GENDER-SPECIFIC ADVERTISING IN THE AUTOMOTIVE INDUSTRY: CAN IT MAKE A DIFFERENCE?
by
GENDER-SPECIFIC ADVERTISING IN THE AUTOMOTIVE INDUSTRY: CAN IT MAKE A DIFFERENCE?
by
NATALIE KROEZEN
University of Groningen Faculty of Economics and Business
Master Thesis Msc Marketing Intelligence
22nd of June, 2015
Supervisor: dr. ir. M.J. Gijsenberg
Celebesstraat 57a 9715 JC Groningen
(06) 13183549 n.j.kroezen@student.rug.nl
Table of Contents MANAGEMENT SUMMARY 4 PREFACE 6 1.INTRODUCTION 7 2.LITERATURE REVIEW 10 2.1BRAND LIKING 10 2.2BRAND CHARACTERISTICS 11 2.3GENDER 13 2.4CONSUMER CONFIDENCE 15 3.METHODOLOGY 17 3.1DATA DESCRIPTION 17 3.2METHOD 19 3.3MODEL 20 4.RESULTS 22
4.1GENERAL DESCRIPTIVE INSIGHTS:FEMALES 22
4.2GENERAL DESCRIPTIVE INSIGHTS:MALES 23
4.3INSIGHTS 23
4.3.1INSIGHTS FOR FEMALES: DIRECT EFFECTS 23
4.3.2INSIGHTS FOR FEMALES: INTERACTION EFFECTS 26
4.3.3INSIGHTS FOR MALES: DIRECT EFFECTS 27
4.3.4INSIGHTS FOR MALES: INTERACTION EFFECTS 29
4.4ADDED Z METHOD: WEIGHTED BETAS FOR VARIABLES 30
4.5GREEN CLASSIFICATION 31
4.6ADDED Z METHOD:GREEN BRANDS VERSUS BROWN BRANDS 33
5.DISCUSSION 36
5.1SUMMARY 36
5.2MANAGERIAL IMPLICATIONS 37
5.3LIMITATIONS AND FUTURE RESEARCH DIRECTIONS 39
6.REFERENCES 41
Management summary
Much research on marketing effectiveness and consumer’s choices regarding FMCG products on the short term has been done. Durables on the other hand received much less attention and little is known about effective marketing strategies. This research focused on how to effectively advertise cars to their potential customers by investigating what factors influence the brand liking score of a car the most. Especially the difference between females and males was investigated to see whether gender-specific advertising can make a difference and could improve the current way of homogeneous advertising.
The difference between gender and their importance of characteristics in cars was measured by three factors: hedonic, utilitarian and ecological motives. Also, the influence of the economic situation plays an important role in the purchase of a durable, expensive good. Therefore, this was the most important moderating effect in this study. Moreover, the introduction of cars with a “green”, i.e. environmentally friendly, image make their way into the automotive market, and is therefore elaborated on extensively. Brand managers must know what the most important characteristics are for customers, which are highly heterogeneous, especially for females and males.
hand, favor ecological motives over all other characteristics, regardless of the segment a car belongs to. Also, during expansions of the economy it is beneficial for brand managers to adjust their advertisements for certain target groups since this will result in a larger effect in the brand liking score of a car. The 18 brands included in this research can benefit tremendously from the outcomes of the individual brand-level outcomes and are advised to adjust their advertisements accordingly.
Preface
This thesis was written for the purpose of successfully obtaining a Master’s degree in Marketing Intelligence at the University of Groningen. The report was written in a period of five months from February till June in the year 2015. For this research the main topic was built on marketing effectiveness for durables, especially cars, which has been researched relatively little. The author proposed a research to investigate whether gender-specific advertising in the automotive industry could make a difference in effective marketing. During the time of writing this thesis, the author was supervised by assistant professor of the department of Marketing, dr. ir. M.J. Gijsenberg. The author wants to express she really valued all the feedback she received during the process and gained many new insights in doing effective research. Also, special thanks for the other five students that were in the same topic group as the author since their teamwork in the beginning of the process is highly appreciated by the author.
