Abstract
This article reviews the literature that evaluates WalMart’s impacts on local economies. The authors first describe the methods used to account for potential reverse causality of WalMart’s store location decisions, and then they discuss the literature assessing the company’s effect on three aspects of community life: (a) retail (and nonretail) businesses, across large- and small-sized stores and in different business environments; (b) retail workers, wages, and types of jobs; and (c) producer and consumer welfare through the company’s price-decreasing effect and other potential indirect effects. Last, articles focusing on a broad spectrum of local conditions that could be affected by the company, including poverty rates, social capital, food insecurity, policy effectiveness, and obesity are reviewed. For each dimension, evidence is found of both positive and negative effects, suggesting that we are still far from truly understanding the net effect of WalMart on local economies, let alone the overall consequences in the long run.
There is perhaps no more divisive economic development strategy than that of locating a new WalMart store in a community. More than with any other chain, the announcement of WalMart’s arrival immediately divides community residents into opponents concerned about the loss of businesses on Main Street as well as other local businesses, on one hand, and advocates who point to new jobs, higher tax revenues, and lower consumer prices, on the other. An extensive literature has emerged on the local impacts of such stores, spanning economics, sociology, and regional science among other fields. This article reviews the approaches used by academics for analyzing the economic impact of WalMart on the communities where it locates and the most salient findings of such approaches. The primary focus of this survey is on the academic literature, including articles in peer-reviewed journals, working papers, and proceedings of professional conferences. 1 The survey updates and complements the literature review by Basker (2007), covering more recent articles on the subject while at the same time providing a different contextualization and perspective of some of the literature already reviewed. 2
We first highlight the analytical challenges that arise in assessing the independent effects of WalMart’s presence because of potential reverse causation (explained below). We next present the stylized facts about the effects of the chain both globally and within a given community—illustrating why it is so contentious. We explore research findings on the impact of the company on local employment, wages, businesses, income, output prices, social capital, and other factors. We divide this literature into several categories: impact on the local business environment, labor conditions (jobs and wages), consumers, and other factors such as local voting behavior and social capital.
Measuring the WalMart Effect: Estimation Issues
Studies assessing WalMart’s effects on local economies confront the empirical problem of having to isolate whether an economic outcome (e.g., higher poverty rates, as in Goetz & Swaminathan, 2006) is caused by the company’s presence or whether it instead drives the company’s location decision. For example, a negative relationship between WalMart’s presence and retail wages or employment levels could be due both to the company’s presence and to WalMart locating in areas where wages or employment are low. This would imply a reverse causation or endogeneity bias, which would make statistical estimates of the WalMart effect unreliable.
One way to test for such reverse causation is to assess whether WalMart’s entry decision in a particular area is driven by past (or by current) economic conditions. Hicks and Wilburn (2001) and Hicks (2007a, 2008) used the same method as a test for endogeneity of the company’s location decisions and found little evidence of it. Although this is a simple strategy for inferring the determinants of the company’s entry, it does not per se rule out completely the presence of endogeneity bias (as the authors admit); furthermore, as this method is used in state-specific studies (in West Virginia, Maryland, and eight Pennsylvania counties, respectively) the variation in economic conditions across areas could have been too limited to produce significant estimates.
Instead, to address formally reverse causality issues, researchers have used two-stage instrumental variables (IVs) estimation procedures, where the number of stores or some transformation is regressed on predetermined variables, and the predicted values are used in place of actual ones in the “second stage” equation of interest. 3 Basker (2005a) uses company-assigned store numbers as a proxy for planned store opening dates. Treating the total number of stores opened every year as predetermined, she uses the progressive store numbers to assign a “planned year” of opening to each WalMart store in her sample, which covered the period from 1977 to 1998. By aggregating this number to the county level, she uses the number of planned store openings in each county in each year (i.e., the number of stores that would have existed in the county in a given year, if stores had opened in the order in which they were numbered) as an instrument for the actual number of store openings.
Goetz and Swaminathan (2006) obtain an instrumented growth of the number of WalMart stores between 1987 and 1998, using interstate highway access, retail market pull factor, per capita earnings, population density, per capita property tax, average commuting time to work, percentage of households with more than three vehicles, and number of female-headed households. They measure these at the beginning of the growth period to mitigate endogeneity problems.
A different identification strategy finds its rationale in the company’s expansion pattern, which follows the “hub-and-spoke” location/logistic strategy. This strategy is described best in WalMart’s founder Sam Walton’s words:
We figured that we had to build our stores so that our distribution centers, or warehouses, could take care of them, but also those stores could be controlled. . . . Each store had to be within a day’s drive of a distribution center. So we would go as far as we could from a warehouse and put in a store. Then we would fill in the map of that territory, state by state, county seat by county seat, until we had saturated that market. (Walton & Huey, 1992, pp. 140-141)
This resulted in new stores being located in areas surrounding the company’s headquarters in Benton County, Arkansas, in the early years, while in later years the company tended to open more stores at greater distances. This pattern changed after 1996, when new store openings appear to be less related to distance from the company’s headquarters, perhaps because of the company diversifying into food retailing with its Supercenters (Bonanno, 2010). Neumark, Zhang, and Ciccarella (2008) exploited the fact that at least until 1995, the likelihood of new store openings followed a specific pattern consistent with the hub-and-spoke logistics. They used the inverse of distance from Benton County, Arkansas, and year dummies to isolate exogenous variation in the number of store openings (i.e., the difference in number of stores from one year to another), so as to remove endogeneity. Employing a similar approach, Dube, Lester, and Eidlin (2007) used different rings of distance from Benton County interacted with year dummies as explanatory variables in their first-stage equation, along with linear and quadratic distance terms.
