THE AUTOMOBILE INDUSTRY:
PRICING THE TRIM LEVELS
By
SVEN STEIJN
Supervisor
TUDOR BODEA
University of Groningen
Faculty of Economics and Business
Msc. Supply Chain Management
ABSTRACT
Volvo cars wants to review other pricing methodologies for their trim levels (large packages of single options for a car). The possibilities of applying hedonic pricing as a trim level pricing method are reviewed by studying the following hypotheses: Hedonic pricing applied to the trim levels of Volvo provides different values for the trim levels and attributes as compared to what is currently used by Volvo. Trim levels of Volvo, Audi and BMW in the important European markets are the input for the hedonic pricing regression model. This model uses a linear regression on the physical characteristics of the trim levels to explain the price. The results show that a small number of characteristics can explain the majority of the price. The values for these characteristics differ from the values currently applied by Volvo and therefore the hypotheses is accepted. A decision support tool that applies the values found in this research is designed. This tool supports new trim level pricing decisions at Volvo cars.
Key words:
- Price management - Automobile industry - Hedonic pricing - Trim levels
Supervisor: Tudor Bodea
Second supervisor: Stuart X, Zhu
TABLE OF CONTENT
ABSTRACT _____________________________________________________________________________ 2 TABLE OF CONTENT ___________________________________________________________________ 3 1. INTRODUCTION ___________________________________________________________________ 4
1.1INTRODUCTION TO THE TOPIC ___________________________________________________________ 4 1.2TOPIC _____________________________________________________________________________ 7 1.3UNIT OF ANALYSIS ___________________________________________________________________ 8 1.4THE GOAL __________________________________________________________________________ 8 1.5RESEARCH QUESTION _________________________________________________________________ 8 1.6VISUALIZATION OF THE RESEARCH DOMAIN ________________________________________________ 9 1.7PRACTICAL AND SCIENTIFIC CONTRIBUTION ________________________________________________ 9
2. THEORETICAL BACKGROUND ____________________________________________________ 10 3. METHODOLOGY _________________________________________________________________ 13
3.1WHAT DATA TO INCLUDE _____________________________________________________________ 13 3.2REQUIRED DATA ____________________________________________________________________ 14 3.3RESEARCH DESIGN __________________________________________________________________ 15 3.4THE VARIABLES _____________________________________________________________________ 17 3.5DATA COLLECTION __________________________________________________________________ 19 4. DATA ANALYSIS __________________________________________________________________ 20 4.1DATA ISSUES _______________________________________________________________________ 20 4.2VARIABLE SELECTION PROCESS _________________________________________________________ 20
5. RESULTS _________________________________________________________________________ 24
1. INTRODUCTION
1.1 Introduction to the topic
Warren Buffet once said: “The single most important decision in evaluating a business is
pricing power. If you have got the power to raise prices without losing business to a competitor, you have got a very good business. If you have to have a prayer session before raising the price by 10%, then you have got a terrible business” (Hinterhuber & Liozu, 2012).
Warren Buffet has done quite well for himself and it might be a good idea to do something with his statement. Buffet however, is not the only one that argues that pricing is of great importance. The average income statement of the global 1200 (an aggregation of 1200 large, publically held companies from around the world), shows how quickly the right price can create profit. The example shown in Figure 1. uses the 5 year average of this income statement to reduce sensitivity to yearly economic variations (Baker, Marn & Zawada, 2010).
Figure 1. The power of 1 Per cent (Baker, Marn & Zawada, 2010)
Moreover, as a profit increasing method, pricing has multiple advantages over cost reductions. First, pricing requires almost no investments. Second, pricing is much quicker to implement. Finally, the profit advantage is higher as compared to the cost reduction profit advantage or increased volume profit advantage (Simon, Butscher, & Sebastian, 2003). Several other studies show the value of setting the prices right (Biller, Chan, Simchi-Levi, & Swann, 2005; Curtis & Wright, 2004; Hinterhuber & Liozu, 2012; Marn, Roegner, & Zawada, 2003; Marn & Rosiello, 1992; Verboven, 1999).
On the other hand, pricing methods can also have disadvantages. When a firm tries to optimize its discounting methods the danger of starting a price war exists (Baker et al., 2010; Hinterhuber & Liozu, 2012). There are examples of price wars within the automobile industry (Hinterhuber & Liozu, 2012). To avoid the danger of starting price wars this research is not focussing on discounting but on value. However, focussing on finding value to the customer comes with difficulties. Among the difficulties is the gathering of the required data on different aspects. Aspects such as willingness to pay, the size of market segments and price elasticity (Hinterhuber & Liozu, 2012). The hedonic pricing method applied in this research only requires data on the price and content of trim levels.
possibility to select single options. These single options are necessary to satisfy the specific needs of the customer. An example of such a single option is a navigation system, in the case that this is not included in the trim level preferred by the customer. As a result, a trim level is a package of different automobile characteristics that are preferred by many customers.
The following central concept in this paper is the concept “hedonic pricing”. Hedonic pricing is a term used in economics to estimate demand or value (Baltas & Saridakis, 2009; Cowling & Cubbin, 1972). This estimation of value has been used in the automobile industry before, for example by Cowling & Cubbin (1972). “Virtually all the variation in prices among
different models of cars could be explained by observable physical characteristics” (Cowling
& Cubbin, 1972). Where the study of Cowling and Cubbin (1972) explained the base car price, this study explains the trim level price. In other words, hedonic pricing is a method to explain the differences in price by assigning values to the characteristics of the product. The following definition for the hedonic price of a characteristic is applied in this research. “The
hedonic price of a characteristic is defined as the derivative of the product price in relation to the corresponding attribute.” In other words, it is the value that consumers attribute to a
supplementary unit of the characteristic (Le Gall-Ely, 2009).