1. Introduction
Advertising within the automobile industry is done at a regular basis by the use of different types of media (Kien, 2014). Approximately 400.000 cars were sold in the Netherlands in 2014 (Autoweek, 2014). Companies spend numerous amounts of budget on advertising, in order to reach a lot of consumers and raise awareness, which ultimately increases future sales (Barosso & Llobet, 2012). However, since much research on advertising effectiveness in the FMCG has been done, it is not clear whether these marketing activities are also efficient with durable consumer goods. Durables are goods that have a long wear-out period and are not purchased frequently, such as cars (Forbes, 2015).
Traditionally companies made advertisements and commercials that were meant to reach as many consumers as possible. The same message was sent to all potential consumers of a product or service, however this method of advertising has been loosing its effectiveness (Trusov, Bucklin, & Pauwels, 2009). It is said that 80% of all advertisements are targeted to everyone, 15% is targeted especially to women and less than 5% of the advertisements focuses solely on men (Cramphorn, 2011). With increasing demands in customization of products and services, it is interesting to see whether this would be effective to apply in advertisements for cars. Especially, effective advertising between men and women gives valuable insights in efficient allocation of a companies marketing budget. Also, since women tend to make more important decisions compared to men regarding money spending, advertisements that stress the empowerment of women are said to be more effective (Sandberg, 2014). According to Tucker (2014), customized advertisements have twice as much effect as traditional advertisements without the focus on a specific target group have. However, the question that rises is what specific kind of information in an advertisement for cars makes it appealing to its potential target group?
Nevertheless, economic evolutions represented by consumer confidence levels have an impact on the brand liking score of a product too. How does a dimension such as time influence a brand liking score when economy levels fluctuate and confidence levels are lower than normal? Another important factor influencing the liking scores of a brand is whether a company is following trends. One of the most important trends is the “being green” trend. Green cars, also known as environmentally friendly cars with low CO2 emissions are being produced with steady rates (Autoweek, 2014). The KPMG Global Automotive Executive Survey by KPMG International Cooperative mentioned that in 2013, more consumers’ purchase decisions are based upon factors that desire green technology and environmentally friendly cars (Ashton, 2013). Since no research before has addressed this difference between gender and this “green” matter when identifying important brand characteristics and its influence on the brand liking score over time, this research will do so.
2. Literature review
2.1 Brand liking
In literature, many subjects on consumers’ preferences in car choices have been explored. Mainly on the point what kind of car has the highest liking score with the use of multiple attributes in the form of a conjoint analysis. A brand liking score is explained as in what way consumers prefer this brand to another when they hypothetically had to make a purchase of that product or service. The actual purchase of a car is a well thought decision that involves a lot of trade-offs between different brands and models of cars (Wu, Liao, & Chatwuthikrai, 2014). Brand characteristics play an important role in car brand liking, hence the Dutch are slowly moving away from greater size, weight and horsepower as important characteristics (Kok, 2013). Baltas & Saridakis (2013, p.92) found the following: “Consumer preferences for car
types vary greatly from time to time”. This statement is supported by Kok (2013); his
article describes the change in preference for cars since 2008 until 2011, with the Figure 1: conceptual framework
main focus on the emission of CO2. Within this time period the consumer’s mindset has changed; consumers moved away from the large, heavy and powerful cars and preferred smaller, less CO2 emission cars instead (Kok, 2013). Another trend that emerged in car brand likings is that consumers prefer compact and subcompact cars with low amounts of CO2 emissions to SUV’s and mid-class cars (Wu, Liao, & Chatwuthikrai, 2014). Yanoff & Hanssen (2008) found that the options and choices in cars that are currently available will influence their liking score and actually might change their initial car brand choice over time.
According to Hoen & Koetse (2014), liking scores of car buyers are substantially heterogeneous, indicating that there are differences between consumers. These differences can be found in the brand’s characteristics or the consumer’s characteristics. Also, consumers tend to choose the option that is the easiest, which might be the brand of car that has the least amount of hassle or fits the budget at that moment in time (Novemsky et al., 2007).
