Measuring and predicting job choices:
isolating and decomposing brand equity
effects into perceived benefit components
Date:
June 18, 2018
2
Measuring and predicting job choices:
isolating and decomposing brand equity
effects into perceived benefit components
Master thesis, MSc Marketing,
Marketing Intelligence
University of Groningen, Faculty of Economics and Business
June 18, 2018
First supervisor
Dr. Felix Egger
Second supervisor
Dr. Erjen van Nierop
3
Abstract
Competition for talented staff has been increasing over the years. Therefore, research into the field of employer branding has become increasingly important. This paper contributes to the field of research by isolating the effects of brand equity from their respective location and industry effects. Moreover, the concept of employee based brand equity is split into four image association constructs. Using two extended choice based conjoint models, the location and industry effects are proven to be confounding variables of the brand equity concept. Furthermore, it is found that the brand equity concept can effectively be split up into the concepts of brand awareness and functional, economic and psychological benefits. In addition, interaction effects are found between the annual salary and two of the brand equity components. The results of this research imply that managers and researchers should take into account the effects of locations and industries in terms of employer branding. Moreover, whereas expected effects are found for functional and psychological benefits, an unexpected negative effect is found for economic benefits. Lastly, limitations and suggestions for further research are discussed.
4
Contents
1. Introduction ... 5 2. Theoretical framework ... 7 2.1 Job choice ... 7 2.2 Conceptual model ... 8 2.3 Brand equity ... 8 2.3.1 Brand awareness ... 9 2.3.2 Functional association ... 10 2.3.3 Economic associations ... 11 2.3.4 Psychological associations ... 12 2.4 Starting salary ... 132.5 Location and industry ... 14
3. Methodology ... 15 3.1 Research design ... 15 3.2 Experimental design ... 17 3.3 Measurement items ... 18 3.4 Data collection ... 19 3.5 Model specification ... 19 4. Results ... 21 4.1 Sample statistics ... 21 4.2 Data transformation ... 22
4.3 Comparison of estimation methods for joint and separated models ... 23
4.4 Comparing different model extensions ... 25
4.5 Interpretation of benchmark model ... 27
4.6 Estimating the effects of brand specific constructs ... 27
4.7 Investigation of interaction effects ... 30
5. Discussion ... 32
5.1 Discussion of findings ... 32
5.2 Theoretical and managerial implications ... 34
5.3 Limitations and further research ... 35
6. References ... 36
7. Appendices ... 38
7.1 Appendix-A: Example survey ... 38
5
1. Introduction
Brand equity as a concept has been known for several years. Keller (1993) introduced the customer based version of the concept and implied that it is made up of a combination of high brand awareness paired with a positive brand image. High levels of brand equity have shown in other research to have several positive effects, such as an increase in the intention to purchase (Cobb-Walgren et al., 1995). This implies that building a strong brand name for the company is relevant for increasing firm performance. However, recent academic research has noted that the effects of brand equity do not only influence firm performance through effects on customer behavior but also from an employee perspective (Berthon, Hewing and Ha, 2005; Theurer et al., 2016; Lievens and Highhouse, 2003). Therefore, the focus of research has shifted away from the customer side of the spectrum and started investigating the effects of employee-based brand equity (EBBE). A concept that has been defined by Ambler and Barrow (1996, p. 8) as a “package of functional, economic and psychological benefits provided by employment, and identified with the employing company”. Therefore, the individual’s perception of receiving certain benefits from employment at a company might influence the value derived from job offers. This makes the definition similar to the customer brand equity perceptive as image perceptions will provide additional value to products outside of its functional characteristics (Aaker, 1991). As this is the case, companies with high EBBE could have a better position in attracting qualified staff
Attracting the right employees has been shown to be relevant for a multitude of reasons. Tavassoli, Sorescu and Chandy (2014) note that in general, the salary of employees accounts for twenty to fifty percent of the total costs of a company. Therefore, the employees are generally viewed as an important resource. Following the resource based view (RBV) by Wernerfelt (1984), it is thus implied that employees are a relevant factor in creating a competitive advantage. Especially when they can be seen as valuable, rare, inimitable and non-substitutable which allows the competitive advantage to be sustainable. Moreover, in earlier research the internal performance of employees has been linked to a more positive evaluation by the customer, which subsequently is linked to better firm performance (Maxham, Netemeyer and Lichtenstein, 2008). Consequently, firms are competing to attract talented staff, whom can subsequently be employed in order to gain a competitive advantage. Leveraging the effects of a strong employer brand is one of the possible ways for firms to engage this competition. Therefore, understanding drivers and effects of employee based brand equity can be considered as relevant.
6
can decrease uncertainty about the economic benefits that are provided by the company. Together with the psychological benefit of self-identification, this is assumed to influence the amount of salary the employee is willing to sacrifice. Furthermore, the concept of brand equity is linked to functional benefits through more opportunities for internal advancement, training and skill-improvement. Moreover, it is assumed to have additional psychological benefits by implying a willingness to work hard. The combination of these benefits is assumed as a way to improve the power of the individual’s résumé (DelVecchio et al., 2007). However, in previous research, the effects of brand equity are researched in conjunction with the related industry and location. These factors are relatively fixed for the brand as they are often tied to a specific industry or location. Unilever for example, has recently been in the news after moving their HQ from London to Rotterdam (Independent, 2018). Therefore, the company will be associated with the city of Rotterdam as well as their respective industry of consumer goods. Industry and location effects are thus considered to be confounding the effects of brand equity in earlier works. Thus, this research aims at filling the research gap identified by Theurer et al. (2016) in isolating the effects of brand equity from its confounding variables and examining how brand equity effects relate to other job characteristics. The measurement of brand equity in relation to other characteristics is done by means of a choice based conjoint. This means of analysis enables the realistic evaluation of multiple job offers simultaneously, while also considering co-existence of several job characteristics (Eggers and Sattler, 2001).
The conjoint analysis includes brand equity and salary, as well as the confounding effects of location and industry. The relevance of separating the location effects can be illustrated by the importance placed on the location of the recently announced second headquarter of Amazon (New York Times, 2017). The company has opened a bidding war in which cities can come up with an offer and proposition for why Amazon should choose their city as the location for the new headquarters. Moving to the right location could be relevant for the company as more talented staff are located in the region. As potential employees might value working closer to where they live, this could be a way for the company to compete for talented staff. Moreover, people might think certain working characteristics belong to a certain industry rather than the company. Therefore, the present research makes a methodological contribution to the field of research, by being the first research to separate the effects of industry and location from that of brand equity. Consequently, the first avenue of research of this paper is captured with the following research question.
7
In addition to evaluating the effects of several job characteristics, a second avenue of research in this paper is focused on the composition of the brand equity concept. The paper aims to build on the customer based brand equity research to find the drivers of employer based brand equity. Decomposing the concept of employer based brand equity can have important implications for managers in recruiting activities. If managers know how to influence their brand equity then it is possible to effectively market the company as an attractive employer. In addition to distinguishing the effects of brand awareness and associations, brand associations are also further separated following the before mentioned definition by Ambler and Barrow (1996). As a result the second question discussed in this paper is as follows:
RQ (2): What are the effects of brand awareness and functional, economic and psychological associations on brand equity?
In order to answer the questions mentioned above, the paper will be structured in the following way. In section 2, the theoretical framework will be provided in which concepts are defined and based on existing literature, hypotheses are formed. In section 3, information will be provided about the methodology of this paper including data collection and analysis techniques. Section 4 includes the actual analysis of the data and provides the results. The results are further discussed in section 5, where managerial implications will be provided along with the research’s limitations and avenues for further research.
