1
Employer Branding
How important is the employer’s brand name and what is its influence on job seekers’ employment decision?
2
Employer Branding
How important is the employer’s brand name and what is its influence on job seekers’ employment decision?
University of Groningen Faculty of Economics and Business
Master Thesis
Master Marketing Management & Marketing Intelligence
June 22, 2015
Irene Suzan Gercama Lissabonstraat 41 9718 AX, Groningen S1996665 i.s.gercama@student.rug.nl +31 (0)6 28 44 25 66 Supervisors
3
Abstract
Designing job offers can be done in two ways, the first scenario is characterized by a vacancy in which is stated for which company someone would work, the second scenario is the so-called headhunter- or anonymous scenario, which is characterized by not revealing the company from which the job offer is originated. This study aims at finding out whether an employer brand causes a differential effect on job seekers’ reaction to job attributes such as salary, location, promotion possibilities, and job security. Existing literature about the effects of brands and about job attributes is discussed in the literature review. In order to identify which of these job attributes has the highest influence on perceived job utility and how the effect of these attributes changes when an employer brand is included, a conjoint experiment is conducted for the two different scenarios. Results demonstrate that job security is the most important attribute of a job offer. Furthermore, it is found that the reaction of job seekers does not differ across the two different scenarios. It can thus be concluded that employer brand equity does not have a significant differential effect on how job seekers respond to the included job attributes. Only one maringal effect was found, which states that when an
employer brand name is given in the vacancy, it becomes less important for the job seeker that the job is located in the same area as where he is currently living. However, since this is just a marginal effect and only holds in specific situations, one should be cautious in deriving conclusions from this. Novertheless, this research still has some important managerial
implications since this study reveals that job security is the most important determinant of the job decision. In this way, human resource managers can better tailor their job offers to the needs of the job seeker.
4
1. Introduction
Due to the aging population, the supply of skilled, value-adding, specialist employees will decrease (Moroko & Uncles, 2008). It is expected by the OECD that by the year of 2050, ten active workers have to support more than seven inactive people, while that ratio was much lower in previous years (Moroko & Uncles, 2008). It is expected that this deceleration in labor supply will last for yet another half-century (Fallick & Pingle, 2008). While supply will decrease over the upcoming years, the demand for skilled and specialized labor will increase, especially because emerging markets are developing and will demand skilled labor more and more (Moroko & Uncles, 2008).
These developments increase the importance for companies to differentiate themselves in order to attract talented employees. In principle, job seekers are interested in working in a particular industry, but it might be hard for them to see the differences between certain companies within that industry, as jobs and organizations within the same industry are often quite similar (Maurer, Howe, & Lee, 1992). This increases the need for companies to stand out in order to be the first choice of talented job seekers. Standing out could be achieved by an effective use of the brand.
As the marketing literature has shown (Ahmad & Thyagaraj, 2014), the use of brands is not anymore just a manner to differentiate one’s product or service, but brands became vital for the success of an organization. Due to the brand’s increased importance, it is important to engage in effective brand management in order to maximize profits (Keller, 2008). This brand management is not only important for consumer goods, but it is important for the employer brand as well, especially now that the labor supply is decreasing and companies still want to attract talented employees. This employer branding can be defined as the sum of a company’s efforts to communicate that it is a desirable place to work to current- and potential staff (Lloyd, 2002).
5 1993). Examples of such outcomes that are attributable to the brand name are repeat purchases (Leone et al., 2006) and the attraction of more, and more talented job seekers (Turban & Cable, 2003).
It is interesting to research whether such job seekers’ knowledge about an employer brand causes a differential effect on the response to job elements such as salary, promotion possibilities, and job security. In this way, it can be determined whether the importance and the effect of these job attributes changes when there is an employer involved with a high employee-based brand equity.
Even though research demonstrates that employer brand names matter (Turban & Cable, 2003), there are cases where outstanding vacancies do not reveal the name of the employer. This is the so-called headhunter- or anonymous scenario. A reason why companies do not wish to reveal their name in the vacancy could be that they do not want to show the competitors in the marketplace that there is a weak spot in the organization that they are trying to fill (Clark, 2011). It seems that such headhunter vacancies are at a disadvantage, since literature has found that a company with a high brand employer equity will attract more, and more talented job seekers (Turban & Cable, 2003). This makes it very interesting to isolate the effect of the employer brand name and see how job seekers react to vacancies where the employer brand name is revealed versus unrevealed employer brand vacancies. By comparing these two scenarios, it will become clear whether and how an employer brand name influences a job-seeker’s response towards other job attributes such as location and salary, and how this will influence employment decisions.
Whereas previous studies focused on how working for a high employer brand might influence the future career (DelVecchio et al., 2007) and on the perception of the employer (Cable & Graham, 2000), this research will contribute to the existing literature by focusing on the concept of employee-based brand equity. Anonymous scenarios will be compared with branded-scenarios in order to isolate this employer brand equity and assess its effect. Other job attributes that contribute to the job-utility will also be taken into account and it will be researched how the importance and the effect of these attributes changes when there is a high equity employer brand involved compared to no brand at all. This is methodologically interesting since isolating and calculating employer brand equity is not trivial. The
comparison of these two scenarios will shed light on the question whether job seekers respond differently towards the remaining job attributes and thus whether, and to what extent,
headhunter vacancies are at a disadvantage compared to branded vacancies.
6
(Cable & Graham, 2000) and that job seekers are more likely to apply for a company with a strong brand (Turban & Cable, 2003). However, it has not yet been researched how a strong employer brand influences a job seeker’s reaction to the remaining job attributes. The research question that this research will try to answer is thus as follows:
How does employer brand equity influence a job seeker’s response to the remaining job attributes?
From this main research questions, the following two sub questions can be derived: 1. How does employer brand equity influence a job seeker’s choice concerning which
vacancy to apply for?
2. If job seekers’ knowledge about an employer brand influences the effect of the remaining job attributes, what would this imply for the headhunter scenario? This research has important managerial relevance, as it will show how important the
employer’s name is when job-seekers need to make a decision as to for which company they want to apply. It will also demonstrate whether the employer’s brand name have a differential effect on the remaining job attributes, such as promotion opportunities and job security. These results would have implications for both the human resource department and the marketing department, as it could turn out that companies need to focus more on managing their employment brand and use branded vacancies rather than headhunter vacancies.
The next section will provide a theoretical framework where the job attributes that influence the perceived job utility, the moderators, and the hypotheses will be described. That part will be closed by providing the conceptual model. In the third section, the methodology used for conducting this research will be described. This part will be followed up by
describing the results of the study and discussing the hypotheses’ outcomes. The fifth chapter will be devoted to the conclusion of this research, to managerial recommendations and to directions for future research.
