Professional Documents
Culture Documents
Graduation candidate
Nicolò Wojewoda, stud. nr. 1334662
Faculty of Technology, Policy and Management
Delft University of Technology
nicolo.wojewoda@gmail.com
Thesis committee
prof. dr. Marina van Geenhuizen, chair
dr. Ronald Dekker, 1st reader
dr. ir. Bert Enserink, 2nd reader
Richard Florida said it best, when he wrote that the what (what do you want to do as a
career) and the who (whom to choose as your life partner) are the two biggest decisions in
your life, but perhaps the where has an equal if not bigger effect, since it often determines
an answer to the other two.
Had it not been a graduation project, my research would have still had a strong appeal, for
me to understand better what to consider and how to do it, when deciding upon my next
step in life and work. Fortunately, it was even more than that, providing me with a learning
experience and a chance to make a real impact with the concepts, tools and methods Iʼve
learned in these two years as a student of Engineering & Policy Analysis. I sincerely hope
the results of this study will give interested stakeholders a glimpse of what the current
generation of young graduates has on their minds, allowing them to shape more relevant
and focused policies for the places they live and work in.
This research could have not happened without the precious cooperation of Stefanie
Kirwald and the Alumnivereniging TU Delft, the indispensable feedback of Marina van
Geenhuizen, Ronald Dekker and Bert Enserink and the loving support of my friends and
family. I owe them all a big debt of gratitude and I hope this study stands as a testimony of
the experience we enjoyed together in these last five months.
This work is dedicated to my mom and dad, whom I love and thank for supporting me
throughout these years and who, I hope, will finally understand what their son is studying
and be able to explain that to their relatives and friends!
2 Methodology
8
2.1 Overview
8
2.2 Literature review
9
2.3 Conceptual approach
10
2.4 Survey design
10
2.5 Survey analysis
14
3 Theoretical background
16
3.1 Existing theories and models
16
3.2 A new conceptual approach
19
3.3 Hypotheses
26
5 Conclusion
51
5.1 Results of the research
51
5.2 Policy implications
53
5.3 Reflection on theoretical insights
54
5.4 Strengths and weaknesses of the study
55
References 57
Appendix A: Keywords 60
When it comes to location decisions, existing theories and models canʼt provide an all-
encompassing solution to understand the problem at hand. Most of the existing studies
highlight specific factors that have an influence in the decision (e.g. utility, abstract needs),
others present statistics on the decisionsʼ outcomes, others instead focus on the process
of the decision and the context in which it takes place.
Therefore, a new conceptual approach was introduced, which would interpret and
synthetise the existing research themes into a new perspective that could be eventually
used to study the case under analyis. In this new approach, there are three basic
components: the factors that influence a graduateʼs decision on where to locate next for
living or working purposes; the decision-making process itself and its elements; the
decision-making outcomes and the related attributes. Such new perspective would have
allowed an empirical study of the links between factors and outcomes, among factors
themselves, or simply the understanding of outcome values. To this goal, the elements
involved had been operationalized in order to allow statistical analyses to be performed.
Also, a few hypotheses were introduced, that formalized existing claims from previous
studies or from common knowledge.
In order to put numbers in those variables and test the hypotheses associated with this
new perspective, a questionnaire was delivered to a subgroup of the TU Delft graduatesʼ
population, representing the cohorts of 2007 and 2002. The questionnaire, in the form of
an online survey, had 142 full respondents, which considering the total number of
graduates invited to answer the questionnaire, corresponds to a response rate of around
20%. Although bias was present in many forms, the resulting sample was fairly
representative of the TU Delft graduatsʼ population, allowing a few basic statistical
analyses to be performed on the survey data.
The results of the survey showed first that, when it comes to residential location choices,
Randstad (84.4% of graduates locating there) and especially South Holland (67.2%) are
the most attractive regions, whereas less than half (46.1%) of the graduates decides to
locate in Haaglanden and even less (30.5%) to stay in Delft.
Analyses show that having oneʼs partner/spouse close and housing opportunities are the
two most important living environment-related factors in the choice of a residential location.
Delft scored quite low on both aspects, since only about half of the graduates admitted
they were satisfied with them.
First employment location choices show even more preoccupying results for the local
stakeholders, with Randstad (81.5%) and South Holland (53.2%) still managing to attract a
considerable portion of the graduatesʼ population, but Haaglanden (29%) and Delft
(16.9%) falling way behind their figures for residential location choices.
In fact, an analysis of the perceptions of factors that influence the graduatesʼ decisions
shows that having an interesting and challenging job, as well as opportunities for career
development, although being the most important criteria for the choice of a first
employment location (respectively, 82% and 73% of the respondents consider them very
important or more), are among the factors that graduates are mostly dissatisfied with in
Delft (only about half of the respondents are satisfied with them).
What emerges from the results of this study is a picture that highlights the incapacity of
Delft and Haaglanden to retain their graduates, especially when it comes to working
purposes. Unfortunately, analyses havenʼt provided any useful insights on factors that can
be affected by policies and therefore the room for action seems to be pretty narrow at this
point. However, the limitations of the study have to be considered and further research will
hopefully shed some light on possible graduate retention strategies for the benefit of the
local economy.
1 Introduction
1.1 Research context
Economic development is often mentioned as one of the pillars of social and individual
well-being: increasing the wealth of a nation by creating jobs, growing incomes, improving
literacy rates and life expectancy and, in general, achieving a better quality of life is both
on most political agendas and in the minds and hearts of individuals, families and
communities around the planet.
Throughout the millennia of human existence and especially in the recent centuries of
accelerated growth of the worldʼs economies, various factors have been put on the
spotlight as key enablers of economic development: what for Adam Smith was the division
of labour and what for David Ricardo was the comparative advantage between nations,
todayʼs key ingredient in economic development is often considered to be knowledge, so
much so as to generate a related buzzword, knowledge economy, which policy-makers
and researchers alike have been focusing increasingly more on in the last decades.
The European Union has been pursuing since the year 2000 the goal of becoming “the
most competitive and dynamic knowledge-based economy in the world” (European
Council, 2000) and the Netherlands has been following suit since then. In fact, a series of
government policies (e.g. the Ministry of Economic Affairʼs “Action for innovation”) and
public debates initiated by non-governmental actors (e.g. Kennislandʼs “Knowledge
Economy Monitor 2003”) have been discussing and providing suggestions for progress on
this matter. But what is really the knowledge economy all about?
1
more, as high labour productivity per unit hour still remains one of the sources of
competitiveness (cf. the work of Robert Reich on the topic), the pressure on adding value
is on the high end of the labour productivity spectrum, where higher education institutions
operate and where universities are the central factor in the development of a better
knowledge economy.
2
can be found in examples such as the case of Twente (OECD, 2006), underscoring that
the new frame of spatial analysis constituted by cities and regions is a meaningful one for
the analysis of all sorts of economic impacts in the area, most of them related to achieving
progress on the topic of the knowledge economy.
3
city of Delft to fulfill its attractive role for TU Delft graduates and to grow as it pleases,
posing the first boundary to the cityʼs current plans.
Combining this attribution of roles and responsibilities from the Haaglanden planners to the
structural limitations of residential locations in the city of Delft itself, makes it very clear that
Delft does not compete at the same level of other Dutch cities, when it comes to having
graduates living there; similarly, Delft does not strive to be specialized in areas where
other cities in the Haaglanden consortium are specializing in instead.
However, even when considering the policy-makersʼ limitations in terms of growth of both
the job market and the housing market, the implications that arise when looking at mobility
of people in and out of the region are still meaningful and it is still compelling to
understand what are the mechanisms that drive people away from or make them stay in
Delft, or in Haaglanden, after graduation and what are the outcomes of their functioning.
As a matter of fact, even considering the structural limitations, the city of Delft and the
region of Haaglanden might still be able to meet their objectives more effectively, by
countering possible negative migration patterns by knowledge workers in the area.
4
Knowledge gaps and the role of this study
Due to the importance and relevance of the presented reflections, the current research
deals with the issue of first employment location and residential location decisions of TU
Delft graduates. Several knowledge gaps can be readily identified:
• thereʼs no knowledge about the first employment location and residential location
choices by TU Delft students after graduation; no research has ever studied where
these alumni decide to live and work right after their graduation;
• thereʼs little understanding about the reasons for TU Delft graduates to stay in or
leave Delft/Haaglanden; no systematic investigation on the factors affecting their
choices has been carried out so far;
• thereʼs lack of knowledge about the decision-making process itself that results TU
Delft graduates in moving or staying: whoʼs involved, when does it happen, how
does it happen;
The answer to this question will contribute to the achievement of the research objective nr.
1, with the goal of clarifying whether there really is or not an issue of TU Delft graduates
moving away from Delft/Haaglanden in favour of other places to work and/or live in, and
whether thereʼs consequently a reasonable concern or not for the loss or lack of
knowledge workers and/or residents in the local knowledge economy and/or living
environment.
5
Research question #2:
What insights can be derived from the decisions of TU Delft graduates on their first
employment location and associated residential location?
a. What are the factors that affect TU Delft graduatesʼ first employment location and
residential location decisions?
b. Do these decisions and factors influencing them differ, within the TU Delft
graduates population?
c. What are the decision-making processes of TU Delft graduates when choosing a
first employment and related housing location?