2. Theoretical framework
2.1 Job choice
8
2.2 Conceptual model
In order to guide the reader throughout the remainder of this chapter, the conceptual model is presented below in figure 1. The model shows that there are four components that influence the brand equity of the employer. Thus, these constructs are assumed to indirectly influence the utility derived from job offer as components of the employee based brand equity concept. Moreover, a direct effect is expected from the starting salary on the perception of utility. This effect is however proposed to be moderated by the levels of brand equity. The model also shows that the location and industry influence the utility, although no specific hypotheses are formed. The perceived utility is subsequently assumed to affect the behavior of the respondent in terms of choosing a job offer. The specific hypotheses and their substantiation from extant research are discussed in the following sections.
Figure 1 : Conceptual model 2.3 Brand equity
9
firm that differentiates offers by the focal firm from its competition and is influential in the attraction, motivation and retention of employees. When specifically looking at the attraction of employees it is thus implied that the employer brand can have positive effects on employer attractiveness.
Several other authors have made an attempt at explaining the underlying reasoning behind the existence of these effects. For example, as mentioned in the introduction, Research by DelVecchio and colleagues (2007) has found increased power of an individual’s résumé to be a result of working for well-known brands. Other research has provided explanations of brand equity effects through uncertainty avoidance and organizational identification. Uncertainty avoidance refers to risk aversion by younger employees due to a lack of experience. As they are unexperienced and uncertain about their identity, they would rather work for familiar brands as they perceive that there are greater opportunities for self-enhancement. Organizational identification refers to the degree to which the individual identifies with the focal firm. Both these concepts are suggested to have a positive effect on the employer attractiveness (Tavassoli, Sorescu and Chandy, 2014). In addition, according to Lievens and Highhouse (2003), the symbolic effects of brand equity within job offers is assumed to positively influence the individual’s perception of a job-offer. Although the current research tests brand effects in conjunction with other characteristics, this is not expected to change the direction of effects and therefore the following hypothesis is proposed:
H1: Higher levels of brand equity will increase the utility of job offers.
In order to allow managers to market the brand to potential employees, this paper proposes decomposing the concept of brand equity. In addition to distinguishing brand awareness and image, this paper further decomposes the concept of brand image following the definition of the employer brand by Ambler and Barrow (2003). They propose that brand associations consist of functional, economic and psychological dimensions. Based on these associations an individual job seeker can gain perceived benefits from being employed at a certain brand or organization. In the following sections, brand awareness is identified as a prerequisite of brand image and the components of brand image and their respective effects are discussed.
2.3.1 Brand awareness
10
which it does so”. The reason this concept is discussed first is because it can be considered as a prerequisite of brand associations. In order to have associations with a company, first the individual has to be aware of the company. One of the effects of brand awareness in the consumer context is that higher levels of brand awareness lead to increased probability of inclusion in the consideration set (Baker et al. 1986). Furthermore, when under situations of low involvement in the elaboration likelihood model by Petty and Cacioppo (1983), the presence of brand awareness is assumed to be strongly influential in the decision process. However, in this particular context the individual’s involvement is likely to be higher since the decision has more important implications. Therefore, the decision process of choosing an employer is expected to fall on the high elaboration side of the continuum (petty and Cacioppo, 1983)
Although this is the case, brand awareness has also shown to be influential in the employment context. The before mentioned effects on the power of an employee’s résumé are mainly based around the fact that the brand is well-known, which implies that there is high brand awareness (DelVecchio et al. 2007). Thus, the attractiveness of an employer is influenced by the brand awareness. As a result, companies with lower levels of brand awareness are restrained in their ability to attract human capital (Wilden, Gudergang and Lings, 2010). In addition, the positive effects of brand awareness can be explained by mere exposure effects (Gordon and Holyoak, 1983). This theory implies that the repeated exposure to certain stimuli will result in a positive affect towards that stimulus. These findings are expanded on by Reber, Winkielman and Schwartz (1998), who state that positive affect is increased by higher frequency and longer exposure as mediated by perceptual fluency. Therefore, they suggest that the ease with which a certain stimulus can be interpreted, increases feelings of liking. Based on the above findings, this paper assumes that higher levels of brand awareness will result in higher brand equity and therefore adopts the following hypothesis:
H2: Higher levels of brand awareness will result in higher levels of brand equity
2.3.2 Functional association
11
by that particular organization. This notion is supported by Berthon, Hewing and Ha (2005). Using factor analysis, they come up with five dimensions that positively influence employer attractiveness. One of these dimensions is labeled development and is concerned with factors that help the individual with self-development and career enhancement.
Sivertzen and colleagues (2013) refined the EmpAt measure introduced by Berthon, Hewing and Ha (2005). They use the same approach to finding five underlying dimensions of employer attractiveness. Although they find a lower importance of economic and social value, both researches find that the development of the individual is an important factor in increasing the attractiveness of job offers. This could either be through self-enhancement of confidence or personal growth. Additionally, the opportunity for personal development was also identified as one of the three perceptions to be important in job offers for private companies in China (Jiang and Iles, 2011). Another dimension that both the EmpAt papers agree on, is that of application value. Meaning that when the work environment enables the application of gathered knowledge, the job offer is assumed to be more attractive (Berthon, Hewing and Ha, 2005; Silvertzen, Nilsen, Olafsen, 2013). As a result of these earlier findings, it is expected that if an individual perceives to receive more functional benefits from employment by an organization, the individual will derive more value from working for that organization. Therefore, the following hypothesis is formed:
H3a: Associating an employer with perceived functional benefits will have a positive effect on brand equity.
2.3.3 Economic associations
The second component of brand associations is related to the economic benefits associated with working for a certain organization. Economic benefits are defined in the paper by Ambler and Barrow (2003) as material or monetary rewards. Consequently, the concept does not only consider salary but also other parts of the compensation package, such as rewards systems. The quality of the compensation package has been identified as one of the five dimensions positively influencing employer attractiveness (Berthon, Hewing and Ha, 2005).
12
having a high corporate social performance (CSP), signals to potential employees that the company has good working conditions and a fair compensation package. The perception of receiving a fair compensation package subsequently leads to higher attractiveness of working for that organization. In other research, implications are made that potential employees prefer a high compensation package over an average level compensation package in traditional as well as realistic job preview settings (Saks, Wiesner and Summer, 1996). With these findings in mind, it is expected that when the job seeker perceives to attain more economic benefits from employment at certain organizations, the job is perceived to be more attractive. Formally, leading to the following hypothesis:
H3b: Associating an employer with perceived economic benefits will have a positive effect on brand equity.
2.3.4 Psychological associations
The third and final proposed component of brand image is that of psychological associations. These associations are related to feelings of belonging and having a purpose. In that light, the research by Backhaus and colleagues (2002) also provides evidence for the positive effects of psychological associations on employer attractiveness. An additional theory they use to explain CSP effects on employer attractiveness is social identity theory. This theory assumes that there is a general perception that the company someone is employed to, says something about the individual’s identity. Therefore, if the individual feels the image of the employing organization will improve their own social identity, the job will be perceived as more attractive. A theory that is similar to the social identity theory, is that of organizational identification. According to the paper of Tavassoli and colleagues (2014), the more an individual identifies with the organization at question, the more attractive the organization becomes. The reason behind this, is because individuals look to affirm, express and enhance their image through the usage of brand affiliations.