The following example from Volvo cars is used to explain what hedonic pricing is. The trim levels of a car consist of different characteristics. These characteristics can be many things, such as wheels, leather seats, a slightly different exterior or sports interior. These characteristics possess a certain value to the customer (Baltas & Saridakis, 2009). This value is not necessarily the same as the price for the single option. Hedonic pricing is a method used to determine the value of a characteristic by defining it as a part of the whole product. For instance a trim level with a price of 2660 Euros can have the following characteristics: 18 Inch wheels, Leather seats and Interior sports design. These characteristics have to possess a certain value that adds to 2660 Euros for the whole trim level. For instance: 18 Inch wheels 560 Euros, Leather seats 1100 Euros, Interior sports design 1000 Euros. These values are found by a regression model which combines sales prices with the characteristics of the product. The regression model is described in more detail in the methodology section of this paper.
Saridakis, 2009). Applying attributes with this low cost/ high value combination more often can be good for profits since these attributes have higher profit margins.
1.2 Topic
The topic of this research is applying hedonic pricing to the trim levels of automobiles. This topic is particularly interesting since the field is relatively new. No academic research seems to have been conducted with respect to hedonic pricing in combination with automobile trim levels. Several authors use hedonic pricing methods in combination with price analysis of automobiles (Bhowmick, 2001; Cowling & Cubbin, 1972; Hulten, 2003; Reis & Santos Silva, 2006). On the other hand, none of these authors apply the hedonic pricing theory to trim level pricing. Alternatively, the base car version was always used for the analysis. Over the last two decades, virtually all manufacturers of automobiles started to offer trim levels for their cars. For this reason, applying the hedonic pricing theory to trim levels is interesting and relevant.
A customers decision to buy a car with or without a specific trim level is mostly determined by price, or value. Because of this, the question one should logically ask is: what is the customer receiving for the stated price? The price should represent the value offered to the customer. The underlying theory that this research uses to find this value is hedonic pricing. The theoretical foundation of hedonic pricing originates from Lancaster (Lancaster, 1966; Lancaster, 1971). Lancaster came up with the idea that characteristics (or attributes, or components) of the product, are the source for the demand (Lancaster, 1966). The characteristics form the reason to purchase the product. Therefore, all characteristics have a price attached to them, this price is called the implicit price. This way of reasoning leads to a price for the whole product (in this case the trim level). This price is the sum of implicit prices from the product characteristics (Bhowmick, 2001).
theory in his study. As a result, this study focusses on the approach used by Bhowmick (2001). This approach is explained in the methodology section.
1.3 Unit of analysis
As stated before, this study wants to improve the trim levels and the trim level pricing of Volvo cars. This is done by obtaining more knowledge on the trim level attributes and their prices. Moreover, this is done by obtaining knowledge on what attributes can be placed in a trim level. This results in knowledge on which attributes or characteristics lead to a high value trim level. The trim levels from Volvo and their competitors (Audi and BMW) are the input to gain this knowledge. The unit of analysis of this research is the following: The trim levels of the Volvo V40, V60, V70, XC60 and the corresponding models and trim levels from Audi (A3, A4, A6 and Q5 trim levels) and BMW (1 series, 3 series, 5 series and X3 trim levels). Involving the data of competitors is important. By doing involving the competitors, a more accurate view is obtained. A more accurate view on what the values for the price determining characteristics are.
1.4 The goal
The aim of this research is to come up with accurate values for the price determining attributes of a trim level. These values can be used for future trim level creation. With these accurate values for the important attributes the future trim level prices can be set. Moreover, high profit combinations can be found. These are the combinations where high values are found for attributes that have a low cost price. These attributes can be used more frequently in new trim levels, which will lead to a higher trim level profitability (Baltas & Saridakis, 2009).
1.5 Research question
The research question that provides the basis for this research is as follows:
What are the effects of applying hedonic pricing to trim levels of the premium car manufacturers in the European markets?
The hypotheses that are formulated to be able to answer the research question are as follows:
H0: Hedonic pricing applied to the trim levels of Volvo leads to the same value for the trim levels and attributes as already used by Volvo.
1.6 Visualization of the research domain
Figure 2. Visualization of the research domain
In Figure 2. we indicate where the values for the attributes found with the hedonic pricing method are used. The found attribute values are used to value the attributes of new trim levels. Valuing the attributes leads to valuing the new trim levels. This value for the new trim level is incorporated in setting the new trim level price.
1.7 Practical and scientific contribution
This research contributes both practice and science. The contribution to practice (Volvo cars) is that hedonic pricing provides another method for composing and pricing the trim levels. Furthermore, a decision support tool is delivered. This tool enables the employees to add attributes to a trim level, the tool then returns a value for the trim level. This value can be used as a reference point for the final price that is set for the trim level. This approach may lead to changes in the trim levels and possibly to a higher profitability for the Volvo trim levels. The contribution to science is found in filling a gap in the literature. As mentioned, several authors use hedonic pricing methods on different industries, the automobile industry is one of these industries. However, none of these authors apply hedonic pricing to trim level pricing (Bhowmick, 2001; Cowling & Cubbin, 1972; Hulten, 2003; Reis & Santos Silva, 2006). Therefore, the contribution to science is that hedonic pricing is applied to trim level pricing. It is researched whether or not hedonic pricing can be applied to trim levels. In addition, it is researched whether or not this new trim level valuing method will lead to new insights.
