Working title: Merger simulation in a differentiated product
market - Ecuador
Estefania Perez Benitez, 11087765 August 11, 2017
University of Amsterdam, Amsterdam Business School MSc Economics
Master Thesis
Abstract
Mergers affect the whole dynamic of a market, from the behavior of the rival and merging
firms, its business strategy, prices, supply and demand. This thesis examines the acquisition of SAB Miller by AB InBev in the Ecuadorian market. It will be seen that the merger will
dramatically increase beer price in Ecuador. These findings are in line with the assessment of the E.U. and the Ecuadorian competition authority, which proposed remedies as the main
Acknowledgements
I would like to express my gratitude to my supervisor Dr. Jo Seldeslachts for the useful
Statement of Originality
This document is written by student Estefania Perez Benitez who declares to take full
responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no
sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of com-pletion of the work, not for the contents.
Contents
1 Introduction 1
2 Beer market in Ecuador 3
3 Literature Review 10 4 Methodology 15 5 Data 19 5.1 Beer characteristics . . . 19 5.2 Variables . . . 21 6 Empirical Results 23 6.1 Nested Logit Model . . . 23
6.2 Merger Simulation . . . 29
6.2.1 Merger simulation without cost efficiencies. . . 30
6.2.2 Merger simulation with cost efficiencies. . . 31
7 Competition Overview 34
8 Conclusions 37
Appendix 39
1
Introduction
In November 2015, SAB Miller, the world’s second largest brewer proposed the acquisition
of AB InBev, the world’s largest brewer and therefore creating a global market leader. At a global level the merged entity will sell twice as much beer, earn four times more profit
than Heineken, (currently the third largest brewer), five times more beer and twelve times more profit than Carlsberg (currently the fourth largest brewer). Such merger raised many
concerns about the impacts on consumers and competition. The European Commission suggested that as result of the merger, even a relative small price increase could cause
considerable harm to consumers. (EuropeanCommission, 2016).
The European Commission and other competition authorities (including the
Ecuado-rian) have already approved the merger subject to remedies. In most of these legislations, competition concerns were found with respect to the merger and therefore, the approval
was conditional on AB InBev selling practically the entire SABMiller beer business in those countries.
This thesis focuses on a competitive assessment of the mentioned merger in Ecuador. I will reassess the merger and conclude if the price effect is large enough to impose structural
remedies. In Ecuador, competition law is a new field. Ecuador finally enacted its first domestic competition law on October 13, 2011. That law’s authority was appointed in
September 2012. With less than 5 years of experience in competition policy, Ecuador is immersed in the first stages of implementing competition law and the economic analyses
behind it.
The Ecuadorian beer market has an ownership structure where, SAB Miller (through
Cervecera Nacional and Dinadec S.A.) holds the strongest market position with approxi-mately 90% of market share. AB Inbev (through Ambev) is the second important firm in the brewing industry with nearly 6% of market share. The difference of the market share is
distributed over other small breweries. Therefore, this transaction may have led to higher beer prices and narrow the market structure even more (Euromonitor, 2015).
Is the effect of the merger between Anheuser-Busch InBev and SAB Miller on the price of beer in Ecuador large enough to impose important structural remedies?
This thesis focuses on the effect of the proposed merger on the price of beer in Ecuador. The proposition is that since the beer market is already highly concentrated by the merging
firms, the effect on prices might be severe. Therefore, structural remedies would need to be imposed.
The structure of this thesis is set out as followed. Chapter 2 provides an introduction to the Ecuadorian beer market. How it operates is essential to understand the dynamic
of a market (demand, supply, products, brands, distribution channels, etc.). The market ownership structure before the merger and the Ecuadorian firms involved in the operation
are also explained. This overview is meant to help the understanding of the methodology and the demand estimation itself.
In chapter 3, an outlay of the most relevant literature is provided. The thesis is po-sitioned in the field of competition and merger simulation and the main theories existing
in this field are explained. The main goal of this research is to contribute to the existing literature by observing consumer patterns in Ecuador. A merger simulation in the beer
industry is introduced which has never been assessed in Ecuador.
Chapter 4 describes the methodology used in this research. As mentioned earlier, the
main purpose of this thesis is the assessment of a merger. The thesis follows two methods to assess the effects of the merger. First, and probably the most important step is the demand
estimation. Since company revenues depend on the preferences of consumers, demand is a central element in shaping market outcomes. There are numerous methods of demand
estimation, nonetheless, the methodology applied relies on the nested logit model, which has frequently been applied by researches such as McFadden(1978), Verboven (1996) and
Slade (2004). The preferences of consumers seem to be more in line with the U.S. beer market rather than the European. Therefore, the nesting structure of Hausman, Leonard
and Zona (1994) is used: Non-premium, premium, Light and Imported. The second step to assess a merger simulation is to assume a multi-product Bertrand-Nash equilibrium and
constant marginal costs in order to compute the products current profit and predict the price variation after the merger.
Chapter 5 presents the description of the data. The sample covers a period from 2007 to 2015. A number of essential beer characteristics has been collected and constructed into
a number of variables that will be explained.
In chapter 6 the results of the analysis are presented. The analysis is split into two parts.
First, the results of the demand estimation, showing the own and cross-price elasticities. Second, the results of the merger simulation and different scenarios.
Chapter 7 covers a short overview of competition law in Ecuador and the decision that was taken by the competition authority (Superintendency for Market Power Control) concerning the merger case.
Lastly, chapter 8 shows the conclusion drawn from this analysis and provides readers with possible extensions to this analysis. It is found that the merger will indeed increase
the price of beer in Ecuador. In the Appendix additional regression results are shown and tests made for the validity of the model.
2
Beer market in Ecuador
This section presents a brief description of the beer market in Ecuador. This includes
the ownership structure, evolution of demand, market share per beer segment and popular brands. The data source comes from the three biggest supermarket chains in Ecuador and
consists of data from 2007 to 2015 of all beer brands sold. Even though independent small grocery shops are the main supply source for beer, the supermarket data set seems to be
appropriate since it contains information of commercialized brands and allow to understand the demand structure.
In Ecuador, the ownership structure (before the merger of SABMiller and AB InBev) is as followed:
Figure 1: Ownership structure SABMiller - ABInBev
SAB Miller plc. was a multinational brewing and beverage company that was acquired by Anheuser-Busch InBev in October 2016. Prior to the date of the acquisition, it was
the worlds second largest brewer. In Ecuador it owned three different firms through its subsidiary SABMiller Southern Investments Ltd.: Cervecera Nacional CN S.A., Dinadec
S.A. and Cernyt S.A..
Cervecera Nacional CN S.A. (Cervecera Nacional) is the primary beer producer in
Ecuador and the owner of Pilsener, the market leader and major national brand. Dinadec S.A. is a distributor of beverages and is mainly in charge of the distribution of Cervecera
Nacional products. Cernyt S.A. used to be involved in the rice production, however it was dissolved in 2014 and is not an existing firm anymore.