13
Lastly, another type of psychological benefits could be to have an exciting work environment. Both papers by Berthon (2005) and Sivertzen (2013) identify the importance of this dimension and label it as interest and innovation respectively. The latter of these labels, emphasizes how both authors assume that having an exciting work environment mainly results from the company being orientated towards innovation. Other researchers have also shown how innovative aspects lead to increased satisfaction and enjoyment, which subsequently is linked to employer attractiveness (Simpson, Siguaw and Enz, 2006). They argue that these effects exist because the work is perceived to be more challenging and therefore more rewarding to carry out. Resulting from these findings, the following hypothesis is suggested:
H3c: Associating an employer with perceived psychological benefits will have a positive effect on brand equity.
2.4 Starting salary
Apart from brand equity and its components, another concept that is part of the job’s characteristics is that of salary. The concept of salary itself is quite straightforward and therefore, although mentioned a lot, has not been specifically defined. Thus, in order to be clear about what this paper is measuring, salary will be defined as the financial amount paid to the employee on an annual rate as compensation for their work or services. This concept has similarities with the economic benefits that is discussed as a component of brand associations. However, there are still differences between the two concepts. One of these differences is that whereas economic benefits consider the compensation package as a whole, this concept only considers the fixed salary. Furthermore, this concept measures the actual economic benefit of salary whereas the associations measure the perceived benefits. Still, one of the measures included within the economic dimension of the EmpAt measure is whether the company pays an above average salary (Berthon, Hewing and Ha, 2005). This implies that at least part of the positive effects of economic benefits are explained by the salary. Although economic benefits did not have a significant impact on job attractiveness in the EmpAt refinement effort (Sivertzen, Nilsen and Olafsen, 2013), starting salary has been identified as one of the three most relevant job attributes for young professionals (Fung et al. 1996).
14
following the definition of customer based brand equity by Keller (1993). This definition assumes that people will respond more favorable to marketing elements of branded products than they do for unnamed or fictitious products. This assumption is based on the fact that marketing response is implied to be a result of the customer’s attitudes towards the brand (Rossiter and Percy, 1987). Moreover, the implication is made that customers are more willing to pay a price premium for the same product when they have a favorable associations with the brand name (Starr and Rubinson, 1978). As a result, the following two hypotheses are suggested:
H4a: A higher starting salary will increase the utility of job offers.
H4b: The effects for starting salary will be weaker under higher levels of brand equity.
2.5 Location and industry
Lastly, two other factors are included in the analysis. These two factors are the location and the industry that are related to the job offer. The industry refers to the general business activity that will be performed at the company concerned in the job offer. Likewise, the location refers to the geographical location in which the companies’ headquarter is located. So far, these concepts have not been distinguished from the effects of brand equity but rather been measured as part of the brand equity. However, the industry might have their own effect on job attractiveness as people can have either positive or negative associations with a certain industry. For example, Borah and Tellis (2016) speak of ‘guilt by association’ effects, which imply that a crisis at a focal firm can cause harm to rival firms as people associate the brands with each other through the industry.
15
3. Methodology
3.1 Research design
In order to separate the brand equity effects from their respective location and industry effects, this paper uses two separate choice based conjoint models. The usage of a choice based conjoint model allows for the simultaneous comparison of multiple sets of attributes in terms of their perceived utility. Since few people actually rate a job offer on a certain scale in real life, this method is the preferred option over rating based conjoint studies. This method asks for the respondent to choose between multiple offers that consist of a set of attributes, which is much closer to a real life situation. The data on the preferred option can subsequently be decomposed in order to find the effects of the individual attributes (Eggers and Sattler, 2011). Therefore, this method is appropriate for the separation of brand effects from their respective industries and locations. This is done by estimating the effects of location and industry in a separate conjoint, as estimating the effects together with their respective brand will result in intercorrelation. Several methods for the estimation of these effects will be discussed and compared in the results section.
16
Industry Location Starting salary
Finance Consumer goods Electronic Energy Utrecht Amsterdam The Hague Rotterdam -15% -7.5% Expected salary +7.5% +15%
Table 1: Attributes and levels conjoint part 1
The second conjoint is similar to the first. As in the first model, the starting salary, locations and industries are included as variables. However, in this model the specific brands are added. In this study, the effects of the location and the industry are tied to a brand and therefore remain fixed. Therefore, the effects of the brand, location and industry are combined to form a single attribute. The brands included in the model are chosen to correspond to the industries and locations of the first conjoint experiment. Moreover, in order to include some variation in brand equity, not only the largest brand for each industry were included. In addition, some brands were also selected that were expected to have lower levels of brand equity due to lower awareness or having a negative brand image. Furthermore, a fictional brand was also implemented in order to control for the situation of no brand equity. Since the company does not exist, respondents are not expected to be aware of the brand and therefore have no brand image associations. A complete list of all the attributes and their respective levels is included in table 2.
Brand (part-worth) Location (part-worth) Industry (part-worth) Salary (linear) - Rabobank - ING - ASN Bank - Utrecht - Amsterdam - The Hague - Finance - -15% - Unilever - Douwe Egberts - Heineken - Rotterdam - Utrecht - Amsterdam - Consumer goods - -7.5% - KPN - TomTom - Electrolight (control) - The Hague - Amsterdam - Utrecht - Electronic - Expected salary - Eneco - NUON - Royal Dutch Shell - Rotterdam - Amsterdam - The Hague - Energy - +7.5% - +15%
17
3.2 Experimental design
As mentioned in the section above, the first model consists of three different attributes, of which two of them have four levels and the other has five. Therefore, there is a total of 80 different possible combinations of attributes. For each choice set there were 3 different alternatives from which the respondents could choose. This leads to a total 82.160 ((240−3)!∙3!240! ) possible choice sets, making a single respondent incapable of answering all of them. Therefore, each respondent was asked to answer 6 choice sets from the first conjoint model. Out of the alternatives, the respondent was asked to the select the job offer that provides the highest utility to them. The no choice option was added as a fourth alternative. Although a dual response means of implementing the no choice option allows for the gathering of data even if the respondents would not consider taking either of the alternatives (Eggers and Sattler, 2011), this method was chosen in order to lower the amount of decisions that had to be made. Figure 2 shows an example of what the choice sets for this first part looked like.
Figure 2: Example conjoint part 1
18
Figure 3: Example conjoint part 2 3.3 Measurement items
19
construct Measurement item Measurement
scale Brand awareness Please select all brands that you are familiar with, for example, you have
previously seen or heard of them. Binary (0,1) Functional
benefits
Please select all brands that would provide functional benefits as a result of being employed by that brand, for
example, stronger résumé, personal growth, opportunity to apply learned knowledge.
Binary (0,1)
Economic benefits
Please select all brands that would provide economic benefits as a result of being employed by that brand, for
example, above average salary, job security, fair compensation package.
Binary (0,1)
Psychological benefits
Please select all brands that would provide psychological benefits as a result of being employed by that brand, for
example, exciting work environment, having good relationships, belonging and acceptance.
Binary (0,1)
Table 3: Measurement items image association constructs 3.4 Data collection
The testing of the model was done by conducting the survey among a total of 337 respondents. The gathering of respondents started at the 7th of May and ended on the 23rd of May 2018. The process of gathering respondents was done through the usage of convenience sampling in combination with snowball sampling. The link to the survey was provided on the Facebook and LinkedIn pages of the researchers in order to reach a large network of potential respondents. Additionally, respondents were asked to share the link to expand the network even further. Lastly, the survey link was shared in several Facebook groups. These groups allow for the trading of respondents among students looking for respondents. In addition to increasing the number of respondents, this also increases the diversity of the sample, making the findings more generalizable. The downside to using this type of sampling is that it allows for little to no control on the respondents that are included in the sample. Therefore, respondents are informed beforehand of the fact that the survey is in English to make sure they will understand the questions. Furthermore, there were no other restrictions put on the sample.
