Professional Documents
Culture Documents
Previous academic research on the expansion of dominant retailers such as Wal-Mart has
looked at implications for incumbent retailers, consumers, and the local community. Little
is known, however, about Wal-Mart’s influence on suppliers’ performance. Manufacturers
suggest Wal-Mart uses its power to squeeze their profits. In this paper we study the validity
of that claim. We investigate the underlying mechanisms that may cause changes in manu-
facturer profits following Wal-Mart market entry. Our data contains information on supplier
interactions with retail stores, including Wal-Mart, for a period of five years.
We find that post-entry supplier profits increased by almost 18% on average, whereas
profits derived from incumbent retailers decreased only marginally. Contrary to predictions
from analytical work our results show wholesale prices are not the main driver of post-entry
supplier profit changes; market expansion is. We observe a significant increase in shipments
to 45% of markets studied. Furthermore, our analysis demonstrates supplier shipment and
profit increases are highest for markets in which incumbents offer a wide variety of products
and carry items that Wal-Mart does not sell.
Key words: Wal-Mart, market entry, supplier profits, wholesale prices, shipments, prod-
uct line, assortment.
“It’s a complicated question, but for business people, it is the essential question about Wal-
Mart: Will doing business with Wal-Mart help my business or hurt it?”
1 Introduction
Wal-Mart, reporting over $400 billion in net sales for 2008 (Wal-Mart Stores Inc. 2009), has
been ranked the number one non-oil company on the FORTUNE 500 for the past nine years.
As the world’s largest retailer Wal-Mart has become the biggest buyer for manufacturers
such as Disney, Gillette, Kellogg’s, Mattel, Procter & Gamble, and Sarah Lee (Useem 2003).
Passing-through cost savings and quantity discounts, the low-overhead, low-inventory retailer
won the battle for the grocery dollar by offering consumers its (in)famous “Every Day Low
Prices”.
Although over 200 million people shop at Wal-Mart’s 4000+ U.S. locations each year
(Wal-Mart Stores Inc. 2008), public opinion is not all favorable. Trade unions and special
interest groups (e.g., McGee and Festervand 1998, Kinzer 2004), local politicians (Moreton
2006, Parsons 2006), and the media eagerly portray the retailer as a “world-wide chain of
exploitation” destroying communities while squeezing its employees and suppliers alike (e.g.,
Akron Beacon Journal 2000). Its sheer size and power compelled manufacturers to ask
whether or not to “... say no to Wal-Mart?” (Bowman 1997). The demise of Rubbermaid is
an often cited example of the retailer’s power. When the durable housewares producer, then
one of America’s most reputable companies, tried to pass-on a cost based price increase,
Wal-Mart pulled all Rubbermaid products from its shelves in 1994. As the big-box retailer
was Rubbermaid’s largest outlet, the manufacturer could not recuperate and was forced to
sell-out to Newell several years later. Over 700 suppliers have now built offices close to
1
Previous academic research on discount chains mainly focuses on implications for in-
cumbent retailers (e.g., Stone 1988, 1995, Basker 2005, Gielens et al. 2008, Ailawadi et al.
2009), consumers (e.g., Basker 2005, Hausman and Leibtag 2005, Luchs 2008), and the local
community (e.g., Bianchi and Swinney 2004, Moreton 2006, Goetz and Rupasingha 2006,
Neumark and Ciccarella 2008, Halebsky 2009). Five recent empirical studies on Wal-Mart in
marketing and economics emphasize its market entry impact on incumbents and consumers.
First, studying seventeen products in ten categories, Basker (2005) reports a post-entry de-
cline in market prices. Effects for cigarettes, cola, pants, shirts, and underwear are, however,
not significant. Singh et al. (2006) – focusing on changes in sales and consumer purchase be-
havior following one Wal-Mart entry – conclude the incumbent supermarket studied suffers
a 17% volume loss, mainly resulting from fewer consumer visits. Interestingly, the authors
detect little change in basket size; the number of products bought by the supermarket’s
loyal customers is comparable to pre-Wal-Mart conditions. Third, Jia (2008) finds that the
multinational’s expansion in the 1980s and 1990s accounted for a 34% to 41% drop in the
number of discount stores. Gielens et al. (2008) study incumbents’ stockprice changes after
Wal-Mart acquired UK based Asda supermarket group. The authors demonstrate assort-
ment as well as positioning similarities between Wal-Mart and incumbents are detrimental
to the latter’s post-entry stock performance. Finally, a recent paper on competitive reactions
to eleven Wal-Mart entries by Ailawadi et al. (2009) reports negative effects on incumbent
retailers’ revenues. Moreover, the authors show increasing assortment size can successfully
mitigate entry impact.
Despite discount chains more than doubling their market share since the late 1960s (Jia
2008), analytical and empirical work analyzing entry implications for manufacturers is scarce
and provides inconsistent results. The trade press proclaims suppliers are forced to offer at-
tractive trade deals, merchandise support, and slotting allowances to please powerful retailers
(Bloom and Perry 2001). Although suppliers have reported for years that big-box retailers
demand financially-damaging concessions that undermine channel profits (Huey 1998, Brunn
2
2006), current analytical studies provide limited support for a directional hypothesis on how
dominant retailers’ actions affect supplier performance.