2. Literature review
In this section, all the relevant job attributes and moderators will be discussed by using the existing literature. The job attributes that will be focused on are: employer brand name, as it was found that a good employer brand will attract a lot- and qualified job applicants
7
2003). Location is included since it is found that this attribute is among the most important attributes affecting job utility (Lievens & Highhouse, 2003). The promotion possibilities attribute is included as it is demonstrated that job seekers are more eager to accept a job offer that offers promotion possibilities (Isenhour, Lukaszweski, & Stone, 2014; Arachchige & Robertson, 2013). Lastly, job security is taken into account as the research of Noe et al. (2013) demonstrate that this attribute is one of the most important influencers of vacancy attractiveness. Since the aforementioned job attributes received a lot of attention in previous researches, they are included in this research as well.
The first moderator included in this research is brand name, where there are two scenarios, namely a branded scenario and an anonymous scenario. This moderator will make it possible to determine the differential impact of employer brand name on the response of job seekers to the remaining job attributes. The second moderator is brand equity, which is
included in order to determine whether the level of employer brand equity will have an influence on the reaction to the included job attributes.
2.1 Employer Brand Name
Previous literature found that employer names matter, as having a high employer brand equity leads to more and better job applicants, higher quality interviews, and thus better qualified employees (Turban & Cable, 2003). Not only do organizations with a high brand employer equity attract talented job seekers, but such a talented workforce also has a positive direct impact on organizational performance (Cable & Graham, 2000).
8
significantly in order to compete against a high equity brand. In the human resource area, this could mean that competitors of a high equity employer brand might need to provide higher salaries to compete with the high equity brand. These findings illustrate that having a strong equity brand is important and brings many advantages with it. The value of a brand to consumers is defined as consumer-based brand equity (Keller, 1993). Thus, employee-based brand equity reflects the value of a brand to employees.
Being a consumer’s first-choice-brand has also important implications for employers who are looking for qualified employees (Rampl, 2014). The current literature confirms that that being an attractive employer brand is important when job seekers have to choose a future employer (Edwards, 2010; Wilden, Gudergan, & Lings, 2010). This is for example shown by the study performed by Turban and Cable (2003). They demonstrated that being a good employer brand leads to more- and more qualified job applicants. Because of this high importance attached to employer brand name, companies are putting a lot of effort in
developing and presenting an image which will serve to attract both many- and qualified job applicants (Arachchice & Robertson, 2013). They do this, as it is important for an employer to be known, noticeable, and different from your competitors (Moroko & Uncles, 2008).
Managing the employer brand does pay-off: job seekers pay a lot of attention to the employer’s brand and they prefer to work for a certain company when working for that company will benefit their self-esteem (Cable & Turban, 2003) and social standing (Tajfel, 1978). A reason for this preference for an organization with a good reputation, might be that a job seeker is firstly attracted by an organization’s job offer because of its favorable perception of the reputation of the organization (Belt & Paolillo, 1982; Gatewood et al., 1993). Cable and Turban (2003) confirm that organizational reputation is a very important brand association and that it has an effect on employer brand attractiveness. This employer brand attractiveness will lead to a desire of the job seeker to work for that specific brand (Collins & Stevens, 2002). Consequently, the organization’s image is a very strong predictor of job pursuit intention (Chapman, 2005).
Although many researchers state that brand name is positively influencing job attractiveness, Arachchice and Robertson (2013) found that being a well-known employer is one of the least preferred organizational characteristics. Their research shows that job
9
following the definition of Keller (2008), which states that people are reacting differently to certain aspects when a particular brand is involved.
H1: Brand employer name has a positive influence on job utility.
H2: If the level of employee-based brand equity increases, the perceived level of job utility will increase as well.
2.2 Moderating role of brand name
Marketing managers nowadays recognize the value of possessing a strong brand, a strong brand being one that creates positive associations and has a high level of awareness (Keller, 2003). They are willing to dedicate a lot of resources to build and maintain such brand
strength. The current academic literature about the value of brands has mainly focused on how companies can use their brands in order to win customers, however, such a focus is actually incomplete because it underestimates a brand’s real contribution to an organization
(Tavassoli, Sorescu, & Chandy, 2014). Firms not only compete for customers, but they also try to win the battle for employees. The benefits of branding can thus be extended by the concept of employee-based brand equity (Tavassoli, Sorescu, & Chandy, 2014).
When a job seeker is looking for a job, it is impossible to fully understand the quality and the characteristics of a prospective employer beforehand (Wilden, Gudergan, & Lings, 2010). Due to the long-term implications of accepting a job offer, it is important that the job seeker looks for ways in which this information asymmetry can be reduced. Previous research demonstrated that investigating recruiting from a marketing perspective is valuable, as the employer brand sends appropriate signals which enables potential employees to reduce information costs associated with gaining information about prospective employers (Wilden, Gudergan, & Lings, 2010). The employer brand communicates messages to job seekers about the quality of a company as an employer (Wilden, Gudergan, & Lings, 2010). An employer brand name signals to the employment market what the competencies and characteristics are of the company as an employer (Spence, 1974).
When no employer brand name is revealed in a vacancy, job seekers can not
10
divided into two scenarios. In this way, it is possible to assess whether job seekers respond differently to the job attributes in the two different scenarios and thus discover whether a job seekers’ knowledge about an employer brand influences how the remaining job attributes are perceived. The hypothesis for this moderator is thus as follows:
H3: Including an employer brand name causes the job seekers to respond differently to the remaining job attributes compared to when no employer brand name is included. 2.3 Moderating role of Employer Brand Equity
Since more and more brands are emerging and companies are developing and introducing many brand extensions, consumers have increased choices when it comes to choosing which product or service to purchase (Ekström, 2010), or which company to work for. In order for a company to stand out of the crowd, it should focus on building and maintaining brand
strength via increasing brand familiarity and creating positive brand associations. This will increase brand equity, and as Keller (1993) stated, brand equity will lead to more positive reactions to certain elements of the concerned brand compared to the same element of another brand. In order to find out whether Keller’s (2003) definition about brand equity holds equally well in the human resource area as in the consumer market, employer brand equity is included as a moderator in this research as well.
In the current literature, there is some evidence that a high equity employer brand does have a differential impact on a job seeker’s response to certain job attributes. Tavassoli, Sorescu and Chandy (2014) found that when someone has the opportunity to work for a respected and strong employer brand, he is willing to sacrifice other job-related variables, such as accepting a lower salary. This would mean that when a job seeker has the opportunity to work for a high equity brand employer, the employer brand name will influence the job seeker’s response to the other job attributes which will result in these job attributes becoming less important.