The use of IV methods has not proven conclusive in resolving the issue of endogeneity of WalMart’s store locations, due mostly to the lack of unequivocal acceptance of any one method. For example, Neumark et al. (2008) questioned the approaches of Basker (2005a), Goetz and Swaminathan (2006), and Dube et al. (2007). In response, Basker (2006) showed that the instruments used by Neumark et al. are likely correlated with the unobservable drivers of WalMart’s location decision, as they are likely to capture variations in local business cycles. 4 Basker concludes that, even if distance and time are strictly exogenous variables, “exogeneity does not automatically mean an instrument satisfies the exclusion restriction” (p. 20). To the extent that most markets are now saturated with WalMart stores, distance from Bentonville has become questionable as an instrument.
Some researchers have implemented both approaches described above to evaluate whether they lead to different results. For example, Hicks (2008), in his analysis of WalMart’s impact on labor in Maryland counties (discussed below in the section on workers and wages), uses three identification strategies to control for the company’s location decision endogeneity—one inspired by Basker’s (2005a) approach, one by that of Neumark et al. (2008), and a third that used the sum of personal income in the county of interest and adjacent ones as a predictor of WalMart’s entry. He finds that the identification strategy chosen did not affect his results, although he does not provide econometric tests to evaluate the advantages of one approach versus another.
As result of this debate, researchers are exploring different approaches to circumvent the endogeneity issue. For example, Drewianka and Johnson (2010) use a carefully crafted “event analysis” that takes into account preexisting local market dynamics. They argue that their “approach identifies WalMart’s effect as a change that occurs when a store opens, adjusted for both changes that occur at the same time in other counties and for the county’s ongoing trend” (p. 13).
Last, for studies that measure the impact of the company’s food stores (the Supercenters), other approaches have been proposed. Artz and Stone (2006), in their study of WalMart Supercenters’ impact on food stores’ sales in Mississippi, use the number of years since the first discount stores had opened to predict the timing of Supercenters opening. 5 Basker and Noel (2009), in their assessment of WalMart Supercenters’ impact on supermarket prices, use different models to assess the potential bias of their estimates. By using cross-sectional, long-difference (changes over a 4-year period), and panel estimation methods and controlling for the city and product-specific time effects, they assess potential biases of their regressions obtained by ordinary least squares. As a further validation of their approach, they use the lagged number of WalMart discount stores (the mass-merchandise retail operation of the company) in different base years as an instrument for the number of Supercenters and reestimate the cross-sectional model, obtaining results that are closer to (although, as the authors point out, not the same as) those of the long-difference and panel models. Furthermore, consistent with Basker’s (2006) criticism of distance-based instruments, Basker and Noel find that the use of an instrument based on the distance from the company’s food distribution centers may not be appropriate in a cross-sectional context.
The WalMart Effect
The study of WalMart has been of particular interest to researchers primarily because of the chain’s enormous scale of operation. While previous retailers have threatened existing paradigms, such as the Sears catalogs, which allowed consumers to bypass rural retailers as early as 1888, none acquired WalMart’s scale, scope, and prominence in the everyday life of consumers. The fact that WalMart allegedly “destroyed” mom-and-pop businesses (Boyd, 1997), the bedrock of the American economy, and that it did so in perhaps unfair ways by exploiting both suppliers and employees, further fueled that interest. More recently, the chain has been found to contribute to the trade deficit (Basker & Pham, 2008) and outsourcing of manufacturing jobs, among other economic issues facing the U.S. economy.
In trying to understand the economic impacts of WalMart on local areas, each likely positive (negative) effect is accompanied by a negative (positive) counterpart, underscoring the difficulty of deriving blanket conclusions about the chain’s net effect (Irwin & Clark, 2006). For example, WalMart is argued to lower prices but exploit labor, increase business efficiency but reduce wages, provide jobs but increase unemployment, and to be an industry leader while eroding downtown business centers (Gereffi & Christian, 2009).
Perhaps the most important impact on retail resulting from WalMart’s entrance into a community is its effect on the support services provided to downtown retailers by myriad businesses, often independently owned and operated. Hidden from view in the traditional supply chain or Main Street retail operation, these include lawyers, accountants, bookkeepers, commercial printers, advertising services, truckers, and other service providers. In the WalMart world, all of these businesses become superfluous, as the functions are provided out of Bentonville headquarters or outsourced overseas. Because these functions, even though they were provided locally, usually were not visible or transparent in the communities affected by the opening of a WalMart, this indirect impact of WalMart is rarely discussed by the public, media, or academics.
The question remains, of course, whether the net impacts of a new WalMart store within a community are positive or negative. Several studies have evaluated the different components of such economic impacts, often producing contrasting empirical evidence. A different but related research question is based on providing counterfactual evidence with which one can infer whether local areas where WalMart chose to locate would have been better or worse off had the company not located there. We start our illustration of the “WalMart effect” with the effect on existing retail businesses.