Answering this question will provide an analysis of the reasons behind TU Delft graduates
moving away from or staying in Delft/Haaglanden, therefore contributing to the
achievement of research objective nr. 2.
6
has ever really provided an overall picture of the major factors and influences together in
the same model.
7
2 Methodology
The following sections will outline the tools and methods that have been adopted in this
study, in order to answer the previously mentioned research questions and achieve the
research objectives. At first, an overview of the whole process will be presented (Section
2.1) and subsequently the elements behind the four steps that brought to the conclusions
of this study will be outlined: a review of relevant literature (Section 2.2), the design of a
conceptual approach on the issue and of relevant test hypotheses (Section 2.3), the
development of a survey based on a questionnaire that tests such approach and
hypotheses (Section 2.4) and the analysis of the surveyʼs results in order to draw
meaningful answers to the research questions (Section 2.5).
2.1 Overview
The following phases were carried out in the study:
• determining a methodology to answer the research questions
• surveying literature and deriving theoretical insights
• designing a survey to test the validity of those insights
• intepreting the surveyʼs results in order to answer the research questions
However, before beginning with the first phase, as with any other scientific inquiry, the first
step of the study has been trying to understand whether the questions being posed had
already been answered or not. The first part of the research therefore focused on finding
relevant sources, where details on first employment location and residential location
decisions of TU Delft graduates were either explicitly mentioned, or could have been easily
derived. Throughout the duration of the project, no such material has been found. The
actors mainly involved in the issue (e.g. Alumnivereniging TU Delft, the universityʼs
campus-wide official alumni organization) donʼt have any precise info on location choices
of TU Delft graduates. Other actors (e.g. Gemeente Delft, Kennisalliantie and Delft
Kennistad) were simply not available to answer requests for information.
The next step, therefore, was to understand how to construct the research process. As a
useful reference for a standard methodology in the investigation of social issues, such as
the location decisions analyzed in this study, Lawrence Neuman (2002) summarized it
quite well when writing the steps to a successful research in social sciences: “First, a
theoretical problem is formulated. Next, an appropriate site and method are selected.
Then, data are collected and analyzed. Finally, the theoretical proposition with which the
research was launched is either challenged or supported.”
Literature on the issue was not lacking. It is mainly from there that a new conceptual
approach was derived, that could accurately represent the location decisions of TU Delft
graduates and the factors that influence them and illustrate them. From this approach,
hypotheses were made on some of the relations within the model between independent
and dependent variables and on some of the possible outcomes of the independent
8
variables. However, once again, there was no data available to validate them, which made
it a necessity to gather relevant data through primary sources.
In the early stages of the research, it was evident that since the location decision rested
within the TU Delft graduates themselves and due to the fact that no other info was
available from other parties (e.g. media, literature, documents), a data-gathering method
that involved those graduates as the target population was the only reasonable option.
Due to the time limits of the project, as well as its need to produce results relevant for the
whole TU Delft graduate population, observation was rapidly discarded as a data-
gathering option: results would have represented a very limited subset of the possible
behaviors/processes that the population as a whole would usually adopt in a location
decision after graduation; furthermore, observation would not have answered the
questions of what graduates perceive as the factors of their location decisions. Also, for
similar issues of representativeness of the whole population, interviews with TU Delft
alumni were also discarded as an option to obtain an answer to the proposed research
questions, being too restricted and qualitative in character. A survey, in the form of an
online questionnaire, was finally chosen as the preferred research method in order to
answer the research questions. In fact, opting for “breadth and generalisation, rather than
for depth and specificity” (Verschuren et al., 1999) entails the use of a survey as a data
gathering method and implies the use of “quantitative processing and analysis of the data”.
Also, all similar studies encountered so far used questionnaires (amongst others: Hansen,
et al., 2003 and Tarant, 2001) in order to know about graduatesʼ choices as well as their
opinions on certain topics related to their decisions.
Finally, the results of the survey provided the answers to the research questions and the
confirmation or rejection of the conceptual approach and associated hypotheses
developed earlier in the research, thus completing the research process.
The next sections will guide the reader into the specifics of those methodological choices
and provide a framework for understanding the following chapters.
9
in the collection of around 150-170 publications that have been analysed in the first phase
of the research, in order to add depth to the research context (Chapter 1) and to the
theoretical background (Chapter 3). References have been screened by both the author
and his thesis committee and come mostly from highly reliable journals and other
publications.
Additional information has been gathered from other stakeholders in the proces (e.g.
Kennisalliantie, Delft Kennisstadt, Gemeente Delft). Data sources in this case include
publications from the relevant organizations. Examples of this are the general statistics
that were obtained from the Dutch national statistics office (CBS), with some useful
information on Delft and its demographic data, and the Kennisalliantie promotional
pamphlet, with an overview on their main organizational interests and objectives.
10
and offered as opt-in choice at the time of graduation. This results in the register
containing information about a smaller group than the actual target group of TU Delft
alumni the survey would be targeted for. For example, graduates from master programmes
in the year 2006 (data for the year 2007 is not available yet) have been 1680 (TU Delft,
2006), whereas only 661 of them chose to register in the alumni register (Stefanie Kirwald,
Marketing Manager Alumni, personal communication) and some of them donʼt provide an
email address or donʼt keep their contact info up to date along the years.
The reasons for this situation are unknown and havenʼt been researched thoroughly yet,
but they constitute nevertheless a bias towards the final results of the survey: in fact, the
group of TU Delft graduates who chose to enter their information in the alumni register
might have particular characteristics that would have an influence in their location decision.
For example, people willing to stay in the Netherlands for their careers might more likely
apply to the register in order to find meaningful local connections and networking
opportunities that they would receive by virtue of being in the alumni register (e.g.
newsletters, invitation to events); non-Dutch students, instead, might not apply to the
register because they already decided to go back to their home countries anyway and they
donʼt see the added value of being in the register in the first place. Nevertheless, due to
the availability of the alumni register as the one and only pool of respondents for the
survey, understanding the reasons behind the bias is out of the scope of the current
research and a possible future research might consider complementing data from the
alumni register with other data about graduates who chose not to apply for it, and
subsequently check whether the aggregated results differ significantly from the results
presented in this research.
Other biases were present in the selection of the sample: for example, a reasonable
expectation is that respondents from 2002 suffer memory and history bias, also referred to
as “cognitive dissonance” (Oswald et al., 2002); people might not remember their choices
or the reasons behind them, as well as interpret them differently due to the fact that a lot of
time passed since then and they might consider their ideas or opinions differently from
what they thought at the time. Anther bias was constituted by a misunderstanding with the
TU Delft Alumni Association, following which many of the TU Delft graduates who
relocated abroad after their studies were not included in the surveyed sample.
Finally, a bias was expected considering the fact that several members of the target
population might not respond to the survey, for various reasons: wrong contact
information, lack of interest in the issue, lack of motivation to contribute to the study, lack
of time. The conclusions of the study will try to reflect these and the other biases coming
from the choice of sample population as well as the particular approach and associated
simplifications that have been put to the test.
Once a target population had been selected, a choice had to be made on which subgroup
of the target group would be involved in the research. A few requirements were necessary
for the subgroup to reflect characteristics of the general population: (1) personal and
educational characteristics of graduates from that subgroup should reflect the variety of
personal and educational characteristics present in any given year at TU Delft; (2)
11
graduates of the subgroup should be ideally closer in time to the present, in order for the
results to constitute a good predictor of future location decisions. Therefore, two
subgroups were selected: all graduates of TU Delft in 2007 and all graduates of TU Delft in
2002. Selecting all kind of graduates (i.e. male, female; Dutch or non-Dutch; etc.) coming
from all possible educational backgrounds (i.e. all educational programmes) was a
necessary condition for having results significantly representative of the general
population. The choice of selecting the cohort of 2007 followed from the requirement of
having a recent sample, whereas the choice of adding the cohort of 2002 was made with a
possible comparison with the two groups in mind, as already experimented in previous
research (Hansen et al., 2003). Finally, only Master of Science graduates for the year 2007
were selected: graduates in 2002 had 5 years of education as their only educational
option, due to the difference in structure of educational systems from 2002 to 2007, and
the necessity to have comparable samples made it obvious to discard the group of
graduates of Bachelors of Science in the year 2007, since they wouldnʼt have any 2002-
equivalent to be compared to.
12
The second section was aimed at understanding the perception of the importance of
certain factors in graduatesʼ decisions in where to locate next, as well as how satisfied
they were with the same factors when considering the city of Delft. These questions were
modeled as Likert scales, as standard methodologies on the subject of evaluating
perceptions suggest (Churchill, 1991).
The third section was aimed at understanding the decision-making process of TU Delft
graduates. It was mostly designed as a set of open questions, in order to give the
respondents the flexibility of a nuanced answer, that would be used to complement the
quantitative data gathered in the other sections.
Sections four and five were aimed at knowing where did the graduates locate for a job and
for living after graduation. Also, additional information about their job and housing (e.g.
sector of occupation, type of housing) were gathered, in order to try and link those
characteristics with the outcomes of the graduatesʼ decisions.
Finally, section six concluded the survey and invited the respondent for comments,
entering their email address and asking whether theyʼd be willing to receive more news
about the study.