Trim level
price
Attributes
2. THEORETICAL BACKGROUND
The automobile industry has always consumed a lot of attention from research. This attention started with the birth of the ideas of Henry Ford, in the early 1900’s. In this chapter an overview of the relevant literature for this research is given.
In order to solve a pricing issue, different kinds of problems have to be overcome. Some of these problems are obtaining large amounts of data and finding consumer insights. Another problem to overcome is the wrong mind-set. The mind-set should not be that prices are set by the market. A difference can be made and influencing the price is possible (Baker et al., 2010).
Figure 3. Market strategy, customer value, transactions, pricing infrastructure (Baker et al., 2010)
In Figure 3. is shown what types of decisions are made in what stage. The figure is shown to indicate in what area this research operates. The area in which this research operates is the Customer Value area. In this area the optimal price/ benefits ratio is studied. (Baker et al., 2010)
hedonic pricing. For these reasons, the study of Bhowmick (2001) provides a starting point for the methodology of this research. This study focusses on finding the value that is attributed to different characteristics of trim levels.
Hedonic pricing does not provide the willingness to pay, it rather provides information on elements of the offer that are valued by the customers (Desmet & Hendaoui, 2000). “The
hedonic price of a characteristic is defined as the derivative of the product price in relation to the corresponding attribute.” In other words, it is the value that consumers attribute to a
supplementary unit of the characteristic. Therefore, if the result is zero the characteristic is perceived by the customers as to have no value (Le Gall-Ely, 2009).
Price hedonics, as a statistical technique, already exists for over 80 years (Hulten, 2003). Within this period the theory has been improved by different authors. These improvements are applied in Bhowmick (2001) and in the article of Hulten (2003). This study applies this improved theory used in Bhowmick (2001) and Hulten (2003).
3. METHODOLOGY
As mentioned before, the work of Bhowmick (2001) and Lancaster (1966) has provided the basis for hedonic pricing in the automobile industry. Where Lancaster (1966) provided the basis for the method of hedonic pricing, Bhowmick (2001) provided the application of the theory to real data in the automobile industry. The methodology of Bhowmick (2001) is taken as a starting point for the methodology of this research and is modified where necessary. In this way the methodology is aligned with the goal of this research and the data available to perform this research. Moreover, the methodology is supported by the work of others, which improves the reliability.
3.1 What data to include
because of this, countries for which Volvo has the most detailed data available. The most detailed data for Volvo and for Audi and BMW. Audi and BMW are the most important competitors of Volvo. Both competitors are active and relevant players in the selected countries. All the selected countries have markets that are well comparable. Therefore the results of the research are more reliable and applicable.
3.2 Required data
In order to get to the goal of values for attributes, large amounts of data are required. Below we have stated the data that are gathered to conduct this research.
First, the trim levels from Volvo and from the competitors are mapped. The data required to do this are obtained from the most recent (2nd half 2012 and 1st half 2013) list prices of the models V40, V60, V70, XC60, (Volvo) A3, A4, A6, Q5, (Audi) 1 series, 3 series, 5 series and X3, (BMW) from The United Kingdom, Germany, France, Italy, Sweden, Spain, Belgium and The Netherlands.
Second, the sales data for the cars and trim levels ordered on those cars are gathered. With the cars we refer to the Volvo V40, V60, V70, XC60, the Audi A3, A4, A6, Q5, the BMW 1 series, 3 series, 5 series and X3 from the The United Kingdom, Germany, France, Italy, Sweden, Spain, Belgium and The Netherlands. The historic sales data for the most recent periods that are available for all brands and models are used. These most recent data consist of the third and fourth quarter of 2012 and the first quarter of 2013. With these data the trim level mix (what trim level of the available trim levels for a car is chosen by what percentage of the customers) by brand, model and trim level are derived. These mix data are used to add weights to the attributes. It is however determined that the mix data are not appropriate and not reliable enough to use for the final results in this research. In the analysis part of this paper more depth on this decision is provided.
in the regression. The results from the performed regressions help to select the final set of variables which determine the value of a trim level.
In short the data to gather consists of the following:
- The trim levels, their price and their content (list prices, real data) - Sales data as specific as possible (per trim level, external research data)
- List of key attributes (as a starting point for the regression, knowledge from experience)
To obtain the information, publicly available data are used for the list prices. The trim level sales data and the list of important attributes are obtained from Volvo cars.
3.3 Research design
The model used is based on the Bhowmick (2001) regression model. Adjustments are made where necessary to make the regression model applicable to trim levels.