3.5 Model specification
As mentioned earlier, to analyze the effects of job attributes, this paper uses the random utility theory. This theory assumes that the total utility U of a job offer j for customer i is a latent value that consists of a vector of parameters V and an error component ɛ. The vector component represents the effects of the constructs in the conceptual model, while the error component captures all the effects that are not accounted for in the model (Manski, 1977).
20
In order to estimate the impact of the systematic components of the model, a multinomial logit (MNL) model is used. Within this model, the probability of choosing a job offer ‘a’ from a choice set with ‘J’ alternatives is the result of the exponentiated utility of alternative ‘a’, divided by the exponentiated sum of utilities of all the alternatives (Eggers, Eggers & Kraus, 2016).
Eq. (2): 𝑝𝑟𝑜𝑏(𝑎|𝐽) = exp (𝑉𝑖𝑎)
∑ exp (𝑉𝐽𝑗 𝑖𝑗)
Therefore, modeling the utility of a certain job offer allows for the interpretation of choice probabilities. Considering that up to now the effects of brand equity are analyzed as confounded by industry and location effects. The formula for the systematic utility of the extant literature and the second conjoint consists of two parts, the brand effects b and the salary effects ‘sal’.
Eq. (3): 𝑉𝑖𝑗 = 𝛽𝑏𝑏𝑗+ 𝛽𝑠𝑎𝑙𝑠𝑎𝑙𝑗
This present research however, considers the effects of brands, location and industry separately. For the model, this means that the first conjoint is used to deduct the location and industry effects from that of the brand. What remains of the brand effects after accounting for the confounding effects, are the isolated effects of brand equity. As a result, the model now includes location ‘loc’ and industry ‘ind’.
Eq. (4): 𝑉𝑖𝑗= 𝛽𝐵𝑏𝑗+ 𝛽𝑠𝑎𝑙𝑠𝑎𝑙𝑗 + 𝛽𝑙𝑜𝑐𝑙𝑜𝑐𝑗 +𝛽𝑖𝑛𝑑𝑖𝑛𝑑𝑗
Lastly, the individual’s perception of image associations has to be included into the model. As depicted earlier, this paper assumes that the brands differ in their perceptions about functional (fun), psychological (psych) and economic (econ) associations with brand awareness (aware) as a prerequisite of these associations. These perceptions are accounted for in our model as alternative specific covariates. Here, γ can be interpreted as the estimate of the marginal effects of the brand-specific image associations. Furthermore, the estimate for the brand β𝑏𝑟 should be interpreted as the residual of the brand effects
after accounting for brand associations (Eggers, Eggers & Kraus, 2016). Therefore, equation 5 shows the model after accounting for the image association constructs. Within this equation, the total brand effect can be considered to be the sum of the brand residual effect and the marginal effects of the respective covariates, as shown in equation 6:
Eq. (5): 𝑉𝑖𝑗= β𝑏𝑟 𝑏𝑗+ 𝛽𝑠𝑎𝑙𝑠𝑎𝑙𝑗 + 𝛽𝑙𝑜𝑐𝑙𝑜𝑐𝑗 +𝛽𝑖𝑛𝑑𝑖𝑛𝑑𝑗 + 𝛾𝑓𝑢𝑛𝑓𝑢𝑛𝑖𝑏 + 𝛾𝑝𝑠𝑦𝑐ℎ𝑝𝑠𝑦𝑐ℎ𝑖𝑏
+ 𝛾𝑒𝑐𝑜𝑛𝑒𝑐𝑜𝑛𝑖𝑏+ 𝛾𝑎𝑤𝑎𝑟𝑒𝑎𝑤𝑎𝑟𝑒𝑖𝑏
21
Lastly, the model needs to take the interaction effects between annual salary and brand equity into account. In order to do so, the model includes interaction effects between the four image association constructs and the annual salary. Therefore, the final model is represented in equation 7, in which 𝜂 represents the estimated coefficient of the interaction effect.
Eq. (7) 𝑉𝑖𝑗 = β𝑏𝑟 𝑏𝑗+ 𝛽𝑠𝑎𝑙𝑠𝑎𝑙𝑗 + 𝛽𝑙𝑜𝑐𝑙𝑜𝑐𝑗 +𝛽𝑖𝑛𝑑𝑖𝑛𝑑𝑗 + 𝛾𝑓𝑢𝑛𝑓𝑢𝑛𝑖𝑏 + 𝛾𝑝𝑠𝑦𝑐ℎ𝑝𝑠𝑦𝑐ℎ𝑖𝑏
+ 𝛾𝑒𝑐𝑜𝑛𝑒𝑐𝑜𝑛𝑖𝑏+ 𝛾𝑎𝑤𝑎𝑟𝑒𝑎𝑤𝑎𝑟𝑒𝑖𝑏 + 𝜂𝑓𝑢𝑛∗𝑠𝑎𝑙𝑓𝑢𝑛 ∗ 𝑠𝑎𝑙𝑖𝑏+𝜂𝑝𝑠𝑦𝑐ℎ∗𝑠𝑎𝑙𝑝𝑠𝑦𝑐ℎ ∗ 𝑠𝑎𝑙𝑖𝑏+𝜂𝑒𝑐𝑜𝑛∗𝑠𝑎𝑙𝑒𝑐𝑜𝑛 ∗
𝑠𝑎𝑙𝑖𝑏+𝜂𝑎𝑤𝑎𝑟𝑒∗𝑠𝑎𝑙𝑓𝑢𝑛 ∗ 𝑎𝑤𝑎𝑟𝑒𝑖𝑏
4.
Results
4.1 Sample statistics
22 Variables Answer possibilities
Mean (range) Sample (%) Age Gender Nationality Level of education Occupation in years male female
other/prefer not to answer Dutch
Chinese Other
No schooling completed High school graduate
Trade/technical/vocational training Bachelor’s degree Master’s degree Professional degree Doctorate degree Other Student Not a student 28 (18-70) 51,5 47,7 0.8 54,8 30,5 14,7 0,4 12,1 2,9 51,5 28,5 2,9 0,4 1,3 64 36
Table 4: Sample statistics 4.2 Data transformation
23
4.3 Comparison of estimation methods for joint and separated models
The estimation of effects was done by conducting several multinomial logistic regression analyses. Effect coding was used for the levels of employers, locations and industries. This allows for the interpretation of coefficients as a hypothetical deviation from the mean, which is zero (Eggers, Eggers & Kraus, 2016). Separating the effects of brand, location and industry can be done in multiple ways. The first of these methods involves estimating the entire model over the both the conjoint models. This is the simplest method to estimate the effects, however it does not account for the correlation between brand, industry and location effects in the second conjoint study. The second and third method both involve splitting up the estimation into two parts. The first part of the estimation for both these methods, estimates the brand and salary effects over the second conjoint model. The second part of estimation concerns the estimation of industry and location effects, for which the approach differs between the two methods. Whereas the second method estimates the effects over the first conjoint study, the third method estimates them over the second conjoint. This third method is possible since the estimation of the second part does not involve the estimation of employer effects. Therefore, the correlation between employers and their respective locations and industries is also not included. The comparison of the coefficients of the three methods can be found in table 5.