The Effect of WalMart on Retail Businesses
The seminal work on the local economic effects of WalMart’s expansion was carried out from the late 1980s through the mid-1990s by Kenneth Stone (Stone, 1988, 1995, 1997). His work is summarized elsewhere in this special issue. In the same period, antitrust investigations surrounding the chain and other large-scale retailers were published in the Shils Report (Shils & Taylor, 1997), but these had little public policy impact, given the political climate at that time. Another important paper from this period is from Boyd (1997), who points to the decline in total U.S. retail establishments and the rise in real sales per establishment between 1954 and 1992 as evidence of WalMart’s impact. 6
To fully assess whether WalMart’s arrival in a community causes a decline in the local business environment, one can start by measuring how incumbent retailers react to WalMart’s entry. Empirical evidence exists that other retail businesses react differently to the company’s entry depending on their size, scope, and the importance of a particular local market to the firm that is facing competition from WalMart. In one of the first econometric analyses of the impact of the company on other stores’ performances, Capps and Griffin (1998) estimate how WalMart’s (and other competitors’) presence affected the sales of one traditional grocery chain (David’s Supermarkets) in the rural areas surrounding the Dallas–Fort Worth area for the period 1987-1994. Their findings indicate that the presence of one WalMart store (on average) reduced sales of the traditional grocer by 21%, an effect larger than those of all other competitors combined (−19%). Because in this study, however, endogeneity of the company’s location decision is not accounted for, and because the reaction of David’s Supermarkets to WalMart is not considered, such estimates could be biased downward.
Khanna and Tice (2000) used zip code–level data over 1975-1996 to identify 862 local markets with one discount store in at least one year of the period considered and 69 different chains competing with WalMart locally. They find that firms more focused on the discount business tended to invest more heavily after WalMart entered. Also, the more a retail company depended on a particular local market, the more likely it was to invest aggressively to fight WalMart’s entry. Less crucial markets were more likely to see disinvestments. Similarly, Franklin (2001), who focused on WalMart’s impact on food-retailing concentration during 1993-1999, finds that in 13 out of 19 metropolitan areas where WalMart Supercenters had more than 5% of grocery market share, food retail chains increased capital investments and built more stores. Neither of these explicitly considers issues of location endogeneity (which could result in sample selection bias), although Franklin uses an approach similar to that of Hicks and Wilburn (2001)—illustrated above—by treating the company’s location as if it were independent of regional growth conditions.
Artz and Stone (2006) study the impact of WalMart Supercenters’ expansion on food store sales in Mississippi’s 82 counties, using 16 years of data (1990-2005) and a difference-in-differences approach. Their findings indicate that, within 2 years of opening the first Supercenter in that state, WalMart was able to capture approximately 17% of grocery sales in nonmetropolitan areas, while in metropolitan ones the company was able to capture 4% of grocery sales within 1 year of entering the market. Artz and Stone correct for endogeneity in the timing of the Supercenters’ openings using years of presence of discount stores.
Recent studies have included greater data detail or spatial resolution but have not been national in scope, instead presenting more detailed estimates of the impact of the company on the structure and performance of local businesses. Singh, Hansen, and Blattberg (2006) investigate how the entry of a WalMart Supercenter affected sales of a grocery store belonging to a large supermarket chain located in a small suburban town in the Northeast. They use a database of shopping information obtained from the frequency shopper cards of 10,000 customers for the 20-month period from November 1999 through June 2001. Accounting for interpurchase time and basket size, and using a hierarchical Bayesian econometric approach, they find that after WalMart located a Supercenter in the proximity (2.1 miles) of the store in August 2000, the store lost 17% of its volume (equal to approximately a quarter million dollars of monthly sales). Such losses came from a reduction in the frequency of purchases (fewer store visits) rather than by changes in the composition of the market basket (i.e., consumers changed patronage but not their overall shopping habits).
Paruchuri, Baum, and Potere (2009) use 25 years of data on 1,603 Florida zip codes to examine the effect on business creation and destruction of WalMart’s expansion from one store in Florida in 1983 to 133 stores in 2004. They carefully control for endogeneity, after reviewing earlier approaches in the literature, and identify “strong endogeneity effects on both the entry and exit of . . . [other] retailers” (p. 223). They find that within the same zip code, WalMart suppresses the entry of new retailers but does not cause more businesses to fail. In adjacent areas, on the other hand, business failure rates rise by more than do new business entry rates. This is consistent with the chain “sucking” retail dollars into a more centralized business hub, similar to Stone’s results in Iowa, although effects vary depending on whether the competing retailer represents furnishings, salons, antiques, or photo studios. Paruchuri et al. conclude that WalMart “reduces the local class of entrepreneurs” (p. 228). Investigating the relatively newer phenomenon of WalMart’s expansion into urban areas, Davis et al. (2009) conduct an in-depth analysis of the chain’s penetration into Chicago, the first truly urban location. This article appears separately in this special issue.
Ailawadi, Zhang, Krishna, and Kruger (2010) selected seven markets in the northeastern United States, where WalMart opened its first stores between 2000 and 2002, to investigate how different types of stores (supermarkets, drug stores, and mass merchandisers) operating in proximity (less than 15 miles) to the new WalMart stores reacted to the company’s entry, comparing their strategies with those of stores that had not been exposed to WalMart or that had enough time to adjust to the company’s presence. Their findings indicate that, although in many cases retailers did not significantly change their marketing strategy, different reactions to WalMart’s entry could be observed: Stores competing with WalMart for the first time tended to cut both prices and assortments; larger stores tended instead to resist price cuts and excessive promotions, while supermarkets were (on average) more likely to cut prices and mass merchandisers to increase promotions. These authors also indicate that sales decreased for more than 65% of the mass merchandisers analyzed (median decrease 40%), while smaller decreases in sales are observed in supermarkets (median decrease 17%) and drug stores (6%). Similar findings are reported by Cleary and Lopez (2011), who examined the impact of a Supercenter’s entry on supermarket conduct in the Dallas–Fort Worth area using scanner data on milk sales over the 5-year period of 1996-2000. Using a structural model and accounting for endogeneity of Supercenter location, they found that WalMart attracts price-sensitive consumers away from traditional retailers, who react mostly by reducing prices (competition in promotion is found to not be statistically significant).
