Competition and Product Quality in the Supermarket Industry


Download Competition and Product Quality in the Supermarket Industry


Preview text

Competition and Product Quality in the Supermarket Industry
David A. Matsa0F*
July 6, 2009
Abstract This paper analyzes the effect of competition on a supermarket firm’s incentive to provide product quality. In the supermarket industry, product availability is an important measure of quality. Using U.S. consumer price index microdata to track inventory shortfalls, I find that stores facing more intense competition have fewer shortfalls. Competition from Wal-Mart – the most significant shock to industry market structure in half a century – decreased shortfalls by up to 24 percent. The risk of customers switching stores appears to provide strong incentives for investments in product quality.
JEL Classifications: D40, G31, L15, L81 Keywords: product quality, competition, monopoly, big-box, inventory management, stockout
* Kellogg School of Management, Northwestern University, [email protected] This work was conducted in coordination with the U.S. Bureau of Labor Statistics (BLS) under an Agency Agreement. I am indebted to Bill Cook, Craig Brown, Mark Bowman, Dan Ginsburg, and others at the BLS who have been extremely helpful throughout this project, and to Mark Bils for sharing portions of his computer code and insight into the data. I also want to thank Mike Mazzeo and Nancy Rose for helpful comments. I am grateful to Garry Van Siclen and Trade Dimensions for providing the supermarket establishment data, and to Thomas Holmes for the information on Wal-Mart store openings. This research would not have been possible without the support of the George and Obie Shultz Fund and the National Science Foundation (Grant No. SES-0551097). The BLS pledges confidentiality to voluntary respondents in the Consumer Price Index (CPI) sample; no inferences should be made from this paper as to whether or not a specific firm is included in the CPI sample. Any opinions, findings, or conclusions expressed herein are my own and do not necessarily reflect the views of the BLS.

Increasingly over the last few years, competition for consumers’ food dollars has intensified....The Company believes its competitive advantages include convenient locations, the quality of service it provides its customers, competitive pricing, product variety and quality and a pleasant shopping environment.
Ingles Markets, Inc. 10-K, September 27, 2008
1. Introduction In most consumer markets, there is more to shopping than finding the lowest price. A product’s
quality can have large effects on demand and consumer welfare. Although theory has long recognized that firms with market power may reduce their products’ quality in order to save costs and maximize their profits (Chamberlin 1933; Abbott 1955), empirical research and competition policy tend to focus almost exclusively on price setting (Draganska, Mazzeo, and Seim 2009). While the potential for competition to have a significant effect on product quality is recognized in theory, its empirical importance is much less clear.
This study focuses on the connection between competition and product quality in the supermarket industry. In the retail sector, a firm’s “product” is the shopping experience it provides its customers. Like for physical products, retail product quality has many dimensions, including the store’s cleanliness, its checkout speed, the courteousness of its staff, and the depth of its product assortment. The particular dimension of quality that I examine in this paper is whether a supermarket reliably has its customers’ preferred products in stock when they want to make a purchase.
As an important determinant of customer satisfaction, maintaining a reputation for product availability is a key strategic issue throughout the retail sector. Estimates suggest that 8.2 percent of the items a grocery retailer carries are not in inventory on a typical afternoon and that this is a major cause of customer dissatisfaction (Andersen 1996; Roland Berger 2002). In fact, consumer surveys find that frequent inventory shortfalls (“stockouts”) and limited product variety are the number one cause of dissatisfaction among supermarket shoppers (FMI 1994b).
Stockouts can be costly to retailers in the current period and especially in future periods. Customers’ immediate substitution to other products or stores upon encountering out-of-stock items is estimated to cost the average grocery retailer 1.7 to 3.1 percent of current period sales (e.g., Andersen 1996; Emmelhainz et al. 1991; Anupindi et al. 1998). Aggregating over the supermarket industry, this translates to $6 billion to $12 billion in lost sales per year.1 An even greater risk to the firm is that some
1F
consumers will shift their future shopping to competing stores. Although these future costs are difficult to
1 This figure aggregates retailers’ lost intended purchase expenditure, but it does not include the sales gained by retailers because of inventory shortfalls at other stores. The number is meaningful from the perspective of an individual profit-maximizing retailer, but not from the perspective of the industry. The decrease in aggregate supermarket sales, which includes gains in addition to losses, is likely much lower.
1

