23 Fats 2.29 0.43 1.87 0.22 1.58 0.39 0.49 Sugar Drinks 5.52 8.99 6.31 7.95 3.73 17.16 8.53 Sweet Packaged 12.37 34.34 17.95 45.71 11.24 18.16 32.93 Savory Packaged 23.22 22.05 24.02 12.43 26.23 14.01 19.91 Notes: The reported figures pertain to calculations made on the various food categories. For example, 6.32% of all supermarket expenditures are on fruits. The expenditures in each column will not sum to one because not all food categories are reported. Table 3 Average USDAScore and Shares of Total Expenditures Attributed to Selected Food Categories, by Store Format Supermarkets Mass. Merchandise Supercenters Drug Stores Club Stores Convenience Stores Other USDAScore 7.43 3.55 6.94 3.55 7.50 4.27 6.41 Fruits 6.32 2.12 4.70 2.65 9.24 1.79 3.39 Vegetables 7.39 1.23 4.93 0.70 4.98 0.70 3.56 Whole Grains 2.01 2.09 2.02 1.19 1.53 0.39 1.71 Refined Grains 8.64 5.98 7.65 2.27 4.95 5.02 7.94 Low-fat Dairy 4.60 1.51 3.37 2.25 2.26 10.36 1.24 Regular Dairy 8.42 2.78 6.53 4.44 6.70 15.84 3.03 Low-fat Meats 1.10 0.29 1.42 0.04 0.94 0.07 0.21 Regular Meats 5.24 1.04 4.89 0.72 4.64 0.063 1.91 Poultry 0.66 0.15 0.61 0.02 2.03 0.03 0.23 Fats 2.29 0.43 1.87 0.22 1.58 0.39 0.49 Sugar Drinks 5.52 8.99 6.31 7.95 3.73 17.16 8.53 Sweet Packaged 12.37 34.34 17.95 45.71 11.24 18.16 32.93 Savory Packaged 23.22 22.05 24.02 12.43 26.23 14.01 19.91 Supermarkets Mass. Merchandise Supercenters Drug Stores Club Stores Convenience Stores Other USDAScore 7.43 3.55 6.94 3.55 7.50 4.27 6.41 Fruits 6.32 2.12 4.70 2.65 9.24 1.79 3.39 Vegetables 7.39 1.23 4.93 0.70 4.98 0.70 3.56 Whole Grains 2.01 2.09 2.02 1.19 1.53 0.39 1.71 Refined Grains 8.64 5.98 7.65 2.27 4.95 5.02 7.94 Low-fat Dairy 4.60 1.51 3.37 2.25 2.26 10.36 1.24 Regular Dairy 8.42 2.78 6.53 4.44 6.70 15.84 3.03 Low-fat Meats 1.10 0.29 1.42 0.04 0.94 0.07 0.21 Regular Meats 5.24 1.04 4.89 0.72 4.64 0.063 1.91 Poultry 0.66 0.15 0.61 0.02 2.03 0.03 0.23 Fats 2.29 0.43 1.87 0.22 1.58 0.39 0.49 Sugar Drinks 5.52 8.99 6.31 7.95 3.73 17.16 8.53 Sweet Packaged 12.37 34.34 17.95 45.71 11.24 18.16 32.93 Savory Packaged 23.22 22.05 24.02 12.43 26.23 14.01 19.91 Notes: The reported figures pertain to calculations made on the various food categories. For example, 6.32% of all supermarket expenditures are on fruits. The expenditures in each column will not sum to one because not all food categories are reported. An examination of food categories, which mostly correspond with or are aggregations of the CNPP food groups, helps to explain the differences in healthfulness across formats. The DGA recommends that people consume increased quantities of fruits and vegetables while consuming reduced quantities of food prepared with added sugar or sodium. The expenditure shares for fruits and vegetables fall in a distinct hierarchy across the formats, with the shares being the highest at supermarkets and club stores. Shares on produce are lower at supercenters and other stores, and they are very low, between 0 and 2%, at the mass merchandisers, convenience stores, and drug stores. On average, households spend a far greater share of their food dollars on sweet and savory packaged foods than on fruits and vegetables. However, the hierarchy is repeated, in reverse, for these foods. It is also notable that over 17% of all spending at convenience stores goes to sugar drinks, including soda and sports drinks, and this is double the next-highest expenditure share for this category among formats. The differences across formats for the remaining categories are generally not striking and do not follow a clear pattern. For example, whole grains are also recommended by the DGA, and the shares for these foods are comparably low across all formats. Thus, it seems that the variation in overall healthfulness across formats is driven mostly by the differences in fruits, vegetables, and packaged foods. Figure 1 presents another way to depict the relationship between households’ USDAScores and their predominant store format choice. For each of the seven store formats, we calculate the yearly average USDAScore among households with a store format share greater than or equal to the 75th percentile among all households in our sample.4Figure 1 plots these averages for USDAScore and shows that shopping baskets are healthiest for households above the 75th percentile shares for club stores, followed by households above the 75th percentile shares for supermarkets. Conversely, shopping baskets are unhealthiest for convenience, other, and drug stores. Figure 1 View largeDownload slide USDAScore by Store Channel Notes: The figure reflects the average USDAScore 1 for households with a store channel share greater than or equal to the 75th percentile share, except Convenience stores, which uses the 90th percentile (because the 75th percentile for convenience stores is a zero share). Figure 1 View largeDownload slide USDAScore by Store Channel Notes: The figure reflects the average USDAScore 1 for households with a store channel share greater than or equal to the 75th percentile share, except Convenience stores, which uses the 90th percentile (because the 75th percentile for convenience stores is a zero share). The USDAScore serves as the dependent variable in equation (2), but regressors in equations (1) and (2) still must be specified before estimating the system. The demographic covariates in D are based on a systematic review of the literature on health, obesity, and food choices ranging across a number of disciplines. Blisard et al. (2004) and Dong and Lin (2009) are examples of applied research that suggest controlling for income, with findings showing that lower-income households purchase and consume fewer healthful foods, on average. A similarly rich line of research has generated a similar consensus with respect to education, food choices, and health outcomes (Garn et al. 1977; Galobardes et al. 2000; Xie et al. 2003). Researchers have also uncovered persistent and significant differences in dietary quality and related health outcomes by race and ethnicity (Block et al. 2004; Zenk et al. 2005). We use Homescan variables for household income, household size, education, and race. We also make efforts to account for the role of time constraints due to the fact that households vary in terms of the time available to prepare food at home, as implied by Cutler et al. (2003) and Mancino and Newman (2007). We therefore include an unemployment variable as well as indicator variables for an array of occupations. In equation (1), Smt is meant to reflect the retail food-at-home food environment. A number of specifications are possible in this setting (including simple store counts per capita). All of the models we estimated yielded substantively similar results with respect to our key hypotheses, and ultimately we employ two measures: the ratio of supermarket counts to convenience store counts and the ratio of supermarket counts to supercenter counts. These ratios are calculated by market and year using store counts drawn from the Homescan data.5 The impacts of convenience stores and supercenters on consumers generate the most attention, by a wide margin, in the research literature and popular press. For market structure (M), we use the TDLinx data set, which contains detailed store characteristics, including location, ownership, size, and categorical annual sales, to calculate the Herfindahl-Hirschman Index (HHI) among food retailers as a measure of concentration as well as the market share controlled by individual store formats, by year and by market.6 The coverage is comprehensive of US food retailers. Unfortunately, the store formats identified in TDLinx do not correspond cleanly to those in Homescan, and food sales data are not available for all of the formats studied in the Homescan data.7 Recognizing that shoppers choose store formats based partly on price, PRmtj is a proxy for the vector of average relative prices of the formats. Given the complexities involved in comparing prices across store types, we simplify and use only the format-specific price of bottled drinking water, which is one of the few products with available UPCs in identical form across all formats throughout the data. Therefore, for PR we use the average per ounce price of bottled water, by format, market, and quarter, normalized such as the price for other stores is one. Finally, equation (2) contains Phmt, an aggregate price for healthy food products, to signify the importance of prices on dietary quality as research suggests (e.g., Kuchler and Stewart 2008; Carlson and Frazao 2012). We use publicly available data from the USDA Quarterly Food-at-Home Price Database (QFAHPD), which provides market-level prices per 100 grams, by food group (see Todd et al. [2010] for more information on the QFAHPD). This resource is ideal for our purposes, given that the QFAPHD is calculated using Homescan, and the QFAHPD food groups were selected with the DGA in mind. For each market and quarter, we calculate a healthy food price and an unhealthy food price as the expenditure-share weighted average of the QFAHPD prices for those food groups falling in the respective categories. Volpe et al. (2013, 571) provide a listing of all 52 food crops along with their designation as healthy or unhealthy. However, because the unhealthy food groups in the QFAHPD are fewer in number and more commonly found across all store formats, we focus only on the aggregate unhealthy price, which we refer to as PUnHealthy. Because of endogeneity concerns mentioned above, we use three sets of instruments to predict PR and PUnHealthy. One set is lagged prices, which follows a long literature on the topic of endogenous regressors (Murray 2006). A second set capture marginal costs, or input prices, that might affect retail prices (e.g.,Villas-Boas 2009). For these, we employ a series of indexes from the Bureau of Labor Statistics (BLS). The BLS database includes both Consumer Price Indexes (CPIs), which measure retail price changes, and Producer Price Indexes (PPIs), which measure farm and wholesale price changes. The CPIs were available regionally, while the PPIs are national. Specifically, we use the following indexes: motor fuel CPI, fuels and utilities CPI, food manufacturing PPI, electric power generation PPI, truck transportation PPI, farm product warehousing and storage PPI, all commodities PPI, and grocery stores PPI. The last set of instruments intended to capture the role of input prices is rooted in Hausman’s (1996) argument that if demographics are properly accounted for, then the prices of a good in one city may be instrumented for using the prices for that same good in another city. Nevo (2001) and others use this technique as well. However, our data contain more cities than quarters, making it impossible to exploit all potential instruments. We instead regress our price variables against 100 random draws of 20 cities and yield 100 expected prices, which we then average. These average expected prices, following the Hausman framework, are cleaned of endogeneity. Following these methods, we rely on the predicted prices P^R and P^UnHealthy. Table 4 presents summary statistics for all the variables used in our analysis. Table 4 Summary Statistics for Variables in Our Final Data Set Variable Description Mean Std. Dev. Min Max USDAScore Basket healthfulness 7.345 2.476 0.898 19.995 FSupermarket Supermarket expenditure share 0.656 0.294 0 1 FDrug Drug store expenditure share 0.026 0.067 0 1 FMass Mass merchandizer expend. share 0.035 0.089 0 1 FSupercenter Supercenter expenditure share 0.134 0.242 0 1 FClub Club store expenditure share 0.089 0.170 0 1 FConvenience Convenience store expend. share 0.005 0.033 0 1 FOther Other stores expenditure share 0.056 0.134 0 1 PUnHealthy Unhealthy price (based on QFAHPD) 0.417 0.036 0.336 0.520 HHIFood Herfindahl-Hirschman Index (food retailers) 0.381 0.229 0.041 1 PSupermarket Avg. supermarket bottled water price/oz 0.022 0.004 0.014 0.032 PDrug Avg. drug store bottled water price/oz 0.030 0.006 0.006 0.059 PMass Avg. mass merch. bottled water price/oz 0.025 0.006 0.011 0.067 PSupermcenter Avg. supercenter bottled water price/oz 0.023 0.009 0.005 0.099 PClub Avg. club store bottled water price/oz 0.013 0.004 0.010 0.034 PConvenience Avg. c-store bottled water price/oz 0.051 0.013 0.009 0.115 POther Avg. other store bottled water price/oz 0.032 0.008 0.007 0.066 SSMarket/Conven Supermarket to convenience store count 1.584 0.923 0.593 6.941 SSMarket/SCent Supermarket to supercenter store count 9.897 12.396 2.54 188 HH-Inc Categorical HH Income 19.895 6.050 3 30 HH-Size Categorical HH Size 2.361 1.298 1 9 Black Back/African American status 0.100 0.300 0 1 Asian Asian American status 0.027 0.163 0 1 Other Race Other race status 0.047 0.211 0 1 Unemployed Unemployed status 0.331 0.471 0 1 MH-Educ Categorical male head education level 3.066 2.079 0 6 MH-Empl. Categorical male head employment status 3.483 3.244 0 9 WIC-Ever Every participate in WIC status 0.059 0.237 0 1 Variable Description Mean Std. Dev. Min Max USDAScore Basket healthfulness 7.345 2.476 0.898 19.995 FSupermarket Supermarket expenditure share 0.656 0.294 0 1 FDrug Drug store expenditure share 0.026 0.067 0 1 FMass Mass merchandizer expend. share 0.035 0.089 0 1 FSupercenter Supercenter expenditure share 0.134 0.242 0 1 FClub Club store expenditure share 0.089 0.170 0 1 FConvenience Convenience store expend. share 0.005 0.033 0 1 FOther Other stores expenditure share 0.056 0.134 0 1 PUnHealthy Unhealthy price (based on QFAHPD) 0.417 0.036 0.336 0.520 HHIFood Herfindahl-Hirschman Index (food retailers) 0.381 0.229 0.041 1 PSupermarket Avg. supermarket bottled water price/oz 0.022 0.004 0.014 0.032 PDrug Avg. drug store bottled water price/oz 0.030 0.006 0.006 0.059 PMass Avg. mass merch. bottled water price/oz 0.025 0.006 0.011 0.067 PSupermcenter Avg. supercenter bottled water price/oz 0.023 0.009 0.005 0.099 PClub Avg. club store bottled water price/oz 0.013 0.004 0.010 0.034 PConvenience Avg. c-store bottled water price/oz 0.051 0.013 0.009 0.115 POther Avg. other store bottled water price/oz 0.032 0.008 0.007 0.066 SSMarket/Conven Supermarket to convenience store count 1.584 0.923 0.593 6.941 SSMarket/SCent Supermarket to supercenter store count 9.897 12.396 2.54 188 HH-Inc Categorical HH Income 19.895 6.050 3 30 HH-Size Categorical HH Size 2.361 1.298 1 9 Black Back/African American status 0.100 0.300 0 1 Asian Asian American status 0.027 0.163 0 1 Other Race Other race status 0.047 0.211 0 1 Unemployed Unemployed status 0.331 0.471 0 1 MH-Educ Categorical male head education level 3.066 2.079 0 6 MH-Empl. Categorical male head employment status 3.483 3.244 0 9 WIC-Ever Every participate in WIC status 0.059 0.237 0 1 Table 4 Summary Statistics for Variables in Our Final Data Set Variable Description Mean Std. Dev. Min Max USDAScore Basket healthfulness 7.345 2.476 0.898 19.995 FSupermarket Supermarket expenditure share 0.656 0.294 0 1 FDrug Drug store expenditure share 0.026 0.067 0 1 FMass Mass merchandizer expend. share 0.035 0.089 0 1 FSupercenter Supercenter expenditure share 0.134 0.242 0 1 FClub Club store expenditure share 0.089 0.170 0 1 FConvenience Convenience store expend. share 0.005 0.033 0 1 FOther Other stores expenditure share 0.056 0.134 0 1 PUnHealthy Unhealthy price (based on QFAHPD) 0.417 0.036 0.336 0.520 HHIFood Herfindahl-Hirschman Index (food retailers) 0.381 0.229 0.041 1 PSupermarket Avg. supermarket bottled water price/oz 0.022 0.004 0.014 0.032 PDrug Avg. drug store bottled water price/oz 0.030 0.006 0.006 0.059 PMass Avg. mass merch. bottled water price/oz 0.025 0.006 0.011 0.067 PSupermcenter Avg. supercenter bottled water price/oz 0.023 0.009 0.005 0.099 PClub Avg. club store bottled water price/oz 0.013 0.004 0.010 0.034 PConvenience Avg. c-store bottled water price/oz 0.051 0.013 0.009 0.115 POther Avg. other store bottled water price/oz 0.032 0.008 0.007 0.066 SSMarket/Conven Supermarket to convenience store count 1.584 0.923 0.593 6.941 SSMarket/SCent Supermarket to supercenter store count 9.897 12.396 2.54 188 HH-Inc Categorical HH Income 19.895 6.050 3 30 HH-Size Categorical HH Size 2.361 1.298 1 9 Black Back/African American status 0.100 0.300 0 1 Asian Asian American status 0.027 0.163 0 1 Other Race Other race status 0.047 0.211 0 1 Unemployed Unemployed status 0.331 0.471 0 1 MH-Educ Categorical male head education level 3.066 2.079 0 6 MH-Empl. Categorical male head employment status 3.483 3.244 0 9 WIC-Ever Every participate in WIC status 0.059 0.237 0 1 Variable Description Mean Std. Dev. Min Max USDAScore Basket healthfulness 7.345 2.476 0.898 19.995 FSupermarket Supermarket expenditure share 0.656 0.294 0 1 FDrug Drug store expenditure share 0.026 0.067 0 1 FMass Mass merchandizer expend. share 0.035 0.089 0 1 FSupercenter Supercenter expenditure share 0.134 0.242 0 1 FClub Club store expenditure share 0.089 0.170 0 1 FConvenience Convenience store expend. share 0.005 0.033 0 1 FOther Other stores expenditure share 0.056 0.134 0 1 PUnHealthy Unhealthy price (based on QFAHPD) 0.417 0.036 0.336 0.520 HHIFood Herfindahl-Hirschman Index (food retailers) 0.381 0.229 0.041 1 PSupermarket Avg. supermarket bottled water price/oz 0.022 0.004 0.014 0.032 PDrug Avg. drug store bottled water price/oz 0.030 0.006 0.006 0.059 PMass Avg. mass merch. bottled water price/oz 0.025 0.006 0.011 0.067 PSupermcenter Avg. supercenter bottled water price/oz 0.023 0.009 0.005 0.099 PClub Avg. club store bottled water price/oz 0.013 0.004 0.010 0.034 PConvenience Avg. c-store bottled water price/oz 0.051 0.013 0.009 0.115 POther Avg. other store bottled water price/oz 0.032 0.008 0.007 0.066 SSMarket/Conven Supermarket to convenience store count 1.584 0.923 0.593 6.941 SSMarket/SCent Supermarket to supercenter store count 9.897 12.396 2.54 188 HH-Inc Categorical HH Income 19.895 6.050 3 30 HH-Size Categorical HH Size 2.361 1.298 1 9 Black Back/African American status 0.100 0.300 0 1 Asian Asian American status 0.027 0.163 0 1 Other Race Other race status 0.047 0.211 0 1 Unemployed Unemployed status 0.331 0.471 0 1 MH-Educ Categorical male head education level 3.066 2.079 0 6 MH-Empl. Categorical male head employment status 3.483 3.244 0 9 WIC-Ever Every participate in WIC status 0.059 0.237 0 1 Results Based on the specifications discussed above, we estimate equations (1) and (2) simultaneously using, in part, the Shonkwiler and Yen (1999) two-step procedure. More specifically, the Inverse Mills ratios are recovered from the first-step probit estimations corresponding to equation (1); then the system of equations (1) and (2) are estimated simultaneously with 3SLS. Results from this second step are presented in table 5. Table 5 3SLS Results, 2004–2010, Full Sample USDAScore FSupermarket FDrug FMass FSupercenter FClub FConvenience Equation (2): Est. Coeff. Equations (1): Est. Coeff. Est. Coeff. Est. Coeff. Est. Coeff. Est. Coeff. Est. Coeff. FSupermarket 1.859*** P^Supermarket 4.707*** 1.140*** 3.636*** 0.029*** 3.596*** −0.215*** FDrug −7.668*** P^Drug 4.158*** −0.452*** −0.827*** −0.0003*** −4.184*** 0.168*** FMass 0.728 P^Mass −4.522*** −0.104** −0.230*** −0.0002*** 3.566*** −0.234*** FSupercenter 2.336*** P^Supermcenter 3.379*** −0.147*** 0.566*** 0.005*** −0.460*** 0.059*** FClub −0.682 P^Club 2.531*** 1.196*** −1.234*** −0.009*** −2.538*** 0.026 FConvenience −20.48*** P^Convenience 1.424*** 0.013 0.313*** −0.018*** −0.208*** −0.062*** P^UnHealthy 2.246*** P^Other 4.073*** 0.585*** −0.238*** −0.017*** −2.470*** 0.247*** HHIFood −0.111*** SSMarket/Conven −0.041*** −0.004*** −0.002*** −0.006** 0.0233*** −0.002*** SSMarket/SCent 0.0003*** 0.00003*** −0.00003*** −0.016*** 0.0002*** −0.00001 HH-Inc 0.005*** HH-Inc −0.00003 −0.00004*** 0.0000* −0.003*** 0.0004*** −0.00004*** HH-Size 0.022*** HH-Size −0.006*** −0.003*** 0.002*** −0.028*** 0.006*** −0.0001** Black 0.141*** Black −0.033*** 0.007*** 0.008*** 0.015*** −0.005*** 0.000003 Asian 0.056 Asian −0.085*** 0.017*** 0.004*** 0.004*** 0.052*** −0.0007** Other Race −0.027*** 0.004*** 0.012*** −0.005*** 0.006*** −0.002*** Unemployed 0.0222 Unemployed 