The starting point is that the differences in price for trim levels have to be the result of differences in qualities. These differences in qualities are, as we like to state, the characteristics of the trim level. Therefore, the price of the trim level, “P”, can be written as a function of a set of characteristics or attributes “X” (Bhowmick, 2001). The equation then looks as follows:
P = f(X1,X2,X3,X4,X5,…Xn)
The implicit price of each characteristic is estimated by the basic regression equation from Bhowmick (2001), which is as follows:
P = b0 + b1 X1 + b2 X2 + b3 X3 + bn Xn + b16 X16
Explanation of the symbols:
p The trim level price (Price excluding local tax rates and in Euro’s)
b0 Intercept term (a constant)
X1 Brand BMW
X2 18 Inch wheels (dummy variable, 1 if present, 0 if not present)
X3 Xenon headlights (dummy variable, 1 if present, 0 if not present)
X4 Window lists chrome/ High gloss (dummy variable, 1 if present, 0 if not present)
X5 Folding mirrors, electric (dummy variable, 1 if present, 0 if not present)
X6 Park assist (dummy variable, 1 if present, 0 if not present)
X7 Interior sports design (dummy variable, 1 if present, 0 if not present)
X8 Sports seats (dummy variable, 1 if present, 0 if not present)
X9 Cloth/ leather upholstery (dummy variable, 1 if present, 0 if not present)
X10 Leather upholstery (dummy variable, 1 if present, 0 if not present) (Leather upholstery can only be present if Cloth/ leather upholstery is present as well)
X11 Navigation system (dummy variable, 1 if present, 0 if not present)
X12 Navigation system pro (dummy variable, 1 if present, 0 if not present) (Navigation system pro can only be present if Navigation system is present as well)
X13 Climate control (dummy variable, 1 if present, 0 if not present)
X14 Power seats with memory function for the position of the seat (dummy variable, 1 if present, 0 if not present)
X15 Multiple characteristics: Smart key, Connectivity, Sound system improvements (dummy variable, 1 if present, 0 if not present)
3.4 The variables
The included variables, X1, X2, X3, … X16, are the result from the initial variables that are considered to be included in the final regression. These variables that are considered to be included in the regression are stated below.
Brand Audi
Brand BMW
Brand Volvo
Segment 1 (Volvo V40/ BMW 1 series/ Audi A3)
Segment 2 (Volvo V60/ BMW 3 series/ Audi A4)
Segment 3 (Volvo XC60/ BMW X3/ Audi Q5)
Segment 4 (Volvo V70/ BMW 5 series/ Audi A6)
16 Inch wheels
17 Inch wheels
18 Inch wheels
Sport-urban exterior (minor exterior differentiations compared to the base car)
Exterior sports design (major exterior differentiations, special sports oriented)
Sports suspension
Front fog lights
Xenon headlights (high quality very bright headlights, type Xenon)
Window lists chrome/ Roof rails chrome (represents chrome lists around the windows, or chrome roof rails or high gloss black parts around the windows)
Window lists chrome/ High gloss
Roof rails chrome
Tailpipe chrome
Folding and dimming mirrors (electrically folding exterior mirrors and auto dimming interior mirror)
Dimming inner mirror
Folding exterior mirrors, electric
Cruise control
Park assist (park help, indicates the distance at the back or front of the car to the nearest object)
Damper settings (allows the driver to select different driving modes for suspension and steering, e.g. comfort, sport, sport +)
3 spoke steering wheel leather
4 spoke steering wheel leather
Interior headlining (the inner roof in a different colour and/ or material as compared to standard)
Interior sports design (interior differentiations, the steering wheel, different stitching, badges and other interior special sports differentiations)
Sports seats (front seats with better support)
Textile/ T-Tec upholstery (T-Tec is more advanced than textile but not the same as leather)
Cloth/ Leather upholstery (mixture from textile and leather upholstery)
Leather upholstery (full leather covered seats)
Inlays (a different design for the interior inlays)
Colour board computer
Navigation system
Navigation system Pro (advanced navigation system with a better looking display of the map and easier to use)
Climate control
Climate control multi-zone (temperature settings for driver and passenger side separated)
Bluetooth phone (connects the phone with the multimedia system of the car by Bluetooth)
Ambient lighting package (interior lights to create a better atmosphere)
Connected drive (internet connected and/ or automatic emergency calls)
Light and rain sensor (windscreen wipers switch on automatically when it starts raining and lights switch on automatically in the dark)
Storage package (provides storage compartments that are separated and easier to use)
Alarm (anti-theft alarm of a higher security class)
Improved Audio system (any improvement made to the sound system and speakers)
Metallic colours
Power seats with memory (electrically adjustable seats with a memory function for different positions)
Heated seats
Electric trunk (automatically closing/ opening trunk)
Direction headlights (the headlights turn with the bend for a better view on the road)
Smart key (a key that does not have to be turned to ignite the engine, instead a button needs to be pressed) + Connectivity (Automatic emergency call in case of a crash + Sound system improvements (Audio improvements)
Comfort seats + Panoramic sunroof (an overhead, able to open glass roof)
These variables are considered since they are able to represent all the different items found in trim levels. With the available data, not all of these variables can be applied. Moreover, not all variables can be included in the regression at the same time. This is because some of the variables can overlap. In the analysis part of this paper the full explanation of why variables are in- or excluded is given.
The variables in this research are determined regardless of the previous research from Bhowmick (2001). The research of Bhowmick (2001) analyses whole cars, where this research analyses trim levels. In trim levels different attributes are found. Moreover, the importance of the attributes differs from the whole car attributes.
563. This indicates that selecting the attribute 18 Inch wheels raises the estimated price by 563 Euros. With this method the sum of parts and the intercept value will represent a total value. The addition of 18 Inch wheels to the trim level therefore accounts for 563 Euros of that total value. In other words, the addition of 18 Inch wheels accounts for 563 Euros of the total price of the trim level.
3.5 Data collection
4. DATA ANALYSIS
Several issues in the data have to be solved in order to come to accurate, reliable and valid results. In this section we explain how these issues are solved. In addition is discussed how the variable selection process takes place.
