Understanding the Regional Demand and Price Response of Dealers in China Automotive Aftermarket A Case Study for Original Equipment Supplier

(1)

Understanding the Regional Demand and Price Response

of Dealers in China Automotive Aftermarket

A Case Study for Original Equipment Supplier

Author: Yan Fu

Date: December 2009

Student number: 1741748

Supervisor: Dr. J.E.M. van Nierop

Second supervisor: Dr. S. Gensler

(2)

Management Summary

This thesis studies the channel behavior in a one manufacturer/supplier and multiple dealerships system, particularly focus on dealer‟s demand and price response in China automotive aftermarket. The objective of our research is to discover what and how the product and location factors are influencing dealer‟s demand and price sensitivity to the products from one manufacturer, meanwhile to compare their difference to the retailing market. For the research subject, an automobile lighting products manufacturer in China, this research aims to assist them in learning more about their authorized dealers‟

behaviors and their preference to the products.

This research is implemented based on a sample of 1,498 observations in total, covers the sales data of 84 product items from the research subject to their 43 dealers throughout mainland China. By building separate multiplicative demand model to each of the four distinct product categories, the relevant and available factors are included as predictors to explain dealer‟s demand variation. Subsequently, each product category is classified further into two subgroups according to each observation‟s value in every predictor. The first subgroup contains the observations from smallest to the median value of each predictor, and the second subgroup includes the remaining observations from the category. In this way, price elasticities are obtained per subgroup to investigate the change of price sensitivity of dealers to each product and location factor.

Derived from the above analyses, it is concluded that observed product and local market characteristics have better explanation on dealer‟s demand elasticity, than dealer store characteristics. Product characteristics change dealer‟s price sensitivity in a different way. Dealers are more sensitive to the price of older products because of the declining

(3)

(4)

Preface

I still remember the first day when I came to Groningen in my friend‟s car, a stranger in the street helped us out and directed us to my student house before we got totally lost in the complex single-way roads. With the help of so many people, I completed my master thesis now and draw this rewarding study into the end. Recalling the past two years, there are so many people that I would like to express my sincere gratitude. My family, my parents and Thomas, thank you for your tremendous support and love in every regard. To my friends, Jackie, Dingding, Xiaomei, Vi-in, Spiros, and Xiao, thank you for sharing the tips in writing thesis and leaving me the happy moments as part of my beautiful memory in the Netherlands.

Besides, I would like to give my special thanks to my supervisors. I appreciate the guidance from Dr. Erjen van Nierop, for his patience and professional insights. With his sensible feedbacks, I learned a lot through writing this thesis. My second supervisor, Dr. Sonja Gensler, for her enlightening comments on the structure and contents, so that I could improve this thesis by paying attention on the ignored detailers. Finally, Ms. Gabriela Lo, the Marketing Manager of Company X, who provided me with the data and all the background information, which made this thesis possible.

Yan Fu

(5)

Chapter 1 Introduction

1.1 The Automotive Aftermarket in China

The automotive aftermarket industry provides replacement parts (excluding tires), accessories, maintenance items, batteries, and automotive fluids for cars and light trucks (Mantrala et al., 2006). Replacement parts include, lighting, electronic parts, electric parts, engines, transmissions and so on. For an auto part original equipment manufacturer

(OEM) Company Group, two parallel markets should be managed. One is the OEM market. The OEM entities (factory) produce the auto part products and sell them directly to the car manufacturer for assembling new cars. The other one is the independent aftermarket (IAM), also the topic of this research. Original equipment suppliers (OES) are the key market players in IAM, who purchase the auto part products from OEM entities (internal Company Group business) and sell them to the individual consumers through dealers (retailers). Authorized franchise dealers can carry and display only one manufacturer‟s line, unless other products are not considered as competition

(Rosenbloom, 2004).

In 2009, the automotive aftermarket in China is projected to grow by 13.8 percent, reaching RMB 40 billion ($4.9 billion). China‟s automotive aftermarket is experiencing tremendous levels of growth in recent years, due to the extremely high rates of new passenger vehicle sales1. However, the automotive aftermarket in China is still emerging, combined with the low labor cost in China, do-it-for-me business pattern dominates the industry, while do-it-yourself has been quite common and popular in mature markets, like in Europe and US. This makes consumers sensitive to the marketing support and service provided by dealers in China‟s automotive aftermarket. Looking at the competitive environment, competition between auto parts manufacturers is stronger in the OEM

1

(8)

market, where they compete to be the authorized auto parts supplier for car models produced by a car manufacturer. In the aftermarket, auto parts manufacturers are less competitive to each other because their target market is rather fixed and monopolized. Instead, they are facing pressure from local auto parts producers, who provide

unauthorized auto parts to consumers at a much lower price. These unauthorized auto parts called as fake products or gray market, authorized auto parts manufacturers pay a lot of attention on fighting against them and protect their market. In conclusion, service and price are critical in this price sensitive emerging market.

1.2 Company X China

Company X is an automobile lighting products OE supplier in China, as a branch of a European auto parts manufacturer company group. It was established in order to develop the auto parts independent aftermarket (IAM) business in China. Before 2008, Company X's business was to import their lighting products from headquarter in Europe into China‟s market, in order to serve the owners of the compatible imported car models locally. Company X was also responsible for selling its domestically produced auto parts through a network of dealers, however, only in those regions, in which the OEM entities were not covering themselves, mainly in southern and western China. This means that the manufacturing entities and Company X were competitors to a certain extent due to the disorganization inside of the company group. It was only recently, since 2008 (i.e. the year under research) the headquarter determined that only Company X is responsible for the IAM business, i.e. Company X is the only official supplier of company group‟s products to the aftermarket consumers in China. A flow chart below presents the current supply chain of the manufacturer company group X, where our research subject,

Company X, lies in IAM market (Figure 1.1). Given this organizational change,

(9)

Figure 1.1 Industry supply chain for auto part Manufacturer Company Group X

Company X undertakes the full range of products under their own brand, including lighting, electronics, and electrics products, most of which are compatible for premium cars (e.g., BMW, Audi). Through their distribution channels –authorized dealers- consumers can purchase Company X‟s products in 23 provinces and municipalities in China. Dealers are selected based on meeting certain qualifications, and are reviewed every year. They are important for Company X concerning product‟s availability in the market and convenience for consumers. Further, dealers contribute to brand building among end-users, customer service unauthorized products reporting and market

intelligence (e.g., price response from consumers). High-margin headlights account for the majority of Company X‟s annual sales of several million yuan and are the focus of this research. Among the product lines of Company X, headlights are their major source of profit in China‟s aftermarket and will be the focus of this research work.

