Who decide the prices of lemonade products in the
Netherlands? Market, brands, or chains.
- The application of dynamic hierarchical factor model
University of Groningen
Faculty of Economics and Business
Author: Linping Luo
Student No. S3685586
l.luo.2@student.rug.nl
First Supervisor: Dr. Keyvan Dehmamy
Second Supervisor: Prof. Dr. Wieringa, J.E.
Abstract
Dynamic hierarchical factor model (DHFM) is applied to the Dutch lemonade market to figure out who is the most powerful entity to decide the prices of lemonade products. A three-level DHFM is built, the common level is the whole Dutch lemonade market level, the block level is all brands in five different chains, and the sub-block level is divided each different chain to five brands. Markov Chain Monte Carlo (MCMC), variance decomposition, impulse responses are conducted. In conclusion, different brands and chains have different power to set the prices, and the influence of market adjustment on price is different per chain and brand. In general, brand managers are more powerful to decide price of products than chain managers, and the influence of market changes on the prices of lemonade products is the lowest among brand managers and chain managers.
Keywords: dynamic hierarchical factor model (DHFM), common dynamics, variance
Forward
The finish of this thesis represents the end of my master program. When I was choosing the topic of the thesis, I was quite interested in the application of dynamic hierarchical factor model because of it is a cool model that I have never heard about, and I need to approach a new tool -- MATLAB. All new things make me excited and feel challenged, so I chose the application of dynamic heretical factor model as my thesis. The reason I chose the lemonade dataset is that I was interested in the dataset when I was doing the assignment for the Market Model course, and I want to know more insights of the Dutch lemonade market.
Abbreviations
AH Albert Heijn
KC Karvan Cevitam
MCMC Markov Chain Monte Carlo
HFM Hierarchical Factor Model
DHFM Dynamic Hierarchical Factor Model
VAR Vector Autoregressive
Table of Content
1. Introduction 5
2. Literature Review 8
2.1 Price 8
2.2 The Influence of Brand and Chain Managers on Price 8
2.3 The Influence of Market on Price 10
2.4 Factor Analysis 11
2.5 Common Dynamics 11
2.6 Hierarchical Factor Model 12
3. Theory, Data exploration and Hypotheses 14
3.1 Theory for Hypotheses 14
3.2 Data Exploration for Hypothesis 15
3.3 Conclusion of Hypothesis 18
4. Methodology 20
4.1 Dynamic Hierarchical Factor Model (DHFM) 20
4.2 Estimation Procedures 23
4.3 Factor Augmented Vector Autoregressive Model (FAVAR) 24
5. Results 26
5.1 Common Movements 26
5.2 Variance Decomposition 27
5.2.1 Variance Decomposition of Each Chain 28
5.2.2 Variance Decomposition of Each Brand 30
5.3 Impulse Responses 32
6. Conclusion 35
7. Implication and Limitations 37
7.1 Implication 37
7.2 Limitations 38
1. Introduction
Lemonade is sweetened beverages which can be characterized by lemon flavor. In the Netherlands, the consumption of Sugar Sweetened Beverage has increased from 59 liters per person per year in 1980 to 103 liters in 2009 (Statistic Netherland). The market share of Lemonade in the beverage category is 15%, which is next to the market share of Cola (45%) and fruit carbonates (40%) (Trinh, Dawes & Lockshin 2009). In the Netherlands, people consumed 501.9 million liters of lemonade products in 2016, which is almost double of other European countries (FWS 2017). Therefore, lemonade products are consumed frequently by Dutch people, which makes it is necessary to research the Dutch lemonade products.
According to the research of Gruenewald, Ponicki, Holder, & Romelsjö (2006), the price increasing for beverages causes significant decrease for sales, and customers react to pricing increasing by reducing total consumption and switching to cheaper brands. Lemonade products are sugar-sweetened beverages (SSB), and customer will shift to energy-containing products such as juices and milk-based drinks if the prices of SSB increase (Silver, Ng, Ryan-Ibarra, Taillie, Induni, Miles & Popkin 2017). Furthermore, researchers found that if the price of soft drinks increases with every 10%, the consumption will decrease with 8-10% (Andreyeva, Long & Brownell 2010; Block, Chandra, McManus & Willet 2010). Lemonade is categorized as a large amount of price promotions (Omar, 1994). In conclusion, the prices changed with lemonade products have considerable influence on sales, and lemonade products have promotions frequently. Therefore, it is important to do a research about what drives the prices of lemonade products.
category of products (Baltas 2004; Teng, Laroche, & Zhu 2007). Chains are becoming more and more important (Ailawadi, Lehmann & Neslin 2001; Kahnc & McAlister 1997). And companies with larger scales have more market power, and their size makes they can "administer" prices, so they usually have a higher price in some products (Buzzell, Gale, & Sultan 1975). Albert Heijn (AH) is the leading retailer in the Netherlands, price in AH is above the average price of other retailers, and AH is very powerful in the supply chain process (Bord Bia 2010). Therefore, chain manages might have more power to decide the prices of products, especially for chain with higher market share, such as AH. Besides the influence of chain managers, previous research show that the prices of lemonade products are influenced by lots of marketing effects, such as seasonal effects (Megan 2016), taxes (Powell, Chriqui, Khan, Wada, & Chaloupka 2012; Smith, Lin & Lee 2010), health concern (Brownell, Farley, Willett, Popkin, Chaloupka, Thompson & Ludwig 2009), and financial crisis (Gerstberger and Yaneva 2013; Brinkman, Sanogo, Subran, & Bloem 2010). Therefore, prices of products also influenced by the whole market.