4.1 Data issues
The first issue is to decide what data to include. In total there are 295 trim levels found in the researched countries and all are included in the original dataset. However, some trim levels are incomplete. The true price of these trim levels cannot be completely determined since the pricing is so unclear or un-transparent (most likely done to have the potential customers come to the car dealer before they can determine the actual price). This unclear pricing only occurs in rare occasions. For this reason 8 trim levels are excluded. The remaining 287 trim levels are used in the final dataset as input for the regression.
The second issue is the decision to apply periods or not. To increase the amount of observations the use of more periods is considered. Since the original plan was to include the trim level mix data as well, the use of more periods could be justified. However, the reliability of the mix data turned out to be below acceptable standards. Moreover, for 93 of the 287 data points there was no mix data available, due to the poor quality of the external research data. Furthermore, the way in which the mix data could be formulated was not appropriate for this research. The trim level availability differs across countries and some countries are much larger in terms of car sales volumes than other countries. For this reason, the weighting of the sales volumes in relation to the trim levels researched becomes an increasingly difficult task, perhaps an impossible task. Additionally, when a new trim level is created with the help of the decision support tool, the mix for that new trim level is unknown. Finally, an increase in price leads to a decrease in mix, simply because customers are restricted by their budget. Therefore, the mix data does not represent what we would like to measure. For these reasons, the decision is made to exclude the mix data from this research. As a result, the original data points (287 trim levels) are used as input for the regression.
4.2 Variable selection process
methods and ranking with the help of collinearity (Guyon & Elisseeff, 2003). However, none of the available methods are conclusive (Guyon & Elisseeff, 2003). The most important criteria for selection is to have significant results. Variables that do not significantly contribute to explaining the price, the insignificant variables, should be changed or removed. Moreover, there are values that most likely represent something else as what was intended to be measured. These are called the “strange” values. These strange values can be caused by multicollinearity. Multicollinearity is analysed by the VIF values (Mansfield & Helms, 1982). Variables with VIF values over 5 are considered to be excluded or changed (Mansfield & Helms, 1982). Furthermore, some of the variables that have not led to significant results are removed from the regression. The following variables are excluded for these reasons: Exterior sports design, Sport-urban exterior, Armrest, Tailpipe chrome, 16 Inch wheels, Sports suspension, Light and rain sensor, Front fog lights, Folding and dimming mirrors, 4 Spoke steering wheel leather, Interior headlining, Textile/ T-Tec, and Colour board-computer.
special variables (Comfort seats and Smart key) this is not the case. The attributes are found together in the trim levels, but the attributes that are included are of a slightly different nature. The other groups are formed of attributes that have the same goal or characteristic. For instance the Interior sports design group. This variable Interior sports design, groups different attributes like a sports steer, different sports stitching and other interior sports parts. Grouping attributes with a different goal is not preferred. In the case of grouping attributes with a different goal, the outcome represents different characteristics. This is not a problem when the characteristics are always applied in a trim level at the same time. However, when a new trim level is composed, these characteristics may be split up. For instance, only Panoramic sunroof is included in the trim level and Comfort seats is excluded. With these groups (Comfort seats and Smart key) this separated value cannot be shown. The other groups represent the same characteristic, therefore they are less likely to be split up. For this reason, groups are normally only formed when the attributes represent the same characteristic.
Moreover, three variables are split up. This is done after it was found that these variables would not result in significant values in any of the regressions. The variable “Chrome/ High gloss window lists and Chrome roof rails” is split up into “Chrome roof rails” and “Chrome/ High gloss windows”. “Folding/ dimming mirrors” is split up into “Folding mirrors, electric” and “Dimming interior mirror”. “Navigation” is split up into “Navigation” and “Navigation Pro”. Splitting up these three variables into six new variables has improved the measurements. Significant results can be obtained for these new variables.
Now that the problems involved at analysing the data are solved, the final selection of variables is started. Of the previously stated list of variables (chapter 3.4), the following variables are given priority to be included in the final results: The Brands, The Segments, Navigation, Leather, Climate control, Xenon headlights, Park assist and 18 Inch wheels. These are the variables that a sales employee, a market intelligence employee and a pricing manager from Volvo Cars pointed out as the most important or core attributes. This selection is a reduced selection from the attributes that are listed by Volvo as important attributes. Values for these variables are likely high and the variables are likely important in explaining the trim level price.
examined. Validating with previous research is not possible since previous research with hedonic pricing in this trim level area does not exist.
5. RESULTS
5.1 Regression results
Table 1. shows a good overview of how important only a small part of all variables are in estimating the price of a trim level. 58% Of the price can be explained by only 10 variables. Although some are insignificant, this indicates how important a small portion of the characteristics are in determining the value of a trim level.