Anheuser-Busch InBev (AB InBev) is a multinational beverage and brewing company. It is the largest brewer in the world and is considered one of the largest fast-moving consumer
goods companies worldwide. AB InBev wholly owns Compaa Cervecera Ambev Ecuador S.A. (Ambev) through Monthiers S.A. and Freeville Management Ltd. (two of their
inter-national subsidiaries). Ambev is the second largest beer producer in Ecuador. Cervecera Nacional is the market leader with a market share of 92% in 2014, followed by Ambev
with 5%, according to Euromonitor (2015). SAB Miller plc. was a multinational brew-ing and beverage company that was acquired by Anheuser-Busch InBev in October 2016.
Prior to the date of the acquisition, it was the worlds second largest brewer. In Ecuador it owned three different firms through its subsidiary SABMiller Southern Investments Ltd.:
Cervecera Nacional CN S.A., Dinadec S.A. and Cernyt S.A..
Cervecera Nacional CN S.A. (Cervecera Nacional) is the primary beer producer in
Ecuador and the owner of Pilsener, the market leader and major national brand. Dinadec S.A. is a distributor of beverages and is mainly in charge of the distribution of Cervecera
Nacional products. Cernyt S.A. used to be involved in the rice production, however it was dissolved in 2014 and is not an existing firm anymore.
Anheuser-Busch InBev (AB InBev) is a multinational beverage and brewing company. It is the largest brewer in the world and is considered one of the largest fast-moving consumer goods companies worldwide. AB InBev wholly owns Compaa Cervecera Ambev Ecuador
S.A. (Ambev) through Monthiers S.A. and Freeville Management Ltd. (two of their inter-national subsidiaries). Ambev is the second largest beer producer in Ecuador. Cervecera
Nacional is the market leader with a market share of 92% in 2014, followed by Ambev with 5%, according to Euromonitor (2015).
The following graph displays the evolution of sold beer in supermarket chains (off-premise)1over the period of the data sample (2007 - 2015). A positive and increasing trend
characterize this market. Cervecera Nacional (SAB Miller) appears as the market leader in all of the period.
According to the market analysis of beer in Ecuador made by Euromonitor International 1Off-premise sale of beer covers the case of beer sale that is not followed by immediate consumption
in the place of sale. Alcohol monopoly stores, supermarkets, grocery stores, kiosks and gasoline stations or other similar shops are examples for this type of sale.
On-premise sale of beer to consumers happens when the product is consumed in the place of sale. This would include the sale of beer in restaurants, bars, cafs and other similar places.
In Ecuador, approximately 75% of beer is purchased in the off-premise channel. Off-premise is the largest channel, in which independent small groceries account for the majority of volume sales as they are present all over the country. On-premise outlets mainly offer higher quality beer. Consumers are more willing to consume it because of the relatively lower prices per bottle compared to other categories in alcoholic drinks like wine (Euromonitor, 2015).
Figure 2: Evolution of quantity sold per year
(2015), in 2014 there were two main breweries: Cervecera Nacional and Ambev. However, there were also at least 14 microbreweries producing craft beer. The number of
micro-breweries has increased in the past 10 years as higher income consumers prefer the more specialized craft beer.
There are fifteen suppliers in the sample. However, not all of them are present every year. For instance, in 2014 there were 9 suppliers and in 2015 there were 13 suppliers.
The number of beer brands owned per supplier and the market share is presented in table 1 for 2015. Cervecera Nacional (SABMiller) offered 9 brands in the Ecuadorian market,
representing 20% of the total number of beers. Ambev (AB InBev) offered 5 beer brands, representing 11% of the brands. When revising the quantity sold per supplier, the results
change. These results show Cervecera Nacional (SABMiller) as the market leader, with 90% of the total quantity sold. Following is Ambev with 7.47%. Furthermore, the
Table 1: Beer brands and Quantity sold per supplier (2015)
Notes: The table shows the quantity sold and the number of brands owned per supplier. It covers 2015.
Changes in consumer preferences are also visible in the beer market. The segmentation
for the beer is based in Hausman, Leonard & Zona (1994). Brands are classified into four types of beer: Premium, Non-Premium, Light and Imported. Marketing and difference in
price are taken into consideration to segment each brand.
Imported beers are all brand of beers that have been imported either for breweries
established in Ecuador (Cervecera Nacional or Ambev) or by importers. In general, the price of imported beers, are easily differentiated by the consumer because the price is higher
than the rest of beers.
Light beers are are characterized as having less calories and alcoholic content. In
Ecuador, this category has just recently become popular for the consumers and that is why there is a small number of beers that are offered in this segment. According to
Cerve-cera Nacional, Pilsener Light, which is one of their brands in this segment is considered as a value generator beer, rather than a sales volume generator.
Non-premium beers are the least expensive beers in the market. Generally, this type of beers is mostly offered in small stores (tiendas de barrio). Supermarkets offer only a few
Premium segment are beers brewed in Ecuador and are supposed to satisfy the majority of the consumers in Ecuador. One of the important differences between premium and
non-premium beers, besides the price, is the quality. In general, premium beers contain some special ingredients that tries to make each beer different from one another (package
presentation, additional ingredients, brand).
The following graph shows how the consumption per segment has varied over time.
The premium and light segments show an increase in the consumption over the last years. Regarding the light segment, it is predictable that might continue increasing, since a healthy
life-style is becoming more popular and low-calorie products are preferred. The opposite holds for the remaining segments: Non-premium and Imported beer. The non-premium segment shows a cut down in consumption over the years. This might be the case on
supermarket chains, however if we collect data of small stores (tiendas de barrio), the non-premium segment might be the most popular one. Imported beer segment2 also show a
decrease in consumption. The reason is that imported brands constantly suffer from import taxes and import barriers, which make these beers more expensive and are consequently
not considered substitutes of the Ecuadorian produced beers (Euromonitor, 2015) 3 2Consumers say that the most important reason to choose small stores or neighborhood stores
(tien-das de barrio) as their first choice to purchase alcoholic beverages (including beer) is the closeness to the stores. Specifically in rural areas, it is easier and cheaper to find a small store rather than a super-market. http://investiga.ide.edu.ec/index.php/revista-agosto-2004/836-el-consumo-en-ecuador-indicadores-exclusivos
3
Figure 3: Share of Segments
In the subsequent table the average price and maximum price of beer (300-355ml) per
segment is shown for 2015. As it can be seen, there is an important difference in the prices between the segments.
Table 2: Average price per segment (2015)
Notes: The table shows the average and maximum price paid per segment.
Each beer segment will be considered as a different market, since the substitutability
between them (as it will be seen) are relatively small. Graph 4, shows the market share of Cervecera Nacional (SAB Miller) and Ambev (AB InBev) in each of the segments for the whole period (2007-2015). In the premium segment, Cervecera Nacional is the leader,
with 94,5% of the market share. Ambev possesses 5,4%, meaning that both firms hold in total 99,9% of the market share. In the Non-Premium and light segment, the two firms are
the only competitors. In the Non-premium segment, Ambev is the leader holding 92,5% of the market share and Cervecera Nacional possesses the rest (7,5%). In the Light segments,
Imported segment Cervecera Nacional holds a market share of 11,5% and Ambev 0,12%. Almacenes Juan El Juri is the market leader with almost 80% of the market share.