Auto Part Manufacturer Company Group X

OEM Market _{IAM Market}

OEM Entities

Automobile Manufacturer

Company X (OE Supplier)

Authorized Dealers (Retailers)

Consumers (Car Owners) Supplier

Sup

plie

(10)

Chapter 2 Problem Statement

2.1 Drivers of Dealer’s Demand

(11)

Figure 2.1 Dealer’s demand and locations (3 quintiles in sales value)

High Demand Middle Demand Low Demand

2.2 Dealer’s Price Response

After internal market consolidation since 2008, Company X feels urgency to understand dealer‟s price response. In order to elaborate the problem in this regard, we explain the current price policy first as below.

 Cost-plus principle (see correlation Appendix 4.1)

For the majority of Company X's products, the price asked is equivalent for all dealers across the country, and is set according to the cost-plus principle and the equivalent industry level. Company X purchases from domestic OEM entities and headquarter in Europe for all the products they sell in the aftermarket, which composes the cost for Company X. Since Company X was not the only OE provider in the aftermarket initially, market demand was uncertain and difficult to predict. In order to realize transactional profits in this circumstance, product costs are essential in setting product prices to dealers.

(12)

 Equivalent price to all dealers

Company X offers around 75 percent of their products, in terms of the number of product items, to all dealers at the same price. For the remaining 25 percent of products,

Company X adjusted product price according to dealer‟s overall sales capability. Notably, quantity discount was not applicable to one single purchase/order in order to prevent dealers‟ stockpiling. The equivalent price to all dealers was aiming to support fair competition between dealers.

 Limited control over retail price

Given the same price to all dealers in most cases, Company X allowed dealers to

accommodate local market characteristics in their retail price to consumers, such as local purchasing power and market competition. Company X did not have an explicit control and management on retail prices, and dealers did not report retail sales and price data to Company X.

This price policy raised a couple of problems after the organizational change. Firstly, an increasing number of dealers questioned about the price fairness under current pricing policy. Dealers claimed that current pricing policy disregards the differing consumer perceptions of price in distinctive regional markets. For instance, more premium cars are in use in affluent regions, where brand X is widely known in the marketplace and

enjoying an image of high price and unique product performance. In these markets, product costs were lower for the dealers, as less expense were required to promote the expensive products to consumers. Moreover, it was easier for them to close the deals, as consumers are usually more willing to pay for high prices. Secondly, Company X was at risk to lose a consistent and favorable brand and price image to consumers, as dealers were able to manage their retail price subject to local market conditions. This turned out to be particularly critical when Company X starts to manage the whole aftermarket in China and designs the national sales plan strategically.

In conclusion, in this emerging, region-distinct and price sensitive market, it is especially true for Company X to retain valuable channel partners by realizing a distinguished

(13)

Company X urges to understand dealer‟s demand and price response of their products at different locations in order to set up an optimal price plan in near future, to meet distinct dealer and market demand after organizational change.

2.2 Research Questions

Derived from the problems, we conclude the problem statement as following,

What are dealer’s price responses, accounting for demand heterogeneity across locations and product categories?

To address the problem statement more specific, we propose two research questions. Sub question 1:

What characteristics of dealers influence their demand and price sensitivity at a location?

Sub question 2:

How does dealer’s price sensitivity change to different products?

2.3 Research Objectives and Contribution

Company X‟s management wants to identify the heterogeneity of price responses from dealers. Therefore, our research goal is to develop an implementable model to analyze dealer‟s demand on headlight products, and the change of dealer‟s price sensitivity, by which Company X is able to increases their knowledge on dealers‟ behavior and understand the regional aftermarkets better.

The primary contribution of this article is to provide Company X with a quantitative insight into the determinants of dealers‟ pricing response, and a tool for predicting

(14)

moving consumer goods. Although until recently research on durable goods retailing is available, for instance the optimal pricing study in automotive aftermarket by Mantrala et al. (2006), much less attention has been paid to the pricing in channel management (Rosenbloom, 2004), in terms of one manufacturer/supplier pricing to its multiple

independent resellers. Considering these two research objectives, this thesis will provide a convergence of contribution in scientific disciplines, involving statistical modeling and estimation of channel pricing response, and social practice, providing Company X more insights of their emerging distribution networks in China‟s automotive aftermarket.

(15)

Chapter 3 Theoretical Framework

In this chapter, we provide the findings from previous researches, in three aspects of interest to this thesis. First, we demonstrate some distinctions of retailers‟ perception of price to consumers‟, and optimal wholesale pricing in channel management. Secondly, we focus on the price elasticity of auto part products in retailing. In the third aspect, we demonstrate the relevant determinants of price sensitivity. All these valuable viewpoints provide the fundament of our hypothesis and conceptual model, which are prepared at the end.

3.1 Pricing to Channel Members

Retailer’s Price Perception

Since independent dealers (retailers) are one type of business customers (Vitale and Giglierano, 2002), we first review the literatures on the general differences between pricing in the business-to-business (B2B) market and the business-to-consumer (B2C) market. Price is a deliberate factor to consumer‟s final decision. Other elements like retail location, product assortment, and payment method, may be part of the shopping

experience but are not usually foremost in the decision process (Vitale and Giglierano, 2002). Instead of price, rational business customers predominantly focus on product functionality or performance, and usually evaluate the price accordingly by applying their industry resources or professional knowledge (Anderson et al., 2009).