2.2 Brand characteristics
Nowadays there are a lot of different brands and all have their own brand identity. Brands are the most important marketing tool for a company; this is what segmentation, targeting and positioning strategies are built on (Lovett, Peres, & Shachar, 2014). Moreover, when a brand is successful, this is the most valuable resource for a company (Elangeswaran & Ragel, 2014). Brands have several characteristics and they must be unique and consistent (DeMers, 2013). Brand characteristics make it an appealing or non-appealing brand for a consumer. However, this differs for every product and type of consumer. Also a consumer’s motivation for choosing a specific car brand is built on different characteristics of a car (Emile & Craig-Lees, 2011). According to Wiedman et al. (2007), the value of a car can be measured in relation to the characteristics of the car brand. These characteristics attract a consumer to a brand or not. These will be referred to as motivations, which can either be hedonistic, utilitarian or ecological.
therefore not worthy enough as a large influencer on the brand liking score. Utilitarian motives on the other hand refer to the ease and functionality of the purchase and use of a product (Anderson et al., 2014). With regard to the purchase of a car, characteristics such as the possibility to trade in your old car, trustworthiness, experiences of friends and low in costs can be linked to utilitarian motives. Within the automobile industry there is a wide diversity in types of cars and there are many different brands that all focus on different characteristics. For example, Kapferer (1998) mentions the characteristics of a luxury car; Rolls Royce might be luxurious for one consumer because of its design and not for another consumer. Typical brand characteristics for Rolls Royce are the car aesthetics, uniqueness, quality and exclusivity. However, for a car such as Ford or Peugeot the focus may lie on different characteristics such as safety and price-quality ratios, which can be defined as utilitarian characteristics. Therefore it is important for companies to keep this heterogeneity in consumer’s motivations in mind when advertising for a brand. Bucklin & Gupta (1992), researched this effect in the fast moving consumer goods category, they found that multiple characteristics of a brand influence consumers purchase incidence and brand choice. New dimensions in how to influence consumer responses with marketing activities came forward, which possibly could also be applied in the durables product category. Moreover, Baltas & Saridakis (2013) researched the effect of consumer characteristics on car brand choice and found that car brand characteristics have a significant effect on the eventual choice of car brand. However, consumers do mostly look at all attributes together when making a choice, rather than evaluating each attribute independently (Wu, Liao, & Chatwuthikrai, 2014).
alternatives. In reality, a huge gap still exists between pro-green attitude and the rare actual pro-green behavior (Bamberg, 2003; Barr, 2006; Rokka & Uusitalo, 2008). Nevertheless, Olsen (2013) found that marketers should address the environmentally friendly characteristics, i.e. the green image, in their advertisements since consumers appreciate and financially value the green choice options. However, only 55% of the consumers in the European Union feel informed about the amount of environmental impact of the products they buy or use (Gershoff & Frels, 2015). Plus, consumers find green products less effective than regular products, which might indicate that regular cars are more effective and less of a hassle concerning price and charging the electric car f.e. (Lin & Chang, 2012). Limited driving ranges, scarce availability of refueling places and long refueling times are indicators that car owners do not suddenly change their preference from a regular car to an environmentally friendly car (Hoen & Koetse, 2014). Research done by Baltas & Saridakis (2013), showed that consumer’s proneness towards buying a green car would significantly affect the car brand choice. Nevertheless, does this ecological trend also show differences in car brand liking scores? Therefore the following hypotheses will be tested for this research:
H1a: Hedonistic motives in car preferences have a negative effect on the car brand liking score.
H1b: Utilitarian motives in car preferences have a positive effect on the car brand liking score.
H1c: Ecological motives in car preferences have a positive influence on the car brand liking score.
2.3 Gender
name and price range. On the other hand, women tend to be less involved in car brand choice than men and showed that style, color, automobile brand and transmission method were on the same importance level as men (Forbes, 2014). Also, women traditionally focus more on decisions relating to color and style, whilst men focus more on the durability and quality aspects of a product (Crosby & Taylor, 1981). Besides, women tend to place more emphasis on health factors when evaluating a product, indicating that ecological factors might influence women in car brand liking more than men (Orth, 2005). According to Qualls (1987), women are significantly influenced by men in terms of brand liking when being in a marital relationship. Moreover, in the fashion industry, research showed that there were significant differences between men and women in brand choices (Kamineni, 2005). Since little is known about this in the automobile industry this could be hypothetically true for this market as well.