Chen (2003) concludes that a dominant retailer’s bargaining power reduces a manufac-
turer’s share of joint profits. He theorizes that manufacturers should decrease incumbent
wholesale prices, who, in turn, will lower retail prices to boost sales and counter the domi-
nant retailer’s rise. Dukes et al. (2006), on the other hand, suggest that a dominant retailer’s
power need not diminish supplier profits. The authors argue that suppliers should increase
rather than lower wholesale prices charged to incumbents. Their analytical results suggest
that while rival retailers may suffer, manufacturers can achieve enhanced performance by
the incumbents’ assortment and its overlap with the dominant retailer’s product selection
affects their performance. In addition, depending on assortment costs, suppliers may choose
to distribute specialty items through incumbents only. Selling to Wal-Mart, which carries
a shallow assortment in many categories, will therefore impact manufacturers’ profits and
and collaboration with Wal-Mart for most manufacturers who self-identified the retailer as
3
a ’primary customer’. However, Mottner and Smith (2009), who test Bloom and Perry’s
model on a longer time-series of Compustat data, conclude working with Wal-Mart will not
affect supplier performance. In contrast, Gosman and Kohlbeck (2006), using Compustat
and CRSP data, estimate higher gross-margins for Wal-Mart vendors.
In sum, as studies on the costs and benefits for manufacturers of working with Wal-
Mart are scarce and inconclusive and empirical research on the impact of Wal-Mart entry
on supplier performance is lacking, we propose the following two research questions:
The fact that some retailers do not share data with third parties, such as Nielsen and
IRI, is one of the most important reasons empirical work on big-box retailers is still lacking
at present. As Wal-Mart has refused to contribute to national consumer sales databases
since 2001 (Sachdev 2001), our study significantly adds to the scarce empirical literature
examining the company’s impact on suppliers by directly analyzing historical data collected
by a vendor. Our unique database contains detailed records on profits, wholesale prices,
and shipments for a major consumer packaged goods manufacturer, allowing us to quantify
changes in supplier performance following Wal-Mart entry. Our data not only captures the
items shipped to each Wal-Mart every week, but also their price. As our dataset spans
thousands of retail stores and hundreds of products for a period of five years, we are able to
to analyze Wal-Mart entry impact in a broad set of geographical markets while accounting
for market structure effects.
4
model and link entry impact variation to retailer assortment profiles in the second stage
(Dukes et al. 2009).
The remainder of the paper is structured as follows. We introduce the data and the
propensity score method in Section 2 and discuss the hierarchical Bayesian model in Section
4. Section 5 describes our empirical results while Section 6 provides managerial implications.
2 Data
In this study we use weekly data for a packaged goods category offered by a major man-
ufacturer in the industry. Detailed information is available on supplier profits, wholesale
prices, and shipments for over 200 products from December 1999 to January 2005.1 The
data provides complete retail coverage in the U.S. including supermarkets, convenience and
drug stores, and mass-merchandisers. Figure 1 shows the spatial distribution of the 759
Wal-Mart market entries observed in the data. None of these markets has another Wal-Mart
store in the same zip-code area. The number of entries by year are shown in Table 1.
For each market we observe the key dependent variables (profits, wholesale prices, and
shipments) both before and after Wal-Mart entry. Shipments are the number of standard
units of goods transported to a market in a week. Wholesale price reflects the average weekly
price charged to retailers across all goods sold in a market. Supplier profit is measured as
1
For confidentiality reasons we can disclose neither manufacturer nor industry.
5
wholesale price minus supplier cost of goods times shipments. As an initial estimate of entry
impact we calculate the change in weekly supplier profits, wholesale prices, and shipments
in pre- versus post-entry periods (see Table 2).
% change
Supplier profits +22.011%
Wholesale prices +6.247%
Shipments +13.509%
Table 2: Pre versus post Wal-Mart entry change in weekly supplier profits, wholesale prices,
and shipments
Supplier profits increased by 22% on average, shipments to entry markets by nearly 14%,
and wholesale prices charged to retailers by just over 6%. Note that the results in Table 2
should be interpreted with caution. For example, demand for the supplier’s products might
have increased regardless of entry. Since data on the same markets over the same time period
but without entry do not exist, we compare outcome data for markets with and without Wal-
Mart entry. To the extent that markets with and without entry experience similar changes
over time the latter serve as valid controls. Table 3 shows changes in outcome variables for
Table 3: Before and after Wal-Mart entry measures of profits, wholesale prices, and shipments
for entry and non-entry markets
changes common to both types of markets (Angrist and Krueger 1999). Impact estimates on
supplier profits, wholesale prices, and shipments, expressed as a percent of the average value
6
in entry markets before Wal-Mart, are 19% (0.637 / 3.350), 0.2% (0.003 / 1.037), and 12.6%
(0.399 / 3.172), respectively. As each of these effects is smaller compared to the weekly
pre versus post entry changes (Table 2), particularly wholesale price for which the impact
estimate is near zero, the results reported in Table 2 may not be attributable to entry.
Even though these numbers are more reliable than those in Table 2, they are not without
limitations. They have a causal interpretation only if we can reasonably assume that, except
for entry, markets with and without entry are comparable. This would imply that Wal-
Mart selects entry markets at random. The fact that entry markets appear, on average, to
generate much higher supplier profits and shipments suggests such an assumption is highly
questionable. If Wal-Mart is strategic in its selection of markets to enter, the results in Table
3 are invalid. Our approach to dealing with potential selection bias is described next.