It is also found that job seekers, and especially those who do not have any work experience yet, are eager to apply for a job offer from a company with a strong corporate brand. (Wilden, Gudergan, & Lings, 2010). The driving force behind this is that such a high equity employer brand would look good on their CV. When employees have been employed for a high equity employer, they will give a powerful signal of competency, which makes them more attractive to future employers (DelVecchio et al., 2007). As Keller (2003)
11
mix elements of that brand, but also to job offers from that brand. It is expected that the higher the brand equity is of the employer, the stronger this differential effect will be. In line with the research of Tavassoli et al. (2014), it is expected that a vacancy of a high equity employer brand, will make the effect that the remaining job attributes have on job utility, less strong. This gives the following hypothesis:
H4: The higher the brand equity of the included employer brand, the less strong will the relationship be between the remaining job attributes and perceived job utility. 2.4 Salary
It has already been found decades ago that payment is an important job attribute (Jurgensen, 1978). Job seekers perceive salary as important and this influences the attractiveness of a certain job offer (Rynes, 1978). It is much more likely that a job applicant will perceive the job and employer as positive when the salary is well above average (Honeycutt & Rosen, 1997). This also increases the probability of the job seeker to apply for the respective company (Honeycutt & Rosen, 1997). Lievens and Highhouse (2003) even found in their research that salary is one of the best predictors of employer brand attractiveness. And as previous research has shown, when the employer brand attractiveness increases, the desire of to work for that brand increases (Collins & Stevens, 2002), and this will in turn increase the likelihood of application.
While some research stated that a high pay level would result in a higher likelihood of starting to work for that company, there are also researches conducted that do not find the same results. DelVecchio et al. (2007) found that undergraduate students would accept a lower salary if they would work for a strong brand compared to a weak brand. This latter research is in line with the findings of Tavassoli et al. (2014), who found that top executives are paid less when they are working for an employer with high brand equity. Since salaries make up approximately 20-50 per cent of the total operating expenses (Society for Human Resource Management, 2008), being a high employer brand and being able to pay your executives less, really does have its perks.
12 H5: The higher the salary level, the higher the perceived utility of the job. 2.5 Location
It stems from the literature that location, among other job attributes, has the biggest impact on job acceptance decisions (Taylor & Bergmann, 1987). It is even said that when these job attributes are known and taken into account, recruitment activities do not influence job acceptance decisions anymore (Turban, 2001). This is confirmed by Boswell et al. (2003), who state that as job seekers move through the stages of the job-choice process, important factors such as the location influence job seekers’ decisions. It is even found that locational aspects are one of the most important factors influencing employer brand attractiveness (Lievens & Highhouse, 2003). Barber and Roehling (1993) stated that location is probably used as a characteristic to screen out job opportunities, as they found that in almost half of the cases in which an applicant turned down a job opportunity, location was the reason.
Apparently, a lot of attention is paid to location when a job seeker reviews a job posting. It turns out that people want to work close to where one lives. Having to relocate is perceived as negative, which makes the job thus less attractive (Grossman & Magnus, 1988).
Although above researches all state that location is an important factor influencing attractiveness of a job, not every researcher agrees on this. Turban and Keon (1993) found no significant results regarding location. Although not all researchers agree on the importance of location regarding making application decisions, in line with the researches described above (Taylor & Bergmann, 1978; Turban, 2001; Barber & Roehling, 1993; Lievens & Highhouse, 2003; Grossmann & Magnus, 1988), it is expected that when the location is close to the place where the job seeker is currently living, the utility of the job offer is influenced positively.
H6: When the location of the job is close to the current living area of the job seeker, then location has a positive relation with perceived job utility.
2.6 Promotion possibilities
Newly graduated job seekers are not very likely to immediately get an executive position in a company, they rather first receive a position in the lower management of a company. From that position, they could grow further into the company. This makes it important for these ambitious people that the company they would like to work for offers promotion possibilities. It is even found that when a company does not offer promotion possibilities, the number of average days to fill a vacancy increases (Isenhour, Lukaszweski, & Stone, 2014). Job seekers are more likely to accept a job when promotion possibilities are offered (Isenhour,
13 are no advancement opportunities, then that would be a deal-breaker (Osborn, 1990).
Opportunities for advancement was generally given top priority in the literature, but business executives could not agree with this priority, as they saw payment as a more important job attribute (Jurgensen, 1978). Recent research confirms that for students who are seeking a job, personal growth and opportunities to get promoted are the most important job attributes (Arachchige & Robertson, 2013).
On the other hand, Chapman et al. (2005) found that advancement opportunities is indeed a predictor of job pursuit intentions, however, to a much lesser extent than most other job attributes. The importance of advancement opportunities is decreasing over the years, while other job characteristics such as type of work are becoming more important (Jurgensen, 1978). Following the recent research of Arachchige and Robertson (2013), it is expected that when there are promotion possibilities, the attractiveness of the job will increase, which in turn will increase the likelihood of applying for that vacancy.
H7: Promotion possibilities will have a positive impact on perceived job utility. 2.7 Job security
In the existing literature, job security is often cited as an important job attribute. Job security, or employment security, means that individuals are convinced that they are enabled to
continue their employment career, with the same employer in another job or with another employer and another job (Dekker, 2010). An example of job security is a formal no-layoff policy (Gramm & Schnell, 2013). The importance of job security, however, varies over the different researches. Noe et al. (2013) state that employment security is perceived by potential employees as one of the primary job attributes that influence the attractiveness of a vacancy. Gramm and Schnell (2013) stated that when organizations offer their (potential) employees a certain job security, they will experience a competitive advantage over firms that do not offer employment security.
14
less important compared to other job attributes such as pay and promotion opportunities. In the contrary to Jurgensen’s (1978) findings, this research expects that job security will
positively influence the attractiveness of a job offer. This is especially the case because we are living in an era where many graduates are searching for a job and they are facing a hard time in finding one. This makes it extra important for job seekers to have a certain level of
certainty that they do not have to face that struggle any time soon after accepting a job offer. H8: Job security will have a positive influence on the perceived utility of a job offer. 2.8 Utility of a job
The basis of conjoint anlaysis is the Random Utility Theory (RUT), which was already introduced by 1927 by Thurstone. It was proposed as a way to understand and model choices between pairs of stimuli (Louviere, Hensher, & Swait, 2001). Nowadays, RUT is a general framework used for modelling and understanding many different kinds of human behavior (Louviere, Hensher, & Swait, 2001). RUT is basically the general paradigm for almost all conjoint analysis techniques. Conjoint analysis is characterized by choice sets which are ordered sets for analytical reasons. The ordering of the elements of such a set is irrelevant to the decision maker that strives for maximum utility (Manski, 1977). According to classical choice theory, individual behavior is the outcome of a two-step process (Manski, 1977). The first step being a choice problem, being an individual and an associated choice set, the second step being the individual selecting among the alternatives in the defined choice set (Manski, 1977).
Such a choice set consists of several different attributes. Goods and services can be seen as combinations of certain attributes. The same goes for a job offer. A job is actually a combination of certain attributes. Above are these attributes described; salary, location, promotion possibilities, brand name, and job security. All these attributes make up the job. Consumers attach utilities to each of the product’s (or service’s) attribute, which means that they attach a certain weight to each product (service) attribute. This happens with job attributes as well. For certain job seekers, salary might receive a higher weight than job
15 2.9 Conceptual model
Brand name
Salary
Location
Promotion
Job security
Perceived job utility
Probability of applying
for the job
16
3. Methodology
In this section, the research methodology will be discussed. First, the choice-based conjoint analysis is introduced, then the levels and attributes will be elaborated upon, the data collection will be explained, and the model will be introduced.