Besides the previous literature, data exploration was done to support the hypotheses. The results show that there are no significant differences among prices of lemonade products in different chains, but it does have considerable difference among prices of lemonade products from each brand. The average prices from brand Raak is only 1.3 euro, and the average price of brand Karvan Cevitam (KC) is twice more than of brand Raak. Besides, average prices of lemonade products are higher during the summer or holidays. Therefore, the prices of lemonade products might maily influenced by brand managers, and also influenced by the whole market.
adjustment on price is different per chain or per brand. Brand managers are more powerful to decide price of products than chain managers. Pricing managers in chains with large market share have more power to influence the prices of lemonade products than in other chains.
In this paper, the influences of market on prices cannot be directly observed. To estimate it, a three-level dynamic hierarchical factor model (DHFM) will be used. DHFM is modified based on the factor model by structuring the data into blocks and sub-blocks. Two-level dynamic factor model (Geweke,1977; Sargent and Sims, 1977) that captured the variations by using common factor and block- specific factors is introduced by Moench, Ng, and Potter (2013). When it comes to the hierarchy of DHFM, the Common level (F) represents the common dynamics of the entire lemonade industry in the Netherlands market, and the block level (G) represents the chain-specific common dynamics of each supermarket (E.g., the variance of all brands in AH over time t). The final level represents the brand-specific common dynamics of each brand in each chain (E.g., the variance of KC in AH over time t).
2. Literature Review
2.1 Price
Price is the payment given by customers for goods or services (Rwegasira 2017), and it depends on the marginal cost and revenue in conventional theory (Hall and Hitch 1939). Generally, there are two types of prices, one is the regular price or original price, and the other one is the promotion price. The promotion price is the discounted price offered in a certain period for all customers (Blattberg, Briesch & Fox 1995), and price promotion is accepted by marketing managers as the fast and most reliable tactic to encourage customers to purchase (Aydinli, Bertini & Lambrecht 2014). The regular price or original price is the long-term price without price promotion, and marketer can change the regular price to lower or higher as one of the marketing tools (Blattberg & Neslin 1990). Price promotion usually comes with different types of communication (retailer ad. or others), and long-term price reduction could be the signal of the decrease of the regular price (Blattberg et al. 1990). Lemonade products have a large amount of price promotions (Omar, 1994). In this paper, the price is the promotion price with weighted average of all stores.
2.2 The Influence of Brand and Chain Managers on Price
Rohloff 1967; Duncan, Hollander, & Savitt 1983; Mason & Mayer 1984), and Walters and MacKenzie (1988) found that the chain traffic, sales and profits are stable when retailers promoted with different combinations. Therefore, both brand managers and chain managers can influence the prices of products by doing promotion.
According to research, chains are become more and more powerful in the supply chain process. According to the research of Ailawadi et al. (2001) and Kahnc et al. (1997), the power balance between manufacturers (brands) and retailers(chains) has changed significantly in past decades, and retailers (chains) are becoming more and more important. At the same time, brand power is different from the product categories, even for AH, which is the largest and most powerful retailer in the Netherlands. At some product categories, AH does well, but some other product categories not (Steenkamp et al. 1997). Therefore, literature show that chain managers might become more powerful to influence the prices of products than before, and different brands and chains could have different influence on the price of products even for the same category.
(DutchNews.nl 2017). In conclusion, literature show that chains with higher market share (e.g., AH) might have more power to influence the prices of products.
2.3 The Influence of Market on Price
Besides chain managers and brand managers have the power to influence the prices of lemonade products, market is also able to influence the price of lemonade from lots of aspects, such as taxes, financial crises, and healthy concern.
Tax has positive influences on the prices of lemonade products. The price of beverages increases because of higher tax, and the consumption of beverage goes down (Powell et al. 2009; Smith et al. 2010). According to the research of Powell et al. (2012), there is an increasing trend for government to use taxes and subsidies as a potential policy to influence customers’ food and beverages pattern. The VAT (Value Added Tax) of food and drink in the Netherlands increased from 6% in 2018 to 9% in 2019 (Belastingdienst of the Netherlands).
Based on the research of Gerstberger et al. (2013), financial crises have significant negative effects on the food and non-alcoholic beverages, which leading people are willing to pay less on lemonade products during financial crises. During the financial crisis, even the consumption of food and nutrition products are significantly decreased because of the bad economy (Brinkman et al. 2010). Lemonade products, as the non-necessary nutrition products for daily life, is also influenced by macroeconomy. According to the supply and demand theory (Arrow 1959), the price of products is influenced by both the need and supply. Therefore, the prices of lemonade products are influenced by the financial crisis.
it also could result in an overweight issue because it contains too much sugar. Sugar-sweetened beverages (SSBs) has been considered as one of the important reasons for obesity, heart disease and diabetes (Brownell et al. 2009). Customers are becoming more and more care about health lifestyle, and there is an increasing trend for an official organization to use pricing policy as the potential policy tool to protect citizens from obesity (Powell et al. 2009). Therefore, the prices of lemonade products might also be influenced by the heath concern.
2.4 Factor Analysis
DHFM is modified based on factor analysis, and the goal of factor analysis is to reduce a large quantity of data by finding common variance to retrieve underlying dimensions in the dataset (Widaman 1993), and test if the hypothesized dimensions also exist in the dataset (Malhotra & Birks 2007). Boivin & Ng (2006) stated that factor analysis can help to find the type of variation which is relevant for the data as a whole. Thompson (2007) stated that there are 3 purposes of factor analysis, the first one is creating a theory of structure based on experience (e.g., Cattell's Structure of Intellect model. The second one is evaluating if factored entities (e.g., variables) cluster in a theoretically expected way. The third one is estimating latent variables scores (i.e., factor scores). In this case, only normal factor model is not enough because it cannot provide insights of common movements and hierarchy structure of the whole Dutch lemonade market.