(16)

Optimal Wholesale Pricing

Company X and its dealers compose a two-level vertical channel, in which one

manufacturer selling through multiple independent retailers. In this section, we review the literatures on the wholesale pricing behavior within this type of channel. Optimal wholesale pricing strategy should maximize whole channel profits, which is termed as “channel coordination” (Jeuland and Shugan, 1983, 1988). Ingene and Parry (1995) question this viewpoint and argue that establishing a wholesale pricing policy that coordinates retailer behavior is generally not profit maximizing for the manufacturer. Rather, the optimal two-part tariff (consisting of a wholesale price and a fixed fee) wholesale pricing policy generates greater manufacturer profits. Moreover, the manufacturer-optimal wholesale pricing is dependent on the retailers‟ fixed costs, the relative size of the retailers, and the degree of inter-retailer competition (Ingene and Parry, 1998, 2000). Additionally, Ingene and Parry (1995) do not recommend non-linear

wholesale pricing schedule due to three reasons. Firstly, complex schedules involve greater administration, bargaining, and contract development costs (Lafontaine, 1990). Secondly, determining optimal quantity levels imposes significant information

acquisition costs on manufactures, especially as the number of retailers increases. Finally, complex schedules may generate negative goodwill, and may even lead to lawsuits.

Based on the pricing to channel members, we decide to review the findings on price response in retailing sector further. Because resellers demands are driven almost entirely by the consumer needs (Rosenbloom, 2004; Blythe and Zimmerman, 2005), and a

significant amount of literatures reveal the impact of product and retailer‟s characteristics on price elasticity.

3.2 Price Elasticity for Auto Parts in Retailing

(17)

“Good” brand to “Best” brand. The three quality variants from one subclass are

compatible to only one automobile model. Their research reveals that the mean own-price elasticity across variants is -2.22, smaller than the -2.5 in grocery retailing sector

(Hanssens et al., 2001) in absolute value. The average own-price elasticities of Good, Better, and Best variants across all subclasses are -1.26, -1.80, and -3.60. The smaller own-price elasticity for cheaper products is due to the favorable price-quality positions of “Good” brand in the perception of consumers. The cross-price elasticity further confirms the importance of price-quality perception on consumer‟s price sensitivity. Mantrala et al. (2006) suggests that for some subclasses, the price reductions on a higher-priced (quality) brand affect a lower-priced (quality) brand more so than the reverse. For these Better-Good product pair (see result 1 in Table 3.1), it is explained that consumers favor the Better variant‟ price reductions, because the quality gap between these brands is larger than their price gap. In some other subclasses (e.g., result 2 in Table 3.1), the Better variants are closer in quality to the Best than to the Good variants, giving the Better variants a positioning advantage relative to both the Good and the Best variants. Given the high price of the Best variants, their quality appears not to be sufficiently superior to the Good variants, giving the latter a positioning advantage relative to the Best. Overall, the ratio of price-quality is essential in determining consumer‟s price sensitivity to auto parts in aftermarket.

Table 3.1 Comparisons of the average cross-price elasticities

(adopted from Mantrala et a.l, 2006)

Result 1: average Eb→G = 1.24 > average EG→b = 1.13,

Result 2: average Eb→B = 1.24 > average EB→b = .49,

Result 3: average EG→B = 1.13 > average EB→G = .49,

Result 4: average Eb→B = 1.24 > average EG→B = 1.13,

Result 5: average Eb→G = 1.24 > average EB→G = .49, and

Result 6: average EG→b = 1.13 > average EB→b = .49.

(18)

3.3 Relevant Determinants of Dealer’s Price Elasticity

A rich literature exists in casting the determinants of consumer‟s price elasticity. The determinants include brand loyalty, product involvement and consumer‟s demographics among others (e.g., Hutchinson et al., 2000 and Bijmolt et al., 2005). The decision to include a determinant of price elasticity to explore dealer‟s price sensitivity is based on three criteria. First, the determinants should represent the auto parts product functionality and performance, instead of factors that related to aesthetics or taste that consumers emphasize (Anderson et al., 2009). Second, we examine the determinants of price elasticity in a trading area/market. Third, we consider the determinants of price elasticity in a store. The following sections discuss the determinants of price elasticity in these three aspects.

Product Characteristics Stage of Product Life Cycle

(19)

margins for serving the older and slower moving parts, accounting for the higher inventory expense of the reseller.

Brand Ownership

In the discussion of consumer‟s price response to manufacturer brand and private label, brand strength is the main concern that differentiates these two brand ownerships. On the one hand, smaller price elasticity is considered for manufacturer brand, because they are well recognized with stronger market positions, and their price are higher than that of private labels (Danaher and Brodie, 2000). On the other hand, Bijmolt et al.‟s (2005) finding imply that the effect of brand ownership on elasticity is not significant, the explanation is when brand differentiation is accommodated as the own price elasticity, manufacturer brands are not necessarily different from private labels (Boulding et al., 1994). Recently, Baltas and Saridakis (2009) conduct a research on the brand-name effects on pricing in automotive market. Their study reveals that automobile prices are determined more by functional characteristics in mainstream segments, but more by brand in high-end segments. Since incremental value is added to a car by its prestige brand name. In addition, high-end automobile brands not only earn brand-name premium but also seize high-margin.

Market Characteristics Climate

Climate is one of the important drivers of failure rates for auto parts, and thus influences consumer‟s price elasticity. Mantrala et al. (2006) apply geographic latitude to

accommodate the effect of climate, and find a positive relation between geographic latitude and price sensitivity. Based on their data from 3,400 retailer stores in the U.S., consumers in the north of U.S. are more price sensitive than in the south. Unfortunately, researchers did not elaborate the relation between climate and price sensitivity further.

Demographics

(20)

The overall store sales are positively correlated with the local potential market and buying power, and socio-demographic variables, like family size, age distribution, income, employment status, education and ethnicity, are the indicators of local buying power and, hence, influence store and category performance (e.g., Dhar and Hoch, 1997). Demographics influence local spending levels and the allocation of resources over

distinguished income-elastic product categories (Campo et al., 2000). They result in distinctive consumer shopping behavior and store selection, by affecting price sensitivity, store characteristics preference, mobility and time cost (Bell and Lattin, 1998). Higher household disposable income reduces consumer price elasticity, because search costs excesses the expected benefits of examining price information for these consumers, so that their motivation and need to look for low prices is reduced (Estelami et al., 2001). Khan and Jain (2005) confirm this finding. In their research, they apply home value, specifically the percentage of houses in the store area valued at more than $150,000, is used to measure the affluence level of the trading area. Consumers are less sensitive to price in affluent areas as well. Product category may moderate the income effect on price elasticity. Categories that call for considerable financial buying power, for instance liquor in Mulhern et al.‟s (1998) research, have a price sensitivity increase with income, since affluent consumers are more likely to respond to price promotions. Further research by Campo et al. (2000) confirms these findings and reveals the location characteristics are particularly relevant for clothing and luxury products in supermarket (e.g.,

audio/video/micro-electronics, household products, and leisure), while much less so for core assortment (groceries and health & beauty care).