Variable
4 variables 10 variables Coefficients Significance Coefficients Significance
Intercept 4058,062 0,000 1863,727 0,000 Brand Audi -488,103 0,006 Brand BMW Brand Volvo 168,891 0,425 Segment 1 (V40/1series/A3) -1542,815 0,000 -433,584 0,065 Segment 2 (V60/3series/A4) -1194,741 0,000 -712,600 0,001 Segment 3 (XC60/X3/Q5) -466,062 0,138 -410,722 0,075 17 Inch wheels 1195,907 0,000 18 Inch wheels 1319,640 0,000 Navigation 1765,394 0,000 Climate control 898,156 0,000 R² 0,103 0,580
Table 1. A small number of variables
Variable ''Automatic modelling'' Coefficients Significance Intercept 941,177 0,000 Segment 3 (XC60/X3/Q5) 272,063 0,043 18 Inch wheels 445,404 0,013 Xenon headlights 537,968 0,000 Window lists Chrome/ High gloss 307,891 0,005 Folding mirrors, electric 481,078 0,002 Park distance control 457,062 0,000 Interior sports design 1033,371 0,000
Sports seats 510,005 0,003
Cloth/ Leather upholstery 412,876 0,019 Leather upholstery 646,062 0,002
Navigation 657,109 0,000
Navigation Pro 961,713 0,000 Climate control multiple zones 923,277 0,000 Ambient lighting package 262,919 0,035 Storage package 372,254 0,086 Power seats with memory 874,205 0,000
Heated seats 440,183 0,094
Smart key + Connectivity + Sound improvements
1096,281 0,054 Comfort seats + Panoramic sunroof 4137,029 0,000
R² 0,819
Table 2. Automatic modelling results
The expectation was that Segment 4 (representing Volvo V70, BMW 5 series and Audi A6) or Segment 1 (representing the Volvo V40, BMW 1 series and Audi A3) would return a significant coefficient before Segment 3. This because these segments are the segments at each end of the range. From the remaining variables that are returned, all make sense, except for climate control multiple zones. Climate control multiple zones seems to have a value that is unacceptably high. This value could reduce or improve when other variables are applied in the regression, variables that are more likely to represent this higher value.
the statistics programme SPSS. After this time consuming process of optimizing, the final results are found.
Variable Results Coefficients Significance Intercept 918,679 0,000 Brand BMW 295,448 0,014 18 Inch wheels 563,081 0,001 Xenon headlights 584,124 0,000
Window lists Chrome/ High gloss 261,975 0,024
Folding mirrors, electric 367,129 0,018
Park distance control 642,881 0,000
Interior sports design 1017,984 0,000
Sports seats 518,118 0,003
Cloth/ Leather upholstery 400,688 0,024
Leather upholstery 723,893 0,001
Navigation 634,247 0,000
Navigation Pro 1071,329 0,000
Climate control 313,061 0,023
Power seats with memory 992,821 0,000
Smart key + Connectivity + Sound improvements 1295,905 0,021 Comfort seats + Panoramic sunroof 4101,697 0,000
Number of observations 287
R² 0,806
Table 3. Final results
cannot. In addition, some important variables such as Navigation pro, 18 Inch wheels and Power seats with memory are found less often in the segments of the smaller cars. These variables are found more often in the segments of the larger cars and in the more expensive trim levels. This explains that segmented values could exist before other variables were applied. It can also explain that after the other variables are applied, the segments turn out to be insignificant. As a result, the segments for trim levels are excluded.
Segmented values are known to exist in the automobile industry for the base car prices. Therefore a check is performed on this data set. The full car segments should be found when the full car price (instead of the trim level price) is explained in the regression. The results of this check are shown in Table 4.
Variable
Full car pricing segments Coefficients Significance Intercept 39753,462 0,000 Segment 1 (V40/1series/A3) -15514,734 0,000 Segment 2 (V60/3series/A4) -7069,005 0,000 Segment 3 (XC60/X3/Q5) -226,649 0,762 Number of Observations 287 R Squared 0,693
Table 4. Full car pricing segments
As can be seen from Table 4. Full car pricing segments, segments are found in this data set. This table is the result of the linear regression with the full car price as the dependent variable and the segments as independent variables. Segment 3 and Segment 4 are very close to each other and therefore this leads to no significant difference (Segment 4 is not shown in the table since Segment 4 is used as the reference group). In the price lists, Segment 3 and Segment 4 are indeed close and often comparable in price. Therefore, the outcome is of this test is acceptable. Since we have proof that segments in full car pricing exist, it was expected that this would hold for trim level pricing as well. If this would be the case, different values for the same characteristics are applied in other segments. However, it turns out that this relationship is not significant and that the price differences in trim levels are better explained by changes in attributes.
characteristics found in this research. Accurate values for the Volvo suggested prices are not given to respect the strict rules for confidentiality from Volvo. The Volvo suggested prices shown can mostly be found by averaging single option prices from the price lists.
Variable Volvo suggested prices Results Coefficients Significance Intercept 918,679 0,000 - Brand BMW 295,448 0,014 - 18 Inch wheels 563,081 0,001 1000,000 Xenon headlights 584,124 0,000 700,000
Window lists Chrome/ High gloss 261,975 0,024 200,000 Folding mirrors, electric 367,129 0,018 200,000
Park distance control 642,881 0,000 350,000
Interior sports design 1017,984 0,000 -
Sports seats 518,118 0,003 -
Cloth/ Leather upholstery 400,688 0,024 350,000
Leather upholstery 723,893 0,001 1000,000
Navigation 634,247 0,000 950,000
Navigation Pro 1071,329 0,000 -
Climate control 313,061 0,023 250,000
Power seats with memory 992,821 0,000 -
Smart key + Connectivity + Sound improvements 1295,905 0,021 650,000 Comfort seats + Panoramic sunroof 4101,697 0,000 -
Number of Observations 287 287
R² 0,806
Table 5. Volvo single option pricing compared to the results
original Volvo value for Navigation and the results of this research, Volvo can set a guideline value for the attribute Navigation pro. There is not yet a Volvo trim level available with Navigation pro included. With the help of the presented decision support tool, the value for a new trim level that includes this Navigation pro attribute can be determined.