3 Selection bias
Matching and Instrumental Variables are two common techniques to correct for selection bias
(Angrist and Krueger 1999). In the context of our study it is difficult to identify instruments
that are correlated with Wal-Mart’s entry decisions but uncorrelated with supplier outcomes
(see Qian 2007, Gensler et al. 2009, Tripathi 2009 for similar arguments). Matching replicates
a randomized experiment by using covariates to pair experimental (EM) and control markets
(CM) (Rubin 2006, Gensler et al. 2009). It ensures that, conditional on covariates, the
Gensler et al. 2009). Propensity Score Matching (PSM) is a commonly used method to reduce
the dimensionality of the matching problem (Rubin 2006). Rosenbaum and Rubin (1983)
have shown that if the conditional independence assumption is satisfied by conditioning on
7
Variable Estimate Standard error
Intercept -22.666∗∗ 2.998
Median age -0.024∗ 0.011
log(Population density) 2.423∗∗ 0.222
log(Population density)2 -0.211∗∗ 0.018
log(Income per capita) 9.919∗∗ 1.976
log(Income per capita)2 -1.519∗∗ 0.329
No. of other supercenters -0.021∗∗ 0.005
Herfindahl index -10.592∗∗ 0.574
N 22186
Nagelkerke’s R2 0.360
∗∗
p-value < .01, ∗ p-value < .05
Table 4: Logit estimates for propensity score matching
covariates (X), it is also satisfied by conditioning on the propensity score P (X). When the
propensity scores for two markets are identical, they are equally likely to receive a treatment
because “as far as we can tell from the values of the confounding covariates, a coin was tossed
to decide who received treatment 1 and who received treatment 2” (Rubin 2006, p. 448).
In our study the propensity score is the probability that Wal-Mart will enter a market
given the value of observables.2 Propensity scores are calculated as the predicted value from
a logistic regression with market treatment as the dependent variable (i.e., 1 if Wal-Mart
entered a market, 0 otherwise) (Angrist and Krueger 1999). To minimize selection bias and
ensure only relevant covariates are included in the model a stepwise estimation procedure
was employed (Rosenbaum and Rubin 1984). Table 4 shows the estimated coefficients and
standard errors for the selected variables.
Population size has been used in previous research on Wal-Mart entry (Jia 2008, Zhu
and Singh 2009). Our estimated non-linear effect suggests entry is less likely for markets
with extremely low or extremely high population density. Although Wal-Mart is known to
prefer lower income markets (Graff and Ashton 1994, Moreton 2006, Vedder and Cox 2006,
Halebsky 2009) we find the retailer avoids both the lowest and the highest income markets.
The negative coefficient for age suggests Wal-Mart opts for areas with younger families (see
2
In this study we equate markets to zip-code areas.
8
alsoSingh et al. 2006). The number of non-Wal-Mart supercenters within a 20-mile radius
captures competitive interaction with Target and K-Mart (Jia 2008, Zhu and Singh 2009).
The coefficient for the Herfindahl index indicates a preference for markets with more but
smaller competitors. We also include state fixed-effects to control for unobserved regional
CMs with propensity scores below the 1st percentile of P (X) for EM and excluded EMs with
propensity scores above the 99th percentile of P (X) for CM. Trimming reduced our sample
size to 629 EM and 10,728 CM. Figure 3 shows the adjusted propensity score distributions.
After ensuring selected markets lie on the overlapping support of observables, markets
with and without entry were paired based on propensity score similarity (Gensler et al. 2009)
using nearest available matching (Rosenbaum and Rubin 1985). The steps in this procedure
are as follows: 1) EMs and CMs are listed in random order; 2) when the first EM is matched
to the nearest CM based on P (X) both markets are removed from the list; 3) repeat step 2
until every EM is matched. After matching the propensity score distributions of EM and CM
are virtually identical (see Figure 4). Matching each EM with one CM allows us to avoid
bias in the estimated treatment effect that may occur when linking multiple, potentially
dissimilar, CMs to an EM (Smith 1997). Moreover, by treating each EM-CM pair as a
separate experiment, we are able to investigate variability in entry effects.
9
Before matching After matching
with entry without entry t-statistic with entry without entry t-statistic
Median age 35.502 37.693 -11.608∗∗ 35.783 35.855 -0.251
log(Population density) 5.887 4.893 18.289∗∗ 5.935 5.922 0.150
log(Income per capita) 2.943 2.893 4.824∗∗ 2.936 2.935 0.044
No. of other supercenters 7.721 5.577 5.600∗∗ 8.378 8.394 -0.028
Herfindahl index 0.119 0.466 -78.605∗∗ 0.119 0.123 -1.008
N 759 21427 629 629
∗∗
p-value < .01, ∗
p-value < .05
Table 5: Variable means before and after matching
The sample means reported in Table 5 show that experimental and control groups were
successfully matched (Rubin 2006). By using PSM, “the observational study equivalent of
randomization in an experiment” (Rubin 2006, p. 461), our data were transformed into a
quasi-experimental design.