0.035*** 0.005*** −0.001 −0.006** 0.020*** −0.001** Max-Educ 0.174*** Max-Educ 0.003*** −0.0003 −0.005*** 0.029*** 0.007*** −0.0003*** NoMaleH −1.167*** NoMaleH 0.022*** 0.008*** 0.010***  −−0.0003*** −0.033*** 0.0001 NoFemH −0.793*** NoFemH 0.035*** 0.008*** −0.004*** −0.0002*** −0.024*** 0.008*** MEmp-FT 0.00511 MEmp-FT −0.006*** −0.007*** 0.006*** 0.005*** 0.003** 0.002*** FEmp-FT −0.138*** FEmp-FT 0.0008 −0.005*** 0.007*** −0.009*** 0.005*** 0.002*** MFEmp-FT −0.087*** MFEmp-FT 0.014*** 0.004*** −0.009*** −0.018*** −0.006*** −0.002*** WIC-Ever 0.0308 Inv. Mills −29.40*** 0.819*** 11.41*** −0.017*** 105.8*** −2.749*** Constant yes Constant yes yes yes yes yes yes Y and Q FE yes Y and Q FE yes yes yes yes yes yes Occup. FE yes yes yes yes yes yes USDAScore FSupermarket FDrug FMass FSupercenter FClub FConvenience Equation (2): Est. Coeff. Equations (1): Est. Coeff. Est. Coeff. Est. Coeff. Est. Coeff. Est. Coeff. Est. Coeff. FSupermarket 1.859*** P^Supermarket 4.707*** 1.140*** 3.636*** 0.029*** 3.596*** −0.215*** FDrug −7.668*** P^Drug 4.158*** −0.452*** −0.827*** −0.0003*** −4.184*** 0.168*** FMass 0.728 P^Mass −4.522*** −0.104** −0.230*** −0.0002*** 3.566*** −0.234*** FSupercenter 2.336*** P^Supermcenter 3.379*** −0.147*** 0.566*** 0.005*** −0.460*** 0.059*** FClub −0.682 P^Club 2.531*** 1.196*** −1.234*** −0.009*** −2.538*** 0.026 FConvenience −20.48*** P^Convenience 1.424*** 0.013 0.313*** −0.018*** −0.208*** −0.062*** P^UnHealthy 2.246*** P^Other 4.073*** 0.585*** −0.238*** −0.017*** −2.470*** 0.247*** HHIFood −0.111*** SSMarket/Conven −0.041*** −0.004*** −0.002*** −0.006** 0.0233*** −0.002*** SSMarket/SCent 0.0003*** 0.00003*** −0.00003*** −0.016*** 0.0002*** −0.00001 HH-Inc 0.005*** HH-Inc −0.00003 −0.00004*** 0.0000* −0.003*** 0.0004*** −0.00004*** HH-Size 0.022*** HH-Size −0.006*** −0.003*** 0.002*** −0.028*** 0.006*** −0.0001** Black 0.141*** Black −0.033*** 0.007*** 0.008*** 0.015*** −0.005*** 0.000003 Asian 0.056 Asian −0.085*** 0.017*** 0.004*** 0.004*** 0.052*** −0.0007** Other Race −0.027*** 0.004*** 0.012*** −0.005*** 0.006*** −0.002*** Unemployed 0.0222 Unemployed 0.035*** 0.005*** −0.001 −0.006** 0.020*** −0.001** Max-Educ 0.174*** Max-Educ 0.003*** −0.0003 −0.005*** 0.029*** 0.007*** −0.0003*** NoMaleH −1.167*** NoMaleH 0.022*** 0.008*** 0.010***  −−0.0003*** −0.033*** 0.0001 NoFemH −0.793*** NoFemH 0.035*** 0.008*** −0.004*** −0.0002*** −0.024*** 0.008*** MEmp-FT 0.00511 MEmp-FT −0.006*** −0.007*** 0.006*** 0.005*** 0.003** 0.002*** FEmp-FT −0.138*** FEmp-FT 0.0008 −0.005*** 0.007*** −0.009*** 0.005*** 0.002*** MFEmp-FT −0.087*** MFEmp-FT 0.014*** 0.004*** −0.009*** −0.018*** −0.006*** −0.002*** WIC-Ever 0.0308 Inv. Mills −29.40*** 0.819*** 11.41*** −0.017*** 105.8*** −2.749*** Constant yes Constant yes yes yes yes yes yes Y and Q FE yes Y and Q FE yes yes yes yes yes yes Occup. FE yes yes yes yes yes yes Notes: *** Statistically significant at the 99% level; ** at the 95% level; and * at the 90% level. First-stage results for F variables in equation (1) not presented. Table 5 3SLS Results, 2004–2010, Full Sample USDAScore FSupermarket FDrug FMass FSupercenter FClub FConvenience Equation (2): Est. Coeff. Equations (1): Est. Coeff. Est. Coeff. Est. Coeff. Est. Coeff. Est. Coeff. Est. Coeff. FSupermarket 1.859*** P^Supermarket 4.707*** 1.140*** 3.636*** 0.029*** 3.596*** −0.215*** FDrug −7.668*** P^Drug 4.158*** −0.452*** −0.827*** −0.0003*** −4.184*** 0.168*** FMass 0.728 P^Mass −4.522*** −0.104** −0.230*** −0.0002*** 3.566*** −0.234*** FSupercenter 2.336*** P^Supermcenter 3.379*** −0.147*** 0.566*** 0.005*** −0.460*** 0.059*** FClub −0.682 P^Club 2.531*** 1.196*** −1.234*** −0.009*** −2.538*** 0.026 FConvenience −20.48*** P^Convenience 1.424*** 0.013 0.313*** −0.018*** −0.208*** −0.062*** P^UnHealthy 2.246*** P^Other 4.073*** 0.585*** −0.238*** −0.017*** −2.470*** 0.247*** HHIFood −0.111*** SSMarket/Conven −0.041*** −0.004*** −0.002*** −0.006** 0.0233*** −0.002*** SSMarket/SCent 0.0003*** 0.00003*** −0.00003*** −0.016*** 0.0002*** −0.00001 HH-Inc 0.005*** HH-Inc −0.00003 −0.00004*** 0.0000* −0.003*** 0.0004*** −0.00004*** HH-Size 0.022*** HH-Size −0.006*** −0.003*** 0.002*** −0.028*** 0.006*** −0.0001** Black 0.141*** Black −0.033*** 0.007*** 0.008*** 0.015*** −0.005*** 0.000003 Asian 0.056 Asian −0.085*** 0.017*** 0.004*** 0.004*** 0.052*** −0.0007** Other Race −0.027*** 0.004*** 0.012*** −0.005*** 0.006*** −0.002*** Unemployed 0.0222 Unemployed 0.035*** 0.005*** −0.001 −0.006** 0.020*** −0.001** Max-Educ 0.174*** Max-Educ 0.003*** −0.0003 −0.005*** 0.029*** 0.007*** −0.0003*** NoMaleH −1.167*** NoMaleH 0.022*** 0.008*** 0.010***  −−0.0003*** −0.033*** 0.0001 NoFemH −0.793*** NoFemH 0.035*** 0.008*** −0.004*** −0.0002*** −0.024*** 0.008*** MEmp-FT 0.00511 MEmp-FT −0.006*** −0.007*** 0.006*** 0.005*** 0.003** 0.002*** FEmp-FT −0.138*** FEmp-FT 0.0008 −0.005*** 0.007*** −0.009*** 0.005*** 0.002*** MFEmp-FT −0.087*** MFEmp-FT 0.014*** 0.004*** −0.009*** −0.018*** −0.006*** −0.002*** WIC-Ever 0.0308 Inv. Mills −29.40*** 0.819*** 11.41*** −0.017*** 105.8*** −2.749*** Constant yes Constant yes yes yes yes yes yes Y and Q FE yes Y and Q FE yes yes yes yes yes yes Occup. FE yes yes yes yes yes yes USDAScore FSupermarket FDrug FMass FSupercenter FClub FConvenience Equation (2): Est. Coeff. Equations (1): Est. Coeff. Est. Coeff. Est. Coeff. Est. Coeff. Est. Coeff. Est. Coeff. FSupermarket 1.859*** P^Supermarket 4.707*** 1.140*** 3.636*** 0.029*** 3.596*** −0.215*** FDrug −7.668*** P^Drug 4.158*** −0.452*** −0.827*** −0.0003*** −4.184*** 0.168*** FMass 0.728 P^Mass −4.522*** −0.104** −0.230*** −0.0002*** 3.566*** −0.234*** FSupercenter 2.336*** P^Supermcenter 3.379*** −0.147*** 0.566*** 0.005*** −0.460*** 0.059*** FClub −0.682 P^Club 2.531*** 1.196*** −1.234*** −0.009*** −2.538*** 0.026 FConvenience −20.48*** P^Convenience 1.424*** 0.013 0.313*** −0.018*** −0.208*** −0.062*** P^UnHealthy 2.246*** P^Other 4.073*** 0.585*** −0.238*** −0.017*** −2.470*** 0.247*** HHIFood −0.111*** SSMarket/Conven −0.041*** −0.004*** −0.002*** −0.006** 0.0233*** −0.002*** SSMarket/SCent 0.0003*** 0.00003*** −0.00003*** −0.016*** 0.0002*** −0.00001 HH-Inc 0.005*** HH-Inc −0.00003 −0.00004*** 0.0000* −0.003*** 0.0004*** −0.00004*** HH-Size 0.022*** HH-Size −0.006*** −0.003*** 0.002*** −0.028*** 0.006*** −0.0001** Black 0.141*** Black −0.033*** 0.007*** 0.008*** 0.015*** −0.005*** 0.000003 Asian 0.056 Asian −0.085*** 0.017*** 0.004*** 0.004*** 0.052*** −0.0007** Other Race −0.027*** 0.004*** 0.012*** −0.005*** 0.006*** −0.002*** Unemployed 0.0222 Unemployed 0.035*** 0.005*** −0.001 −0.006** 0.020*** −0.001** Max-Educ 0.174*** Max-Educ 0.003*** −0.0003 −0.005*** 0.029*** 0.007*** −0.0003*** NoMaleH −1.167*** NoMaleH 0.022*** 0.008*** 0.010***  −−0.0003*** −0.033*** 0.0001 NoFemH −0.793*** NoFemH 0.035*** 0.008*** −0.004*** −0.0002*** −0.024*** 0.008*** MEmp-FT 0.00511 MEmp-FT −0.006*** −0.007*** 0.006*** 0.005*** 0.003** 0.002*** FEmp-FT −0.138*** FEmp-FT 0.0008 −0.005*** 0.007*** −0.009*** 0.005*** 0.002*** MFEmp-FT −0.087*** MFEmp-FT 0.014*** 0.004*** −0.009*** −0.018*** −0.006*** −0.002*** WIC-Ever 0.0308 Inv. Mills −29.40*** 0.819*** 11.41*** −0.017*** 105.8*** −2.749*** Constant yes Constant yes yes yes yes yes yes Y and Q FE yes Y and Q FE yes yes yes yes yes yes Occup. FE yes yes yes yes yes yes Notes: *** Statistically significant at the 99% level; ** at the 95% level; and * at the 90% level. First-stage results for F variables in equation (1) not presented. Our main interest focuses on results in the first and second columns of table 5, which correspond to equation (2) and the factors that influence the USDAScore.8 Here, regression coefficients can be interpreted as the impacts on food-purchase healthfulness. The impact of increasing store-format shares (F) are divided into positive and negative effects on the USDAScore. Increasing the expenditure shares at supermarkets and supercenters is associated with increased healthfulness; conversely, increasing expenditure shares at drug stores and convenience stores is associated with decreased healthfulness. Some of these results, such as the positive estimate for FSupermarket and the negative estimate for FConvenience, corroborate most previous research. However, the supercenter and club store results are noteworthy, each for different reasons. The FClub result is noteworthy because it shows an increasing expenditure share at club stores has no statistical impact on the USDAScore despite the fact that table 3 and figure 1 suggest that club store purchases are healthier, on average, than those made at other formats. The lack of statistical significance may, in part, reflect a low overall share of food purchases in club stores. On the other hand, the positive association between USDAScore and FSupercenter speaks to the open empirical question of whether or not supercenters are associated with healthy or unhealthy food purchases. Our results suggest increasing the share of the food dollar spent at supercenters can lead to healthier food purchases. This result is discussed in the next section. Most other results for estimated equation (1) are generally straightforward: (i) A higher price for unhealthy food ( P^Unhealthy) is associated with healthier food purchases.9 (ii) Increased concentration in food retailing (HHIFood) is associated with less healthier food purchases. (iii) Higher household income, larger household size, and higher education are all associated with healthier food purchases. (iv) The lack of a male or female household head has a negative association with healthy food purchases. (v) Full-time employment by a female head, either alone or in conjunction with full-time employment by a male head, is associated with less healthy purchases. (v) Finally, neither unemployment status nor current or former WIC program status has a statistical significance for the full data sample. The last seven columns of table 5, which correspond to equation (1), provide insight into the factors that influence the household-level expenditure shares across store formats. With two exceptions, the own-price coefficients are negative and statistically different from zero. The own-price estimate for club stores has the greatest magnitude, suggesting that this expenditure share is most sensitive to own price. The own-price coefficient for supermarkets and supercenters, the two exceptions, are positive, suggesting that higher supermarket and supercenter prices are associated with higher shares. This result may be an artifact of the high expenditure share for supermarkets and supercenters relative to other store formats. Cross-price estimates are a mixture of positive and negative, indicating that store formats can be considered substitutes and complements. For example, in the FMass column, the cross price estimate for P^Supermcenter is 0.566, indicating that households will substitute towards mass merchandizers if prices at supercenters increase. On the other hand, in the FSupercenter column, the cross price estimate for club stores is −0.009, indicating that households will buy slightly less from supercenters if prices at club stores increase. A number of store formats, that is, drug stores, mass merchandisers, club stores, convenience stores, and other stores all appear to be complementary to supercenters. The food environment variables also have both positive and negative effects depending on the store-format share. Increasing the supermarket-to-convenience store count ratio leads to higher expenditure shares at club stores, but lower shares at all other formats. However, increasing the supermarket-to-supercenter store count ratio leads to higher expenditure shares at supermarkets, drug stores, and club stores, but lower shares at mass merchandisers and supercenters. Other household demographic variables also have mixed results depending on the particular store format. Household size leads to higher shares at mass merchandisers and club stores, and lower shares at other formats. Household income leads to higher shares at club stores and mass merchandisers, and lower shares at all other formats. We investigate household income’s role in greater detail by reestimating equations (1) and (2) simultaneously for three subsamples based on income percentiles: one subsample for household incomes less than or equal to the 25th percentile, one where income is within the two middle quartiles, and the final one where income is greater than or equal to the 75th percentile. Abridged results just for equation (1) are presented in table 6. Table 6. 3SLS Results for USDAScore, Full Sample and Household Income Subsamples USDAScore Full Sample < 25th percentile HH Inc Middle 50 percentile HH Inc ≥ 75th percentile HH Inc FSupermarket 1.859*** 3.829** 2.044** −1.874* FDrug −7.668*** 3.209 −1.031 −7.573* FMass 0.728 7.720* 4.874*** 3.217 FSupercenter 2.336*** 0.803 3.386*** 0.903 FClub −0.682 3.329 −0.729 −5.243*** FConvenience −20.48*** 2.839 −18.50** −70.89*** P^UnHealthy 2.246*** 0.629 1.175* 2.466*** HHIFood −0.111*** 0.0427 −0.0974* −0.238*** HH-Inc 0.005*** 0.0198*** 0.0101*** 0.00258*** HH-Size 0.022*** 0.144*** 0.00629 0.0469*** Black 0.141*** 0.108 0.078* −0.129** Asian 0.056 −0.468 −0.0445 −0.0959* Unemployed 0.0222 −0.137** 0.0645** 0.146** Max-Educ 0.174*** 0.159*** 0.167*** 0.187*** NoMaleH −1.167*** −0.992*** −1.222*** −1.313*** NoFemH −0.793*** −0.595*** −0.853*** −0.546*** MEmp-FT 0.00511 −0.035 −0.031 0.0775 FEmp-FT −0.138*** −0.309*** −0.202*** −0.084** MFEmp-FT −0.087*** 0.167 −0.076** −0.102** WIC-Ever 0.0308 −0.032 0.050 0.036 Constant yes yes yes yes Y and Q FE yes yes yes yes USDAScore Full Sample < 25th percentile HH Inc Middle 50 percentile HH Inc ≥ 75th percentile HH Inc FSupermarket 1.859*** 3.829** 2.044** −1.874* FDrug −7.668*** 3.209 −1.031 −7.573* FMass 0.728 7.720* 4.874*** 3.217 FSupercenter 2.336*** 0.803 3.386*** 0.903 FClub −0.682 3.329 −0.729 −5.243*** FConvenience −20.48*** 2.839 −18.50** −70.89*** P^UnHealthy 2.246*** 0.629 1.175* 2.466*** HHIFood −0.111*** 0.0427 −0.0974* −0.238*** HH-Inc 0.005*** 0.0198*** 0.0101*** 0.00258*** HH-Size 0.022*** 0.144*** 0.00629 0.0469*** Black 0.141*** 0.108 0.078* −0.129** Asian 0.056 −0.468 −0.0445 −0.0959* Unemployed 0.0222 −0.137** 0.0645** 0.146** Max-Educ 0.174*** 0.159*** 0.167*** 0.187*** NoMaleH −1.167*** −0.992*** −1.222*** −1.313*** NoFemH −0.793*** −0.595*** −0.853*** −0.546*** MEmp-FT 0.00511 −0.035 −0.031 0.0775 FEmp-FT −0.138*** −0.309*** −0.202*** −0.084** MFEmp-FT −0.087*** 0.167 −0.076** −0.102** WIC-Ever 0.0308 −0.032 0.050 0.036 Constant yes yes yes yes Y and Q FE yes yes yes yes Notes: *** Statistically significant at the 99% level; ** at the 95% level; and * at the 90% level. Neither first-stage nor 3SLS results for equation (1)’s F variables are presented. Table 6. 3SLS Results for USDAScore, Full Sample and Household Income Subsamples USDAScore Full Sample < 25th percentile HH Inc Middle 50 percentile HH Inc ≥ 75th percentile HH Inc FSupermarket 1.859*** 3.829** 2.044** −1.874* FDrug −7.668*** 3.209 −1.031 −7.573* FMass 0.728 7.720* 4.874*** 3.217 FSupercenter 2.336*** 0.803 3.386*** 0.903 FClub −0.682 3.329 −0.729 −5.243*** FConvenience −20.48*** 2.839 −18.50** −70.89*** P^UnHealthy 2.246*** 0.629 1.175* 2.466*** HHIFood −0.111*** 0.0427 −0.0974* −0.238*** HH-Inc 0.005*** 0.0198*** 0.0101*** 0.00258*** HH-Size 0.022*** 0.144*** 0.00629 0.0469*** Black 0.141*** 0.108 0.078* −0.129** Asian 0.056 −0.468 −0.0445 −0.0959* Unemployed 0.0222 −0.137** 0.0645** 0.146** Max-Educ 0.174*** 0.159*** 0.167*** 0.187*** NoMaleH −1.167*** −0.992*** −1.222*** −1.313*** NoFemH −0.793*** −0.595*** −0.853*** −0.546*** MEmp-FT 0.00511 −0.035 −0.031 0.0775 FEmp-FT −0.138*** −0.309*** −0.202*** −0.084** MFEmp-FT −0.087*** 0.167 −0.076** −0.102** WIC-Ever 0.0308 −0.032 0.050 0.036 Constant yes yes yes yes Y and Q FE yes yes yes yes USDAScore Full Sample < 25th percentile HH Inc Middle 50 percentile HH Inc ≥ 75th percentile HH Inc FSupermarket 1.859*** 3.829** 2.044** −1.874* FDrug −7.668*** 3.209 −1.031 −7.573* FMass 0.728 7.720* 4.874*** 3.217 FSupercenter 2.336*** 0.803 3.386*** 0.903 FClub −0.682 3.329 −0.729 −5.243*** FConvenience −20.48*** 2.839 −18.50** −70.89*** P^UnHealthy 2.246*** 0.629 1.175* 2.466*** HHIFood −0.111*** 0.0427 −0.0974* −0.238*** HH-Inc 0.005*** 0.0198*** 0.0101*** 0.00258*** HH-Size 0.022*** 0.144*** 0.00629 0.0469*** Black 0.141*** 0.108 0.078* −0.129** Asian 0.056 −0.468 −0.0445 −0.0959* Unemployed 0.0222 −0.137** 0.0645** 0.146** Max-Educ 0.174*** 0.159*** 0.167*** 0.187*** NoMaleH −1.167*** −0.992*** −1.222*** −1.313*** NoFemH −0.793*** −0.595*** −0.853*** −0.546*** MEmp-FT 0.00511 −0.035 −0.031 0.0775 FEmp-FT −0.138*** −0.309*** −0.202*** −0.084** MFEmp-FT −0.087*** 0.167 −0.076** −0.102** WIC-Ever 0.0308 −0.032 0.050 0.036 Constant yes yes yes yes Y and Q FE yes yes yes yes Notes: *** Statistically significant at the 99% level; ** at the 95% level; and * at the 90% level. Neither first-stage nor 3SLS results for equation (1)’s F variables are presented. In almost every case, the middle subsample generates similar results for the full sample. However, results for the low-income and high-income subsamples vary considerably, both compared to each other and compared to the middle-income subsample. For example, low-income subsample results, like the middle-quartile subsample results, show that increased expenditure shares at supermarkets are associated with a higher USDAScore. However, the high-income subsample results show that increased shares at supermarkets are unexpectedly associated with a lower USDAScore. For the low-income subsample, increased shares at supermarkets and supercenters are associated with healthier food purchases. Results for all other formats are not statistically significant. Alternatively, for the high-income subsample, increased shares at drug stores, club stores, convenience supermarkets, and even supermarkets are all associated with less healthy food purchases. These results suggest that high-income consumers may be getting healthy food from places not in the dataset. Table 6 also shows that unhealthy prices do not play a strong role for the low-income subsample. However, they do play an increasing strong role in the middle- and high-income subsamples. Industry concentration, as reflected by HHIFood, has a negative effect on USDAScore only for the two higher income subsamples. Education is consistently positive across all subsamples. The lack of a male or female head and full-time employment status is also generally consistent across all subsamples. Discussion and Future Work American shoppers increasingly have a large number of retail formats from which to choose when purchasing food at home. Our study is the first, of which we are aware, to investigate and quantify the impact that alternative store formats have on the healthfulness of households’ food purchases. We extend related streams of research in two important ways. First, we relax the assumption that all stores other than conventional supermarkets, or all stores of the same approximate size, have the same impacts on food purchases. And second, we examine total quarterly household shopping baskets, as opposed to studying select product categories, to measure overall adherence to the DGA in order to get a clearer picture of how purchase decisions may be shaped by store choice. Using a large and highly detailed data set, we find that increased purchases at convenience stores and drug stores result in food choices less in line with the DGA. This finding is notable as these small-store formats, especially convenience stores, have been the subject of a number of nutritional, sociological, and economics studies on dietary quality and food access, and we discuss a number of reasons why this finding might have arisen. Additionally, we corroborate the consensus of research across a range of disciplines which argues that consumers purchase more healthful foods at larger stores with greater selection and, in many cases, lower prices per unit. Dietary quality, as reflected by the USDAScore, is positively associated with higher expenditure shares