Figure 4: Market sharer per segments
Table 3 presents the most popular brands sold on the market. Club and Pilsener are
considered the local beers and are the most popular and commercialized brands in Ecuador (on-premise and off-premise). Brand-wise, Club is a stable leader with a market share
around 56% which doubles the market share of the second leader Pilsener. The market becomes even more concentrated and asymmetric from the company level perspective. For
instance, in table 3 only Heineken is supplied by a different company. (Almacenes Juan El Juri Cia. Ltda.)
3
Literature Review
Microeconomic theory stipulates that competition can be harmed by high industry concen-tration. A merger leads firms to have a unilateral incentive to increase prices and reduce
quantity (Salant, Switzer, & Reynolds, 1983). Moreover, the merger can induce firms to collude. On the other side, the merger produces synergies which allow the firms to reduce
Table 3: 10 popular brands (2015).
Notes: The table presents the most popular brands sold on the supermarket chains.
surplus is larger than the fall in consumer surplus due to higher prices. (Williamson, 1968). Mergers in concentrated industries are an important issue for competition authorities.
Au-thorities may be averse to mergers in industries where concentration is already high, as the mergers may substantially increase firms abilities to exercise market power to the detriment
of consumers.
Merger simulation models have been employed both by antitrust authorities, merging
companies and by courts to assess the pro or anticompetitive character of the proposed mergers. Despite some setbacks, it seems likely that the merger policy tool simulation
models will play a more important role in the future (Budzinski & Rukmer, 2009). Merger simulations are useful therefore to quantify and disentangle the impacts a merger is likely
to have on concentration and efficiencies. Impacts of proposed merger remedies can be estimated likewise.
Because the various effects of mergers depend considerably on the characteristics of each market, Budzinski and Ruhner (2009) describe the general process followed in a Merger
Simulation exercise:
1. Estimation of the demand. Frequently proposed models are linear, log-linear, logit, nested logit, Almost Ideal Demand System (AIDS) and multi stage demand. On the
basis of the assumed demand function, cross prices and own price elasticities can be estimated as well.
2. Calibration of the demand system. The parameters are specified so that elasticities yield the prices and market shares in the pre-merger market.
3. The supply side is modelled by assuming an oligopoly model. In most cases Bertrand
competition is assumed because it allows inferring marginal costs.
4. A new equilibrium after the merger can be simulated using the pre-merger empirical
data and adjusting market shares to the post-merger situation.
The econometric literature on merger analysis has grown over the last decade. In
partic-ular, (AIDS), logit and nested logit models have been extensively used to analyze markets with differentiated consumer products and their merger simulation.
The Nested Logit Model was first motivated by McFadden, Kirshner, & Puig (1978) by assuming consumers to undertake a two-stage decision purchasing a good. In the first stage
the consumers decide which group of goods to buy and in the second stage they choose between goods within that group. The main assumption is that the groups are mutually
exclusive and an exhaustive collection of products.
Hausman, Leonard and Zona (1994) also calculate the price effects of beer mergers in
the U.S. They assume a three-multilevel stage, the almost ideal demand system (AIDS) and estimate the elasticities, assuming that costs across cities do not vary. The basic idea
of the system is to have the top level correspond to overall demand for the product (beer). The middle level corresponds to different segments for the products: premium beer, light
(low calorie) beer, imported beer and non-premium (popular) beer. The bottom level of the demand system corresponds to competition among brands in a given segment. The
estimates from each level are used to estimate the overall own and cross price elasticities for each brand.
A different approach was then taken by Berry (1994) and Berry, Levinson, and Pakes (1995) in analyzing the automobile market. Consumers are assumed to base their purchase
choices on observables product characteristics. These characteristics are supposed to be independent of demand shocks and therefore a given products equilibrium price is affected
by variation in the characteristics of competing products.
Werden and Froeb (1994) used the logit model to simulate a merger in the U.S. long
distance carriers industry. They assumed Nash equilibrium in prices and constant marginal cost. They then explored the effects of mergers in the mentioned market and found that
mergers involving AT&T would have lessened welfare.
Verboven (1996) used the nested logit model in analyzing the car market and divided
the set of cars into different groups and subgroups. The nest parameter(s) control(s) how close the goods in a nest are substitutes for each other and how close they all are substitutes
for goods outside the nest. Verboven divided the cars into classes and subgroups according to the country of origin.
In more recent studies, Nevo (2000) estimates a random coefficients model to study the
effects of mergers in the U.S. ready-to-eat cereal industry. His strategy is to model demand as a function of product characteristics, heterogeneous consumer preferences and unknown
parameters by the econometrician. The estimation of the structural parameters that govern demand and supply are used to simulate the post-merger equilibrium.
Slade (2004) analyzes the market power and joint dominance in U.K. brewing industry. Once again, she used the nested logit model where partitions n brands of beer into G groups,
and an outside good. The partition is chosen so that alike products are in the same group. The groups that were considered are lagers, stouts, keg ales, and real ales. Besides the
nested logit approach, Slade also study the merger using a distance metric approach. Bjornerstedt and Verboven (2013) implemented a merger simulation after estimating an
aggregate nested logit demand system in the European car market. The merger simulation in their empirical analysis aims to predict the price effects of a merger after estimating
the demand system assuming the behavior of the firms as Bertrand-Nash. Assuming this behavior allows computing the products current profit margins and their implied marginal
costs. Finally, assuming constant marginal costs, the authors predict how prices will change after the merger.
very well. It is still gaining popularity among researches over the world as this field has reached a high degree of complexity and requires refined econometric techniques (Giacomo,
2004). Budzinski and Ruhner (2009) described the commonly used econometric techniques in merger simulation analysis as well as their limitations.
Linear and log-linear demand models are considered the simplest but however unrealistic demand models. The predictions of the elasticities of demand are assumed constant and as
a result these models have received much criticism. It seems unrealistic that when prices and market shares change, the demand elasticities do not. (Budzinski & Rukmer, 2009)
At the moment, discrete choice demand models seem more accurate of estimating the demand. The general idea is that the consumer utility depends on observable and unob-servable product characteristics. An outside good is usually considered to represent the
option of not purchasing the product. The logit model and nested logit model are the main examples of discrete choice demand models. However, the logit model presents a
limita-tion. The logit specification assumes the Independence of Irrelevant Alternatives (IIA), which implies that consumers switch to other products in reaction to a price increase for
one product in proportion to the relative shares of these products. Nevertheless, it seems to violate the core characteristics of heterogeneous markets, mainly because consumers do
not consider all products as equally substitutable. The nested logit model on the other hand overcomes the problem of the IIA. It accounts different levels of substitutability by
separating them into an identically and independently distributed nest-specific component, which is allowed to be correlated across products that are grouped into one of the multiple
predetermined exhaustive and exclusive nests. Consequently, the nested logit allows a more realistic demand estimation. (Budzinski & Rukmer, 2009)
Finally, AIDS model specifications consider prices, quantities and market shares in levels i.e. the whole market, market segments and individual products at the lowest level. However
a problem might arise. Underlying theory of individuals assumes no corner solution, which might not be true for specific products. (Budzinski & Rukmer, 2009).
be applied in this thesis.