Choosing the appropriate scales and measures for each one of the questions in the
questionnaire was done by mostly relying on existing questionnaires used in similar
research (amongst others: Hansen, et al., 2003 and Tarant, 2001) and by verifying
reliability and validity of the questions. Regarding reliability, two important indicators are
often mentioned: test-retest reliability (also referred to as ʻtemporal stabilityʼ) and internal
consistency. Temporal stability is, as already mentioned, an issue when it comes to
graduates of the 2002 sample, but it shouldnʼt be a problem for the cohort of 2007. Internal
consistency has been checked by using Cronbachʼs coefficient alpha on the questionsʼ
scales and making sure that the values of the coefficient are close to 1. Validity, instead,
refers to the the accuracy of the measurements; it has been tested mainly through face
validity (test survey sample and thesis committee) and by carefully wording the
questionnaire as suggested by relevant literature (e.g. Churchill, 1991).
Once completed, a test survey based on the questionnaire that had been developed ran
for 5 days, targeted to fellow class mates in the Engineering & Policy Analysis master
programme. The test survey provided the possibility to express comments on each one of
the surveyʼs sections, as well as a general comment on the survey. Through the feedback
of my class mates and the thesis committee, an improved and final version of the
questionnaire was published online on the authorʼs website, using the open-source
software LimeSurvey.
The link to the survey was sent by the TU Delft Alumni Association to the selected sample
of alumni who had a valid email address, together with a short message that explained the
character of the questionnaire and would invite people to fill it in in the next 2 weeks.
However, due to the initial lack of responses, the period in which the survey was active
finally resulted in 18 days (28th April - 16th May). Participation was encouraged by providing
a lottery prize of 3 gift vouchers of 25€ each on bol.com, randomly assigned among those
who would complete the survey.
13
2.5 Survey analysis
The phases of survey analysis were:
• importing the data into SPSS
• cleaning the data
• obtaining descriptive statistics of the data
• performing statistical analysis on the data
The survey results were imported in the program SPSS, which was used to have an
overview of the survey results and to perform the various statistical analyses. First of all,
only the complete answers to the survey were considered for further elaboration: in fact
incomplete answers didnʼt allow to draw any significant conclusion because of their
incompleteness. Most of them stopped at the very first questions about gender and
country of birth.
Then, the data entered into the program was cleaned from inconsistencies. The responses
to a certain question were given the same format (e.g. somebody entered the full date
when the questionnaire asked for the year only, so the year was extracted from the full
date). Also, some inconsistencies were most certainly typos or simple mistakes in
selecting the correct options. For example, somebody indicating the full address of work
and housing in the Netherlands are expected to indicate ʻNetherlandsʼ as country of first
employment, but a respondent instead selected ʻNepalʼ (the country right under
ʻNetherlandsʼ in the list of countries).
The next step was getting descriptive statistics from the available data. The objective with
regards to that was to provide a picture of the sample, as well as their overall preferences
in terms of choices, as well as behaviours in deciding a place to live and work in. Maps
were generated, where each respondent would be a dot on the map and the maps would
show where they had lived before coming to Delft, during their studies and where did they
work and live after them. The maps were drawn through the help of a list of Dutch post
codes found on the Internet, as well as the mapping capabilities of Google Maps.
Finally, multivariate analysis provided an even richer picture, by highlighting which
particular subgroups of the sample showed particular features (e.g. movers or stayers) or
displayed certain perceptions. This was accomplished by understanding which factors
were more likely to have an influence in the final variable (movers or stayers) through a
series of bivariate analyses; subsequently, these factors were grouped together in a
multivariate analysis that yielded the correlations explaining which of them is a stronger
influence.
The methods used for multivariate analysis were logistical and linear regression, due to
the, respectively, categorical and numerical nature of the dependent variable. All
conditions for using those methods were checked before proceeding.
Throughout the process, the term ʻmoversʼ and ʻstayersʼ were adopted, which described
people staying in or moving away from Delft for first employment or residential reasons.
Additional indicators were produced, in order to show not only whether respondents were
movers or stayers with regards to Delft proper (postcodes from 2600 to 2629), but with
14
regards to surrounding towns (e.g. Den Hoorn, Pijnacker) (mention Haaglanden here) and
with regards to the South Holland province. These three indicators relate to the interest of
not only the city of Delft, but of other local and regional actors, as described in Chapter 1.
The results of the statistical analysis were complemented by the comments left by the
respondents at the end of the survey, as well as the answers to the open questions in the
section regarding their decision-making process, thus completing the analysis of the data
and providing an answer to the research questions of this study.
15
3 Theoretical background
This chapter relies on previous studies on and around the subject of this report. It will first
outline the existing theories and models found in literature, along with their possible
contribution to the current research (Section 3.1) and it will then present and describe a
new conceptual approach developed by the author, based on the integration and
interpretation of those theories and models (Section 3.2). Finally, the new approach will
allow the formulation of several hypotheses (Section 3.3) that will be tested through the
survey.
16
connections that increase its competitiveness by exposing it to new markets, new sources
of knowledge and innovation and in general better networking opportunities. A similar
observation can be made of individuals, whose location decisions, as will be made clear by
the referenced literature, can also be considered in terms of resources (e.g. a high quality
living environment, job opportunities) and connectiveness (e.g. a vibrant social life, links
with friends and family).
When considering an individualʼs location decision, tt the very basic level, the all-
encompassing factor is considered to be utility; the graduate, as in general any person
facing that decision “will choose a residential location by weighing the attributes of each
available alternative - accessibility of workplace, shopping and schools; quality of [...] and
so forth; and picking the alternative that maximizes utility” (McFadden, 1977). However,
theories involving a rational choice are far from popular in most recent literature: first of all,
the information the choice is based on is considered often to be imperfect (van Ommeren,
1997) and the moving/switching costs involved in the process are often excluded from the
balance, too. Finally, when it comes to motivations, thereʼs a “difference between content
(person, physical environment, social environment, other external or societal.) and level of
need (basic needs, anticipated basic needs, higher-order needs)” (Oswald et al., 2002):
not only tangible and visible characteristics of the two or more locations presented as
alternatives are to be considered, but also the relation of those locations with the needs,
attitudes and behaviors of the person. Higher-order needs (Oswald et al., 2002) like need
for independence (in the case of graduates, perhaps independence from parents) and the
need to change are factors that enter the equation when deciding on a future residential
and job location. Anticipated needs also play a role, especially in the case of graduates,
where a change in needs might occur following the future expected condition of being able
to earn a salary (and being therefore entitled to a greater financial independence and
purchasing power). Models that highlight these needs usually descend from the
“Complementary-Congruence Model (Carp & Carp, 1984)” (as noted by Oswald et al.,
2002).
On another level, the factors most commonly listed are:
- life cycle circumstances (Oswald et al., 2002; Clark et al., 1999)
- safety place (Oswald et al., 2002)
- pull from other cities or push from the current one [as in (Oswald et al., 2002)]
- lifestyle approach (Scheiner et al., 2003)
- journey to work (Kain, 1962)
- mutual influence of firm and household location decisions (do jobs follow people or the
other way around?) (Deitz, 1998)
The timeline of such a decision varies, in some cases years (Oswald et al., 2002) but
probably in the case of the graduate the choice is made in the final part of oneʼs studies
(i.e. final year). Also, age is important (Oswald et al., 2002) but in our case itʼs a very
homogeneous variable and same goes for level of education.
17
When it comes to the investigation of first employment location and residential location
choices, similar research (e.g. Hansen et al., 2003; Harren, 1979) stresses the line of
inquiry related to patterns of individual decision-making by the worker. In the case of a
graduate student seeking first employment, the motivations are very varied, as they relate,
for instance, with unemployment reasons (Böheim et al., 1999) or expected income
(Kennan et al., 2003) or others.
Despite the similarities, however, the theme of location and relocation usually presents
characteristics that differ from the features of the current research:
- it is not dealing with first employment location only, but with employment locations at any
point in life (or at least, more than one point in life)
- it is not always dealing with employment location and residential location at the same
time; usually the focus is on one of the two only;
- it is dealing mostly with relocation, therefore examining mostly the cases in which the
decision of individuals or group of individuals resulted in a change from the current/
previous location for work and housing; factors and decisions that drive people to stay
are rarely explicitly mentioned and are often portrayed as simply the negation of the
aspects that drove people to relocate;
3.1.2 Mobility
The second main area of interest that intersects with the topic of this research is mobility.
Contrarily to location and relocation decisions and outcomes, which describe both
behaviors and final states of individuals, mobility is often a characteristic associated to a
system or a population (e.g. mobility of a certain social class, mobility of workers, etc.). It
explains both the degree of change in residential and/or work locations, as well as the
general tendency to travel or commute. It is in many cases associated with the concept of
migration, which will be explained in the following section.
For example, the research of mobility introduces an important factor: young educated
individuals are one of the most mobile sections of the population (Nijkamp et al., 1992).
3.1.3 Migration
The keyword migration is also very significant in previous research on the topic. In fact, it
highlights, with a term that is probably more susceptible to political debate, the fact that
common relocation decisions of a big enough group of people results in a migration flow
from an area to another either in the same country (internal migration) or among different
countries (international migration).