4 Model
We use a hierarchical Bayesian model to estimate Wal-Mart entry effects on supplier per-
formance. Our model provides a differences-in-differences estimator, as extensively used in
economics (Angrist and Krueger 1999) and marketing (e.g., Ailawadi et al. 2009, Tripathi
2009). Our dependent variables are supplier profits, wholesale prices, and shipments in week
t. For each pair of matched experimental and control markets we have
P ROF ITiet = α0i + α1i EMie + α2i W Miet + α3i EMie × W Miet + "1iet , (1)
W Piet = β0i + β1i EMie + β2i W Miet + β3i EMie × W Miet + "2iet , (2)
SHIPiet = γ0i + γ1i EMie + γ2i W Miet + γ3i EMie × W Miet + "3iet , (3)
market; t indexes time t = 1, 2, ..., T ; P ROF ITiet is supplier profit; W Piet is the wholesale
price charged; SHIPiet is shipment volume; EMie is an experimental market dummy (1 for
10
a market with entry, 0 otherwise); and W Miet is the Wal-Mart entry dummy (1 if week t
is after entry, 0 otherwise). Note that the times series data for the matched experimental
and control markets are stacked for estimation. We assume ("1iet , "2iet , "3iet )" ∼ N(0, Σi ) for
i = 1, 2, ..., N. The coefficients α3i , β3i , γ3i capture the treatment effect (i.e., Wal-Mart entry)
θi = ∆" Zi + ξi , (4)
where θi" is the vector of parameters in equations (1-3); ∆ = [δ1 , δ2 , ..., δnz ], ξi ∼ N(0, Vθ ); and
Zi is an nz × 1 vector of covariates used to capture heterogeneity in entry effects. We explore
whether cross-sectional impact differences can be linked to variation in the products offered
to, and carried by, incumbent retailers and Wal-Mart. We also include the covariates from
the propensity score model in the Z matrix as controls. Additional details on the estimation
algorithm are provided in Appendix A.
5 Results
We estimate equations (1-4) twice: First with data from all retailers in a market combined,
including Wal-mart (total market), and again with data from incumbents only (incumbents).
By isolating the impact of entry on, for example, supplier profits generated from incumbents,
we are able to investigate a potential source of changes in the outcome metrics for markets
as a whole. For example, profits from a market might increase after entry due to a direct
boost from Wal-Mart at the expense of profits from incumbents. Providing results for both
total markets and incumbents separately provides additional insight into processes affecting
supplier performance following entry.
11
5.1 Wal-Mart entry impact
The mean percentage change in manufacturer total market profits is +17.77% (see Table 7)
across all markets with Wal-Mart entry; profits from incumbents drop only slightly (-1.34%).3
The supplier studied clearly benefits from the collaboration with Wal-Mart. Compared to the
size of Wal-Mart entry effects on manufacturer profits the magnitude of impact on wholesale
prices is much smaller. On average, wholesale prices increased only 0.55% and 0.32% for
total market and incumbents, respectively. Interestingly, shipments to incumbents drop
only 2.52% following entry. When we include Wal-Mart, however, total market shipments
increase by 14.95%, on average, which is similar in magnitude to the post-entry profit change
mentioned above. As the estimated Wal-Mart entry effects differ substantially from those
reported in Table 2, controlling for selection bias is clearly important in our application.
The magnitudes of entry impact on supplier profits and shipments are intriguing. To
ensure these effects are not an artifact of the estimation procedure used we conducted
several robustness checks (see Table 7). A fixed-effects differences-in-differences estimator
was employed to calculate the main effects at the total market level (Angrist and Krueger
1999). The estimates for profits, wholesale prices, and shipments were 19.01%, 0.32%, and
12.59% respectively; very similar to the results based on matching. We also estimated two
differences-in-differences models with alternative matching procedures. First, we used the
3
For confidentiality reasons we do not report the parameter estimates from Equations 1-3 directly. Rather,
we transform them to a percentage change after entry relative to the average weekly profits, wholesale prices,
and shipments before Wal-Mart entry.
12
same covariates in the logit model as reported in Table 4 but estimated the model five times,
once for each year in our dataset. Markets were matched based on the parameters of the
logit model for the year in which entry occurred. If Wal-Mart’s entry strategy changes over
time this matching procedure should produce results different from those reported in Table
5.1. The estimates for profits, wholesale prices, and shipments derived using the ’matching
by year’ procedure were 17.52%, 0.33%, and 13.91% respectively; again, very similar to our
earlier results. Finally, we estimated a matching model with a broader set of covariates
including quadratic terms for all variables but excluding state-fixed effects, regardless of
statistical significance. Note that we did not use trimming as part of this matching proce-
dure. The estimates for profits, wholesale prices, and shipments from the ’matching without
trimming’ model were 16.95%, -0.25%, and 15.53% respectively; again, very similar to the
results reported in Table 5.1.
Figures 5, 6, and 7 contain histograms of the θi estimates derived from equations 1-4. All
three graphs show that entry impact estimates vary significantly; the vertical black line in
each figure is drawn at the median value. We investigate plausible causes of this variability
in Section 5.3 below.
13
Table 8 shows the percentage of positive, non-significant, and negative estimates of entry
impact on supplier profits from total markets (i.e., incumbents plus Wal-Mart) and incum-
bents respectively. In nearly 56% of the markets studied post-entry market profits increased
significantly; they decreased in only 9% of markets. Supplier profits from incumbent retail-
ers are down in nearly a third of post-entry markets, whereas in 20% of cases incumbent
contributions increased. Interestingly, over two-thirds of markets generate as much, if not
more, profits for the supplier after entry even when excluding contributions from Wal-Mart.
% (non)significant effects
+ ns -
Total market 55.92% 34.99% 9.09%
Incumbents 19.90% 47.17% 32.93%
For significant estimates zero is not contained in the 95% credibility region.
The three scatter plots in Figure 8depict the relationship between supplier profits from
Wal-Mart, incumbents, and the total market following entry. The first panel shows the
correlation between profits from incumbents and Wal-Mart is limited (r = 0.065), demon-
strating that the benefits derived from Wal-Mart’s market presence need not come at the
expense of profits the manufacturer generates from other retailers. Interestingly, the correla-
tion between profits from Wal-Mart and the total market, shown in the second panel, is not
especially strong either (r = 0.361). As Wal-Mart accounts, on average, for 19.19% of total
post-entry market profits, this result is surprising. In contrast, the correlation between prof-
its from incumbents and total market, shown in the bottom panel, is very high (r = .954),
which suggests maintaining profit levels generated from incumbents is key to the supplier’s
post-entry performance.