quantify, studies find they have the potential to be substantial (e.g., Straughn 1991; Fitzsimons 2000; Anderson et al. 2006). To avoid these costs, a typical supermarket firm spends twice as much on inventory as on advertising.2
2F
Pressure from competition places countervailing incentives on a retail firm’s optimal level of availability. This tradeoff and other factors affecting optimal availability are discussed in Section 2. On one hand, lower profit margins reduce the immediate costs of losing a sale because an item is out of stock. But greater competition also increases the future costs if dissatisfied consumers are more likely to switch stores for their future shopping needs. Which of these incentives dominates is an empirical question. This ambiguity is common in theories of quality competition (for example, see Spence 1975). Schmalensee’s (1979) review of the theoretical literature on quality competition concludes “there is an obvious need for empirical work to confront the implications of the theoretical literature with data.”
To measure inventory shortfalls, this study takes advantage of a unique data set that was hand collected by the U.S. Bureau of Labor Statistics for use in constructing the consumer price index (CPI). Section 3 describes these data. The data are unique in that they contain information that can be used to measure supermarket product availability at the store-product level, even for firms that do not report financial information to the Securities and Exchange Commission (SEC). These disaggregated microdata allow me to examine variation in product availability across products and stores – even within the same firm – to shed light on the economic forces that affect expenditures on product quality.
I examine the empirical connection between competition and supermarket quality using two complementary approaches. First, in a descriptive analysis, I measure the conditional correlation between local retail competition and stockouts. These results control for market and store characteristics as well as an extensive set of fixed effects, including indicators for firm, product category, metropolitan area, and date. I find that stockouts are negatively correlated with competition. Supermarkets that face competition in their census tract average 5 percent lower stockout rates than otherwise similar stores, suggesting that competition increases firms’ incentive to provide quality.
To address the possibility that these correlations are confounded by endogeneity or an omitted local market characteristic, I employ a second empirical approach. I examine the effects of Wal-Mart’s entry into the supermarket industry – the most significant shock to industry market structure in half a century. Section 4 describes these developments. Within fourteen years of opening its first supercenter (stores selling both general merchandise and a full-line of groceries) in 1988, Wal-Mart has become the largest grocery retailer in the U.S. and the first truly national grocery chain. I exploit time series patterns in Wal-Mart’s entry into local grocery markets to examine how competing supermarkets adjust their
2 Using an inventory carrying cost of 25 percent (Brooks 1972), the average Compustat firm in 2005 spent 1.6 percent of sales on financing and maintain inventory, compared to 0.8 percent on advertising.
2