The phenomenon of international migration is often associated with the buzzword brain
drain, which often stands for the migration of skilled workers from a less developed region/
country of the world to another. Its relevance is applicable to our case, due to the fact that
the city of Delft would like to retain as much skilled workers (i.e. TU Delft graduates) to
other cities or countries, therefore trying to avoid a brain drain from Delft.
18
An important focus is interregional labour migration, presented amongst others by Evers et
al. (1984).
19
Figure 3.1: Conceptual approach
20
Figure 3.2: Conceptual approach - constituting elements
3.2.1 Factors
Factors here are listed in two main categories: individual characteristics and external
factors. Individual characteristics elements are what drives most of the research on
mobility, migration and relocation. Differences in gender, age, etc., present different
patterns of mobility and of location decisions.
The other group of factors is the one including external factors, related to living
environment and job market. Most of these factors have been derived from previous
studies, while a few ones have been suggested by common knowledge and by the
feedback of the thesis committee.
3.2.2 Decision
The second element in the model is the decision-making process itself. We can identify the
timeline of the decision, its degree of rationality/irrationality, as well as the search for input
and advice on the assessment of the different external factors already presented, both in
the place where the person il living and in the potential pool of places the person could
relocate to. Finally, the trigger of the decision and the process/procedure are relevant to
know how.
21
Finally, the outcome of the decision is not only where the graduate chooses to live and
work, but also what will his/her mobility pattern be through time (previous choices affect
future ones), how do place of work and place to live relate with each other (a golden rule
reference is that no more than 45 minutes are to be spent commuting from one to the
other) and finally what are the characteristics of the first employment and what are the
ones of the related residential choice.
The elements in this new conceptual approach represent the basis on which the
questionnaire and related survey were designed. In following sections the underlying
assumptions and related simplifications will be explained in detail.
3.2.5 Assumptions
These are some of the assumptions used to generate the components of the conceptual
approach.
Relation firms-graduates
Although the mutual influence between firms and graduates in a certain area is an
important research topic as well as an issue tested by several models/theories, the current
research is not a longitudal study on the topic and therefore considers the job market and
1 In the survey, these definitions are not provided to the respondent, as that would probably only cause
confusion; instead, they have been free to interpret the phrase “where were you living, at the time of your
first employment”, in their most common sense of the phrase.
22
firm situation as an external factor, given as a static input factor to our decision model, with
the feedback from the graduatesʼ population out of the scope of this research.
3.2.6 Operationalization
Having tailored the model to the specific characteristics of the population that will be
surveyed, the next step is the operationalization of the modelʼs elements.
Operationalization will indicate the best way to measure each one of the variables in order
to ensure that statistical analyses can be carried out on them and on their interactions.
Tables 3.1 to 3.6 outline the variables that are present in the model, the method that is
used to measure them (both in terms of level of measurement and possible answers to the
question) and the related question that was included in the online questionnaire. For the
full questionnaire, please refer to Appendix B.
Variable Measurement
Gender nominal
Nationality nominal
23
Variable Measurement
Ambition nominal
Table 3.1: Operationalization of individual characteristics
Variable Measurement
24
Variable Measurement
Variable Measurement
Variable Measurement
occupation nominal
sector nominal
25
Variable Measurement
Variable Measurement
3.3 Hypotheses
The operationalization of the conceptual approach allows to test some basic hypotheses
through the results of the survey. These hypotheses have been derived either as already
existing hypotheses in previous theories and models, or have been drawn from common
sense notions that have emerged throughout my research.
A multitude of hypotheses could be generated from relevant literature and common sense,
due to the fact that countless combinations of independent variables (factors and decision-
making process elements) can be linked to countless combinations of dependent factors
(outcomes of decision-making processes); similarly hypotheses can be made on the
values of the variables themselves and on the relation among independent variables as
well as among dependent ones. However, this study will test only a handful (Table 3.7),
due to the pressing time-constraints and in order to keep the focus of the research sharp,
focused on answering the research questions.
26
# Hypothesis Rationale Relation with research
questions
27
4 Results of the empirical study
The chapter deals with the results of the survey and their interpretation in light of the
conceptual approach and associated hypotheses outlined in the previous chapter. It will
first provide an overview (Section 4.1) of the survey results, including descriptive statistics
on the individual characteristics of the survey respondents; then, it will present and
analyze the first employment location and residential location choices of the respondents
(Section 4.2), the factors behind their decisions (Section 4.3), the resulting profiles of
movers and stayers (Section 4.4) and finally illustrate the decision-making processes
described by the respondents (Section 4.5).
4.1 Overview
The invitation to participate in the survey was sent out to a total of 1109 TU Delft
graduates. The TU Delft Alumni office that sent the invitation counted 128 bounces to their
email, indicating that those addresses were not reachable, for various reasons (e.g. “email
address not known”, “undefined reason”, “mailbox full”, etc.). Out of the remaining ones,
as illustrated in Figure 4.1, the survey delivered a total number of 186 responses: 44 of
them were partial answers (most of them stopped at the first couple of questions) and 142
were full answers to the questionnaire. This does not mean that those 142 respondents
filled in all of the questions, but that they went through the 6 steps of the online
questionnaire and finally submitted their answers.
didnʼt respond
partial answers 12%
full answers 13%
email bounces 4%
72%
Over two thirds of the respondents are male (Figure 4.2), reflecting the general trend of the
TU Delft student population. Most of the graduates are either Dutch-born or Dutch citizens
(Figures 4.3-4.4): around 13% of the survey sample is foreign-born, whereas the number
is around 10% for the TU Delft student population.
28
Female
Male 1%
didnʼt specify 28%
71%
Netherlands 118
EU 8
rest of the world 16
Netherlands 129
EU 5
rest of the world 8
At the time of their graduation, most of the respondents lived with other students or young
professionals, as customary for TU Delft students (Figure 4.5).
The majority of the respondents, at the time of their graduation, were either single or,
although having a partner, didnʼt live together with him/her (Figure 4.6).
29
single 56
with partner, not living together 55
with partner, living together 23
married 8
The year of graduation was included in the questionnaire, in order to accomodate for
possible comparisons between the two target groups: graduates of 2002 and graduates of
2007. The results on this variable reflect the expectation of graduates indeed belonging to
those groups, although some outliers exist (Figure 4.7).
2001 2
2002 53
2003 2
2004
2005
2006 2
2007 79
2008 4
Finally, grade averages and thesis grade of the respondents shown in Table 4.1, present a
great variance (especially in the case of the thesis).
Table 4.1: Respondents of the survey - final grade average and thesis grade statistics
30
4.2 Movers and stayers
The respondents of the survey provided data about the place they lived in before coming
to study at TU Delft, during their studies there, as well as where they had their first
employment and related residence after their graduation.
The first element of interest is that, before their studies at TU Delft, respondents lived all
over the Netherlands (Figure 4.8), in all provinces, although most of them came from the
areas where statistics show that thereʼs a higher population (e.g. South Holland, North
Holland).
After graduation, the location decisions of the respondents brought them mostly in South
Holland and neighbouring regions, both for housing (Figure 4.9) and work (Figure 4.10).
The most popular destinations were Delft, Rotterdam, The Hague and Amsterdam.
Itʼs interesting to notice the differences between the places graduates decide to live in and
places they decide to work in: whereas Delft, The Hague, Rotterdam and in general
31
southern destinations are more popular to live in, Amsterdam, Utrecht and in general
northern destinations are favoured as places to work in.
32
Figure 4.10: Places of first employment after graduation from TU Delft
The distributions by relevant region of analysis can be found in Tables 4.2 and 4.3.
33
Delft Elsewhere Total
Also, short-stay housing or rental are the most preferred types of accomodation in the
places graduate choose to live in (Figure 4.11). Some of them own property (perhaps
referring to family property), whereas some others still stay in student accomodations
(perhaps the ones that chose to stay in Delft and either study or work at the university).
short-stay housing/rental 78
own property 31
student accomodation 22
missing 11
For the purpose of testing hypotheses and, in general dependences and correlations, first
of all, a series of dummy variables were created: livedInDelft_before, livedInDelft_during,
livedInDelft_after and workedInDelft_after, respectively indicating whether the respondent
had been living in Delft before his/her studies, during his/her studies, after his/her studies
or working there after his/her studies. Those dummy variables have been assigned the
value 1 if the respondent indicated Delft (postcodes 2600-2629) in the relevant answer to
the questionnaire and 0 otherwise. Due to the nominal nature of those variables, it would
be possible to execute the relevant tests (e.g. Chi-square) and determine any correlation/
dependence among them.
The first hypothesis could then be translated into more formal statistical terms.
34
Hypothesis #1 TU Delft graduates are more likely to relocate after their studies to
another city than to stay in Delft, for either employment or
residence
This hypothesis was easily confirmed, both for residential and first employment locations,
as shown in Tables 4.4-4.5.