5.2 The decision support tool
The final results lead to the decision support tool shown in Table 6. First, this tool allows the price managers from Volvo cars to include items in a trim level, simply by adding a “1” behind the item. In the bottom row the resulting price for a trim level with the selected attributes is shown. This price can now be used as a reference point to price the trim level.
Items To include item = 1 Brand BMW 18 Inch wheels 1 Xenon headlights Window lists chrome/ High gloss
Folding mirrors, electric 1 Park distance control
Interior sports design 1
Sports seats 1
Cloth/ Leather upholstery 1 Leather upholstery
Navigation
Navigation Pro Climate control Power seats with memory Smart key + Connectivity + Sound system upgrade Comfort seats + Sunroof
Suggested trim level price € pre-tax: 3786 Table 6. Trim level price decision support tool
analyse the value of the new trim levels from the competition. This can be done by adding 1’s behind the items that are found in the tool and that are included in the trim level from Audi or BMW. This can be of use when the specific market (in Europe) seems to differ from the other markets. By analysing the competitors trim levels in this way, the difference in price can be seen as a guide for the difference in value for the market. When the tool is used to find this specific market value it is advised to analyse at least 5 different trim levels to improve the accuracy of the results.
5.3 Research quality
The quality of the research is assessed by the different types of validity.
Internal validity: Part of the internal validity is the construct validity. Construct validity is the answer to the question: Does the measure behave like the theory says it should behave? At first this was a major issue for this research. The results of the regression showed strange values and many problems were found with the significance of the values. After the variable selection process this has improved significantly. Since there is no reference point other than the existing Volvo data for single option prices, we are unable to know how the measures should behave and when this behaviour becomes unacceptable. However, guidelines from previous research were used to assess the construct validity.
The time given to gather the data and the results has caused a challenge. It would have been better to include more European countries in this research. In this way a more accurate view of all involved variables and measures of the variables could be obtained. However, this was not possible with the time and resource constraints. Overall, the internal validity is judged as moderate/ high to high.
The reliability of the used data: The list price reliability is high, these are real applied list prices and actually used at the time of this research and therefore the most accurate and recent data possible.
6. CONCLUSIONS
In this section the conclusions from this research are drawn. Moreover, the limitations of this research are shown and future research opportunities are given.
6.1 Conclusions
The first and perhaps one of the most important conclusions is that there is no significant result on segmented values for trim levels. In other words, trim levels with the same content have a similar price across car segments. However, it is found that the average price for trim levels is higher on cars in a larger segment. This indicates that the content or characteristics of the trim levels determine the differences in price.
The second conclusion regards the brand values. The brands Audi and Volvo do not significantly differ in pricing. BMW is the only brand of the involved car companies that shows a significant result on brand value. The BMW brand value found in this research is 295 Euros. This value is the brand value for the trim levels, in Euros and excluding local tax rates. This value is compared to the Audi and Volvo brand value. This indicates that BMW demands 295 Euros more for the average trim level compared to Audi and Volvo. Therefore the conclusion is made that the brand value of BMW is higher.The knowledge that the Volvo and Audi trim level prices do not differ significantly is a very important finding that should be emphasised. Volvo intends to price their trim levels with prices comparable to Audi and BMW. For this price “position” Audi is regarded as more important to Volvo. Since the brand Volvo and the brand Audi do not differ significantly, Volvo is at the price position they want to be. This price position might seem to be something that can easily be spotted by comparing price lists. However, this is not the case. The contents of trim levels always differ across brands. This makes it difficult to analyse a brands price position. To analyse this price position, the truly important parts or characteristics of the trim levels have to be valued. This research found the price position of the brands Audi, BMW and Volvo in relation to each other. This is done by finding the most important trim level characteristics and coupling these to the values of the trim levels and the brands.
least no accurate value. This research found values for these previously not (accurately) valued attributes. Examples of such attributes are Sports seats and Interior sports design. First, these attributes are difficult to value because these attributes are only offered in trim levels and not available as single options. Second, these items are difficult to value because they are mostly about looks, or a feeling that is obtained by having these items. The regression method provides a way to find the value for such items. Until now, Volvo has not offered Sports seats on any other car than cars with the special sports trim level. However, the competitors do apply Sports seats and similar attributes separately in other trim levels than their special sports trim level. Therefore, Volvo should consider to offer sports seats and similar attributes in other trim levels as well. The found values in this research for these attributes can be used to calculate if a more profitable trim level structure is possible. This might be possible when attributes such as Sports seats are applied in other trim levels as the special sports trim level as well.
All the findings of this research are new and cannot be compared or validated with earlier scientific research on the same topic. However, there is some previous research that can be applied to the results. From the literature it is suggested to apply outliers (attributes that seem to have no value or extremely high value) as revenue maximizing opportunities (Ellison, 2005). Ellison (2005) suggests that the outliers, can be put in the base car if they turn out to have almost no value and no costs to produce. The other option is to take the high valued characteristics out of the base car version. These high valued items can now fully extract their value in a trim level which has to be ordered in addition to the base car. The trim levels can be sold with a higher mark-up as compared to the base car. The results from this research can be used for this purpose. We cannot provide the answer to where these changes can be made. This, because the cost prices are confidential and we are not allowed to use that information in this paper.