2.5 Common Dynamics
in the commodity prices and identified a common factor, which conclude that it is necessary to include common variations at different levels because comparable trends and heterogeneity can be shared by different levels.
Diebold and Yue (2008) built a comparable model and compared the common and idiosyncratic dynamics over time, and which shows that all country-specific dynamics moved similar to the global dynamics. To help marketing managers to decide the price strategy, dynamic composite in common movement is necessary to analyze because the variance of different levels’ common dynamics shows the variance of affected price promotion depth (Barberis, Shleifer & Wurgler 2002). Bernanke, Boivin & Eliasz (2005) stated that in dynamic factor model, few estimates indexes and factors can be generated from lots of time series information. To predict future, Quah & Sargent (1993) summarized information from extensive time series information by building an approximate dynamic factor model. Therefore, common dynamics are necessary to contain in this paper because it can provide insights of the common movement of the whole Dutch lemonade market.
2.6 Hierarchical Factor Model
3. Theory, Data exploration and Hypotheses
3.1 Theory for Hypotheses
Brand manager and chain manager are able to influence prices from different ways. Besides they can set the base price (Lau et al. 2007), they also motivated and capable to do price promotions. Price promotion has significant effects on brand sales (Walters 1991), especially in short-term (Wilkinson et al. 1982). Chain manager provide short-term promotion to attract consumers to stores and stimulate them to buy products with a regular price (Mulhern et al. 1995; Kuehn et al. 1967; Duncan et al. 1983; Mason et al. 1984). Furthermore, the category of lemonades is characterized by a large amount of price promotions (Omar, 1994). In conclusion, brand managers and chain managers have the power to influence the prices of products.
Besides chain managers and brand managers have the power to influence the prices of products, market also can influence the prices of products from different ways. Taxes have significant positive influences on the prices of products, and there is an increasing trend for government to use taxes and subsidies as a potential policy to influence customers’ food and beverages pattern (Powell et al. 2012; Smith et al. 2010). The VAT (Value Added Tax) of food and drink in the Netherlands increased from 6% in 2018 to 9% in 2019 (Belastingdienst of the Netherlands). Besides taxes, the health concern also influences the prices of products. Lemonade belongs to sugar-sweetened beverages (SSBs), and which has been considered as one of the important reasons for obesity, heart disease and diabetes (Brownell et al. 2009). There is an increasing trend for an official organization to use pricing policy as the potential policy tool to protect citizens from obesity (Powell et al. 2009). Furthermore, final crisis has significant negative influence on the prices of products (Gerstberger et al. 2013; Brinkman et al. 2010).
To conclude, previous research show that chain managers, brand managers and market have power to influence the prices of lemonade products. Chains are becoming more powerful to set prices than before, and chains with higher market share are more likely to have large influence on the pricing decision.
3.2 Data Exploration for Hypothesis
The dataset, provided by the Nielsen Company, contains price data in the Netherlands which are collected from five lemonade brands within five supermarket chains, based on 169 weeks (week 41 in 2013 to week 52 in 2016). The prices of lemonade in this case is the weighted average prices of all stores.
which is one week before the Charismas, and this finding is supported by literature that sales and prices become higher before Christmas (Arthur 2017). This also indicates that the prices of lemonade products were influenced by the festivals. As shown in figure 1, the prices at the beginning were quite low, and it increased gradually. The reason of this might be that the prices of lemonade products are affected by the seasonal changing, the start week of the dataset was winter, and the prices was low, and then the prices increased when summer was coming. This finding also supported by previous research; the consumption of lemonade might different between winter and summer in place with the clear seasonal difference (Kim et al. 2001; Yang et al. 2007). To conclude, market can influence the prices of lemonade products by seasonal changes or festival.
Figure 1 Average price trend in lemonade market
power, and their size makes they can "administer" prices, so they usually have a higher price in some products (Buzzell et al. 1975). At the same time, there is not too much differences among prices of lemonade products in different chains.
Figure 2 Price map of different chains
Figure 3 illustrates the prices of lemonade products from five brands. The average prices of KC and Teisseire are the highest (round 2.7 Euro per product) in the Dutch lemonade market, which is twice higher than the price of brand Raak. To conclude, the prices of different brands are quite different.
Figure 3 Price map of different lemonade brands
3.3 Conclusion of Hypothesis
Based on literature mentioned before, chain managers are becoming more and more powerful in supply chain. But the data exploration shows that the prices of lemonade products in different chains does not have considerable differences, and there are significant differences among prices of lemonade products from different brands. To conclude, the hypotheses for this paper is that although chains are becoming more and more powerful, it does not mean chain managers are more powerful than brand managers to set the prices of lemonade products.
other chains. Therefore, chain managers in AH might have more power to set prices than other chains.
Both literature and data exploration show that the prices of lemonade products are influenced by the whole market. The prices and sales of lemonade products increase when it comes to festivals and summer (Megan 2016). Based on literature and data exploration, hypotheses for this paper are:
Hypothesis 1: Different brands have different power to set the prices.
Hypothesis 2: Different chains have different power to set the prices of products. Hypothesis 3: The influence of market adjustment on price is different per chain. Hypothesis 4: The influence of market adjustment on price is different per brand.
Hypothesis 5: Brand managers are more powerful to decide price of products than chain managers.