35
To support the picture of TU Delft graduates moving away from Delft, to live and work in
other places, are the other descriptive statistics about the types of jobs they get and the
types of accomodations they settle in. What emerges from Figures 4.12-4.15 is that most
of the graduates find a job matching their technical/business/managerial background
acquired at TU Delft, in private companies, most of them with more than 500 employees,
securing temporary contracts as employees. This profile matches the initial impressions
outlined in Chapter 1, where the assumption was made that TU Delft graduates look for
challenging jobs in big corporations in the main cities of the Netherlands (Rotterdam, The
Hague, Amsterdam) in order to boost their opportunities of career development.
yes 120
no 18
missing 4
private 98
public 32
nonprofit 8
missing 4
36
< 50 employees 31
50 - 500 employees 28
> 500 employees 81
missing 2
Hypothesis #2 TU Delft graduates are more likely to relocate after their studies to
another city region than to stay in Haaglanden, for either first
employment or housing
Results, in Table 4.6, show that most graduates (53.9% in the case of residence, 71.0% in
the case of employment) choose to live/work in Haaglanden after having studied in Delft.
This confirms the incapacity of the city region to retain its pool of graduates, at least
considering the ones from TU Delft.
37
4.3 Factors behind graduatesʼ decision
The second important aspect of the survey results is the indication on what are the factors
that graduates perceive as important in their decision on where to locate next. In an
attempt to compile a ranking among them, based on the respondentsʼ stated ratings, a
count was made of the times the importance of a factor was perceived as very important or
fundamental: this count can be found in Figure 4.16, as a percentage of the total number
of valid responses on that factor.
interesting/challenging job 82
opportunity for career development 73
closeness to partner 50
housing opportunities 44
closeness to friends 44
commuting experience to work/study 42
flexible job or working hours 30
social events and night life 29
cost of living 28
safety of the area 25
starting salary 24
proximity to other places of interest 22
outdoor recreation 22
closeness to family 22
quality of public services 21
job benefits 20
cultural attractions 17
proximity to other important employers 15
physical setting 13
job for spouse/partner 12
opportunity to raise a family 11
Figure 4.16: Ranking of factors by importance
The ranking shows two factors that clearly stand above all others: the search for an
interesting and challenging job and the opportunity for career development. Both are
related to work issues and both score very high among all respondents. These results
connect quite clearly with the choices in first employment outlined in the previous section:
TU Delft graduates move to places where they can find a challenging job that opens the
doors for a steady career development and, apparently, they canʼt find that in Delft, so they
38
move to main economic centres of the Netherlands (Rotterdam, The Hague, Amsterdam)
instead.
The second batch rates moderately high among respondents. First of all, graduates donʼt
want to miss the connections with the ones they most probably got to know during their
studies in Delft (friends, partner); this element is probably the emotional balance needed to
contrast the strong commitment to enter a challenging career. Then, housing and
commuting experience constitute the practical element in the equation: respondents still
want to be living with reasonable standards and not spend too much time traveling from
their house to work and back.
The third group of factors, from flexible job or working hours to the opportunity to raise a
family, score quite low among respondents. The interesting note, however, is that the
element of finding a job for a spouse/partner is the second-lowest important factor in the
ranking, whereas being together with the spouse/partner is among the highest. This
certainly shows that even for the dearest ones, the degree of independence in living the
life one wants is pretty high among the target population.
When it comes to factors ranking, the second hypothesis of this study comes into play:
Clearly, the top factors are job-related ones and not ones relating to the living environment,
but housing opportunities are still a fundamental factor (44% of the respondents consider it
very important or more) and, as the data displayed in Figure 4.17 shows, only 38.5% of the
respondents are satisfied with it in Delft.
When we compare the importance of these factors with the degree of satisfaction in them
when it comes to the time the graduates lived in Delft, other interesting results emerge.
First of all, as shown in Figure 4.17, when it comes to the living environment Delft rates
quite high on many fronts. Although in a couple of occasions (commuting experience and
closeness to friends) the satisfaction strikes factors that are also important in a graduate
location decision, in other cases, where more than 50% of the graduates have declared to
be moderately satisfied or very satisfied, that satisfaction strikes factors that are not very
important in a location decision. For example, although Delft rates high (more than 70%)
on public services, proximity of other places of interest and safety of the area, those
factors are considered very important or fundamental by less than 25% of the
respondents. On the other hand, Delft rates quite poorly (around or under 50% of the
respondents declare to be satisfied) when it comes to important factors such as housing
opportunities and closeness to partner.
39
very important or fundamental (%) moderately satisfied or very satisfied (%)
closeness to partner
housing opportunities
closeness to friends
cost of living
outdoor recreation
closeness to family
cultural attractions
0
.
.
22
45
67
90
%
Figure 4.17: Comparison between importance of living environment factors and the satisfaction of
those factors in Delft
Considering work-related factors instead (Figure 4.18), Delft scores moderately good
(slightly above 50%) on the two most important factors in a location decision: challenging
job and opportunity for career development. Other factors, instead, are considered
satisfactory only by less than half of the respondents.
40
very important or fundamental (%) moderately satisfied or very satisfied (%)
interesting/challenging job
starting salary
job benefits
0
.
.
22
45
67
90
%
Figure 4.18: Comparison between importance of job market factors and satisfaction of those
factors in Delft and surroundings
The analysis so far has shown us that Delft fails to fully satisfy graduates from TU Delft,
resulting in most of the students to move away from the city for work or living opportunities.
When it comes to housing, opportunities are limited and when it comes to work, more
challenging jobs and more career development opportunities are perceived to be found
elsewhere (Amsterdam, The Hague, Rotterdam).
41
Table 4.7: Results of dependence test between livedInDelft_before and livedInDelft_after
Hypothesis #4 Graduates with higher grades (both grade average and thesis
grade) tend to relocate more than graduates with lower grades.
This hypothesis can be verified through a logistic regression model (Tables 4.9 and 4.10),
due to the fact that both variables final grade average and final thesis grade have a
numerical level of measurement and the dependent variable has a nominal level of
measurement.
42
The results show that whereas for living in Delft thereʼs no significant dependence, working
in Delft is very correlated with the two variables, especially with the final thesis grade. This
means that people with higher final thesis grades are more likely to stay in Delft for work.
The result makes sense, by comparing it to the answers of respondents who obtained
higher grades: in most of the cases their career has in fact been a PhD at TU Delft.
Table 4.9: Results of dependence test between final grade average, final thesis grade and
livedInDelft_after
Table 4.10: Results of dependence test between final grade average, final thesis grade and
workedInDelft_after
Finally, the last comparison is between graduates in 2002 and graduates in 2007 (Tables
4.11 and 4.12). No significant correlation can be found, meaning that no particular
differences in moving vs. staying can be seen between the two target populations.
43
Table 4.11: Results of dependence test between the approximate year of graduation and
livedInDelft_after
Table 4.12: Results of dependence test between the approximate year of graduation and
workedInDelft_after
Although many times indicative, bivariate analysis doesnʼt allow a full understanding of the
influence a certain variable has on one of the previously specified outcome indicators
(dichotomous variables indicating whether the respondent lived/worked or not in Delft/
Haaglanden/Randstad before/during/after their studies at TU Delft); in fact such influence
might be covered by the effect of other variables. This is the reason for a more reliable
model to be developed, that can account for a more precise analysis and prediction of the
location decisions of TU Delft graduates and the factors that influence them.
Two types of models have been developed throughout the study: the first kind has as
outcome indicators dichotomous variables such as the previously mentioned
44
livedInDelft_after or workedInDelft_after; the second kind has instead numerical
continuous variables as the main outcome indicators. In the second type of models, the
distance between Delft and the places of first employment and residence was calculated
and used as a value in those indicators.
The two types of models allowed for different methods of statistical analysis: in the first
kind, a logistic regression method was used, due to the categorical nature of the
independent variable and the mix of numerical and categorical variables that compose the
model; in the second kind, linear regression was used to develop a model, due to the
numerical nature of the independent variable and after carefully reducing the others to
dichotomous categorical variables in order to ensure proper interpretation of the model.
In both models, the SPSS Enter method has been used, that forced all variables at once in
the same model. Also, in the logistic regression, categorical variables with more than 2
values were reduced to more variables with 2 values automatically by the software.
A first model of the independent variable livedInDelft_after can be seen in Table 4.13. The
variables entered in the model are either the ones that provided a more significant
dependence with livedInDelft_after in previous bivariate analysis, or that are conceptually
linked to the location decision, due to the models and theories presented in Chapter 3.
This accounts for both variables that, as other results have shown, are highly likely to be
influential on the independent variable, as well as variables that were not likely influential
according to the results, but are likely to have their effect covered by other factors and
their inclusion in the model makes sense to the light of previous theoretical work. As an
example, satisfaction with commuting experience and satisfaction with career development
opportunities were included, respectively, in the models about living in Delft and working in
Delft, due to the high importance they scored among the respondents in determining a
place to live or work in; on the other hand, although results show that gender doesnʼt score
high on significance in bivariate analyses, there are reasons coming from literature that
make it a good candidate for inclusion in our models.
The significance of the various elements in the model is quite high, with the most
significant results being above 20%. The study has a significance level of 10% and
therefore no factor can be concluded to have a significant influence on the independent
variable. We can see that marital status and housing opportunities in Delft are the factors
that are most likely influent in the model outcome, perhaps signifying that single graduates
are more likely (positive coefficient) to stay in Delft and graduates who are not satisfied
with housing opportunities in Delft are less likely (negative coefficient) to live there.
However, no statistical significance on these connections can be drawn, at least with the
current sample.