quality in response. After controlling for market and even store fixed effects, the specific timing of WalMart’s entry represents an exogenous shock to the competitive landscape of these markets. Consistent with this interpretation, there are no pre-existing differences or trends in product availability in these markets before Wal-Mart’s entry.
Section 5 documents how supermarkets’ product availability changes after Wal-Mart comes to town. The analysis finds that pressure from competition can have significant effects on the quality of a firm’s products. When facing competition from Wal-Mart, stockouts decrease on average by about 8 percent. The effect is long-lived and robust to including store fixed effects, indicating that the additional stockouts are not attributable to demand miscalculations or attrition that may have followed Wal-Mart’s entry. Furthermore, I separate products based on their distribution channel and find that competition has no impact on stockouts for product categories for which inventory is directly managed by manufacturers’ distributors, whose optimal stockout rate is unlikely to change with retail market structure. There is also no evidence of incumbents decreasing product variety after competition increases.
Heterogeneity in these effects across stores and markets further supports the role of competition in improving quality. I examine whether there are greater reductions in stockouts among stores that are more likely to lose their customers to Wal-Mart, as measured along two dimensions. First, competition from Wal-Mart typically represents a more direct challenge to large regional chains that feature extensive selections of products than to independent stores that serve niche consumer markets. Second, because Wal-Mart caters to low income consumers with a low price, low service format, supermarkets located in lower income areas within a given supercenter’s catchment area are likely to be most affected when a supercenter opens. I find that the decrease in stockouts is indeed related to both factors. In fact, the average stockout rate decreases by 24 percent among chain stores in low income areas.
This analysis focuses on product availability in the supermarket industry, but it likely also has implications for other dimensions of product quality at these and other firms. Supermarket executives see the stockout rate as a barometer of the whole operational side of the business (Progressive Grocer 1968, p. S4). Empirically, I confirm that supermarket inventories are correlated with other dimensions of product quality: chains that rarely run out of inventory are also cleaner, have more courteous staff, and have faster checkouts. If competition reduces stockouts, these other dimensions of product quality may improve as well. In fact, case studies suggest that supermarket chains attempt to differentiate themselves from WalMart by improving quality along many dimensions, including increasing product assortments, offering fresher meat and produce, and maintaining cleaner, more attractive displays (Carré, Tilly, and Holgate 2009).
This article complements a small number of empirical papers that examine the connection between quality provision and market concentration in other settings. For example, Domberger and Sherr
3

(1989) examine markets for real estate transfer services in England and Wales, and Mazzeo (2003) examines on-time performance in the US passenger airline industry. These studies tend to find positive correlations between competition and quality, but the data analyzed leave some questions about causal inference.3F3 This paper addresses those concerns by exploiting variation at the product level in individual stores that experience different shocks to their competitive environment. The shocks vary in their magnitude and hit at different times. By studying how competition affects inventory strategies across these markets and stores, and for products with different characteristics, I am able to make a strong argument for competition causing supermarkets to improve their quality.
This work also extends previous research on retail inventory strategy. An extensive theoretical literature investigates optimal retail inventory policies and stockout avoidance (e.g., Hall and Porteus 2000; Dana 2001; Dana and Petruzzi 2001; Gaur and Park 2007), but until recently there has been little data on the empirical relevance of these models. Based on a large-scale field test at a catalog company, Anderson, Fitzsimons, and Simester (2006) conclude that inventory policy at the firm they study fails to account for the future opportunity costs of a stockout. Building on that result, Matsa (2009) shows that highly leveraged supermarket firms, which are likely to heavily discount the future costs, are more likely to run out of inventory. This is the first empirical paper (that I know of) to examine the connection between market structure and inventory policy.
2. Retail product availability and competition Maintaining optimal product availability is an important dimension of store quality in the retail
sector. Studies indicate that 8.2 percent of a grocery retailer’s items are out of stock on a typical afternoon and that frequent stockouts and limited product variety are the number one cause of dissatisfaction among supermarket shoppers (Andersen 1996; Roland Berger 2002; FMI 1994b). Reducing stockouts is cited as being as important to store image as flashy customer attractions such as new buildings, remodeling, attractive decor, advertising, promotion, and merchandising (Progressive Grocer 1968, p. S1). Nevertheless, stockout rates vary dramatically across stores and markets, and high in-stock levels are reported to provide retailers with a significant competitive advantage (Andersen 1996).
The frequency of retail stockouts is largely determined by ordering and stocking decisions made at the regional and store levels. Studies find that the vast majority of supermarket stockouts are caused
3 Domberger and Sherr (1989) find that the threat of new entry caused by liberalization was associated with increases in customers’ self-reported satisfaction with their attorney for property sales but not for property purchases. Based on only time series variation, it is unclear to whether these measures of satisfaction reflect the rapid acceleration in house prices over the period rather than an actual change in the quality of legal services provided. Mazzeo (2003) finds that average flight delays are longer in more concentrated airline markets. Based only on cross-sectional variation, it is possible that unobserved market characteristics that cause delays also deter entry from additional carriers.
4