at supermarkets and supercenters. This positive association is well known for supermarkets, but a somewhat open question for supercenters. From a policy perspective, our study has the potential to inform the unanswered questions regarding supermarket “interventions.” A number of studies (Kristal et al. 1997; Wrigley et al. 2003; Cummins et al. 2005; Escaron et al. 2015) have examined the ex-post impacts of the subsidized introduction of large supermarkets into geographic areas where there previously were not any. Interventions and the interest surrounding them stem primary from the argument that low adherence to the DGA and related adverse health outcomes in the United States can be attributed, in part, to lack of food access. Most intervention analyses have identified little to no significant impact on food choices or dietary quality. Our findings suggest that store format or simply store characteristics may play an important role in determining the economic and health impacts of new stores in markets. In addition, as the food retailing landscape continues to evolve, supercenters, especially Wal-Mart, have been subject to much scrutiny and controversy as they have expanded throughout the United States. Basker (2007) notes that local governments have frequently attempted to block Wal-Mart’s entry into markets using zoning regulations or wage ordinances. From a policy perspective, however, the impacts of supercenters on factors such as food access, diet quality, and health outcomes remain unclear. Supercenters, which sell full lines of fruits, vegetables, and all foods necessary to meet the DGA, have been shown to result in lower average food prices within markets (Woo et al. 2001; Hausman and Leibtag 2007; Volpe and Lavoie 2008; Basker and Noel 2009). In this article, we find that increased expenditure shares supercenters (including Wal-Mart) can lead to healthier food purchases. Both the low- and high-income quartiles suggest that our results may be sensitive to income, and our supercenter result is one such case. The positive effect of supercenter expenditure share on food-purchase healthiness holds only for the full sample and the middle-income subsample but not the low- or high-income subsamples. Increased convenience store shares are strongly associated with lower food-purchase healthiness in all subsamples except the bottom quartile, whereas increased club store shares are strongly associated with lower food-purchase healthiness in only the high-income subsample. Similarly, higher unhealthy prices are positively associated with food-purchase healthiness in all samples but in low-income subsample. Industry concentration is negatively associated with food-purchase healthiness in all samples but the low-income subsample. Lastly, and perhaps most surprisingly, increased supermarket shares are negatively related to dietary quality in the top income quartile. Our findings leave much room for discussion and in many cases call for further research. The next important step in this line of inquiry is to understand the mechanisms that are driving our results. It seems plausible that supermarkets and supercenters stores offer larger selections of fresh produce and other healthful food options than do most other formats, but this may not be true across all stores or for all households. In addition, pilot studies of the Healthy Corner Store Initiative have conducted at several cities across the United States, and the goal of this initiative is to increase the availability and awareness of healthy foods in urban corner stores. Our convenience-store results suggest that initiatives of this sort may have success for the lowest income quartile. To fully characterize the relationship between dietary quality and food purchasing choices, however, we would need to increase the focus on heterogeneity across households or across store types. For example, to tease out the supercenter result from a policy perspective, we would need a sample more representative of low-income households. Additional work may also investigate potential differences across other demographic splits such as specific education, or racial or ethnic subsamples. We hope that our results therefore can motivate further research using store scanner data to investigate the link between store choices and food choices. Disclaimer A portion of this research was undertaken while Richard Volpe was an economist with the USDA Economic Research Service. The views expressed in this paper are those of the authors and may not be attributed to USDA or ERS. The authors accessed the Nielsen data via a third-party agreement with the Economic Research Service. Footnotes 1 An anonymous reviewer correctly notes that USDAScore may misrepresent food choices as it has the potential to incorrectly penalize households who “overshoot” (i.e., have higher observed shares than recommended shares) for CNPP food categories deemed healthy, or undershoot for CNPP categories deemed unhealthy. For this reason, we also calculate a calculate USDAScore+, which first designates CNPP categories as healthy or unhealthy and then disallows penalties for being “too healthy” or not “unhealthy enough”. In the results section, we note that penalties of these types are extremely rare in the Homescan data, so our USDAScore and USDAScore+ perform almost identically. 2 Technically, the store-format shares are also bounded by one. However, the number of quarterly household-level store-format shares attaining this upper bound is extremely small, so generally ignore right-censoring concerns. 3 Additionally, it is uncommon for households to attribute a large portion of their quarterly food dollars to some of the smaller nontraditional formats. Though not reported, we examined quarterly shopping baskets for those households that spend at least 90% of their food dollars at a single store format. For example, there were only 80 quarterly baskets in sample consisting predominantly of purchases made from convenience stores. 4 For one store format, convenience stores, we use the 90th percentile instead of the 75th percentile. The reason is that most expenditures shares for convenience stores are zero, i.e., both the median and 75th percentile share for convenience stores are zero. On the other hand, the 90th percentile share for convenience stores is 0.57%. 5 Our store counts, by format, are drawn directly from the Homescan data. Other data sources, such as the US Census of Business or Nielsen TDLinx, raise questions about the applicability to our sample in terms of geography or format definition. The major drawback of using Homescan store counts is that small formats outside of traditional grocery, particularly convenience stores, are necessarily undercounted. However, this is only an issue if the rate of undercounting varies in a systematic way across Scantrack markets, and we have no evidence that this is the case. 6 The HHI is calculated as the sum of squared market shares for the firms operating within an industry and market. Technically we calculated the HHI (10), using only the ten largest firms in each market. Given that food retail is highly concentrated locally, HHI exhibits much more variation than other commonly used concentration measures, such as the four-firm concentration ratio (CR4). 7 We experimented with using store counts, as drawn from the TDLinx data, to account for formats such as convenience stores and club stores, for which food sales data are not available. However, we encountered two issues with this strategy. First, these variables were almost never significant in determining USDAScore. And second, we deemed them too similar to the food environment variables included in (S), in that they measure food access and proximity rather than market structure or share. 8 We also estimated equations (1) and (2) simultaneously using USDAScore+, which was defined in an earlier footnote. Because the correlation between USDAScore and USDAScore+ is 0.94, and because 3SLS coefficients are extremely similar in both cases, we do not present the results with USDAScore+ in this article. However, these results are available from the authors upon request. 9 Our instrumental variables for prices pass the standard battery of tests, including the Hausman endogeneity test and the Sargan overidentification test, indicating that they are preferred to OLS. However, the usual caveats and assumptions to these tests apply. These results are available upon request. References Basker E. 2007 . The Causes and Consequences of Wal-Mart's Growth. The Journal of Economic Perspectives 21 3 : 177 – 98 . Google Scholar CrossRef Search ADS Basker E. , Noel M. . 2009 . The Evolving Food Chain: Competitive Effects of Wal‐Mart's Entry into the Supermarket Industry . Journal of Economics and Management Strategy 18 4 : 977 – 1009 . Google Scholar CrossRef Search ADS Bhattacharya J. , Bundorf M.K. . 2009 . The Incidence of the Healthcare Costs of Obesity . Journal of Health Economics 28 3 : 649 – 58 . Google Scholar CrossRef Search ADS PubMed Blattberg R.C. , Briesch R. , Fox E.J. . 1995 . How Promotions Work . Marketing Science 14 3 : G122 – G132 . Google Scholar CrossRef Search ADS Blisard N. , Stewart H. , Jolliffe D. . 2004 . Low-Income Households' Expenditures on Fruits and Vegetables. US Department of Agriculture, Economic Research Service, Agriculture Information Bulletin No. 792-5. Block J.P. , Scribner R.A. , DeSalvo K.B. . 2004 . Fast Food, Race/Ethnicity, and Income: A Geographic Analysis . American Journal of Preventative Medicine 27 3 : 211 – 17 . Bonanno A. , Lopez R.A. . 2009 . Competition Effects of Supermarket Services . American Journal of Agricultural Economics 91 3 : 555 – 68 . Google Scholar CrossRef Search ADS Bustillos B. , Sharkey J.R. , Anding J. , McIntosh A. . 2008 . Availability of More Healthful Food Alternatives in Traditional, Convenience, and Nontraditional Types of Food Stores in Two Rural Texas Counties . Journal of the American Dietetic Association 109 5 : 883 – 89 . Google Scholar CrossRef Search ADS Carlson A. , Frazao E. . 2012 . Are Healthy Foods More Expensive? It Depends on How You Measure the Price. EIB-96, US Department of Agriculture, Economic Research Service, October 2010. Carlson A. , Lino M. , Fungwe T. . 2007 . The Low-Cost, Moderate-Cost, and Liberal Food Plans. US Department of Agriculture, Center for Nutrition Policy and Promotion. Carpenter J.M. , Moore M. . 2006 . Consumer Demographics, Store Attributes, and Retail Format Choice in the US Grocery Market . International Journal of Retail and Distribution Management 34 6 : 434 – 52 . Google Scholar CrossRef Search ADS Chen D. , Jaenicke E.C. , Volpe R. . 2016 . Relationships between Home and Neighborhood Food Environments and Obesity: A National Multilevel Analysis Using Scanner Data . American Journal of Public Health . 106 5 : 881 – 88 . Google Scholar CrossRef Search ADS PubMed Chung C. , Myers S.L. . 1999 . Do the Poor Pay More for Food? An Analysis of Grocery Store Availability and Food Disparaties . Journal of Consumer Affairs 33 : 276 – 96 . Google Scholar CrossRef Search ADS Courtemanche C. , Carden A. . 2011 . Supersizing Supercenters? The Impact of Wal-Mart Supercenters on Body Mass Index and Obesity . Journal of Urban Economics 69 2 : 165 – 81 . Google Scholar CrossRef Search ADS Cummins S. , Pettigrew M. , Higgins C. , Findlay A. , Sparks L. 2005 . Large Scale Food Retailing as an Intervention for Diet and Health: Quasi-experimental Evaluation of a Natural Experiment . Journal of Epidemiology and Community Health 59 : 1035 – 104 . Google Scholar CrossRef Search ADS PubMed Cutler D.M. , Glaeser E.L. , Shapiro J.M. . 2003 . Why Have Americans Become More Obese? Journal of Economic Perspectives 17 3 : 93 – 118 . Google Scholar CrossRef Search ADS Dong D. , Lin B. . 2009 . Fruit and Vegetable Consumption by Low-Income Americans: Would a Price Reduction Make a Difference? ERR-70, US Department of Agriculture, Economic Research Service, January, 2009. Escaron A.L. , Martinez-Donate A.P. , Riggall A.J. , Meinen A. , Hall B. , Nieto F.J. , Nitzke S. . 2015. Developing and Implementing “Waupaca Eating Smart” A Restaurant and Supermarket Intervention to Promote Healthy Eating Through Changes in the Food Environment . Health Promotion Practice , p.1524839915612742. Einav L. , Leibtag E. , Nevo A. . 2008 . On the Accuracy of Nielsen Homescan Data. ERR-69, US Department of Agriculture, Economic Research Service, November 2008. Einav L. , Leibtag E. , Nevo A. 2010 . Recoding Dicrepancies in Nielsen Homescan Data: Are They Present and Do They Matter? Quantitative Marketing and Economics 8 2 : 207 – 39 . Google Scholar CrossRef Search ADS Food Institute. 2010 . “The Food Industry Review.” (2010 Edition). Fox E.J. , Montgomery A.L. , Lodish L.M. . 2004 . Consumer Shopping and Spending across Retail Formats . Journal of Business 77 2 : S25 – S60 . Google Scholar CrossRef Search ADS Galobardes B. , Morabia A. , Bernstein M.S. . 2000 . The Differential Effect of Education and Occupation on Body Mass and Overweight in a Sample of Working People of the General Population . Annals of Epidemiology 10 8 : 532 – 37 . Google Scholar CrossRef Search ADS PubMed Garn S.M. , Bailey S.M. , Cole P.E. , Higgins I.T. . 1977 . Level of Education, Level of Income, and Level of Fatness in Adults . American Journal of Clinical Nutrition 30 5 : 721 – 25 . Google Scholar CrossRef Search ADS PubMed Gauri D.K. , Trivedi M. , Grewal D. . 2008 . Understanding the Determinants of Retail Strategy: An Empirical Analysis . Journal of Retailing 84 3 : 256 – 67 . Google Scholar CrossRef Search ADS Goodman C.S. 1968 . Do the poor pay more? Journal of Marketing , 32 : 18 – 24 . Google Scholar CrossRef Search ADS Hausman J. 1996 . “The Valuation of Goods under Perfect and Imperfect Information.” Book chapter. The Economics of New Goods. Hausman J. , Leibtag E. , 2007 . Consumer benefits from increased competition in shopping outlets: Measuring the effect of Wal‐Mart . Journal of Applied Econometrics , 22 7 : 1157 – 77 . Google Scholar CrossRef Search ADS Kaufman P.R. , MacDonald J.M. , Lutz S.M. , Smallwood D.M. . 1997 . Do the Poor Pay More for Food? Item Selection and Price Differences Affect Low-Income Household Food Costs. Agricultural Economics Report No. (AER759) 32 pp, December 1997. Kristal A.R. , Goldenhar L. , Muldoon J. , Morton R.F. . 1997 . Evaluation of a Supermarket Intervention to Increase Consumption of Fruits and Vegetables. American Journal of Health Promotion 11 6 : 422 – 25 . Google Scholar CrossRef Search ADS PubMed Kuchler F. , Stewart H. . 2008 . Price Trends Are Similar for Fruits, Vegetables, and Snack Foods. US Department of Agriculture, Economic Research Service, Economic Research Report No. 55, March 2008. Lamm R.M. 1981 . Prices and Concentration in the Food Retailing Industry . Journal of Industrial Economics 30 1 : 67 – 78 . Google Scholar CrossRef Search ADS Leibtag E. , Barker C. , Dutko P. . 2010 . How Much Lower Are Prices at Discount Stores? ERR-105, US Department of Agriculture, Economic Research Service. Liese A.D. , Weis K.E. , Pluto D. , Smith E. , Lawson A. . 2007 . Food Store Types, Availability, and Cost of Foods in a Rural Environment . Journal of the American Dietetic Association 107 11 : 1916 – 23 . Google Scholar CrossRef Search ADS PubMed MacDonald J.M. , Nelson P.E. . 1991 . Do the Poor Still Pay More? Food Price Variations in Large Metropolitan Areas . Journal of Urban Economics 30 3 : 344 – 359 . Google Scholar CrossRef Search ADS Major M. 2013 . Supermarkets Emerge as Grocery Share Leaders. http://www.progressivegrocer.com/top-stories/headlines/trending-topics/id39214/supermarkets-emerge-as-grocery-share-leaders/. Accessed February 21, 2018. Mancino L. , Kuchler F. , Leibtag E. . 2008 . Getting Consumers to Eat More Whole-grains: The Role of Policy, Information, and Food Manufacturers . Food Policy 33 6 : 489 – 96 . Google Scholar CrossRef Search ADS Mancino L. , Newman C. . 2007 . Who Has Time to Cook? How Family Resources Influence Food Preparation. ERR-40, US Department of Agriculture, Economic Research Service, May, 2007. Market Force . 2011 . Market Force Study Finds Consumers Perceive ALDI as Grocery Price Leader. Martinez S.W. 2007 . The US Food Marketing System: Recent Developments, 1997–2006. ERR-42. USDept of Agriculture, Econ. Res. Serv. May 2007. Matsa D.A. 2011 . Competition and Product Quality in the Supermarket Industry . Quarterly Journal of Economics 126 3 : 1539 – 91 . Google Scholar CrossRef Search ADS PubMed Murray M.P. 2006 . Avoiding Invalid Instruments and Coping with Weak Instruments . The Journal of Economic Perspectives 20 4 : 111 – 32 . Google Scholar CrossRef Search ADS Powell L.M. , Bao Y. . 2009 . Food Prices, Access to Food Outlets, and Child Weight . Economics and Human Biology 7 1 : 64 – 72 . Google Scholar CrossRef Search ADS PubMed Richards T.J. , Hamilton S.F. . 2006 . Rivalry in Price and Variety Among Supermarket Retailers . American Journal of Agricultural Economics 88 3 : 710 – 26 . Google Scholar CrossRef Search ADS Rose D. , Richards R. . 2004 . Food Store Access and Household Fruit and Vegetable Use Among Participants in the US Food Stamp Program . Public Health Nutrition 7 8 : 1081 – 1088 . Google Scholar CrossRef Search ADS PubMed Seiders K. , Simonides C. , Tigert D.J. . 2000 . The Impact of Supercenters on Traditional Food Retailers in Four Markets . International Journal of Retail and Distribution Management 28 45 : 181 – 93 . Google Scholar CrossRef Search ADS Shonkwiler J.S. , Yen S.T. . 1999 . Two-Step Estimtion of a Censored System of Equations . American Journal of Agricultural Economics 81 4 : 972 – 82 . Google Scholar CrossRef Search ADS Todd J.E. , Mancino L. , Leibtag E. , Tripodo C. . 2010 . Methodology Behind the Quarterly Food-at-Home Price Database. US Department of Agriculture, Economic Research Service, Technical Bulletin No. 1926. Villas-Boas S. 2009 . An Empirical Investigation of the Welfare Effects of Banning Wholesale Price Discrimination . RAND Journal of Economics 40 1 : 20 – 46 . Google Scholar CrossRef Search ADS Volpe R. J. , Lavoie N. . 2008 . The Effect of Wal-Mart Supercenters on Grocery Prices in New England . Applied Economic Perspectives and Policy 30 1 : 4 – 26 . Volpe R.J. , Okrent A. . 2012 . Assessing the Healthfulness of Consumers’ Grocery Purchases. EIB-102, US Department of Agriculture, Economic Research Service, November 2012. Volpe R.J. , Okrent A. , Leibtag E. . 2013 . The Effect of Supercenter-format Stores on the Healthfulness of Consumers’ Grocery Purchases . American Journal of Agricultural Economics 95 3 : 568 – 89 . Google Scholar CrossRef Search ADS Wrigley N. , Warm D. , Margetts B. . 2003 . Deprivation, Diet, and Food-retail Access: Findings from the Leeds ‘food deserts' Study. Environment and Planning A 35 1 : 151 – 88 . Google Scholar CrossRef Search ADS Woo B. J. , Huang C. L. , Epperson J. E. , Cude B. . 2001 . Effect of a New Wal-Mart Supercenter on Local Retail Food Prices . Journal of Food Distribution Research 32 1 : 173 – 81 . Xie B. , Gilliland F.D. , Ly Y. , Rockett H.R. . 2003 . Effects of Ethnicity, Family Income, and Education on Dietary Intake among Adolescents . Preventive Medicine 36 1 : 30 – 40 . Google Scholar CrossRef Search ADS PubMed Yu C.-C. , Connor J.M. . 2002 . The Price-Concentration Relationship in Retailing: Retesting Newmark . Agribusiness 18 4 : 413 – 26 . Google Scholar CrossRef Search ADS Zenk S.N. , Schulz A.J. , Hollis-Neely T. , Campbell R.T. , Holmes N. , Watkins G. , Nwankwo R. , Odoms-Young A. . 2005 . Fruit and Vegetable Intake in African Americans: Income and Store Characteristics. American Journal of Preventative Medicine 29 1 : 1 – 9 . Google Scholar CrossRef Search ADS © The Author(s) 2018. Published by Oxford University Press on behalf of the Agricultural and Applied Economics Association. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Applied Economic Perspectives and Policy Oxford University Press

Store Formats, Market Structure, and Consumers’ Food Shopping Decisions

Loading next page...