Market Competition

Price sensitivity can be reflected in switching among brands of the same product category, between products categories in the same store, and across different stores in retailing. As the number of alternative retailer increases, consumers have more substitution

(21)

alternatives are available in the trading area. Referring to the store performance, Campo et al. (2000) argue a conflicting viewpoint. They find a positive relation between

competition and store sales value in one trading area. The presence of more or larger stores increases the competitive pressures, but is indicative of high economic potential and buying power as well and thus positively related to store performance. In an

aggregate level, which is the overall sales or market share performance for a multi-store retailer or a manufacturer with distribution networks in one trading area, different perspectives and findings are provided. Coughlan et al. (2006) reveals that market response to distribution intensity is likely to be concave for shopping goods. For

shopping goods such as automobiles and consumer durables, consumers may do research about products before purchase. The role of resellers is more than providing convenience and assortment. They provide additional support, such as sales assistance and product demonstrations. Assisting them to cover these costs, suppliers limit intrabrand

competition by restricting the number of outlets in a trading zone. This finding is in line with Bucklin et al.‟s (2008) opinion, dealer intensity increases the intrabrand competition, result in lower levels of dealer profitability due to higher price sensitivity.

Store Characteristics

(22)

estimated price sensitivities across stores. Belonging to one single retailer, smaller sized stores, with a larger concentration in markets are associated with lower sales, due to greater interstore cannibalization. Consumers show less price sensitivity in these smaller stores as well, because competition from other retailers is less in the same region, and the prices at all stores within a given market are similar, so that consumers have no real alternatives to switch to even if the price is increased at a given store.

Differential assortment category is exploited to influence individual retailer store performance in terms of two principal ways. One is distinctive assortment category can develop retailer store a competitive advantage, by offering tailored assortment

combination to fit the local needs (Grewal et al., 1999). The other way is to help retailer to allocating scarce resources available at the store. By providing only appealing

assortment category at store, retailers can arrange the resources more effectively and efficiently, for example promotions, local personnel, and store space (Campo et al., 2000). In consequence, the space allocated to a category can have a positive impact on category performance, because it increases the visibility of the category and attracts more

consumers (e.g., Desmet and Renaudin, 1998). The space can be measured as the absolute space allocated to the category, or the category‟s share in the store‟s total sales surface when accommodating the competition between categories for customer attention.

3.4 Conceptual Model

Although the available literature reviews are mostly in retailing, we apply these findings to propose the conceptual model on retailer‟s (dealer‟s) behavior in this section. It provides an overview on the key relations between product characteristics and dealer‟s price sensitivity, and the impacts of market and store characteristics on dealer‟s demand elasticity that is relevant in this study. A summary of the hypotheses is reported

(23)

Figure 3.1 Conceptual model

Anderson et al. (2009) and Mantrala et al. (2006) both mention the value and the

importance of product functionality and quality performance in business market and price elasticity in automotive aftermarket retailing. Their opinions bring us to consider that the quality of car parts is not only sensible to dealers, due to their business interest, but also critical for consumers to justify the price-quality position of a product. In retailing studies, product cost is not involved in price elasticity estimation, because consumers generally do not have inside knowledge to justify product quality by retailer‟s product cost.

However, for rational business buyers like retailers, who are sophisticated market players in the industry, who usually have sufficient resources and information to justify the value of the brand/product in their perception. In automotive aftermarket, the product cost for OE supplier like Company X, represents the price offered by the manufacturer factory (see Figure 1.1), often indicates the quality, function complexity, and brand image. Retailers (dealers) often have an insight on the cost and the value of a product than consumers, and thus the price sensitivity for different brand/product will be influenced by the cost. In this case, dealers are less sensitive to the price of higher cost products,

because dealers value the higher quality.

(24)

Two key points are concluded from the rich literature concerning the effect of product life cycle on price elasticity (e.g., Parker and Neelamegham, 1997; Simon, 1979). One is the value of product differentiation for consumers, and the other is product/market

competition. Regardless the stage of product life time, price elasticity is low (versus high) when product differentiation value is high for consumers. Besides, higher level of market competition provides consumers more alternatives, and induces greater price elasticity. Recall from section 1.1, automotive aftermarket for OE supplier is technically free of competition, and product differentiation is not important for consumers or dealers, because no alternative is available when there is a product failure. Rather, older products are likely to require longer shelf time before sold, which increases dealer‟s costs and thus increases their price sensitivity.

Hypothesis 1.2 Dealers are more sensitive to the price of longer lifetime products.

Brand ownership is not a relevant issue for Company X at a first glance, whose products are all under their Brand X. However, their products are available for multiple car brands belonging to various automobile segments. For example, Volkswagen (VW) Golf from compact car segment, BMW from premium car segment. To accommodate the

heterogeneity of brand preference of dealers is of interest in estimating price elasticity. Exclusive brands are regarded as status symbols with low function to price ratios (Strach and Everett, 2006). Brand values arise from strong, favorable and unique associations in consumers‟ minds between brand names and products quality (Keller, 1993). When consumers purchase a car part replacement, the value of the offering is regarded to be consistent with its compatible car value. Expensive and premium car owners may willing to pay more for the stronger brands and the expected higher quality, which induce less price sensitivity, and this consumer reaction will influence dealer‟s price sensitivity accordingly.

(25)

There is no clear conclusion of climate impact on car parts price elasticity in previous researches. Company X explains their observation on the climate as following. Colder environment could cause a higher failure rate, in terms of unfavorable working conditions for auto parts and higher possibility of car accident from ice sliding for example. In Mantrala et al.‟s (2006) research, consumers in the north of U.S. are more price

sensitivity, we assume this result is because of the average environment temperature is lower than in the north than in south, this raise the failure rate of auto parts and increase the frequency of purchasing replacement parts for consumers. In consequence, the higher consumption of auto parts makes consumer more sensitive to price in lower temperature seasons. In this research, we consider the climate isolated from the geographic location and propose our hypothesis in temperature, which is more straightforward to explore the climate condition influence on price elasticity.

