new the “wal-mart effect” in central and eastern europe · 2011. 9. 4. · retailers’ entry...
TRANSCRIPT
The “Wal-Mart Effect” in Central and Eastern Europe*
Deniz Igan†
International Monetary Fund‡
Junichi Suzuki University of Toronto
First draft: February 2007 This draft: August 2011
Abstract
Prior to the recent global financial crisis altering the price dynamics, inflation in central and eastern European countries was characterized by a sharp drop pattern. Several factors contributed to this decline such as benign global economic conditions and growing credibility of domestic macroeconomic policies, yet factors related to industrial organization and market competition have been somewhat overlooked. Using measures of retail competition intensity reflecting the prevalence of high-productivity, modern-format retailers, we find that the decline in retail goods price inflation, and particularly food price inflation, was associated with increased retail competition. This may be linked to improved living conditions as households allocate more to health and education. JEL Classification Numbers: E31, L16 Keywords: Wal-Mart effect, retail competition, inflation
* This research was undertaken while Suzuki was a graduate student at the University of Minnesota and a summer intern at the European Department of the IMF. We would like to thank Juan José Fernández-Ansola, Marcelo Pinheiro, Natalia Tamirisa, Subhash Thakur, participants at the IMF Brown Bag seminar, the editor, and two anonymous referees for useful comments and suggestions. All remaining errors are our own.
† Corresponding author. Mailing address: IMF, Research Department, 700 19th Street NW, Washington, DC 20431. Phone: +1 (202) 623-4743. Fax: +1 (202) 589-4743. Email: [email protected].
‡ The views expressed in this paper are those of the authors and do not necessarily represent those of the IMF or IMF policy.
I. INTRODUCTION
Central and eastern European countries experienced considerable decline in their inflation
rates in the period starting with their transition from socialism to capitalism. The pattern
was particularly striking among the countries in the region that became some of the new
member states of the European Union (EU).1 Starting from relatively high levels in 1997,
the inflation rates in these countries actually stood almost at the same level as the EU-15
average at the end of 2005 (Figure 1). Hungary, Poland, the Czech Republic and the
Slovak Republic led a group of European countries in terms of decline in inflation
relative to the initial level between 1997 and 2005 (Figure 2).
Several factors contributed to the decline in inflation. First, the external
conditions were, in general, benign. Strong capital inflows at a time of high liquidity in
international financial markets were reflected in the appreciating exchange rate.
Moreover, increased trade relations with low-inflation western European countries as part
of the greater integration with the region moderated imported inflation. Increase in trade
openness also dampened prices through substitution of domestically produced goods with
cheaper foreign alternatives. Second, on the domestic front, somewhat suppressed
demand, and a series of positive supply shocks (low input prices, exchange rate
appreciation, and large productivity gains) relaxed the pressure on prices. Last but not
least, decreasing inflation expectations as a result of growing credibility of the monetary
policy frameworks also helped decrease inflation.
1 The new member states of the European Union consist of those that joined in 2004, namely, Cyprus, the Czech Republic, Estonia, Hungary, Latvia, Lithuania, Malta, Poland, the Slovak Republic, and Slovenia.
A less studied factor, related to the increasing trade openness and integration
of consumption and production networks (globalization), is the entry of international
retailers.2 Changing demographical, economic, and institutional conditions trigger entry
to a particular market, leading to intensified competition and, in turn, lower prices and
consequently lower inflation. A potential difference in the magnitude of this effect across
countries may emerge since, as retail markets in developed countries become saturated,
retail market chains in these countries start looking for new targets. Naturally, central and
eastern European countries entered on the radar screen of large international retail market
chains in the 1990s (Figure 3). Over the next decade, many major retailers had entered
these nascent markets.
Existing studies point to a significant effect on prices associated with the entry
of low-cost, high-efficiency modern retailers. In the case of Wal-Mart, the largest and
most extensively studied global retailer, researchers find that entry of Wal-Mart into a
market contributes to lower consumer prices both in the short and the long run (Basker,
2005a, b; Global Insight, 2005). The term “Wal-Mart effect” was coined to refer to the
downward pressure on prices documented in these studies. We borrow this term in
referring to the effect on inflation of the changing retail industry landscape in Europe.3
To the best of our knowledge, despite the desirable empirical setting in central and
eastern European countries to study this phenomenon due to the rapid and significant
transition, this is the first formal analysis of the Wal-Mart effect in the region.
2 Allard (2006) looks at the role of globalization in bringing down inflation in the Polish case.
This paper empirically examines the contribution of retail competition to
inflation in Europe between 2000 and 2004, by comparing central and eastern European
countries, given their distinctive experience in retail trade developments over the past
couple of decades, to western European countries. We find that a 10 percent increase in
retail competition reduces the inflation rate for goods that are traded on retail by around
0.5 percentage points. The decline in food inflation associated with retail competition is
estimated to be close to 0.6 percentage points, or roughly 30 percent of the average
observed inflation rate among the countries in the sample in 2005. These results are
robust to alternative measures of retail competition intensity and controlling for other
factors that had a downward effect on inflation. Taking into account the possibility that
retailers’ entry choice may depend on a country’s socio-economic conditions including
inflation dynamics by estimating a simultaneous equation system does not alter the
findings.
These results may have real implications on living conditions and welfare.
Food constitutes an important category in household expenditures and income effects
from lower food prices may be quite large. The econometric analysis suggests that the
portion of expenditures allocated to food in the consumption basket of all households
could have been reduced by 3 percent during our sample period.4 The households in four
central and eastern European countries in our sample used to allocate almost a third of
3 Wal-Mart is not necessarily among the retailers active in these markets but the effect we study relies on similar economic foundations.
4 This is calculated as the cumulative difference from 2000 to 2004 between the cost of the food consumption basket under the actual foods inflation rate and that under the 0.6 percentage points lower inflation rate due to higher retail competition intensity.
their overall expenditure to food in 2000. By 2008, this was reduced to roughly a fifth,
although the physical amounts consumed did not change much, leaving more of the
disposable income to be spent on housing, health, education, and leisure.
Quantifying the impact of retail competition on inflation may also be useful in
order to effectively implement inflation targeting policy. For instance, in the few couple
of years prior to the global financial crisis, the Czech National Bank undershoot its
inflation target. Examination of changes in the retail industry and sectoral decomposition
of inflation suggest that a potentially important component related to industrial
organization and market competition might be missing when only macroeconomic factors
are considered. Thus, understanding the quantitative impact of retail competition on
inflation could improve monetary policy management. For an inflation-targeting country
like the Czech Republic, the challenge is to determine the nominal interest rate necessary
to reach a particular target in the presence of downward pressure on consumer prices due
to intense retail competition. An appropriate choice of nominal anchors may need to take
this pressure into account.
The rest of the paper is organized as follows. Section II introduces the
measures for the degree of retail competition and documents the transformation of central
and eastern Europe in terms of intensity of competition. Section III studies the empirical
relationship between increases in retail goods prices and the degree of retail
modernization. Section IV discusses several issues related to the robustness and
interpretation of the findings. Section V concludes.