3.1 Choice-Based Conjoint Analysis
Conjoint analysis (CA), as a method for measuring and analyzing consumer preferences for goods and services, has gained in popularity over the past two decades (Akaichi, Nayga, & Gil, 2013). CA is defined as “a decomposition into part-worth utilities or values of a set of individual evaluations of, or discrete choices from, a designed set of multi-attribute
alternatives (Louviere, 1988).It is thus used to measure trade-offs among multi-attributed products and services (Green & Rao, 1971; Johnson, 1974; Green & Srinivasan, 1990). There are two widely-used CA methods, namely choice-based conjoint (CBC) and ranking conjoint analysis. Raking based conjoint analysis has as limitation that consumers can not really quantify their utility and it does not really mimic real life, as consumers do not really rate products in real life. With CBC analysis, respondents receive a choice set of combinations of attributes and they have to indicate which combination they would choose (Akaichi, Nayga, & Gil, 2013). This task is basically the same as what buyers do in the real market place and because CBC mimics this purchasing process, CBC is so popular in preference elicitation research (Louviere, 1988). In this research, the choice is made to use CBC, because this research aims to find out how the preference for certain job attributes changes when a high equity brand is involved. One additional advantage of CBC over other conjoint analysis formats, is that the surveys used in CBC analysis are much shorter compared to surveys used in other conjoint analysis formats (Johnson & Orme, 1996). A shorter survey has as an advantage that the likelihood that respondents will suffer from fatigue effects is diminished and with this the chance that respondents will answer the questions randomly will decrease. 3.2 Attributes and Levels
In CBC analysis, products and services are seen as attribute bundles. A job offer in this case is existing of attributes as well. The sum of the utilities of the attribute levels is the utility of the job offer. Finding relevant attributes is the first step in developing a CBC study (Orme, 2005) and this could be done in either a structured- or unstructured way. In this research, the
17
Attribute Level 1 Level 2 Level 3 Level 4
Company KLM Arke Transavia Corendon Dutch
Airlines
Salary €2.100 €2.300 €2.500 €2.700
City Groningen Utrecht Amsterdam Eindhoven
Job security 6 month contract 1 year contract 1.5 year contract Permanent contract
Promotion After 1 year After 2 years After 3 years After 4 years
Table 1: Attributes and attribute levels
The four aviation brands are chosen based on the DCPI measurements (2012). These measurements demonstrate that KLM is delivering the best customer performance in the aviation industry, followed by Arke, and Transavia can be found on the third place. According to this research, Corendon Dutch Airlines is not among the 80 best performing Dutch
companies, so it can be assumed that the customer performance of Corendon Dutch Airlines is at a lower level than the former three brands. In order to avoid that respondents will base their job choice based on the industry included in the job offer, only one industry is included in this research.
Being a management graduate will give you a starter’s salary of approximately €2.300 per month and being a business graduate will give you a somewhat higher starter’s salary, namely €2.500 (Gemiddeld Inkomen, 2015). These levels are chosen for the attribute salary, together with one higher- and one lower amount of salary.
According to a research conducted by Intermediair (2013), Utrecht, Amsterdam, Groningen, and Eindhoven are among the best cities to work in. These four cities are chosen because they are spread throughout The Netherlands and thus represent the different
provinces quite well.
The number of hours worked under temporary contracts has increased significantly last year (CBS, 2014). It is shown that employers are increasingly offering temporary
contracts instead of permanent contracts (CBS, 2014). Due to this, there is chosen to include three temporary contract types and one permanent contract type in this research.
18 starting point.
As can be seen from table 1, all the attributes have the same amount of levels. This is done in order to avoid the so-called “number of levels effect”. In this way, respondents will not perceive an attribute with e.g. five levels as more important than an attribute with four levels. The attributes are independent, which is required in order to combine the levels freely with one another. Moreover, the attribute levels are all mutually exclusive, which means that they can not occur at the same time.
3.3 Choice Experiment
Since this research uses five different attributes with each four levels, it would mean that there are 1.024 possible job offers. No respondent would like to review 1.024 job offers, so a full factorial design is not an option. Hence, a fractional factorial design is used, which uses only a subset of the full factorial design.
An efficient choice based conjoint design is balanced and orthogonal, which means that each attribute level is shown approximately an equal number of times and that attribute levels are chosen independently of other attribute levels. In this way, an attribute level’s utility is measured independently of other effects (Sawtooth Software, 2013). With the
programme Preferencelab this is achieved by randomly selecting the stimuli. Furthermore, the design uses minimal overlap, which means that each attribute level is shown as few times possible in a single choice set.
The design consists of three alternatives per choice set. There is no no-choice option included, even though a design with a no-choice option shows increased efficiency (Brazell et al., 2006). The reason why not including a no-choice option, is because when the no-choice option is selected, no information is retrieved regarding the attractiveness of the available alternatives. Since no no-choice option is included, this design is a so called forced-choice task, where the respondents have to choose an option among a set of available alternatives. This study design matches real life, where students graduate and need to apply for a job and where they do not have a possibility to not apply for any vacancy either.
19 Figure 1: Choice set example in the branded scenario
Figure 2: Choice set example in the anonymous scenario 3.4 Sampling and Procedure
The survey is constructed within the programme Preferencelab. The target population for this research are Dutch master students, as they are in the life phase where they have to think about the kind of job that they would like to pursue. In order to reach this target population, convenience sampling will be used; the survey link will be provided on social media and there is aimed for snowball sampling, in which the link will be forwarded by respondents that already participated in the survey. At the beginning of the survey the participants are told that they should choose the most attractive job offer, so the main purpose of the research is not revealed, this is done in order to prevent any biased results. The survey starts with several questions about the respondent’s demographics, followed by 10 branded choice sets, followed by seven questions about the four included brands, followed by 10 anonymous choice sets.
3.5 Measurement Level
20
consumer brands. Moreover, the overall brand equity level can already be derived from the analysis by using two different scenarios and brand name as a moderator. The following questions about brand equity are included in the research: a) Some characteristics of this brand come to my mind quickly, b) I can recognize this brand quickly among other competing brands, c) I am familiar with this brand, d) This brand has a very unique brand image
compared to competing brands, e) I like the brand image of this brand, f) I like and trust this company, and g) I respect and admire people who use this brand. All these questions had to be answered using a five point Likert scale (1= totally disagree, 5= totally agree).