24 Joint
estimation
Separate estimation with both conjoint parts
Separate estimation with only part 2 separate combined separate combined Brand Rabobank .02(.13) .20(.10)* -.04 .20(.10)* .12 ING -.02(.13) .17(.10) -.04 .17(.10) .08 ASN BANK -.06(.14) -.17(.11) -.07 -.17(.11) -.22 Unilever .31(.13)* .52(.09)* .24 .52(.09)* .07 Douwe Egbert’s -.09(.12) .39(.09)* -.15 .39(.09)* -.12 Heineken .11(.12) .57(.09)* .06 .57(.09)* .05 KPN .10(.15) -.43(.12)* .13 -.43(.12)* -.06 TomTom -.05(.14) -.28(.11)* -.03 -.28(.11)* .05 Electrolight -.12(.14) -.34(.11)* -.12 -.34(.11)* .00 Eneco -.12(.14) -.33(.11)* -.15 -.33(.11)* -.10 NUON -.43(.14)* -.38(.11)* -.43 -.38(.11)* -.21 Shell .34(.13)* .07(.10) .33 .07(.10) .27 Industry Finance .07(.06) .08(.06) .06(.05) Consumer goods .36(.05)* .38(.06)* .49(.05)* Electronics -.36(.06)* -.38(.07)* -.36(.06)* Energy -.08(.06) -.0.08(.06) -.19(.06)* Location Rotterdam -.12(.06)* -.10(.06) -.04(.07) Amsterdam .14(.06)* .13(.06)* .03(.05) The Hague -.16(.06)* -.18(.06)* -.01(.06) Utrecht .15(.06)* .16(.06)* .02(.06) Salary 4.84(.23)* 5.45(.35)* 4.41(.29)* No-choice 5.02(.24)* 5.84(.38)* 4.40(.31)*
Note: *significant at 0.05 level Standard error in parentheses
25
4.4 Comparing different model extensions
In order to estimate the different equations mentioned in the methodology section, three different models were estimated. The first of these model is the benchmark model, which includes the effects for the brands, industries, locations and the annual salary. Therefore, the benchmark model represents the estimation of equation 4 from the methodology section. The second model is an extended version of the benchmark model and represents equation 5. The model extends the original model by including the brand specific covariates that relate to the brand image associations. Thus, this model takes the effects of brand awareness as well as functional, economic and psychological benefits into account. The final model further extends the extended model by accounting for the interaction effects between the brand equity constructs and the annual salary. This model is thus a representation of the 7th and final equation in the methodology section.
A comparison of the different models in terms of their model fit can be found in table 6. This table clearly shows that the model that accounts for the image associations and their respective interaction effects performs best on all the model fit criteria. Although, the final model has a fairly low R-square, it has a mean absolute error of 2.3%. This means that the predictions over the hold-out set based on the model are 2.3% off on average. Therefore, the model can be assumed to perform fairly well in predicting the job seeker’s perceived utility of a job offer. Based on likelihood ratio tests, it can be concluded that all of the models perform significantly better than a null model (all p<.001). Moreover, the extended model outperforms the benchmark model significantly (Chisq = 241.7, p< .001) in terms of log likelihood values. Finally, the extended model with interaction has a significantly better model fit than the extended model (Chisq= 13.0, p = .011). The estimations of each of the three models are represented in table 7, after which the results will be discussed
Benchmark model eq.(4) Extended model eq.(5) Extended + Interaction model eq.(7)
AIC (Akaike information criterion)
8615.4 8381.7 8376.7
Log-Likelihood -4288.7 -4167.9 -4161.4
McFadden 𝑹𝟐 .07 .10 .10
Mean absolute error 3.6% 2.6% 2.3%
26 Attribute
Benchmark model Extended model Extended + interaction model Employer Rabobank .02(.13) -.03(.13) -.04(.13) ING -.02(.13) -.18(.13) -.18(.13) ASN Bank -.06(.14) .11(.14 .12(.14) Unilever .31(.13)* -.10(.13) -.10(.13) Douwe Egbert’s -.09(.12) -.16(.13) -.17(.13) Heineken .11(.12) -.13(.12) -.15(.13) KPN .10(.15) .22(.15) .24(.15) TomTom -.05(.14) .10(.14) .11(.14) Electrolight -.12(.14) .16(.15) .17(.15) Eneco -.12(.14) .05(.15) .05(.15) NUON -.43(.14)* -.20(.14) -.19(.14) Shell .34(.13)* .15(.14) .15(.14) Industry Finance .07(.06) .08(.06) .08(.06) Consumer goods .36(.05)* .38(.06)* .38(.06) Electronics -.36(.06)* -.38(.07)* -.38(.07) Energy -.08(.06) -.08(.06) -.08(.06) Location Rotterdam -.12(.06)* -.10(.06)* -.10(.06)* Amsterdam .14(.06)* .12(.06)* .12(.06)* The Hague -.16(.06)* -.18(.06* -.18(.06* Utrecht .15(.06)* .15(.06)* .16(.06)* Annual salary 4.84(.23)* 4.91(.23)* 5.24(.30)* No – choice 5.02(.24)* 5.31(.24)* 5.64(.32)* Image associations Brand awareness -.13(.08) 1.67(.60)* Functional benefits .70(.09)* 0.64(.20)* Economic benefits .03(.09) -1.79(.72)* Psychological benefits .70(.08)* .49(.20)* Interaction effects Awareness * Salary -1.79(.57)* Functional * Salary .07(.22) Economic * Salary 1.79(.70)* Psychological * Salary .25(.22)
Note: *significant at 0.05 level Standard error in parentheses
27
4.5 Interpretation of benchmark model
The benchmark model, which estimates the effects of equation 4, shows overall face validity. The annual salary has a positive effect on the utility of a job offer (β= 4.84, p < .001), as is to be expected. This means that as the annual salary belonging to a certain job offer increases, the utility of that job offer increases as well. In this case a value of 1 represents an annual salary that is equal to 100 % of the expected salary indicated by the respondent. Under the assumption that the salary has a linear effect, an increase of 1 % in salary therefore equals a .04 increase in the job offer’s utility. Whereas in the separated estimation of brand equity effects most brands have a significant influence on the job utility, this model shows that only a few brands remain that have a significant influence after accounting for the location and industry effects. The brands Unilever (β=.31, p = .013) and Shell (β= 34, p = .004) seem to still positively influence the individuals perception of the job offer. In contrast, the brand of NUON (β=.43, p= .001) negatively influences this perception. The fact that only three of the original nine brands remain to have a significant impact on the job offer’s utility confirms the assumption that brand effects in extant literature are confounded by the effects of industry and location
In terms of industry effects, the industry of consumer goods seems to be most preferred among the respondents and significantly improves the perceptions of the job offers’ utility (β=.36, p < .001). This means that people have an increased preference for a job offer when it concerns a position in the industry of consumer goods. The opposite is true for the industry of electronics (β= -.36, p < .001). The prospect of working in this industry therefore decreases the perception of a job offer’s utility. For the industry of finance (β=.07, p= .203) and energy (β= -.08, p = .085) no significant effects were found. Concerning the locations, all of them seem to significantly influence job perceptions. Amsterdam (β=.14, p = .015) and Utrecht (β= .15, p = .009) have a significant positive influence on the job seeker’s perceptions. Thus, on average, people prefer to work in one of those cities relative to working in Rotterdam (β= -.12, p = .021) and The Hague (β= -.17, p = .005). Lastly, the no-choice option has a significantly positive effect on the job utility (β= 5.03, p < .001). This means that the total utility of the job offer should meet the threshold of 5.03 in order to be preferred over not choosing any of the job offers
4.6 Estimating the effects of brand specific constructs
28
is sufficient variation between the brands concerning their mean rating. However, it also hints towards some correlation between the constructs.