II. INTENSITY OF RETAIL COMPETITION AND PRICE DYNAMICS: BACKGROUND
A. Transformation of the Retail Landscape in Central and Eastern Europe
Modern stores with high measured output per worker dominate the retail industry in
richer countries while less-productive traditional stores are the norm in developing
countries (Lagakos, 2009). But, with considerable potential for growth in per capita
income, unsaturated markets, relatively low level of political and macroeconomic risk,
and an ambitious agenda for reforms to improve business environment, central and
eastern European countries offered one of the best opportunities for any retailer that
would consider international expansion in the 1990s. As a result, major retailers started
entering these markets as early as 1991. The number of retail operations in eastern
Europe owned by Austrian, Belgian, British, Dutch, French, German, and Italian firms
increased from 28 in 1991 to approximately 1800 in 2002 (Alexander and Myers, 1997;
Dries et al., 2004). Meanwhile, the shopping habits of households shifted in favor of
larger retail formats (INCOMA Research, 2006). Such developments are observed
especially, in the Czech Republic, Poland, Hungary, and Slovakia, and more recently in
Bulgaria, Croatia, Romania, Russia, and Ukraine.5 The entry of international retail market
chains in emerging countries and the increased competition that comes with it could have
significant impact on price dynamics, similar to what was observed in developed
countries.
In documentation of the impact of changes in retail industry on price
dynamics, the decline in inflation is mostly attributable to the decrease in the prices of
5 See Dries et al. (2004) for a detailed description of the retail sector developments in central and eastern Europe.
goods traditionally sold by retailers. Food, clothing, and furnishings were the driving
force for the decline in inflation, especially in the Czech Republic and Poland (Figure 4).
Negative or small positive contributions to inflation from food prices were recorded first
in the Czech Republic and Poland, and more recently in the Slovak Republic and
Slovenia. It is interesting to note that this timing coincides almost perfectly with the
ranking of these countries according to their attractiveness for international retailers in
1995 (shown in Figure 3). Negative contributions from clothing were also observed in
EU-15, yet, they were stronger in the Czech Republic and Poland.6 Across central
European countries, increases in retail goods prices (food, beverages, and clothing) are
noticeably lower than increases in non-retail goods prices (transport and non-tradables,
such as housing, health, and services) (Figure 5).
B. Why Does Retail Competition Reduce Pressure on Consumer Prices?
There are two main channels through which intensified retail competition can lead to
decreased consumer prices. The first of these channels works in a static way by lowering
mark-up rates. With more firms entering into the market, each firm finds it more difficult
to differentiate its characteristic (e.g., quality, variety, or location) from its rivals.
Consequently, firms start bidding harder against each other, ending up with a lower
market price (Bresnahan and Reiss, 1991). This bidding pressure, if continuous, may also
put a limit on how fast consumer prices increase over time. The second channel works in
a more dynamic way. As intense competition forces less productive firms to exit the
6 Clothing, or textiles in general, is probably the category with most ease to transport, and, hence, to trade internationally. Fierce international competition in textiles, induced by east Asian producers, suggests that the impact of globalization is likely to be exceptionally strong in this sector. This could explain the observation of negative contributions in EU-15. Food, on the other hand, is more “local” because of its perishable nature.
market and only productive firms consider entry, the industry as a whole becomes more
productive (Hopenhayn, 1992). Indeed, Javorcik, Keller, and Tybout (2008) and Javorcik
and Li (2008) provide direct evidence of the productivity and efficiency gains due to the
entrance of large retailers on the manufacturing sectors in less-developed countries. Since
productive firms have lower cost structures and better ability to accommodate supply
shocks, they help increase the economy’s resistance to transitory shocks by avoiding
price hikes that may turn out to be temporary. Therefore, inflation becomes more stable.
To the extent that higher productivity at the retail sector absorbs or limits the increases in
the prices set by the suppliers, consumer price inflation would also be lower.7
In central and eastern European countries, both channels are likely to be in
effect. Several analysts reported the transition of the retail industry in the region from a
centralized state-owned system into a modern, highly competitive marketplace. For
instance, in the Czech Republic, which was one of the “first wave” countries (Dries et al.,
2004) in the expansion of western retailers into emerging markets, the number of
hypermarkets grew from a mere 7 in 1997 to 192 in 2005. Thanks to the increased use of
cars for shopping (INCOMA Research, 2006), retailers have a larger pool of potential
customers. As a result, more and more households identify hypermarkets, and
7 On one hand, one could imagine that the latter effect might attenuate or eliminate the former: As the number of firms decrease through the dynamic channel, an oligopolistic market structure might emerge and offset the aforementioned static channel. On the other hand, the number of firms does not necessarily decrease as long as entry takes place simultaneously with exit so that the end result is a turnover of retailers. For all practical purposes, we assume that the degree of competition among the remaining retailers and prospective entrants is high enough to ensure a downward pressure on prices in the period we are considering.
increasingly, discount markets, as their main source for food shopping.8 Since eastern
Europeans consumers are reported to care more about the price than their western
counterparts (Ewing, 2005), price wars are more common.9 Furthermore, the retail market
has already shown signs of coming to its saturation point with some retailers (e.g.,
Carrefour) pulling out of some markets (e.g., the Czech Republic) because of fierce
competition (Drtina, 2006) and others expanding into less urban areas (Deloitte, 2006).
This creates a more concentrated market where the most productive retailers stay in
business.
C. Measuring Retail Competition Intensity
Our main measure of retail competition intensity is modern retailers’ grocery sales share.
Modern retailers are defined as chain stores whose formats fall into: (1) hypermarkets, (2)
supermarkets, or (3) discount stores.10 General definitions of these three store formats are
as follows: (a) Hypermarkets sell large variety of food as well as non-food items; (b)
Supermarkets sell large variety of food and limited variety of non-food items; (c)
Discount stores sell limited variety of goods (mainly grocery items) at low prices. Among
these three, hypermarkets are the largest store format. The share of these modern retail
stores in grocery sales is expected to show the retail competition level at least indirectly
8 According to a survey by the private market research companies Incoma and GfK, the proportion of Czech households whose main venue for food shopping is hypermarkets grew from a mere 1 percent in 1997 to 37 percent in 2003.
9 FleetSheet, a news bulletin in the Czech Republic, reported that major retailers, such as Tesco, Globus, and Delvita, track prices so that no competitor within 10 kilometers of each of their stores has lower prices on certain products.
10 See Appendix for more details on the statistical definitions and examples of each format. Although the difference of these three definitions are somewhat vague, it is not a serious problem since we use the total sales share of these three store formats as the benchmark retail competition index.
since, as competition becomes intense, less productive stores are forced to exit and only
modern retailers, which are supposedly more productive, can remain in the market.