4
Methodology
The basic idea to assess the effects of a merger is to estimate the demand of a product and
to simulate a merger to be able to observe changes in prices ex-post.
Berry, Levinsohn, & Pakes (1995) studied the demand analysis in differentiated products,
in which is assumed that the demand can be described by a discrete-choice model and that prices are endogenously determined by price-setting firms. The utility of consumers depends
on product characteristics and individual taste parameters. The utility of consumer i for product j depends on the characteristics of the product and the consumer taste.
U (xj, ξj, vi, θd)
The terms in the equation are observed and unobserved (by the econometrician) product characteristics, price and demand parameters respectively. The term v captures specific
terms that are not observed by the econometrician.
With this specification, the utility of consumer i for product j is given by
uij = xjβi− αpj+ ξj + εij
where the consumer specific taste parameters are β1 and εij (observed and unobserved
characteristics for the econometrician). ξj, is the mean of consumers valuations of an
unobserved product characteristic such as the quality of the product, and is unobserved. In the following equation,δj, is the mean utility to all consumers and εij, is the
individual-specific utility term. The mean utility δj, depends on xj product characteristics, on pj, the
price of product j , ξj, unobserved valuation. And is given by
δj = xjβi− αpj+ ξj
As for now, the consumer has two options, purchasing the product or not. When consumers decide not to purchase the product (beer), they purchase an outside good (0)
for which the mean part of the indirect utility δ0} is 0, so that uio= εi0. For this thesis,
the outside good will be alcoholic beverages that are not beer (whiskey, wine, rum, etc.)
According to Encuesta de Ingresos y Gastos del Hogar (Censos(INEC), 2014), from the population that consumes alcoholic beverages, 79% prefers beer. Therefore, the demand for
the outside good will be equal to 21% of the total alcoholic demand in Ecuador. Regarding the total alcoholic demand in Ecuador, I collected demographic information per year, the
age range considered is between 15 to 64 years old.4
In the demand specification, δj is the dependent variable - the natural logarithm of
the brands overall market share where the market includes the outside good. To define
the potential beer market size, I specify it as the beer demand (79% of alcoholic demand) multiplied by twelve, a crude proxy for the number of beers the consumers purchase.5
The distribution of the utility will follow the assumptions of a one-level nested logit model. The market of product j can be divided into G different groups. Therefore, each
subgroup h contains hg products.
J =
G
X
g = 1Jg
The intuition behind the subdivision of the product J, is that consumers may have correlated preferences across all products of the same group and (no stronger) correlated
preferences across all products of the same group but a different subgroup. Marketing 4The reason of not considering population above 18 years old in Ecuador (legal drinking age) is mainly
because this information was not available. The Word Bank data base (which is my source) has the following age ranges information: 0-14; 15 -64; 64 and above. Even though the legal drinking age in Ecuador is 18, it might be important to mention that there is a lenient attitude towards underage alcohol consumption. (drinkingmap, n.d)
5
(Bjornerstedt & Verboven, 2013) To specify the potential demand of a market, it is common to use common proxies. For instance, in the European car market, the population is divided by four as an approxi-mate number of households. In the beer market, the population is multiplied by twelve, assuming that each consumer purchases at least 12 units of beer.
classifications normally partition the markets in different subgroups. In the beer market, Hausman, Leonard and Zona (1994), considered the following segments: premium beer, light
(low calorie) beer, imported beer and non-premium (popular) beer. Slade (2004) considered a different segmentation for the UK beer market: lagers, stouts, keg ales, and real ales. In
general, products in different nests are supposed to have less in common, and therefore are not good substitutes. In this thesis, the nesting classification will be: premium,
non-premium, low calories and imported, since it is more in line with the Ecuadorian consumer preferences.
With the nested logit model, the IIA property relaxes by using nests. The IIA implies that for a specific individual, the ratio of the choice probabilities of any alternatives is entirely unaffected by the presence or absence of any other alternatives in the choice set and
by their systematic utilities. The four nests that will be considered in this thesis (premium, non-premium, imported and light beer) are independent and are not close substitutes. Even
though, imported beers share some attributes with the other alternatives, at least in the Ecuadorian market they are not close substitutes and will never affect the choice of the
other nests (IIA holds). The reason is mainly because high import taxes and barriers make the imported beers highly expensive compared to the rest of the alternatives.
Finally, the demand equation will look like follows:
ln(sjt) − ln(s0t) = δjt = xjtβ − αpjt+ σhln(sj/ht) + ξj
where σh is the nested logit coefficient associated to the four nests; and which measure
the degree of correlation of consumer preferences for the product belonging to the same subgroups. sj/ht, is the market share of product j in its subgroup h ˙Finally, xjtβ are the
beer characteristics: alcohol content, calories, IBU, style, color and serving way. (this characteristics will be described in the following chapters)
In most cases, the error term may be correlated with price and market shares, so that instrumental variables should be used. (Bjornerstedt & Verboven, 2013) Good instruments
must explain variation in price given the variation already explained by the included exoge-nous variable and be uncorrelated with unobserved determinants of demand. For instance,
determinants of supply that has no role on the demand side can play a suitable role as in-struments, since shifts in the supply side will identify or trace out the demand curve. Cost
shifters would qualify as instruments, but these are typically not available at the product level. Berry, Levinsohn and Pakes (1995), suggest to use sums of the other products
char-acteristics (over the firm and the entire market). Verboven (1996) adds sums of the other product characteristics by subgroup and group. I will also use rival product characteristics
as instruments.
The instrumental variables used are explained below:
IV between alcohol: The average of alcoholic content of competing brands of beer be-longing to the same segment and across the different segments. IV between IBU: The average of IBU of competing brands of beer belonging to the same segment and across the
different segments. IV between calories: The average of calories of competing brands of beer belonging to the same segment and across the different segments.
Lastly the price elasticity of demand is as follows and reflects the percentage change in market share of brands with respect to a change in price.
nij = (∂qi/∂pj)(pj/qi)
The own price elasticity
nii= αipi " si+ 1 1 − σg + σg 1 − σg si/g #
and the cross-price elasticities, 1) If j 6= i and j ∈ g nij = αjpj " sj+ σg 1 − σg sj/g # 2) If j 6= i and j /∈ g nij = αjpjsj
After estimating the demand system, I will implement a merger simulation in Stata, following Bjornerstedt and Verboven (2013) paper on Merger Simulation with Nested Logit
Demand- Implementation using Stata. The merger simulation results will depend crucially on the parameters: α, σg, price and quantity data per product.
5
Data
The data 6 set was provided by the three biggest supermarket chains in Ecuador and con-sists of data from 2007 to 2015 of all beer brands sold in these supermarkets in Ecuador.