We find that Wal-Mart entry can affect wholesale prices charged to incumbent retailers,
even though the size of the effect is, on average, small (Table 7). Table 9 shows the percentage
14
of markets with a significant increase, no significant change, and a significant decrease in
wholesale prices. The distribution of wholesale price changes is clearly very balanced: we
observe each effect in approximately one-third of markets. Even though previous research
has suggested that Wal-Mart entry may depress retail prices (Basker 2005, Ailawadi et al.
% (non)significant effects
+ ns -
Total market 37.39% 32.08% 30.53%
Incumbents 36.02% 31.73% 32.25%
For significant estimates zero is not contained in the 95% credibility region.
nearly 70% of markets incumbents generate as much, if not more, supplier shipment volume
post-entry.
% (non)significant effects
+ ns -
Total market 44.77% 44.60% 10.63%
Incumbents 18.35% 49.23% 32.42%
For significant estimates zero is not contained in the 95% credibility region.
The three scatter plots in Figure 9depict the relationship between supplier shipments to
Wal-Mart, incumbents, and the total market following entry. The first panel demonstrates
for 18.85% of total market shipments after entry, this effect is surprising. In contrast, the
15
correlation between shipments to incumbents and total market, shown in the bottom panel,
is very high (r = .963), demonstrating the importance of maintaining post-entry shipment
levels to incumbents.
As mentioned above, Dukes et al. (2006) argue that manufacturers can boost profits when
faced with a dominant retailer by charging higher wholesale prices to incumbents, whereas
Chen (2003) theorizes they should lower them. We use the correlation between the total
market estimate for α3 and the β3 estimate for incumbents to evaluate their contradictory
hypotheses (see equations 1 and 2). A strong positive correlation between the parameters
would provide support for Dukes et al. (2006)’s hypothesis, whereas a strong negative correla-
tion would affirm Chen (2003)’s theory. Figure 10 shows a scatter plot of Wal-Mart’s impact
on both wholesale prices charged to incumbents and manufacturer profits. The loosely scat-
tered points clearly indicate that the correlation between manufacturer profit changes and
wholesale price changes is negligible (r = −0.024). Interestingly, the results from our em-
pirical analysis support neither hypothesis. Although the results in Table 9 show wholesale
prices can change, Figure 10 clearly demonstrates they are not the key driver of manufacturer
profit changes following Wal-Mart entry.
The fact that wholesale prices are, on average, only marginally affected by Wal-Mart
entry implies that shipments must be the key driver of the manufacturer profit changes
reported in Table 7. The scatter plot in the left-hand panel of Figure 11 depicts the link
between changes in profits and total market shipments following Wal-Mart entry; the one
on the right shows the correlation between profit and shipments to incumbent retailers. In
16
contrast to Figure 10 both plots show a positive relationship, confirming that shipments are
indeed the prominent driver of supplier profit change. The slope in the right-hand panel
(r = 0.884) clearly demonstrates that incumbents are important to the supplier’s overall
profitability in post-entry periods. While selling to Wal-Mart generates financial benefits,
the manufacturer obviously fairs best when incumbent shipments either increase or remain
unchanged following entry.
Results reported in Section 5.1 show considerable variation in Wal-Mart entry effects on
supplier performance. Manufacturers could learn how to influence outcomes in their favor by
understanding the sources of variation. As mentioned in our introduction, Dukes et al. (2009)
theorize incumbents retailers should carry a broader assortment in the presence of a dominant
retailer that chooses to offer a limited product selection per category. Their results imply that
both the incumbents’ assortment size and the overlap with Wal-Mart impacts performance.
Recent work by Ailawadi et al. (2009) confirms that incumbents can mitigate Wal-Mart entry
effects by increasing assortment size.4 Retailers who’s assortments overlap substantially with
Wal-Mart’s are vulnerable according to Gielens et al. (2008). Moreover, Dukes et al. (2009)
also suggest that a supplier may choose to sell its specialty items only to incumbents in the
presence of a dominant retailer. Even though a dominant retailer will only stock the most
popular items to reduce assortment costs, incumbents will benefit from carrying a larger
assortment despite the additional cost. In sum, previous research implies that, because Wal-
Mart carries a limited selection of goods in most categories (e.g.,Stone 1988, O’Keefe 2002),
incumbents should offer a relatively large assortment focused on products Wal-Mart does
4
Since the authors do not have information on products carried by Wal-Mart, they cannot address the
implications of assortment overlap.
17
not sell. Therefore, we expect that the impact of Wal-Mart entry on supplier profits will be
moderated by incumbents’ assortment choices.
Figure 12 depicts the distribution of assortment overlap between Wal-Mart and incum-
bents in experimental markets pre- and post-entry. Surprisingly, it seems that incumbent
retailers’ product assortment becomes more like Wal-Mart’s after entry. Since this effect
could, at least partly, be explained by changes in the supplier’s product line we express the
assortment characteristics for experimental markets relative to their matched control mar-
ket. In fact, our focal supplier’s product line contracted by 35% as depicted in Figure 13.5
We define change in assortment overlap as (ope1 − ope0 ) − (opc1 − opc0 ) where e (c) identifies
an experimental (control) market. op1 is the percent overlap in post-entry assortment be-
tween incumbents and Wal-Mart, whereas op0 represents the assortment similarity between
incumbents before entry and Wal-Mart upon entry. Although Dukes et al. (2009) suggest
incumbents should diversify, the median assortment overlap change is +3.3%, even after
controlling for variation in the manufacturer’s product line. We define change in assortment
ne1 −ne0 nc1 −nc0
size as ( ne0
) −( nc0
) where e (c) is defined as before and n1 (n0 ) captures the num-
ber of products in the incumbents assortment after (before) Wal-Mart entry. The median
assortment size change is -0.02%.