Hypothesis 2.1 Dealers are more sensitive to price during lower temperature seasons.

Next, we have our attention on how the market characteristics moderate the price sensitivity in dealer stores, by influencing local dealer‟s demand. To begin with, demographics have been proved to be relevant in determining local demand more than once, which are often linked to the purchasing power (e.g., Khan and Jain, 2005; Campo et al., 2000). Here, we select the variables that are considered to be relevant to car parts consumption, and express the relations between the local demographics and dealer store‟s price sensitivity. Compatible car ownership is a special factor that is not available in other researches but suggested by Company X. This factor records the number of cars in use for the specific car part and represents the largest potential market for one car part product. In general, larger car capacity is expected to cheaper cars, because they are affordable by more consumers. The demand of these compatible car parts is also higher. Dealers may be more sensitive to the price as they purchase in great volume and also because of the higher consumer‟s price elasticity.

(26)

Hypothesis 2.3 Dealers are more sensitive to the price of car parts whose compatible car ownership is larger.

Intrabrand competition in one trading area will increase consumer‟s price sensitivity, because of more substitution possibilities for consumers. When dealers are selling

products which need additional marketing support to consumers, cost will increase due to the lower margin. In this case, dealers will become more sensitive to the price as well, to respond to higher competition in the market.

Hypothesis 2.4.a Dealers’ distribution density has negative effect on dealer’s demand.

Hypothesis 2.4.b Dealers are more sensitive to the price in higher distribution density trading areas.

(27)

change dealer‟s price sensitivity accordingly. Extensive assortment stores may show higher price sensitivity, and specialty stores show less price sensitivity, especially on the product category they specialized in.

Hypothesis 3.1.a Dealers’ store size has positive effect on their demand.

Hypothesis 3.1.b Dealers possessing fewer assortments are more sensitive to price.

Hypothesis 3.2.a Dealers’ assortment category share has positive effect on their demand of the particular category.

(28)

Chapter 4 Research Design

4.1 Methodology

Aggregate demand models can either describe market behavior directly, or indirectly through individual behavior models from which outcomes are aggregated to determine market response (Lilien, Kotler, and Moorthy, 1992, p. 672). A directly postulated aggregate response model is applied to aggregate data and has its own component of response uncertainty. The indirectly specified aggregate model with probabilistic

properties is derived from the perspectives of the individual component models (Leeflang et al., 2000). For this study, we apply the directly postulated aggregate response model. We choose log-log demand model (e.g., Reibstein and Gatignon, 1984) instead of the widely used multinomial logit (MNL) model (e.g., Guadagni and Little, 1983; Mantrala et al., 2006) in retailing research. Because the MNL model is applied to reflect the

individual consumer or households‟ preference for brand by capturing their heterogonous sensitivities to marketing variables, and modeling the aggregate response from the

(29)

4.2 Data Collection

Product Data

The data provided for this research compose of one year historical sales figures from Company X to their 43 dealers across 23 provinces and municipalities in China, includes the sales, price and cost (the price asked by OEM entities to Company X) of 84 different headlight products and their compatible car brands. Considering the importance of headlight product for Company X in terms of its profitability and sales volume, we are required to concentrate our research on this product type particularly. Being different from the panel data that is usually applied in the retailing research, this dataset from Company X only records the actual sales transaction information with dealers on a

monthly basis. In another word, for a given product whose sales is zero in a certain month, the according price and cost data are not available in Company X‟s sales system, and thus are not included in our dataset either. For a given product, the sales data are composed of the non-zero sales figure (in units) and the number of orders within the year under

research. The production year of each product‟s compatible car is also provided by Company X. We transform this data into the product age by calculating the discrepancy between the start year of car production and the 2008. This variable shows the product life time at the year of this sales data. Temperature data is collected on the monthly basis for each dealer‟s location from China Meteorological Administration at www.cma.gov.cn.

Dealer Trading Area Data

(30)

product failure. The compatible car ownership, is the accumulated sales volume of the compatible car model, demonstrates the largest possible demand for a certain product. For instance, there are 5,000 units of Audi A6 in use in Shanghai by the end of 2006, representing the market potential for all the Audi A6‟s lighting products in the

aftermarket in 2008. Because new cars are generally sold with one year warranty, with which car owners will not access to automotive aftermarket for car parts replacement, thus the sales of Audi A6 in 2007 is not relevant in estimating the market capacity. We are provided with these car capacity data at the province-level. With regard to the market competition, the number of dealers (stores) in the province that are in direct competition with the product is applied to express the product-specific competition in one trading area.

Company X provided data on dealer (store) characteristic variables as well. The

assortment records the number of product items from Company X that each dealer store carries. This figure does not refer to the assortment depth or the width, since product items in every dealer store are across product categories. Also, dealers‟ assortment category combination is provided, which shows the percentage of each product category in their whole assortments. The specific assortment categories are presented in the next section.

4.3 Product Categories

(31)

Therefore, we classify all the headlight products into four product categories with the help of Company X, according to their price level, functionality, and product age as well as the market segment and price level of their compatible automobiles. The average price of product categories is in an ascending sequence. Table 4.1 presents an index of product categories.

Table 4.1 Categories descriptive statistics

Category 1 2 3 4 Mean Std. Deviation Mean Std. Deviation Mean Std. Deviation Mean Std. Deviation Sales 20.77 41.89 5.10 13.08 5.80 7.06 12.59 31.14 Price 307.31 101.23 640.19 200.94 819.69 206.78 2072.64 887.59 Cost Ratio 0.86 0.16 0.72 0.14 0.76 0.21 0.71 0.16 Product Age 6.74 1.94 16.12 3.55 7.40 3.37 5.50 2.17

Compatible Car Ownership 15,999 14,932 1,725 1,890 4,816 3,302 4,363 3,932

Assortment Depth 72.50 27.32 64.76 31.50 70.26 26.63 59.40 31.54 Category Share_1 0.54 0.23 0.28 0.15 0.35 0.13 0.32 0.47 Category Share_2 0.06 0.06 0.17 0.16 0.09 0.06 0.10 0.08 Category Share_3 0.22 0.12 0.27 0.14 0.32 0.08 0.23 0.17 Category Share_4 0.19 0.11 0.28 0.20 0.24 0.10 0.40 0.26 Distribution Density 1.30 0.46 1.33 0.47 1.36 0.48 1.24 0.43 # of Orders 2.01 0.94 2.54 1.14 3.20 1.91 3.24 1.91 N 505 111 416 466