Even though independent small grocery shops are the main supply source for beer, this information is clearly impossible to obtain, primarily because most of the stores are
infor-mal and sinfor-mall familiar businesses without publicly accessible sales data. Nevertheless, the supermarket data set seems to be appropriate since it contains information of all
commer-cialized brands, including beers owned by the merging firms, namely Ambev and Cervecera Nacional, imported beers and craft beers which represent the whole beer industry in the
country.
The raw dataset includes 407 observations. Each observation is a brand of beer sold
in each year and is differentiated by package size in ml. and price. The range of package size is from 207 ml to 5000 ml. Nonetheless, to estimate the demand and compare across
brands, beers between 300 ml to 355ml are considered. In 2015, beers between 300 ml. to 355 ml. had a market share of 76.35% .7
In addition to the price, quantity and brand, beer characteristics are introduced. In the following subsections, the variables used for the demand estimation are explicated.
5.1 Beer characteristics
The beer characteristics information was collected from RateBeer, a recognized source for
beer information. The beer characteristics included are alcohol content, calories, IBU (Bit-6Since the data was provided by the competition authority in Ecuador, the data and this thesis will be
confidential.
7
This shows that this specific package size is the most preferred for the consumers. This procedure reduced the sample.
terness range), style, color and way of serving.8 All of the beer characteristics are string variables.
ABV (Alcohol by Volume): Alcohol by volume is a measure of how much alcohol is contained in a given alcoholic beverage. Is expressed as a volume percent. In general, the
alcoholic content of beer is between 2% to 12%. The range of alcohol content in the data set goes from 0.5% to 10%. And the average of alcoholic content is 4,7.
Calories:Estimated calories for a12 fluid oz. a unit (ounce) serving. Calories range is between 15 and 300 in the data set. The average of contained calories is 140.
IBU (International Bittering Unit): IBU is a unit that quantifies the bitterness of beer. Light beers without much bitterness would generally have 8-20 IBU, while Indian Ale may have 40 IBU or more. In the data set of this thesis, beers range between 0 to 45 IBU.
Style: The style of a beer is simply a label given to a beer to describe its overall character and to categorize beers by some factors including appearance, flavor, ingredients, production
method or history. Some famous beer styles are: American Pale Ale, Golden Ale, Pilsener, Pale Lager, Stout or Witbier. In the Ecuadorian market, 19 styles of beers were accounted
for in the analyzed period and the pale lager style is rated as the favorite by the consumer. Color: Color of the beer is another important characteristic, especially when it comes
to not experienced or non-beer fanatics kind of consumers. The Lovibond Degree is the scale that measures the color of beer. In Ecuador, there are accounted 5 categories. Golden
is the favorite option.
Serving way: refers to the appropriate glass for each beer which will accentuate the
characteristics of each beer. There are 7 groups of glasses and the Dimpled mug, Lager glass and shaker are the most preferred.
The following table indicates the pairwise relationship between some characteristics and the price. All of the characteristics are positively related with the price. Alcohol and calories
are positively and strongly related, as expected. Beer containing a higher ABV, contains more calories.
8
Table 4: Correlation table
Notes: The table shows the correlation between the price and alcoholic content, calories and IBU.
5.2 Variables
Table 5 gives summary statistics of the data and the overall state of variables before any
regression analysis is performed. Meaning and full description, including units of measure, are given in the tables footnote.9
9
The name of quantity in the data set can be found as QUAN1. The minimum quantity sold is 1 unit of beer and the maximum amount is 19.400.000 sold beers. Price can be found in the data set as PRICE1and the average price is $1.46 with a $1.011 standard deviation. Brand and Supplier were used to distinguish each brand of beer in the different segments and to calculate market shares per segment and across segments. In the data base, brand is called BRAND1and Suppliers as Supplier1. The segments are divided in 4 categories (Premium, Non-Premium, Imported and Light) and the variable that describes the segments is called Nest1. ABV (alcohol by volume) is the alcoholic content characteristic. The maximum alcoholic content in the data is 10%. This variable is called alcohol in the data. The average calories contained per beer is 140 with a standard deviation of 31 calories. IBU shoes the bitterness of the beer, and shows a maximum of 45 points. The beer characteristics style, color and serving way can be found in the data set as style2, color1and servedin1. This variables are categorical variables, there are 19 styles of beer, 5 colors and 7 serving ways in the data set. An interaction of the price with ABV was also created and is called prialc. Dummy of style is a dummy variable indicating whether the beer is classified as ales/pale or a different category. In the data set, this variable is distinguished as Ales. Finally, dummy of ABV is a dummy variable that takes a value of 1 if the ABV is less than 150, or 0 otherwise. The name of the variable is highalcohol2.
Table 5: Summary Statistics
Notes: The table shows summary statistics. Data covers the period from 2007 to 2015. Price: Unit price expressed in USD. The price of beer varies dramatically, the lowest price in the data is 0.38 USD and the highest is 7.53USD. The difference of price mainly depends on branding and whether the beers are imported or produced in Ecuador. Quantity: The variable is the sum of units of beer that were sold in a specific time period, brand and supplier. Brand: String variable that shows the brands of beer commercialized per period. For instance, in 2015 there were 64 brands commercialized. Supplier: String variable that indicates the supplier of each brand of beer. In the data set 15 suppliers were reported in the analyzed period. Segments: Variable that shows 4 different segments of beer: Premium, Non-Premium, Imported, Light. In the data set the variable is reported as nest1 Price interaction: Interaction of the price with the alcoholic content. Denotes price times alcohol content in each year. Dummy for Style: The string variable of style is converted into a dichotomous variable that equals 1 if the style of the beer is either ale or pale, and 0 otherwise. Dummy for ABV: The string variable of ABV is converted into a dummy variable, which takes the value 1 for brands whose alcohol content is greater than 5Dummy for Color: For the string variable color, a dummy variable assigned a value of 1 for brands whose color is categorized as golden or blonde, and 0 otherwise. (not used in the final estimation as it was not significant) Dummy for Calories: The string variable of calories is converted into a dichotomous variable where brands of beer whose calories are less than 150 are considered low calorie and have a value of 1, while brands of beer with more calories have a value of 0. (not used in the final estimation as it was not significant) Dummy for IBU: The string variable of IBU is converted into a dummy variable with value1 if the level of bitterness is less than 30 and 0 otherwise. (not used in the final estimation as it was not significant)
6
Empirical Results
6.1 Nested Logit Model
In this section, the econometric specification of the Nested Logit empirical model is provided.
The aim of the model is to determine the demand of the beer industry and with it the own-price elasticities and cross-own-price elasticities. The dependent variable is the natural logarithm
of the brands overall market share (δjt), σh is the nested logit coefficient associated to the
four segment and measure the degree of correlation between the groups. sj/ht, is the market share of product j in its subgroup h. α, is the price coefficient.
As it was mentioned before, a key issue that has been addressed is that of endogeneity.