3.6 Random Utility Model
Consumers, and also job seekers, make their choices based on overall utilities of alternatives. Goods, services, and also job offers are combinations of attributes and people attach part-worth utilities to each attribute. In this study, the random utility model is used. The utility of job applicant n a job offer i is as follows:
Uni = Vni + εni
Where:
V = systematic utility component, rational utility ε = stochastic utility component, error term
The systematic utility is actually the sum of the part-worth utilities:
Vni = ∑𝑘=1𝐾 𝛽nk xik Where:
k = (1, …, K) number of attributes
x = dummy indicating the specific attribute level of product i β = part-worth utility of consumer n for attribute k
The systematic utility for a job offer in this research would be as follows:
21
As this research also includes two moderators, who can not be tested within the same model due to multi-collinearity issues, two other formulas including the interaction effects are also displayed:
Vi = 𝛽brand BRAND + 𝛽brand-equity BRAND-EQUITY + 𝛽location LOCATION + 𝛽brand*location BRAND*LOCATION + 𝛽promotion PROMOTION + 𝛽brand*promotion BRAND*PROMOTION + 𝛽security SECURITY + 𝛽brand*security BRAND*SECURITY + 𝛽salary + SALARY + 𝛽brand*salary BRAND*SALARY
And the formula where brand equity is included as a moderator:
Vi = 𝛽brand BRAND + 𝛽brand-equity BRAND-EQUITY + 𝛽location LOCATION + 𝛽brand-equity*location BRAND-EQUITY*LOCATION + 𝛽promotion PROMOTION + 𝛽brand-equity*promotion BRAND-EQUITY*PROMOTION + 𝛽security SECURITY + 𝛽brand-equity*security
BRAND-EQUITY*SECURITY + 𝛽salary + SALARY + 𝛽brand-equity*salary BRAND-EQUITY*SALARY The dependent variable here is the chosen job offer and it is assumed that job applicants choose the alternative with the highest utility.
By using different company brands, who differ in their brand equity levels, it is possible to determine the probability of choosing a job offer from a high equity brand instead of choosing that same job offer from a low equity employer brand. This probability can be derived from using the formula 1:
Formula 1: N = 𝐸𝑋𝑃(𝑉ℎ𝑖𝑔ℎ 𝑒𝑞𝑢𝑖𝑡𝑦 𝑏𝑟𝑎𝑛𝑑)+𝐸𝑋𝑃(𝑉𝑙𝑜𝑤 𝑒𝑞𝑢𝑖𝑡𝑦 𝑏𝑟𝑎𝑛𝑑)𝐸𝑋𝑃(𝑉ℎ𝑖𝑔ℎ 𝑒𝑞𝑢𝑖𝑡𝑦 𝑏𝑟𝑎𝑛𝑑)
Where,
𝑁 = probability of choosing a job offer from a high employer equity brand over the same offer from a low equity employer brand
22 3.7 Model Specification
There are three different model specifications; vector specification (linear), ideal points specification (quadratic), and part-worth model (nominal). An attribute can be thus have one of the three formats. The format does not have to be the same for all the attributes. A linear model estimates only one parameter, which is then multiplied with the level of that specific attribute. An advantage of this model is that it allows for interpolation and some
extrapolation. A part-worth model estimates a parameter for every attribute level. Nominal attributes are treated as part-worth, whereas numeric attributes are treated as linear. In this research, the attributes brand, job security, and location are treated as part-worth. The numeric variables promotion possibilities, and salary will be firstly treated as part-worth, but when, after plotting the parameters, it turns out that they can be treated as numeric, then the parameters will be re-estimated.
3.8 Model Fit
In order to assess the model fit, it is important to consider the information criteria, namely the values of AIC, AIC3, BIC, and CAIC. These log-likelihood-based measure penalize for the number of parameters. The model has to be chosen where these information criteria have the lowest values. In this research, the decision regarding which model to work with will be based on BIC and CAIC as these measures are generally preferred.
4. Results
In this section, the characteristics of the sample will first be discussed, followed by the conjoint analysis for both the branded and the anonymous scenario.
4.1 Sample Characteristics
This research is conducted among Dutch students who are currently following a master programme. A total of 63 students participated in this research. The average age of the respondents is 24 years old, 58,7 per cent of the respondents is male and 41,3 of the
23 4.2 Reliability Analysis
In this research, the concept brand equity consists of several questions. In order to find out whether these different questions really do measure the same construct, the Cronbach’s Alpha will be used. According to Nunnally (1978), there is internal consistency when the
Cronbach’s Alpha has a value which is higher than 0.7. Additionally, a factor analysis will be performed in order to identify underlying constructs. When the Bartlett’s test of sphericity is significant and the Kaiser-Meyer-Olkin has a higher value than 0.5, then the factor analysis can be performed. If the factor analysis demonstrates that there indeed is an underlying construct, then the separate questions can be combined in one single variable.
In order to derive to one brand equity measure for each of the four included brands, a reliability analysis is performed. This showed that the Cronbach´s Alpha for all the concepts representing the brand equity for the four brands is high enough (KLM = 0,883, Arke = 0,925, Transavia = 0,922 and Corendon Dutch Airlines = 0,910). All these values are above the threshold of 0,7 (Nunnally, 1978). In order to find common variance and retrieve underlying dimensions, a factor analysis is performed. Before a factor analysis can actually be performed, one should check whether the KMO measure of sampling adequacy and Bartlett’s test of sphericity have the required values. KMO predicts based on correlation and partial correlation if data are likely to factor well. This measure should have a value of above 0,5. The Bartlett’s test of sphericity checks whether the variables are correlated, H0 being: variables are
uncorrelated, H1 being: variables are correlated. The values of the former two measures that correspond with the brand equity questions for each of the four respective brands can be found in table 2.
KMO Bartlett’s test
KLM questions 0,807 0,000 < p0,01
Arke questions 0,867 0,000 < p0,01
Transavia questions 0,867 0,000 < p0,01
Corendon questions 0,895 0,000 < p0,01
Table 2: Factor analysis statistics
24 4.3 Aggregate choice model anonymous scenario
In this section, an aggregate choice based conjoint (CBC) model for the anonymous scenario will be estimated. In a later section, the CBC model for the branded scenario will be
estimated, followed by a model for the combined scenarios. In that way it is possible to draw conclusions whether or not job seekers respond differently to certain job offers when an employer brand is included.
Firstly, the values of the attribute promotion possibilities are recoded. As a promotion after four years is the worst scenario, while a promotion after one year would be the best scenario. In order to have all the variables measured in the same direction, where the highest value means something good, this variable is recoded. Moreover, a dummy variable is created by comparing whether the respondent lives in the same area as where the job is located. The value of 0 represents that the location where the respondent is living is different than the location in the chosen job offer, and 1 represents that this location is the same. Then, all the main effects are included in Latent Gold and they are all set as nominal. In this way, a separate parameter will be estimated for all the different levels of a certain variable. The results of this can be found in table 3.
Attributes Wald p-value
Salary 92,8383 5,4E-20
City 40,0279 1,10E-08
Job security 89,576 2,70E-19
Job promotion 93,3522 4,20E-20
Table 3: Nominal variable statistics.