Remark: Cost ratio is presented instead of cost to facilitate the margin rate comparison (Cost ratio = cost/price)

Category 1

There are 505 observations in this category, representing the number of transactions to all dealers within the one year. The average sales are the highest (20.77 units, Table 4.1) among all the categories, which is partially due to its lowest average price (RMB307.31) and largest car capacity (15,999 units) in the regional market comparing with all the other Company X‟s products. This category is aggregated with two compatible car brands, VW Jetta and VW Bora, both belong to compact cars. The cost ratio is the highest among categories, meaning its margin rate on price is the lowest. Dealers prefer to make fewer orders per year (average 2.01 orders per dealer per product) but larger amount per order, representing a possibility that dealers like to keep more inventory of this product category. Because of its low price and large market capacity, risk is relatively low. Store

(32)

Category 2

This category is composed with the oldest products from Company X, because their compatible cars are imported car models that were produced 16 years ago on average. Therefore, such cars are much less yet in use (1,725 units regionally) and makes this category with the smallest number of observations at 111. On average, the price of these products is RMB640, more than double the price of Category 1. The cost ratio is the second lowest among categories (0.72), meaning a higher margin rate on price. Dealers who buy this category have no specific preference on their assortment categories, because category shares are more or less evenly distributed, from the lowest at 17% of Category 2, to the highest at 28% in .Category 1 and 4.

Category 3

The products in this category are compatible for only one brand of premium cars, Audi A6 with 416 observations. This category has the second largest market potential (4,816 units) and higher price at RMB820. According to the figure of distribution density (1.36), this category is popular in the trading area with more dealers. Dealer who buy this

category are favorable to Category 3 and 1, whose category shares are approximately 35%.

Category 4

(33)

popularity of this subclass products among dealers. Dealers who provide this category have a large assortment share of Category 1 as well. This most expensive product

category has relatively large demand from dealers, because it faces less competition from fake market.

4.4 Multiplicative Demand Model

After reviewing the correlations between independent variables (Appendix 4.1), we find a common strong correlation (>.6) between price and cost across categories, which is a consistent illustration of the cost-plus pricing strategy mentioned in chapter 2. To avoid the potential collinearity and the resulted unstable estimates (Leeflang et al, 2000) if include both variables, product cost is transformed into cost ratio for the model, calculated as cost divided by price.

The research is undertaken in two steps. Firstly, multiplicative demand model is applied to each category. In this way, the brand preference and the effects of product and location characteristics on dealer‟s demand of each category products can be identified. In the four category models, we focus on dealer‟s demand elasticity. Next, we separate each

(34)

Category Demand Model

Multiplicative form:

Sitk = β0 Pitkβ1 CRitkβ2 Akβ3 Tikβ4 β5CBk ICiβ6RCiβ7 NDiβ8 ADiβ9 CS_1iβ10 CS_2iβ11 CS_3iβ12 CS_4iβ13 NOikβ14 εitk

Log - log form:

ln(Sitk) = ln(β0) + β1 ln(Pitk) + β2 ln(CRitk) + β3 ln(Ak) + β4 ln(Tik) + CBk ln(β5) + +β6 ln(ICi) +β7 ln(RCik) + β8 ln(NDi) + β9 ln(ADi)+ β10 ln(CS_1i) + β11 ln(CS_2i) + β12 ln(CS_3i) + β13 ln(CS_4i) + εitk

with i = 1,…, 43 (dealer) store t = 1,…, 12 month k = 1,…, k product

Subgroup Demand Model

Multiplicative form:

Sitk = β0 Pitkβ1 CRitkβ2 Akβ3 Tikβ4 ICiβ5RCiβ6 NDiβ7 ADiβ8 CS_1iβ9 CS_2iβ10 CS_3iβ11 CS_4iβ12 NOikβ13 εitk

Log - log form:

ln(Sitk) = ln(β0) + β1 ln(Pitk) + β2 ln(CRitk) + β3 ln(Ak) + β4 ln(Tik) + β5 ln(ICi) +β6 ln(RCik) + β7 ln(NDi) + β8 ln(ADi)+ β9 ln(CS_1i) + β10 ln(CS_2i) + β11 ln(CS_3i) + β12 ln(CS_4i) + εitk

(35)

Table 4.2 A list of variables in multiplicative demand models

Sitk Sales in units of product k in month t to dealer i Pitk Price of product k in month t to dealer i CRitk Cost ratio of product k in month t to dealer i

Ak Age of product k

Tit Environment temperature in month t in dealer i's province CB_k Dummy variables for compatible car brands

IC_i Median income of the dealer i 's province

RCik Compatible car ownership for product k in dealer i's province NDi Number of dealers in the province dealer i located

ADi Dealer i's store assortment

CS_1i Percentage of Category 1 in dealer i's assortment CS_2i Percentage of Category 2 in dealer i's assortment CS_3_i Percentage of Category 3 in dealer i's assortment CS_4_i Percentage of Category 4 in dealer i's assortment

NOik Number of orders for product k made by dealer i within one year

P rod u ct L oc at ion

We estimate the proposed multiplicative demand models separately by applying the log-log form in SPSS using the ordinary least squares approach, linear regression models will be run to identify the heterogeneous price elasticity and effects of product and location characteristics on dealer‟s demand across categories and subgroups. There are 505 observations in Category 1 model, 111 observations in Category 2 model, 416

(36)

Chapter 5 Results and analysis

5.1 Assumptions

The log-log demand model shares the typical assumptions of the linear regression. There are several assumptions about the disturbance terms should be satisfied in order to ensure the statistical precision of the OLS parameter estimates (Leeflang et al, 2000). In this section, important assumptions will be investigated for our transformed linear regression model. Constraining by the length of this thesis, we typically elaborate the assumption investigation on four category demand models. The same methods are applied to the subgroup demand models as well to test and ensure their assumptions are satisfied.

Assumption 1: Disturbances are normally distributed.