4. Methodology
4.1 Dynamic Hierarchical Factor Model (DHFM)
The DHFM is modified from the factor analysis, and the goal of factor analysis is to reduce a large quantity of data by finding common variance to retrieve underlying dimensions in dataset (Widaman 1993) and test if the hypothesized dimensions also exist in the dataset (Malhotra et al. 2007). Factor model assumes that common variance exists among a set of variables. However, basic factor analysis cannot take dynamic hierarchical effects into account. DHFM not only can reduce dimension in the same way as the basic factor model does, but also simultaneously contains the hierarchical structure of factors. Sims & Sargent (1977) and Geweke (1977) applied a two-level dynamic factor model in the economic analysis by extending the static strict factor model to allow for dynamics, called the dynamic factor model (DFM). Moench et al. (2013) extended the DFM to two level DHFM by adding the hierarchy into the model, which includes the common level and block level. When it comes to model with three or higher level, each block will be divided into sub-blocks to reach the third hierarchical framework and all time-varying intercepts should be included in equations because the current level is depending on previous hierarchical level (Moench et al. 2013).
AH over time t). And the third level (X) represents each brand in each chain (E.g., the variance of KC in AH over time t). Graph 4 shows the hierarchy of the DHFM.
Figure 4 DHFM Conceptual Model
five lemonade brands (𝑁𝑏 = 5). Therefore, 25 time series (N=B* 𝑁𝑏) with 169 observations (T=169) are included in this model.
In the three-level DHFM, the variation of each time series n in a given block b was decomposed into a serially correlated idiosyncratic component and a common component. Xbnt refers to the total movement of each brand in each chain (e.g., KC in AH). 𝐺𝑏𝑛(𝐿)𝐺𝑏𝑡is the common component, 𝐺𝑏𝑛(𝐿) is the polynomial of lag order L, and it stands for the variation that time series n at block b shares with the other time series at block b. 𝑒𝑋𝑏𝑛𝑡 refers to the idiosyncratic components, which describe the variation that is unique for the time series n at the third level x and block b. To conclude, the equation is:
𝑋𝑏𝑛𝑡 = 𝐺𝑏𝑛(𝐿)𝐺𝑏𝑡+ 𝑒𝑋𝑏𝑛𝑡 (1)
The variation of block level can be decomposed into a serially correlated block specific component and a common component. 𝐺𝑏𝑡 refers to the total movement of all brands in each chain (e.g., five brands in AH). 𝐹𝑏(𝐿)𝐹𝑡 is the comment component, 𝐹𝑏(𝐿) is the polynomial of lag order L, and it has l_F lags. 𝑒𝐺𝑏𝑡 is the block-specific component, which denotes the variation that is unique for any block (e.g., AH). To conclude, the equation is:
𝐺𝑏𝑡 = 𝐹𝑏(𝐿)𝐹𝑡+ 𝑒𝐺𝑏𝑡 (2)
The top level of hierarchy for the DHFM is 𝐹𝑡, which represents the variation of all factors in the Dutch lemonade market. 𝐹(𝐿) is a distributed of loading on the common factors of the whole Dutch lemonade market, and it has q_F lags. The market level is assumed to be serially correlated, and the equation is:
We assume 𝑒𝑋𝑏𝑛𝑡, 𝑒𝐺𝑏𝑡, and 𝑒𝐹𝑡 that to be stationary, normally distributed autoregressive processes.
𝑒𝑋𝑏𝑛𝑡 = Xbn(𝐿)𝑒𝑋𝑏𝑛𝑡 𝑒𝑋𝑏𝑛𝑡~N (0,𝜎𝑋𝑏𝑛2 ) (4)
𝑒𝐺𝑏𝑡 = Gb(𝐿)𝑒𝐺𝑏𝑡 𝑒𝐺𝑏𝑡~N (0,𝜎𝐺𝑏2 ) (5)
𝑒𝐹𝑡 = F(𝐿)𝐹𝑡 𝑒𝐹𝑡~N (0,𝜎𝐹2) (6)
Lots of insights of the prices can be found by estimating this model. Firstly, check whether prices of the all brands in all chains follow common dynamics, chain specific or brand specific dynamics. Secondly, the share of F (market level), G (chain level) and X (brand level) can be observed by doing variance decomposition. Finally, the reaction of prices to an exogenous shock in the Dutch lemonade market can be measured on market, chain and brand level by estimating impulse responses.
4.2 Estimation Procedures
are obtained by backwards smoothing. Three common factors chose, so 𝐾𝐹 = 3; the specific factor for block is 𝐾𝐺,𝑏 = 1, and the lag order 𝑞𝐹,𝐺,𝑋 = 2. 100,000 draws are estimated to get
an equally convergence sample, the first 50,000 draws are burn-in and not used to minimize their effect on the other 50,000 draws.
Secondly, the variance decomposition will be conducted to provide insights into the importance of each level in DHFM and add all these decomposed variances up is the total variance in price. The variance decomposition analysis is widely applied in economics, Campbell (1990) used it to figure out what moves the stock market. According to the research of Moench, Ng, & Potter (2013), the equation of share of common level, block level and sub-block level are as following:
Var (𝑋𝑏𝑛) = 𝛾𝐹(Var (F)) + 𝛾𝐺(Var (𝑒𝐺𝑏)) + 𝛾𝑋(Var (𝑒𝑋𝑏𝑛)) (7)
𝑆ℎ𝑎𝑟𝑒𝐹 = 𝛾𝐹(Var (F)) / Var (𝑋𝑏𝑛) (8) 𝑆ℎ𝑎𝑟𝑒𝐺 = 𝛾𝐺(Var (𝑒𝐺𝑏)) / Var (𝑋𝑏𝑛) (9)
𝑆ℎ𝑎𝑟𝑒𝑋 = 𝛾𝑋(Var (𝑒𝑋𝑏𝑛)) / Var (𝑋𝑏𝑛) (10)
4.3 Factor Augmented Vector Autoregressive Model (FAVAR)
which solved the degrees of freedom problem in VARs models by combining the VAR model with factor analysis (Belviso & Milani 2006; Wagan 2011).