This table demonstrates that all the main effects are significant, as all p values are <0,01. In order to check whether some of these nominal variables could be treated as numeric, the below graphs are plotted. This demonstrates that the variable salary and job promotion can be treated as numeric, which is beneficial since there is then the opportunity to interpolate and extrapolate. To formally test whether this if treating these two variables as numeric yields to a better model than were these two attributes are treated as nominal a chi-square test is
25
effect regarding these two variables will be calculated with the numeric value from promotion and salary. For the other moderation effects, the interaction is still calculated with the effect scores of the respective variables.
Figure 3: Linear representation of promotion Figure 4: Linear representation of salary
In order to select the best model, several models are estimated. This is done for both a main effects model in order to estimate the parameters, and for an interaction model in order to test the hypothesized moderation effects between the variables. The results of these different estimations can be found in table 4 and 5. The estimated models will be compared using the adjusted R2, the hit rate, BIC, and CAIC. One should compare models with the adjusted R2 instead of the regular R2, as the value of the latter one always increases when more parameters are added. All the estimated models have the same loglikelihood null model, namely
-692,126 (LL(0) 63 consumers * 10 choice sets * ln(1/3 alternatives)). The hit rate reveals how many of the observed choices are predicted correctly with the selected model. Model 2 has a hit rate of 59,05 per cent, which means that 59,05 per cent of the observed choices is predicted correctly. When looking at the adjusted R2 of the main effect models, the one where salary and promotion are numeric, performs better than the first model, as there the value is higher than in model 1. When considering the BIC and CAIC score, one should chose the model where these values are the lowest. Model 2 has the lowest scores on these two information criteria. Log-likelihood LL (β*) R2 R2adj Chi-square Nr of parameters Hit rate BIC CAIC Model 1a -560,734 0,1898 0,1754 262,7828 10 58,41% 1162,9000 1172,9000 Model 2b -563,593 0,1857 0,1770 257,0664 6 59,05% 1152,0440 1158,0440
Table 4: Information criteria main effects models.
a: All variables (salary, promotion, location, job security) are part-worth b: Salary and promotion are numeric, rest part-worth
26 The parameters of the best performing model, model 2, can be found in table 5.
Table 5: Parameters of main effects model and interaction effects model. ***α<0,001
4.4 Attribute Importance
In order to define which of the job attributes are most important for the job applicant, the parameters from model 2 are used in order to calculate the relative importance. These can be found in table 6. This importance is determined by calculating the range between the highest vale and the lowest value of each attribute.
Attributes Range Importance
Salary 1,08 26,42%
Location 0,6626 16,21%
Job promotion 1,0362 25,35%
Job security 1,3093 32,03%
Table 6: Relative importance levels.
As can be seen from table 6, job security is the most important attribute; the longer the period of the contract, the higher the job utility will be. Regarding salary, which is the second most important job attribute; the higher the salary, the higher the utility attached to that job offer. Closely following up salary; job promotion is the third most important attribute; when a job will offer you soon new promotion possibilities, then the utility of the job increases. Lastly, whether the job is located in the same area as where the job applicant is living is the least important job attribute. Apparently, job seekers do not consider moving to another city as a problem, as they attach the least importance to this attribute.
4.5 Aggregate choice model branded scenario
The model of the branded scenario will be estimated in the same way as was done with the anonymous scenario. Since the preference function of salary and promotion seem to be numeric as is shown in figure 5 and 6, it is tested with the chi-square whether a model with
Attribute Main effects model
Salary
Same location Other location
Promotion possibilities Job security 6 months Job security 1 year Job security 1.5 year Job security permanent
27
numeric values for these attributes performs better compared to a nominal model. It turns out that with four degrees of freedom, χ2 should be above 13,277. This means that the test is not significant, as p(3,26)>0,01 and that the models perform equally well, which means that the redundant parameters can be eliminated (Werner & Engel, 2010). Hence, from now on, salary and promotion will be treated as numeric. Brand equity is a numeric covariate as no fixed levels of brand equity were manipulated in the experiment. To demonstrate visually that these attributes indeed do show a linear relationship, they are plotted in figures 5 and 6.
Figure 5: Linear representation of promotion Figure 6: Linear representation of salary Since salary and promotion will be treated as numeric variables, the interaction effects with
these variables are calculated using their numeric value. The remaining attributes will still be nominal.In order to select the best performing model, several models are estimated. This is done separately for the main effect model as well as for the interaction model. The results of these estimated models can be found in table 7 and table 8.
Log-likelihood LL (β*) R2 R2 adj Chi-square Nr of parameters Hit rate BIC CAIC Model 1a -524,999 0,2282 0,1689 315,8442 14 61,90% 1126,4110 1140,4110 Model 2b -535,834 0,2258 0,2114 312,5842 10 61,90% 1113,0990 1123,0990 Model 3c -536,494 0,2249 0,2119 311,2638 9 60,79% 1110,2760 1119,2760 Model 4d -539,774 0,2201 0,2100 304,7042 7 61,11% 1108,5491 1115,5491
Table 7: Information criteria main effects models. a: All variables, except brand equity are part-worth.
b: Salary, brand equity, and promotion numeric, rest part-worth. c. Salary and promotion numeric, rest part-worth. Brand equity excluded.
d. Salary, brand equity, promotion numeric, rest part-worth. Company name excluded.
Log-likelihood LL (β*) R2 R2 adj Chi-square Nr of parameters Hit rate BIC CAIC Model 5a -535,242 0,2267 0,2050 313,7674 15 61,74% 1132,6310 1147,6310 Table 8: Information criteria interaction effects models.
a: Brand equity moderator included for all attributes.
28
Regarding the main effect models; model 3 has the highest adjusted R2, but model 4 has the lowest BIC and CAIC values. However, in model 2, where both company name and brand equity are included, company name is significant (p(0,049<0,05), whereas brand equity is not significant (p(0,25>0,05). It seems managerially more relevant to include the attribute
company name, than excluding that attribute and including the brand equity attribute, since a company name can give information about more aspects than just about brand equity. In order to check whether a model with company name included but brand equity excluded is not performing worse than the model where both job attributes are included, a chi-square test is performed. With one degree of freedom, χ2 should be above 6,635. This means that the test is not significant, as p(1,32)>0,01 and that the models perform equally well, which means that the redundant parameters can be eliminated (Werner & Engel, 2010). So, a model where brand equity is not included performs equally well as a model where that attribute is included. It is thus decided to exclude this variable from the model. Together with the fact that the adjusted R2 is the highest at model 3, decided is to continue working with a model where salary and promotion are numeric and company, location, and job security are nominal, and where brand equity is not included.