within the store rather than by other parties in the vertical chain: 51 to 73 percent of out-of-stocks due to inaccurate forecasting (e.g., underestimating demand) and ordering errors (e.g., failing to sufficiently monitor the shelf inventory and not reordering when demand exceeds forecast) and another 8 to 22 percent due to failing to restock shelves with available backroom or display inventory (Andersen Consulting 1996; Gruen et al. 2002). A store’s investments in inventory and staff assigned to shelf monitoring thus have the most direct impact on its product availability. Because these decisions can be adjusted in real time, an optimizing retailer typically takes retail market structure as given when making investments in product availability.
Although no retailer typically wants to run out of inventory, reducing stockouts is costly and maintaining 100 percent product availability in a supermarket is certainly not optimal. Optimal stocking decisions trade off expenditures on both inventory costs and shelf monitoring costs for the present value of expected lost profits from running out of inventory. Running out of inventory can reduce profits in both the short and long run. Upon finding that an item is out of stock, consumers substitute by purchasing an alternative item (which reduces the intended purchase expenditure by 0.4 percent), shopping at another store or forgoing the purchase (a 1.3 percent expenditure reduction), or delaying purchase (a 1.3 percent reduction; Andersen 1996). Lost sales from immediate substitution on out-of-stock items is estimated to cost the grocery industry $6 to 12 billion annually.1
0H
Importantly, a store that runs out of inventory not only loses profits from current purchases of the product but also risks affecting consumers’ future shopping behavior. Switching supermarkets is thought to be costly for shoppers, who are accustomed to a particular store’s layout and a regular food shopping routine. Because consumers only reluctantly change their routines, a small but important risk of running out of inventory is that it may cause some customers to switch their regular business to another retailer (Anderson et al. 2006). Consumer surveys find that stockouts are a significant factor in consumers’ decisions to switch primary grocers (Andersen 1996). For this reason a store’s provision of reliable product availability can be thought of as an investment in future market share.
A store’s optimal level of product availability thus depends on customers’ propensities, when facing an out-of-stock item, to reduce current expenditures and to switch stores on future shopping trips. An increase in retail competition can affect these risks. Competition may reduce the short-run costs of running out of stock if profit margins decrease, making it less costly on the margin for customers to substitute to a lower margin product, delay purchase, or search elsewhere. But competition may also increase the long-run costs, because consumers will have additional options for their future shopping needs. Although a reputation for running out of stock frequently may irritate customers, it is not so costly for a store when customers other food shopping options are limited. Thus once another store opens
5

nearby, a strong incentive can develop to improve quality and reputation for reliable in-stock performance.
Retailers’ incentives in inventory management differ from those of the product manufacturers. When customers substitute brands upon encountering an out-of-stock item, the retailer only loses any difference in margin between the products, but the manufacturer loses profits from the sale entirely. On the other hand, the manufacturer may lose little when customers switch stores, while the retailer loses the sale entirely. Therefore, manufacturers do not share stores’ incentive to offer greater product availability when retail competition increases. If manufacturers managed shelf inventory, increased retail competition would be unlikely to affect product availability.
Many factors besides competition and customer substitution and switching also affect a particular item’s optimal stockout rate, including the price elasticity of demand, the wholesale cost, the inventory cost, and the variability of demand. Products for which demand is less elastic earn greater markups and are more valuable to keep in stock. Some products, such as refrigerated items, are more costly to inventory than shelf-stable products. And some products, such as many seasonal items, have less predictable demand.
The optimal rate also varies across retailers. In large stores, returns to scale in demand forecasting and order management may reduce the cost of providing product availability. Greater product variety may also reduce the inconvenience for consumers of encountering a stockout in these stores. The distance from a store’s primary supplier may affect the store’s optimal stockout rate, and technological advances and other changes over time may affect inventory costs. To control for these and other factors that affect the optimal rate of product availability, the empirical analyses of stockouts presented below controls for product category, store, and year-month fixed effects as well as a number of market, item, and time-varying store characteristics.
3. Data and descriptive analysis A. Data on retail stockouts
Reliable data on supermarket product availability are rare; most food stores do not even keep systematic records on their availability (Andersen 1996). The most frequently cited statistics on the prevalence of stockouts come from an Andersen (1996) study, sponsored by the Coca-Cola Retailing Research Council. The authors performed daily audits of 7,000 items in eight product categories in ten demographically and regionally diverse stores for one month. Such isolated (and often localized) studies do not lend themselves to either cross-sectional or longitudinal analysis at the store or market level. Due to the cost of conducting wider-scale audits, some studies have attempted to measure out-of-stocks using
6