Equation 4 describes the second stage model used to correlate entry effects on supplier
performance to changes in incumbents’ assortment. Table 11 contains the ∆ estimates for
the total market and incumbent level analyses.6 The impact of Wal-Mart entry on supplier
5
Although the percentage of total shipments accounted for by Wal-Mart increases over time, it never exceeds
5% in the time span of our data. Hence, we assume the reduction in assortment size is not related to
Wal-Mart.
6
To simplify exposition we do not report parameter estimates for the control variables included in the Z
matrix.
18
profits from the total market and incumbents are both positively correlated with incumbent
assortment size changes. Consistent with predictions by Dukes et al. (2009) and extending
results reported in Ailawadi et al. (2009), we demonstrate that carrying a wider assortment
not only benefits incumbents but also boosts suppliers’ overall profits.
*
zero is not contained in the 99% credibility region.
Parameters were converted to percentages for reasons of confidentiality.
Confirming Dukes et al. (2009) Table 11 shows the manufacturer benefits most, at both
the market and incumbent level, when incumbent retailers assortment overlap with Wal-
Mart is limited. Holding other variables constant, a 1% decrease in assortment overlap
increases weekly post-entry supplier profits from incumbents by 0.517% and shipments to
size are statistically significant, effect sizes are small. A 1% increase in assortment size
increases weekly post-entry supplier market profits by 0.039% and shipments by 0.041%.
The impact on profits from and shipments to incumbents are similar in magnitude (0.037%
and 0.038% respectively).
19
6 Conclusion
In this paper, we study a broad set of geographical markets to determine whether Wal-Mart
entry impacts supplier profits and, if so, what processes drive post-entry profitability change.
Our unique database, collected by a Wal-Mart vendor, spans thousands of retail stores and
hundreds of products for a period of five years. We employed propensity score matching to
control for potential selection bias before analyzing the drivers of profit change: wholesale
prices and shipments. A hierarchical Bayesian model was used to quantify the effects of
entry and link the results to differences in retailer assortment characteristics across markets.
We find that mean post-entry supplier profit increased by 17.77% for the total market,
whereas profits derived from incumbents decreased only marginally. While the size of this
effect is surprising, various analyses demonstrate its robustness. Interestingly, over two-
thirds of markets generate as much, if not more, profits for the supplier after entry even
when excluding contributions from Wal-Mart. Contrary to both Chen (2003) and Dukes
et al. (2006) our results clearly show that changes in wholesale prices are not the main driver
of post-entry supplier profit changes; market expansion is. We observe a significant increase
in shipments to 45% of markets studied. Surprisingly, as post-entry incumbent shipments
drop less than 3%, on average, cannibalization by Wal-Mart is limited. Furthermore, our
analysis demonstrates that both supplier shipments and profits increases are highest for
markets in which incumbents offer a wider array of products and carry items that Wal-Mart
selling to Wal-Mart can directly boost suppliers’ bottom line. Even though news stories
threaded with manufacturer complaints about Wal-Mart resonate with critics and general
public alike, we argue that some complaints about selling to Wal-Mart may be overstated as
the supplier studied is clearly better off post-entry.
20
Second, we show that while suppliers benefit from the additional volume Wal-Mart gen-
erates they perform best when post-entry shipments to incumbents increase or remain un-
changed. In addition, the strong link between assortment characteristics and shipments to
incumbents suggests suppliers should encourage them to carry larger and non-overlapping
assortments, which could have important implications for retail competition. Wal-Mart is
known for its aggressive competitive strategy, as David Glass (CEO Wal-Mart Stores Inc.
1988-2000) explained: “We want everybody to be selling the same stuff, and we want to
compete on a price basis, and they will go broke 5 percent before we will.”(Fishman 2006,
p. 48, 68). Trying to beat Wal-Mart at the pricing game is infeasible (Bergdahl 2004)
but, as one grocer put it, “They can’t beat our price on items they don’t have.” (O’Keefe
2002). Retailers selling the same goods as Wal-Mart not only put themselves in jeopardy
(e.g., Stone 1995, Gielens et al. 2008), our results demonstrate they also bring down supplier
shipments and profits. Increased retail differentiation will not only boost supplier profits
from incumbents but, surprisingly, also from Wal-Mart. Therefore, we suggest that sup-
pliers motivate incumbents to diversify, for example, by offering them specialty products,
slotting allowances, services, or training that Wal-Mart does not need or want.
Third, to enhance incumbents assortment options manufacturers should maintain prod-
uct lines. Wal-Mart has become the primary customer for many suppliers, who devote a
large chunk of their marketing resources to serve the retailer (Useem 2003, Fishman 2006).
Some manufacturers even prune their product lines to better target Wal-Mart’s customers
(Fishman 2006), offering incumbents little room to diversify. They are setting up a “lose-
lose” situation, as retailers cannot but carry the same products Wal-Mart does. Just as
Wal-Mart continually improves itself by studying its customers buying habits (Huey 1998),
manufacturers should not only develop and market products for Wal-Mart, but also learn
from and cater to incumbent retailers’ consumers.