29
Attribute Main effects model Interaction effects model
KLM Arke Transavia Corendon Salary Brand Equity*Salary Same location Other location
Brand Equity*Same location Brand Equity*Other location Promotion possibilities
Brand Equity*Promotion Job security 6 months Job security 1 year Job security 1,5 year Job security permanent
Brand Equity*6 months contract Brand Equity*1 year contract Brand Equity*1,5 year contract Brand Equity*permanent contract
0,3724*** -0,1827 -0,0220 -0,1677 0,0022*** x 0,2189*** -0,2189 x x 0,4436*** x -0,7495*** -0,0103 0,0606 0,6991 x x x x 0,2799* -0,1553 -0,0220 -0,1025 0,0019*** 0,0001 0,2564*** -0,2564 -0,0114 0,0114 (reference) 0,0015** -0,0270 -0,8921** -0,1401 0,1234 0,9089 0,0385 0,0380 -0,0178 -0,0587 (reference)
Table 9: Parameters of main effects model and interaction effects model. ***α<0,001 **α<0,05 *α<0,1
4.6 Attribute Importance
Also for the branded scenario, the relative importance levels are calculated. The outcomes can be found in table 10.
Attributes Range Importance
Company 0,5551 10,90%
Salary 1,32 25,92%
Location 0,4378 8,59%
Job promotion 1,3308 26,13%
Job security 1,4486 28,45%
Table 10: Relative importance
30
period of time, then the utility of the job increases. Salary is the third most important job attribute; the higher the salary, the higher the utility attached to that job offer. The fourth most important job attribute is company name, although this attribute receives a lot less attention compared to the other job attributes. Lastly, whether the job is located in the same area as where the job applicant is living is the least important job attribute.
4.7 Comparison between anonymous and branded scenario
In order to compare the two different scenarios, the two estimated main effects models are represented in table 11.In both models, the main effects all significantly influence job utility (p<0,01). The first notable aspect is the parameter of location. The effect that same location has on job utility is larger in the anonymous scenario than in the branded scenario.
Apparently, when the company is unknown, job applicants are not very willing to move to another area than where they are living now. The other parameter values are quite similar regardless of which scenario the vacancy is displayed in. In order to test whether brand name has a real, significant differential effect on respondents’ reaction towards the job attributes, brand name will be included as a moderator in the combined scenario in the next section.
Attribute Anonymous main effects
model
Branded main effects model KLM Arke Transavia Corendon Salary Same location Other location Promotion possibilities Job security 6 months Job security 1 year Job security 1.5 year Job security permanent
x x x x 0,0018*** 0,3313*** -0,3313 0,3454*** -0,6225*** -0,1469 0,0826 0,6868 0,3724*** -0,1827 -0,0220 -0,1677 0,0022*** 0,2189*** -0,2189 0,4436*** -0,7495*** -0,0103 0,0606 0,6991
Table 11: Parameters of anonymous scenario and branded scenario. ***α<0,001 **α<0,05 * α<0,1
31
determining the attractiveness of a job. Moreover, a notable difference is the low percentage score of location in the branded scenario compared to the anonymous scenario where the percentages are closer to one another. This seems to suggest that when someone does not know for which company to work for, he or she is less likely to move to another area for that job.
Attributes Importance branded Importance anonymous
Company 10,90% x
Salary 25,92% 26,42%
Location 8,59% 16,21%
Job promotion 26,13% 25,35%
Job security 28,45% 32,03%
Table 12: Relative importance
4.8 Aggregate choice model combined scenarios
Here, a main effect model as well as interaction models will be estimated in which the two scenarios are combined. In this model salary and promotion will be treated as numeric, as the chi-square test turned out not to be significant, as the χ2 value for four degrees of freedom should be above 13,277 and this is not the case since p(2,80)>0,01. So, the model where salary and promotion are numeric performs equally well as the model in which they are nominal, which means that the redundant parameters can be eliminated (Werner & Engel, 2010). Hence, from now on the two variables will be treated as numeric. A visual
representation can be found in figures 7 and 8.
Figure 7: Linear representation promotion Figure 8: Linear representation salary
From the information criteria can be derived which of the main effect models and which of the interaction effect models is performing the best. The results of these estimated models can be found in table 13 and 14.
32 Log-likelihood LL (β*) R2 R2 adj Chi-square Nr of parameters Hit rate BIC CAIC Model 1a -1101,8953 0,2103 0,1939 564,7124 14 60,16% 2261,7944 2275,7944 Model 2b -1103,2934 0,2030 0,1957 561,9162 10 59,68% 2248,0180 2258,0180 Model 3c -1103,9098 0,2025 0,1960 560,6834 9 59,37% 2245,1078 2254,1078 Model 4d -1107,0483 0,2003 0,1952 554,4064 7 59,21% 2243,0985 2250,0985
Table 13: Information criteria main effects model. a: All variables except brand equity are part-worth.
b: Salary, brand equity, and promotion numeric, rest part-worth. c. Salary and promotion numeric, rest part-worth. Brand equity excluded.
d. Salary, brand equity, and promotion numeric, rest part-worth. Company name excluded.
Log-likelihood LL (β*) R2 R2adj Chi-square Nr of parameters Hit rate BIC CAIC Model 5a -1100,09 0,2053 0,1923 568,3324 15 59,92% 2262,3176 2277,3176 Model 6b -1100,14 0,2052 0,1922 568,2236 15 60,16% 2262,4265 2277,4265 Model 7c -1101,50 0,2043 0,1963 565,4992 11 60,08% 2248,5783 2259,5783 Model 8d -1102,36 0,2036 0,1957 563,7912 11 59,44% 2246,1431 2256,1431
Table 14: Information criteria interaction effects model. a: Brand name moderator included for all attributes. b. Brand equity moderator included for all attributes. c. Only significant brand moderators included. d. Only significant brand equity moderators included.
In order to decide which main effect model is performing best, four models are estimated of which the information criteria can be found in table 13. In model 2, where both company name and brand equity are included, brand equity is highly insignificant (p(0,27)>0,05), thus a model is estimated without brand equity. That model, model 3, has the highest adjusted R2 value, which would suggest that that is the best performing model. Only model 4, where brand equity is included but company name is excluded, has lower BIC and CAIC values compared to model 3. However, the difference between these values is rather small, and since the adjusted R2 value is the highest at model 3, it is decided that that is the best performing model. The parameters of this model can be found in table 15.
33
neither brand name, nor brand equity is significantly moderating any of the relationships between any job attribute and perceived job utility. When the moderating effects are estimated separately, e.g. first the moderating effect of brand name on the relation between salary and utility, then the effect on the relation between promotion and utility, it turns out that both brand name and brand equity are significantly moderating the effect between location and perceived job utility (brandname*location=p<0,1, brandequity*location=p<0,1). The precise parameters of these models can be found in respectively appendix 2 and appendix 3. The parameters of the interaction effect are in both cases negative (-0,1275 and -0,0349 respectively) which means that when a brand name is included in the job offer, the positive effect that same location has on job utility becomes less strong. Regarding brand equity: when the level of brand equity increases, the positive effect that same location has on perceived job utility, becomes less strong. However, since these effects are not significant in the full model, not much attention should be paid to these results.
