Content uploaded by Conny Roggeband
Author content
All content in this area was uploaded by Conny Roggeband on Apr 24, 2015
Content may be subject to copyright.
VRIJE UNIVERSITEIT
Framing Immigration and Integration
Facts, Parliament, Media and Anti-Immigrant Party Support in the Netherlands
ACADEMISCH PROEFSCHRIFT
ter verkrijging van de graad Doctor aan
de Vrije Universiteit Amsterdam,
op gezag van de rector magnificus
prof.dr. L.M. Bouter,
in het openbaar te verdedigen
ten overstaan van de promotiecommissie
van de faculteit der Sociale Wetenschappen
op vrijdag 26 oktober 2007 om 10.45 uur
in de aula van de universiteit,
De Boelelaan 1105
door
Rens Vliegenthart
geboren te Oudewater
promotor: prof.dr. P.G. Klandermans
copromotoren: dr. D. Oegema
dr. C.M. Roggeband
ISBN: 978-90-9022279-0
Uitgave: Rens Vliegenthart
Omslagontwerp: S. van der Ploeg, Room for
ID's, Nieuwegein
Foto's omslag: M. Meijer, M. Nuytemans,
K. Van Berendoncks
Druk: Ridderprint Offsetdrukkerij BV,
Ridderkerk
Deze dissertatie is de 15e in de Dissertatie Serie van de
Faculteit der Sociale Wetenschappen,
Vrije Universiteit, Amsterdam
©2007 Rens Vliegenthart. Niets uit deze uitgave mag worden
verveelvoudigd, opgeslagen in een geautomatiseerd
gegevensbestand of openbaar gemaakt, in enige vorm of op enige
wijze, zonder voorafgaande schriftelijke toestemming van
rechthebbende: Rens Vliegenthart, www.rensvliegenthart.com.
iii
Voorwoord/Foreword
Vaak wordt beweerd dat het werk als promovendus een eenzame aangelegenheid is.
Natuurlijk heb ik de afgelopen jaren vele uren in afzondering achter mijn computer gezeten,
maar ik heb toch ook vreselijk plezierig gewerkt met een grote groep verschillende mensen,
waarvan ik er een aantal in dit voorwoord wil noemen.
Een eerste woord van dank gaat uit naar mijn begeleiders Bert Klandermans en Dirk
Oegema, die beiden op een verschillende manier zeer betrokken zijn geweest bij mijn
promotietraject. Bert bedank ik in het bijzonder voor de ruimte en vrijheid die hij mij
gegeven heeft zonder daarbij niet thuis te geven als ik hulp nodig had. Dirk was een
geweldig dagelijks begeleider en zijn enthousiasme en luisterend oor zijn onmisbaar
geweest. Ten tweede bedank ik Conny Roggeband en Hajo Boomgaarden. Een groot deel
van dit proefschrift komt voort uit twee onderzoeksprojecten die ik met elk van hen
afzonderlijk ben begonnen. Conny heeft me daarbij enthousiast gemaakt voor het
integratieonderwerp en ondanks het feit dat onze onderzoekstradities zeer sterk verschillen
is de samenwerking meer dan succesvol en plezierig te noemen. Hajo heeft zich gedurende
de afgelopen jaren ontwikkeld tot een echte ‘partner in crime’, waarbij een paperidee voor
een congres in Oxford is uitgegroeid tot hechte samenwerking en een goed persoonlijk
contact. Ik vind het geweldig dat we nu ‘echt’ collega’s zijn.
De afdeling Sociologie en in het bijzonder de ‘Social Conflict and Change’ groep heeft een
inspirerende werkomgeving gevormd. Ik wil in het bijzonder Ruud Koopmans bedanken
voor zijn betrokkenheid bij het project en zijn nooit ontbrekende kritische blik. Mijn
kamergenoten van Z-403, Evelyn Ersanilli, Heike Schröder, Inger Plaisier en Robert Braun,
verdienen het ook om genoemd te worden: zij boden een omgeving waarin
wetenschappelijke discussies nooit ver weg waren en spontane samenwerkingen
ontstonden, maar waar het toch vooral ook erg gezellig was. Ook David Hollanders was
daar deelgenoot van. Wij waren slechts korte tijd collega’s, maar hij heeft mij gedurende
het hele traject (statistisch) bijgestaan en veel van mijn kennis over econometrische
tijdreeksanalyses is te danken aan zijn directe nabijheid. Ook van de afdeling
Communicatiewetenschap zijn er een aantal mensen van grote waarden geweest.