𝑌𝑡 is vector of an observable economic variable, which represents for the dynamic of the whole market. In this case, 𝑌𝑡 is the weekly sales of per unit of the different lemonade brands. 𝐹𝑡is an additional economic information, and it is a vector of latent factors. Φ(L) is a conformable lag polynomial of finite order p, which might contain a priori restrictions (Bernanke, Boivin, & Eliazs, 2005). 𝑣𝑡 is the error term with means zero (Vargas-Silva 2008; Bernanke et al. 2005). Therefore, the equation is:
[𝐹𝑌𝑡
𝑡] =(𝐿) [
𝐹𝑡−1
5. Results
The results of estimation and the interpretation of estimation will be discussed in this part. A three-level DHFM model will be estimated, the Common level (F) represents the common dynamics of the entire lemonade industry in the Netherlands market. The block level (G) represents the chain-specific common dynamics of each supermarket (e.g., the variance of all brands in AH over time t), and the final level (X) represents the brand-specific common dynamics of each brand in each chain (e.g., the variance of KC in AH over time t). In this part, the results and interpretation of common movements, variance decomposition and impulse response will be discussed.
5.1 Common Movements
Gibbs sampling, the Markov Chain Monte Carlo (MCMC) iterative method, is used to get 100,000 draws, and the first 50,000 draws are used to minimize their effect on the other 50,000 draws. For the second 50,000 draws, every 50th draw is stored, so there are 1,000 draws
can be used for the posterior analysis.
or a negative rate of the price decrease. Besides, the third factor has negative values, which shows a negative rate of general price increase or positive rate of a general price decrease.
Figure 5 Common movements of all lemonade products in all chains
5.2 Variance Decomposition
The variance decomposition analysis is conducted in this step to compare the variance of prices explained by each level in DHFM. According to the research of Moench, Ng, & Potter (2013), the common share F = 𝛾𝐹(Var (F)) / Var (𝑋𝑏𝑛), the block share 𝐺 = 𝛾𝐺(Var (𝑒𝐺𝑏)) / Var
5.2.1 Variance Decomposition of Each Chain
Table 1 shows the share of three different DHFM levels for five different chains. High variance in share F means a strong reaction to changes in the market. AH has the largest value (15.88%) on the common level than other four chains, which means AH reactive the most strong than other chains when there is a change in the Dutch lemonade market, e.g., seasonal changes. On the contrary, Jumbo (1.72%) reactive the least strong than other chains.
High variance in share G means strong reactions at the chain level. Block level (G) explains the 49.05% of the total variance of Jumbo, which is much higher than AH (14.25%), Plus (32.09%), Coop (34.18%) and EMTE (32.33%). These finding indicates that pricing managers in Jumbo is more powerful to decide the prices of lemonade products than all other chains, and pricing managers in AH have the least power to determine the prices of lemonade products. It is surprised to find that only 14.25% of prices of lemonade products in AH is explained by the chain level, which is against the results of literature. Previous research show that companies with larger scales have more market power, and their size makes they can "administer" prices, so they usually have a higher price in some products (Buzzell et al. 1975). AH is the leading retailer in the Netherlands (Bord Bia 2010), and it has the largest market share (35.4% in 2016), which is obviously higher than the second largest group Super Unine (29.6%), which includes then Plus, Deen and Spar Labels (DutchNews.nl 2017). Theoretically, the prices of lemonade products in AH should most explained by the chain level than other chains.
Literature shows chain are becoming more and more important (Ailawadi et al. 2001; Kahnc et al. 1997; Steenkamp et al. 1997; Lehmann et al. 2001). Combing the finding from this paper, the conclusion is that although chains are becoming more and more powerful, the prices of products still mainly depend on brand managers instead of chain managers. Furthermore, the total variance explained by common level (Dutch lemonade market level) is lower than the variance explained by chain level and brand level. This finding could also be a supplement of previous research. Literature shows that the prices of lemonade products are influenced by lots of marketing effects, such as seasonal effects (Megan 2016), taxes (Powell et al. 2012; Smith et al. 2010), health concern (Brownell et al. 2009), and financial crisis (Gerstberger et al. 2013; Brinkman et al. 2010). Combing the estimation results and literature, the prices of lemonade products mainly depend on the brand level, and the influences of marketing on the prices of lemonade products are limited.
Table 1 Variance decomposition of each chain
Chain Share F Share G Share X
Albert Heijn 0.1588 (0.0297) 0.1425 (0.0214) 0.6987 (0.0439) Jumbo 0.0172 (0.0087) 0.4905 (0.0434) 0.4923 (0.0433) Plus 0.0407 (0.0196) 0.3209 (0.0646) 0.6383 (0.0657) Coop 0.0313 (0.0163) 0.3418 (0.0696) 0.6268 (0.0723) EMTE 0.0219 (0.0104) 0.3233 (0.0322) 0.6548 (0.0337)
5.2.2 Variance Decomposition of Each Brand
most actively to the shock on brands level. It is interesting to find that the prices of lemonade products for all brands are barely explained by the common level (Dutch lemonade market), the maximum case is 6.73% for KC. Besides, the prices of brand with highest average prices (KC) are mainly depend on chains, and all other four brands mainly depend on brand level.
In conclusion, hypothesis 1 and 4 are supported by the variance decomposition analysis. Different brands have a different power to set the prices of products, and the influence of market adjustment on price is different among brands. Besides the hypothesis mentioned before, there is a new finding which is the prices of lemonade products mainly depend on the brand level, and the influences of marketing on the prices of lemonade products are limited.