 
/lp/ou_press/store-formats-market-structure-and-consumers-food-shopping-decisions-3oDc9OlwAH
Publisher
Oxford University Press
Copyright
© The Author(s) 2018. Published by Oxford University Press on behalf of the Agricultural and Applied Economics Association. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com
ISSN
2040-5790
eISSN
2040-5804
D.O.I.
10.1093/aepp/ppx033
Publisher site
See Article on Publisher Site

Abstract

Abstract A growing literature in health and nutrition suggests that healthy foods are less available and more expensive at nontraditional store formats such as supercenters, convenience stores, and drug stores. We use Nielsen Homescan data to investigate the relationship between store format and the healthfulness of consumers’ grocery shopping. Accounting for a rich set of controls, as well as food retail market structure, we simultaneously estimate the healthfulness of consumers’ food purchases and the shares of food expenditure at traditional and nontraditional store formations. We find that healthier food choices are generally associated with higher food expenditure shares at supermarkets and supercenters and lower shares at drug stores and convenience stores. In addition, market concentration has a negative effect on shopping healthfulness. dietary quality, consumer choices, food expenditures, store formats, market structure, household scanner data D12, L11, Q12, I10 Introduction Obesity, overweight status, and diabetes are three examples of prevalent health concerns that can be attributed, at least in part, to people’s food choices. Each has been cited as a significant cause of death and source of healthcare costs in the United States (Bhattacharya and Bundorf 2009). It is widely understood that consumers do not, on average, meet USDA guidelines for healthy eating. Americans eat too few fruits, vegetables, and whole grains while eating too many fats, added sugars, and refined grains (Mancino et al. 2008; Dong and Lin 2009). Economists, social scientists, and nutritionists have devoted much research to the determinants of poor diet choices and possible solutions. One area of research that has received relatively little attention is the impact of store formats and store characteristics on consumers’ food purchasing decisions. Consumers are increasingly shopping at nontraditional store formats, which include supercenters, club stores, mass merchandisers that sell food, drug stores, and convenience stores. Martinez (2007) showed that the share of consumers’ food-at-home expenditures at outlets other than supermarkets increased from 17% to 31% from 1994 to 2005. This study uses Nielsen Homescan consumer purchase data to investigate the relationship between store formats and the healthfulness of consumers’ grocery shopping. Because store-format choice and food-purchase healthfulness are interrelated decisions, we estimate them in a simultaneous system. Taking care to control for probable confounding factors such as the food environment, demographics, and relative prices, we estimate how store format choices, as reflected by format-specific food expenditure shares, affect the healthfulness of households’ food purchases. We generally find that healthier food choices are associated with higher food expenditure shares at supermarkets and supercenters and lower shares at drug stores and convenience stores. In addition, increased retail food industry concentration has a negative effect on shopping healthfulness. Background on Store Choice and Food Shopping Decisions There are a number of reasons to expect that store formats might significantly shape consumers’ food choices. For instance, product assortments, average prices, and promotional strategies differ substantially across store formats (e.g., Blattberg et al. 1995; Seiders et al. 2000; Leibtag et al. 2010; Market Force 2011). A considerable literature in health and nutrition (e.g., Liese et al. 2007) is devoted to the availability and price of foods at different types of stores as they differ according to socioeconomic indicators. A regularity uncovered by a review of these studies is that healthful options, measured broadly by fruits, vegetables, and other whole foods, are fewer in number and higher in price at small retailers such a drug stores and convenience stores than they are at supermarkets. In rural Texas, Bustillos et al. (2008) find that the variety of healthful foods, including canned and frozen options, is significantly higher at supermarkets than at convenience stores, dollar stores, or mass merchandisers. For rural South Carolina, Liese et al. (2007) find that healthful food options are significantly less numerous and higher priced at convenience stores and other smaller formats. Studying a nationally representative of Food Stamp participants, Rose and Richards (2004) conclude that access to large supermarkets is directly correlated with the consumption of fruits and vegetables. Zenk et al. (2005), studying the purchase patterns of African American women, find that lower-income shoppers are more likely to shop at nontraditional formats and, as a result of differences in product availability across formats, purchase less fresh produce. A few studies go further and link the food environment to health outcomes. Powell and Bao (2009) find an inverse relationship between body mass index (BMI) and supermarket availability across metropolitan areas in the United States. Chen et al. (2016) show that household members residing in food deserts, areas defined in part by poor access to supermarkets, are more likely to be obese. A closely related stream of literature, which fits within in this category of studies on food choices by store format, examines food expenditures by income group. Goodman (1968), MacDonald and Nelson (1991), Kaufman et al. (1997), and Chung and Myers (1999) use a variety of empirical approaches to investigate whether poor households pay more for food, per unit, than do higher-income households. At the heart of each study is the notion that low-income households often lack access to large supermarkets and therefore shop at smaller stores with fewer choices and higher prices. Another relevant research topic, grounded in economics and marketing, investigates the determinants of store choice, particularly among formats. The study most similar in spirit and approach to our own is Fox et al. (2004), who find that much of the variation in household food expenditures, about 31%, can be explained by format choice, which in turn is determined more by promotional activities than by prices or location. Using a survey of representative households, Carpenter and Moore (2006) find that product assortment is more important for frequent shoppers at both supermarkets and club stores than for supercenter shoppers, whereas frequent supercenter shoppers value price competitiveness more than shoppers of alternative formats. Finally, a small group of research papers links market structure to food choice. For example, several studies identify a positive connection between food prices and market concentration (e.g., Lamm 1981; Yu and Connor 2002). However, supermarkets also compete along a variety of other dimensions, including variety (Richards and Hamilton 2006) and quality (Matsa 2011). Therefore, market concentration may shape consumers’ food choices through a variety of mechanisms, under the presumption that competition across firms decreases with concentration. In addition, the particular impacts of nontraditional store formats such as supercenters is a relatively new area of study. For example, Courtemanche and Carden (2011) attribute an increase in BMI and obesity in American cities to the opening of Wal-Mart Supercenters, while Volpe et al. (2013) find that increased supercenter market share within metropolitan areas is associated with decreased dietary quality among grocery purchases. Alternative Store Formats and Consumers’ Evolving Store Choices Table 1 summarizes the Food Institute’s (2010) descriptions of each of the store formats used in our study: supermarkets, drug stores, mass merchandisers, supercenters, club stores, convenience stores, and a category of other stores that consists mostly of dollar stores and military commissaries. While there are no universally accepted definitions and classifications of food retail store formats, throughout this study we use the store-format names provided by Nielsen, who collects the Homescan data. The format descriptions are drawn from a consensus of Progressive Grocer, the Food Marketing Institute, Nielsen, and the US Census for retail. Table 1 Descriptive Statistics for Food Retailing Store Formats, 2009 Format Description Number Stores (US) Avg. Store Area (sq. feet) Avg. No. SKUsa Avg. Weekly Sales Δ HH Penetration 2001–‘09 Supermarket Food retailer with greater than 9,000 sq. feet of selling space and $2 million annually in food sales. 40,205 52,100 45,500 $221,345 −2% Mass Merchandiser Department stores selling primarily general merchandise as well as a limited selection of foods, including perishables 3,683 66,400 95,000 $228,639 −18% Supercenter Large stores that combine department stores with full supermarkets 3,366 184,100 100,000 $903,311 19% Drug Store Stores featuring prescription-based pharmacies and generating at least 20% of revenues from food and general merchandise 24,660 12,300 5,400 $46,229 −6% Club Store Large stores selling food and general merchandise in bulk to consumers with paid memberships 1,304 130,500 5,100 $1,108,329 0% Convenience Store Similar to drug stores but without pharmacies. May or may not feature gas stations 150,704 2,600 4,200 $19,538 −7% Othersd A collection of smaller formats including dollar stores, direct-to-consumer outlets, and hospitals. 22,407 7,600 5,500 $20,228 5% Format Description Number Stores (US) Avg. Store Area (sq. feet) Avg. No. SKUsa Avg. Weekly Sales Δ HH Penetration 2001–‘09 Supermarket Food retailer with greater than 9,000 sq. feet of selling space and $2 million annually in food sales. 40,205 52,100 45,500 $221,345 −2% Mass Merchandiser Department stores selling primarily general merchandise as well as a limited selection of foods, including perishables 3,683 66,400 95,000 $228,639 −18% Supercenter Large stores that combine department stores with full supermarkets 3,366 184,100 100,000 $903,311 19% Drug Store Stores featuring prescription-based pharmacies and generating at least 20% of revenues from food and general merchandise 24,660 12,300 5,400 $46,229 −6% Club Store Large stores selling food and general merchandise in bulk to consumers with paid memberships 1,304 130,500 5,100 $1,108,329 0% Convenience Store Similar to drug stores but without pharmacies. May or may not feature gas stations 150,704 2,600 4,200 $19,538 −7% Othersd A collection of smaller formats including dollar stores, direct-to-consumer outlets, and hospitals. 22,407 7,600 5,500 $20,228 5% Source: The Food Institute Industry Review (2010). a The number of SKUs and average weekly sales include nonfood items and products not available at supermarkets. b In the 2010 Review this is a category only includes dollar stores when measuring penetration. Otherwise it reports a weighted average of dollar stores and military commissaries based on their expenditure shares in the Homescan data. Table 1 Descriptive Statistics for Food Retailing Store Formats, 2009 Format Description Number Stores (US) Avg. Store Area (sq. feet) Avg. No. SKUsa Avg. Weekly Sales Δ HH Penetration 2001–‘09 Supermarket Food retailer with greater than 9,000 sq. feet of selling space and $2 million annually in food sales. 40,205 52,100 45,500 $221,345 −2% Mass Merchandiser Department stores selling primarily general merchandise as well as a limited selection of foods, including perishables 3,683 66,400 95,000 $228,639 −18% Supercenter Large stores that combine department stores with full supermarkets 3,366 184,100 100,000 $903,311 19% Drug Store Stores featuring prescription-based pharmacies and generating at least 20% of revenues from food and general merchandise 24,660 12,300 5,400 $46,229 −6% Club Store Large stores selling food and general merchandise in bulk to consumers with paid memberships 1,304 130,500 5,100 $1,108,329 0% Convenience Store Similar to drug stores but without pharmacies. May or may not feature gas stations 150,704 2,600 4,200 $19,538 −7% Othersd A collection of smaller formats including dollar stores, direct-to-consumer outlets, and hospitals. 22,407 7,600 5,500 $20,228 5% Format Description Number Stores (US) Avg. Store Area (sq. feet) Avg. No. SKUsa Avg. Weekly Sales Δ HH Penetration 2001–‘09 Supermarket Food retailer with greater than 9,000 sq. feet of selling space and $2 million annually in food sales. 40,205 52,100 45,500 $221,345 −2% Mass Merchandiser Department stores selling primarily general merchandise as well as a limited selection of foods, including perishables 3,683 66,400 95,000 $228,639 −18% Supercenter Large stores that combine department stores with full supermarkets 3,366 184,100 100,000 $903,311 19% Drug Store Stores featuring prescription-based pharmacies and generating at least 20% of revenues from food and general merchandise 24,660 12,300 5,400 $46,229 −6% Club Store Large stores selling food and general merchandise in bulk to consumers with paid memberships 1,304 130,500 5,100 $1,108,329 0% Convenience Store Similar to drug stores but without pharmacies. May or may not feature gas stations 150,704 2,600 4,200 $19,538 −7% Othersd A collection of smaller formats including dollar stores, direct-to-consumer outlets, and hospitals. 22,407 7,600 5,500 $20,228 5% Source: The Food Institute Industry Review (2010). a The number of SKUs and average weekly sales include nonfood items and products not available at supermarkets. b In the 2010 Review this is a category only includes dollar stores when measuring penetration. Otherwise it reports a weighted average of dollar stores and military commissaries based on their expenditure shares in the Homescan data. In addition to the format descriptions, table 1 also includes format-specific descriptive statistics as well as estimates of the changes in household penetration, or the share of US households shopping at least some of the time at each of the formats, for the years 2000 through 2009. This statistic supports the narrative that Americans are increasingly shopping among nontraditional formats for their groceries. The total penetration for supermarkets fell slightly over this time, while it increased substantially for supercenters. Much of the increase in supercenter penetration is explained by the large-scale conversion of mass merchandisers to supercenters. The final store-format category, “others,” is an amalgamation of smaller formats defined by Nielsen. As of 2010, the breakdown of this category in Homescan is as follows: Forty percent of the receipts in this category come from dollar stores. Military commissaries and specialized department stores such as sporting goods and electronic stores account for 15% each. Online food shopping accounts for another 10%, and the remaining 20% is attributed to a collection of outlets such hospitals, vending machines, and fast food restaurants where consumers are able to buy snacks and drinks that are suitable for consumption at home. This category saw an increase in household penetration of 5%, the timing of which coincides with the rapid expansion of dollar stores during the recession of 2007–2009. Table 2 reports how consumers’ food spending has changed since 1999 based on data from the Census of Retail Trade, as compiled and organized by the USDA Economics Research Service (ERS). The table also includes the most comparable numbers as compiled for Progressive Grocer (PG) in Major (2013). There are obvious and important differences in the way the two sources chronicled in table 2 categorize and define retail food formats. However, both data sources corroborate the trend away from supermarkets and toward supercenters and club stores. The ERS numbers show the supermarket share of the US grocery dollar falling from 72.7% in 1999 to 63.8% in 2011. Meanwhile, the share for supercenters and club stores combined rose from 6% to 16%. Martinez (2007), also using US Census data, confirms that this trend began in earnest in 1994. The PG numbers show that the supermarket share falls from 72.3% in 1997 to 58.9% in 2011, while the share for warehouse clubs and supercenters grows from 9% to 23%. Table 2 Shares of Consumers’ Food Expenditures, by Store Format and Data Source Economic Research Service Year Supermarkets Convenience Stores Other Grocery Specialty Food Stores Warehouse Clubs and Supercenters Mass Merchandise Other Stores Home Deliveries, Mail Orders Direct to Consumera 1999 72.7 2.5 1.1 2.0 5.8 1.8 4.7 3.6 5.8 2000 70.9 2.5 1.4 2.0 7.2 1.7 4.9 3.4 6.0 2001 70.1 2.6 1.2 2.0 8.9 1.7 4.8 3.1 5.5 2002 67.4 2.6 1.3 2.0 11.7 1.6 4.8 3.0 5.6 2003 66.8 2.6 1.3 2.1 12.6 1.4 4.9 3.0 5.7 2004 66.3 2.7 0.8 2.3 13.4 1.2 4.9 2.8 5.6 2005 65.8 2.6 0.7 2.3 14.1 1.0 4.9 2.8 5.8 2006 65.5 2.6 0.8 2.3 14.7 0.9 4.9 2.8 5.5 2007 65.2 2.6 0.6 2.4 15.1 0.7 4.8 2.7 5.9 2008 64.9 2.5 0.7 2.4 15.6 0.7 4.8 2.6 5.9 2009 65.0 2.5 0.6 2.3 15.9 0.7 4.7 2.4 5.9 2010 64.4 2.5 0.9 2.3 16.1 0.6 4.8 2.4 5.9 2011 63.8 2.4 1.4 2.3 16.0 0.6 4.9 2.7 5.9 Progressive Grocer Supermarkets Pharmacy and Drug Stores Discount Deparment Stores Warehouse Clubs and Supercenters Mass Merchandise 1997 72.3 1.2 2.3 9.1 N/A 2002 66.0 1.2 2.2 15.6 0.7 2007 60.2 1.5 2.1 22.0 1.4 2011 58.9 1.5 1.7 23.2 1.5 Economic Research Service Year Supermarkets Convenience Stores Other Grocery Specialty Food Stores Warehouse Clubs and Supercenters Mass Merchandise Other Stores Home Deliveries, Mail Orders Direct to Consumera 1999 72.7 2.5 1.1 2.0 5.8 1.8 4.7 3.6 5.8 2000 70.9 2.5 1.4 2.0 7.2 1.7 4.9 3.4 6.0 2001 70.1 2.6 1.2 2.0 8.9 1.7 4.8 3.1 5.5 2002 67.4 2.6 1.3 2.0 11.7 1.6 4.8 3.0 5.6 2003 66.8 2.6 1.3 2.1 12.6 1.4 4.9 3.0 5.7 2004 66.3 2.7 0.8 2.3 13.4 1.2 4.9 2.8 5.6 2005 65.8 2.6 0.7 2.3 14.1 1.0 4.9 2.8 5.8 2006 65.5 2.6 0.8 2.3 14.7 0.9 4.9 2.8 5.5 2007 65.2 2.6 0.6 2.4 15.1 0.7 4.8 2.7 5.9 2008 64.9 2.5 0.7 2.4 15.6 0.7 4.8 2.6 5.9 2009 65.0 2.5 0.6 2.3 15.9 0.7 4.7 2.4 5.9 2010 64.4 2.5 0.9 2.3 16.1 