purchase scanner data (e.g., Gruen et al. 2002). But such studies risk confusing low availability with low demand, which would bias estimates (Dorgan 1997).
I obtain reliable data on stockouts from the CPI Commodity and Services Survey, which is used by the U.S. Bureau of Labor Statistics (BLS) to compute the consumer price index (CPI). To calculate the CPI, the BLS sends surveyors to record the prices of about 30,000 items sold at grocery stores each month, where each price is specific to a particular product at a particular establishment. Generally, a product must be available for purchase at the time the surveyor visits the establishment in order to be included in the CPI. If the product is unavailable for sale, the surveyor determines whether the establishment expects to carry the item in the future. Thus a product may be considered out of stock if it is not available for sale, it is continuing to be carried by the outlet, and it is not seasonally unavailable (Bils 2005). More details on the construction of these data are described in Appendix A.
Using these microdata, I examine observations on product availability at the item-store-month level from January 1988 through December 2004. Any particular item is sampled for at most 5 years: the full data set includes about 5 million observations on availability for almost 220,000 unique items at 11,500 stores in more than 8,000 census tracts and 147 metropolitan areas. I augment the CPI data with detailed store-level information from the Trade Dimensions Retail Site Database and demographic information on each store’s census tract from the 2000 U.S. Census of Population and Housing. Summary statistics are reported in Table 1. The average stockout rate among supermarket respondents is 4.3 to 5.3 percent. Two factors relating to BLS data collection procedures seem to explain why these estimates are lower than the 1996 industry estimate of 8.2 percent. First, CPI data are usually collected throughout the day on weekdays, whereas out-of-stocks are most prevalent in the afternoon and on Sundays (when they reach an estimated 10.9 percent; Andersen 1996). Second, for food items consumed at home, the BLS effectively does not record stockouts caused by store shelving issues (8 to 22 percent of stockouts; see Appendix A for details). While these factors affect estimates of the absolute level of stockouts, they are unlikely to bias estimates of changes in product availability caused by changes in stores’ competitive environments.
Average stockout rates by day of the week are reported in Figure 1. Average stockout rates are greatest on Sundays (4.8 percent) and decline throughout the rest of the week. This general pattern is consistent with 13 different industry studies surveyed by (Gruen et al. 2002, p.14). Because the heaviest shopping usually takes place over the weekend, re-ordering and deliveries typically occur on Monday and Tuesday. Throughout the week, restocking and preparations for Saturday and Sunday promotions lead to lower stockout rates. After the heavy shopping on Saturday, stockout rates increase sharply on Sundays when labor is reduced, stores seldom take deliveries, and inventories runs out for items that are more popular than expected.
7