Table 2 Variance decomposition of each brand
Brands Share F Share G Share X
5.3 Impulse Responses
Impulse response analysis is conducted to show how a shock in the whole Dutch lemonade market can affect the chain and brand level in lemonade price per unit.
5.3.1 Shock on Common Factor Level F
A standard deviation shock on the common factor level F is set. Figure 6 illustrates the positive responses of common F with shock on the first common factor, and it shows that the first common factor goes up when there is a shock, and then it will come back to steady after five weeks.
Figure 6 The impulse response of common F with the shock on common F
5.3.2 Shock on Block Level G
Figure 7 Impulse response of five blocks with shock on 3 common factors
5.3.3 Shock on level X
6. Conclusion
In this paper, DHFM is applied to the Dutch lemonade dataset to analyze what drives the prices of lemonade products. To get the results, Gibbs sampling, which is the Markov Chain Monte Carlo (MCMC) iterative method, is conducted to figure out the common movement of the whole Dutch lemonade market; variance decomposition is conducted to provide insights of the importance of different levels in DHFM on prices decision-making process; and impulse response is conducted to show how a market shock affects the prices of different chains and different brands in chains.
In conclusion, different brands and chains have different power to set the prices, and the influence of market adjustment on price is different across per chain and per brand. Additionally, there are some findings from this paper, which provide some supplements for previous research. According to literature, chains are becoming more and more powerful and important (Ailawadi et al. 2001; Kahnc et al. 1997; Steenkamp et al. 1997; Lehmann et al. 2001). This paper shows that although chain managers play an important role in pricing decision (around 30% of prices explained by chain level), the prices of lemonade products mainly depend on brand managers instead of chain managers. On the other side, literature also shows that marketing factors also can influence the prices or products from different aspects, such as seasonal effects (Megan 2016), taxes (Powell et al. 2012; Smith et al. 2010), health concern (Brownell et al. 2009), and financial crisis (Gerstberger et al. 2013; Brinkman et al. 2010). The result shows that market indeed has some influence on the prices of lemonade products, but the influence of market on prices is quite limited (only 5% of prices explained by market for most chains).
7. Implication and Limitations
7.1 Implication
• Implication for Academic
Although previous research show that chains are becoming more and more powerful (Ailawadi, et al. 2001; Kahnc et al. 1997; Lehmann et al. 2001), it does not mean chain managers are more powerful to influence the prices of lemonade products than brand manages. According to the estimation results, above 50% of price of lemonade products are explained by the brand level instead of chain level.
Literature state that the prices of lemonade products are influenced by lots of marketing effects, such as seasonal effects (Megan 2016), taxes (Powell et al. 2012; Smith, et al. 2010), health concern (Brownell et al. 2009), and financial crisis (Gerstberger et al. 2013; Brinkman, et al. 2010). The estimation results from this paper show that the prices of lemonade products mainly depend on the brand level, and the influences of marketing on the prices of lemonade products are limited.
• Implication for Business
The prices of lemonade products mainly depend on brand managers, and the prices have significant influence on the sales of products. Therefore, brand managers need to be sure that the prices of products are matched with the target customers when they are doing pricing.
This paper shows that chain with large market share does not mean the pricing manager in this chain have more power to influence the prices of products, and the variance of prices explained by chain can provide some insights of the store management. Take AH as an example, only 14.25% of prices were explained by chain itself, which might indicate that store managers in AH don not have much power to set the prices of products. On the contrast, store managers in Jumbo (49.05%) do have more flexibility to set the prices of products. These funds provide insights of the management of different chains.
7.2 Limitations
The model is not fully complete. In this case, only the prices of five brands in five chains over 169 weeks variable were included. However, there are some other variables can influence the prices of lemonade products, such as advertisement and weather.
The hierarchy structure could be different. In this case, the structure of the hierarchy is Dutch lemonade market-chains-brands. Besides this way, the block level could be the brands instead of chains. Then information criteria can be used to see which combination of lags would improve the factor model.
Furthermore, higher levels of DHFM model could be built. In this case, a three-level model was built, which includes the Dutch lemonade market, five chains and five brands in each chain. A higher-level model can provide more insights into common movements of different levels, such as add one level of premium or normal brand, or national brand/international brand. Last but not least, factors in a regression analysis can be used to predict sales.
References
Gruenewald, P. J., Ponicki, W. R., Holder, H. D., & Romelsjö, A. (2006). Alcohol prices, beverage quality, and the demand for alcohol: quality substitutions and price
elasticities. Alcoholism: Clinical and Experimental Research, 30(1), 96-105.
Megan, C. 2016. How the change in seasons can affect your clients’ spending habits? https://www.nationalbusinesscapital.com/change-seasons-spending-habits/ Retrieved on 27-11-2018.
Andreyeva, T., Long, M. W., & Brownell, K. D. (2010). The impact of food prices on consumption: a systematic review of research on the price elasticity of demand for food. American journal of public health, 100(2), 216-222.
Block, J.P., Chandra, A., McManus, K.D., Willet, W.C., 2010. Point-of-purchase price and education intervention to reduce consumption of sugary soft drinks. Am. J. Public Health 100, 1427–1433.
Silver, L. D., Ng, S. W., Ryan-Ibarra, S., Taillie, L. S., Induni, M., Miles, D. R., ... & Popkin, B. M. (2017). Changes in prices, sales, consumer spending, and beverage consumption one year after a tax on sugar-sweetened beverages in Berkeley, California, US: A before-and-after study. PLoS medicine, 14(4), e1002283.