Focusing on the Washington, D. C. area, Haltiwanger, Jarmin, and Krizan (2010) ask whether big-box stores complement or substitute for mom-and-pop stores. They find that the entry of big-box stores in general has a “substantial negative impact on employment growth and survival of single unit and smaller chain stores that operate in the same detailed industry as the Big-Box” (p.129), but that this negative effect declines with distance and it varies by sector. In this vein, Jarmin, Klimek, and Miranda (2009) argue that big-box retailers are only imperfect substitutes for “single unit retailers,” which continue to show high entry rates nationally. Chiou (2009), however, using national survey data to analyze the impact of WalMart’s presence on DVD sales by other retailers and focusing on central and southern California for a counterfactual experiment, finds that WalMart’s entry not only affects market shares of stores more similar to the chain, such as Costco (−12%), Target (−10%), and Kmart (−8%), but also those of specialized stores (ranging from −16% for Media Play to −6% for Best Buy). Such results suggest that WalMart not only competes head-to-head with unspecialized large-scale retailers but also that it wins sales from other, specialized stores. 7
An exception to the trend of using data at a higher level of resolution, but for limited geographic areas, is Sobel and Dean’s (2008) study of WalMart’s impact on the overall size of small business using state-level observations. Focusing, among other measures, on the number of small-sized retail establishments (1-4 employees and 5-9 employees) per 100,000 people, as well as the self-employment rate, they found that WalMart had on average no statistically significant detrimental effect on mom-and-pop stores. They conclude that
while the entry of a specific WalMart store might cause some individual, small mom and pop businesses to fail, consistent with Schumpeter’s theory of creative destruction, these failures are completely offset by the entry of other new small businesses somewhere else in the economy. (p. 692)
Last, it is worth pointing out that, given the evidence of the company charging lower prices (discussed in the section on consumers below), consumers could find its stores more appealing during economic downturns. This provides relief to cash-strapped consumers hit by economic hardship and it could, in turn, hurt other retailers. Although no direct evidence exists, Basker (2011) estimates that the (aggregate) revenue elasticity at WalMart is negative (in a range between −0.5 and −0.7), suggesting that (with everything else being constant), the merchandise, and hence the shopping experience, at WalMart is characterized by inferior goods. One of the possible reasons for this is that households perceiving the discounter as normal are less numerous than those perceiving it as inferior. If this is the case, WalMart sales could have benefitted (relatively speaking) from shrinking consumer incomes during the most recent economic downturn, whereas other retailers struggled to cope.
The studies outlined above provide evidence that WalMart has a strong impact on the texture and structure of local retail businesses environments, and that in studies of limited geographic scope, such impacts appear to be detrimental (in particular for small businesses). Such results suggest that WalMart not only competes head-to-head with unspecialized large-scale retailers but that it also wins sales from other, specialized stores. Analyses focusing on larger areas, however, found that WalMart had a zero-sum effect on the structure of retail trade, which suggests that positive effects may be in place in other areas, perhaps because of the effective reactions that other leading retail chains implement to counteract the company’s expansion.
The Effect of WalMart on Workers and Wages
Given WalMart’s impact on the structure of the local retail environment described above, one would expect that local retail labor conditions would change as well. The expansion of WalMart could be of particular concern for workers, especially if the company pays lower wages than its competitors and offers fewer benefits (Boarnet, Crane, Chatman, & Manville, 2005; Miller, 2004). Shils and Taylor (1997) report, for example, that in the mid-1990s, one half of WalMart associates received wages only slightly above the prevailing Federal minimum wage of $4.25/hour, and that many of the company’s full-time employees were food stamp recipients. Local policymakers have sought unsuccessfully to pass regulations targeting big-box retailers to improve workers’ conditions by imposing either higher wages and hourly benefits (e.g., Chicago’s “living wage” ordinance) or a mandatory contribution to public health care expenditures (The Maryland Fair Share Health Act). Boarnet et al. present detailed benefit and cost calculations associated with the arrival of WalMart in the San Francisco Bay Area to show the distributional effects on consumers (in terms of price savings) and on workers employed in retailing (changes in wages and benefits). They outline planning strategies to help communities decide whether or not to allow a store to be located within a particular community.
As a result of this interest by policymakers, the economic literature aiming to measure the impact of WalMart on retail labor and wages has flourished. Many studies have focused on assessing the impact of the company on workers (mostly retail labor) using individual states (or parts of them) as case studies, while more recently there has also been an increase in nationwide analyses. We discuss the state-level case studies first, moving to nationwide ones afterward. In proceeding through this section, the reader should be aware that the findings of most of the articles discussed below “are relative to a counterfactual of what would have happened to retail employment absent the effects of WalMart” (Neumark et al., 2008, p. 486). This means that those studies comparing pre– and post–WalMart entry labor conditions would capture how employment figures in areas with WalMart compare with those of a hypothetical scenario without the company, all else being equal.
WalMart and workers: State-specific studies
In one of the first investigations of WalMart’s impact on labor, Ketchum and Hughes (1997) focused on Maine’s 16 counties during the period 1990-1994, using data from the Maine Department of Labor. They estimate the impact of the company’s entry on per capita employment (employment per 1,000 population) and on average wages for retail, services, and the manufacturing sectors. They determine that, across sectors, there is not a statistically significant difference in per capita employment and wages across the 12 counties where WalMart located in Maine and those where it did not. Given the early date of this study, there were no explicit controls for location endogeneity. Hicks and Wilburn (2001) studied the impact of WalMart on retail employment in West Virginia, using county-level data covering the period 1988-2000, correcting for spatial autocorrelation, and controlling for WalMart’s presence in neighboring counties. They find that the company’s presence had no impact on average retail wages, leading instead to a small increase in permanent retail employment (of approximately 54 workers) while supporting growth in the number of retail establishments.