45
Table 4.13: Logistic regression model for the variable livedInDelft_after
When it comes to a place of work, instead, some factors (Table 4.14) are well below the
significance level of 10%: gender, living in Delft before oneʼs studies, marital status, year of
graduation and final thesis grade. Particularly, this model shows that:
• female graduates are more likely (positive coefficient) to find their first job in Delft;
• graduates who lived in Delft before their studies are more likely to work there
afterwards;
• that single graduates are more likely to stay in Delft for first employment;
• that graduates of the cohort of 2007 have been more likely to relocate than
graduates of the 2002 cohort;
• that graduates with a higher final thesis grade are more likely to stay in Delft; this
might be explained by the fact that theyʼre usually offered a PhD position at the
university;
46
Table 4.14: Logistic regression model for the variable workedInDelft_after
Showing a connection between the same factors and the distance from Delft that
graduates decide to relocate to, for work or residence reasons, results in being a harder
challenge. In fact, both models (Tables 4.15 and 4.16) have no coefficients with a
significance below 10%. Similarly to the previous case, though, we can show that there are
variables that could have a significance, which could probably be shown with a more
refined model (e.g. with a larger sample): marital status, satisfaction with closeness to
oneʼs partner and satisfaction related to housing opportunities are the factors closer to a
statistical significance than others, in the model of distance of the graduateʼs place of
residence from Delft; in the other model, referring to the distance of oneʼs first employment
location, the gender, marital status and final grade average are instead the ones closer to
significance (but still above 10%). In both models, there is a constant factor that is close or
over 200, which shows a high significance: this accounts for the fact that in average a
graduate moves from Delft many kilometres away.
47
Table 4.15: Linear regression model for the variable distanceFromDelft_live
Unfortunately, none of the models provides a high explanatory power of either the final
dichotomous location decision (living in Delft or not; working in Delft or not; etc.): in the
case of the logistic regression, the models are able to predict correctly up to 80% of the
cases (compared with a starting 70% of explanatory power they start with); in the case of
the linear regression, this figure is even worse, since only 30-40% of the variance in the
final results is covered by the factors in the model, most of which do not have a significant
influence on the independent variable.
48
However, it is evident that models regarding first employment location decisions, especially
in the logistic regression case, provide more hints as to the factors that influence them,
compared to the models regarding residential location decisions. In other words, it appears
to be easier to analyze and predict first employment location choices of graduates, than to
analyze and predict their residential location choices.
The first employment location model shows that the so-called individual characteristics (i.e.
gender, marital status, year of graduation, final thesis grade) are found to influence the
location decision more significantly than the two factors related to how satisfied the
respondents were with the Delft job market: career development opportunities and finding
an interesting/challenging job. Although these two factors were considered the most
influential in a decision on where to locate for a first employment, they donʼt have a proven
influence on the final first employment location decision. This might relate to the fact that,
as described in the push-pull model in Chapter 3, although Delft satisfies graduates in its
career development opportunities and availability of interesting/challenging jobs, other
cities might simply score higher on those indicators and therefore exert the necessary
attraction to drive people away from a job in Delft.
On the other hand, although not proven to have a significant influence on the residential
location decision, the two factors related to the satisfaction of housing opportunities in Delft
and the possibility to be close to oneʼs partner scored better than the significance levels of
all other satisfaction factors and individual characteristics. In this case, though, housing
opportunities had a slightly negative correlation with the final decision, an element which
can be very hard to explain, but that could perhaps be better understood with a greater
sample and therefore a higher number of cases that the analysis draws upon.
49
first job choice, then choice of housing 78
first housing choice, then choice of job 31
housing and job choices at the same time 30
other 2
missing 1
The decision was not only a result of the balance and rationalization of the various factors
presented in the previous sections, but it often relied on the input of other people. As
shown in Figure 4.20, people emotionally closer to the respondent were more often
consulted than other people in the university environment (classmates, faculty staff,
professors). Relying on advice from oneʼs partner and friends can be connected with the
previous finding of being close to oneʼs partner and friends as some of the most important
factors in the location decisions: these might be perhaps interpreted as consultations
aimed at finding a place to work that is close to oneʼs partner and reasonable close to
oneʼs friends. Finally, media and other sources were rarely considered.
family 48
friends 70
partner 71
children 1
classmates 18
professor/faculty 18
media (TV, news, Internet, ...) 5
50
5 Conclusion
This final chapter will provide the answers to the research questions and evaluate the
achievement of the research objectives (Section 5.1). It will then complement those results
with policy implications relevant to the background of the study (section 5.2) and insights
derived from the statistical analysis that will contribute to the previously presented models
and theories (Section 5.3). Finally, an overview of the strengths and weaknesses of this
study will be given, in order to pave the way to future research on this or similar issues
(Section 5.4).
51
Living Working
The second objective and related research questions required a more in depth analysis of
the factors that influenced those decisions and the processes that generate them, as well
as whether these factors and processes are the same or whether they differ for specific
subgroups of the TU Delft graduates population. This was achieved through three steps:
• first, surveying the graduates on the perceived importance of those factors in their
decision-making process;
• then, through statistical analysis, understanding whether those factors and other
individual characteristics can indeed have a significant influence in these
decisions;
• finally, surveying the graduates on what have been their decision-making process
with regards to first employment location and residential loacation;
Regarding the first point, results show that finding an interesting and challenging job, as
well as the availability of career development opportunities were perceived as the most
influent elements in the decision (respectively, 82% and 73% of the respondents
considered them as very important or more); being close to oneʼs partner (50%) and
friends (44%), as well as housing opportunities (44%) and commuting experience (42%)
were also perceived as major elements in a graduatesʼ decision. This contributed to the
explanation of why most graduates left Delft after graduation (69.5% for living reasons,
83.1% for working reasons), since those factors were some of the ones the respondents
were less satisfied with, when it came to the perception of their experience in Delft; for
example, only 38.5% of the respondents were satisfied by the housing opportunities in the
city.
The statistical analysis that followed showed that housing opportunities, commuting
experience, as well as closeness to oneʼs partner and friends were not found to be
significantly influencing a graduatesʼ decision on where to live next, despite being
52
indicated as somewhat important by respondents themselves, however they showed to be
at least more infuent than other individual characteristics such as previous living location
or final grade average, believed to have an influence in the decision according to previous
studies. The former group of elements, relating to job-market characteristics, was not
found to have a direct influence on the decision of first employment location: statistical
analysis showed that being satisfied with the offer of interesting jobs and career
development opportunities in Delft didnʼt result in more people staying in Delft for work
reasons; this might due to the fact that although Delft can be somewhat satisfying in
certain aspects, other cities exert a bigger attraction and therefore manage to motivate
graduates to relocate for their first employment.
Individual characteristics, however, proved to be more influential in the choice, denoting a
core difference in the profiles of people that choose to live or work somewhere else than in
Delft. Whereas for the residential location decision there was no definite proof that any of
the individual characteristics play a role in the choice, when it comes to first employment
location, gender, marital status, having lived in Delft before the studies, year of graduation
and final thesis grade explained most of the final choices of the respondents, resulting in
the correct estimation or prediction of 70-80% of the existing cases. Changes in such
factors, therefore, can change significantly the outcome of the location decision; for
example, having a very high final thesis grade will often result in an offer of a PhD position
at the university, therefore motivating graduates to stay in the city for work purposes.
Finally, decision processes are mostly taking job choice as the priority, with residential
location choice following accordingly. This lines up with the result of job-related factors
scoring higher on the perceived importance in the graduatesʼ decision. Also, as
demonstrated by the high perceived importance of a partner and friends in a location
decision, oneʼs partner, friends and family are also the major sources of input and advice
when it comes to the graduatesʼ decision-making processes.
53
city: in fact, whereas now Delft is very attractive to students with high academic profile, the
city might want to make itself attractive to other kinds of students, that look for different
employment opportunities and that perhaps want different things from the city, too.
However, it will be up to the stakeholders of the city deciding which direction they want the
city to develop towards, which specialization should the cityʼs economic activities point to,
and accordingly Delft will be able to develop the necessary policies.
On a bigger scale, though, it is worth considering the impact that graduatesʼ decisions
have on not only Delft, but on Haaglanden and on the whole Randstad region. In this case,
there are less reasons to be worried about, since most of the graduates relocate in this
area for living or work reasons after their studies. Considering Haaglanden, the biggest
concentrations of TU Delft graduates are in Den Haag and Delft, whereas very few of them
relocate to the other 7 municipalities of the consortium. A more in depth and focused study
might be able to reveal the reasons why graduates decide to move there and see what are
these places lacking in order to boost the economic potential of the project. Considering
Randstad instead, the graduates are pretty much distributed all over the area, with
concentration on the bigger cities, but with coverage of the whole region.
When it comes to the decision-making process, this study provides some insights for
policy-makers who want to contrast the negative trend of relocation, if present and if it is a
matter of concern for the stakeholders. First of all, the choice of a job is first considered in
a location decision. This means that if the city of Delft or the city region of Haaglanden
could provide a challenging environment where the graduate student would find an
interesting occupation with enough opportunity for career development, there would be a
higher chance for him/her to stay. Also, graduates rely a lot on advice from close ones:
family, friends and partner. This means that big marketing campaigns and strong
promotion through media or university communication channels wonʼt be received often as
good as some word of mouth from, for example, friends who can share with the graduate a
success story about working and/or living in one place or another.
54
suitable first employment opportunity. An investigation of mobility patterns according to
academic performance within the highly skilled student population would be interesting to
perform in order to better understand the patterns that guide students with higher
performance and students with lower performance, possibly leading to a new and more
refined model that links level of education with mobility.
Always relating to mobility, studies show that having been previously mobile before makes
people more likely to be mobile afterwards (at least when theyʼre young). This statement is
indeed confirmed, since many of the students who moved to Delft for their studies, moved
again afterwards.
When it comes to the push-pull model, no real confirmation or rejection can be done, on
the basis of the survey results, since no data regarding the satisfaction of graduates with
their first employment location and residential location choices was gathered. However, the
model is a very good explanation of the research results and is therefore worth
investigating in further research.
55
and economic centres, being in between The Hague and Rotterdam and quite close to
other high-tech economic areas in Haaglanden. In comparison, the other two universities
in the Netherlands (TU Twente, TU Eindhoven) are quite isolated from the Randstad and
this is perhaps why they have resorted and currently resort to more localized economic
areas (e.g. Philips developed a business/research area near Eindhoven; TU Twente has a
strong incubator programme that incentives the creation of start-ups in the region). The
nature of Delft as one of the nodes in the knowledge network of Haaglanden/Randstad
results in its specialization on some particular disciplines and economic activities that are
the strength of its knowledge economy (e.g. hydraulics, aerospace engineering, high tech
research).
The feature of Delft as a node very close to other cities in the Haaglanden/Randstad hubs
provides also certain restrictions in its spatial development. In fact, the area available to
the municipality is quite limited and, even if willing, the city would not be able to
accomodate, year after year, all the graduates who want to live and work there. Other
areas in the region have been growing and are being developed, that either focus on
residence or on work opportunities for graduates coming from Delft and neighbouring
universities in Den Haag or Rotterdam. This results in a natural tendency of Delft
graduates to move, since the constraints of the system are such that Delft wonʼt be able to
fit all needs.
Besides all this, due to the fact that the study utilizes an analytical approach on the
individual decisions of the graduate student, despite the cultural differences that could be
accounted for in several successive surveys, it can be assumed that many of the
concerns, factors and attitudes of the surveyed population can be indeed observed in
other student populations from universities around the world, particularly in countries with
similar kind of developed knowledge economies whose universities are driving the
development of a new knowledge-intensive workforce. As many scholars and experts put
it, the new class of creative young knowledge workers has more similarities across country
borders than with other classes in their own country. That suggests that the applicability of
this studyʼs methodology can be easily done, with the modification of a few factors that
account for cultural elements, as well as for specific factors that are relevant to the place
the study is being made.
56
References
Journal articles
Bartik T. J. (1985), ʻBusiness Location Decisions in the United States: Estimates of the
Effects of Unionization, Taxes and Other Characteristics of Statesʼ, Journal of
Business & Economic Statistics, vol. 3, no. 1, pp. 14-22
Benneworth P., Charles D. (2005), ʻUniversity spin-off policies and economic development
in Less successful regions: Learning from two decades of policy practiceʼ, European
Planning Studies, vol. 13, no. 4, pp. 537-557
Benneworth P., Hospers G.-J. (2007), ʻThe new economic geography of old industrial
regions: universities as global - local pipelinesʼ, Environment and Planning C:
Government and Policy, vol. 25, pp. 779-802
Cappellin R. (2007), ʻThe Territorial Dimension of the Knowledge Economy: Collective
Learning, Spatial Changes and Regional and Urban Policiesʼ, American Behavioral
Scientist, vol. 50, no. 7, pp. 897-921
Clark W. A. V., Withers S. D. (1999), ʻChanging jobs and changing houses: mobility
outcomes of employment transitionsʼ, Journal of Regional Science, vol. 39, no. 4,
pp. 653-673
Conceição P., Heitor M. V (1999), ʻOn the role of the university in the knowledge
economyʼ, Science and Public Policy, vol. 26, no. 1, pp. 37-51
Deitz R. (1998), ʻA joint model of residential and employment location in urban areasʼ,
Journal of Urban Economics, vol. 44, pp. 197-215
Devereux M. P., Griffith R. (2003), ʻEvaluating Tax Policy for Location Decisionsʼ,
International Tax and Public Finance, vol. 10, pp. 107-126
Eliasson K., Lindgren U., Westerlund O. (2003), ʻGeographical Labour Mobility: Migration
or Commuting?ʼ, Regional Studies, vol. 37, no. 8, pp. 827-837
Evers G.H.M., van der Veen A. (1984), ʻA Simultaneous Non-Linear Model for Labour
Migration and Commutingʼ, Regional Studies, vol. 19, no. 3, pp. 217-229
Godin B. (2006), ʻThe Knowledge-Based Economy: Conceptual Framework or
Buzzword?ʼ, Journal on Technology Transfer, vol. 31, pp. 17-30
Goldstein H., Renault C. (2004), ʻContributions of Universities to Regional Economic
Development: A Quasi-Experimental Approachʼ, Regional Studies, vol. 38, n. 7, pp.
733-746
Hansen S. B., Ban C., Huggins L. (2003), ʻExplaining the “Brain Drain” from Older
Industrial Cities: The Pittsburgh Regionʼ, Economic Development Quarterly, vol. 17,
no. 2, pp. 132-147
Harloe M., Perry B. (2004), ʻUniversities, Localities and Regional Development: The
Emergence of the ʻMode 2ʼ University?ʼ, International Journal of Urban and
Regional Research, vol. 28, no. 1, pp. 212-23
Harren V. A. (1979), ʻA Model of Career Decision Making for College Studentsʼ, Journal of
Vocational Behavior, vol. 14, pp. 119-133
57
Head K., Ries J., Swenson D. (1995), ʻAgglomeration benefits and location choice:
Evidence from Japanese manufacturing investments in the United Statesʼ, Journal
of International Economics, vol. 38, pp. 223-247
Kain J. F. (1962), ʻThe journey-to-work as a determinant of residential locationʼ, Papers
and proceedings of the Regional Science Association, vol. 9
McFadden D. (1977), ʻModelling the choice of residential locationʼ, Cowles Foundation
Discussion paper no. 477, Cowles Foundation for Research in Economics at Yale
University
Oswald F., Schilling O., Wahl H.-W., Gäng K. (2002), ʻTrouble in paradise? Reasons to
relocate and objective environmental changes among well-off older adultsʼ, Journal
of Environmental Psychology, vol. 22, pp. 273-288
Raspe O., van Oort F. (2006), ʻThe Knowledge Economy and Urban Economic Growthʼ,
European Planning Studies, vol. 14, no. 9, pp. 1209-1234
Scheiner J., Kasper B. (2003), ʻLifestyles, choice of housing location and daily mobility: the
lifestyle approach in the context of spatial mobility and planningʼ, International
Social Science Journal, vol. 55, issue 176, pp. 319-332
van der Meer E. (1997), ʻThe University as a local source of expertiseʼ, GeoJournal, vol.
41, no. 4, pp. 359-367
van Geenhuizen M., Nijkamp P., Rijckenberg H. (1997), ʻUniversities and knowledge-
based economic growth: the case of Delft (NL)ʼ, GeoJournal, vol. 41, no. 4, pp.
369-377
van Ommeren J., Rietveld P., Nijkamp P. (1997), ʻCommuting: in search of jobs and
residencesʼ, Journal of Urban Economics, vol. 42, pp. 402-421
Other publications
Böheim R., Taylor M. (1999), ʻResidential mobility, housing tenure and the labour market in
Britainʼ, Institute for Social and Economic Research and Institute for Labour
Research, University of Essex
Bolz U., Ford A., Gourley M., Magee C., Castilla Porquet M., Rakel J., Radovanovic D.,
Sieverdink A., Sivertsen T., Sturesson J., Teunisse P., Toussing J. (2005) , ʻCities of
the future - global competition, local leadership*ʼ, PricewaterhouseCoopers
Churchill G. A. Jr. (1991), ʻMarketing research - methodological foundationsʼ, 5th ed., The
Dryden Press International Edition
European Council (2000), ʻPresidency Conclusionsʼ, Lisbon European Council, 23-24
March 2000, available at http://europa.eu/european-council/index_en.htm
Florax R. (1992), ʻThe university: a regional booster? Economic impacts of academic
knowledge infrastructureʼ, Ashgate
Florida R., Gates G., Knudsen B., Stolarick K. (2006), ʻThe University and the Creative
Economyʼ
Jones A., Williams L., Lee N., Coats D., Cowling M. (2006), ʻIdeopolis: Knowledge
City-Regionsʼ, The Work Foundation
58
Kennan J., Walker J. R. (2003), ʻThe effect of expected income on individual migration
decisionsʼ, National Bureau of Economic Research
Lawrence Neuman W. (2002), ʻSocial Research Methods: Qualitative and Quantitative
Approachesʼ, 5th ed., Allyn & Bacon
Mellander C., Florida R. (2006), ʻThe Creative Class or Human Capital? Explaining
regional development in Swedenʼ, The Martin Prosperity Institute
Nijkamp P., van Wissen L., Rima A. (1992), ʻA Household Life Cycle Model for Residential
Relocation Behaviorʼ, Serie Research Memoranda, Vrije Universiteit Amsterdam
OECD (2006), ʻSupporting the Contribution of Higher Education Institutions to Regional
Development - Peer review report: Twente in the Netherlandsʼ
Stadgewest Haaglanden (2008A), ʻRegionaal Structuurplan Haaglanden 2020ʼ
Tarant S. A. (2001), ʻPredicting Retention of Recent College Graduates in Science and
Engineering: Implications for States and Organizational Recruiting Practicesʼ, MSc
Thesis, North Carolina State University
TU Delft (2006), ʻTU Delft Highlights 2006ʼ
Varga A. (1998), ʻUniversity research and regional innovation: a spatial econometric
analysis of academic technology transfersʼ, Kluwer Academic Publishers
Verschuren P., Doorewaard H. (1999), ʻDesigning a Research Projectʼ, Lemma
World Bank (1998), World Development Report 1998, ʻKnowledge for Developmentʼ
59
Appendix A: Keywords
The following keywords have been used to find published materials on the study issue, as
described in Section 2.2:
relocation decision
relocation model
relocation decision model
relocation decision-making
relocation decision-making model
relocation factors
relocation reasons
relocation
graduate* relocation
alumni relocation
youth relocation
relocation young
residential relocation
relocation university
relocation college
determinants relocation
motivation relocation
relocation graduation
career relocation
residential mobility
housing mobility
housing job mobility
geographical mobility
mobility university
mobility college
determinants mobility
motivation mobility
mobility graduation
career mobility
migration
migration decision
migration model
migration decision model
migration decision-making
migration decision-making model
migration factors
60
migration reasons
graduate migration
college migration
university migration
migration graduates
migration alumni
migration youth
migration young
skilled migration
residential migration
housing migration
housing job migration
internal migration
determinants migration
motivation migration
migration graduation
graduation migrants
graduate migrants
knowledge migrant*
graduate retention
graduation retention
workplace location
workplace location decision
workplace location choice
job location
job location decision
job location choice
career location
career location decision
career location choice
employment location
employment location decision
employment location choice
housing location
housing location decision
housing location choice
residential location
residential location decision
residential location choice
job housing location
workplace housing location
career housing location
61
employment housing location
job residential location
workplace residential location
career residential location
employment residential location
career migration
knowledge workers
graduate retention
knowledge economy
knowledge economy geograph*
knowledge economy region*
location choice
economic geography
migration knowledge
talent attraction
talent retention
graduate attraction
talent retention
graduate retention
alumni retention
student retention
talent involvement
graduate involvement
alumni involvement
student involvement
urban economy
regional economy
local human capital
human capital flows
brain drain
local knowledge
urban knowledge economy
regional knowledge economy
urban knowledge economy innovation
regional knowledge economy innovation
tiebout
tiebout relocation
62
Appendix B: Survey transcript
This is a page-by-page transcript of the online survey delivered to TU Delft graduates.
When available, the possible answers are indicated in square brackets after each
question. Comments meant to help the respondent are preceded by the symbol “[?]”.
PAGE 1
Welcome!
My name is Nicolò Wojewoda and this short survey is part of my graduation project on the
issue of first employment location and housing location decisions of TU Delft alumni after
graduation. Iʼm working on the project through TU Delft and the support of other
stakeholders, under the supervision of prof. Marina van Geenhuizen (Professor of
Innovation and Innovation Policy in the Urban Economy at the Faculty of Technology,
Policy and Management).
The questionnaire is divided in five main parts: your individual characteristics, your
perceptions of Delft as a living and working environment, your decision on whether to stay
in or leave Delft after graduation, your first employment after graduation and your first
housing after graduation.
If you'll leave your email address at the end of the survey (optional - you can also decide
not to do so), I'll make sure to contact you personally and let you know the results of my
research. Also, youʼll qualify for a prize (3 bol.com gift-vouchers of 25 EUR each),
randomly assigned among the provided email addresses. No link will be established
between your answers and your email address: the survey is completely anonymous.
63
Nicolò Wojewoda
MSc in Engineering & Policy Analysis
Faculty of Technology, Policy & Management
Delft University of Technology
[Next >>]
PAGE 2
Individual characteristics
This part of the survey will allow the research to understand what characteristics do the
"movers" and "stayers" possess (e.g. are women more likely to move out of Delft than
men?)
64
[<< Previous] [Next >>]
PAGE 3
1. At the time of your graduation, how satisfied were you by the following elements of
Delft? [from 1 = “not satisfied at all” to 5 = “very satisfied” + 0 = “no opinion at the time/not
applicable”]
a. closeness to friends
b. closeness to family
c. closeness to partner
d. cultural attractions
e. outdoor recreation
f. social events and night life
g. cost of living
h. housing opportunities
i. physical setting (i.e. geography, climate)
j. commuting experience to work/study
k. proximity of other places of interest
l. quality of public services
m.opportunity to raise a family
n. safety of the area (e.g. in terms of crime)
2. At the time of your graduation, how important were the following elements, in your
decision on where to locate next? [from 1 = “not important at all” to 5 = “fundamental” + 0 =
“no opinion at the time/not applicable”]
a. closeness to friends
b. closeness to family
c. closeness to partner
d. cultural attractions
e. outdoor recreation
f. social events and night life
g. cost of living
h. housing opportunities
i. physical setting (i.e. geography, climate)
j. commuting experience to work/study
k. proximity of other places of interest
65
l. quality of public services
m.opportunity to raise a family
n. safety of the area (e.g. in terms of crime)
3. At the time of your graduation, how satisfied were you by the job market in Delft and
neighbouring areas? [from 1 = “not satisfied at all” to 5 = “very satisfied” + 0 = “no opinion
at the time/not applicable”]
a. starting salary
b. job benefits (e.g. retirement plan, travel assistance, etc.)
c. job for spouse/partner
d. interesting/challenging job
e. opportunity for career development
f. flexible job or working hours
g. proximity to other important employers
[?] Neighbouring areas are municipalities in close proximity of Delft (e.g. Pijnacker,
Rijswijk).
4. At the time of your graduation, how important were the following factors in choosing
your first employment? [from 1 = “not important at all” to 5 = “fundamental” + 0 = “no
opinion at the time/not applicable”]
a. starting salary
b. job benefits (e.g. retirement plan, travel assistance, etc.)
c. job for spouse/partner
d. interesting/challenging job
e. opportunity for career development
f. flexible job or working hours
g. proximity to other important employers
PAGE 4
Decision
The purpose of this section is to better understand the process of your decision-making
regarding your first employment and related housing.
66
3. Who provided major input/guidance in your choice? Please select all that apply. [friends;
family; partner or spouse; children; classmates at TU Delft; professor/faculty staff; media
(TV, newspapers, Internet, ...); other]
4. What process or procedure did you follow in order to decide on your future first
employment and related housing? _____
[?] What was the timeline of the decision? Was it a "rational" or "emotional" one? What
were the steps involved?
PAGE 5
First employment
In this section I'll ask you about the characteristics of your first job after graduation.
1. What type of job was your first one after graduation? [employee, permanent contract;
employee, temporary contract; employee, other type of contract; self-employed,
freelance; self-employed, owner/manager; Iʼm still unemployed]
2. Did your job match your recenty acquired qualification by TU Delft? [yes; no]
[?] e.g. a computer engineer working for MicroSoft would answer YES, whereas a civil
engineer working at McDonald's would answer NO
3. What was your occupation in your first employment? [list derived from the 2002
Standard Occupational Classification (SOC), by the US Department of Labor, Bureau of
Labor Statistics2]
4. In which sector were you employed? [private; public; nonprofit]
5. In what country was your first job located? [dropdown list with all countries in the world]
6. If your first job was in the Netherlands, please indicate the postal code of your work
location
numbers (e.g. 2625) _____ letters (e.g. JM) _____
7. What was the size of your employer? [< 50 employees; 50 - 500 employees; > 500
employees]
PAGE 6
Housing
In this section I'll ask you about the characteristics of your housing at the time of your first
job after graduation.
2 http://www.bls.gov/soc/
67
1. In what country did you live in, during your first job after graduation? [dropdown list with
all countries in the world]
2. If you lived in the Netherlands, please indicate the postal code of your housing location.
numbers (e.g. 2625) _____ letters (e.g. JM) _____
3. What kind of accomodation were you staying in? [student accomodation; short-stay
housing/rental; own property; other]
PAGE 7
Conclusion
1. Before submitting your answers, if you would like to be kept informed about the results
of this research, please enter your email address. Your contact info will be used
exclusively for informing you about my graduation project and/or for entering the lottery
assigning 3 bol.com gift-vouchers for 25 EUR each; at any time youʼll be free to contact
me (nicolo.wojewoda@gmail.com) and ask to be removed from my contact list.
_____
2. Would you be willing to be contacted again, regarding this research? (e.g. receiving the
final report) [yes; no]
3. Would you like your email address to enter the lottery that is going to randomly award 3
bol.com gift-vouchers of 25 EUR each? [yes; no]
4. If you want, please report here any comments you would like to make about this study
or the survey.
_____
PAGE 8
[Exit]
68