Hence, we can expect that intense competition brings high sales share of modern
retailers. While this index is less direct than other candidates such as mark-up rates or
retailers’ productivity level, it is the only readily available one. European Marketing
Data and Statistics is the primary data source. This annually published book provides
retail industry data in comparable format over time and across countries. The sample
consists of eighteen EU member countries for 2000, 2001, 2002, and 2004. The countries
in the sample are Austria, Belgium, Czech Republic, Denmark, Finland, France,
Germany, Greece, Hungary, Ireland, Italy, Netherlands, Poland, Portugal, Slovak
Republic, Spain, Sweden and United Kingdom. Due to the lack of retail data, the sample
leaves out seven of the EU countries (Cyprus, Estonia, Latvia, Lithuania, Luxemburg,
Malta and Slovenia). The final data set used in the econometric analysis is balanced by
filling in the missing data in 2003 with the average grocery sales share of modern
retailers for each country based on the observations available in 2000, 2001, 2002, and
2004.11
This benchmark index provides a convenient way to summarize the
competition of modern-format retailers vis-à-vis small, “mom-and-pop” retailers. But, it
does not distinguish between store formats or reveal information about competition
among modern-format retailers. Due to the limited availability of data, it is difficult to
11 To address any concerns that filling in the data for 2003 may introduce measurement error and, hence, bias the coefficient estimates, we also re-do the analysis by dropping 2003 from the final data set altogether. The results, not presented for sake of brevity, remain the same and are available from the authors upon request.
argue that we construct the perfect measure of retail competition intensity, or even a more
conventional one such as the Herfindahl-Hirschman index. Instead, we use three
alternative measures to ensure that the choice of indices does not significantly affect the
results: (1) the sales share of hypermarkets and discount stores but not supermarkets, (2)
the sales share of hypermarkets only, and (3) the number of modern retail stores per
million population.
The first and second alternative indices aim to reflect the differences in store
formats. Generally, hypermarkets are considered to constitute a more advanced store
format than supermarkets. If this claim is true, strong retail competition eventually
replaces supermarkets with hypermarkets (similar to mom-and-pop shops being replaced
by the modern store formats). Hence, high sales share of hypermarkets and low sales
share of supermarkets can be interpreted as signals of strong retail competition. Similarly,
discount stores tend to have less overhead and more aggressive pricing strategies. Similar
arguments are possible for other combinations of these three store formats; however, this
combinations involving hypermarkets are likely to be the more important because (1)
hypermarkets have all the advantages of supermarkets and (2) most hypermarkets are
owned by retail giants that operate internationally, and, hence, have good logistics while
some supermarkets operate domestically and are rarely different from small, traditional
grocers. Whether the benchmark index is more suitable than these alternatives may also
depend on the transition stage the retail industry is in. If the transition is still from small,
traditional stores to modern retail outlets, the benchmark index is more appropriate.
However, if this stage has already been completed and the current transition is from
supermarkets to hypermarkets, the alternative index based on sales share of hypermarkets
and discount stores only is more appropriate. Additionally, the differences in the range of
the product lines (e.g., hypermarkets being more likely to sell clothes or big-ticket items
than supermarkets or discount stores) may help distinguish the effect of retail competition
on the price dynamics of different goods categories.
The third alternative index relies on store numbers rather than sales share to
get some sense of competition among the modern-format retailers since it is correlated
with the number of firms active in the retail industry in a given country. Hence, it may
capture different competitive implications because, for example, a 50 percent market
share for hypermarkets could be consistent with one firm selling the entire amount or a
large number of firms with equal shares, making the aggregate market share a potentially
noisy measure of competition. By definition, this index does not discriminate among
three different store formats although each format might have a different impact on the
degree of competition in the retail industry. For example, consider a country that has a
population of one million and ten independent grocery stores. Suppose that the entry of
one hypermarket leads to the closing of four independent grocers while that of one
supermarket leads to the closing of two grocers, reflecting the difference in market
coverage between these two store formats. When each independent grocery store
maintains 10 percent of population as its customer base, the entry of one supermarket
increases grocery sales share of modern retail stores from 0 to 20 percent while that of
one hypermarket increases this share to 40 percent. However, in both cases, the number
of modern retail stores per million population is one. Thus, the store-number-based index
does not distinguish between these two cases while the sales-share-based index does. The
desirable feature of the store-number-based index is that it captures the changes in
competition intensity stemming from the tension between incumbents and entrants.
However, it fails to capture the fact that a large-format entrant is likely to increase
competition intensity more than a small-format entrant because it can reach a larger
customer base and utilize returns to scale and networking better.
The benchmark retail competition index indicates that substantial
heterogeneity exists across countries both in a given year and in their experience over
time (Figure 6). Cross-sectional variation appears to be more important than panel
variation based on a comparison of the horizontal range and the distance of the
observations from the 45-degree line. Retail competition intensity in central and eastern
Europe started from lower levels but had almost reached the average level of other EU
member countries by 2004. That year, about 58 percent of grocery sales on average in the
four central and eastern European countries in the sample (Czech Republic, Hungary,
Poland, and Slovak Republic) took place in modern retail stores. This ratio is slightly
below the average of EU-15 countries (65 percent).
The data also confirm the prevalence of different store formats across
countries. For instance, the rise of hypermarkets and the fall of supermarkets characterize
the retail landscape in the Czech Republic. In 2004, 34 percent of Czech grocery sales
took place in hypermarkets. This is the second highest hypermarket share in the sample
(just behind the French) while the supermarket share (15 percent) is the second lowest. In
Hungary, the reverse was the case: while hypermarkets left the Hungarian retail market,
supermarkets rose to prominence with their share in grocery sales increasing from 16 to
59 percent from 2000 to 2004.
The benchmark index has moderate correlation with the alternative indices.
Correlation coefficient for the grocery sales share of hypermarkets and that of
supermarkets in 2004 is -0.63 while those of other combinations are relatively low (Table
1). This negative correlation supports the idea that hypermarkets replace supermarkets
but not discount stores. Also, the correlation between the benchmark index and the three
alternative indices are 0.57, 0.42, and 0.36, respectively (Table 2), suggesting that the
choice of retail competition index would have limited effects on the results of the
statistical analysis.
III. ECONOMETRIC ANALYSIS
A. Empirical Methodology
The empirical framework is based on two recent studies about the effects of Wal-Mart on
the local economy in the U.S. (Basker, 2005a, b). These studies aim to quantify the
dynamic effects of Wal-Mart entry on local employment (Basker, 2005a) and retail price
of several consumer goods such as shampoo (Basker, 2005b). Regressing these variables
on an index that represents a history of Wal-Mart’s presence and several other variables
to control for the economic factors that are traditionally assumed to be behind
employment and inflation dynamics, she finds that Wal-Mart entry contributes both to a
net increase in local employment and to a decrease in retail price.
We estimate a modified version of Basker’s model. The main difference stems
from the fact that we express our specification in terms of inflation, rather than price
levels. To put it more precisely, we subtract the log of the lagged price level from each
side of Equation (1) in Basker (2005b) to obtain
ln CPIit – ln CPIit-1 = β0+ β1 compit+ γ`Xit + δ1Ci+ δ2Yt+ εit (1)
where CPIit is the price index for a given category of goods and compit is the measure of
retail competition intensity, i.e., the benchmark index described in Section II.C. 12, 13, 14
Subscripts i and t represent country i and year t, respectively. Xit is a set of control
variables, which we specify below. Ci and Yt are country and year fixed effects,
respectively. εit is presumably an i.i.d. normal error term; we discuss the implications of
potential serial correlation in the error term below when we introduce an alternative
specification.
The reason we convert Basker’s specification from levels to changes is two-
fold. First, the focus of our investigation is inflation rather than price levels. Second, we
have data on the price indices, rather than price levels of particular goods comparable
across countries. Note that this data issue does not affect the comparability of inflation
rates since we use the indices harmonized across countries according to Eurostat
guidelines.