According to Lieven et al. (2015), brand designs significantly influence the perceptions of a brand, both for men and women and these significantly relate to consumer preferences. This could mean that men and women have different levels of importance in brand characteristics and may therefore have non-similar liking scores for car brands. Ignoring these heterogeneous appeals can result in overconfidence that profit will be made when applying one-to-one marketing policies (Chintagunta, Dube, & Goh, 2005). Also, for both men and women it is important to have a visual presentation of the car to even come close to a possible purchase of a car (McShane, Bradlow, & Berger, 2012). This indicates that an effective advertisement should focus on the exterior look of the car, which could point to the fact that women prefer these cars more compared to men since they are focused more on style. For this reason, the following hypothesis will be tested:
H2a: Women choosing a car based on hedonic motives have a stronger, positive effect on the brand liking score than women who have utilitarian or ecological motives.
2.4 Consumer confidence
The concept of consumer confidence has been used tremendously to forecast consumer spending in several economical activities. However, in the evolution of changing brand liking scores due to the fluctuating state of the economy, little is known. According to Statistics the Netherlands (CBS, 2015), the definition of consumer confidence is as follows: “The indicator providing information on the
confidence and expectations consumers have about developments in the Dutch economy. Together with the indicators on economic climate and willingness to purchase, these contribute to the prediction of short-term fluctuations in the consumption by private individuals, particularly about the demand for and purchase of durable consumer goods.” Comparatively, de Boef & Kellstedt (2004) state that
consumer confidence is explained as; the way consumers view the economy, which can be either optimistically or pessimistically. When the present is good, the future appears to be viewed at with a more optimistic eye. Nevertheless, with higher inflation rates and increasing unemployment numbers a pessimistic viewpoint of the economy is more likely (de Boef & Kellstedt, 2004).
which changes over time, influences the consumers’ spending behavior over time (Al-Eyd, Barrell, & Davis, 2009). Moreover, in a study of Lyziak & Mackiewicz-Lyziak (2014), it became clear that habitants of the Netherlands are found to be in the top of consumers who are anticipating future inflations. Likewise, this could be explained as cautiousness of the Dutch in their spending abilities and brand choices, which are not the most exclusive and expensive cars. For this reason the following hypotheses are formulated:
H3a: Ecological motives moderated by higher consumer confidence levels, compared to the previous period, have a stronger, positive effect on the car brand liking score than hedonic and utilitarian motives moderated by higher consumer confidence levels.
H3b: Periods with lower consumer confidence levels compared to the previous period, have a negative influence on the car brand liking score.
17 3. Methodology
The research objectives of this study were aiming on finding a clear view of what factors were important in advertising, regarding gender in the automotive industry. In order to find answers on the hypothesized relationships, multiple moderated time series regression models were estimated. First, the models should provide estimates for the three factors on an individual brand level. Second, it should allow for interactions between the consumer confidence level and the three factors. Third, since there is a time span of 4 years in the data, the models should allow the possibility of a dynamic effect in the form of a trend line. Fourth, since the brands used in this research are different from each other, the parameter estimates should vary. Fifth, caution should be taken for the effect of multicollinearity and autocorrelation, since there are eight variables included in the model. Sixth, to see whether the effects can also be generalized for the industry, weighted betas of the variables should be calculated to get the complete effect of a variable. Seventh, since this research wants to investigate whether the “green” trend is influencing the brand liking score of cars, a ranking should be made according to the yearly average liking scores.
3.1 Data description
data set in order to accurately account for the effect of gender. This gave 23567 data points for males and 21218 for females to analyze. In total, 26 car brands were evaluated to identify the first preference of a car, which was translated into the brand liking score. This score is based on several car characteristics such as price/quality ratios, engine power, technology, design, trustworthiness and space in the car. For this research, 18 car brands were used for analyzing the effects of these characteristics since these gave enough information on a weekly basis. The main reason for excluding 8 brands from the analysis is that they had over 80 weeks of missing data points, out of a total of 203 weeks. Manually manipulating these missing data points by entering the mean value might give wrong insights. Therefore, the 18 brands that are included can give reliable insights, as this is the data that is provided by the respondents themselves. The average amount of respondents per week is 118. The car brands that are used are shown in table 1, which can be seen on the previous page. Also, to get a clear view on what factors in choosing a car are the most important, the characteristics for having a certain car preference were factorized. This was done to group them into three main factors, which was better to interpret. In total, 23 characteristics that are important in choosing a car were listed in the survey. The three factors each contain three of these characteristics for the reason they can be closely linked to each other and tell something about hedonic, utilitarian and ecological motives in car preferences and their liking scores. Consumers that are more focused on hedonic motives find the look of a car more important. Utilitarian motivated consumers find the price/quality of the car important whilst the ecological motivated consumers care for the environment. The three factors are shown in table 2 on the previous page.