Keil and Spector (2005) examined the effect of WalMart on labor market outcomes of African Americans in Alabama, as well as relative income differentials across socioeconomic groups. Although they discuss in-migration of workers with lower skills into communities containing a WalMart store, they fail to address the implicit endogeneity concern that this raises. The authors conclude that the presence of a WalMart increases employment opportunities for minorities, thus reducing their unemployment rates, and while the chain’s presence has a statistically significant impact on income differentials (which are reduced), that effect is weak.
Focusing on eight Pennsylvania counties in which at least one WalMart store opened during 2001-2005, Hicks (2007a) uses quarterly workforce indicators data to illustrate the impact of the company’s entry and expansion on labor dynamics. Discarding issues of endogeneity of location by virtue of the fact that all the counties in the analysis experienced entry by WalMart (instead of appropriately testing for it), Hicks finds that WalMart’s entry has little impact on average retail earnings, while it is associated with an increase of approximately $90 in monthly earnings for new hires. His results indicate that net job flows (i.e., new jobs created) were unaffected by the company’s presence; job separation (i.e., job losses) was unaffected by WalMart’s entry but declined in the quarters after entry (by approximately 150 jobs); job creation decreased with entry by approximately 144, although new hires at entry were equal to 180 jobs (approximately the number of workers hired by WalMart when a store opens); and the company’s presence is associated with a marked reduction in job turnover. Such figures result in a decline of total employment equal to 181 workers at entry, with the effect wearing off in subsequent quarters.
Last, Hicks (2008) tested for the impact of the company’s expansion on labor figures (retail employment, retail earnings, and aggregate employment) in Maryland’s 23 counties during 1988-2003, using data from the Regional Economic Information System, while controlling for potential endogeneity of a WalMart’s exposure variable by means of the different approaches illustrated in the first section of the article, on estimation issues. His findings indicate that, regardless of whether endogeneity is accounted for, the presence of the company in Maryland results in a loss in countywide annual retail employment of between 248 and 408 workers and an increase in hourly retail wages of between $1.07 and $1.94 per hour, suggesting that both effects could be due to increased marginal productivity of labor. He reports no evidence of the company’s expansion affecting aggregate employment.
WalMart and workers: Nationwide studies
The first large-scale and statistically rigorous analysis of WalMart’s effect on net job changes was Basker’s (2005a). Basker combined self-collected WalMart location store data with publicly available county-level data (primarily from the County Business Patterns) for the contiguous United States, using an identification strategy based on planned store openings, as described earlier. Basker estimates that the arrival of a WalMart is associated with a net immediate-term gain of 100 jobs, which, in the longer run, declines over 5 years to an average of 50 jobs (similar to Hicks & Wilburn’s, 2001 findings); no discernible effect was found on retail employment in neighboring counties. Along with these retail figures, she finds that the presence of the company caused a loss of 20 higher paying wholesaling jobs and a slight increase in restaurant employment. A shortcoming of Basker’s analysis is that the data do not allow her to distinguish between full- and part-time employment, which is likely critical in this context. Another shortcoming is that Basker’s sample only includes counties with employment levels above 1,500 in 1964 and with positive employment growth, which may lead to sample selection bias, as Goetz and Swaminathan (2006) point out.
Neumark et al. (2008) investigated instead the impact of WalMart’s entry on both retail labor and per capita retail earnings, as a proxy for retail wages. They combined administrative WalMart store location data with County Business Patterns data, focusing mostly on the period 1977-1995, for which their identification strategy (discussed above) seems more effective. For each job created by WalMart, 1.4 jobs are estimated to be lost in the local economy, resulting in a county-level reduction of approximately 146 workers, or 2.7% of the average retail employment. In addition, aggregate retail payrolls decline by $1.1 to $1.7 million, or approximately 1.5%, while retail earnings per worker appear to be unaffected. However, the authors point out that, given the strong growth in retail employment during the period considered, the “employment effects of WalMart that we estimate simply imply that retail employment growth was a bit more modest than it would otherwise have been” (p. 428). It should be noted that using similar identification assumptions (which, as discussed in the first section of this review, have been criticized in Basker, 2006), Dube, Lester, and Eidlin (2007) found that a new WalMart store reduces county-level average retail earnings by 0.5%, while at the state level, 10 new WalMart stores cause a reduction of per capita earnings by 0.5% to 2.0%, mostly because of a reduction in labor market rents. As in Basker (2005a), both Dube et al. (2007) and Neumark et al. (2008) fail to distinguish between full- and part-time employment.
Bonanno and Lopez (2012) approach the issue of WalMart’s impact on workers from a different standpoint. Using a 2006 cross-section of county-level observations for the contiguous United States, they evaluated whether WalMart is able to exert monopsony power over retail workers. 8 Bonanno and Lopez found that, even though the amount of monopsony power of WalMart over workers is on average limited (per capita retail earnings are less than 3 percentage points below the competitive level), in the rural South, where the company has the largest and longest presence, its anticompetitive behavior is more substantial. While these results may potentially call for scrutiny by antitrust authorities, Bonanno and Lopez caution that, in the absence of WalMart, retail labor figures could in fact be even lower, and that the increased consumer surplus from lower prices (discussed in the next section) also needs to be accounted for to determine the impact of any policy intervention. As in other studies cited, the scope of Bonanno and Lopez is limited to retail workers alone, and the authors do not account for spillover effects of industries outside retailing. Also, they treat retail labor as homogenous, disregarding not only the distinction between part-time and full-time employees but also the existence of jobs for which different skills are needed.