12 Basker’s equation in its original form is pkjt = αk+ βk pkjt-1 + θkWMjt+ Σjγkjcj+ Σtδktqt+Σj τkjtt + εkjt,where pkjt is the natural log of the price of product k in city j in quarter t. cj is a city indicator, qt is a quarter indicator, tt is a linear time trend, and WMjt is the Wal-Mart indicator (it equals 1 if city j has a Wal-Mart store in quarter t). Suppressing the product subscript, k, imposing the assumption that β equals 1, and rearranging with a slightly different set of control variables delivers the baseline regression equation we have in (1).
13 We repeat the analysis with the alternative measures of retail competition intensity. The results are not reported here for brevity but are available upon request.
14 Another difference between the reduced form here and the ones in Basker (2005a, b) is the choice of proxy for measuring competition effects. Here, the proxy for retail competition index is the grocery sales share of modern retail stores whereas Basker uses a Wal-Mart entry dummy. To be more precise, Baker (2005b) uses a conventional dummy variable whereas Basker (2005a) uses the number of Wal-Mart stores per capita as an independent variable to capture the Wal-Mart effect on local employment. While the latter is not a dummy variable, it is very close to being binary since a significant number of locations do not have any Wal-Mart stores.
We rewrite the benchmark specification, using ΔCPIit ≈ ln CPIit – ln CPIit-1
and spelling out Xit , as
ΔCPIit = β0+ β1 compit+ γ1 gapit+ γ2 CEEi+ γ3 EUROit + δ1Ci+ δ2Yt + εit (2)
where ΔCPIit is the inflation rate for a given category of goods, calculated using
Harmonized Indices of Consumer Prices (HICP) from Eurostat. gapit is the deviation of
output from its trend, calculated using the Hodrick-Prescott filter15; CEEi is a dummy
variable that takes on value 1 if country i is 1 of the central and eastern European
countries; EUROit is another dummy variable whose value is 1 if country i uses euro as its
currency at time t and 0 otherwise. As an alternative specification, we add the lagged
inflation rate on the right-hand side to account for persistence in the inflation dynamics to
get
ΔCPIit = β0+ β1 compit+β2 ΔCPIit-1+ γ1 gapit+ γ2 CEEi+ γ3 EUROit + δ1Ci+ δ2Yt + εit (3)
β1 shows the marginal impact of retail competition on inflation and is our main
coefficient of interest. We expect β1<0 since retail competition decreases retailers’ mark-
up and forces less productive firms to shut down their businesses, easing pressure on
consumer prices. β2 , when it is included, shows whether the inflation rate is persistent
and, at the same time, stationary. 0< β2 <1 implies that past increases in prices affect the
current increases (e.g., through expectations) but past shocks to the inflationary process
eventually disappear. Output gap, gapit, shows how far GDP deviates from its trend. This
15 The following is the detailed procedure to calculate output gap. First, using Hodrick-Prescott filter, we separate country GDP yt into its trend y*
t and its deviation from the trend εt, i.e., yt=y*t+ εt. Output gap is
then calculated as the ratio of deviation to trend, i.e., gapt= εt/ y*t.
captures the business cycle effects on inflation. We expect that β3>0 because positive
deviation of GDP from its trend reflects high demand, and, hence, generates upward
pressure on prices translating into a higher inflation rate. CEE dummy is intended to
capture any intrinsic difference between the old EU member states and the new member
states from central and eastern Europe. Euro dummy captures any common persistent
inflation pressure received by countries that have adopted euro as their currency. We also
employ time dummies for the impact of temporary shocks that uniformly affect all
countries in the sample and country dummies that would account for any country-specific
factors.
Before we move on to the results of the analysis, a brief discussion of the error
term is warranted. If εit is not i.i.d. and exhibits dependence over time due to, e.g.,
measurement error, the coefficient estimates may be biased when there is the lagged
inflation term on the right-hand side. In that case, the baseline specification omitting
lagged inflation provides a cross-check on the reliability of the estimates in addition to
the usual diagnostic tests we perform.
B. Regression Results
Table 3 presents the summary statistics used in the econometric analysis. There is enough
variation in the panel data set both for retail competition and inflation rates. Scatter plots
suggest a negative relation between the intensity of retail competition and inflation
(Figure 7). In what follows, we present the findings from econometric analysis aiming to
isolate this relation from the influence of other factors. We first estimate the equation for
retail inflation.16 To explore industry-specific differences, we look at the estimates for
two separate retail inflation rates: food inflation and clothing inflation.17 Moreover, to
empirically evaluate the impact of retail competition on sectors other than retail, the
regressions are repeated using overall inflation rate as well as non-retail inflation rate.
Retail competition has a sizeable downward impact on food inflation while its
impact on clothing inflation is not as clear (Table 4; column 1 compared to column 5).
The parameter estimate for retail competition is negative and significant only when food
inflation is regressed. Adding lagged inflation rate on the right-hand side does not alter
this result (columns 3-4 and 5-6). Unit root and normality tests confirm that the error term
is normal i.i.d., further alleviating the concern that the estimates may be biased. We
would expect the coefficient on retail competition to be significant in the regression using
clothing inflation as the dependent variable. One reason for the insignificance of the
coefficients in that case is perhaps the way the retail competition index is constructed.
Since this index is based on grocery sales share, it might not properly reflect the
importance of modern-format retailers or, more broadly, competition level in sales of
clothing.
The negative effect of retail competition on inflation remains when food,
beverages, and clothing items are combined under retail goods (Table 5, columns 1-4). It
is not surprising that the coefficient on retail competition is not highly significant when
non-retail inflation is the dependent variable since, arguably, retail competition does not
16 The term “retail inflation” is used for increases in the price of food, beverages, and clothing.
17 The term “food inflation” is used for increases in the price of food and beverages.
have direct effects on the price of non-retail goods (columns 5-8). This also suggests that
the effect we are focusing on is unlikely to be due to a general tendency for inflation to
decline as it falsifies the notion that intensified retail competition is associated with a
decline in inflation even for non-retail goods.
The significance of the retail competition coefficient shows a pattern similar
to the non-retail inflation case in the regressions using overall inflation as the dependent
variable (Table 6). Since retail inflation is a mere fraction of overall inflation, its effects
are difficult to capture in a regression analysis using this aggregate inflation measure, let
alone with so few observations.
According to the estimates, if another 10 percent of total grocery shopping
took place at modern retail stores during a given year, then food inflation rate in that year
would drop by 0.58 percentage points. The actual annual average inflation rate for food
and beverages between 2000 and 2004 was 1.95. Hence, the impact of retail competition
in the form of high-productivity, modern-format retailers might have helped food
inflation be almost 30 percent lower than it would otherwise be.
IV. DISCUSSION
This section discusses the relevance of the results presented by examining robustness and
potential identification problems. We start with the use of alternative indices as a
robustness check. Then, we inquire whether endogeneity of retail competition intensity
and inflation developments is a concern.