3.2 Method
The modeling approach is a moderated time-series regression model. The OLS method is used to estimate the effects. First, the model should include the three factors named before, which are the explanatory variables (HM, UM & EM). These factors were checked by reliability analyses, however not all of the brands gave conclusive crohnbach alpha scores for the factors. The outcomes of the reliability test can be found in the appendix. Therefore the factors are based upon theoretical reasoning and that they can be closely linked to each other. Also, consumer confidence is included as a separate explanatory variable (CC). Since the initial confidence levels were all negative for the time period used in this research, a constant was added to create a positive number, this made it more applicable to use in the analysis. After this, the confidence level was dummy coded based on the first-difference. When there was an increase in the confidence level compared to the previous period, the corresponding week received a 1 in the dummy. Contrarily, when the confidence level was less than the previous period it received a zero in the dummy. For the dependent variable, the mean brand liking score of a brand is used. These scores are based on weekly data after being aggregated on weeks running from 1 (3rd of January 2011) till 203 (21st of December 2014). This gives a mean preference for a brand per week ranging from 0 to 1. Indicating that for example a score of .13 meant a total of 13% of all observations in that week preferred that specific brand. Since the data file is split on gender, gender was not included as a separate explanatory moderating variable. Splitting up the file and running separate analyses gave estimates per brand based on gender.
solve this problem the lagged version of the dependent variable was added in the model as an explanatory variable. For these specific cases the Durbin h statistic is calculated, since the regression included one lagged dependent variable. This statistic accounts for the augmenting effect of the lagged explanatory variable. The Durbin h was calculated manually according to the equation shown below.
The outcome of the h statistic needed to be < 1.645 for the assumption to hold that autocorrelation is absent (Pennings, Keman, & Kleijnnenhuis, 2006). This resulted in all cases in no autocorrelation. The outcome of the Durbin Watson and h tests can be found in the appendix.
Third, to analyze whether there was a significant difference in the brand liking score over time between men and women, a trend line was added in the model. This trend line will tell if time had an influence on the brand liking score.
Fourth, to see whether the variables have a collective effect as well, and to give empirical generalizations for the automotive industry, the added Z method is applied (Rosenthal, 1991). This was done by recalculating the Z-score for each brand-specific p-value. Subsequently the Z’s were summed and divided by the square root of the number of included brands. This new Z-score was used to derive the associated p-values per variable. When this gave significant results, the associated effect size was calculated by retrieving the weighted average response parameters across the included brands.
3.3 Model
Since this research makes use of two separate datasets, i.e. split on gender, separate models needed to be estimated. The model includes all the variables that are mentioned before. The model is a unit-by-unit model since it is built per brand and on an individual level. The model can be used to predict the brand liking score when the estimates of the variables are known. Below the equation of the model is visualized:
Yi = β0i + βiHMi + βiUMi + βiEMi + βiCCi + βiHMCCi + βiUMCCi + βiEMCCi + βiTi +
where
Yi = brand liking score of brand i
β0 = intercept of brand i
βiHMi = the main effect of hedonic motives on brand i
βiUMi = the main effect of utilitarian motives on brand i
βiEMi = the main effect of ecological motives on brand i
βiCCi = the main effect of consumer confidence on brand i
βiHMCCi = the moderated effect of consumer confidence and hedonic motives on
brand i
βiUMCCi = the moderated effect of consumer confidence and utilitarian motives on
brand i
βiEMCCi = the moderated effect of consumer confidence and ecological motives on
brand i
βiTi = the direct effect of time on brand i
βiBLSt-1i = the direct effect of the lagged brand liking score on brand i
4. Results
Before running the several necessary analyses, an independent samples t-test was done to see whether there was a significant difference between men and women in their car brand preference and if gender influences this accordingly. The independent samples t-test was significant with t(23186) = -12,846 and p = .000. Also, a Chi-square test with cross tables was done to see if men and women differ in car brand preference when taking all brands separately. The Chi-square test was significant with Chi-square(17) = 432,87, p = .000. Therefore further analysis can be done to see what factors influence this significant difference between females and males in car brand preference, i.e. brand liking score.