Using an “event analysis” approach instead of a two-stage procedure on a panel of county-level observations in the continental United States, Drewianka and Johnson (2010, p. 38) conclude that the presence of WalMart slightly increases local employment in the retail sector (up to approximately 155 to 162 additional jobs in the long run), and that establishment counts in other retail subsectors are essentially not affected by the presence of WalMart. While their study is remarkable in terms of the empirical effort, questions about the methods persist: For example, their findings suggest that WalMart tends to build more stores when retail employment is falling (relative to other controls), which implicitly highlights potential issues of reverse causality. Also, similar to most of the studies illustrated above, Drewianka and Johnson’s work ignores spillover effects on and from contiguous counties, and they are unable to distinguish between full- and part-time employment.
A different aspect of the impact of WalMart’s presence on workers is the trade-off between jobs of different types that are created or displaced in the presence of the company. Goetz and Shrestha (2009) find that the presence of the chain is associated with higher earnings from self-employment, which they attribute to the creative destruction process, whereby more marginal small businesses are driven out first by the arrival of the chain. Although this study falls short of investigating whether and how the mix of self-employed workers changes in affected communities, its results are consistent with findings of large-scale studies investigating labor productivity growth in retailing during the 1990s, which point to more productive establishments having replaced less productive ones (see, e.g., Foster, Haltiwanger & Krizan, 2006).
Kolko and Neumark (2010) investigate whether locally owned businesses provide more employment stability during economic downturns, protecting the local community from layoffs and related externalities. If this were the case, a local WalMart store would likely eliminate more jobs in a recession than would a similar store owned by a local resident. The authors find evidence that local ownership leads to more stability, and that headquarters of major chains are the most attractive to a community in terms of providing job stability over time, but the authors do not provide sufficient sector detail to draw conclusions for big-box retailers.
Summarizing, the literature on WalMart’s impact on workers provides conflicting findings. While early studies pointed to either no or a small positive impact, as noted by Basker (2007) in her literature review, as the empirical analyses of the effect of the company on local labor figures proliferate, a consensus remains elusive.
The Effect of WalMart on Consumers
In this section, we examine the effects of WalMart on consumers, mainly through the lower prices it offers. As the company focuses its marketing effort on attracting consumers into stores by highlighting “low prices” (the company has recently moved away from the “Always low prices—always” slogan to a perhaps even more effective one: “Save money. Live better”), the alleged low prices the company touts have become an empirical conundrum worth investigating by academics. In particular, if other retailers adjust to WalMart’s presence by modifying their pricing strategies (as pointed out in the section on WalMart’s effect on retail businesses, above), the company’s presence could also alter the overall price levels faced by consumers. The effect of WalMart on retail prices and inflation rates (Hausman & Leibtag, 2005) comes from two distinct but related sources: an indirect “aggregate mechanism” and a direct “market mechanism,” as explained by Basker (2005b):
The aggregate mechanism works through WalMart’s interactions with both suppliers (manufacturers and importers) and other large retail chains. This mechanism can lower prices in communities not served by WalMart if it leads to lower costs for other retailers. The market-specific mechanism works through competition (and possibly learning) at the local level. (p. 204)
Both effects can influence the economic conditions of local areas in ways that extend well beyond those areas where WalMart operates. While the focus on understanding the supply-side aggregate mechanism has been very modest so far, several studies have analyzed the impact of WalMart on prices through the “market mechanism.” The practical implication of WalMart’s “pro-competitive” effect is that, by pushing other retailers’ prices down (Basker, 2005b; Basker & Noel, 2009; Cleary & Lopez; 2011; Volpe & Lavoie, 2008), consumer welfare increases (Cleary & Lopez, 2011; Hausman & Leibtag, 2007), which benefits the entire local area and not only WalMart shoppers. 9
Not enough evidence exists as to whether the benefit for shoppers results in an overall increase in total welfare, mainly because a lowering of prices could merely represent a pecuniary transfer from producers to consumers. 10 If, in fact, there are no market frictions and cost reductions are passed on to consumers (or, alternatively, if lower prices are achieved only by reducing input costs, i.e., the alleged lower wages and poorer worker conditions), then the overall impact of lower prices is zero because welfare is simply redistributed. This is unlikely to be the case, however, given real-world frictions in vertical channels and imperfect competition (or lack of perfect information and foresight). As retailers usually benefit from a certain level of pricing power with respect to consumers, WalMart’s presence could push them to become either more competitive or more efficient (see more below), thus producing a positive welfare effect.
Other types of frictions may also exist. As retailers compete with WalMart, they need to restructure their operations, which may result in job losses or displacement of business opportunities (see previous sections). Similarly, the “aggregate effect” hypothesized by Basker (2005b) may perversely lead other retailers—in particular smaller ones—to raise prices. Foer (2007) illustrates this potential indirect price-increasing effect of large retailers on their competitors by using Dobson’s (2005) “waterbed effect.” As large retailers gain bargaining power over their suppliers by buying goods at the marginal production cost, the same suppliers may be forced to charge higher prices to smaller retailers, who have little to no bargaining power, to cover fixed costs. As a result, although consumers may be worse off in the short run because of the higher prices, in a long-run scenario they will likely switch to the low-priced retailers (such as WalMart), and as a consequence, small retailers, mostly managed and operated locally, are more likely to be hurt. Whether WalMart is responsible for “gouging” or for reducing other retailers’ costs has yet to be established. With these caveats in mind, the procompetitive effect of WalMart’s presence has been documented on several occasions, across data sources, methodologies, and scopes of analyses. A brief overview of relevant findings follows.