0.6 4.8 2.4 5.9 2011 63.8 2.4 1.4 2.3 16.0 0.6 4.9 2.7 5.9 Progressive Grocer Supermarkets Pharmacy and Drug Stores Discount Deparment Stores Warehouse Clubs and Supercenters Mass Merchandise 1997 72.3 1.2 2.3 9.1 N/A 2002 66.0 1.2 2.2 15.6 0.7 2007 60.2 1.5 2.1 22.0 1.4 2011 58.9 1.5 1.7 23.2 1.5 Source: The Economic Research Service estimates are calculated using data from the US Census Bureau and the Bureau of Labor Statistics. They are available at the Food Expenditures Topic Page, http://www.ers.usda.gov/data-products/food-expenditures.aspx. The Progressive Grocer estimates are drawn from Major (2013) and are available here: http://www.progressivegrocer.com/top-stories/headlines/trending-topics/id39214/supermarkets-emerge-as-grocery-share-leaders/. a Via farmers, processors, and wholesalers. Table 2 Shares of Consumers’ Food Expenditures, by Store Format and Data Source Economic Research Service Year Supermarkets Convenience Stores Other Grocery Specialty Food Stores Warehouse Clubs and Supercenters Mass Merchandise Other Stores Home Deliveries, Mail Orders Direct to Consumera 1999 72.7 2.5 1.1 2.0 5.8 1.8 4.7 3.6 5.8 2000 70.9 2.5 1.4 2.0 7.2 1.7 4.9 3.4 6.0 2001 70.1 2.6 1.2 2.0 8.9 1.7 4.8 3.1 5.5 2002 67.4 2.6 1.3 2.0 11.7 1.6 4.8 3.0 5.6 2003 66.8 2.6 1.3 2.1 12.6 1.4 4.9 3.0 5.7 2004 66.3 2.7 0.8 2.3 13.4 1.2 4.9 2.8 5.6 2005 65.8 2.6 0.7 2.3 14.1 1.0 4.9 2.8 5.8 2006 65.5 2.6 0.8 2.3 14.7 0.9 4.9 2.8 5.5 2007 65.2 2.6 0.6 2.4 15.1 0.7 4.8 2.7 5.9 2008 64.9 2.5 0.7 2.4 15.6 0.7 4.8 2.6 5.9 2009 65.0 2.5 0.6 2.3 15.9 0.7 4.7 2.4 5.9 2010 64.4 2.5 0.9 2.3 16.1 0.6 4.8 2.4 5.9 2011 63.8 2.4 1.4 2.3 16.0 0.6 4.9 2.7 5.9 Progressive Grocer Supermarkets Pharmacy and Drug Stores Discount Deparment Stores Warehouse Clubs and Supercenters Mass Merchandise 1997 72.3 1.2 2.3 9.1 N/A 2002 66.0 1.2 2.2 15.6 0.7 2007 60.2 1.5 2.1 22.0 1.4 2011 58.9 1.5 1.7 23.2 1.5 Economic Research Service Year Supermarkets Convenience Stores Other Grocery Specialty Food Stores Warehouse Clubs and Supercenters Mass Merchandise Other Stores Home Deliveries, Mail Orders Direct to Consumera 1999 72.7 2.5 1.1 2.0 5.8 1.8 4.7 3.6 5.8 2000 70.9 2.5 1.4 2.0 7.2 1.7 4.9 3.4 6.0 2001 70.1 2.6 1.2 2.0 8.9 1.7 4.8 3.1 5.5 2002 67.4 2.6 1.3 2.0 11.7 1.6 4.8 3.0 5.6 2003 66.8 2.6 1.3 2.1 12.6 1.4 4.9 3.0 5.7 2004 66.3 2.7 0.8 2.3 13.4 1.2 4.9 2.8 5.6 2005 65.8 2.6 0.7 2.3 14.1 1.0 4.9 2.8 5.8 2006 65.5 2.6 0.8 2.3 14.7 0.9 4.9 2.8 5.5 2007 65.2 2.6 0.6 2.4 15.1 0.7 4.8 2.7 5.9 2008 64.9 2.5 0.7 2.4 15.6 0.7 4.8 2.6 5.9 2009 65.0 2.5 0.6 2.3 15.9 0.7 4.7 2.4 5.9 2010 64.4 2.5 0.9 2.3 16.1 0.6 4.8 2.4 5.9 2011 63.8 2.4 1.4 2.3 16.0 0.6 4.9 2.7 5.9 Progressive Grocer Supermarkets Pharmacy and Drug Stores Discount Deparment Stores Warehouse Clubs and Supercenters Mass Merchandise 1997 72.3 1.2 2.3 9.1 N/A 2002 66.0 1.2 2.2 15.6 0.7 2007 60.2 1.5 2.1 22.0 1.4 2011 58.9 1.5 1.7 23.2 1.5 Source: The Economic Research Service estimates are calculated using data from the US Census Bureau and the Bureau of Labor Statistics. They are available at the Food Expenditures Topic Page, http://www.ers.usda.gov/data-products/food-expenditures.aspx. The Progressive Grocer estimates are drawn from Major (2013) and are available here: http://www.progressivegrocer.com/top-stories/headlines/trending-topics/id39214/supermarkets-emerge-as-grocery-share-leaders/. a Via farmers, processors, and wholesalers. Methodological Approach To investigate the healthfulness of households’ food purchases, we adopt the methodology developed by Volpe and Okrent (2012) and assign a healthfulness score, hereafter called the USDAScore, to the quarterly shopping baskets of each household. This score relies on the classification of all grocery products, undertaken by the USDA Center for Nutrition Policy and Promotion (CNPP) and reported by Carlson et al. (2007), into a series of broader food categories. These categories are comprehensive of all products in the Nielsen Homescan data and are nonoverlapping. The Carlson et al. (2007) report provides spending recommendations, allotting food dollars across these CNPP categories, to assist consumers in making grocery purchases that abide by the Dietary Guidelines for Americans (DGA). Moreover, these recommendations are distinct according to age and gender. With the rich demographics of the Homescan data, we are able to construct household-specific USDA spending recommendations based on household composition. The USDAScore is based on the differences between the category-specific observed expenditure shares and the USDA recommended expenditure shares. More specifically, the USDAScore for household i in quarter q is given by USDAScoreicq=∑c((expshareicq−USDAexpshareic)2)−1 where c denotes the CNPP food categories. The complete details behind the construction of this score, and information on the CNPP food categories (including how food items in Homescan are mapped to these categories), are available in Volpe and Okrent (2012). The USDAScore is unitless and should be thought of as an index, where higher scores indicate greater adherence to the DGA.1 Our general goal is to investigate how store format decisions and other factors affect the household-specific USDAScore. Despite the rich set of demographics afforded by the Homescan data, a single-equation approach modeling USDAScore as a function of format expenditure shares and controls could be subject to potential endogeneity. Shoppers may select a store format based on their food preferences, particularly their demand for healthful food. For example, if supermarkets offer the healthiest food, then it would stand to reason that people seeking a healthy lifestyle will choose these stores. With these complexities in mind, we use a system of equations approach to account for the potential endogeneity of format shares. In this setting, we simultaneously model two consumer choices: what store format to shop at and how healthy the shopping basket will be. Drawing on the theoretical and applied literature on store choice and food choice, we specify other covariates that may affect simultaneously a household’s observed store-format shares and USDAScore. Our empirical framework relies on eight reduced-form equations, seven that model store-format expenditure share as a function of prices, the food environment, and household demographics, and one that models the USDAScore as a function of the format expenditure shares, prices, market structure, and household demographics. These eight equations are represented by Fimtj=fPRmtj,Smt,Dimt+υimt,forj=1to7,and (1) USDAScoreimt=fFimt,Phmt,Mmt,Dimt+ωimt, (2) where Fimtj is the format share for household i, in market m and quarter t, for store format j. Fimtj is a function of demographics (D), including proxies for time constraints, the local food environment (S), and retail-format prices (PR). USDAScoreimtj is a function of prices for healthy foods (Ph), market structure (M), store format shares (F), and household demographics (D). Estimating the simultaneous equation system given by equations (1) and (2) presents a number of empirical concerns. First, we drop one of the format share equations to avoid a linear dependency. A second concern involves specifying S, D, and M. Here, we draw on related literature and describe how these vectors are specified in the next section. A third concern is that the store format shares, Fimtj, are bounded by zero, thus making the equations given by (1) left censored.2 This censoring makes estimating the system computationally intensive. To overcome this problem, we rely on the general concepts found in Shonkwiler and Yen (1999), who employ a two-step procedure to estimate a left-censored system. To apply their procedure to our system, we construct and estimate individual probit equations corresponding to the equations in (1), recover the Inverse Mills ratios, then estimate uncensored linear equations for (1) and (2) as a system, with the Inverse Mills ratios included as extra terms in (1). The last concern involves prices, which pose several empirical difficulties. One difficulty, discussed in the next section, involves the question of how to specify PR and Ph. However, a second and especially troublesome difficulty, potential endogeneity, is discussed here. If underlying determinants of prices are unobserved by the researcher but correlated with the error terms, then the coefficient estimates in the simultaneous system can be biased. A general strategy to overcome this endogeneity issue is to use instruments to estimate PR and Ph. We follow this strategy and use Hausman-style instruments composed of prices from outside markets that have uncorrelated demand or supply shocks. Thus PR and Ph are replaced by their predicted values from instrumental variable estimations. Data and Variable Specification The Homescan data set consists of the self-scanned purchases of a sample of households who are asked to record their complete purchases of each shopping trip by scanning UPCs. For the years 2004–2010 the database contains over 50,000 households (most of which are not in the sample the entire period) that span nearly the entire continental United States. The Homescan data include detailed information on product names and characteristics as well as household demographics, thus allowing for an in-depth analysis of quality of food purchases while controlling for extensive confounding variables. While the Homescan data are self-reported, robustness checks and calibration efforts have demonstrated the data to contain errors in line with those found in major government-collected data sets (Einav et al. 2008, 2010). We structure the data into cumulative shopping baskets by household and quarter (1998Q1–2010Q4). The final data set consists of more than 1.1 million quarterly baskets from over 90,000 unique households. A key aspect to every basket is that expenditures are broken down by store format, a treatment that allows for the precise measurement of the contribution each store format has on shopping baskets. Table 3 provides USDAScore and expenditure shares for selected food categories, by store format, and with the exception of club stores, it shows that foods purchased at nontraditional formats are less healthful. It is important to keep in mind throughout that all shares presented at table 3 are averages across formats. That is, if we treat all expenditures at supermarkets as a single shopping basket, it has an average USDAScore of 7.4 and fruits constitute 6.3% of the expenditures. This average does not necessarily reflect the purchasing habits of households in general or those that shop entirely or primarily at supermarkets, as consumers are able to shop among formats to complete their food shopping within any given time period, and in practice often do so (Fox et al. 2004; Gauri et al. 2008).3 Supermarket expenditures have the highest USDAScore save for club stores, which score slightly higher at 7.5. Foods purchased at mass merchandisers and drug stores are the least healthful, with each format scoring 3.6. Convenience stores, the subject of much of the literature that motivates this study, score 4.3. Other stores score 6.4 and supercenter expenditures are slightly less healthful on average than those made at supermarkets, scoring 6.9. Table 3 Average USDAScore and Shares of Total Expenditures Attributed to Selected Food Categories, by Store Format Supermarkets Mass. Merchandise Supercenters Drug Stores Club Stores Convenience Stores Other USDAScore 7.43 3.55 6.94 3.55 7.50 4.27 6.41 Fruits 6.32 2.12 4.70 2.65 9.24 1.79 3.39 Vegetables 7.39 1.23 4.93 0.70 4.98 0.70 3.56 Whole Grains 2.01 2.09 2.02 1.19 1.53 0.39 1.71 Refined Grains 8.64 5.98 7.65 2.27 4.95 5.02 7.94 Low-fat Dairy 4.60 1.51 3.37 2.25 2.26 10.36 1.24 Regular Dairy 8.42 2.78 6.53 4.44 6.70 15.84 3.03 Low-fat Meats 1.10 0.29 1.42 0.04 0.94 0.07 0.21 Regular Meats 5.24 1.04 4.89 0.72 4.64 0.063 1.91 Poultry 0.66 0.15 0.61 0.02 2.03 0.03 0.23 Fats 2.29 0.43 1.87 0.22 1.58 0.39 0.49 Sugar Drinks 5.52 8.99 6.31 7.95 3.73 17.16 8.53 Sweet Packaged 12.37 34.34 17.95 45.71 11.24 18.16 32.93 Savory Packaged 23.22 22.05 24.02 12.43 26.23 14.01 19.91 Supermarkets Mass. Merchandise Supercenters Drug Stores Club Stores Convenience Stores Other USDAScore 7.43 3.55 6.94 3.55 7.50 4.27 6.41 Fruits 6.32 2.12 4.70 2.65 9.24 1.79 3.39 Vegetables 7.39 1.23 4.93 0.70 4.98 0.70 3.56 Whole Grains 2.01 2.09 2.02 1.19 1.53 0.39 1.71 Refined Grains 8.64 5.98 7.65 2.27 4.95 5.02 7.94 Low-fat Dairy 4.60 1.51 3.37 2.25 2.26 10.36 1.24 Regular Dairy 8.42 2.78 6.53 4.44 6.70 15.84 3.03 Low-fat Meats 1.10 0.29 1.42 0.04 0.94 0.07 0.21 Regular Meats 5.24 1.04 4.89 0.72 4.64 0.063 1.91 Poultry 0.66 0.15 0.61 0.02 2.03 0.03 0.23 Fats 2.29 0.43 1.87 0.22 1.58 0.39 0.49 Sugar Drinks 5.52 8.99 6.31 7.95 3.73 17.16 8.53 Sweet Packaged 12.37 34.34 17.95 45.71 11.24 18.16 32.93 Savory Packaged 23.22 22.05 24.02 12.43 26.23 14.01 19.91 Notes: The reported figures pertain to calculations made on the various food categories. For example, 6.32% of all supermarket expenditures are on fruits. The expenditures in each column will not sum to one because not all food categories are reported. Table 3 Average USDAScore and Shares of Total Expenditures Attributed to Selected Food Categories, by Store Format Supermarkets Mass. Merchandise Supercenters Drug Stores Club Stores Convenience Stores Other USDAScore 7.43 3.55 6.94 3.55 7.50 4.27 6.41 Fruits 6.32 2.12 4.70 2.65 9.24 1.79 3.39 Vegetables 7.39 1.23 4.93 0.70 4.98 0.70 3.56 Whole Grains 2.01 2.09 2.02 1.19 1.53 0.39 1.71 Refined Grains 8.64 5.98 7.65 2.27 4.95 5.02 7.94 Low-fat Dairy 4.60 1.51 3.37 2.25 2.26 10.36 1.24 Regular Dairy 8.42 2.78 6.53 4.44 6.70 15.84 3.03 Low-fat Meats 1.10 0.29 1.42 0.04 0.94 0.07 0.21 Regular Meats 5.24 1.04 4.89 0.72 4.64 0.063 1.91 Poultry 0.66 0.15 0.61 0.02 2.03 0.03 0.23 Fats 2.29 0.43 1.87 0.22 1.58 0.39 0.49 Sugar Drinks 5.52 8.99 6.31 7.95 3.73 17.16 8.53 Sweet Packaged 12.37 34.34 17.95 45.71 11.24 18.16 32.93 Savory Packaged 23.22 22.05 24.02 12.43 26.23 14.01 19.91 Supermarkets Mass. Merchandise Supercenters Drug Stores Club Stores Convenience Stores Other USDAScore 7.43 3.55 6.94 3.55 7.50 4.27 6.41 Fruits 6.32 2.12 4.70 2.65 9.24 1.79 3.39 Vegetables 7.39 1.23 4.93 0.70 4.98 0.70 3.56 Whole Grains 2.01 2.09 2.02 1.19 1.53 0.39 1.71 Refined Grains 8.64 5.98 7.65 2.27 4.95 5.02 7.94 Low-fat Dairy 4.60 1.51 3.37 2.25 2.26 10.36 1.24 Regular Dairy 8.42 2.78 6.53 4.44 6.70 15.84 3.03 Low-fat Meats 1.10 0.29 1.42 0.04 0.94 0.07 0.21 Regular Meats 5.24 1.04 4.89 0.72 4.64 0.063 1.91 Poultry 0.66 0.15 0.61 0.02 2.03 0.03 0.23 Fats 2.29 0.43 1.87 0.22 1.58 0.39 0.49 Sugar Drinks 5.52 8.99 6.31 7.95 3.73 17.16 8.53 Sweet Packaged 12.37 34.34 17.95 45.71 11.24 18.16 32.93 Savory Packaged 23.22 22.05 24.02 12.43 26.23 14.01 19.91 Notes: The reported figures pertain to calculations made on the various food categories. For example, 6.32% of all supermarket expenditures are on fruits. The expenditures in each column will not sum to one because not all food categories are reported. An examination of food categories, which mostly correspond with or are aggregations of the CNPP food groups, helps to explain the differences in healthfulness across formats. The DGA recommends that people consume increased quantities of fruits and vegetables while consuming reduced quantities of food prepared with added sugar or sodium. The expenditure shares for fruits and vegetables fall in a distinct hierarchy across the formats, with the shares being the highest at supermarkets and club stores. Shares on produce are lower at supercenters and other stores, and they are very low, between 0 and 2%, at the mass merchandisers, convenience stores, and drug stores. On average, households spend a far greater share of their food dollars on sweet and savory packaged foods than on fruits and vegetables. However, the hierarchy is repeated, in reverse, for these foods. It is also notable that over 17% of all spending at convenience stores goes to sugar drinks, including soda and sports drinks, and this is double the next-highest expenditure share for this category among formats. The differences across formats for the remaining categories are generally not striking and do not follow a clear pattern. For example, whole grains are also recommended by the DGA, and the shares for these foods are comparably low across all formats. Thus, it seems that the variation in overall healthfulness across formats is driven mostly by the differences in fruits, vegetables, and packaged foods. Figure 1 presents another way to depict the relationship between households’ USDAScores and their predominant store format choice. For each of the seven store formats, we calculate the yearly average USDAScore among households with a store format share greater than or equal to the 75th percentile among all households in our sample.4Figure 1 plots these averages for USDAScore and shows that shopping baskets are healthiest for households above the 75th percentile shares for club stores, followed by households above the 75th percentile shares for supermarkets. Conversely, shopping baskets are unhealthiest for convenience, other, and drug stores. Figure 1 View largeDownload slide USDAScore by Store Channel Notes: The figure reflects the average USDAScore 1 for households with a store channel share greater than or equal to the 75th percentile share, except Convenience stores, which uses the 90th percentile (because the 75th percentile for convenience stores is a zero share). Figure 1 View largeDownload slide USDAScore by Store Channel Notes: The figure reflects the average USDAScore 1 for households with a store channel share greater than or equal to the 75th percentile share, except Convenience stores, which uses the 90th percentile (because the 75th percentile for convenience stores is a zero share). The USDAScore serves as the dependent variable in equation (2), but regressors in equations (1) and (2) still must be specified before estimating the system. The demographic covariates in D are based on a systematic review of the literature on health, obesity, and food choices ranging across a number of disciplines. Blisard et al. (2004) and Dong and Lin (2009) are examples of applied research