4.1 General descriptive insights: Females
The female sample comprised of 21218 respondents, which is 47.3% of the total sample. The division between males and females was nearly even, which makes the outcomes comparable to each other. The division in age between the female respondents is as follows: 12.6% is younger than 30, 20.9% is between 30 and 39, 25.3% is between 40 and 49, 24.7% is between 50 and 59, 12.3% is between 60 and 65 and 4.3% is over 65 years of age. Another important descriptive is that nearly 50% of the female respondents live with a family household without children under the age of 18. Females education levels are as follows: low level 30.2%, medium level 44.2% and a high education level 25.6%. In total nearly 60% of the females have a full-time or part-time job.
4.2 General descriptive insights: Males
The data set had a total of 23567 observations, of which all data necessary was filled in accordingly. This resulted in the fact that 52.7% of the respondents were males. Out of this sample, the division in age is as follows: 15.0% is younger than 30, 21.4% is between 30 and 39, 23.6% is between 40 and 49, 23.2% is between 50 and 59, 12.4% is between 60 and 65 and 4.4% is over 65 years of age. Also, nearly 50% of the male respondents are living with a family household without children under the age of 18. The respondents education levels were as follows: low level 18.8%, medium level 43.4% and a high educational level 37.8%. Out of the complete sample, 80.0% had a full-time or part-time job.
4.3 Insights
Table 3 shows the parameter estimates on brand level that were estimated via multiple ordinary least square regressions. First, all significant main effects on the brand liking score will be evaluated, followed by the significant interaction effects. The statistics for females will be given first, followed by the outcomes for males.
4.3.1 Insights for females: direct effects
The main effects of the motives in preferring a car brand are evaluated with the use of a significance level of .05. Regarding the direct effect of hedonic motives on the brand liking scores it shows that for Audi (β = .008, p < .05), BMW (β = .007,
p < .01), Citroen (β = .017, p < .01), Fiat (β = .013, p < .05), Hyundai (β = .010, p <
.01), Mercedes (β = .011, p < .01), Nissan (β = .012, p < .01), Peugeot (β = .014, p < .05), Seat (β = .008, p < .01), Skoda (β = .012, p < .01), Volvo (β = .011, p < .05) and
0,0% 10,0% 20,0% 30,0% 40,0% 50,0%
Low level Medium level High level
Educational level
Education level Females Education level MalesVolkswagen (β = .016, p < .05) a positive effect has been found. The largest effect is for Citroen, whilst the smallest effect on the liking score is for BMW. When the hedonic motives for a Citroen go up by one unit, the brand liking score of Citroen increases with .017 point, whilst the intercept of Citroen is .022 when all others are held at zero. The increase is almost at the same size of the constant, which could indicate a substantial increase in the brand liking score. In total, 12 brands have significant effects on the brand liking score regarding hedonic motives.
The utilitarian motive factor showed positive results for BMW (β = .011, p < .05), Citroen (β = .010, p < .05), Ford (β = .014, p < .05), Hyundai (β = .009, p < .01), Kia (β = .012, p < .01), Nissan (β = .011, p < .05), Seat (β = .010, p < .01), Skoda (β = .009, p < .01), Suzuki (β = .010, p < .01) and Volkswagen (β = .034, p < .05). For the utilitarian motives, the largest effect was found for Volkswagen with an effect of .034 with a one-unit increase. The smallest effect was for Hyundai and Skoda with respectively an effect size of .009. In total, 10 brands show effects on the brand liking score.
Ecological motives show positive results for Ford (β = .030, p < .05), Hyundai (β = .009, p < .05), Mercedes (β = .009, p < .05) and Volvo (β = .016, p < .05). The effect of ecological motives for Ford is the highest with a .030 point increase in the brand liking score of Ford with a constant at .046 which is an increase of 75.0%.
The direct effect of consumer confidence on the brand liking score, which is dummy coded, with 1 representing an upward confidence level based on the first-difference method and 0 a decline, showed that consumer confidence had negative results for BMW (β = -.005, p < .05), Fiat (β = -.005, p < .05) and Seat (β = -.001, p < .05). This means that during an expansion period, compared to a decline period in the previous month, the brand liking score of BMW will decrease with .005 points.