Regressions using alternative retail indices also indicate downward impact on
inflation. To check the robustness of the regression estimates, we run the same
regressions but use the three alternative retail competition indices described in Section
II.C.18 First, we find that the estimation results using the grocery sales share of
hypermarkets and discount stores (but not supermarkets) and that of hypermarkets only
are similar to the benchmark results. Again, the results indicate that 10 percentage points
increase in grocery sales share of hypermarkets and discount stores decreases food
inflation rate by 0.31 percentage points. Second, regression results using the number of
modern retail stores per million population are somewhat different from the benchmark
regression results. The main difference is that, although the signs of estimates are the
same, most of the coefficients are insignificant. This difference may reflect the
shortcomings of an index based on store numbers. As discussed earlier, the indifference
of this index to relative store sizes among the three store formats might bias the
estimation results. In particular, the competitive influence of different store formats may
be considerably different (e.g., a single hypermarket may introduce more competitive
pressure than multiple supermarkets due to its ability to exploit returns to scale at a
higher rate) and not accounting for this may obscure the effect of retail competition on
prices and inflation. At the end, what matters may be whether modern retailers are
entering a market or not rather than how extensive the modern retail network is in a
country.
18 These regressions are not reported here for sake of brevity, but are available upon request.
There are other explanations that are consistent with the estimation results.
While we consider the case in which retail competition affects inflation, the direction of
causality could be the opposite. Suppose, for example, as most modern retailers are
foreign-owned, that stable inflation is a key factor for their entry decision. If this is the
case, inflation rates could negatively correlate with a measure of retail competition even
when retail competition itself does not affect inflation. To examine this possibility, we
need to consider the entry behavior of modern retailers.
To verify this explanation, we examine whether price volatility affects the
presence of modern retailers. Following the specification used in foreign direct
investment literature (e.g., Bevan and Estrin, 2004), we estimate a simple entry model of
modern retailers:
compit=φ0+ φ1ΔCPIit-1+ φ2volit+φ3insit+φ4incit+φ5popit+ φ6ΔGDPit+νit (4)
where vol is the level of price volatility, ins represents the development level of the
business environment (e.g., legal and regulatory system), inc is the income level, and pop
measures how large a country is. We also include the growth rate of real GDP
considering that modern retailers might be more willing to enter into more rapidly
growing markets. Since firm-level entry data are not available, we use the sales-share-
based index compit as a proxy for the development of modern retailers.19 We use the
standard deviation of monthly HICP series calculated over the months in each year for
19 Several major retail chains (e.g., ALDI) are privately held (as opposed to being publicly traded) and do not disclose the information on the development of their store network.
vol, Ease of Doing Business Ranking20 developed by the World Bank for ins, GDP per
capita in purchasing power parity terms for inc, and the log of population for pop. Both
GDP per capita and population data come from the IMF’s International Financial
Statistics. We are interested in whether φ1 is significantly below zero and φ2 is
significantly above zero. If this is the case, as inflation falls and the price level becomes
less volatile in a country, modern retailers are less likely to start business there.
There is evidence of a negative relationship between entry of retailers and
inflation (Table 7). Although the magnitude of the coefficient estimate varies
considerably depending on which other variables are included on the right hand side, the
results suggest that retail competition tends to be less intense in countries with higher
inflation. However, there is no evidence of a significant relationship between entry and
volatility. Higher income levels attract modern retailers, but size is not an as important
factor. The only counterintuitive sign is for the coefficient on real GDP growth: faster
growing countries appear to have a shrinking share of modern retailers.
To complete the analysis one needs to take into account that the impact of
retail competition on inflation and that of inflation on retail competition work
simultaneously. Hence, we estimate the two regression equations we have estimated
separately before in a seemingly unrelated equations framework. Results presented in
Table 8 indicate that the previous findings are not altered: there is empirical evidence that
entry decision of modern retailers depends on inflation, yet their entry still has an impact
on inflation. Interestingly, when the simultaneity is taken into account, the coefficient for
20 We take the inverse of a country’s ranking. Therefore, a better business environment corresponds to a higher institution score. Raw data can be downloaded at http://www.doingbusiness.org/.
overall inflation also becomes significant, but the impact on food inflation is larger in
magnitude.
An alternative method to look at the potentially two-way relationship between
inflation and the retail environment is to control for “inflationary conditions” in a country
by including non-retail inflation rate in the main regression.21 The results, not reported
here for brevity but available upon request, support the conclusions from the baseline
specification. This is consistent with retail competition having an effect on inflation,
particularly for food items, on top of what would be expected by the general inflation
dynamics in the country although does not establish causality beyond doubt.
V. CONCLUSION
We document evidence suggesting that more intense retail competition in the form of
more prevalence in the market by modern retailers contributed to inflation moderation in
the sectors that are directly affected by the presence of modern retailers in Europe
between 2000 and 2004. The regression results imply that another 10 percent increase in
the grocery sales share of modern retail stores reduces food inflation rate by more than
0.6 percentage points. Lower mark-up and productivity increase in retail stores may
explain the dampening of price pressures. These findings may have real implications for
welfare if households use the “savings” from lower price inflation in one of the major
items in their budget to spend more on goods and services that improve their living
conditions.
21 We thank an anonymous referee for this suggestion.
Appendix: Definition of Store Formats
According to the documentation provided by Euromonitor, each store format is defined as follows. Hypermarkets: Retail outlets selling groceries and non-food merchandise with a
retail sales area of over 2,500 square meters. Frequently located on out-of-town sites or as the anchor store in a shopping centre. Example brands include Carrefour, Tesco Extra, Géant, E Leclerc, Intermarché, Auchan. Excludes cash and carry, warehouse clubs and mass merchandisers.
Supermarkets: Retail outlets selling groceries with a selling space of between 400 and 2,500 square meters, selling at least 70 percent groceries. Excludes discounters, convenience stores and independent grocery stores. Example brands include Champion, Tesco, Casino.
Discounters: Discounters comprises hard discounters and soft discounters. Hard discounter: first introduced by Aldi in Germany, and also known as limited-line discounters. Retail outlets, typically of 300-900 square meters, stocking fewer than 1,000 product lines, largely in packaged groceries. Goods are mainly private-label or budget brands. Soft discounter: usually slightly larger than hard discounters, and also known as extended-range discounters. Retail outlets typically stocking 1,000-4,000 product lines. As well as private-label and budget brands. Stores commonly carry leading brands at discounted prices. Discounters exclude mass merchandisers and warehouse clubs. Example brands include Aldi, Lidl, Plus, Penny, Netto.
References Alexander, N. and H. Myers, 1997. Food Retailing Opportunities in Eastern Europe. European Business Review, 97 (3): 124-133. Allard, C. 2006. Inflation in Poland: How Much Can Globalization Explain? Poland:
2006 Article IV Consultation – Selected Issues, IMF. Basker, E., 2005a. Job Creation or Destruction? Labor-Market Effects of Wal-Mart Expansions. Review of Economics and Statistics, 87: 174-183. Basker, E., 2005b. Selling a Cheaper Mousetrap: Wal-Mart’s Effect on Retail Prices. Journal of Urban Economics, 58: 203-229. Bevan, A. and S. Estrin, 2004. The Determinants of Foreign Direct Investment into
European Transition Economies. Journal of Comparative Economics, 32: 775-787. Bresnahan, T. and P. Reiss, 1991. Entry and Competition in Concentrated Markets. Journal of Political Economy, 99 (5): 977-1009. Deloitte, 2006. Retail in Central Europe. Consumer Business Market Review. Dries, L., T. Reardon, and J. Swinnen, 2004. The Rapid Rise of Supermarkets in Central
and Eastern Europe: Implications for the Agrifood Sector and Rural Development. Development Policy Review, 22 (5): 525-556.