Waarschijnlijk had ik de afgelopen jaren het grootste deel van mijn tijd besteed aan het
handmatig coderen van krantenartikelen als Wouter van Atteveldt niet de omgeving had
gecreëerd waarin de mogelijkheden van de computer echt goed benut worden. Zijn
technische kennis, maar ook zijn niet-aflatende hulpvaardigheid, zijn echt onmisbaar
gebleken. Met Lonneke van Noije deelde ik frustratie als de techniek ons in de steek liet en
euforie als het allemaal wel werkte. Mijn wetenschappelijke carrière is eigenlijk begonnen
iv
onder de vleugels van Jan Kleinnijenhuis en hij heeft mij nooit uit het oog verloren,
waarvan akte! De ASCoR PhD-club op de UvA, waar ik mij altijd zeer welkom heb
gevoeld, vormde een ideale plek om onderzoeksideeën en eerste versies van papers te
presenteren. Jochen Peter leidde deze groep met niet aflatend enthousiasme en wist als geen
ander duidelijk te maken dat wetenschap een mix van principe en pragmatisme is.
The department of Sociology at the University of California, Irvine hosted me for six
months during the fall of 2005 and spring of 2006. Dave Snow offered everything one
could wish for when entering an unknown scientific environment: hospitality and
intellectual inspiration. Catherine Corrigall-Brown, Steve Boutcher and Kelsy Kretschmer,
among others, made these six months probably the best of the past years.
Met Rik de Ruiter deelde ik gedurende mijn studietijd al een groot aantal interesses en het
was een waar genoegen om ook samen het promotietraject te doorlopen, hoewel de
Italiaanse zon wel eens wat jaloezie opwekte. Mijn familie en in het bijzonder mijn ouders
hebben altijd een grote interesse voor dat waar ik mee bezig was getoond. Mijn broer Arjan
was altijd letterlijk dichtbij en de vele koppen koffie en korte babbeltjes vormden een
welkome afwisseling voor het wetenschappelijke geploeter. Tenslotte kan één persoon niet
ongenoemd blijven: Marjolein is in vele opzichten onmisbaar geweest maar heeft mij
vooral altijd voorgehouden dat wetenschap meer dan data is en het leven veel meer dan
wetenschap. En gelijk heeft ze!
Ondanks de betrokkenheid van al deze mensen bij dit proefschrift blijft de inhoud en ook
eventuele fouten geheel mijn eigen verantwoordelijkheid.
Rens Vliegenthart
Amsterdam/Hilversum
Augustus 2007
Note: All data used in this dissertation are available at my personal website
www.rensvliegenthart.com.
v
Table of Contents
VOORWOORD/FOREWORD ........................................................................................III
TABLE OF CONTENTS ................................................................................................... V
LIST OF FIGURES........................................................................................................VIII
LIST OF TABLES.............................................................................................................IX
LIST OF ABBREVIATIONS ............................................................................................ X
CHAPTER 1. INTRODUCTION....................................................................................... 1
1.1
S
ETTING THE SCENE
...................................................................................................... 1
1.2
T
HIS RESEARCH
............................................................................................................. 2
1.3
T
HEORETICAL FRAMEWORK
.......................................................................................... 4
1.4
O
UTLINE OF THE DISSERTATION
.................................................................................. 12
CHAPTER 2. METHODOLOGICAL CONSIDERATIONS ....................................... 15
2.1
M
EASURING THE CONCEPTS
........................................................................................ 15
2.2
A
PPLYING TIME SERIES TECHNIQUES ON AGGREGATE LEVEL DATA
............................. 18
CHAPTER 3. DIVERGENT FRAMING: THE PUBLIC DEBATE ON MIGRATION
IN THE DUTCH PARLIAMENT AND MEDIA ........................................................... 31
3.1
A
BSTRACT
................................................................................................................... 31
3.2
I
NTRODUCTION
............................................................................................................ 31
3.3
M
EDIA AND POLICY FRAMING
..................................................................................... 32
3.4
H
YPOTHESES
............................................................................................................... 34
3.5
D
ATA
.......................................................................................................................... 35
3.6
M
ETHODS
.................................................................................................................... 36
3.7
R
ESULTS
..................................................................................................................... 39
3.8
C
ONCLUSIONS
............................................................................................................. 50
CHAPTER 4. PRESS AND PARLIAMENT: RELATIONSHIPS BETWEEN THE
DEBATE IN MEDIA AND POLITICS........................................................................... 53
4.1
A
BSTRACT
................................................................................................................... 53
4.2
I
NTRODUCTION
............................................................................................................ 53
4.3
R
ELATIONSHIPS BETWEEN PARLIAMENTARY AND MEDIA FRAMING
............................. 54
4.4
E
VENTS
....................................................................................................................... 57