Brand Standard deviation in parentheses Awareness (-.13) Functional (.70)* Economic (.03) Psychological (.70)* Rabobank ,73(.44) ,48(.49) ,49(.50) ,23(.42) ING ,79(.41) ,58(.49) ,60(.49) ,33(.47) ASN bank ,54(.50) ,21(.41) ,24(.43) ,17(.38) Unilever ,85(.35) ,68(.47) ,50(.50) ,51(.50) Douwe Egbert’s ,68(.48) ,35(.48) ,23(.42) ,36(.48) Heineken ,81(.39) ,53(.50) ,46(.50) ,50(.50) KPN ,71(.46) ,36(.48) ,26(.44) ,20(.40) TomTom ,65(.48) ,27(.45) ,19(.39) ,23(.42) Electrolight ,06(.24) ,11(.32) ,08(.27) ,10(.30) Eneco ,58(.49) ,22(.41) ,18(.39) ,17(.38) NUON ,56(.50) ,20(.40) ,18(.38) ,15(.36) Shell ,88(.33) ,59(.49) ,59(.49) ,33(.47)
Table 8: Brand specific rating on image associations constructs
Table 9 shows that there is indeed significant correlation between the constructs. Although this is the case, including the constructs into the model does not result in multicollinearity issues (𝑉𝐼𝐹𝑎𝑤𝑎𝑟𝑒 =
1.82, 𝑉𝐼𝐹𝑓𝑢𝑛 =2.18, 𝑉𝐼𝐹𝑒𝑐𝑜𝑛 = 1.85, 𝑉𝐼𝐹𝑝𝑠𝑦𝑐ℎ =1.56). Therefore, the addition of the four constructs into
the model seems to be justifiable. Moreover, the earlier comparison of the models already showed that accounting for the constructs significantly improved the performance of the model.
** correlation significant at 0.01 level
Awareness Functional Economic Psychological
Awareness 1 0.62** 0.56** 0.48**
Functional 1 0.63** 0.56**
Economic 1 0.49**
Psychological 1
Table 9: Correlation matrix image associations constructs
29
effect was found. Thus, economic benefits cannot be assumed to have a significant influence on the job seeker’s perceived utility of a job offer. The same applies for the concept of brand awareness (β= -.13, p = .116). Consequently, being aware of the brand does not significantly influence perceptions of a job offer. In addition to the inclusion of the alternative specific covariates, some other differences between the extended and benchmark should also be mentioned. Whereas the effects for industry and locations are relatively unaffected, the estimates for the employers are vastly different from the benchmark model. However, as depicted in equation 6 in the methodology, the total brand effect can be calculated by adding the marginal effects of the covariates to the brand residual. Thus, the total brand effect for Rabobank is calculated as follows:
-.03 (brand residual for Rabobank) - .13 * .73 (effect of brand awareness) + .70 * .48 (effect of functional benefits) + .03 * 49 (effect for economic benefits) + .70 * .23 (effect for psychological benefits) = .39 (total brand effect for Rabobank)
30
Brand Residual
value
Total brand effect Brand effect zero centered Benchmark brand estimate Rabobank -0,03 0,39 0,00 0,02 ING -0,18 0,37 -0,01 -0,02 ASN bank 0,11 0,31 -0,07 -0,06 Unilever -0,10 0,64 0,25 0,31 Douwe Egbert’s -0,16 0,26 -0,13 -0,09 Heineken -0,13 0,50 0,12 0,11 KPN 0,22 0,53 0,14 0,1 TomTom 0,10 0,37 -0,01 -0,05 Electrolight 0,16 0,30 -0,08 -0,12 Eneco 0,05 0,25 -0,13 -0,12 NUON -0,20 -0,02 -0,41 -0,43 Shell 0,15 0,70 0,31 0,34
Table 10: Comparison of brand effect for extended and benchmark model 4.7 Investigation of interaction effects
31
salary. This means that the importance of the annual salary is rather unaffected by whether the job seeker perceives to receive either functional or psychological benefits.
Lastly, the type of analysis used allows for the interpretation of relative importance by dividing the range of an attribute by the sum of the ranges of all attributes. These importance scores can be found in table 11. The table shows that the annual salary is the most important attribute of a job offer. The total importance of the brand effects however, is only slightly smaller at 44.2 %, considering that the image associations constructs are brand specific covariates. The importance for industry and location are considerably lower. However as discussed earlier, these attributes can still significantly influence the perception of a job offer
Attribute attribute range attribute importance
Brand residuals 0,43 3,8% industry 0,76 6,7% location 0,34 3,0% Salary 5,24 46,1% Awareness 1,67 14,7% Functional 0,64 5,6% Economic 1,79 15,8% Psychological 0,49 4,3%
Table 11: Relative importance scores
Table 12 provides a full overview of all the hypotheses and whether they are supported or not. The findings from this table will then be further discussed in more detail in the discussion section.
Hypothesis/ Supported/rejected
H1 Higher levels of brand equity will increase the utility of job offers. supported H2 Higher levels of brand awareness will result in higher levels of brand
equity
supported
H3a The amount of perceived functional benefits will have a positive effect on brand equity.
Supported
H3b The amount of perceived economic benefits will have a positive effect on brand equity.
rejected
H3c The amount of perceived psychological benefits will have a positive effect on brand equity.
supported
H4a A higher starting salary will increase the utility of job offers. supported H4b The effects for starting salary will be weaker under higher levels of
brand equity.
Partially supported
32
5. Discussion
5.1 Discussion of findings
The aim of this study was to find out how brand equity effects influence the utility of a job offer in relation with other job characteristics. In particular, separating the brand effects from the effects of their respective industry and location so only the net effect of the brand remains. Moreover, this study tried to split up the construct of brand equity into several components to give insights into how managers could improve their employer branding efforts. In order to study these effects, a survey including two conjoint studies was conducted amongst a group 241 respondents. After extending the model with brand specific image associations and their respective interaction effects, the final model was estimated following a multinomial logit procedure. When comparing this model to a model where only brand effects are included, it can be concluded that the effects of brand equity are indeed confounded by the effects of their respective location and industry. Although after accounting for these effects only a few brands remain significant, the model does show that having higher levels of brand equity can significantly influence the job seekers perception of the job offer’s utility. Therefore, the employee based brand equity can be considered to have similar effects as customer based brand equity on an individual’s perception of a certain offer. As Backhaus and Tikoo (2004) implied, the job seeker can react differently to similar job offers based on the concept of brand equity.
33
from employment at a certain company, their perception of the job offer improves. Both of these effects were as expected and therefore their respective hypotheses (H3a and H3c) are supported.
The hypothesis for the economic benefits however could not be supported. This construct, mainly concerning whether the individual expects to be fairly compensated, was found to have no significant impact on the perception of utility for a job offer. This would support the finding by Sivertzen, Nilsen and Olafsen (2013), who state that the economic dimension from the EmpAt measure by Berthon (2005) does not have a significant impact. However, after accounting for the interaction effect with annual salary, the effect for economic benefits even becomes significantly negative. This might be accountable for by the fact that economic benefits include the expected salary. Therefore, after accounting for the effect of the actual salary, the utility of perceiving to attain economic benefits might decrease. Moreover, investigating the willingness to pay for the less preferred brands would result in a negative value, meaning that people want to be paid more in order to work at those companies. Therefore, people might be expecting more economic benefits for the less preferred brands. This is common practice in real life situations, as companies from unattractive industries pay more to attract qualified staff. Lastly, another possible explanation could be that there are unmet expectations. As the respondent expects to be fairly compensated but is then presented with a salary that is below expectation, the individual might be dissatisfied. Dissatisfaction as a result of unmet expectations was shown in previous research to negatively influence an individual’s behavioral intentions (Gupta & Stewart, 1996). The interaction effect itself is significantly positive, which could again be attributable to the expectations that the job seeker has of a certain brand. Thus, when the individual expects to be fairly compensated, the actual value of the annual salary becomes more important to them.