Drtina, T., 2006. Hypermarkets Face Saturation Points. Czech Business Weekly, 27. Euromonitor, 2005. European Marketing Data and Statistics. Ewing, J., 2005. Hungry for Discounts, Not Delicacies. Business Week, May 23. Global Insight, 2005. The Economic Impact of Wal-Mart. Report by Advisory Services Division. Hopenhayn, H., 1992. Entry, Exit, and Firm Dynamics in Long Run Equilibrium. Econometrica, 60 (5): 1127-1150. Javorcik, B., W. Keller, and J. Tybout, 2008. Openness and Industrial Response in a Wal-
Mart World: A Case Study of Mexican Soaps, Detergents and Surfactant Producers. The World Economy, 31(12): 1558-1580.
Javorcik, B. and Y. Li, 2008. Do the Biggest Aisles Serve a Brighter Future?
Global Retail Chains and Their Implications for Romania. World Bank Working Paper No. 4650.
INCOMA Research, 2006. Price Sensitivity in Central and Eastern Europe: Shopper Segmentation According to the INCOMA Research and GFK Study. Press Release.
Lagakos, D., 2009. Superstores or Mom and Pops? Technology Adoption and
Productivity Differences in Retail Trade. Federal Reserve Bank of Minneapolis Research Department Staff Report 428.
Figure 1. Central European Countries: HICP Inflation, 1997-2005(In percent)
Source: Eurostat.
-5
0
5
10
15
20
1997 1998 1999 2000 2001 2002 2003 2004 2005
CZE HUN
POL SVK
SVN EU15
Figure 2. Selected European Countries: Inflation Developments, 1997 versus 2005
Source: Eurostat.
0
4
8
12
16
Bel
gium
Ger
man
y
Swed
en
Uni
ted
Kin
gdom
Ital
y
Net
herl
ands
Spai
n
Port
ugal
Den
mar
k
Gre
ece
Slov
ak R
epub
lic
Cze
ch R
epub
lic
Pola
nd
Hun
gary
HICP inflation, in percent1997
Average = 3.76
0
4
8
12
16
Finl
and
Swed
enN
ethe
rlan
dsC
zech
Rep
ublic
Den
mar
kFr
ance
Ger
man
yU
nite
d K
ingd
omA
ustr
iaPo
rtug
alIr
elan
dPo
land
Ital
yB
elgi
umSl
ovak
Rep
ublic
Spai
nG
reec
eH
unga
ry
HICP inflation, in percent2005
Average = 2.10
-15
-12
-9
-6
-3
0
3
Hun
gary
Pola
nd
Cze
ch R
epub
lic
Slov
ak R
epub
lic
Gre
ece
Swed
en
Finl
and
Net
herl
ands
Den
mar
k
Uni
ted
Kin
gdom
Port
ugal
Ger
man
y
Ital
y
Fran
ce
Irel
and
Aus
tria
Bel
gium
Spai
n
Change in HICP inflation from 1997 to 2005, in percentage points
HungaryPoland
Czech Republic
Slovak Republic
GreeceSweden
FinlandNetherlandsDenmarkUnited KingdomPortugalGermanyItalyFranceIrelandAustria
BelgiumSpain
-15
-12
-9
-6
-3
0
3
0 2 4 6 8 10 12 14 16
Cha
nge
in H
ICP
infl
atio
nfr
om 1
997
to 2
005,
in p
erce
ntag
e po
ints
HICP inflation in 1997, in percent
70
75
80
85
90
95
100
Sout
h K
orea
Pola
nd
Bra
zil
Chi
le
Tai
wan
Indo
nesi
a
Mal
aysi
a
Arg
entin
a
Tha
iland
Cze
ch R
epub
lic
Tur
key
Vie
tnam
Phili
ppin
es
Chi
na
Saud
i Ara
bia
Indi
a
Mex
ico
Uru
guay
Rus
sia
Rom
ania
Col
ombi
a
Isra
el
Slov
ak R
epub
lic
Egy
pt
Sout
h A
fric
a
Hon
g K
ong
Slov
enia
Ven
ezue
la
Bul
garia
Hun
gary
Figure 3. Top 30 Most Attractive Emerging Markets for International Retailers, 1995 1/
Source: A. T. Kearney.1/ A. T. Kearney, a consulting company, calculates the Global Retail Development Index (GRDI) to evaluate the attractiveness of emerging markets for international retailers, based on an assessment of country and market risks. The data used in the calculations come from UN Population Division Database, the World Economic Froum's Global Competitiveness Report, national statistics, Euromoney, World Bank reports, and Euromonitor and Planet Retail databases.
Figure 4. Central European Countries: Contributions to HICP Inflation, 1997-2005(In percent)
Source: Eurostat.
-0.6
-0.3
0.0
0.3
0.6
0.9
1997 1999 2001 2003 2005
Czech Republic
Food (inc. alcohol and tobacco) Clothing Furnishings Transport Non-tradables
-0.6
-0.3
0.0
0.3
0.6
0.9
1997 1999 2001 2003 2005
EU-15
-0.6
-0.3
0.0
0.3
0.6
0.9
1997 1999 2001 2003 2005
Hungary
-0.6
-0.3
0.0
0.3
0.6
0.9
1997 1999 2001 2003 2005
Poland
-0.6
-0.3
0.0
0.3
0.6
0.9
1997 1999 2001 2003 2005
Slovenia
-0.6
-0.3
0.0
0.3
0.6
0.9
1997 1999 2001 2003 2005
Slovak Republic
Figure 5. Central European Countries: HICP Inflation by Sectors, 1997-2005 1/(Year-on-year percent change)
Source: Eurostat.1/ Retail sector consists of food, beverages, and clothing. Non-retail includes housing, transport, and services.
-5%
0%
5%
10%
15%
20%
1997 1999 2001 2003 2005
Czech Republic
-5%
0%
5%
10%
15%
20%
1997 1999 2001 2003 2005
Hungary
-5%
0%
5%
10%
15%
20%
1997 1999 2001 2003 2005
Poland
-5%
0%
5%
10%
15%
20%
1997 1999 2001 2003 2005
Slovak Republic
-5%
0%
5%
10%
15%
20%
1997 1999 2001 2003 2005
Slovenia
Non_Retail
Overall
Retail
Figure 6. Evolution of Grocery Sales Share by Store Format, 2000 versus 2004(In percent)
Source: European Marketing Data and Statistics.