Disturbance term in the log-linear form of the model should be normally distributed for the standard test statistics for hypothesis testing and applied confidence intervals. This assumption may be violated by model misspecification (Leeflang et al, 2000). Through drawing the standardized residual‟s histogram and P-P plot, we could examine the normality of the residuals, and thus validate this assumption. Appendix 5.1 demonstrates these graphs for all the regressions models, in which no significant skewness or kurtosis is shown to suggest deviations from normality. Therefore, we confirm the error terms are normally distributed.

Assumption 2: Disturbance is homoscedastic

(37)

that indicate heteroscedasticity. Leeflang et al (2000) mention that heteroscedasticity occurs especially when cross-sectional data are used for estimation. We further assess the equality of variances in a statistic manner, by classifying the each category‟s data into groups according to the standardized predicted value in regression, and compare the mean of the unstandardized residuals from each group. Appendix 5.3 provides the result of the ANOVA test, which allows us to conclude that no significant difference in the variance has been found between groups, as both p values are larger than 0.05. As a result, homoscedasticity is justified.

Assumption 3: There is no relation between the predictors and distanbance This assumption requires the predictors are nonstochastic or „fixed‟ (Leeflang et al., 2000). The model assumes the price is exogenous, there should be no correlation between price and disturbance term. If this assumption is violated, which means price is actually endogenous, the resulting price elasticity is biased close to zero, since it represents the sales response given demand-side and supply-side reactions (Bijmolt et al., 2005). We examine this assumption by checking the correlation between the error term and the logarithm of price (Reibstein and Gatignon, 1984). A plot of the unstandardized regression residual on the logarithm of price assists in identifying the violation of this assumption. Appendix 5.4 shows no apparent curves in the scatter plot may defy this assumption, additionally the correlations between the estimated unstandardized residuals of the demand function and logarithm of price to be small for all subclasses, and the insignificant correlation (p>0.05) between them (Appendix 5.5) assure this assumption. In other words, it is not necessary to include price endogeneity in the estimation.

Assumption 4: Independent variables are uncorrelated.

(38)

These two variables do not appear a high VIF in the other category models. We check the correlations between predictors in Category 1 and 2 (Appendix 4.1), and find that product age has a high correlation with price that greater than .7 in absolute value (p<0.01) in Category 1. In Category 4, category share_1 is correlated with product age and category share_4 at approximately .6 in absolute value (p<0.01). This could be caused by a lack of data and typical product characteristics in these two categories, so we decide to delete these two variables in the first two category models, and keep them in the other two categories. In the coefficients output (Table 5.6) from SPSS afterwards, no

multicollinearity is identified as VIF for predictors in all models are smaller than 10.

Assumption 5: There is no autocorrelation between disturbance terms.

This assumption is not examined, because it is often applied to the time series data to test whether the error terms are independent (Leeflang et al, 2000). Considering our dataset is not time series data, but only involve discontinuous non-zero sales data, this assumption is not examined.

5.2 Model Reliability

Reliability of Overall Model

(39)

Table 5.1 Category demand model summary Model for Category R R Square Adjusted R Square Std. Error of the Estimate F Sig. 1 .590a .348 .332 1.14711 21.931 .000a 2 _.786a .618 .574 .61739 13.891 .000a 3 .547a .299 .277 .80151 13.212 .000a 4 .727a .528 .510 .93356 29.459 .000a

Reliability of Individual Parameters

Table 5.2 shows the estimated effects of 12 product and location variables and 4 car brand dummy variables in four category models. We summarize the unstandardized coefficient betas, and their p value (Sig.) in each category in one table for easy comparison. In the following short list, we report an overview on the importance of independent variables in category and subgroup demand estimation, by checking their p values (Table 5.2 and Appendix 5.7-15). Any p value that larger than 0.10 is considered to be insignificant.

Product Characteristics

1. Price elasticity: 53 of 67 estimates are significant. 2. Cost ratio: 23 of 65 estimates are significant. 3. Product age: 30 of 49 estimates are significant. 4. Temperature: 43 of 67 estimates are significant. 5. Car brand: no estimate is significant.

Location Characteristics

6. Income: 52 of 67 estimates are significant.

7. Car ownership: 27 of 66 estimates are significant. 8. Distribution intensity: 31 of 59 estimates are significant. 9. Assortment: 30 of 65 estimates are significant.

(40)

For product characteristics, 149 estimates out of 248 are significant (i.e., 60%). And 251 estimates out of 483 are significant (i.e., 52%) for location characteristics. Overall, the selected independent variables have good explanatory power on the sales of Company X to dealers, especially the product characteristics. For location characteristics, more unobserved elements seem to be partly responsible for the observed variation in dealer‟s demand, especially the store characteristics.

Table 5.2 Summary of parameters in category models

Category 1 2 3 4

Unstandardized

Coefficients Beta Sig. Beta Sig. Beta Sig. Beta Sig.

Brand Preference

VW Golf -0.38 0.28

VW Bora VIF>10

VW Jetta VIF>10

Mercedes-Benz 0.31 0.47

Audi Audi 100 Audi A6 Audi A6 (base case)

VW Passat -0.30 0.38

BMW -0.09 0.73

Product Characteristics

Ln (price) -0.37 0.32 -1.93 0.00 -1.00 0.00 -1.03 0.00

Ln (cost ratio) 0.32 0.71 1.44 0.03 0.14 0.00 0.28 0.31

Ln (product age) VIF>10 -2.89 0.00 0.13 0.07 -0.20 0.04

Ln (temperature) 1.39 0.00 -0.17 0.45 0.55 0.00 -0.25 0.00 Location Characteristics Ln (income) 1.36 0.00 1.55 0.01 0.57 0.00 1.13 0.00 Ln (car ownership) 0.28 0.00 -0.07 0.35 0.14 0.06 0.06 0.50 Ln (distr. density) -0.58 0.00 -0.27 0.23 -0.32 0.02 -0.09 0.57 Ln (assortment) 0.04 0.79 0.13 0.64 0.36 0.03 -0.14 0.18 Ln (category share_1) -0.19 0.55 0.47 0.02 -0.34 0.06 -0.10 0.23 Ln (category share_2) 0.15 0.05 0.14 0.65 0.06 0.26 0.01 0.80 Ln (category share_3) -0.30 0.00 0.60 0.04 0.51 0.04 0.22 0.01 Ln (category share_4) -0.34 0.00 0.42 0.24 0.36 0.02 0.14 0.31 Ln (# of orders) 0.35 0.01 0.63 0.16 0.19 0.00 1.00 0.00 (Constant) -18.06 0.00 10.54 0.07 -1.45 0.56 -0.55 0.82 Adjusted R-square 0.33 0.57 0.28 0.51 N 505 111 416 466

a. Dependent Variable: Ln (sales)