Basker (2005b) offers the first wide-scale analysis of the impact of WalMart’s entry on prices. Focusing on the 21-year period 1982-2002, she combines quarterly price data for 10 consumer goods (including such items as toothpaste, aspirin, 2-liter bottles of Coca-Cola, and men’s underwear) from the American Chamber of Commerce Research Association with self-collected data on the location of WalMart stores across 165 U.S. cities to capture variations in prices triggered by WalMart’s entry. Basker’s estimated price-decreasing effects are modest in the short run, ranging on average from 1.5% to 3.0%, whereas they become four times as large in the long run.
Hausman and Leibtag (2007) used a customized 4-year (1998-2001) subset of AC Nielsen Homescan data containing actual consumers’ purchase information to investigate whether there is a relationship between consumers’ expenditure share in nontraditional food retailers and price levels. They find a negative and statistically significant relationship between the share of shopping at nontraditional stores and prices across 20 food categories, resulting in a decrease in the average price consumers pay. Such a reduction in price levels is the result of a shift in both consumers’ purchases and retailers’ price competition. These authors also estimated the increase in welfare associated with having access to Supercenters to equal, on average, 25% of food expenditures, made up of a 4.8% component of lower prices due to increased competition and a 20.2% component due to increase in variety. 11
Volpe and Lavoie (2008) analyzed primary price data on 54 food products collected from six WalMart Supercenters, six large chains operating in the proximity of a WalMart Supercenter (each less than 5 miles from a Supercenter and referred to as a “competing store”), and six comparison stores with characteristics similar to those of the competing stores, across Massachusetts, Connecticut, and Rhode Island. On average, food items sold at Supercenters tend to have lower prices than at traditional retailers, but such differences are smaller for the competing traditional stores, suggesting that WalMart’s procompetitive effect is stronger for neighboring stores than for more distant stores.
Basker and Noel (2009) analyzed the impact of WalMart Supercenters on other stores’ prices, using annual store-level data from the American Chamber of Commerce Research Association for 24 categories, including foods such as dairy products, meats, produce, canned goods, and frozen goods. Their analysis covered 175 markets and relied on fixed-effect static models in levels, long-run difference models, and panel models, to estimate both short-run and long-run price effects. 12 Basker and Noel find that on average WalMart’s presence has a price-reducing effect on incumbent supermarkets’ own prices of between 1.0% and 1.2%. They also point out that such a response comes primarily from small competitors, while the response from what they call the Big Three competitors (Albertsons, Safeway, and Kroger) is less than half that size.
Consistent with the hypothesized price-decreasing “market effect,” the papers discussed above measure how the presence of WalMart leads competitors to react by reducing their own prices. Little research has tried to assess how much of such price reduction comes from changes in logistics systems (i.e., cost reductions) or simply from changes in market conduct. Cleary and Lopez (2011), for example, measure the impact of the growth of WalMart Supercenters on the anticompetitive conduct of five traditional retail chains in a limited geographic area (Dallas–Fort Worth), applying a structural model to 5 years of monthly scanner data of sales of a staple food product (milk) as a proxy for their overall conduct. WalMart is found to reduce traditional retailers’ oligopoly power, leading to a significant increase in consumers’ welfare. However, besides allowing for a reaction in promotion strategy, their model does not consider traditional retailers’ other adjustments to WalMart’s entry, such as changes in logistics or other cost-reducing investments, which could potentially result in job losses and reduction in workers’ welfare.
Last, we mention the work of Stater and Visser (2008), who use a standard monocentric city model to analyze how big-box stores may affect prices paid by consumers, consumer utility, and land values. In their model, the big-box locates in a suburb at some distance from the central business district, as is frequently the case in practice; existing retailers in the urban core experience a reduction in profits, also charging lower prices as they are forced to compete with the lower cost discounter (the authors abstract from commuting costs). As prices fall, consumer utility rises, and the cost of land falls in the central business district while rising in the area surrounding the big-box. Although this article does not provide an empirical test, it sketches a useful research agenda for further investigations into the local economic impact of WalMart, with land values being a key variable.
Other Effects of WalMart
Given the enormous interest generated by WalMart, researchers have examined other potential effects of the company on local communities. For example, Goetz and Swaminathan (2006) found that counties with existing and new WalMart stores had greater difficulty reducing poverty rates between 1989 and 1999 than did those counties without WalMart stores. Using an identification strategy discussed in the introduction of this review, their study revealed that while counties not receiving a new store saw the average poverty rate drop by 2.4 percentage points, counties receiving a new store saw a decline of only 2.2 percentage points, for a difference of 0.2 percentage points, or almost 10%. Counties without a WalMart store in 1987 reduced their poverty rates by 0.1 more percentage points than did counties with a store. These authors, however, also acknowledge that their study was unable to answer the question of whether the consumer price-reducing effect of WalMart was big enough to offset this smaller reduction in the poverty rate. Similarly, Fleming and Goetz (2011) report that while smaller, locally owned businesses were associated in a statistically significant way with faster per capita income growth in the period from 2000 to 2007 (everything else being constant), large, nonlocally owned firms (such as WalMart) significantly suppressed local income growth rates over the same period. These authors examine the effect of initial conditions (in 2000) on subsequent economic growth and do not explicitly consider endogeneity, arguing that the baseline conditions in 2000 could not be affected by subsequent economic growth.