34
5.2 Theoretical and managerial implications
In contrast to earlier papers in the field of employee based brand equity, this paper considers the effects of brands, industries and locations on job attractiveness separately. Since the location and industry are tied to a certain brand, generally these constructs are estimated as a single construct. This paper however, shows that the industry and location effects confound the effects of brand equity. This means for the field of research that these effects should be taken into consideration when estimating the effects of brand equity in order to find the true effect of the brand. If these factors are not taken into account, the effects of brand equity might be over or underestimated depending on their respective location or industry. What this means for managers is that they should not only consider the image of the brand, but they should also consider carefully in which city the company should be located as this could significantly influence the attractiveness of a job offer. Therefore, being located in the right city could create a competitive advantage for the brand when competing for talented staff. In terms of industry, companies in the industries that currently negatively affect the individual’s perception of job attractiveness could consider to collectively try and improve the perception of working in that specific industry.
35
5.3 Limitations and further research
This paper has some interesting findings and important implications for the field of research. However, it also has some shortcomings. One of the shortcomings of this paper could be the length of the survey experiment. Although the sample size was sufficient at 243 respondents, there were 91 respondents that started the survey, but did not complete it. A pick-any approach was already applied to decrease the amount of decisions the respondents had to make. Yet, the survey could still be considered to involve a lot of decisions. Therefore, it is assumed that the fact that people did not finish the survey is at least partly due to the survey length. Having a shorter survey could have resulted in a larger sample size and therefore more generalizability of findings. Moreover, it is possible that some respondents lost their attention during the survey which could possibly bias their answers. On the other side, in order to keep the length of the survey short, the image association constructs were measured using a single item scale. This might not accurately represent the underlying aspects of the construct as a scale with multiple items would. In order to confirm the findings of this research, future research might therefore try analyzing the effects separately. This would allow for the usage of multiple item scales without running into problem with survey length.
36
6. References
Aaker, D.A. (1991), Managing Brand Equity: Capitalizing on the Value of a Brand Name, The Free Press, New
York.Ambler, T. and Barrow, S. (1996). The employer brand. Journal of Brand Management, 4, 185–206.
Backhaus, K., & Tikoo, S. (2004). Conceptualizing and researching employer branding. Career development
international, 9(5), 501-517.
Backhaus, K. B., Stone, B. A., & Heiner, K. (2002). Exploring the relationship between corporate social performance and employer attractiveness. Business & Society, 41(3), 292-318.
Baker, W., Hutchinson, J., Moore, D., & Nedungadi, P. (1986). Brand familiarity and advertising: effects on the evoked set and brand preference. Advances in Consumer Research, 13, 637-642
Bech, M., Kjaer, T., & Lauridsen, J. (2011). Does the number of choice sets matter? Results from a web survey applying a discrete choice experiment. Health economics, 20(3), 273-286.
Berthon, P., Ewing, M., & Hah, L. L. (2005). Captivating company: dimensions of attractiveness in employer branding. International journal of advertising, 24(2), 151-172.
Borah, A., & Tellis, G. J. (2016). Halo (spillover) effects in social media: do product recalls of one brand hurt or help rival brands?, Journal of Marketing Research, 53(2), 143-160.
Cobb-Walgren, C. J., Ruble, C. A., & Donthu, N. (1995). Brand equity, brand preference, and purchase intent. Journal of advertising, 24(3), 25-40.
Cox, J. (2018, March 15). Unilever, Britain's third biggest company, moves its headquarters to Rotterdam after almost a century in London. Retrieved June 12, 2018 from:
https://www.independent.co.uk/news/business/news/unilever-moves-headquarters-rotterdam-london-brexit-100-years-a8256736.html
DelVecchio, D., Jarvis, C. B., Klink, R. R., & Dineen, B. R. (2007). Leveraging brand equity to attract human capital. Marketing Letters, 18(3), 149-164.
Eggers, F., Eggers, F., & Kraus, S. (2016). Entrepreneurial branding: measuring consumer preferences through choice-based conjoint analysis. International Entrepreneurship and Management Journal, 12(2), 427-444. Eggers, F., and H. Sattler (2011). Preference Measurement with Conjoint Analysis: Overview of State of-the-art Approaches and Recent Developments. GfK Marketing Intelligence Review, 3(1), 36–47.
Fung, C.K., Min, K.Z. and Yong, W.W. (1996), “Job choice and job mobility for Chinese young professionals”, Hong
Kong Manager, November-December, 20-5.
Gatewood, R. D., Gowan, M. A., & Lautenschlager, G. J. (1993). Corporate image, recruitment image and initial job choice decisions. Academy of Management journal, 36(2), 414-427.
Gordon, P. C., & Holyoak, K. J. (1983). Implicit learning and generalization of the" mere exposure" effect. Journal of
Personality and Social Psychology, 45(3), 492.
Gupta, K., & Stewart, D. W. (1996). Customer satisfaction and customer behavior: the differential role of brand and category expectations. Marketing Letters, 7(3), 249-263.
37
Keller, K.L. (1993). Conceptualizing, Measuring, and Managing Customer-Based Brand Equity, Journal of Marketing, 57 (January), 1–22.
Lievens, F., & Highhouse, S. (2003). The relation of instrumental and symbolic attributes to a company's attractiveness as an employer. Personnel psychology, 56(1), 75-102.
Loo, R. (2002). A caveat on using single-item versus multiple-item scales. Journal of managerial psychology, 17(1), 68-75.
Manski, C. F. (1977). The structure of random utility models. Theory and decision, 8(3), 229-254.
Petty, R. E., Cacioppo, J. T., & Schumann, D. (1983). Central and peripheral routes to advertising effectiveness: The moderating role of involvement. Journal of Consumer Research, 10(2), 135-146.
Reber, R., Winkielman, P., & Schwarz, N. (1998). Effects of Perceptual Fluency on Affective Judgments. Psychological Science, 9(1), 45-48.
Rossiter, J. R. & Percy, L. (1987), Advertising and Promotion Management. New York: McGraw-Hill Book Company. Saks, A. M., Wiesner, W. H., & Summers, R. J. (1996). Effects of job previews and compensation policy on applicant attraction and job choice. Journal of Vocational Behavior, 49(1), 68-85.
Sivertzen, A. M., Nilsen, E. R., & Olafsen, A. H. (2013). Employer branding: employer attractiveness and the use of social media. Journal of Product & Brand Management, 22(7), 473-483.
Simpson, P. M., Siguaw, J. A., & Enz, C. A. (2006). Innovation orientation outcomes: The good and the bad. Journal
of Business Research, 59(10-11), 1133-1141.
Sonnier, G., & Ainslie, A. (2011). Estimating the value of brand-image associations: The role of general and specific brand image. Journal of Marketing Research, 48(3), 518-531.
Starr, M.K., and Rubinson, J.R. (1978), "A Loyalty Group Segmentation Model for Brand Purchasing Simulation," Journal of Marketing Research, 15 (August), 378- 83.
Tavassoli, N. T., Sorescu, A., & Chandy, R. (2014). Employee-Based Brand Equity: Why Firms with Strong Brands Pay Their Executives Less. Journal of Marketing Research, 51(6), 676-690.
Theurer, C. P., Tumasjan, A., Welpe, I. M., & Lievens, F. (2016). Employer Branding: A Brand Equity based Literature Review and Research Agenda. International Journal of Management Reviews
Turban, D. B., & Greening, D. W. (1997). Corporate social performance and organizational attractiveness to prospective employees. Academy of management journal, 40(3), 658-672.