Omar, B. (1994). Can we talk of health promotion at a time of so many
Belviso, F., & Milani, F. (2006). Structural factor-augmented VARs (SFAVARs) and the effects of monetary policy. Topics in Macroeconomics, 6(3).
Wheeler, M., & Chowdhury, A. R. (1993). The housing market, macroeconomic activity and fiancial innovation: an empirical analysis of US data. Applied Economics, 25(11), 1385-1392.
Vargas-Silva, C. (2008). The effect of monetary policy on housing: a factor-augmented vector autoregression (FAVAR) approach. Applied Economics Letters, 15(10), 749-752.
Bernanke, B. S., Boivin, J. (2003). Monetary policy in a data-rich environment. Journal of Monetary Economics, 50, 525–46.
Wagan, H. (2011). Measuring the impact of monetary policy: a factor-augmented vector autoregressive (favar) approach under bayesian framework. Economics Bulletin, 31(4), 1-48.
Bernanke, B. S., Boivin, J., & Eliasz, P. (2005). Measuring the effects of monetary policy: a factor-augmented vector autoregressive (FAVAR) approach. The Quarterly journal of economics, 120(1), 387-422.
Trinh, G., Dawes, J., & Lockshin, L. (2009). Do product variants appeal to different segments of buyers within a category?. Journal of Product & Brand Management, 18(2), 95-105.
Mulhern, F. J., & Padgett, D. T. (1995). The Relationship between Retail Price Promotions and Regular Price Purchases. Journal of Marketing, 59(4), 83–90.
Kuehn, Alfred and Albert Rohloff (1967), Evaluating Promotions Using a Brand Shifting Model "in Promotional Decisions Using Mathematical Models, Patrick Robinson, ed. Boston: Allyn and Bacon, Inc., 50-85.
Duncan, Delbert, Stanley Hollander, and Ronald Savitt (1983), Modern Retailing Management. Homewood, IL: Richard D. Irwin. Inc.
Mason, J. Barry and Morris Mayer (1984), Modern Retailing. Plano, TX: Business Publications, Inc.
Walters, R., & Mackenzie, S. (1988). A structural equations analysis of the impact of price promotions on store performance. Journal of Marketing Research, 25(1), 51-63.
doi:10.1177/002224378802500105
Wilkinson, Judy B., J. B. Mason, and Christie Paksoy (1982), "Assessing the Impact of Short-Term Supermarket Strategy Variables," Journal of Marketing Research, 19 (February), 72-86.
Powell, L.M., Chriqui, J.F., Chaloupka, F.J., 2009. Associations between state-level soda taxes and adolescent body mass index. J Adolesc Health. 45 (3), S57–S63.
Smith TA, Lin BH, Lee JY., 2010. Taxing caloric sweetened beverages: potential effects on beverage consumption, calorie intake, and obesity. Economic Research Report 2010. No. (ERR-100).
Gerstberger, C. & Yaneva, D. (2013), Analysis of EU-27 household final consumption expenditure — Baltic countries and Greece still suffering most from the economic and financial crisis, Eurostat Statistics in Focus, no. 2/2013. Retrieved from http://
epp.eurostat.ec.europa.eu/cache/ITY_OFFPUB/KS-SF-13-002/EN/KS-SF-13- 002-EN.PDF [accessed March 2013]
Villarejo-Ramos, A. F., & Sanchez-Franco, M. J. (2005). The impact of marketing communication and price promotion on brand equity. Journal of Brand Management, 12(6), 431-444.
Hall, R., & Hitch, C. (1939). Price Theory and Business Behaviour. Oxford Economic
Teng, L., Laroche, M., & Zhu, H. (2007). The effects of multiple-ads and multiple-brands on consumer attitude and purchase behavior. Journal of Consumer Marketing, 24(1), 27-35.
Baltas, G. (2004). A model for multiple brand choice. European Journal of Operational
Research, 154(1), 144-149. doi:10.1016/S0377-2217(02)00654-9
Steenkamp, J., & Dekimpe, M. (1997). The increasing power of store brands: Building loyalty and market share. Long Range Planning, 30(6), 917-930. doi:10.1016/S0024-6301(97)00077-0
Geweke, J. (1977). The Dynamic Factor Analysis of Economic Time Series. in D. J. Aigner and A. S. Goldberger (eds.), Latent Variables in Socio Economic Models (Amsterdam: North-Holland).
Sargent, T., & Sims. C (1977). Business Cycle Modelling without Pretending to Have Too Much a Priori Economic Theory in New Methods in Business Cycle Research, ed. by C. Sims. Minneapolis: Federal Reserve Bank of Minneapolis.
Stock, J., & Watson, M. (1989). New indexes of coincident and leading economic indicators. Nber Macroeconomics Annual, 4, 351-394. doi:10.1086/654119
El-Ansary, A., & Stern, L. (1972). Power measurement in the distribution channel. Journal of Marketing Research, 9(1), 47-52. doi:10.1177/002224377200900110
Lau, A., Lau, H., & Wang, J. (2007). Pricing and volume discounting for a dominant retailer with uncertain manufacturing cost information. European Journal of Operational
Research, 183(2), 848-870. doi:10.1016/j.ejor.2006.10.017
Pan, K., Lai, K., Leung, S., & Xiao, D. (2010). Revenue-sharing versus wholesale price mechanisms under different channel power structures. European Journal of Operational
Research, 203(2), 532-538. doi:10.1016/j.ejor.2009.08.010
Nederlandse vereniging Frisdranken, Waters, Sappen. (2016). Kerngegevens Basic statistical information. Publicatie van de Nederlandse vereniging Frisdranken, Waters, Sappen
(FWS).