The first study of how WalMart may affect social capital was by Goetz and Rupasingha (2006), who found that the arrival of a store was associated in a later time period with lower stocks of social capital, fewer nonprofit and charitable organizations, lower rates of voting in the presidential election of 2000, and less religious adherence or church attendance in a county. In short, their results suggested less civic participation by local residents, holding constant other pertinent factors. The authors attributed these results to a general withdrawal of individuals affected by the arrival of the chain, or greater apathy. Carden, Courtemanche, and Meiners (2009a, 2009b) reexamine the hypotheses tested by Goetz and Rupasingha and focus on the effect of WalMart on social capital–related factors such as membership in clubs, religious participation, and time spent with friends, using Census data as well as the DDB Needham Lifestyle survey, which was also used by Putnam (2000). They conclude that there is no robust or consistent WalMart effect on social capital: some measures are being positively affected, whereas others decline. Carden and Courtemanche (2009) investigate the chain’s impact on visits to art galleries or classical concerts and find no consistent negative effect. They suggest that their results are consistent with a broadening of consumers’ consumption portfolio, because of the price-reducing effect of the chain. However, although they try to limit the source of omitted variable bias by controlling for county-level fixed effects, they do not control for reverse causality, which could undermine the reliability of their results.
Given its massive presence, it could also be hypothesized that WalMart’s expansion affects other areas of public concern. The presence of the company has been found to have a potential role in contributing to increasing rates of obesity, for example. Courtemanche and Carden (2011) combined micro-level data from the Behavioral Risk Factor Surveillance System matched with county-level WalMart Supercenters’ location and store-opening data, controlling for the endogeneity of store locations using different approaches including one similar to Neumark et al. (2008). They find that the presence of one additional Supercenter per 100,000 residents causes an increase in the average body mass index by 0.25 units, resulting in an increase of the adult obesity rate of approximately 2.4%. However, comparing the potential health care cost associated with the increases in obesity rates with the increased welfare from the price-reducing effects of WalMart, they conclude that “the resulting increase in medical expenditures offsets only 6% of consumers’ savings from shopping at WalMart. The obesity effect alone is therefore not sufficient to conclude that WalMart entry is bad for communities” (p. 4).
Bonanno (2012) examine how different food retailers affect the likelihood of households becoming food insecure by combining the Current Population Survey’s household-level Food Security Supplement data with Metropolitan Statistical Area–level data on food stores’ density (number of food outlets over population) from different sources. 13 Using the lagged number of per capita discount stores and distance from the company’s food distribution centers as instruments, he finds no overall link between the presence of WalMart Supercenters (per 100,000 residents) and the likelihood of being food insecure, as the company’s direct food insecurity mitigating effect (through lower prices and larger assortments) is offset by a detrimental indirect effect via its negative influence on other food stores’ presence. They suggest that the lower prices and one-stop shopping opportunities offered by the company can partially offset travel costs to the store, providing low-income households with a way to satisfy their alimentary needs.
Last, Hicks (2007b) uses observations consisting of a panel of the contiguous 48 states for a 25-year period (1978-2003) to evaluate whether WalMart’s expansion affected expenditures on low-income federal support programs. He finds that WalMart’s presence measured in terms of both number of stores and labor share is not related to an increase in food stamp expenditures. He also finds weak evidence of a decreasing effect on Aid to Families with Dependent Children/Temporary Assistance for Needy Families expenditures and an increasing effect on Medicaid expenditures, of approximately $898 per worker.
Conclusions
Through our literature review, we encountered a wealth of modeling strategies, methodologies, and data sources used, indicating a strong interest in understanding the economic impact of WalMart on local areas. Below, we distill a few facts that illustrate the current level of knowledge on the subject, and we conclude with indications of areas needing additional work.
What We Have Learned So Far
Effect on local businesses: While small businesses are affected negatively, often being forced to close down and divest, large retailers seem to fight back more effectively; the outcome is the restructuring of local retail environments, resulting in the disappearance of local business centers through a process of creative destruction.
Effect on workers: Results with respect to retail employment, wages, and job displacements are mixed and vary depending on sample size and periods, modeling choice, and empirical methods used to control for the endogeneity of the company’s location decision.
Effect on prices: Empirical evidence indicates that WalMart reduces retail prices, igniting a virtuous (in the sense of “potent, efficacious,” as in a virtuous circle) “procompetitive” effect and increasing consumers’ welfare. There is some indication that such an increase comes not only from lower prices but also from increased variety.
Other effects: Research has pointed out that WalMart may have an impact on social capital, on poverty rates, and on policies and public expenditures at the local level.
Future Research Agenda
Identification strategy: The need for a unifying empirical framework and identification strategy to deal with the endogeneity of the company’s store location decision emerges clearly, in particular from the mixed results obtained when looking at its effect on workers and wages. It is not obvious that this issue will be resolved easily or quickly, given the different preferences of researchers. Nevertheless, addressing this issue would appear to be a major priority.
Worker displacement: The current literature has not investigated issues of relocation of those labor resources that are freed up or displaced by WalMart. Nor have potential adjustment mechanisms to help displaced workers find better employment opportunities been analyzed.
Welfare analysis: While there is research assessing either consumers’ or workers’ welfare changes due to the company’s presence, no comprehensive analysis exists of the overall welfare changes triggered by the company—comparing, for example, how losses due to job displacement compare with savings from lower prices. This would require detailed and intricate knowledge of supply and demand curves and their elasticities.
Consumer spending: More research is needed to understand what consumers’ savings from lower prices translate into.
The big picture: Little is known about the long-term dynamic consequences of WalMart’s growing power, and about how that power influences the overall economy.
In sum, despite the richness of this branch of literature, we are still quite far from having a full understanding of the economic impact of WalMart on local economies.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.