Finally, the hypothesis is analysed with the knowledge gained and the conclusions made.
H1: Hedonic pricing applied to the trim levels of Volvo provides different values for the trim levels and attributes as compared to what is currently used by Volvo.
6.2 Relevance to practice
Resulting from the conclusions, many recommendations for practice can be made. First, an important contribution to practice is the decision support tool. This tool can be applied in three ways. First, the tool can be applied to create a reference point for the value of future trim levels. Second, the tool can be applied to analyse trim levels from the competitors. We recommend to apply the decision support tool to European markets and to the models involved in this research. The guidelines that come with the decision support tool explain the tool. They explain the tool in such a way that it can be used by anyone who has never seen the decision support tool before. Moreover, no knowledge of how the values are found is required to use the decision support tool.
6.3 Relevance to literature
The relevance of this research to the literature is moderate to high. The literature can build on this research by applying it to other manufacturers and other markets. Before this research it was not known if the trim level value could be explained by hedonic pricing. Therefore, it was not known how well this could be done by hedonic pricing. This research found that with a small number of characteristics the value of trim levels can be largely explained (80,6% of the value is explained with 16 variables and the intercept value). Moreover, the results lead to different values when compared to the single option values. As a result, an important finding to literature is that the methodology of this research is applicable to trim levels of automobiles. Moreover, the methodology of this research is highly generalizable within the automobile industry. In addition, conducting this kind of research is relevant. For trim level pricing, not the same, but new values are found. Values that provide opportunities to improve the pricing. In terms of results the generalizability is restricted to Europe and the researched car models.
6.4 Limitations
perhaps the biggest limitation to this research. These constraints result in not including trim levels from all over the world but only from Europe. For this reason, only Volvo and the most important competitors are studied. This provides opportunities for future research.
6.5 Future research
ACKNOWLEDGEMENTS
REFERENCES
ArticlesBaker, W. L., Marn, M. V, & Zawada, C. C. 2010. Pocket Price and Pocket Margin Waterfalls, 307–325.
Baltas, G., & Saridakis, C. 2009. Brand-name effects, segment differences, and product characteristics: an integrated model of the car market. Journal of Product & Brand
Management, 18(2), 143–151.
Bhowmick, B. P. 2001. Using hedonic prices to estimate quality changes in American and Japanese cars. Fordham University
Biller, S., Chan, L., Simchi-Levi, D., & Swann, J. 2005. Dynamic pricing and the direct-to-customer model in the automotive industry. Electronic Commerce, 5(2), 309–334.
Cowling, K., & Cubbin, J. 1972. Hedonic price indexes for United Kingdom cars. The
Economic Journal. Retrieved from http://www.jstor.org/stable/10.2307/2230261
Curtis, E., & Wright, R. 2004. Price setting, price dispersion, and the value of money: or, the law of two prices. Journal of Monetary Economics, 51(8), 1599–1621.
Deltas, G., & Zacharias, E. 2011. Product line pricing in a vertically differentiated oligopoly.
Canadian Journal of economics, 44(3), 907–930.
Ellison, G. 2005. A model of add-on pricing. The Quarterly Journal of Economics, (May), 585–638.
Guyon, I., Elisseeff, A. 2003. An Introduction to Variable and Feature Selection. Journal of
Machine Learning research, 3,1157-1182.
Hinterhuber, A., & Liozu, S. 2012. Is It Time to Rethink Your Pricing Strategy? MIT Sloan
Management Review, 53(4), 68-77.
Hulten, C. R. 2003. Price Hedonics: A Critical Review, (September), 5–15.
Lancaster, K. 1966. A new approach to consumer theory. The journal of political economy, 74(2), 132–157.
Lancaster, K. 1971. Consumer demand: a new approach. New york, 177.
Le Gall-Ely, M. 2009. Definition, Measurement and Determinants of the Consumer’s
Willingness to Pay: A Critical Synthesis and Avenues for Further Research. Recherche et
Applications en Marketing (English Edition), 24(2), 91–112.
Marn, M., Roegner, E., & Zawada, C. 2003. Pricing new products. McKinsey Quarterly.
Marn, M., & Rosiello, R. 1992. Managing price, gaining profit. Harvard Business
Review.(September), 84-94.
Ohta, M., & Griliches, Z. 1976. Automobile prices revisited: Extensions of the hedonic hypothesis. Household production and consumption, I, 325–398.
Reis, H. J., & Santos Silva, J. M. C. 2006. Hedonic prices indexes for new passenger cars in Portugal (1997–2001). Economic Modelling, 23(6), 890–908.
Simon, H., Butscher, S. a, & Sebastian, K.-H. 2003. Better Pricing Processes for Higher Profits. Business Strategy Review, 14(2), 63–67.
Verboven, F. 1999. Product Line Rivalry and Market Segmentation—With An Application to Automobile Optional Engine Pricing. The Journal of Industrial Economics, XLVII(4), 399–426.
Books
Baker. W.L., Marn. M.V., Zawada. C.C. 2010. The Price Advantage, Second edition. John Wiley and Sons.
Websites
Audi UK. 2013. Available at:
http://configurator.audi.co.uk/entry?mandant=accx-uk&pr=8V1ABC%5C1%5CMRADC2L%5CMRAOI8D%5CGPNVPNV%5CGWB1WB1%5C MLRA2ZQ%5CMMFA9S5&next=model-page (Accessed March 17 2013)
BMW UK. 2013. Available at:
