Finding the good in goodbye

(1)

Finding the good in goodbye

determining validity, form and content of exit interviews and surveys

Master thesis, specialization Human Resource Management

University of Groningen, Faculty of Economics and Business

Preliminary Draft

May, 2010

MICHEL A.H. HAMS Student number: 1472879 Paterswoldseweg 69A 9727 BA Groningen Telephone number: +31 624962091 E-mail: a.h.hams.1@student.rug.nl Supervisor/ university: Drs. J. van Polen Second assessor: Drs. A.J.E. Schilder Supervisor/ field of study:

(2)

Exit interviews and surveys (EIS) are used by companies for several purposes; its main purpose is to provide data about the causes and reasons behind turnover. Ultimately this data could be used to decrease non-desirable turnover. Previous research results on the validity of EIS have been somewhat contradicting. Theory shows there are several flaws which can possibly decrease validity of EIS and research results. The biggest issue seems to be employees having incentives that prevent them from being completely honest. The content of which an EIS should consist of in order to reach its goals are determined by the use of turnover related research and interviews with the end users of EIS data (e.g. HR-Managers). In addition several methods of conducting exit interviews are analyzed. The selected content led to the creation of a new exit survey which is largely based on measuring aspects that have been shown to be linked to turnover. Questions on job satisfaction and information on the employees‟ future employer are likely the most informative measures. Furthermore the use of a computerized exit survey is argued to be the most beneficial method for conducting EIS, because of increased validity of data and for practical and economic reasons. Recommendations are made for future research to gain empirically based answers on the effectiveness and usefulness of EIS.

Keywords

Exit Interview; Exit Survey; Voluntary Turnover; Validity;

(5)

1. INTRODUCTION

The aging workforce in European nations and the existing deficit of employees in vital sectors

make the attraction of skilled employees a challenging human resource problem and a critical

issue for decades to come (Tanova & Holtom, 2008). In a tight labour market, it is also logical

to also give as much attention to retaining employees as toattracting employees. One of the

ways organizations can deal with the problem of a deficit of employees is by reducing the

amount of turnover. Turnover is the withdrawal of an employee from an organization

(Feinberg & Jeppeson, 2000). Although turnover is considered to have negative effects on

organizational performance, organizations always have to deal with a certain amount of

turnover. Turnover it is reckoned to be an inevitable phenomenon in any organization

(Kirschenbaum & Weisberg, 2002).

Managers in, for example, industrial organizations have always known that employee

turnover can be very expensive (Hinrichs, 1975). Organizations lose the investments in

training, the experience and the knowledge of the departing employee. In addition, there are

the costs of selecting, recruiting and training new personnel. These are only the tangible costs

of employee turnover; there are also intangible costs, e.g., a decrease in moral of the

employees who stay at the company or the effects on social capital (Morrell, Loan-Clarke, &

Wilkinson, 2004). Organizations benefit by instruments that can help decrease turnover

because it prevents them from spending a substantial amount of time, effort and money on the

negative effects of turnover.

(6)

gather information from employees regarding the impressions and experiences they have had

with the organizations they are leaving (Giacalone & Duhon, 1991). The EIS is conducted

during the last working days of an employee that is voluntarily departing his or her

organization. EIS methodology is designed for voluntary separations, because of issues that

are raised by layoffs and terminations. Non-voluntary separations will require a different

approach (Frase-Blunt, 2004). The exit interview is conducted in the form of a face-to-face

conversation with a member of the organization (usually either the person‟s direct supervisor or a member of the human resource department) or an outside consultant hired by the

organization. An exit survey can take the form of a paper and pencil survey or an online

survey. The exit interview or exit survey can also be completed over telephone. Combinations

of different forms and different methods to gather information are possible. The EIS is meant

to generate information about the reasons why employees leave their current organization.

This information could prove to be vital in retaining the employees that are talented, because

the employees that can afford to leave the organization are thought to posses the most

valuable skills and abilities (Tanova & Holtom, 2008). The ultimate goal of the EIS process is

to use the generated information so working conditions can be changed in such a way that

employees stay motivated and remain loyal to their organization. In short: the EIS can, if the

data that was collected is valid and reliable, give information about the causes and reasons

behind turnover and ultimately help to decrease turnover.

When organizations first started conducting exit interview and exit surveys, it was not long

before researchers and practitioners realized there could be problems with the validity of data

collected by EIS. In early research on the subject McMurry (1943) already stated: “When an

(7)

with other methods used by organizations to generate information, the usefulness of this data

is partly determined by its validity. “All data collection and analysis is carried out in the

context of a model or set of assumptions about the process being observed” (Russ & Hoover, 2005). If these assumptions are wrong, the information that is collected by the EIS will be

meaningless.

Researchers have indeed found problems with the validity and reliability because of biased

responses in EIS (Lefkowitz & Katz, 1969; Hinrichs, 1975; Giacalone & Duhon, 1991). Part

of this problem might be the fact that in academic literature there is no consensus on what the

nature of the EIS data, that is collected, should consist of (Giacalone, Elig, Ginexi, & Bright,

1995). Although attempts have been made to standardize the EIS (Goodale, 1982; Hilb, 1978;

in Knouse, Beard, Pollard, & Giacalone, 1996), the questions surrounding its validity and reliability have never fully been answered. There have been studies that focused on the

willingness to report on certain issues during the EIS (Giacole, Knouse & Montagliani 1997;

Knouse et al., 1996). Although measuring willingness provides some sense about the degree

of validity, it is evidently not the same as an actual report. Earlier literature (Lefkowitz &

Katz, 1969) investigated the validity of exit interviews by conducting a follow up procedure

that measured the reasons for turnover for the second time, a few months after the exit

interview. After this was carried out, the reasons for departure from the follow up procedure

were compared to the reasons for departure given on the interview. In many of the cases the

reasons for departure given on follow up questionnaire, differentiated from the reasons given

on the exit interview. More precisely, the follow up data differed significantly from the initial

data collected. This was demonstrated by the reported rank order correlation coefficients of

(8)

invalid data.The question whether these results were found because of the difference between

methods of administration or because of the difference in time of administration remained

uncertain. It could also be a matter of unreliability because of information falsification by

employees or the use of unstructured interviews (Giacole & Duhon, 2009). Furthermore the

literature on EIS shows that a lack of success with exit interviewing has been largely the

result of ineffective data monitoring and misunderstanding of the end users‟(e.g. management,

HR departments, company policy makers) needs (Giacalone, 1999). If the data collected in

the EIS is not useable for end users, the information is bound to be neglected.

To date, an EIS instrument that takes into account research findings on EIS, as well as

provides with a sound policy for conducting EIS, has not yet been developed. Neither is there

a concise theoretical foundation for the content that EIS should consist of. Most organizations

develop their own instrument, because academic literature is not clear about what data should

be collected in EIS (Giacalone et al., 1999).

When no valid and reliable method for conducting EIS is developed, it is difficult for

organizations to use EIS effectively as an instrument to reduce turnover. Without at least

some consensus or reasoning behind the information the EIS should collect, its chances of

being an effective instrument to decrease turnover are bound to be ruined.

The main question that the research at hand aims to answer is:

To what degree and in which form and content is conducting EIS a valid and reliable method to obtain information about why employees leave an organization?

The first objective is to provide a critical and extensive review of research that has been

(9)

main question and to also give a more conclusive and practical answer than previous research

on this subject, the author intents to develop a systematic and reliable procedure that provides

the means to conduct a valid and reliable EIS. There are several methods and forms for

conducting EIS, so the second objective will be determining what would be the most valid

and reliable method to conduct the EIS. After this is accomplished, the actual content of the

EIS can be gathered and selected. These objectives are to be achieved in the context of the

ultimate goal of EIS: decreasing turnover.

In finding a solution to the presented problem it is useful to answer a number of sub

questions. To begin developing a research instrument, deducting the criteria which it should

adhere to is an important step:

What are criteria for the EIS to be a valid and reliable information gathering instrument?

After determining the criteria this will help answering questions about form and content of the

EIS:

In which form should the EIS be conducted to obtain valid and reliable information?

The following step is determining information that is useful to end users of the results of EIS,

e.g., human resource employees or management. This is to be done partly theoretical as well

as empirically by interviewing some members of the EIS end user group. All this should lead

to an answer on the question:

(10)

Previous research on the topic of EIS has always shown great duality in its opinion on the

value and the use of data collected by EIS. On one hand researchers have plead for its ability

to play a sizeable role in reducing turnover while on the other hand its value, for a variety of

reasons, is argued to be very questionable (Garretson & Steel, 1982). The research at hand

intents to develop a valid and reliable method and instrument for conducting EIS; one that is

capable of gathering valuable information from employees who are in the process of leaving

(11)

2. THEORY

2.1 Background of EIS

2.1.1 Purpose and Espoused Benefits of Exit Interviews and Surveys

Giacalone et al. (1995) report three distinct purposes for conducting EIS: diagnosis and

strategy (to uncover problems and reasons behind turnover), public relations (to provide

separating employees the opportunity to raise issues that are important to them) and finally

personal reasons (to give employees the opportunity to voice frustrations). Although EIS are

traditionally used to uncover issues that make an organization susceptible to voluntary

turnover, it can also be used as a method to uncover information about organizational issues

that are unrelated to turnover (Giacalone, Knouse, & Pollard, 1999). The advantage of this is

that EIS could not only help to reduce turnover, but also improve performance of the

organization in general. It has the potential to uncover hidden sources of organizational

problems and concerns (Mok & Luk, 1995). Furthermore, organizations putting effort in

separating from employees on good terms gives departing employees feelings of respect and

appreciation. It sends a message that the opinions of employees are valued, regardless of the

fact that they are leaving the organization. This is also considered to be important in terms of

employment branding, which is about the organizations percieved identity and managing the

organizations image in its role as an employer (Backhaus & Surinder, 2004).

The moment an employee changes its role from being an active member of an organization to

an individual who is leaving the organization, he or she has the opportunity to give more

(12)

repercussions when talking about the organization in comparison to a continuing employee

(Hinrichs, 1975). The exit interview could possibly motivate a departing employee to open up

and speak more freely about their organization as when compared to a situation where the

employee is actually part of the organization (Hilb, 1978; Giacalone, 1989). In some cases

this is thought to be because employees know that, after giving their opinions, they do not

have to confront the other members or supervisors of the organization (Knouse et al., 1996).

Of course this is only the case when the EIS is not conducted by the employees‟ direct supervisor and the procedure takes place outside of work or after the last working day.

EIS could also help to fine tune the recruitment and selection process. The information that is

given by the departing employee can give a realistic view of the job requirements and the

common problems faced in the job; which is known to be helpful information for improving

the selection and recruitment process and hiring new employees (Sherwood, 1983). EIS could

also serve to uncover unwanted business practices, such as sexual harassment, discrimination,

deceptive public statements and unethical practices (Giacalone et al., 1999). All of the topics

in the EIS are dealt with under the assumption that information that is conveyed by the

employees is honest and actually used by end users of the data (e.g. HR departments and

management) to improve their organization. As mentioned before in the introduction,

literature on EIS indicates there are still several problems to overcome.

2.1.2 Shortcomings of Exit Interviews and Surveys

The literature indicates two categories of flaws that the EIS reportedly possesses. First there

are administrative issues; flaws that result from the way EIS are administered or the way the

(13)

1991). Garretson & Steel (1982) concluded that too often organizations fail to use the

collected information. EIS data was generally obtained and stored in such a way that it was

impossible to observe trends in the data; consequently it proved difficult to draw sound

conclusions from the data. Mok & Luk (1995) stated that organizations that do not make an

analysis, that do not compile summaries and simply put the data in their employee archives

make EIS merely a symbolic gesture. In many cases it is advisable for organizations to act on

trends in their collected data, rather than on the data of a single employee since more

calculated and grounded actions can be taken this way (Fottler, Hernandez, & Joiner, 1994).

The use of data in this fashion also ensures that the effects of outliers, the unrealistic views of

extremely positive employees and very disgruntled employees on the other hand, are

accounted for.

The second category of flaws that EIS is susceptible to are methodological issues. Most of the

critiques about methodological flaws are focused on the distorted information that departing

employees could give when conducting the EIS (Giacalone, Knouse, & Montagliani, 1997).

There are a multitude of reasons why employees are, either consciously or unconsciously, not

or not completely honest when providing answers on the questions of their exit interview or

exit survey. One of the reasons for this fact is that departing employees do not have a real

incentive for giving honest answers (Giacalone & Duhon, 1991). Departing employees are

leaving the organization and organizational changes that could originate from their answers

on EIS questions would not affect them. An employee could consciously decide not to

criticize the organization because he or she might want to work for the same company again

in the future. The employee could also choose to be dishonest because of fear for negative

(14)

colleagues could for example be a change in labour conditions or layoffs. It is also possible

that departing employees think, because of earlier experiences, the organization will not make

use of the information they provide. It is hard to imagine they will put much effort in

completing the exit interview or survey under this condition (Zarandona & Camuso, 1995).

Another reason that is mentioned by researchers as a possible cause of distortion is the

concern that employees will not be honest because of fear of not getting a good reference

from their employer (Hinrichs, 1975). Although this argument is probably not of much

relevance for employees anymore, because in this day and age references are generally

checked poorly and sometimes references are not asked for or checked at all (Walcot, 2004).

Still employees, especially those that leave voluntarily, may have a desire to depart on good

terms and give their answers from a somewhat positively and defensively biased attitudinal

approach. One of the possible explanations researchers present to account for the lack of

openness or honesty shown by employees in EIS situations is based on the theory of

impression management (Leary & Kowalski, 1990). Impression management in an

organizational context is a process in which the employee intentionally sends out distorted

information to create or maintain a favorable image (Brennan, Guillamon-Saorin, & Pierce,

2009).

2.1.3 What EIS Measure

The primary goal of the EIS procedure is to determine why employees leave an organization

(Feinberg & Jeppeson, 2000). To reach this goal literature points out several subjects that can

(15)

adequacy of training, working conditions and job ratings (Giacalone & Duhon, 1991).

Questions serving the EIS‟ primary goal and that are typically asked in EIS are for example: “What are your reasons for termination?”, “Was your termination preventable? Explain why or why not.” (Sherwood, 1983). Furthermore general information is gathered by means of asking about their opinions and attitudes towards different aspects of the organization.

Examples of questions serving this goal are: “What is your opinion about your working

conditions?”, “What is your opinion about your relationship with co-workers”, “Do you have comments on other company policies and procedures?” (Sherwood, 1983).

What de facto is measured by EIS are reasons why employees, according to themselves, have

chosen to leave their organization. EIS forces the employee to make causal attributions.

Causal attributions are explanations for the occurrence of an event (Bradbury and Fincham,

1990). In the case of EIS explanations for why the employee has made the choice to leave the

organization. Research that is focused on measuring employee attitudes and their antecedents

(e.g. opinions and causal attributions) do so by using self-report techniques that rely on

deliberative processing which in turn emphasize explicit mental processes (Johnson &

Steinman, 2009). Subsequently, an assumption that end users of the EIS data make and rely

on, is that employees have deliberately processed or are during the EIS sufficiently able to

deliberately process what their reasons for leaving are. Furthermore they assume these reasons

(16)

2.2 Validity and Reliability

2.2.1 Previous research on validity and reliability

Many researchers seeking to evaluate the quality of EIS information do so by conducting post

termination EIS and use data thereof as a criterion to validate the reasons given by employees

in their exit interview (Steel, Griffith, & Hom, 2002). This type of research conducted by

Feinberg and Jeppeson (2000) indicated there was no significant correspondence between

reasons for leaving given on the exit interview and reasons for leaving given on the exit

survey. They concluded that exit interviews lack validity and are unreliable. They also stated

that although exit interviews have become the main weapon in the fight against turnover,

organizations should consider adapting a more independent method of collecting data. An exit

interview done by company management does provide some information about the reasons

for turnover but is especially deficient on the subjects of job dissatisfaction and conflict with

management (Hinrichs, 1975). In light of these research outcomes it is relevant to realize that

the exit interviews were conducted in an unstructured manor and compared to answers given

in a paper and pencil questionnaire that had to be returned by mail. The reasons for leaving

are measured on two occasions and with two different instruments, first by conducting an

interview and six months later by using a survey. The follow up surveys in the research by

Hinrichs (1975) provided more negative responses about the organization in comparison to

the exit interview; 76% of the surveys contained negative responses compared to 4% of the

interviews. In contrary to the interview the surveys were returned anonymously. This could

mean that anonymity guarantees more valid responses. Research on the role of anonymity by

Giacalone et al. (1995) showed that data from respondents, who had to put their social

(17)

collected through exit surveys were this was not mandatory. These results suggested that

anonymity plays an important role in gathering valid and reliable information.

Pre-post comparisons in quasi-experimental research are susceptible to a number of plausible

and potentially severe biases (Wholey, Hatry, Newcomer, 2004). Ironically, research on the

validity of exit interviews conducted by Lefkowitz and Katz (1969) and Hinrichs (1975) the

differences in pre and post outcomes of the exit interviews and surveys in their research could

be also be attributed to two common threats to internal validity of experiments. Threats to

internal validity are those factors that have the potential to provide alternate explanations for

the observed effects (Christ, 2007). The threats in these cases originate from a change of

measurement instrument and the presence of history effects. An example of a history effect in

this case is that the data results only indicate that a majority of the departed employees realize,

during their new job in the months after their departure, that in fact it were other aspects of the

organization that made them leave. Change of measurement instrument is another plausible

possibility for the differences between reasons provided by departing and departed

employees. The cause of this fact might be that either the interview or the survey is more

prone to distortion. The setting of the exit interview could for example lead to social anxiety

and distort the process (Gardner, 1988). Ultimately, there is no way to objectively determine

which information is more valid; the very nature of the EIS makes it impossible to empirically

generate conclusions on this matter. For example, consider the following experiment to gain

insight in the effects of maturation on the reasons why employees choose to leave their

organization. One should first obtain the reasons for leaving at the point in time of actually

leaving the organization (e.g. the last day of work), then change the organization based on the

(18)

these outcomes should be compared to a group that has left and gave different reasons on the

follow up procedure six months later, then change the organization (while keeping in mind

these reasons), rehire them, and see if they will stay this time. All this is done while

controlling for other external factors which might pose a threat to the internal validity of a

field experiment (Tybout & Calder, 1977). Finally it is possible to make a comparison and see

which reasons were more accurate. This example makes it clear that it is very difficult to

empirically investigate whether the reasons that employees give on EIS are valid and reliable.

2.2.2 Defining validity and reliability

Modern models of validity have concluded that validity is a unitary construct with different

types of evidence that are supportive of validity (Messick, 1995). Cascio & Aguinis (2005)

state: “Validity is also not a dichotomous variable (i.e. valid or not valid) but is a matter of degree”. There is little certainty to the degree of validity of EIS and research provides mixed results (Knouse et al., 1996). Standards for educational and psychological measurements

(AERA, APA, & NCME, 1999) indicate that to determine the degree of validity, there are

three types of principal strategies: content-related evidence, construct-related evidence and

criterion related evidence. Content related evidence is about whether or not the measurement

procedure contains a fair sample of the situations it is supposed to represent (Cascio &

Aguinis, 2005). As stated before, a question that remains unanswered by literature is what

exactly the content of the EIS should be. The author is determined to specify the content of

EIS and provide content related evidence for these contents based on turnover theory and

(19)

Construct related evidence is closely related to content validity but deals with inferences

about test construction while construct related validity deals with the inferences that can be

made from scores on an instrument (Cascio & Aguinis, 2005). In the case of EIS, this means

that if EIS has a certain amount of content related evidence, certain statements can be made

about the how scores (i.e. the collected data) can be interpreted. For example an inference that

can be made is the use of the EIS for collecting information on certain aspects of the

organization. The ability of EIS in attaining this goal is well supported (Giacalone,

Jurkiewicz, & Knouse, 2003). Other possible inferences still show a lack of construct

evidence.

What is perhaps more important in relation to the goals of end-users of EIS, is to determine if

the data that is collected by EIS can be a valid and reliable source of information for making

decisions regarding organizational measures that could decrease turnover. Criterion related

evidence has to be gathered in order to determine the empirical relationship between predictor

and a criterion (Cascio & Aguinis, 2005). In terms of EIS this means asking: does the use of

EIS data measurably contribute in decreasing turnover? It is somewhat disappointing to

acknowledge that Zarandonna and Camuso‟s (1985) plea for controlled studies in organizations, to concretely indicate EIS accuracy as a basis for change, has not yet been

answered. Conducting these kind of studies will contribute in providing criterion related

evidence and determining the degree of EIS validity and reliability. These kinds of studies are

beyond the scope of this paper, but by providing content related evidence for conducting EIS,

the first essential step for gathering criterion related evidence is made. Determining what

procedure and which contents of EIS is most likely to provide valid and reliable results will

(20)

constructed exit interview or survey could be used to gather construct related evidence and

criterion related evidence of validity.

Although validity and reliability are related to each other, the reliability of EIS results is an

aspect that has gotten less attention in literature. The data collected from interviews and

surveys should, if properly constructed and are conducted under the right conditions, provide

reliable data. The degree of reliability of, for example, surveys can easily be tested through

use of common scientific methods, but the question remains if they do not provide a false

sense of security. General assumptions are usually made like demanding high test-retest

reliability. For example, in the case of exit surveys test-retest reliability should be high if the

same survey is taken four days apart. But a low test-retest reliability of surveys taken months

apart may simply mean former employees have changed their views and opinions on what

caused them to leave the organization. Note that this does not imply that an exit survey is not

a reliable method and does not provide reliable data. Reliability simply assumes the variables

that are measured do not significantly change over time. Which why it is important to realize

that what EIS measures are the opinions of people and those are more susceptible to change

than for example an adult person‟s height in centimeters. Consider the following quote by Wikman (2006): “In sum, we can say that a not inconsiderable part of the precariousness of survey responses does not have anything at all to do with measuring instruments and

measurements. It is present as a natural element in the reality of what is being measured – due

to linguistic prerequisites and unclear ideas on part of the individual.” If one is aware of the fact that opinions can change over a certain period of time, a somewhat lower degree of

reliability does not have to be a problem given the goals of EIS. A bigger problem here is that,

(21)

21

internal consistency and therefore a certain degree of reliability, the instrument itself may not

be valid. An example is collecting an employees name on exit surveys; results will likely be

repeatable and somewhat consistent, but will not be valid because employees are likely to

withhold the same kind of information. The importance of collecting reliable data should not

be underestimated and undervalued, but considering past research on EIS, ensuring valid

responses has higher priority and is probably a much bigger problem.

For answering the second and third sub question of this paper, it is relevant to determine the

content of the EIS. Furthermore it is important to determine the method and conditions in

which known negative effects of external factors on validity and reliability of the EIS data are

minimized. Previous research on the validity of EIS data indicate that there are several factors

and conditions that may have a tremendous impact on the information that is collected:

interview or survey (Feinberg & Jeppeson, 2000), anonymous or not (Giacalone, 1993),

conducted by the company itself or by an external consultant (Hinrichs, 1975), content and

willingness to report on subjects (Giacalone & Knouse, 1990). Figure 1 represents a

conceptual framework of the process of collecting of EIS data, more specifically the reason(s)

for leaving the organization. It indicates the external factors that are of influence on the

validity, reliability and thereby use of the EIS data.

Figure 1: Conceptual framework of factors influencing validity EIS data

Exiting employee

reason(s) for leaving and attitudes concerning the organization

Reason(s) for leaving

organization as interpreted by end users of EIS

External factors (e.g. method, procedures and content of EIS)

Formation of attitudes and judgements

EIS data Exiting employee‟s stated

reason(s) for leaving and attitudes concerning the organization

Exiting employee reason(s) for leaving and attitudes concerning the organization

(22)

The assumption that is made from here on is that if EIS procedures present the „right‟

questions and control sufficiently for external factors, organizations are able to react on trends

in reasons that employees give to decrease turnover.

2.3 Content of EIS

2.3.1 Determining content

A question that remains unanswered by literature is precisely what content the EIS should

consist of. Giacalone & Knouse (1989) point out several subjects that can be handled in EIS,

for instance: reasons for leaving, quality of supervision, pay, performance appraisals,

adequacy of training, and working conditions. Yet theory on EIS provides no clarity about

behind what data, exactly, should be collected (Giacalone & Duhon, 1991). Considering a

purpose of EIS is diagnosis and strategy (Giacalone, 1995), it is relevant to acknowledge that

available literature on the validity and reliability of EIS data only compares the main reason

of leaving (Lefkowitz & Katz, 1969; Hinrich, 1975; Feinberg & Jeppeson, 2000). But this

research neglects the validity of other, possibly reliable and valuable, information that has

been collected during the EIS. This information could also benefit the organization in its

effort to reduce turnover. It is imaginable that the main reason for leaving is a salary increase

at another job, but if their direct supervisor had not been an emotionally unstable person they

would have stayed, regardless of the possibility attaining a higher salary elsewhere. Another

example is a situation where employees indicate the amount of work stress is their main

(23)

opportunities is what makes the amount of stress, and staying at the organization, intolerable.

In theory there is an endless amount of factors or combinations of factors that could lead to a

person‟s departure. The ever-expanding amount of theoretical constructs that explains turnover and its causes substantiates this argument (Hom, Kinicki, & Domm, 1989). It is

possible that just one factor weighing very heavily (e.g. discrimination) is responsible for the

departure but it could also be a combination of six factors that individually are of minor

importance but combined led to the employee‟s departure. These kinds of examples suggest

that not just the main reason for leaving has to be measured and taken into account for, but

other factors might also be of importance in determining the reason behind an employee‟s departure. Since EIS are about explaining turnover and with the intention to produce an

informed and reasoned answer of what the content of EIS should be, an analysis of previous

research that is focused on explaining turnover seems justified.

2.3.2 Turnover in relation to content of EIS

Research focused on analyzing the factors that contribute to the departure of employees from

an organization, led to the development of turnover theory. There are two schools of thought

in this field, the „psychological school‟ and the „labour market school‟ (Morrell, Loan-Clarke, & Wilkinson, 2001). The labour market school focuses on subjects like job search (Fahr &

Schneider), labour market flexibility (Davia 2005) and unemployment level (Trevor 2001);

subjects that are generally outside the influence of individual organizations. The

psychological school focuses on explaining or predicting behavior of employees leaving an

organization (Morrel et al., 2001). The subjects that are frequently covered by this school of

(24)

and job involvement (Hom and Kinicki, 2001). Considering the goals and nature of EIS, but

also the possible influence that individual organizations could have on its employees, it seems

sensible to focus the content of EIS on the subjects researched in the psychological school of

turnover theory.

There has been a considerable amount of research on the complex subject of turnover. To

exemplify the complexity and amount of factors involved a count of turnover models reveals

there are currently 24 models in print, most of which are based on correlational research

(Steel & Lounsbury, 2009). The following quotation from Holtom, Mitchell & Lee‟s (2006) paper indicates the variables that are most frequently linked to turnover: “Over the past half

century, psychologists and management researchers have focused on two major factors as

causes of employee retention: job satisfaction and job alternatives. People who are satisfied

with their jobs (e.g., evaluate positively their pay, supervision, chances for promotion, work

environment and tasks) will stay, and those who aren‟t will leave. In addition, people with who have more job alternatives will be more likely to leave than those with fewer

alternatives.” These two variables, job satisfaction and job alternatives, are at the base of many theoretical models that explain the causes of turnover (Steel & Lounsbury, 2009).

Until recently there were generally two types of models: turnover process models which are

focused on how people quit and turnover content models which are focused on why people

quit (Meartz, 2004). In the past decade management researchers are trying to integrate both

process and content turnover models. This has led to a better understanding of interventions

that successfully decrease turnover (Brooks, 2007). Tea Heon et al. (2008) have recently

developed a model that has more success in explaining voluntary turnover than traditional

(25)

Mitchell (1994) because it takes into account that the importance of job satisfaction and ease

of movement (e.g. job alternatives) depend on the group of leavers studied. According to their

research, based on a sample of 6,198 representative respondents that are part of the American

workforce, four distinct groups of leavers could be identified. The distinction is based on the

path they take when leaving the organization and the most important variables involved in

their decision.

Group 1: Individuals who quit after successful searching for another job.

Following the path: dissatisfaction → search → new job→ quit. Group 2: Individuals who quit to search for another job.

Following the path: dissatisfaction→ quit → no new job→ search. Group 3: Individuals who accept unsolicited job offers.

Following the path: no search → new job → quit.

Group 4: Individuals who quit for family-related reasons.

Following the path: quit → no new job → no search.

Group 1 and group 2 experience some form of dissatisfaction with their current job and

groups 3 and 4 do not. In group 1 there are usually a number of factors that, over an extended

period of time, cause dissatisfaction which in turn instigates job search and eventually make

them change jobs. In group 2, 3 and 4 there is a specific event (e.g. conflict situation,

unsolicited job offer, pregnancy) accountable for the decision to leave without already having

(26)

Lee, Inderrieden, 2005). The different groups are susceptible to different interventions that are

focused on preventing turnover. If a large number of leavers are part of group 3 the

organization may choose to see if they are competitive enough on the labour market. Given

the „war on talent‟ and the „raiding of employees‟ it is seems likely that the group receiving unsolicited job offers will grow in the near future (Tea Heon et al., 2008). Group 4 mostly

consisted of employees choosing to take care of children or spend more time with their

family. To decrease turnover for this group the organization has to see if the organization can

for example offer flexible working times. If EIS could collect the right kind of data, suitable

(e.i. turnover reducing) interventions could be implemented. The added value of using the

model of Tea Heon et al. (2008) in relation to EIS is that it determines the prevalence of

certain groups of leavers. With this knowledge group-specific interventions can also be made,

which could lead to a higher reduction of turnover. Furthermore because it relies on more than

the traditional focus on job satisfaction, the model is able to better explain the causes of

turnover.

Still, as with other models explaining turnover, there are signs of incompleteness in the

model. It ignores to account for the effects of for example organizational commitment, which

beside job satisfaction, is a prominent attitudinal predictor of voluntary turnover. In

meta-analysis by Griffeth, Hom, & Gaertner (2000) an average corrected correlation of −.23 is

found for organizational commitment. There has also been a considerable amount of research

on the connection between certain individual factors (e.g. personality traits) that have shown

an impact on turnover decisions (Zimmerman, 2008). A great deal of research on the subject

of voluntary turnover is still being conducted and there is no universal agreement on the

(27)

Holtom, 2008). This fact might prove to make the selection of content for EIS somewhat more

arbitrary than desired. Since models that can account for all the reasons behind employee

turnover have yet to be developed, data collected by EIS could possibly aid in reaching this

goal.

2.3.3 Further considerations

An option not directly related to turnover literature, but relevant to EIS is the option of adding

company specific or industry specific questions to the EIS. This might prove to be of interest

of the end users of the EIS data because these questions can uncover specific causes of

problems in the organization, which is another purpose of the EIS (Giacalone et al., 1995).

Theoretically this should also lead to information that is adjusted to the needs of the end users

of EIS data. For example questions about the functioning or effects of a specific piece of

software that the organizations employees have to work with. It is reasonable to conclude that

the possibilities for the content, the questions asked in EIS, and the information that could be

gathered is very broad. Relevant when selecting content is Holtom et al.‟s (2005) remark that

(28)

2.4 Methods and Timing

2.4.1 Developments in methods

In addition to the traditional methods of conducting EIS recent technological advancements

increased the possibility for organizations to conduct computerized online exit surveys

(Giacalone et al., 2003). In comparison to an exit interview the exit survey, besides providing

a greater sense of anonymity, is less expensive, less time intensive and makes data collection

and categorization easier (Giacalone et al., 1999). By using the survey method it is easier,

under the presumption of using closed questions, to adhere to Giacalone & Knouses (1989)

caution against making generalizations out of statistically insignificant data. The added

advantage of using computerized surveys is that it is possible to use decision threes that can

distinguish between certain groups of leaving employees and offer the most suitable questions

(Giacalone et al., 1995). This prevents the posing of irrelevant questions and excessively long

surveys and in case of a paper and pencil survey, the need to use large amounts of papers.

Although in some situations using computerized exit surveys is not optimal when it comes to

gathering in depth information, the increased speed, accuracy and adaptability forms a great

incentive for using this method (Giacalone et al., 2003).

Recent research by Westaby (2006) on the measurement of turnover motives by consultants

showed that consultants focused on developing turnover reducing interventions should

measure whether different facets of job satisfaction contribute to the employee‟s decision to leave. This is done by including reason scales into surveys, which means respondents are

asked to indicate whether or not a specific factor was of importance in their decision to leave.

The research outcomes showed that this approach has considerably more predictive validity

(29)

into turnover motives. In addition to these findings, a study by Rice, McFarlin, & Gentil

(1991), showed that including questions that measure the self reported importance of a facet

are significant moderators of outcomes on job satisfaction scales.

2.4.2 Timing

The suggested time EIS should take place as suggested by literature ranges from the last week

before departure up to 60 days after a person leaves the organization (Dwight, Leonard, &

Weichmann 2006). By some practitioners the last day of work is not seen as a suitable day to

conduct EIS because of all kinds of practicalities and lack of sufficient attention (Barada,

1998). In contrary to this assertion, Mok and Luk (1995) state that EIS should take place on

the last day of work because leavers might be less likely to respond honestly if they still have

to work in the organization for a certain period of time. Anecdotal evidence indicates that EIS

conducted two to seven weeks after the employee has left the organization ensures more

meaningful results (Compensation & Benefits for Law Offices, 2005). In short, literature

offers little consensus on answering which time of administration is the most effective; the

lack of empirical research on this topic is likely to be the main cause for this uncertainty.

When using online surveys it is recommended to have the exiting employee complete the

survey by making it a mandatory part of the departure process (Brooks, 2007). This ensures a

high participation rate and should also make the collection of statistically significant

(30)

3. METHOD

3.1 Use of literature

Analysis of both quantitative and qualitative literature on EIS and turnover was done to

provide answers on several of the questions posed in the introduction section. Most but not all

literature was found by using Purple Search, the meta-search engine of the University of

Groningen (http://purplesearch.ub.rug.nl). This search engine enables simultaneous search in

the most important academic databases. Primary, secondary and tertiary literature was found

by using keywords and combinations of keywords related to the research questions (e.g. exit

interview, exit survey, validity, reliability and turnover). Subsequently selection of relevant

articles was done based on titles and abstracts of the articles that were found. Additional

relevant articles were obtained by making use of the snowball method. Besides articles from

peer reviewed journals some qualitative information on EIS was found in specialist magazines

like for example “HR Magazine”. Given the limited amount of research on EIS available, it

seemed useful to include some of the information in these articles that was provided by

experts or practitioners. The limited amount of academic literature on EIS is also responsible

for the need to use of articles dating back as far as 1943. One governmental document by

Giacalone (1993) which was not freely available in scientific databases was obtained by

contacting the author.

3.2 End user interviews

Three semi-structured interviews have been held with end users of EIS data, in this case HR

(31)

recommendations made by researchers, who have called for making data that is collected by

EIS more relevant and usable for practitioners (Giacalone et al., 1999). The goals of the

interviews were threefold. The first goal was see which exit interview and survey themes or

variables are seen as important and relevant according to end users. The second goal was

attain information on the use of EIS in practice and the gap between theory and practice. The

third goal was to probe which open ended questions in EIS instruments are of particular

interest to end users.

Besides open ended questions, two additional categories of questions that had not been

directly identified in the theory section have been added after the interviews took place. The

first category contains questions that provide information on early turnover (e.g. within one

year). The goal of these questions is to provide information on the introduction of new

employees in the organization and the functioning of recruitment and selection process of the

organization. The second type of questions is about the occurrence of undesired incidents (e.g.

discrimination, bulling, sexual harassment, violence). According to the experiences of an

interviewee these kinds of incidents could prove to be a cause of turnover.

3.3 Selection of content

Testing employees with the use of research instruments is systematic in three areas: content,

administration and scoring (Cascio & Aguinis, 2005). The current article proposes a new way

of looking at EIS by integrating recent developments in turnover research with variables that

(32)

measure to what degree the relevant subjects that were considered in the theory section played

a part in the employees‟ decision to leave the organization. When the impact of each of these constructs is measured, organizations can instigate interventions that can relieve or improve

aspects of the organization that respondents found to be problematic or a reason to leave.

When considering the three purposes of EIS: diagnosis and strategy, public relations and

personal reasons, it becomes clear although party dependent on the content of the EIS, the

mere fact of conducting the EIS will fulfill the last two purposes. The approach that seemed

most suitable for selecting content for the EIS is refraining from designing new questions by

measuring the relevant constructs with existing scales. This way most constructs are measured

through the use of established instruments that have been proven to be valid and reliable in

the past. Because of the setting and nature of the EIS, questions are adapted to past tense. The

respondent is asked to reflect on his period as an employee and the reason(s) that formed the

decision to leave the organization. The method for selecting content is to measure in the

theory section mentioned variables that have been shown to be linked to turnover and could

also possibly serve the goals of EIS. As stated before, several variables that were mentioned

in the end user interviews and that were not of company specific nature, have also been added

to the selection.

3.4 EIS content

3.3.1 General variables

To ensure more candid responses on the questions of the exit survey the respondents name

will not be asked. To provide an even greater degree of anonymity, the variables that are of

(33)

(e.g. age of the respondent are measured in groups and not specific year). The subjects are

made clear that reports will not be based on less than 10 respondents. The general variables

that are measured are: sex, educational level, organizational rank (management or not),

department and tenure.

3.3.2 Tenure

Based on the interviews with end users, there was sufficient reason to include additional

questions for employees who quit within one year after being hired. A total of nine questions

related to the employees‟ experiences during their first period of tenure are added to the EIS. Subjects that are handled are: guidance and instructions, clarity of tasks and responsibilities,

openness of colleagues, fulfillment of job expectations and fulfillment of the expectations

raised by the people responsible for selection and hiring of the employee in question. The

argumentation end users gave for adding these questions is that information of employees

who quitted within one year could help improve the selection, reception and treatment of new

employees.

3.3.3 Shocks and reason for leaving

Literature showed that the occurrence of shocks are responsible or are party responsible for

the decision to leave the organization. In accordance with the study by Holtom et al. (2005)

shocks were measured by asking the following question: “Was there an event that caused you

to start thinking about leaving?” Subsequently three different categories of shocks that have

(34)

offered a better job, offered a job with more promotion opportunities, offered a job that paid

more money, conflict with management, conflict with supervisor, conflict with coworker,

performance issues (encouraged to leave, passed over for promotion), merger or acquisition.

The third category is “personal shocks”, which consists of: spouse or partner employment, family issues, significant illness, moved, back to school, started own business,

and other.

Holtom et al. (2005) do make clear that although shocks can be present, this does not suggest

that employers or consultants should disregard job satisfaction. Because of this fact and the

possibility of attaining additional information, it is still important to measure job satisfaction

in the group of leavers that has chosen to leave after experiencing shock. In addition to the

measurement of shocks other common reasons given for leaving an organization are added.

This allows for easy processing of data and provides end users with added or more specific

information. The respondents are asked to select their main reason for leaving but also for

other reasons that have played a role in their decision to leave the organization. These reasons

are based on the common reasons for employees to quit their jobs that are mentioned by Tea

Heon et al. (2008) and Giacalone & Knouse (1990). An example of a choice that is added

beyond shocks is “length of commute”.

3.3.4 Job satisfaction

For measuring the satisfaction with different aspects of the employees‟ job an adjusted

version of the Job Satisfaction Survey (JSS) is used. This instrument, developed by Spector

(1985), is well established among the other job satisfaction scales (Blau, 1999). The JSS is

(35)

respondents have to reflect on their experiences with a job, which they have chosen to fulfill

no longer. The JSS is a questionnaire that evaluates nine facets of job satisfaction that

contribute to overall satisfaction. In total there are 36 items, 4 for each facet, with a Likert

type response scale ranging from 1 (strongly disagree) to 6 (strongly agree). A higher score

meaning a higher level of satisfaction and because some items are worded negatively some

items must be reversed before summing with the positively worded into facet or total scores.

Because the JSS measures facets of job satisfaction, it allows researchers to find out which

elements of the employees‟ job contributes to either satisfaction or dissatisfaction (Rice, McFarlin, & Gentil, 1991). Measuring these facets is likely to prove useful to end users of EIS

data because it serves the goals of EIS and offers information on how a job was experienced

by the respondent.

The reason approach by Westaby (2001) is used to measure the impact of each JSS subscale

on the decision to leave. The reason approach entails that respondents have to rate the extent

to which each of the scales formed a reason for leaving the organization. This is done by

using the by Westaby suggested 3 point scale with ratings 1 (not a reason), 2 (a reason), and

3 (a strong reason).

3.3.5 Organizational commitment

The measure used to indicate organizational commitment is based on a shortened version of

Porters scale for organizational commitment (Porter, Steer, Mowday, & Boulian, 1974). This

commonly used instrument has been validated by for example Bline, Duchon, & Meixner

(36)

others that I was part of this organization.” The previously mentioned reasoning question measuring the effect organizational commitment had on the decision to leave was added. A

high score on the commitment scale means the employee is more likely to have taken pride in

organizational membership and believed in the goals and values of the organization. A

consequence of a high level of organizational commitment is that the employee is likely to

have exhibited higher levels of performance and productivity (Steinhaus and Perry, 1996).

3.3.6 Home-work interference

Negative work-home interference can measured by an adapted version of a validated subscale

consisting of three items derived from the SWING questionnaire (Geurts, Taris, Kompier,

Dikkers, Van Hooff, & Kinnunen, 2005). Cronbach‟s alpha for the scale was 0.67 in a study conducted by Peters, den Dulk, & van der Lippe (2009). To make the scale suitable for EIS

setting the questions were asked in past tense. A five-point Likert scale is used to record

answers, a high score corresponding with high levels of negative work-home interference. An

example questions is: “How often did it happen that your work schedule made it difficult for you to fulfill your domestic obligations?” The previously mentioned reasoning question, this time measuring the effect home-work interference had on the decision to leave, was added.

3.3.7 Leadership

Since dissatisfaction with leadership by an employee‟s direct supervisor is known for being related to turnover (Giacalone et al., 1995) and to provide information beyond that measured

by the job satisfaction questions, the questions as used by research measuring influence of

(37)

Instituut voor Integratie en Sociale Weerbaarheid, Rijks Universiteit Groningen 2009). For

answering the six questions on style of leadership, respondents have to rate on a five-point

Likert scale the occurrence of the type of leadership, which can either be participative and

directive. An example of a question on participative leadership is: “My direct superior shows respect for my personal feelings”. The previously mentioned reasoning question, this time measuring the effect type of leadership had on the decision to leave, was added.

3.3.8 Stress

Subjective stress has been shown to be related to occupational turnover and was measured by

a 4-item scale developed by Motowidlo, Packard, and Manning (1986). An exemplary item is:

“I feel a great deal of stress because of my job.” Respondents have to indicate agreement with the statement on a 5-point Likert scale ranging from 1 (strongly disagree) to 5 (strongly

agree). Motowidlo et al. (1986) reported a coefficient alpha of .83 for this scale.

3.3.9 Negative incidents

The incidence of violence, harassment and discrimination at work is measured by asking if

certain incidents took place over the past twelve months. The incidents that presented to the

respondent are based on the European Working Conditions Survey (EWCS; Fourth European

Survey on Working Conditions, 2005). An illustrative item is: “Over the past 12 months, have

you or have you not, personally been subjected at work to discrimination linked to gender?”

(38)

3.3.10 Open ended questions

In the EIS the respondent is given the freedom to report in his or her own words their

motivations for leaving the organization. Based on the interviews with end users several open

ended questions have been added to conclude the EIS. For example, one of the end users

stated: “What do employees consider to be the positive aspects of their job?” An open ended question inquiring for these aspects was added. This way the respondents are given the chance

to explain and ventilate their opinions and which aids in reaching several goals of EIS. An

example of an open ended question is: “What could be improved at the company you worked for?”

3.3.11 Positive job aspects

In addition respondents are asked to pinpoint which jobs aspects that are generally appreciated

by employees the organization fails to provide for or take care of. Doing so is also based on

the comments of an end user, who indicated that asking the employees which aspects of their

jobs should be improved could provide relevant information. To obtain this information a list

of the most appreciated and valued job aspects as found in research by Hugick & Leonard

(1991) is presented to the respondent. Subsequently respondents are asked to indicate on

which of these aspects the organization should be improved. The aspects were placed in

random order, an example aspect is: “Possibility to learn new skills.” Adding this list of specific aspects allows for easy processing and should provide clarity on what aspects can be

improved. To obtain more detailed information an open ended question asking the respondent

(39)

4. RESULTS

4.1 Validity and reliability of EIS data

An analysis of earlier research on validity and reliability was made. Flawed research methods

have very likely led to wrong or at least to ungrounded conclusions in early research on the

subject of exit interviews and surveys. The setting and methods that were used have

compromised validity and possibly reliability of the research on validity of EIS data.

Literature showed that especially exit interviews compared to exit surveys are likely to have a

high degree of biased responses due to, amongst other influences, impression management

and a lack of anonymity. These factors have possibly detrimental effects to validity and

reliability of data collected in exit interview situations. Another important aspect that is

contemplated on, is the fact that the data collected by EIS is highly dependent on the

capability of the employee to correctly answer the question why he or she has chosen to leave

an organization. These capabilities are questionable and employees might unintentionally

provide invalid and unreliable responses. The degree of validity and reliability of EIS data can

be considered highly uncertain. By means of a theoretical example it was shown that it is

difficult to provide empirical evidence for validity and reliability of EIS data.

4.2 Input from end users

The interviews that were held with end users, HR managers in this case, served three

purposes. The first goal was to find out which themes or variables are, according to end users,

(40)

that were proposed in the theory section were mentioned by the interviewees, they indicated a

few additional topics that might provide valuable information. This led to the expansion of the

survey by adding several questions based on the comments of these end users. One of the end

users was interested in questioning whether employees were willing to return to the company

and whether they would recommend the company to acquaintances. In similar fashion another

end user stated: “Many of our divisions are quite dependent on the attraction of local employees, which is why it would be interesting to see if the departing employee would

recommend our organization in her or his social circle.” This is measured by adding the

following question to the commitment scale: “I would recommend working at this organization to family, friends and acquaintances.” A reason for adding this question is that it might give an indication on how the organization is perceived (e.g. generally great place to

work) by leaving employees. For similar reasons a question whether employees would

consider returning to the organization was added.

After conducting several interviews, it became clear that almost every aspect the end users

considered useful for adding to content of EIS, had already been considered in the theory

section. The aspects mentioned by end users that were not added to the EIS instrument were

all kinds of company specific information. For developing an EIS instrument that can be used

in every organization adding company specific aspects was considered neither relevant nor

practical. Subsequently, efforts to conduct more interviews with end users were ceased. It has

to be noted however that company specific information might prove to be of great value in

fulfilling EIS goals. Consider for example a situation where the mandatory use of a new

software system leads to allot of job related stress. It would be useful to measure what is

Finding the good in goodbye

Finding the good in goodbye

Contents