Since prices are choice variables for firms, it is likely that they will respond to components of demand that are unobserved to the econometrician, therefore price is considered endogenous.
This also happens with the market share of a product in each subgroup (σh), which is also
considered endogenous. In particular, one needs instruments that vary by brand. The
exogenous demand and cost variables are obvious candidates. However, cost information was not available. Therefore, rival-product characteristics are used as instruments (alcoholic
content, calories and IBU) and includes the average of the characteristic of competing brands of beers belonging to the same segment and across the different segments. The economic
motivation for these instruments comes from Berry, Levinsohn, & Pakes (1995).
Applying a two-stage least square estimator, endogeneity issues are tested and the
va-lidity of the instruments are verified. Wooldridges score test reject the null hypothesis that price and lSjgt ln(sj/ht) are exogenous at conventional levels (p=0.000). Here, both test
statistics are highly significant and as a result the null hypothesis of exogeneity is rejected
so price and lSjgt ln(sj/ht) are treated as endogenous variables as it was already expected.
In addition to the condition that instrumental variables has to be correlated with the
endogenous regressors, the instruments must also be uncorrelated with the structural error term. If the model is over identified, meaning that the number of additional instruments
are uncorrelated with the error term. The test of over identifying restriction shows a Score chi2 of 0.0265 at a 5% significance level, so the null hypothesis that instruments are valid is not rejected. Finally, Sheas partial R2 , indicate that instruments are not weak, at a 5% significance level. 10
After identifying and confirmed the endogeneity problem and found valid and non-weak instruments, a number of specifications are estimated and are shown in Table 4. All
specifications contain the dummy for ABV, serving way, color, dummy for style and the price interaction with ABV. The coefficients of the mentioned variables were always significant
at conventional levels. The variable IBU and the dummy variables for calories, color and IBU were eliminated from the model since none of them were significant. Conceivably, these characteristics are not meaningful for the Ecuadorian consumer it the moment of purchasing
a beer.
When deciding which model to employ for further analysis, two main factors are taken
into consideration. First, the coefficients of interest and must be significant and second, their magnitude should be coherent to economic logic and should correspond to the
find-ings in the relevant literature. The specification (IV-5) seems to be more in line with the economic intuition and shows better estimates. The final equation looks like followed:
δjt = −αpricejt+σhln(sj/gt)+β1Dummyf orABV +β2priceinteractionABV +β3servingway+
β4color + β5Dummyf orstyle + ξ
10
Table 6: Nested logit Demand Equations
Note: The demand specifications try to use all beer characteristics collected. Regressions IV-1 to IV-6 shows different results using different variables. Dummy for ABV and the price interaction are highly significant and appears in all the specifications. Dummy for calories is not significant and was not considered in the final specification (IV-5). The strings variables for serving way and color are highly significant and are considered in all the specifications. There were considered two dummies for style, Dummy for style (which consider the categories ales and pale) and Dummy for style2 (which consider only the category pale). Dummy for style 2 was not significant and eliminated from the specification while Dummy for style is significant and included in the IV-5 specification. Finally Dummy for IBU and IBU were eliminated since they were not significant.
First, the coefficient of price is highly significant and negative. It means that price has a negative effect on consumer mean valuation. Second, when the same specification is
estimated by OLS, the slope is more negative with the IV estimation as one would expect. Nevertheless, due to the presence of the segment-share variables, lSjgt, demand is more
elastic with the latter. The coefficient of lSjgt is positive, less than one, highly significant and assumes a magnitude that implies within-group correlation of tastes. Both coefficients,
are significant and the results are congruent with economic theory and expectation. Regarding the alcoholic content dummy variable, it shows a negative relationship with
beer demand, showing that demand decreases as beers contain a higher alcoholic content. This fact, is confirmed with the interaction of the price with ABV, which shows a positive relation between price and alcoholic content.
Now, that the specification of the model is correct, the own and cross price elasticities are calculated. Brand own-price elasticities are calculated holding the prices of all other brands
constant. When compared to Slade (2004), where the mean own-price elasticity is 3.6, (Slade, 2004), she concludes that the logit-demand specification is not very satisfactory for
this application, since it is too low. Differently, estimates reported in Hausman, Leonard and Zona (1994), where own-price elasticities for brands averaged -5.0, the own price elasticities
seems to be more accurate. For the Ecuadorian beer market, the mean own-price elasticity is -13.63 in 2015. This result shows that the beer market is highly elastic and therefore
agrees with the theory that demand for individual brands should be highly elastic, since there are many close substitutes for a given brand. (Slade, 2004).
Table 7, present an overview of own and cross price elasticities averaged on a segment basis for 2015.
Overall, 1% price increase causes a decrease in demand of 13%. The within-segment cross price elasticities are estimated reasonably and in accordance with economic expectations.
For instance, a percentage increase of price of brand j in the segment Premium, on average increases the demand of another brand in the same segment by 0.122%. An outlier of
Table 7: Elasticities(2015)
Note: The table present an overview of non-weighted own price elasticities and cross price elasticities. It is evident that group elasticities are higher than when comparing outside groups.
segment are commercialized in supermarket chains, consumers perceive the brands as highly
substitutable and sensitive to each-other price changes, and therefore high within-segment cross price elasticity is determined. The cross price elasticities between brands in different segments is low and it averages 0.023. This implies that brands of beer that are categorize
in different segments are independent and not substitutes.
To analyze the different substitution patterns, some beer brands were selected covering
all the different segments.
Table 8: Brands and Segments
Regarding the supply side, table 10 presents average implied marginal costs and the Lerner index for 2015. First, Ambev and Cervecera Nacional are the companies with lower
marginal cost, each with USD 0.607 and USD 0.578 respectively. The Lerner Index deter-mines the market power of a company. In case of having a value of 0, firms marginal cost
Table 9: Own and cross price elasticities
The table presents the elasticities for the selected brands. When comparing different cross-price elasticities, we can observe that the cross-price elasticity of two products within a segment is higher than the cross-price elasticity of two products that do not belong to the same segment. For instance, the cross elasticity of the beer (14) Budweiser Light is equal to 0.105 when compared to (48) Pilsener Light (both beers in the Light segment). While when it is compared to beers from different nests, the elasticity is lower. This is also the case of the other beers. In the imported segment, Kunstmann (33), shows a cross price elasticity of 0.1448 with Corona (25). But when it is compared with Paramo Golden (41) which is a Premium beer, the value decreases to 0.001.
and prices are the same and the firm does not have any market power. On the other hand,
if the Lerner Index equals 1, it corresponds to a monopolistic environment where the firm possesses market power. In this case, the mark ups for Ambev and Cervecera Nacional
are the highest, implying they have some degree of market power, compared to the other competitors in the market.
Table 10: Supply side for 2015
6.2 Merger Simulation
Now that the demand system is estimated, the merger simulation will be completed consid-ering an oligopoly model with multi-product price setting firms who have constant marginal
cost and that can be estimated with a unit demand specification. (Bjornerstedt & Verboven, 2013)
First, the merger simulation was initialized, where a one-level nested logit model is specified, and the groups are the beer segments. The price, quantity, market size and
firm variables are specified. The nested logit model was estimated (as in the previous section) with a two-stage least square regression using instruments. The parameters that
will influence the merger simulation therefore are identical; α= -3.25 and σ = 0.47.