Stockout rates also vary systematically across product categories. Figure 2 graphs average stockout rates by major product group. Average stockout rates are lowest for prescription drugs (0.9 percent) and fats and oils (1.4 percent), and they are highest for fish and seafood (8.3 percent), poultry (7.9 percent), and other fresh meat, fruit, and bakery products. On the whole, stockout rates are lower for shelf-stable than refrigerated or frozen products. Similar patterns are evident across more detailed product categories. Appendix Table 1 reports average stockout rates in each of 76 different product categories. There is significant variation in stockout rates even for product categories within the same major group. For example, cakes are out of stock 10.0 percent of the time, whereas cookies are out of stock only 3.5 percent; and ground beef is out 2.5 percent, while other beef and veal are out 12.5 percent. In the analyses that follow, I control for product category fixed effects at this more detailed level.
B. Stockouts and competition and other market, store, and product characteristics Although rates of product availability calculated from the CPI data are certainly measured with
sampling error, they seem to provide reliable estimates. Table 2 presents the conditional correlations of stockouts with various market, store, and product characteristics in a linear-probability multivariateregression framework.4F4 Each observation represents a particular item in a particular store on a particular day: the dependent variable equals 100 if the item is out of stock and 0 if it is in stock. In addition to the controls shown, the regression reported in Table 2 includes a set of fixed effects for each of the following: the firm, the product category, the metropolitan area, the day of the week, and the year-month.5 The
5F
standard errors are adjusted to account for contemporaneous and intertemporal within-market correlation in the error term.
A number of robust patterns emerge from the estimates presented in Table 2. First, the structure of retail competition is correlated with supermarket quality. As a measure of supermarket competition, I examine whether a store faces any competing stores in its census tract. On average, stores that face competition in their census tract have 5 percent lower stockout rates than other stores (evaluated at the mean; p < 0.01).6 A causal interpretation of this estimate implies that pressure from supermarket
6F
competition encourages improvements in quality. But it is also possible that an omitted variable, such as an unmeasured dimension of consumer preferences, leads to both greater competition and greater product
4 Given the size of the data set and the large number of fixed effects, maximum likelihood estimation of a probit or conditional logit model is computationally infeasible. 5 There is one firm fixed effect for all stores not affiliated with a chain (defined as 11 or more retail stores), which corresponds to 18.9 percent of the sample. Product category fixed effects are at the level of BLS entry level items. The sample includes items from approximately 75 grocery categories, ranging from breakfast cereal to eggs to laundry and cleaning products; Appendix Table A1 lists the complete set of product categories. 6 Stockout rates at these stores average 21 basis points lower – a 5 percent decrease relative to the sample mean of 4.3 percent.
8

availability in some markets. The analysis of Wal-Mart entry in the supermarket industry, presented below, aims to distinguish between these possibilities by examining the competitive responses to plausibly exogenous shocks to the competitive landscape of various local markets over time.
The importance of other market characteristics is also striking. Supermarkets located in census tracts with wealthier consumers offer greater product availability (fewer stockouts). Higher prices (and presumably margins) in these markets give firms an incentive to offer greater availability (Dana 2001). Markets with more senior citizens also have significantly fewer stockouts, likely because senior citizens – with pronounced preferences and low time costs – are more likely to shop elsewhere when their preferred item is out of stock (Peckham 1963).
The stockout rates are also linked to store and product characteristics. Returns to scale seem to lead to lower stockout rates at larger stores, and stockout rates are greatest when a store is located far from its primary supplier. Stockout rates are also greater for seasonal items (items that are not offered for sale year-round), which is consistent with the demand for these items being less predictable.7
7F
C. Stockouts and other dimensions of store quality Observed rates of product availability are also correlated with other dimensions of supermarket
quality. I obtain ratings of major supermarket chains from Consumer Reports magazine, which publishes the ratings based on tens of thousands of readers’ responses to questionnaires (See Appendix A for details). Combining these ratings with the CPI data yields a sample of more than 2 million availability observations for 57 supermarket chains.
Conditional correlations of stockouts with various ratings from Consumer Reports are presented in Table 3.8 I regress the stockout indicator on each rating, fixed effects for product category, metropolitan area, day of the week, and year-month, and controls for the market, store, and product characteristics shown in Table 2. The standard errors are adjusted to account for within-firm correlation in the error term. The results show that the more highly rated chains are less likely to run out of inventory. Consumer Reports’ overall satisfaction score summarizes readers’ average satisfaction with store quality. While overall satisfaction is ostensibly rated on a 100-point scale, the actual ratings range from 63 to 87. The estimate, which is reported in Column (1), suggests that a 10-point greater overall satisfaction rating is associated with a lower stockout rate by 0.6 percentage points – a 13 percent reduction for the average firm (p < 0.01).
7 The algorithm for identifying stockouts is structured to avoid classifying an unavailable item as a stockout when the item is out of season (see Appendix A for details). 8 These results are also reported in Matsa (2009).
9

Preparing to load PDF file. please wait...

0 of 0
100%
Competition and Product Quality in the Supermarket Industry