that suggest controlling for income, with findings showing that lower-income households purchase and consume fewer healthful foods, on average. A similarly rich line of research has generated a similar consensus with respect to education, food choices, and health outcomes (Garn et al. 1977; Galobardes et al. 2000; Xie et al. 2003). Researchers have also uncovered persistent and significant differences in dietary quality and related health outcomes by race and ethnicity (Block et al. 2004; Zenk et al. 2005). We use Homescan variables for household income, household size, education, and race. We also make efforts to account for the role of time constraints due to the fact that households vary in terms of the time available to prepare food at home, as implied by Cutler et al. (2003) and Mancino and Newman (2007). We therefore include an unemployment variable as well as indicator variables for an array of occupations. In equation (1), Smt is meant to reflect the retail food-at-home food environment. A number of specifications are possible in this setting (including simple store counts per capita). All of the models we estimated yielded substantively similar results with respect to our key hypotheses, and ultimately we employ two measures: the ratio of supermarket counts to convenience store counts and the ratio of supermarket counts to supercenter counts. These ratios are calculated by market and year using store counts drawn from the Homescan data.5 The impacts of convenience stores and supercenters on consumers generate the most attention, by a wide margin, in the research literature and popular press. For market structure (M), we use the TDLinx data set, which contains detailed store characteristics, including location, ownership, size, and categorical annual sales, to calculate the Herfindahl-Hirschman Index (HHI) among food retailers as a measure of concentration as well as the market share controlled by individual store formats, by year and by market.6 The coverage is comprehensive of US food retailers. Unfortunately, the store formats identified in TDLinx do not correspond cleanly to those in Homescan, and food sales data are not available for all of the formats studied in the Homescan data.7 Recognizing that shoppers choose store formats based partly on price, PRmtj is a proxy for the vector of average relative prices of the formats. Given the complexities involved in comparing prices across store types, we simplify and use only the format-specific price of bottled drinking water, which is one of the few products with available UPCs in identical form across all formats throughout the data. Therefore, for PR we use the average per ounce price of bottled water, by format, market, and quarter, normalized such as the price for other stores is one. Finally, equation (2) contains Phmt, an aggregate price for healthy food products, to signify the importance of prices on dietary quality as research suggests (e.g., Kuchler and Stewart 2008; Carlson and Frazao 2012). We use publicly available data from the USDA Quarterly Food-at-Home Price Database (QFAHPD), which provides market-level prices per 100 grams, by food group (see Todd et al. [2010] for more information on the QFAHPD). This resource is ideal for our purposes, given that the QFAPHD is calculated using Homescan, and the QFAHPD food groups were selected with the DGA in mind. For each market and quarter, we calculate a healthy food price and an unhealthy food price as the expenditure-share weighted average of the QFAHPD prices for those food groups falling in the respective categories. Volpe et al. (2013, 571) provide a listing of all 52 food crops along with their designation as healthy or unhealthy. However, because the unhealthy food groups in the QFAHPD are fewer in number and more commonly found across all store formats, we focus only on the aggregate unhealthy price, which we refer to as PUnHealthy. Because of endogeneity concerns mentioned above, we use three sets of instruments to predict PR and PUnHealthy. One set is lagged prices, which follows a long literature on the topic of endogenous regressors (Murray 2006). A second set capture marginal costs, or input prices, that might affect retail prices (e.g.,Villas-Boas 2009). For these, we employ a series of indexes from the Bureau of Labor Statistics (BLS). The BLS database includes both Consumer Price Indexes (CPIs), which measure retail price changes, and Producer Price Indexes (PPIs), which measure farm and wholesale price changes. The CPIs were available regionally, while the PPIs are national. Specifically, we use the following indexes: motor fuel CPI, fuels and utilities CPI, food manufacturing PPI, electric power generation PPI, truck transportation PPI, farm product warehousing and storage PPI, all commodities PPI, and grocery stores PPI. The last set of instruments intended to capture the role of input prices is rooted in Hausman’s (1996) argument that if demographics are properly accounted for, then the prices of a good in one city may be instrumented for using the prices for that same good in another city. Nevo (2001) and others use this technique as well. However, our data contain more cities than quarters, making it impossible to exploit all potential instruments. We instead regress our price variables against 100 random draws of 20 cities and yield 100 expected prices, which we then average. These average expected prices, following the Hausman framework, are cleaned of endogeneity. Following these methods, we rely on the predicted prices P^R and P^UnHealthy. Table 4 presents summary statistics for all the variables used in our analysis. Table 4 Summary Statistics for Variables in Our Final Data Set Variable Description Mean Std. Dev. Min Max USDAScore Basket healthfulness 7.345 2.476 0.898 19.995 FSupermarket Supermarket expenditure share 0.656 0.294 0 1 FDrug Drug store expenditure share 0.026 0.067 0 1 FMass Mass merchandizer expend. share 0.035 0.089 0 1 FSupercenter Supercenter expenditure share 0.134 0.242 0 1 FClub Club store expenditure share 0.089 0.170 0 1 FConvenience Convenience store expend. share 0.005 0.033 0 1 FOther Other stores expenditure share 0.056 0.134 0 1 PUnHealthy Unhealthy price (based on QFAHPD) 0.417 0.036 0.336 0.520 HHIFood Herfindahl-Hirschman Index (food retailers) 0.381 0.229 0.041 1 PSupermarket Avg. supermarket bottled water price/oz 0.022 0.004 0.014 0.032 PDrug Avg. drug store bottled water price/oz 0.030 0.006 0.006 0.059 PMass Avg. mass merch. bottled water price/oz 0.025 0.006 0.011 0.067 PSupermcenter Avg. supercenter bottled water price/oz 0.023 0.009 0.005 0.099 PClub Avg. club store bottled water price/oz 0.013 0.004 0.010 0.034 PConvenience Avg. c-store bottled water price/oz 0.051 0.013 0.009 0.115 POther Avg. other store bottled water price/oz 0.032 0.008 0.007 0.066 SSMarket/Conven Supermarket to convenience store count 1.584 0.923 0.593 6.941 SSMarket/SCent Supermarket to supercenter store count 9.897 12.396 2.54 188 HH-Inc Categorical HH Income 19.895 6.050 3 30 HH-Size Categorical HH Size 2.361 1.298 1 9 Black Back/African American status 0.100 0.300 0 1 Asian Asian American status 0.027 0.163 0 1 Other Race Other race status 0.047 0.211 0 1 Unemployed Unemployed status 0.331 0.471 0 1 MH-Educ Categorical male head education level 3.066 2.079 0 6 MH-Empl. Categorical male head employment status 3.483 3.244 0 9 WIC-Ever Every participate in WIC status 0.059 0.237 0 1 Variable Description Mean Std. Dev. Min Max USDAScore Basket healthfulness 7.345 2.476 0.898 19.995 FSupermarket Supermarket expenditure share 0.656 0.294 0 1 FDrug Drug store expenditure share 0.026 0.067 0 1 FMass Mass merchandizer expend. share 0.035 0.089 0 1 FSupercenter Supercenter expenditure share 0.134 0.242 0 1 FClub Club store expenditure share 0.089 0.170 0 1 FConvenience Convenience store expend. share 0.005 0.033 0 1 FOther Other stores expenditure share 0.056 0.134 0 1 PUnHealthy Unhealthy price (based on QFAHPD) 0.417 0.036 0.336 0.520 HHIFood Herfindahl-Hirschman Index (food retailers) 0.381 0.229 0.041 1 PSupermarket Avg. supermarket bottled water price/oz 0.022 0.004 0.014 0.032 PDrug Avg. drug store bottled water price/oz 0.030 0.006 0.006 0.059 PMass Avg. mass merch. bottled water price/oz 0.025 0.006 0.011 0.067 PSupermcenter Avg. supercenter bottled water price/oz 0.023 0.009 0.005 0.099 PClub Avg. club store bottled water price/oz 0.013 0.004 0.010 0.034 PConvenience Avg. c-store bottled water price/oz 0.051 0.013 0.009 0.115 POther Avg. other store bottled water price/oz 0.032 0.008 0.007 0.066 SSMarket/Conven Supermarket to convenience store count 1.584 0.923 0.593 6.941 SSMarket/SCent Supermarket to supercenter store count 9.897 12.396 2.54 188 HH-Inc Categorical HH Income 19.895 6.050 3 30 HH-Size Categorical HH Size 2.361 1.298 1 9 Black Back/African American status 0.100 0.300 0 1 Asian Asian American status 0.027 0.163 0 1 Other Race Other race status 0.047 0.211 0 1 Unemployed Unemployed status 0.331 0.471 0 1 MH-Educ Categorical male head education level 3.066 2.079 0 6 MH-Empl. Categorical male head employment status 3.483 3.244 0 9 WIC-Ever Every participate in WIC status 0.059 0.237 0 1 Table 4 Summary Statistics for Variables in Our Final Data Set Variable Description Mean Std. Dev. Min Max USDAScore Basket healthfulness 7.345 2.476 0.898 19.995 FSupermarket Supermarket expenditure share 0.656 0.294 0 1 FDrug Drug store expenditure share 0.026 0.067 0 1 FMass Mass merchandizer expend. share 0.035 0.089 0 1 FSupercenter Supercenter expenditure share 0.134 0.242 0 1 FClub Club store expenditure share 0.089 0.170 0 1 FConvenience Convenience store expend. share 0.005 0.033 0 1 FOther Other stores expenditure share 0.056 0.134 0 1 PUnHealthy Unhealthy price (based on QFAHPD) 0.417 0.036 0.336 0.520 HHIFood Herfindahl-Hirschman Index (food retailers) 0.381 0.229 0.041 1 PSupermarket Avg. supermarket bottled water price/oz 0.022 0.004 0.014 0.032 PDrug Avg. drug store bottled water price/oz 0.030 0.006 0.006 0.059 PMass Avg. mass merch. bottled water price/oz 0.025 0.006 0.011 0.067 PSupermcenter Avg. supercenter bottled water price/oz 0.023 0.009 0.005 0.099 PClub Avg. club store bottled water price/oz 0.013 0.004 0.010 0.034 PConvenience Avg. c-store bottled water price/oz 0.051 0.013 0.009 0.115 POther Avg. other store bottled water price/oz 0.032 0.008 0.007 0.066 SSMarket/Conven Supermarket to convenience store count 1.584 0.923 0.593 6.941 SSMarket/SCent Supermarket to supercenter store count 9.897 12.396 2.54 188 HH-Inc Categorical HH Income 19.895 6.050 3 30 HH-Size Categorical HH Size 2.361 1.298 1 9 Black Back/African American status 0.100 0.300 0 1 Asian Asian American status 0.027 0.163 0 1 Other Race Other race status 0.047 0.211 0 1 Unemployed Unemployed status 0.331 0.471 0 1 MH-Educ Categorical male head education level 3.066 2.079 0 6 MH-Empl. Categorical male head employment status 3.483 3.244 0 9 WIC-Ever Every participate in WIC status 0.059 0.237 0 1 Variable Description Mean Std. Dev. Min Max USDAScore Basket healthfulness 7.345 2.476 0.898 19.995 FSupermarket Supermarket expenditure share 0.656 0.294 0 1 FDrug Drug store expenditure share 0.026 0.067 0 1 FMass Mass merchandizer expend. share 0.035 0.089 0 1 FSupercenter Supercenter expenditure share 0.134 0.242 0 1 FClub Club store expenditure share 0.089 0.170 0 1 FConvenience Convenience store expend. share 0.005 0.033 0 1 FOther Other stores expenditure share 0.056 0.134 0 1 PUnHealthy Unhealthy price (based on QFAHPD) 0.417 0.036 0.336 0.520 HHIFood Herfindahl-Hirschman Index (food retailers) 0.381 0.229 0.041 1 PSupermarket Avg. supermarket bottled water price/oz 0.022 0.004 0.014 0.032 PDrug Avg. drug store bottled water price/oz 0.030 0.006 0.006 0.059 PMass Avg. mass merch. bottled water price/oz 0.025 0.006 0.011 0.067 PSupermcenter Avg. supercenter bottled water price/oz 0.023 0.009 0.005 0.099 PClub Avg. club store bottled water price/oz 0.013 0.004 0.010 0.034 PConvenience Avg. c-store bottled water price/oz 0.051 0.013 0.009 0.115 POther Avg. other store bottled water price/oz 0.032 0.008 0.007 0.066 SSMarket/Conven Supermarket to convenience store count 1.584 0.923 0.593 6.941 SSMarket/SCent Supermarket to supercenter store count 9.897 12.396 2.54 188 HH-Inc Categorical HH Income 19.895 6.050 3 30 HH-Size Categorical HH Size 2.361 1.298 1 9 Black Back/African American status 0.100 0.300 0 1 Asian Asian American status 0.027 0.163 0 1 Other Race Other race status 0.047 0.211 0 1 Unemployed Unemployed status 0.331 0.471 0 1 MH-Educ Categorical male head education level 3.066 2.079 0 6 MH-Empl. Categorical male head employment status 3.483 3.244 0 9 WIC-Ever Every participate in WIC status 0.059 0.237 0 1 Results Based on the specifications discussed above, we estimate equations (1) and (2) simultaneously using, in part, the Shonkwiler and Yen (1999) two-step procedure. More specifically, the Inverse Mills ratios are recovered from the first-step probit estimations corresponding to equation (1); then the system of equations (1) and (2) are estimated simultaneously with 3SLS. Results from this second step are presented in table 5. Table 5 3SLS Results, 2004–2010, Full Sample USDAScore FSupermarket FDrug FMass FSupercenter FClub FConvenience Equation (2): Est. Coeff. Equations (1): Est. Coeff. Est. Coeff. Est. Coeff. Est. Coeff. Est. Coeff. Est. Coeff. FSupermarket 1.859*** P^Supermarket 4.707*** 1.140*** 3.636*** 0.029*** 3.596*** −0.215*** FDrug −7.668*** P^Drug 4.158*** −0.452*** −0.827*** −0.0003*** −4.184*** 0.168*** FMass 0.728 P^Mass −4.522*** −0.104** −0.230*** −0.0002*** 3.566*** −0.234*** FSupercenter 2.336*** P^Supermcenter 3.379*** −0.147*** 0.566*** 0.005*** −0.460*** 0.059*** FClub −0.682 P^Club 2.531*** 1.196*** −1.234*** −0.009*** −2.538*** 0.026 FConvenience −20.48*** P^Convenience 1.424*** 0.013 0.313*** −0.018*** −0.208*** −0.062*** P^UnHealthy 2.246*** P^Other 4.073*** 0.585*** −0.238*** −0.017*** −2.470*** 0.247*** HHIFood −0.111*** SSMarket/Conven −0.041*** −0.004*** −0.002*** −0.006** 0.0233*** −0.002*** SSMarket/SCent 0.0003*** 0.00003*** −0.00003*** −0.016*** 0.0002*** −0.00001 HH-Inc 0.005*** HH-Inc −0.00003 −0.00004*** 0.0000* −0.003*** 0.0004*** −0.00004*** HH-Size 0.022*** HH-Size −0.006*** −0.003*** 0.002*** −0.028*** 0.006*** −0.0001** Black 0.141*** Black −0.033*** 0.007*** 0.008*** 0.015*** −0.005*** 0.000003 Asian 0.056 Asian −0.085*** 0.017*** 0.004*** 0.004*** 0.052*** −0.0007** Other Race −0.027*** 0.004*** 0.012*** −0.005*** 0.006*** −0.002*** Unemployed 0.0222 Unemployed 0.035*** 0.005*** −0.001 −0.006** 0.020*** −0.001** Max-Educ 0.174*** Max-Educ 0.003*** −0.0003 −0.005*** 0.029*** 0.007*** −0.0003*** NoMaleH −1.167*** NoMaleH 0.022*** 0.008*** 0.010***  −−0.0003*** −0.033*** 0.0001 NoFemH −0.793*** NoFemH 0.035*** 0.008*** −0.004*** −0.0002*** −0.024*** 0.008*** MEmp-FT 0.00511 MEmp-FT −0.006*** −0.007*** 0.006*** 0.005*** 0.003** 0.002*** FEmp-FT −0.138*** FEmp-FT 0.0008 −0.005*** 0.007*** −0.009*** 0.005*** 0.002*** MFEmp-FT −0.087*** MFEmp-FT 0.014*** 0.004*** −0.009*** −0.018*** −0.006*** −0.002*** WIC-Ever 0.0308 Inv. Mills −29.40*** 0.819*** 11.41*** −0.017*** 105.8*** −2.749*** Constant yes Constant yes yes yes yes yes yes Y and Q FE yes Y and Q FE yes yes yes yes yes yes Occup. FE yes yes yes yes yes yes USDAScore FSupermarket FDrug FMass FSupercenter FClub FConvenience Equation (2): Est. Coeff. Equations (1): Est. Coeff. Est. Coeff. Est. Coeff. Est. Coeff. Est. Coeff. Est. Coeff. FSupermarket 1.859*** P^Supermarket 4.707*** 1.140*** 3.636*** 0.029*** 3.596*** −0.215*** FDrug −7.668*** P^Drug 4.158*** −0.452*** −0.827*** −0.0003*** −4.184*** 0.168*** FMass 0.728 P^Mass −4.522*** −0.104** −0.230*** −0.0002*** 3.566*** −0.234*** FSupercenter 2.336*** P^Supermcenter 3.379*** −0.147*** 0.566*** 0.005*** −0.460*** 0.059*** FClub −0.682 P^Club 2.531*** 1.196*** −1.234*** −0.009*** −2.538*** 0.026 FConvenience −20.48*** P^Convenience 1.424*** 0.013 0.313*** −0.018*** −0.208*** −0.062*** P^UnHealthy 2.246*** P^Other 4.073*** 0.585*** −0.238*** −0.017*** −2.470*** 0.247*** HHIFood −0.111*** SSMarket/Conven −0.041*** −0.004*** −0.002*** −0.006** 0.0233*** −0.002*** SSMarket/SCent 0.0003*** 0.00003*** −0.00003*** −0.016*** 0.0002*** −0.00001 HH-Inc 0.005*** HH-Inc −0.00003 −0.00004*** 0.0000* −0.003*** 0.0004*** −0.00004*** HH-Size 0.022*** HH-Size −0.006*** −0.003*** 0.002*** −0.028*** 0.006*** −0.0001** Black 0.141*** Black −0.033*** 0.007*** 0.008*** 0.015*** −0.005*** 0.000003 Asian 0.056 Asian −0.085*** 0.017*** 0.004*** 0.004*** 0.052*** −0.0007** Other Race −0.027*** 0.004*** 0.012*** −0.005*** 0.006*** −0.002*** Unemployed 0.0222 Unemployed 0.035*** 0.005*** −0.001 −0.006** 0.020*** −0.001** Max-Educ 0.174*** Max-Educ 0.003*** −0.0003 −0.005*** 0.029*** 0.007*** −0.0003*** NoMaleH −1.167*** NoMaleH 0.022*** 0.008*** 0.010***  −−0.0003*** −0.033*** 0.0001 NoFemH −0.793*** NoFemH 0.035*** 0.008*** −0.004*** −0.0002*** −0.024*** 0.008*** MEmp-FT 0.00511 MEmp-FT −0.006*** −0.007*** 0.006*** 0.005*** 0.003** 0.002*** FEmp-FT −0.138*** FEmp-FT 0.0008 −0.005*** 0.007*** −0.009*** 0.005*** 0.002*** MFEmp-FT −0.087*** MFEmp-FT 0.014*** 0.004*** −0.009*** −0.018*** −0.006*** −0.002*** WIC-Ever 0.0308 Inv. Mills −29.40*** 0.819*** 11.41*** −0.017*** 105.8*** −2.749*** Constant yes Constant yes yes yes yes yes yes Y and Q FE yes Y and Q FE yes yes yes yes yes yes Occup. FE yes yes yes yes yes yes Notes: *** Statistically significant at the 99% level; ** at the 95% level; and * at the 90% level. First-stage results for F variables in equation (1) not presented. Table 5 3SLS Results, 2004–2010, Full Sample USDAScore FSupermarket FDrug FMass FSupercenter FClub FConvenience Equation (2): Est. Coeff. Equations (1): Est. Coeff. Est. Coeff. Est. Coeff. Est. Coeff. Est. Coeff. Est. Coeff. FSupermarket 1.859*** P^Supermarket 4.707*** 1.140*** 3.636*** 0.029*** 3.596*** −0.215*** FDrug −7.668*** P^Drug 4.158*** −0.452*** −0.827*** −0.0003*** −4.184*** 0.168*** FMass 0.728 P^Mass −4.522*** −0.104** −0.230*** −0.0002*** 3.566*** −0.234*** FSupercenter 2.336*** P^Supermcenter 3.379*** −0.147*** 0.566*** 0.005*** −0.460*** 0.059*** FClub −0.682 P^Club 2.531*** 1.196*** −1.234*** −0.009*** −2.538*** 0.026 FConvenience −20.48*** P^Convenience 