Austria
BelgiumCzech
Republic
Denmark
Finland
France
Germany
Greece
Hungary
Ireland
Italy
Netherlands
PolandPortugal
Slovak Republic
SpainSweden
United Kingdom
0
20
40
60
80
100
0 20 40 60 80 10020
04
2000
Total
Austria
Belgium
Czech Republic
DenmarkFinland
France
Germany
Greece
HungaryIreland
Italy
Netherlands
Poland
Portugal
Slovak Republic Spain
Sweden
United Kingdom
0
10
20
30
40
50
60
0 10 20 30 40 50 60
2004
2000
Hypermarkets
Austria
Belgium
Czech Republic
DenmarkFinland
France
Germany
Greece
Hungary
Ireland
Italy
Netherlands
Poland
Portugal
Slovak Republic
SpainSweden
United Kingdom
0
20
40
60
80
0 20 40 60 80
2004
2000
Supermarkets
Austria
Belgium
Czech Republic
Denmark
FinlandFrance
Germany
Greece
Hungary
IrelandItaly
Netherlands
Poland
Portugal
Slovak Republic
Spain
Sweden
United Kingdom
0
10
20
30
40
0 10 20 30 40
2004
2000
Discount stores
Figure 7. Retail Competition and Inflation 1/
Source: Eurostat and European Marketing Data and Statistics.1/ Retail competition index, measured as the share of modern retailers in total grocery sales, is on the horizontal axis. HICP inflation is shown on the vertical axis. Both are expressed in percent.
-10
-5
0
5
10
15
0 20 40 60 80 100
Overall
-10
-5
0
5
10
15
0 20 40 60 80 100
Non-Retail
-10
-5
0
5
10
15
0 20 40 60 80 100
Food and beverages
-10
-5
0
5
10
15
0 20 40 60 80 100
Clothing
Hypermarkets Supermarkets Discount Stores
Hypermarkets 1.00Supermarkets -0.63 1.00Discount Stores 0.11 -0.11 1.00
Table 1. Correlation Matrix of Sales Share of Each Store Format in 2004
Source: European Marketing Data and Statistics ; authors' calculations.
DHS DH H NUMBER
DHS 1.00DH 0.57 1.00H 0.42 0.81 1.00NUMBER 0.36 0.14 0.06 1.00
Table 2. Correlation Matrix of Alternative Retail Competition Indices
Source: European Marketing Data and Statistics ; authors' calculations.Notes: DHS is the sum of grocery sales share of discount stores, hypermarkets and supermarkets. DH is the sum of grocery sales share of discount stores and hypermarkets. H is the sum of grocery sales share of hypermarkets only. NUMBER is the number of discount stores, hypermarkets, and supermarkets per one million population.
PeripheryAustria France Netherlands GreeceBelgium Germany Spain PortugalDenmark Ireland SwedenFinland Italy United Kingdom
Obs Mean Std. Dev. Min Max
Retail competitionBenchmark 90 64.354 17.395 16.157 91.300Alternative 1 90 29.446 13.838 2.200 65.080Alternative 2 90 18.194 10.503 0.000 51.647Alternative 3 90 163.497 213.880 0 970
Inflation rateOverall 90 3.081 2.211 -0.104 12.193Retail 90 2.250 2.384 -2.291 11.231Non-Retail 90 3.302 2.374 0.646 14.423Food 90 2.958 2.472 -2.597 12.556Clothing 90 0.120 2.877 -7.532 6.104
Output gap 90 0.208 1.613 -3.488 4.018Volatility 90 0.521 0.290 0.124 1.469Institution 90 0.048 0.037 0.009 0.200Income 90 3.118 0.334 2.294 3.652Size 90 2.733 0.970 1.335 4.415Real GDP growth 90 2.733 1.780 -1.120 9.220
Czech RepublicHungaryPoland
Slovak Republic
Table 3. The Sample
Sources: European Marketing Data and Statistics, Eurostat, IMF International Financial Statistics, World Bank; authors' calculations.Notes: Retail competition, inflation rate, output gap, and real GDP growth are expressed in percent. Benchmark retail competition index is DHS (the sum of grocery sales share of discount stores, hypermarkets and supermarkets). Alternative 1 is DH (the sum of grocery sales share of discount stores and hypermarkets). Alternative 2 is H (the sum of grocery sales share of hypermarkets only). Alternative 3 is NUMBER (the number of discount stores, hypermarkets, and supermarkets per one million population). Inflation rates are calculated using the harmonized index of consumer prices (HICP). Output gap is the ratio of deviation of actual GDP from its trend, calculated using the Hodrick-Prescott filter. Volatility is the standard deviation of the monthly price index. Institution is the inverse of the country ranking according to the World Bank's ease of doing business index. Income is the log of GDP per capita in purchasing power parity terms. Size is the log of population. Real GDP growth is the annual change in GDP in real terms.
List of Countries
Summary Statistics
Western EuropeCore
Central and Eastern Europe
(1) (2) (3) (4) (5) (6) (7) (8)
Retail competition -0.059** -0.059** -0.058** -0.058** -0.035 -0.035 -0.007 -0.007[0.026] [0.026] [0.025] [0.025] [0.031] [0.0314] [0.0214] [0.0214]
Output gap 0.800*** 0.800*** 0.801*** 0.801*** 0.189 0.189 0.320 0.320[0.246] [0.246] [0.251] [0.251] [0.249] [0.249] [0.193] [0.193]
Inflation rate, lagged 0.019 0.019 0.460*** 0.460***[0.165] [0.165] [0.126] [0.126]
CEE dummy 0.011 0.011 0.029** 0.020*[0.011] [0.011] [0.012] [0.011]
Euro dummy 0.010** 0.010** 0.063*** 0.031***[0.005] [0.004] [0.010] [0.011]
Constant 0.058** 0.047** 0.057** 0.046** 0.024 -0.039 -0.006 -0.037**[0.024] [0.021] [0.022] [0.019] [0.0273] [0.0252] [0.019] [0.018]
Number of observations 90 90 90 90 90 90 90 90Adjusted R-square 0.69 0.69 0.69 0.69 0.77 0.77 0.83 0.83Time fixed effects yes yes yes yes yes yes yes yesCountry fixed effects yes yes yes yes yes yes yes yes
Notes: All regressions are estimated using OLS. The dependent variable in columns (1) through (4) is the annual inflation rate for food items. The dependent variable in columns (5) through (8) is the annual inflation rate for clothing items. Retail competition is the share of hypermarkets, supermarkets, and discount stores in grocery sales. Output gap is the ratio of deviation of actual GDP from its trend, calculated using the Hodrick-Prescott filter. CEE dummy is a binary variable that is 1 if the country is in central and eastern Europe. Euro dummy is a binary variable that is 1 if the country uses the euro as its currency. Robust standard errors are in brackets. *, **, and *** denote statistical significance at 10, 5, and 1 percent level, respectively. All regressions include time and country fixed effects, coefficients of which are suppressed in the table.