(41)

5.3 Parameter Estimation

In the multiplicative demand model, parameters represent 1 percent change of independent variable leads to the according percent change of dependent variable (Leeflang et al., 2000). Noted that the dependent variable is the unit sales to dealers, positive sign of parameter means independent variable has a positive effect on sales. Focusing on the category‟s price elasticity first, three coefficients out of four are found to be significant (p<0.01, Table 5.2). Category 1‟s price is not related with its sales to dealers, as its p value is larger than 0.10 (p=0.32). Among the other three categories, dealers are most sensitive to the price of Category 2. The estimate of -1.93 indicates that 1 percent increase of price leads to 1.93 percent decrease of sales. This effect is almost doubled comparing to the higher priced Category 3 or 4. Category 3 and 4‟s consumers are higher priced car owners, who may less sensitive to the price of their car parts, which induces smaller price elasticity for dealers as well.

Product Cost

In order to compare the price sensitivity of dealers to higher and lower cost products, we divide each product category into two subgroups by the median value of product cost of the category. The first subgroup is composed of lower cost products, and the second subgroup contains the remaining higher cost products from the same category. Table 5.3 presents the price elasticity per subgroup across categories.

Hypothesis 1.1 Dealers are more sensitive to the price of lower cost products.

The empirical results partially support the hypothesis 1.1.

Only in Category 3, price is significant in both product cost subgroups when determining dealer‟s demand. The result in this category rejects the hypothesis. The lower cost

(42)

elasticity (-5.09). The average price elasticity in Category 3 is -1.21 with the cost range from RMB 480.71 to 1002.25. Finally, the smallest price elasticity (-0.83) is for the highest cost products (RMB 948.16 ~ 2991.46) in Subgroup 4.2. Looking at the big picture, dealers are more sensitive to the price of lower cost products, which supports the hypothesis.

Table 5.3 Price elasticity per product cost subgroup across categories Minimum Maximum

1.1: low cost 163 225 Not Significant (N.S) 0.28 285

1.2: high cost 225 531 N.S 0.50 220 2.1: low cost 281 377 -5.09 0.00 56 2.2: high cost 377 724 N.S 0.16 55 3.1: low cost 481 589 -0.65 0.06 200 3.2: high cost 589 1002 -1.77 0.00 216 4.1: low cost 771 948 N.S 0.28 236 4.2: high cost 948 2991 -0.83 0.00 230

Subgroup Product Cost (RMB) Price Elasticity Sig. # of

Observation

(43)

The fragmental results on the comparison of price sensitivity suggest that product cost seems to be not the only indicator of product value, moreover the need of consumers, in other words how profitable a product is for dealers are very important for determining the product value and their price sensitivity.

Product Age

Due to the short of observations in Subgroup 2.2 and 4.2, we did not run the subgroup models because of a lack of degrees of freedom. The resulted unstable estimates are not applicable to the comparison of price elasticity. The results as follows.

Table 5.4 Price elasticity per product age subgroup across categories Minimum Maximum

1.1: short life time 5 7 N.S 0.12 321

1.2: long life time 7 9 18.85 0.00 184

2.1: short life time 11 16 -2.10 0.00 94

2.2: long life time 16 24 N/A 0.00 17

3.1: short life time 1 7 -0.83 0.06 311

3.2: long life time 7 12 -2.40 0.00 105

4.1: short life time 1 7 -1.02 0.00 462

4.2: long life time 7 11 N/A 0.00 4

Subgroup Product Age (Year) Price Elasticity Sig. # of

Observation

Hypothesis 1.2 Dealers are more sensitive to the price of longer lifetime products.

The empirical results support the hypothesis 1.2.

(44)

average price elasticity of Subgroup 3.1 and 4.1 is -0.93, this represents much lower price sensitivity when dealer purchase products less than 7 years old.

In category models (Table 5.2), product lifetime appears to have negative effect on sales to dealers in Category 2 and 4 (both p<0.05), but positive effect on sales in Category 3 (p=0.07<0.10). Dealers‟ demand for shorter lifetime products are much stronger when they purchase Category 1, as its coefficient is -2.89 (Table 5.2), much greater than -0.20 for Category 4. In comparison, dealers prefer older products from Category 3.

The general estimated life time cycle of a car in China is 10 years. Product with life time that longer than 7 years are more likely in their decline stage. The older the product is, the higher the possibility is that dealers have higher inventory cost due to the unstable and small consumer market (Moore, 2006). For instance, according to the dataset, there are 311 orders from dealers for Subgroup 3.1, and only 105 orders for Subgroup 3.2 (Appendix 5.8). In total, 1865 units of products from Subgroup 3.1 are sold, while only 547 units from Subgroup 3.2. This result may also induced by the price sensitivity of consumers to the products at later stages.

Brand Preference

Car brand is a discrete variable. Each subgroup contains the products that compatible to the particular one car brand. Because of the low demand of products for VW Golf, VW Passat and Mercedes-Benz in Category 4, separate models for these brands are not available because of the small number of observations. Table 5.5 (see next page) shows the price elasticities of the available brands.

Hypothesis 1.3 Dealers are less sensitive to the price of premium car brands.

The empirical results support the hypothesis 1.3.

(45)

be a differential brand preference in regional markets. As business customers, dealers consider all direct and indirect costs and the economic benefits of selling a specific car part, the profit is much more important than the brand of a product. Even though, consumers in end market emphasize on aesthetics or taste (Anderson et al., 2009) could make them prefer one brand than another, this brand preference is very possible to be different across regions, which influences dealer‟s brand preference as well.

Table 5.5 Price elasticity per car brand across categories

1 Jetta 15.25 0.00 336 Bora 5.86 0.09 169 2 Audi 100 -1.93 0.00 111 3 Audi A6 -1.00 0.00 416 4 Audi A6 N.S 0.44 319 BMW -1.43 0.00 120 # of Observation Price Elasticity Sig.

Car Brand

Understanding the Regional Demand and Price Response of Dealers in China Automotive Aftermarket A Case Study for Original Equipment Supplier