Powell, L. M., Chriqui, J. F., Khan, T., Wada, R., & Chaloupka, F. J. (2012). Assessing the potential effectiveness of food and beverage taxes and subsidies for improving public health: a systematic review of prices, demand and body weight outcomes. Obesity reviews : an official journal of the International Association for the Study of Obesity, 14(2), 110–128. doi:10.1111/obr.12002
Arrow, K. J. (1959). Toward a theory of price adjustment. The allocation of economic resources, 41-51.
Rwegasira, M. (2017). Oil prices swing in Tanzania–the social and economic impact to the community: a case study of Dar es Salaam (Doctoral dissertation, The University of Dodoma).
Blattberg, R., Briesch, R., & Fox, E. (1995). How promotions work. Marketing Science, 14(3), 122.
Aydinli, A., Bertini, M., & Lambrecht, A. (2014). Price promotion for emotional impact. Journal of Marketing, 78(4), 80-96.
Blattberg, R. C., & Neslin, S. A. (1990). Sales promotion. Englewood Cliffs.
Carvalho, A. X., & Puterman, M. L. (2005). Dynamic optimization and learning: How should a manager set prices when the demand function is unknown?.
"LEMONADE - meaning in the Cambridge English Dictionary". dictionary.cambridge.org.
Powell, L. M., & Chaloupka, F. J. (2009). Food prices and obesity: evidence and policy implications for taxes and subsidies. The Milbank quarterly, 87(1), 229–257.
doi:10.1111/j.1468-0009.2009.00554.x
Arthur, B. 2017. Alcohol and soft drinks show strong growth in Christmas supermarket.
https://www.beveragedaily.com/Article/2017/01/10/Alcohol-soft-drinks-in-UK-Christmas-supermarket-sales. Retrieved on Nov 24, 2018.
Kim, B. H., Gye, S. H., Lee, H. S., Jang, Y., & Sin, A. J. (2001). 1999 Seasonal Nutrition Survey (I)-Food consumption survey. Journal of the Korean Dietetic Association, 7.
Yang, J. J., Park, S. K., Lim, H. S., Ko, K. P., Ahn, Y., & Ahn, Y. O. (2007). Seasonal Variation of Food Intake in Food Frequency Questionnaire among Workers in a Nuclear Power Plant. Journal of Preventive Medicine and Public Health, 40(3), 239-248.
Widaman, K. F. (1993). Common factor analysis versus principal component analysis: Differential bias in representing model parameters?. Multivariate behavioral research, 28(3), 263-311.
Malhotra, N., & Birks, D. (2007). Marketing Research: an applied approach: 3rd European Edition. Pearson education.
Boivin, J., & Ng., S. (2006). Are more data always better for factor analysis? Journal of Econometrics, 132, pp.169-194
Chadwick, M., Fazilet, F., & Tekatli, N. (2015). Understanding the common dynamics of the emerging market currencies. Economic Modelling, 49, 120-136.
Ojede, A. & Mugera, A. (2018). Co-Movements Between Key Components of Aggregate Productivity And Real Exchange Dynamics In Developing Countries.The Journal of Developing Areas, 52(4), 1-27.
Diebold, F. X., Li, C., & Yue, V. Z. (2008). Global yield curve dynamics and interactions: a dynamic Nelson–Siegel approach. Journal of Econometrics, 146(2), 351-363.
Barberis, N., Shleifer, A., & Wurgler, J. (2002). COMOVEMENT.Working Paper Series, 8895(8895).
Thompson, B. (2007). Factor analysis. The Blackwell Encyclopedia of Sociology.
Quah, D., & Sargent, T. J. (1993). A dynamic index model for large cross sections. In Business cycles, indicators and forecasting (pp. 285-310). University of Chicago Press.
Bernanke, B. S., Boivin, J., & Eliasz, P. (2005). Measuring the effects of monetary policy: a factor-augmented vector autoregressive (FAVAR) approach. The Quarterly journal of economics, 120(1), 387-422.
Bogomolova, S., Szabo, M., & Kennedy, R. (2017). Retailers' and manufacturers' price-promotion decisions: Intuitive or evidence-based?. Journal of Business Research, 76, 189-200.
Ailawadi, K. L., Lehmann, D. R., & Neslin, S. A. (2001). Market response to a major policy change in the marketing mix: Learning from Procter & Gamble's value pricing strategy. Journal of marketing, 65(1), 44-61.
Kahn, B. E., & McAlister, L. (1997). Grocery revolution. Addison-Wesley.
Moench, E., Ng, S., & Potter, S. (2009). Dynamic Hierarchical Factor Models. Federal Reserve Bank of New York Staff Reports, 412, 1-44.
Malhotra, N., & Birks, D. (2007). Marketing Research: an applied approach: 3rd European Edition. Pearson education.
Bord Bia, Irish Food Board, SUCCESSFULLY ENTERING THE DUTCH RETAIL MARKET, AN UNDERSTANDING OF PRICE MARGINS AND VALUE CHAIN MECHANICS, 2010,
https://www.bordbia.ie/industry/events/SpeakerPresentations/2010/MarketplaceSeminar20 10EuropeanGuides/Netherlands%20Guide.pdf.
Buzzell, R. D., Gale, B. T., & Sultan, R. G. (1975). Market share-a key to profitability. Harvard business review, 53(1), 97-106.
Kaufman, J. (2012). The personal mba : Master the art of business (Rev. and expanded ed.). New York, N.Y.: Portfolio/Penguin.
Hess, A., & Hess, J. (2018). Principal component analysis: Principal component analysis. Transfusion, 58(7), 1580-1582. doi:10.1111/trf.14639