Future research could extend our findings in several ways. For example, while we focus on
one major manufacturer in large number of geographical areas in the U.S., researchers could
21
expand our findings to additional industries and countries. Moreover, whereas our study
quantifies Wal-Mart entry impact on supplier profits, wholesale prices, and shipments, its
implications on incumbent retailer profits have yet to be addressed. Since our results clearly
indicate limited cannibalization from incumbents, it would be interesting to investigate the
underlying causes of the Wal-Mart entry shipment boost. Is it an income effect? Is the
increase driven by a disproportionate change in the demand for products in a specific quality
tier? Are incumbent retailers lowering prices for certain products? Furthermore, retailers,
suppliers, and consumers may benefit if future studies could determine how retailers can best
differentiate their assortments from Wal-Mart.
Wal-Mart’s influence on today’s global economy is unlike that of any other retailer. Our
study provides several new and important insights into Wal-Mart’s impact on its suppliers.
We hope that, with manufacturer and retailer cooperation, researchers will be able to paint
a comprehensive picture of the costs and benefits for all constituents when Wal-Mart comes
to town.
22
A Estimation algorithm
A system of regression equations (see equations 1-3) is estimated for each pair of matched
markets, where the errors are assumed to follow a normal distribution with mean 0 and
covariance matrix Σi . In the second stage the θi coefficients from the first stage are linked
to a set of cross-sectional characteristics Zi . Natural conjugate priors are specified for the
¯ Vθ ⊗ A−1 ) where %̄ = 0
hyper prior parameters: Vθ ∼ IW (w0 , W0 ), vec(∆)|Vθ ∼ N(vec(∆),
regression. The sampler is run for a total of 20,000 iterations to simulate the posterior
distributions of the parameters. Every 5th draw of the last 10,000 is selected, leaving 2,000
draws for inference. Since we use a standard linear hierarchical Bayesian model, we refer to
Rossi et al. (2005) for further details.
23
References
Kusum L. Ailawadi, Norm Borin, and Paul W. Farris. Market Power and Performance:
Kusum L. Ailawadi, Jie Zhang, Aradhna Krishna, and Michael Kruger. When Wal-Mart
Enters: How Incumbent Retailers React and How this Affects Their Sales Outcomes.
Emek Basker. Selling a Cheaper Mousetrap: Wal-Mart’s Effect on Retail Prices. Journal of
Urban Economics, 58(2):203–229, 2005.
Michael Bergdahl. What I learned from Sam Walton. How to Compete and Thrive in a
Wal-Mart World. John Wiley & Sons, Inc., Hoboken, New Jersey, 2004.
Dan Bianchi and Dan Swinney. Wal-Mart: A destructive force for Chicago communities and
companies, March 25 2004. Center for Labor and Community Research, Chicago, IL.
Paul N. Bloom and Vanessa G. Perry. Retailer Power and Supplier Welfare: The Case of
Robert J Bowman. Should You just Say No to Wal-Mart? Distribution, 96(12):52, 1997.
S. D. Brunn. Wal-mart world: the world’s biggest corporation in the global economy. CRC
Press, 2006.
24
Meghan Busse, Jorge Silva-Risso, and Florian Zettlemeyer. $1000 cash back: The pass-
through of auto manufacturer promotions. The American Economic Review, 96(4):1253–
1270, 2006.
Zhiqi Chen. Dominant Retailers and the Countervailing-power Hypothesis. RAND Journal
of Economics, 34(4):612–625, 2003.
Anthony J. Dukes, Esther Gal-or, and Kannan Srinivasan. Channel Bargaining with Retailer
Anthony J. Dukes, Tansev Geylani, and Kannan Srinivasan. Strategic Assortment Reduction
by a Dominant Retailer. Marketing Science, 28(2):309–319, 2009.
Charles Fishman. The Wal-Mart Effect. How the World’s Most Powerful Company Really
Works - and How It’s Transforming the American Economy. the Penguin Group, New
York, NY, 2006.
Sonja Gensler, Peter Leeflang, and Bernd Skiera. Influence of Online Use on Product Usage
Behavior and Consequences for Customer Channel Migration Strategies. Working paper,
January 2009.
Paul J. Gertler and Tim S. Simcoe. Disease Management. SSRN eLibrary, 2006.
Katrijn Gielens, Linda M. Van de Gucht, Jan-Benedict E.M. Steenkamp, and Marnik G.
Dekimpe. Dancing with a Giant: The Effect of Wal-Mart’s Entry into the United Kingdom
on the Stock Prices of European Retailers. Journal of Marketing Research, XLV:519–534,
2008.
Stephan J. Goetz and Anil Rupasingha. Wal-Mart and Social Capital. American Journal of
Agricultural Economics, 88(5):1304–1310, December 2006.
Marty Gosman and Mark J. Kohlbeck. The effects of wal-mart on the profitability and
investor valuation of its suppliers. SSRN eLibrary, 2006.
25
O.Z. Graff, Thomas and Dub Ashton. Spatial Diffusion of Wal-Mart: Contagious and Reverse
Hierarchical Elements. The Professional Geographer, 46(1):19–29, 1994.
Stephen Halebsky. Small Towns and Big Business. Challenging Wal-Mart Superstores. Row-
man & Littlefield publishers, Inc., Lanham, MD, 2009.
Jerry Hausman and Ephraim Leibtag. Consumer Benefits from Increased Competition in
Shopping Outlets: Measuring the Effect of Wal-Mart. Working paper, October 2005.