The merger simulation, considers Cervecera Nacional (SABMiller) selling its operations
to Ambev (AB InBev).
Before the results of the merger simulation are shown, the HHI pre-merger and post-merger are considered. In average, the pre post-merger HHI is 8.274 points and the post-post-merger
HHI 9.571. Implying a variation of 1.297 points. Additionally, the producer surplus is 324.268 while the consumer surplus is -1.258.640. These results show that the effects of
the merger does not maximize consumer surplus, mainly because the prices might increase dramatically in the Ecuadorian market.
Table 11, shows the HHI for 2015 per segment. As it is noticeable, the four segments are highly concentrated. However, as a result of the merger, only the Premium segment
shows a high variation of the HHI.11 It is noteworthy that even though the other segments do not show a variation of HHI, these segments are dominated by the two merging firms.
11
Horizontal merger guidelines of the European Commission, suggest safe harbors for merger assessment. When the change in HHI is above 250, the market is considered as problematic and a more detailed analysis would need to be conducted.
Table 11: HHI per segments
Note: The table presents the HHI pre-merger and post-merger per segment. Also, the consumer surplus and Producer surplus were calculated.
6.2.1 Merger simulation without cost efficiencies.
First, it is assumed that there are no marginal cost savings to the seller or the buyer, and
that there is no partial coordination (neither before, nor after the merger).
Table 12, show prices before and after the merger and the percentage price change
averaged by firm. The simulation, predicts that Ambev S.A. will on average increase its price by 27% while Cervecera Nacional will on average raise its price by 2.3%. The rivals
show no response, perhaps by the low market share they possesses. The change in market share for Cervecera Nacional is positive, with a gain of 0.045%.
Table 12: Effects of the merger
Note: The results show prices and market share before and after the merger and the percentage price change averaged by firm. The data that is being analyzed consider supermarket chains which does not represent the whole beer market in Ecuador. However, the price increase after the merger will affect not only supermarket chains but also small groceries. The effect of the merger in the consumer surplus might be tougher.
Table 13 display the results of prices and market share before and after the merger, the different segments. The merger simulation predict that Ambev will on average raise
Premium segment. In the light segment, the price would increase in 17% by Ambev. The other segments did not show an important variation on prices or market shares.
Table 13: Effects of the merger per segments
Note: The table shows results in the premium segment. The merger would arise competition concerns in this segment.
6.2.2 Merger simulation with cost efficiencies.
Considering the current U.S law, a merger should seek to maximize aggregate surplus. A merger should be allowed if and only if the efficiencies are enough to ensure that price
does not increase. For this, it is accounted for the possibility that the merging firms would benefit from a marginal cost saving, that may be passed on into consumer prices.
Cost efficiencies often are the main statement that companies use to defend the merger in front of competition authorities. The merging parties claim that mergers will lead to a
more efficient operation. Although the efficiencies are only a claim and not yet guaranteed, this thesis will assume the merging firms incurs a marginal cost saving of 20%.
in some of the rivals products and shows a price decrease of 5.3% for Cervecera Nacional products. However, the price prediction of Ambev S.A. products would still increase in
17.8%. (which implies approximately a 10% decrease compared to the case where the merg-ing firms does not have efficiencies). Additionally, consumer surplus would be of 833.054,
since it is assumed that thee cost savings are passed on into consumer prices. Even though there is the assumption of cost savings, it is evident that the prices might increase as an
effect of the merger. For the prices to remain unchanged after the merger, the minimum required efficiency per product owned by the merging firms (14) should be on average 22%
(unweighted).
Table 14: Merger Simulation with cost savings
Note: The efficiencies that the merging firms would contribute to are unknown.
The efficiencies realized by the merging parties would affect the segments as well. The
effects of the merger with 20% reduction in marginal cost of SAB Miller and AB InBev are clearly a better scenario than the one without efficiencies.
First, a different conclusion about welfare effects is drawn. The consumer and producer surplus are positive in all segments, implying that the both consumers and producers would
benefit of the merger.
Table 15: Welfare effects with cost efficiencies
the efficiencies would be passed on consumers. In the Non-premium, Light and Imported
segments, the price decreases substantially (8,9%, 10,9% and 14,3% respectively). The Imported segment would experience a drop not only by Cervecera Nacional products but
also by some competitors. In the Premium segment however, the price would increase 14% for Ambev and decrease 5% for Cervecera Nacional.
Table 16: Effects of the merger with cost savings per segment
Note: The table provides the changes in price and market share after the merger, considering cost efficiencies of 20
In the final decision of the Ecuadorian competition authority, regarding this case, it is
mentioned that the efficiencies in this case do not meet the burden of proof to create an incentive for possible anticompetitive practices (Superintendency Resolution case Cerveceria
Nacional - Ambev, 2016)
7
Competition Overview
There are two potential effects in the economy, when the authorities approve a merger. The
merging companies can expand the market and bring benefits to the economy. For instance, companies may develop new products more efficiently or reduce production/ distribution
costs. Through their increased efficiency, consumers benefit from higher quality goods at fairer prices. However, some mergers strengthen the power of a dominant player and
are likely to harm consumers through higher prices, reduced choice or less innovation. For instance, harm to the ability to innovate of the merged entity’s rivals was probed in a number
of European merger cases including Intel/Mc Afee, ARM/Giesecke & Devrient/Gemalto Joint Venture, Telefonica UK/Vodafone UK/Everything Everywhere Joint Venture and
Intel/Altera. (European Commission)
As it was mentioned, the tendency in developing countries which are in the process of
implementing and enforcing competition rules has been to adopt either the U.S. antitrust scheme or the European Unions rules. In Ecuador, the law is a hybrid system between the
latter mentioned (Marn Tobar, 2013).
On October 13, 2011 Ecuador enacted the Organic Law for Market Power Regulation
and Control, whose regulations have been in force since May 7, 2012. The Superintendency for Market Power Control has been the designated authority to regulate competition issues,
such as abuse of dominant position, restrictive agreements and practices, unfair competition and merger and acquisition control.
Regarding to substantive matters, the Law generally follows international rules and principles. With respect to merger and acquisitions, the Superintendency follows the
guide-lines on the assessment of mergers and best practice guideguide-lines published by the European Commission.