Wernerfelt, B. (1984). A resource‐based view of the firm. Strategic management journal, 5(2), 171-180. Wingfield, N., & Cohen, P. (2017, September 07). Amazon Plans Second Headquarters, Opening a Bidding War Among Cities. Retrieved February 27, 2018, from https://www.nytimes.com/2017/09/07/technology/amazon-headquarters-north-america.html
38
7.
Appendices
7.1 Appendix-A: Example survey
Example conjoint part 1
Example conjoint part 2
39 7.2 Appendix-B: R-code: rm(list= ls()) setwd("~/scriptie data") library(mlogit) dat <- read.csv("Employer_Branding_CBC.csv") dat <- dat[!(dat$X9_salary == 0), ] library(readxl) Employer_Branding_Data <- read_excel("Employer_Branding_Data.xlsx") View(Employer_Branding_Data) summary(dat$X9_salary) table(dat$X9_salary) # convert data
dat <- mlogit.data(dat, choice="Selection_Dummy", shape="long", alt.var="Alternative_id")
summary(Employer_Branding_Data$X6_age) table(Employer_Branding_Data$X5_gender) table(Employer_Branding_Data$X10_nationality) table(Employer_Branding_Data$X24_education) table(Employer_Branding_Data$X23_occupation) summary(Employer_Branding_Data$X22_work.experience) # calculate models
ml1<- mlogit(Selection_Dummy ~ Employer.Rabo + Employer.ING + Employer.asn + Employer.Unilever + Employer.DE + Employer.Heineken +
Employer.KPN + Employer.Tomtom + Employer.EL + Employer.Eneco + Employer.Nuon +
Industry.Banking + Industry.Consumer + Industry.Electronics + Location.The.Hague + Location.Amsterdam + Location.Utrecht + Annual.Salary.val + None_option | 0, dat)
summary(ml1)
# Recover reference levels
40
# recover standard errors and ref.level std error covMatrix <- vcov(ml1)
Employer.Shell.Std.Error <- sqrt(sum(covMatrix[1:11, 1:11])) Employer.Shell.zValue <- Employer.Shell / Employer.Shell.Std.Error Industry.Energy.Std.Error <- sqrt(sum(covMatrix[12:14, 12:14])) Industry.Energy.zValue <- Industry.Energy/ Industry.Energy.Std.Error Location.Rotterdam.std.Error <- sqrt(sum(covMatrix[ 15:17, 15:17]))
Location.Rotterdam.zValue <- Location.Rotterdam/ Location.Rotterdam.std.Error
# WTP
coef(ml1)[1:11]/coef(ml1)["Annual.Salary.val"] Employer.Shell/coef(ml1)["Annual.Salary.val"]
### Use only first CBC without Employer Brand
# convert data
dat1 <- mlogit.data(dat[which(dat$CBC_part == 1),], choice="Selection_Dummy", shape="long", alt.var="Alternative_id") # calculate models
ml2<- mlogit(Selection_Dummy ~ Industry.Banking + Industry.Consumer + Industry.Electronics + Location.The.Hague + Location.Amsterdam + Location.Utrecht +
41
# convert data
dat2 <- mlogit.data(dat[which(dat$CBC_part == 2),], choice="Selection_Dummy", shape="long", alt.var="Alternative_id") # calculate models
ml3<- mlogit(Selection_Dummy ~ Employer.Rabo + Employer.ING + Employer.asn + Employer.Unilever + Employer.DE + Employer.Heineken +
Employer.KPN + Employer.Tomtom + Employer.EL + Employer.Eneco + Employer.Nuon +
Annual.Salary.val + None_option | 0, dat2) summary(ml3)
covMatrix4 <- vcov(ml3)
Employer.Shell2 <- (-1)* sum(coef(ml3)[1:11])
Employer.Shell2.Std.error <- sqrt(sum(covMatrix4[1:11,1:11])) Employer.Shell2.zValue <- Employer.Shell2/Employer.Shell2.Std.error # compare estimates from ml2 and ml4
ml4<- mlogit(Selection_Dummy ~ Industry.Banking + Industry.Consumer + Industry.Electronics + Location.The.Hague + Location.Amsterdam + Location.Utrecht +
Annual.Salary.val + None_option | 0, dat2) summary(ml4) cbind(coef(ml2)[1:6], coef(ml4)[1:6]) covMatrix3 <- vcov(ml4) Industry.Energy3 <- (-1) * sum(coef(ml4)[1:3]) Industry.Energy3.Std.Error <- sqrt(sum(covMatrix3[1:3,1:3])) Industry.Energy3.zValue <- Industry.Energy3/Industry.Energy3.Std.Error Location.Rotterdam3 <- (-1) * sum(coef(ml4)[4:6]) Location.Rotterdam3.Std.Error <- sqrt(sum(covMatrix3[4:6,4:6])) Location.Rotterdam3.zValue <- Location.Rotterdam3/Location.Rotterdam3.Std.Error ### Add covariates dat$Awareness <- 0
42
dat$Awareness[which(dat$Employer == 6 & dat$X9_awareness_.img....files.logo_Heineken.png..img. == 1)] <- 1 dat$Awareness[which(dat$Employer == 7 & dat$X9_awareness_.img....files.logo_KPN.png..img. == 1)] <- 1 dat$Awareness[which(dat$Employer == 8 & dat$X9_awareness_.img....files.logo_Tomtom.png..img. == 1)] <- 1 dat$Awareness[which(dat$Employer == 9 & dat$X9_awareness_.img....files.logo_EL.png..img. == 1)] <- 1 dat$Awareness[which(dat$Employer == 10 & dat$X9_awareness_.img....files.logo_Eneco.png..img. == 1)] <- 1 dat$Awareness[which(dat$Employer == 11 & dat$X9_awareness_.img....files.logo_Nuon.png..img. == 1)] <- 1 dat$Awareness[which(dat$Employer == 12 & dat$X9_awareness_.img....files.logo_Shell.png..img. == 1)] <- 1
dat$functional <- 0
dat$functional[which(dat$Employer == 1 & dat$X10_functional.benefits_.img....files.logo_Rabobank.png..img. == 1)] <- 1 dat$functional[which(dat$Employer == 2 & dat$X10_functional.benefits_.img....files.logo_ING.png..img. == 1)] <- 1 dat$functional[which(dat$Employer == 3 & dat$X10_functional.benefits_.img....files.logo_asn_bank.png..img. == 1)] <- 1 dat$functional[which(dat$Employer == 4 & dat$X10_functional.benefits_.img....files.logo_Unilever.png..img. == 1)] <- 1 dat$functional[which(dat$Employer == 5 & dat$X10_functional.benefits_.img....files.logo_DE.png..img. == 1)] <- 1 dat$functional[which(dat$Employer == 6 & dat$X10_functional.benefits_.img....files.logo_Heineken.png..img. == 1)] <- 1 dat$functional[which(dat$Employer == 7 & dat$X10_functional.benefits_.img....files.logo_KPN.png..img. == 1)] <- 1 dat$functional[which(dat$Employer == 8 & dat$X10_functional.benefits_.img....files.logo_Tomtom.png..img. == 1)] <- 1 dat$functional[which(dat$Employer == 9 & dat$X10_functional.benefits_.img....files.logo_EL.png..img. == 1)] <- 1 dat$functional[which(dat$Employer == 10 & dat$X10_functional.benefits_.img....files.logo_Eneco.png..img. == 1)] <- 1 dat$functional[which(dat$Employer == 11 & dat$X10_functional.benefits_.img....files.logo_Nuon.png..img. == 1)] <- 1 dat$functional[which(dat$Employer == 12 & dat$X10_functional.benefits_.img....files.logo_Shell.png..img. == 1)] <- 1
dat$economic <- 0