Attribute Main effects model Brand name interaction
effect model KLM Arke Transavia Corendon Salary Same location Other location Promotion possibilities Job security 6 months Job security 1 year Job security 1,5 year Job security permanent Brand name*Salary
Brand name*Same location Brand name*Other location Brand name*Promotion Brand name*6 months Brand name*1 year Brand name*1,5 year Brand name*permanent 0,3619*** -0,1688 -0,0283 -0,1647 0,0020*** 0,2760*** -0,2760 0,3950*** -0,6890*** -0,0801 0,0753 0,6938 0,3729*** -0,1830 -0,0220 -0,1679 0,0018*** 0,3316*** -0,3316 0,3458*** -0,6232*** -0,1470 0,0827 0,6875 0,0004 -0,1124 0,1124 (reference) 0,0983 -0,1273 0,1368 -0,0220 0,0125 (reference) Table 15: Parameters main effects model and interaction effects model.
34
Since the moderating effects are effect coded, the reference level needs to be recovered, since it is not given. This is achieved by taking the sum of the three levels that are given (e.g. for job security: -0,1273 + 0,1368 + -0,0220) multiplied by -1.
4.9 Attribute Importance
Table 16 demonstrate the relative importance for the combined model and is based on the main effects model. As one can see, it is not very different from the previous relative
importance levels. Job security again receives the greatest importance when reviewing several job offers, salary is placed second and is closely followed by job promotion. The latter two almost have equal importance levels. Company name has the least importance. It obviously is not very important for job seekers for which company they will work, as long as other
attributes are satisfactory.
Table 16: Relative importance 4.10 Willingness to pay
Since the salary attribute is linear, it is possible to calculate the willingness-to-pay for the different job offers. This can be done by dividing the parameters of the different levels by the salary parameter. The below tables demonstrate either positive or negative values. In this case, a job applicant that would accept a job offer at Corendon, wants a higher salary as some sort of compensation for working for a low equity brand. A higher salary is also required when the job is offered in a different area than where the job seeker is currently living. On the other hand, when the job would be permanent, someone is willing to accept a lower salary. This is also implied for working for KLM; one is willing to forgo of €180,95 monthly salary in order to work for KLM.
Table 17: WTP location
Attributes Range Importance
35
Table 18: WTP company Table 19: WTP job security
It would be interesting to find out what the likelihood is of accepting a job offer of the high equity brand compared to the job offer with the same job attribute levels except for the fact that it is from a low equity brand. According to the survey questions about the four brands, KLM is the highest equity brand and Corendon is the lowest equity brand. The utility function for KLM would thus be: 0,3619 + 0,0020*2.700 + 0,2760 + 0,3950*4 +0,6938= 8,3117. Using the same attribute levels but using the parameter of the low equity brand will result in the following utility: -0,1647 + 0,0020*2.700 + 0,2760 + 0,3950*4 +0,6938= 7.7851. In order to find the probability of choosing the same job offer from a high equity brand instead of from a low equity brand, formula 1, depicted in the methods section, is used.
The probability of choosing the same job offer when it is from a high equity brand versus a low equity brand is 0,6287, or 62,87 per cent. So looking at it from another angle, the probability that a job seeker will apply for a job offer from a low equity brand compared to that same offer from a high equity brand is only 0,3713, or 37,13 per cent.
4.11 Latent Class Analysis
A latent class analysis is performed in order to distribute the respondents into different segments. This is done on the basis of attributes that differentiate the segments from one another. In Latent Gold, different models consisting of two to five classes are estimated. It is firstly checked whether certain variables need to be set to class-independent. This has to be done when certain attributes do nog significantly differ across segments. It is found that company is not significant as p>0,05, and thus this variable is set to class-independent. The other variables (salary, location, promotion, job security) are significant (p<0,05), so the segments are significantly different from each other based on these variables. Covariates were also entered in Latent Gold to assign people to segments in a better way, but the variables gender, age, and living area were not significant and where thus not significantly different among the segments. It was therefore chosen not to include these variables into the
36
distribution to the segments. When looking at the information criteria of these models in table 20, the model with the best values will be chosen.
Table 20: Information criteria latent class analysis
When looking at the BIC value, the lowest value is represented at model 4, where there are five classes. This is in correspondence with the CAIC and BIC value, which is also the lowest at model 4. Hence, a model with five latent classes is preferred. In figure 9, the values of BIC and CAIC are plotted. This demonstrates that the lowest point of both BIC and CAIC is at when the individuals are distributed into five segments. When six segments are included, the BIC and CAIC values increase again. So this is in line with the previous conclusion; that a four-class model is the preferred option.
Figure 9: Visual representation of BIC and CAIC values
However, from a managerial perspective it might be a hassle to take into account five different segments. Therefore, the Wald statistics and the p values are compared between a four-and five segments model. It turns out that with a four class model, all the attributes still significantly differ across the segments. Also when looking at the size of the classes, it turns out that when there are five different classes, the fifth class is very small. When four classes are estimated, the distribution among the segments is more equal. It can thus be concluded that working with a four class model is the best solution in this case.
1800 1850 1900 1950 2000 2050 2100 2150
2 class 3 class 4 class 5 class 6 class
BIC CAIC
LL BIC(LL) CAIC(LL) Npar df p-value Class.Err. R²
37
Class 1 Class 2 Class 3 Class 4 Class 5
4 class model 30% 29% 26% 14% x
5 class model 27% 26% 22% 14% 9%
Table 21: Relative sizes of the segments
In table 22 the relative importance of the attributes can be found per class. This demonstrates which segments values which attribute the most. Below, a description of the segments will be provided.
Attribute Class 1 Class 2 Class 3 Class 4
Company 6,35% 18,62% 11,20% 5,51%
Salary 29,83% 32,59% 17,55% 16,74%
Location 1,27% 9,19% 38,45% 0,56%
Job security 31,31% 17,11% 7,09% 67,20%
Promotion 31,24% 22,48% 25,71% 10,00% Table 22: Relative importance
Personal achievers- class 1: this segment clearly places the most value on the three typical job attributes. They value a fast promotion, a long contract, and a high salary. They do not really care where they will receive these benefits, in which location or at which company, as long as the conditions are attractive.
Money seekers- class 2: These people do place the highest value on salary. They want to know that they will earn a good salary when applying for a job. Moreover, this is the only segment that places quite a high value on the company they would like to work for. And in this case, the company that has the highest influence on utility is KLM. These people apparently want to earn a high salary at a high equity company.
Stay at home’ers- class 3: for this segment, location is playing the most important role when comparing the importance of location. They are absolutely reluctant to moving to another area for their work. They mainly select job offers based on whether the job is offered in their living area or not.
Security seekers- class 4: for this segment, it is important that their potential job offers a certain level of security. They do not like to start a job where it is uncertain how long they can stay. Salary is also important for them, but it is secondary compared to job security. The importance of the other job attributes are is negligible.
4.12 Hypotheses testing
Table 23 demonstrates for the aggregate combined model which of the hypotheses are