Panle Jia. What Happens When Wal-Mart Comes to Town: An Empirical Analysis of the
Stephen Kinzer. Wal-Mart’s big-city plans stall again. New York Times, page A.27, May 6,
late edition 2004.
Ryan J. Luchs. Three Essays Addressing Issues in Retail Channels. PhD thesis, the Katz
J.E. McGee and T.A. Festervand. When Wal-Mart comes to town: A look at David’s
response to the arrival of retailing’s Goliath. Journal of Business Strategy, 7(3):15–27,
1998.
Bethany E. Moreton. The Would of the Service Economy: Wal-Mart and the Making of
Christian Free Enterprise, 1929-1994. PhD thesis, Yale University, 2006.
Sandra Mottner and Steve Smith. Wal-Mart: Supplier performance and market power.
Journal of Business Research, 62(5):535–541, May 2009.
Junfu Zhang Neumark, David and Stephen Ciccarella. The effects of Wal-Mart on local
26
Larry Parsons. Vote makes way for Wal-Mart: Marina planning commission’s action can’t
be appealed to city council. The Monterey County Herald, page 1, April 28 2006.
Yi Qian. Do national patent laws stimulate domestic innovation in a global patenting envi-
ronment? a cross-country analysis of pharmaceutical patent protection, 1978 –2002. The
Review of Economics and Statistics, 89(1):436–453, 2007.
Paul R. Rosenbaum and Donald B. Rubin. The Central Role of the Propensity Score in
Paul R. Rosenbaum and Donald B. Rubin. Reducing bias in observational studies using
subclassification on the propensity score. Journal of the American Statistical Association,
79(387):516–524, 1984.
Paul R. Rosenbaum and Donald B. Rubin. Constructing a Control Group Using Multivari-
ate Matched Sampling Methods That Incorporate the Propensity Score. The American
Peter E. Rossi, Greg M. Allenby, and Robert E. McCulloch. Bayesian Statistics and Mar-
keting. John Wiley and Sons, West Sussex, England, 2005.
Donald B. Rubin. Matched Sampling for Causal Effects. Cambridge University Press, New
York, 2006.
Ameet Sachdev. Wal-Mart Cuts Data Supply: Researchers Say Scanner Numbers Hard to
Replace. Chicago Tribune, May 15 2001.
Vishal P. Singh, Karsten T. Hansen, and Robert C. Blattberg. Market Entry and Consumer
Herbert Smith. Matching with multiple controls to estimate treatment effects in observational
studies. Sociological Methodology, 27(1):325–353, 1997.
27
Kenneth E. Stone. The effect of Wal-Mart Stores on businesses in host towns and surrounding
towns in Iowa. Department of Economics, Iowa State University, 1988.
Manish Tripathi. Spillovers from Offline Entry: An Analysis of Buyer and Seller Behavior
on eBay. Working paper, June 2009.
Jerry Useem. One Nation under Wal-Mart. Fortune: New York, 147(4):65, 2003.
Richard Vedder and Wendell Cox. The Wal-Mart Revolution. How Big-box Stores Benefit
Consumers, Workers, and the Economy. American Enterprise Institute for Public Policy
Research, 2006.
Wal-Mart Stores Inc. 2009. Wal-mart reports financial results for fiscal year and fourth
quarter.
Ting Zhu and Vishal Singh. Spatial Competition with Endogenous Location Choices - An
Application to Discount Retailing. Quantitative Marketing and Economics, 7(1):1–35,
2009.
28
Figure 1: Location of Wal-Mart entries
29
80
60
Density
Control Market
Experimental Market
40
20
0
Pr[Entry=1|X]
30
20
15
Control Market
Experimental Market
Density
10
5
0
Pr[Entry=1|X]
Figure 3: Propensity score distribution for experimental and control markets on common
support of P (X)
31
Control Market Experimental Market
4
4
3
3
Density
Density
2
2
1
1
0
0.0 0.1 0.2 0.3 0.4 0.5 0.0 0.1 0.2 0.3 0.4 0.5
Pr[Entry=1|X] Pr[Entry=1|X]
32
Supplier Profits (Total Market) Supplier Profits (Incumbents)
140
150
120
100
100
80
Frequency
Frequency
60
50
40
20
0
− 0+ − 0+
33
Wholesale Prices (Total Market) Wholesale Prices (Incumbents)
80
80
60
60
Frequency
Frequency
40
40
20
20
0
− 0+ − 0 +
34
Shipments (Total Market) Shipments (Incumbents)
120
120
100
100
80
80
Frequency
Frequency
60
60
40
40
20
20
0
− 0+ − 0+
35
Profits from incumbents
0+
−
0 +
0+
−
0 +
0
−
− 0 +
36
Shipments to incumbents
0+
−
0 +
Shipments to Wal−Mart
Shipments to total market
0+
−
0 +
Shipments to Wal−Mart
Shipments to total market
0+
−
− 0 +
Shipments to incumbents
37
Wholesale Prices
0+
−
− 0 +
Supplier Profits
Figure 10: Wal-Mart entry impact on supplier profits versus Wal-Mart entry impact on
wholesale prices charged to incumbents
38
Shipments to Total Market
Shipments to Incumbents
0+
0+
−
− 0+ − 0+
Figure 11: Wal-Mart entry impact on supplier profits versus Wal-Mart entry impact on
shipments to total market (left panel) and incumbents only (right panel)
39
3.0
2.5
2.0
Density
pre−entry
1.5
post−entry
1.0
0.5
0.0
Assortment Overlap
40
110
100
90
Quarterly Assortment Index
80
70
60
50
41