According to the guidelines, market concentration is a strong driver of the estimated
markets and between 10% and 20% in concentrated markets. In general, if the merger analysis demonstrates an average price increase between 5% and 10%, the competition
authorities unconditionally approve the merger. If the price increase is higher and depending on competition characteristics of each market, the merger can be approved with remedies or
blocked. Remedies are therefore considered an important tool to mitigate the post-merger price increase. The review of merger decisions in the EU shows that on average, mergers
were followed by a price increase which remained under 5% in the majority of the cases. The average price increase in unconditionally approved mergers was under 5% and in remedies
mergers around 2%. (Peter Ormosi, Franco Mariuzzo, and Richard Havell, 2015)
As it was seen in the merger simulation, the merger will likely result in a drastic price increase and may potentially harm competition and its consumers. Therefore, the
compe-tition authority should decide whether to approve conditionally or block the merger. In Ecuador, even before the merger the beer market was already not competitive so it seems
logic to block the merger. However, since it is a worldwide merger, which was already ap-proved in the U.S and E.U, conceivably the decision of Ecuador would not affect the merger
itself. Thus, remedies would be the appropriate tool so that former competition can be at least maintained.
Regarding the final resolution on this case, the Superintendency concluded to approve the notified merger by AB InBev, conditionally to the fulfillment of structural remedies. The
most relevant remedies are the ones mentioned below and concern the sale of a divestiture package to a suitable purchaser.
1. Divestiture of Ambevs factory, its distribution system and the selling of the brands
Zenda, Dorada and Biela.
2. License, permit and authorization to use and exploit Brahma and all of the products of this brand.
3. Access and use of the sale channel provided by Dinadec S.A.
Bud66.
5. Hold Separate Agreement, which mentions that the firms will remain competing as separate businesses until the new purchaser enters the Ecuadorian market and that
the remedies mentioned are fulfilled.
Determining what is an acceptable divestiture package is critical in designing an effective relief. This means that the package should be attractive enough for the new purchaser to
compete effectively and thereby maintain or restore competition that would otherwise have been lost as a result of the merger. It is relevant therefore to remember that the final goal of
a divestiture is to ensure that the approved purchaser possesses the means and the intention to effectively maintain or restore the competition in the industry (Network, 2016).
Competition authorities have the ultimate decision of approving the purchaser. There-fore, characteristics as financial capability, managerial expertise and operational capability
must be considered. At the moment, and as far as is known, the Superintendency might have received intentions of purchasers. Nonetheless, there is not yet a new competitor in
the beer market.
There are a few reasons why finding a suitable purchaser might be difficult.
In first place, the incentive for a new purchaser to enter the market should be strong enough, such that being a competitor will result in profitability of the business. In the
Ecuadorian beer market, as it was seen before, Cervecera Nacional has historically held almost the entire market share, mainly because of brand positioning of Pilsener and Club,
which are brands that are strongly related to the Ecuadorian nationality and are the pre-ferred in the on-trade and off-trade channel. Most probably, Cervecera Nacional is the
market leader because of the distribution system which is in charge of Dinadec S.A. (also SAB Miller S.A.). Dinadec distribution system allows Cervecera Nacional products to
be available in every supermarket and mainly in all rural areas and small stores around Ecuador. Despite Ambev having a good distribution system, the coverage of distribution
is better by Cervecera Nacional and Dinadec. For Ambev, entering the Ecuadorian market in 2003 and despite of being a global market leader, it was impossible to compete or gain
market share over the years. Hence, what would be the probability of the new purchaser to compete in fair conditions with the merging firms?
From the side of the Superintendency, it would not be correct to approve any candidate as a new competitor. The purchaser must be strong enough to change the dynamics of the
market and the incentives of the merging parties to increase prices dramatically or harm competition. The authority cannot release its authorization if the divestment will cause
further competition problems or delay the implementation of the remedy. Furthermore, there might be a situation where the purchaser will be unable to carry on the business
going forward. Also, if the size of the buyer is too small in the market, it could prevent from developing the business in a competitive way (Percivalle, 2014).
In Ecuador, with its short experience in competition policy, there has been just one
ad-ditional merger case in which the authority asked for remedies to approve the concentration. The merger consisted in the acquisition of Holding Tonicorp S.A. by Arca Ecuador S.A.
and The Coca Cola Company in the soft drink industry. This case however did not consist in a divestiture package but on arrangements of exclusivity contracts with suppliers and
consumers. Consequently, the beer case is the first merger case in Ecuador that is dealing with divestiture as a remedy.
8
Conclusions
This thesis has investigated the effect of the merger between Anheuser-Busch InBev and
SAB Miller on the beer price in Ecuador. It has used data covering the 2007-2015 time period in a number of nested logit model regressions, in order to estimate the demand of
the beer industry in Ecuador and subsequently, simulate the merger between the mentioned firms. As it was suspected, the proposed merger would increase the prices in a severe way.
According to the simulation, Ambev (Anheuser-Busch InBev) would increase prices by 27% and Cervecera Nacional (SAB Miller) by 2.3%. The Premium segment would be the most affected segment, with a price increase of 26% when efficiencies are not considered and 13%
obtained by the European Commission and the Superintendency for Market Power Control. Both competition authorities decided to approve the merger conditional with the
fulfill-ment of the divestiture of basically the whole Anheuser-Busch InBevs business in Ecuador and SAB Millers business in Europe. Among merger remedies, divestiture is one of the
preferred tools to prevent the worsening of a dominant position. Given these considera-tions, I am able to conclude that the divestiture remedy has been applied coherently by
the Superintendency both at merger and at antitrust level. However, since divestiture is a recent instrument in Ecuador and it is the first time for the competition authority to set
such remedy, it is significant that the authority approves a strong and suitable purchaser and keeps an eye on the development of the beer market in the following years.
In view of the remedies, the proposed transaction would no longer raise competition
concerns and the intensity of competition will remain unchanged. It is desirable that the new competitor can challenge the beer industry and adjust it into a more fair market with
competition, which at last will benefit the consumers.
There is a number of gaps or additional analyses, which could be conducted by other
researchers. First, the use of modeling techniques in merger control is still an open field for further analysis and discussion.
Continuing with the merger case, it would be important to analyze the post-merger performance of the market and the response of prices and demand. Unfortunately, at the
moment there is not available post-merger information.
Also, it might be important to discuss or create a merger guideline for developing
coun-tries. For instance, what would be a reasonable decision for competition authorities in developing countries when the effects of mergers are negative but are not harmful in the
Appendix
Table 18: Appendix 2 Endogeneity test
Note: The command used in STATA is estat endogeneity Unlike the Durbin and Wu-Hausman test, Wooldridges score and the regression-based test do not allow to test a subset of the endogenous regressors in the model, instead, whether all he endogenous regressors are in fact exogenous can be tested. (Stata)
Table 19: Appendix 3 Instrumental Variable validity
Note: estat firststage is the command used in Stata for the Sheas partial R-squared. For the over identifying test, estat overid command is used. By default if the model contains one endogenous regressor, then the firs-stage R2, adjusted R2, partial R2 and f statistics are reported, but since there are 2 endogenous regressors, Sheas R2 and adjusted R2 are reported instead. What really matters for IV estimation is whether the components of the instruments that are orthogonal to the other regressors can be explained by the component of he predicted value of the instruments that is orthogonal to the predicted values of the other regressors in the model. Sheas partial R2 measure this correlation Regarding the score chi2 the Wooldridges score test of overdignifying restrictions is robust to heteroskedasticiy (Stata)