1.424*** 0.013 0.313*** −0.018*** −0.208*** −0.062*** P^UnHealthy 2.246*** P^Other 4.073*** 0.585*** −0.238*** −0.017*** −2.470*** 0.247*** HHIFood −0.111*** SSMarket/Conven −0.041*** −0.004*** −0.002*** −0.006** 0.0233*** −0.002*** SSMarket/SCent 0.0003*** 0.00003*** −0.00003*** −0.016*** 0.0002*** −0.00001 HH-Inc 0.005*** HH-Inc −0.00003 −0.00004*** 0.0000* −0.003*** 0.0004*** −0.00004*** HH-Size 0.022*** HH-Size −0.006*** −0.003*** 0.002*** −0.028*** 0.006*** −0.0001** Black 0.141*** Black −0.033*** 0.007*** 0.008*** 0.015*** −0.005*** 0.000003 Asian 0.056 Asian −0.085*** 0.017*** 0.004*** 0.004*** 0.052*** −0.0007** Other Race −0.027*** 0.004*** 0.012*** −0.005*** 0.006*** −0.002*** Unemployed 0.0222 Unemployed 0.035*** 0.005*** −0.001 −0.006** 0.020*** −0.001** Max-Educ 0.174*** Max-Educ 0.003*** −0.0003 −0.005*** 0.029*** 0.007*** −0.0003*** NoMaleH −1.167*** NoMaleH 0.022*** 0.008*** 0.010***  −−0.0003*** −0.033*** 0.0001 NoFemH −0.793*** NoFemH 0.035*** 0.008*** −0.004*** −0.0002*** −0.024*** 0.008*** MEmp-FT 0.00511 MEmp-FT −0.006*** −0.007*** 0.006*** 0.005*** 0.003** 0.002*** FEmp-FT −0.138*** FEmp-FT 0.0008 −0.005*** 0.007*** −0.009*** 0.005*** 0.002*** MFEmp-FT −0.087*** MFEmp-FT 0.014*** 0.004*** −0.009*** −0.018*** −0.006*** −0.002*** WIC-Ever 0.0308 Inv. Mills −29.40*** 0.819*** 11.41*** −0.017*** 105.8*** −2.749*** Constant yes Constant yes yes yes yes yes yes Y and Q FE yes Y and Q FE yes yes yes yes yes yes Occup. FE yes yes yes yes yes yes USDAScore FSupermarket FDrug FMass FSupercenter FClub FConvenience Equation (2): Est. Coeff. Equations (1): Est. Coeff. Est. Coeff. Est. Coeff. Est. Coeff. Est. Coeff. Est. Coeff. FSupermarket 1.859*** P^Supermarket 4.707*** 1.140*** 3.636*** 0.029*** 3.596*** −0.215*** FDrug −7.668*** P^Drug 4.158*** −0.452*** −0.827*** −0.0003*** −4.184*** 0.168*** FMass 0.728 P^Mass −4.522*** −0.104** −0.230*** −0.0002*** 3.566*** −0.234*** FSupercenter 2.336*** P^Supermcenter 3.379*** −0.147*** 0.566*** 0.005*** −0.460*** 0.059*** FClub −0.682 P^Club 2.531*** 1.196*** −1.234*** −0.009*** −2.538*** 0.026 FConvenience −20.48*** P^Convenience 1.424*** 0.013 0.313*** −0.018*** −0.208*** −0.062*** P^UnHealthy 2.246*** P^Other 4.073*** 0.585*** −0.238*** −0.017*** −2.470*** 0.247*** HHIFood −0.111*** SSMarket/Conven −0.041*** −0.004*** −0.002*** −0.006** 0.0233*** −0.002*** SSMarket/SCent 0.0003*** 0.00003*** −0.00003*** −0.016*** 0.0002*** −0.00001 HH-Inc 0.005*** HH-Inc −0.00003 −0.00004*** 0.0000* −0.003*** 0.0004*** −0.00004*** HH-Size 0.022*** HH-Size −0.006*** −0.003*** 0.002*** −0.028*** 0.006*** −0.0001** Black 0.141*** Black −0.033*** 0.007*** 0.008*** 0.015*** −0.005*** 0.000003 Asian 0.056 Asian −0.085*** 0.017*** 0.004*** 0.004*** 0.052*** −0.0007** Other Race −0.027*** 0.004*** 0.012*** −0.005*** 0.006*** −0.002*** Unemployed 0.0222 Unemployed 0.035*** 0.005*** −0.001 −0.006** 0.020*** −0.001** Max-Educ 0.174*** Max-Educ 0.003*** −0.0003 −0.005*** 0.029*** 0.007*** −0.0003*** NoMaleH −1.167*** NoMaleH 0.022*** 0.008*** 0.010***  −−0.0003*** −0.033*** 0.0001 NoFemH −0.793*** NoFemH 0.035*** 0.008*** −0.004*** −0.0002*** −0.024*** 0.008*** MEmp-FT 0.00511 MEmp-FT −0.006*** −0.007*** 0.006*** 0.005*** 0.003** 0.002*** FEmp-FT −0.138*** FEmp-FT 0.0008 −0.005*** 0.007*** −0.009*** 0.005*** 0.002*** MFEmp-FT −0.087*** MFEmp-FT 0.014*** 0.004*** −0.009*** −0.018*** −0.006*** −0.002*** WIC-Ever 0.0308 Inv. Mills −29.40*** 0.819*** 11.41*** −0.017*** 105.8*** −2.749*** Constant yes Constant yes yes yes yes yes yes Y and Q FE yes Y and Q FE yes yes yes yes yes yes Occup. FE yes yes yes yes yes yes Notes: *** Statistically significant at the 99% level; ** at the 95% level; and * at the 90% level. First-stage results for F variables in equation (1) not presented. Our main interest focuses on results in the first and second columns of table 5, which correspond to equation (2) and the factors that influence the USDAScore.8 Here, regression coefficients can be interpreted as the impacts on food-purchase healthfulness. The impact of increasing store-format shares (F) are divided into positive and negative effects on the USDAScore. Increasing the expenditure shares at supermarkets and supercenters is associated with increased healthfulness; conversely, increasing expenditure shares at drug stores and convenience stores is associated with decreased healthfulness. Some of these results, such as the positive estimate for FSupermarket and the negative estimate for FConvenience, corroborate most previous research. However, the supercenter and club store results are noteworthy, each for different reasons. The FClub result is noteworthy because it shows an increasing expenditure share at club stores has no statistical impact on the USDAScore despite the fact that table 3 and figure 1 suggest that club store purchases are healthier, on average, than those made at other formats. The lack of statistical significance may, in part, reflect a low overall share of food purchases in club stores. On the other hand, the positive association between USDAScore and FSupercenter speaks to the open empirical question of whether or not supercenters are associated with healthy or unhealthy food purchases. Our results suggest increasing the share of the food dollar spent at supercenters can lead to healthier food purchases. This result is discussed in the next section. Most other results for estimated equation (1) are generally straightforward: (i) A higher price for unhealthy food ( P^Unhealthy) is associated with healthier food purchases.9 (ii) Increased concentration in food retailing (HHIFood) is associated with less healthier food purchases. (iii) Higher household income, larger household size, and higher education are all associated with healthier food purchases. (iv) The lack of a male or female household head has a negative association with healthy food purchases. (v) Full-time employment by a female head, either alone or in conjunction with full-time employment by a male head, is associated with less healthy purchases. (v) Finally, neither unemployment status nor current or former WIC program status has a statistical significance for the full data sample. The last seven columns of table 5, which correspond to equation (1), provide insight into the factors that influence the household-level expenditure shares across store formats. With two exceptions, the own-price coefficients are negative and statistically different from zero. The own-price estimate for club stores has the greatest magnitude, suggesting that this expenditure share is most sensitive to own price. The own-price coefficient for supermarkets and supercenters, the two exceptions, are positive, suggesting that higher supermarket and supercenter prices are associated with higher shares. This result may be an artifact of the high expenditure share for supermarkets and supercenters relative to other store formats. Cross-price estimates are a mixture of positive and negative, indicating that store formats can be considered substitutes and complements. For example, in the FMass column, the cross price estimate for P^Supermcenter is 0.566, indicating that households will substitute towards mass merchandizers if prices at supercenters increase. On the other hand, in the FSupercenter column, the cross price estimate for club stores is −0.009, indicating that households will buy slightly less from supercenters if prices at club stores increase. A number of store formats, that is, drug stores, mass merchandisers, club stores, convenience stores, and other stores all appear to be complementary to supercenters. The food environment variables also have both positive and negative effects depending on the store-format share. Increasing the supermarket-to-convenience store count ratio leads to higher expenditure shares at club stores, but lower shares at all other formats. However, increasing the supermarket-to-supercenter store count ratio leads to higher expenditure shares at supermarkets, drug stores, and club stores, but lower shares at mass merchandisers and supercenters. Other household demographic variables also have mixed results depending on the particular store format. Household size leads to higher shares at mass merchandisers and club stores, and lower shares at other formats. Household income leads to higher shares at club stores and mass merchandisers, and lower shares at all other formats. We investigate household income’s role in greater detail by reestimating equations (1) and (2) simultaneously for three subsamples based on income percentiles: one subsample for household incomes less than or equal to the 25th percentile, one where income is within the two middle quartiles, and the final one where income is greater than or equal to the 75th percentile. Abridged results just for equation (1) are presented in table 6. Table 6. 3SLS Results for USDAScore, Full Sample and Household Income Subsamples USDAScore Full Sample < 25th percentile HH Inc Middle 50 percentile HH Inc ≥ 75th percentile HH Inc FSupermarket 1.859*** 3.829** 2.044** −1.874* FDrug −7.668*** 3.209 −1.031 −7.573* FMass 0.728 7.720* 4.874*** 3.217 FSupercenter 2.336*** 0.803 3.386*** 0.903 FClub −0.682 3.329 −0.729 −5.243*** FConvenience −20.48*** 2.839 −18.50** −70.89*** P^UnHealthy 2.246*** 0.629 1.175* 2.466*** HHIFood −0.111*** 0.0427 −0.0974* −0.238*** HH-Inc 0.005*** 0.0198*** 0.0101*** 0.00258*** HH-Size 0.022*** 0.144*** 0.00629 0.0469*** Black 0.141*** 0.108 0.078* −0.129** Asian 0.056 −0.468 −0.0445 −0.0959* Unemployed 0.0222 −0.137** 0.0645** 0.146** Max-Educ 0.174*** 0.159*** 0.167*** 0.187*** NoMaleH −1.167*** −0.992*** −1.222*** −1.313*** NoFemH −0.793*** −0.595*** −0.853*** −0.546*** MEmp-FT 0.00511 −0.035 −0.031 0.0775 FEmp-FT −0.138*** −0.309*** −0.202*** −0.084** MFEmp-FT −0.087*** 0.167 −0.076** −0.102** WIC-Ever 0.0308 −0.032 0.050 0.036 Constant yes yes yes yes Y and Q FE yes yes yes yes USDAScore Full Sample < 25th percentile HH Inc Middle 50 percentile HH Inc ≥ 75th percentile HH Inc FSupermarket 1.859*** 3.829** 2.044** −1.874* FDrug −7.668*** 3.209 −1.031 −7.573* FMass 0.728 7.720* 4.874*** 3.217 FSupercenter 2.336*** 0.803 3.386*** 0.903 FClub −0.682 3.329 −0.729 −5.243*** FConvenience −20.48*** 2.839 −18.50** −70.89*** P^UnHealthy 2.246*** 0.629 1.175* 2.466*** HHIFood −0.111*** 0.0427 −0.0974* −0.238*** HH-Inc 0.005*** 0.0198*** 0.0101*** 0.00258*** HH-Size 0.022*** 0.144*** 0.00629 0.0469*** Black 0.141*** 0.108 0.078* −0.129** Asian 0.056 −0.468 −0.0445 −0.0959* Unemployed 0.0222 −0.137** 0.0645** 0.146** Max-Educ 0.174*** 0.159*** 0.167*** 0.187*** NoMaleH −1.167*** −0.992*** −1.222*** −1.313*** NoFemH −0.793*** −0.595*** −0.853*** −0.546*** MEmp-FT 0.00511 −0.035 −0.031 0.0775 FEmp-FT −0.138*** −0.309*** −0.202*** −0.084** MFEmp-FT −0.087*** 0.167 −0.076** −0.102** WIC-Ever 0.0308 −0.032 0.050 0.036 Constant yes yes yes yes Y and Q FE yes yes yes yes Notes: *** Statistically significant at the 99% level; ** at the 95% level; and * at the 90% level. Neither first-stage nor 3SLS results for equation (1)’s F variables are presented. Table 6. 3SLS Results for USDAScore, Full Sample and Household Income Subsamples USDAScore Full Sample < 25th percentile HH Inc Middle 50 percentile HH Inc ≥ 75th percentile HH Inc FSupermarket 1.859*** 3.829** 2.044** −1.874* FDrug −7.668*** 3.209 −1.031 −7.573* FMass 0.728 7.720* 4.874*** 3.217 FSupercenter 2.336*** 0.803 3.386*** 0.903 FClub −0.682 3.329 −0.729 −5.243*** FConvenience −20.48*** 2.839 −18.50** −70.89*** P^UnHealthy 2.246*** 0.629 1.175* 2.466*** HHIFood −0.111*** 0.0427 −0.0974* −0.238*** HH-Inc 0.005*** 0.0198*** 0.0101*** 0.00258*** HH-Size 0.022*** 0.144*** 0.00629 0.0469*** Black 0.141*** 0.108 0.078* −0.129** Asian 0.056 −0.468 −0.0445 −0.0959* Unemployed 0.0222 −0.137** 0.0645** 0.146** Max-Educ 0.174*** 0.159*** 0.167*** 0.187*** NoMaleH −1.167*** −0.992*** −1.222*** −1.313*** NoFemH −0.793*** −0.595*** −0.853*** −0.546*** MEmp-FT 0.00511 −0.035 −0.031 0.0775 FEmp-FT −0.138*** −0.309*** −0.202*** −0.084** MFEmp-FT −0.087*** 0.167 −0.076** −0.102** WIC-Ever 0.0308 −0.032 0.050 0.036 Constant yes yes yes yes Y and Q FE yes yes yes yes USDAScore Full Sample < 25th percentile HH Inc Middle 50 percentile HH Inc ≥ 75th percentile HH Inc FSupermarket 1.859*** 3.829** 2.044** −1.874* FDrug −7.668*** 3.209 −1.031 −7.573* FMass 0.728 7.720* 4.874*** 3.217 FSupercenter 2.336*** 0.803 3.386*** 0.903 FClub −0.682 3.329 −0.729 −5.243*** FConvenience −20.48*** 2.839 −18.50** −70.89*** P^UnHealthy 2.246*** 0.629 1.175* 2.466*** HHIFood −0.111*** 0.0427 −0.0974* −0.238*** HH-Inc 0.005*** 0.0198*** 0.0101*** 0.00258*** HH-Size 0.022*** 0.144*** 0.00629 0.0469*** Black 0.141*** 0.108 0.078* −0.129** Asian 0.056 −0.468 −0.0445 −0.0959* Unemployed 0.0222 −0.137** 0.0645** 0.146** Max-Educ 0.174*** 0.159*** 0.167*** 0.187*** NoMaleH −1.167*** −0.992*** −1.222*** −1.313*** NoFemH −0.793*** −0.595*** −0.853*** −0.546*** MEmp-FT 0.00511 −0.035 −0.031 0.0775 FEmp-FT −0.138*** −0.309*** −0.202*** −0.084** MFEmp-FT −0.087*** 0.167 −0.076** −0.102** WIC-Ever 0.0308 −0.032 0.050 0.036 Constant yes yes yes yes Y and Q FE yes yes yes yes Notes: *** Statistically significant at the 99% level; ** at the 95% level; and * at the 90% level. Neither first-stage nor 3SLS results for equation (1)’s F variables are presented. In almost every case, the middle subsample generates similar results for the full sample. However, results for the low-income and high-income subsamples vary considerably, both compared to each other and compared to the middle-income subsample. For example, low-income subsample results, like the middle-quartile subsample results, show that increased expenditure shares at supermarkets are associated with a higher USDAScore. However, the high-income subsample results show that increased shares at supermarkets are unexpectedly associated with a lower USDAScore. For the low-income subsample, increased shares at supermarkets and supercenters are associated with healthier food purchases. Results for all other formats are not statistically significant. Alternatively, for the high-income subsample, increased shares at drug stores, club stores, convenience supermarkets, and even supermarkets are all associated with less healthy food purchases. These results suggest that high-income consumers may be getting healthy food from places not in the dataset. Table 6 also shows that unhealthy prices do not play a strong role for the low-income subsample. However, they do play an increasing strong role in the middle- and high-income subsamples. Industry concentration, as reflected by HHIFood, has a negative effect on USDAScore only for the two higher income subsamples. Education is consistently positive across all subsamples. The lack of a male or female head and full-time employment status is also generally consistent across all subsamples. Discussion and Future Work American shoppers increasingly have a large number of retail formats from which to choose when purchasing food at home. Our study is the first, of which we are aware, to investigate and quantify the impact that alternative store formats have on the healthfulness of households’ food purchases. We extend related streams of research in two important ways. First, we relax the assumption that all stores other than conventional supermarkets, or all stores of the same approximate size, have the same impacts on food purchases. And second, we examine total quarterly household shopping baskets, as opposed to studying select product categories, to measure overall adherence to the DGA in order to get a clearer picture of how purchase decisions may be shaped by store choice. Using a large and highly detailed data set, we find that increased purchases at convenience stores and drug stores result in food choices less in line with the DGA. This finding is notable as these small-store formats, especially convenience stores, have been the subject of a number of nutritional, sociological, and economics studies on dietary quality and food access, and we discuss a number of reasons why this finding might have arisen. Additionally, we corroborate the consensus of research across a range of disciplines which argues that consumers purchase more healthful foods at larger stores with greater selection and, in many cases, lower prices per unit. Dietary quality, as reflected by the USDAScore, is positively associated with higher expenditure shares at supermarkets and supercenters. This positive association is well known for supermarkets, but a somewhat open question for supercenters. From a policy perspective, our study has the potential to inform the unanswered questions regarding supermarket “interventions.” A number of studies (Kristal et al. 1997; Wrigley et al. 2003; Cummins et al. 2005; Escaron et al. 2015) have examined the ex-post impacts of the subsidized introduction of large supermarkets into geographic areas where there previously were not any. Interventions and the interest surrounding them stem primary from the argument that low adherence to the DGA and related adverse health outcomes in the United States can be attributed, in part, to lack of food access. Most intervention analyses have identified little to no significant impact on food choices or dietary quality. Our findings suggest that store format or simply store characteristics may play an important role in determining the economic and health impacts of new stores in markets. In addition, as the food retailing landscape continues to evolve, supercenters, especially Wal-Mart, have been subject to much scrutiny and controversy as they have expanded throughout the United States. Basker (2007) notes that local governments have frequently attempted to block Wal-Mart’s entry into markets using zoning regulations or wage ordinances. From a policy perspective, however, the impacts of supercenters on factors such as food access, diet quality, and health outcomes remain unclear. Supercenters, which sell full lines of fruits, vegetables, and all foods necessary to meet the DGA, have been shown to result in lower average food prices within markets (Woo et al. 2001; Hausman and Leibtag 2007; Volpe and Lavoie 2008; Basker and Noel 2009). In this article, we find that increased expenditure shares supercenters (including Wal-Mart) can lead to healthier food purchases. Both the low- and high-income quartiles suggest that our results may be sensitive to income, and our supercenter result is one such case. The positive effect of supercenter expenditure share on food-purchase healthiness holds only for the full sample and the middle-income subsample but not the low- or high-income subsamples. Increased convenience store shares are strongly associated with lower food-purchase healthiness in all subsamples except the bottom quartile, whereas increased club store shares are strongly associated with lower food-purchase healthiness in only the high-income subsample. Similarly, higher unhealthy prices are positively associated with food-purchase healthiness in all samples but in low-income subsample. Industry concentration is negatively associated with food-purchase healthiness in all samples but the low-income subsample. Lastly, and perhaps most surprisingly, increased supermarket shares are negatively related to dietary quality in the top income quartile. Our findings leave much room for discussion and in many cases call for further research. The next important step in this line of inquiry is to understand the mechanisms that are driving our results. It seems plausible that supermarkets and supercenters stores offer larger selections of fresh produce and other healthful food options than do most other formats, but this may not be true across all stores or for all households. In addition, pilot studies of the Healthy Corner Store Initiative have conducted at several cities across the United States, and the goal of this initiative is to increase the availability and awareness of healthy foods in urban corner stores. Our convenience-store results suggest that initiatives of this sort may have success for the lowest income quartile. To fully characterize the relationship between dietary quality and food purchasing choices, however, we would need to increase the focus on heterogeneity across households or across store types. For example, to tease out the supercenter result from a policy perspective, we would need a sample more representative of low-income households. Additional work may also investigate potential differences across other demographic splits such as specific education, or racial or ethnic subsamples. We hope that our results therefore can motivate further research using store scanner data to investigate the link between store choices and food choices. Disclaimer A portion of this research was undertaken while Richard Volpe was an economist with the USDA Economic Research Service. The views expressed in this paper are those of the authors and may not be attributed to USDA or ERS. The authors accessed the Nielsen data via a third-party agreement with the Economic Research Service. Footnotes 1 An anonymous reviewer correctly notes that USDAScore may misrepresent food choices as it has the potential to incorrectly penalize households who “overshoot” (i.e., have higher observed shares than recommended shares) for CNPP food categories deemed healthy, or undershoot for CNPP categories deemed unhealthy. For this reason, we also calculate a calculate USDAScore+, which first designates CNPP categories as healthy or unhealthy and then disallows penalties for being “too healthy” or not “unhealthy enough”. In the results section, we note that penalties of these types are extremely rare in the Homescan data, so our USDAScore and USDAScore+ perform almost identically. 2 Technically, the store-format shares are also bounded by one. However, the number of quarterly household-level store-format shares attaining this upper bound is extremely small, so generally ignore right-censoring concerns. 3 Additionally, it is uncommon for households to attribute a large portion of their quarterly food dollars to some of the smaller nontraditional formats. Though not reported, we examined quarterly shopping baskets for those households that spend at least 90% of their food dollars at a single store format. For example, there were only 80 quarterly baskets in sample consisting predominantly of purchases made from convenience stores. 4 For one store format, convenience stores, we use the 90th percentile instead of the 75th percentile. The reason is that most expenditures shares for convenience stores are zero, i.e., both the median and 75th percentile share for convenience stores are zero. On the other hand, the 90th percentile share for convenience stores is 0.57%. 5 Our store counts, by format, are drawn directly from the Homescan data. Other data sources, such as the US Census of Business or Nielsen TDLinx, raise questions about the applicability to our sample in terms of geography or format definition. The major drawback of using Homescan store counts is that small formats outside of traditional grocery, particularly convenience stores, are necessarily undercounted. However, this is only an issue if the rate of undercounting varies in a systematic way across Scantrack markets, and we have no evidence that this is the case. 6 The HHI is calculated as the sum of squared market shares for the firms operating within an industry and market. Technically we calculated the HHI (10), using only the ten largest firms in each market. Given that food retail is highly concentrated locally, HHI exhibits much more variation than other commonly used concentration measures, such as the four-firm concentration ratio (CR4). 7 We experimented with using store counts, as drawn from the TDLinx data, to account for formats such as convenience stores and club stores, for which food sales data are not available. However, we encountered two issues with this strategy. First, these variables were almost never significant in determining USDAScore. And second, we deemed them too similar to the food environment variables included in (S), in that they measure food access and proximity rather than market structure or share. 8 We also estimated equations (1) and (2) simultaneously using USDAScore+, which was defined in an earlier footnote. Because the correlation between USDAScore and USDAScore+ is 0.94, and because 3SLS coefficients are extremely similar in both cases, we do not present the results with USDAScore+ in this article. However, these results are available from the authors upon request. 9 Our instrumental variables for prices pass the standard battery of tests, including the Hausman endogeneity test and the Sargan overidentification test, indicating that they are preferred to OLS. However, the usual caveats and assumptions to these tests apply. These results are available upon request. References Basker E. 2007 . The Causes and Consequences of Wal-Mart's Growth. The Journal of Economic Perspectives 21 3 : 177 – 98 . Google Scholar CrossRef Search ADS Basker E. , Noel M. . 2009 . The Evolving Food Chain: Competitive Effects of Wal‐Mart's Entry into the Supermarket Industry . Journal of Economics and Management Strategy 18 4 : 977 – 1009 . Google Scholar CrossRef Search ADS Bhattacharya J. , Bundorf M.K. . 2009 . The Incidence of the Healthcare Costs of Obesity . Journal of Health Economics 28 3 : 649 – 58 . Google Scholar CrossRef Search ADS PubMed Blattberg R.C. , Briesch R. , Fox E.J. . 1995 . How Promotions Work . Marketing Science 14 3 : G122 – G132 . Google Scholar CrossRef Search ADS Blisard N. , Stewart H. , Jolliffe D. . 2004 . Low-Income Households' Expenditures on Fruits and Vegetables. US Department of Agriculture, Economic Research Service, Agriculture Information Bulletin No. 792-5. Block J.P. , Scribner R.A. , DeSalvo K.B. . 2004 . Fast Food, Race/Ethnicity, and Income: A Geographic Analysis . American Journal of Preventative Medicine 27 3 : 211 – 17 . Bonanno A. , Lopez R.A. . 2009 . Competition Effects of Supermarket Services . American Journal of Agricultural Economics 91 3 : 555 – 68 . Google Scholar CrossRef Search ADS Bustillos B. , Sharkey J.R. , Anding J. , McIntosh A. . 2008 . Availability of More Healthful Food Alternatives in Traditional, Convenience, and Nontraditional Types of Food Stores in Two Rural Texas Counties . Journal of the American Dietetic Association 109 5 : 883 – 89 . Google Scholar CrossRef Search ADS Carlson A. , Frazao E. . 2012 . Are Healthy Foods More Expensive? It Depends on How You Measure the Price. EIB-96, US Department of Agriculture, Economic Research Service, October 2010. Carlson A. , Lino M. , Fungwe T. . 2007 . The Low-Cost, Moderate-Cost, and Liberal Food Plans. US Department of Agriculture, Center for Nutrition Policy and Promotion. Carpenter J.M. , Moore M. . 2006 . Consumer Demographics, Store Attributes, and Retail Format Choice in the US Grocery Market . International Journal of Retail and Distribution Management 34 6 : 434 – 52 . Google Scholar CrossRef Search ADS Chen D. , Jaenicke E.C. , Volpe R. . 2016 . Relationships between Home and Neighborhood Food Environments and Obesity: A National Multilevel Analysis Using Scanner Data . American Journal of Public Health . 106 5 : 881 – 88 . Google Scholar CrossRef Search ADS PubMed Chung C. , Myers S.L. . 1999 . Do the Poor Pay More for Food? An Analysis of Grocery Store Availability and Food Disparaties . Journal of Consumer Affairs 33 : 276 – 96 . Google Scholar CrossRef Search ADS Courtemanche C. , Carden A. . 2011 . Supersizing Supercenters? The Impact of Wal-Mart Supercenters on Body Mass Index and Obesity . Journal of Urban Economics 69 2 : 165 – 81 . Google Scholar CrossRef Search ADS Cummins S. , Pettigrew M. , Higgins C. , Findlay A. , Sparks L. 2005 . Large Scale Food Retailing as an Intervention for Diet and Health: Quasi-experimental Evaluation of a Natural Experiment . Journal of Epidemiology and Community Health 59 : 1035 – 104 . Google Scholar CrossRef Search ADS PubMed Cutler D.M. , Glaeser E.L. , Shapiro J.M. . 2003 . Why Have Americans Become More Obese? Journal of Economic Perspectives 17 3 : 93 – 118 . Google Scholar CrossRef Search ADS Dong D. , Lin B. . 2009 . Fruit and Vegetable Consumption by Low-Income Americans: Would a Price Reduction Make a Difference? ERR-70, US Department of Agriculture, Economic Research Service, January, 2009. Escaron A.L. , Martinez-Donate A.P. , Riggall A.J. , Meinen A. , Hall B. , Nieto F.J. , Nitzke S. . 2015. Developing and Implementing “Waupaca Eating Smart” A Restaurant and Supermarket Intervention to Promote Healthy Eating Through Changes in the Food Environment . Health Promotion Practice , p.1524839915612742. Einav L. , Leibtag E. , Nevo A. . 2008 . On the Accuracy of Nielsen Homescan Data. ERR-69, US Department of Agriculture, Economic Research Service, November 2008. Einav L. , Leibtag E. , Nevo A. 2010 . Recoding Dicrepancies in Nielsen Homescan Data: Are They Present and Do They Matter? Quantitative Marketing and Economics 8 2 : 207 – 39 . Google Scholar CrossRef Search ADS Food Institute. 2010 . “The Food Industry Review.” (2010 Edition). Fox E.J. , Montgomery A.L. , Lodish L.M. . 2004 . Consumer Shopping and Spending across Retail Formats . Journal of Business 77 2 : S25 – S60 . Google Scholar CrossRef Search ADS Galobardes B. , Morabia A. , Bernstein M.S. . 2000 . The Differential Effect of Education and Occupation on Body Mass and Overweight in a Sample of Working People of the General Population . Annals of Epidemiology 10 8 : 532 – 37 . Google Scholar CrossRef Search ADS PubMed Garn S.M. , Bailey S.M. , Cole P.E. , Higgins I.T. . 1977 . Level of Education, Level of Income, and Level of Fatness in Adults . American Journal of Clinical Nutrition 30 5 : 721 – 25 . Google Scholar CrossRef Search ADS PubMed Gauri D.K. , Trivedi M. , Grewal D. . 2008 . Understanding the Determinants of Retail Strategy: An Empirical Analysis . Journal of Retailing 84 3 : 256 – 67 . Google Scholar CrossRef Search ADS Goodman C.S. 1968 . Do the poor pay more? Journal of Marketing , 32 : 18 – 24 . Google Scholar CrossRef Search ADS Hausman J. 1996 . “The Valuation of Goods under Perfect and Imperfect Information.” Book chapter. The Economics of New Goods. Hausman J. , Leibtag E. , 2007 . Consumer benefits from increased competition in shopping outlets: Measuring the effect of Wal‐Mart . Journal of Applied Econometrics , 22 7 : 1157 – 77 . Google Scholar CrossRef Search ADS Kaufman P.R. , MacDonald J.M. , Lutz S.M. , Smallwood D.M. . 1997 . Do the Poor Pay More for Food? Item Selection and Price Differences Affect Low-Income Household Food Costs. Agricultural Economics Report No. (AER759) 32 pp, December 1997. Kristal A.R. , Goldenhar L. , Muldoon J. , Morton R.F. . 1997 . Evaluation of a Supermarket Intervention to Increase Consumption of Fruits and Vegetables. American Journal of Health Promotion 11 6 : 422 – 25 . Google Scholar CrossRef Search ADS PubMed Kuchler F. , Stewart H. . 2008 . Price Trends Are Similar for Fruits, Vegetables, and Snack Foods. US Department of Agriculture, Economic Research Service, Economic Research Report No. 55, March 2008. Lamm R.M. 1981 . Prices and Concentration in the Food Retailing Industry . Journal of Industrial Economics 30 1 : 67 – 78 . Google Scholar CrossRef Search ADS Leibtag E. , Barker C. , Dutko P. . 2010 . How Much Lower Are Prices at Discount Stores? ERR-105, US Department of Agriculture, Economic Research Service. Liese A.D. , Weis K.E. , Pluto D. , Smith E. , Lawson A. . 2007 . Food Store Types, Availability, and Cost of Foods in a Rural Environment . Journal of the American Dietetic Association 107 11 : 1916 – 23 . Google Scholar CrossRef Search ADS PubMed MacDonald J.M. , Nelson P.E. . 1991 . Do the Poor Still Pay More? Food Price Variations in Large Metropolitan Areas . Journal of Urban Economics 30 3 : 344 – 359 . Google Scholar CrossRef Search ADS Major M. 2013 . Supermarkets Emerge as Grocery Share Leaders. http://www.progressivegrocer.com/top-stories/headlines/trending-topics/id39214/supermarkets-emerge-as-grocery-share-leaders/. Accessed February 21, 2018. Mancino L. , Kuchler F. , Leibtag E. . 2008 . Getting Consumers to Eat More Whole-grains: The Role of Policy, Information, and Food Manufacturers . Food Policy 33 6 : 489 – 96 . Google Scholar CrossRef Search ADS Mancino L. , Newman C. . 2007 . Who Has Time to Cook? How Family Resources Influence Food Preparation. ERR-40, US Department of Agriculture, Economic Research Service, May, 2007. Market Force . 2011 . Market Force Study Finds Consumers Perceive ALDI as Grocery Price Leader. Martinez S.W. 2007 . The US Food Marketing System: Recent Developments, 1997–2006. ERR-42. USDept of Agriculture, Econ. Res. Serv. May 2007. Matsa D.A. 2011 . Competition and Product Quality in the Supermarket Industry . Quarterly Journal of Economics 126 3 : 1539 – 91 . Google Scholar CrossRef Search ADS PubMed Murray M.P. 2006 . Avoiding Invalid Instruments and Coping with Weak Instruments . The Journal of Economic Perspectives 20 4 : 111 – 32 . Google Scholar CrossRef Search ADS Powell L.M. , Bao Y. . 2009 . Food Prices, Access to Food Outlets, and Child Weight . Economics and Human Biology 7 1 : 64 – 72 . Google Scholar CrossRef Search ADS PubMed Richards T.J. , Hamilton S.F. . 2006 . Rivalry in Price and Variety Among Supermarket Retailers . American Journal of Agricultural Economics 88 3 : 710 – 26 . Google Scholar CrossRef Search ADS Rose D. , Richards R. . 2004 . Food Store Access and Household Fruit and Vegetable Use Among Participants in the US Food Stamp Program . Public Health Nutrition 7 8 : 1081 – 1088 . Google Scholar CrossRef Search ADS PubMed Seiders K. , Simonides C. , Tigert D.J. . 2000 . The Impact of Supercenters on Traditional Food Retailers in Four Markets . International Journal of Retail and Distribution Management 28 45 : 181 – 93 . Google Scholar CrossRef Search ADS Shonkwiler J.S. , Yen S.T. . 1999 . Two-Step Estimtion of a Censored System of Equations . American Journal of Agricultural Economics 81 4 : 972 – 82 . Google Scholar CrossRef Search ADS Todd J.E. , Mancino L. , Leibtag E. , Tripodo C. . 2010 . Methodology Behind the Quarterly Food-at-Home Price Database. US Department of Agriculture, Economic Research Service, Technical Bulletin No. 1926. Villas-Boas S. 2009 . An Empirical Investigation of the Welfare Effects of Banning Wholesale Price Discrimination . RAND Journal of Economics 40 1 : 20 – 46 . Google Scholar CrossRef Search ADS Volpe R. J. , Lavoie N. . 2008 . The Effect of Wal-Mart Supercenters on Grocery Prices in New England . Applied Economic Perspectives and Policy 30 1 : 4 – 26 . Volpe R.J. , Okrent A. . 2012 . Assessing the Healthfulness of Consumers’ Grocery Purchases. EIB-102, US Department of Agriculture, Economic Research Service, November 2012. Volpe R.J. , Okrent A. , Leibtag E. . 2013 . The Effect of Supercenter-format Stores on the Healthfulness of Consumers’ Grocery Purchases . American Journal of Agricultural Economics 95 3 : 568 – 89 . Google Scholar CrossRef Search ADS Wrigley N. , Warm D. , Margetts B. . 2003 . Deprivation, Diet, and Food-retail Access: Findings from the Leeds ‘food deserts' Study. Environment and Planning A 35 1 : 151 – 88 . Google Scholar CrossRef Search ADS Woo B. J. , Huang C. L. , Epperson J. E. , Cude B. . 2001 . Effect of a New Wal-Mart Supercenter on Local Retail Food Prices . Journal of Food Distribution Research 32 1 : 173 – 81 . Xie B. , Gilliland F.D. , Ly Y. , Rockett H.R. . 2003 . Effects of Ethnicity, Family Income, and Education on Dietary Intake among Adolescents . Preventive Medicine 36 1 : 30 – 40 . Google Scholar CrossRef Search ADS PubMed Yu C.-C. , Connor J.M. . 2002 . The Price-Concentration Relationship in Retailing: Retesting Newmark . Agribusiness 18 4 : 413 – 26 . Google Scholar CrossRef Search ADS Zenk S.N. , Schulz A.J. , Hollis-Neely T. , Campbell R.T. , Holmes N. , Watkins G. , Nwankwo R. , Odoms-Young A. . 2005 . Fruit and Vegetable Intake in African Americans: Income and Store Characteristics. American Journal of Preventative Medicine 29 1 : 1 – 9 . Google Scholar CrossRef Search ADS © The Author(s) 2018. Published by Oxford University Press on behalf of the Agricultural and Applied Economics Association. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices)

Journal

Applied Economic Perspectives and PolicyOxford University Press

Published: Mar 26, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off