Table 4. Regression Results: Retail Sector
Dependent Variable: Inflation Rate
ClothingFood
(1) (2) (3) (4) (5) (6) (7) (8)
Retail competition -0.055** -0.055** -0.047** -0.047** -0.039* -0.039* -0.007 -0.007[0.025] [0.025] [0.022] [0.022] [0.022] [0.022] [0.017] [0.017]
Output gap 0.699*** 0.699*** 0.716*** 0.716*** 0.078 0.078 0.387* 0.387*[0.228] [0.228] [0.233] [0.233] [0.253] [0.253] [0.219] [0.219]
Inflation rate, lagged 0.133 0.133 0.449** 0.449**[0.170] [0.170] [0.174] [0.174]
CEE dummy 0.028*** 0.028*** 0.011 0.016[0.009] [0.010] [0.010] [0.010]
Euro dummy 0.031*** 0.027*** 0.008 0.003[0.005] [0.006] [0.006 [0.005]
Constant 0.048** 0.017 0.041** 0.014 0.065*** 0.057*** 0.024 0.021[0.022] [0.019] [0.020] [0.018] [0.021] [0.019] [0.017] [0.015]
Number of observations 90 90 90 90 90 90 90 90Adjusted R-square 0.73 0.73 0.73 0.73 0.74 0.74 0.80 0.80Time fixed effects yes yes yes yes yes yes yes yesCountry fixed effects yes yes yes yes yes yes yes yes
Notes: All regressions are estimated using OLS. The dependent variable in columns (1) through (4) is the annual inflation rate for retail goods. The dependent variable in columns (5) through (8) is the annual inflation rate for non-retail goods. Retail competition is the share of hypermarkets, supermarkets, and discount stores in grocery sales. Output gap is the ratio of deviation of actual GDP from its trend, calculated using the Hodrick-Prescott filter. CEE dummy is a binary variable that is 1 if the country is in central and eastern Europe. Euro dummy is a binary variable that is 1 if the country uses the euro as its currency. Robust standard errors are in brackets. *, **, and *** denote statistical significance at 10, 5, and 1 percent level, respectively. All regressions include time and country fixed effects, coefficients of which are suppressed in the table.
Table 5. Regression Results: Retail versus Non-Retail
Dependent Variable: Inflation Rate
Retail Non-Retail
(1) (2) (3) (4)
Retail competition -0.044** -0.044** -0.018 -0.018[0.022] [0.022] [0.016] [0.016]
Output gap 0.267 0.267 0.464** 0.464**[0.238] [0.238] [0.231] [0.231]
Inflation rate, lagged 0.405** 0.405**[0.180] [0.180]
CEE dummy 0.012 0.017*[0.009] [0.009]
Euro dummy 0.012** 0.007[0.004] [0.005]
Constant 0.061*** 0.049*** 0.030* 0.023*[0.020] [0.018] [0.0155] [0.014]
Number of observations 90 90 90 90Adjusted R-square 0.74 0.74 0.78 0.78Time fixed effects yes yes yes yesCountry fixed effects yes yes yes yes
Notes: All regressions are estimated using OLS. The dependent variable in all columns is the annual inflation rate for all goods. Retail competition is the share of hypermarkets, supermarkets, and discount stores in grocery sales. Output gap is the ratio of deviation of actual GDP from its trend, calculated using the Hodrick-Prescott filter. CEE dummy is a binary variable that is 1 if the country is in central and eastern Europe. Euro dummy is a binary variable that is 1 if the country uses the euro as its currency. Robust standard errors are in brackets. *, **, and *** denote statistical significance at 10, 5, and 1 percent level, respectively. All regressions include time and country fixed effects, coefficients of which are suppressed in the table.
Table 6. Regression Results: Overall Inflation
Dependent Variable: Inflation Rate
Overall
(1) (2) (3) (4) (5)
Inflation rate, lagged -2.418*** -2.410*** -1.638*** -1.340** -1.651***[0.505] [0.512] [0.505] [0.571] [0.566]
Price volatility -0.051 -0.052 -0.064 -0.058 -0.088[0.073] [0.073] [0.062] [0.060] [0.058]
Institutional quality -0.133 -0.234 -0.221 -0.268[0.963] [1.016] [1.044] [1.056]
Income 1.057** 1.124** 1.441***[0.466] [0.457] [0.498]
Population -1.171 -2.251*[1.462] [1.262]
Real GDP growth -0.0226**[0.010]
Constant 0.585*** 0.587*** -3.164* 2.235 6.04[0.099] [0.100] [1.623] [6.622] [5.577]
Number of observations 90 90 90 90 90Adjusted R-square 0.86 0.86 0.87 0.88 0.88Time fixed effects yes yes yes yes yesCountry fixed effects yes yes yes yes yes
Overall
Dependent Variable: Retail Competition
Table 7. Regression Results: Modern Retailers' Entry Behavior
Notes: All regressions are estimated using OLS. The dependent variable in all columns is retail competition measured by the share of hypermarkets, supermarkets, and discount stores in grocery sales. Inflation rate is the overall inflation for all goods. Price volatility is the standard deviation of the monthly price index. Institutional quality is the inverse of the country ranking according to the World Bank's ease of doing business index. Income is the log of GDP per capita in purchasing power parity terms. Population is the log of population. Real GDP growth is the annual change in GDP in real terms. Robust standard errors are in brackets. *, **, and *** denote statistical significance at 10, 5, and 1 percent level, respectively. All regressions include time and country fixed effects, coefficients of which are suppressed in the table.
Food Retail Overall(1) (2) (3) (4) (5) (6)
Retail competition -0.086*** -0.071*** -0.035**[0.022] [0.019] [0.017]
Output gap 0.882*** 0.783*** 0.496***[0.189] [0.169] [0.151]
Inflation rate, lagged -0.022 0.094 0.371*** -1.129** -1.171** -1.553**[0.111] [0.107] [0.100] [0.482] [0.545] [0.640]
Price volatility -0.075 -0.071 -0.113**[0.057] [0.057] [0.057]
Institutional quality -0.269 -0.120 -0.268[0.746] [0.760] [0.748]
Income 1.774*** 1.732*** 1.524***[0.362] [0.366] [0.381]
Population -3.084*** -2.888** -2.264*[1.138] [1.153] [1.185]
Real GDP growth -0.020* -0.019* -0.021*[0.011] [0.011] [0.011]
Constant 0.057*** 0.029*** 0.036*** 0 0 5.832[0.016] [0.011] [0.011] [0] [0] [4.972]
Number of observations 90 90 90 90 90 90Adjusted R-square 0.68 0.73 0.78 0.88 0.88 0.88Time fixed effects yes yes yes yes yes yesCountry fixed effects yes yes yes yes yes yes
DV: Retail Competition
Table 8. Regression Results: Joint Estimation
Notes: Regressions are estimated using SURE. Columns (1) and (4) ((2) and (5); (3) and (6)) correspond to the same SURE system. The dependent variable in columns (1), (2), and (3) is the annual inflation rate for food items, retail goods, and all goods, respectively. The dependent variable in columns (4), (5), and (6) is retail competition measured by the share of hypermarkets, supermarkets, and discount stores in grocery sales. Output gap is the ratio of deviation of actual GDP from its trend, calculated using the Hodrick-Prescott filter. Price volatility is the standard deviation of the monthly price index. Institutional quality is the inverse of the country ranking according to the World Bank's ease of doing business index. Income is the log of GDP per capita in purchasing power parity terms. Population is the log of population. Real GDP growth is the annual change in GDP in real terms. Robust standard errors are in brackets. *, **, and *** denote statistical significance at 10, 5, and 1 percent level, respectively. All regressions include time and country fixed effects, coefficients of which are suppressed in the table.
Equation 1: Impact of Retail Competition on Inflation
Equation 2: Impact of Inflation on Retail Competition
DV: Inflation Rate