4.5
M
ETHODS
.................................................................................................................... 58
vi
4.6
A
NALYSIS
.................................................................................................................... 60
4.7
R
ESULTS
..................................................................................................................... 62
4.8
C
ONCLUSION
............................................................................................................... 70
CHAPTER 5. ATTENTION IN THE MEDIA: FACTS AND ISSUE-COVERAGE IN
DUTCH NEWSPAPERS .................................................................................................. 73
5.1
A
BSTRACT
................................................................................................................... 73
5.2
I
NTRODUCTION
............................................................................................................ 73
5.3
N
EWS CONTENT AS A REFLECTION OF REALITY
? ......................................................... 74
5.4
T
HE REPRESENTATION IN THE NEWS OF REAL WORLD DEVELOPMENTS
....................... 75
5.5
T
HE RELATIONSHIP BETWEEN KEY
-
EVENTS AND MEDIA REPORTING
........................... 76
5.6
T
HE
D
UTCH CASE
........................................................................................................ 78
5.7
M
ETHOD
..................................................................................................................... 79
5.8
A
NALYSIS
.................................................................................................................... 82
5.9
R
EAL WORLD DEVELOPMENTS VERSUS EVENTS
........................................................... 83
5.10
T
OWARDS A TYPOLOGY OF EVENT
-
EFFECTS
.............................................................. 88
5.11
C
ONCLUSION
............................................................................................................. 89
CHAPTER 6. EXPLAINING THE RISE OF ANTI-IMMIGRANT PARTIES: THE
ROLE OF NEWS MEDIA CONTENT ........................................................................... 91
6.1
A
BSTRACT
................................................................................................................... 91
6.2
I
NTRODUCTION
............................................................................................................ 91
6.3
T
HE DEPENDENT VARIABLE
:
S
UPPORT FOR ANTI
-
IMMIGRANT PARTIES
........................ 92
6.4
T
HE INDEPENDENT VARIABLE
:
W
HY NEWS CONTENT MATTERS
.................................. 94
6.5
C
ONTROLLING FOR UNEMPLOYMENT
,
IMMIGRATION
,
AND LEADERSHIP
...................... 95
6.6
A
DDITIONAL FACTORS
................................................................................................ 97
6.7
H
YPOTHESES
............................................................................................................... 98
6.8
M
ETHODS
.................................................................................................................... 98
6.9
A
NALYSIS
.................................................................................................................. 100
6.10
R
ESULTS
.................................................................................................................. 100
6.11
C
ONCLUSION
........................................................................................................... 104
CHAPTER 7. FITTING THE PIECES TOGETHER: AN ANATOMY OF A
POLITICAL PROCESS ................................................................................................. 107
7.1
A
BSTRACT
................................................................................................................. 107
7.2
I
NTRODUCTION
.......................................................................................................... 107
7.3
R
EAL WORLD CUES
,
PARLIAMENT
,
MEDIA AND PARTY SUPPORT
:
ABOUT CHICKENS AND
EGGS
............................................................................................................................... 108
7.4
O
PERATIONALIZATION OF THE CONCEPTS
................................................................. 112
7.5
A
NALYSIS
.................................................................................................................. 115
7.6
R
ESULTS
................................................................................................................... 116
vii
7.7
C
ONCLUSION
............................................................................................................. 128
CHAPTER 8. CONCLUSION AND DISCUSSION..................................................... 131
8.1
S
UMMARY
................................................................................................................. 131
8.2
I
MPLICATIONS
:
WHERE DO WE TAKE IT FROM HERE
? ................................................. 135
8.3
P
OSSIBLE LIMITATIONS AND DIRECTIONS FOR FURTHER RESEARCH
........................... 143
ANNEXES ........................................................................................................................ 149
ANNEX 1. LIST OF SENSITIZING QUESTIONS TO CODE FRAMES ................ 151
ANNEX 2. PARLIAMENTARY DOCUMENTS USED IN QUALITATIVE
ANALYSIS....................................................................................................................... 152
ANNEX 3. FRAMING IN QUESTIONS OF POLITICAL PARTIES IN
PARLIAMENT................................................................................................................ 153
ANNEX 4. SEARCH STRINGS FOR CHAPTERS 3, 4 AND 7 ................................. 154
ANNEX 5. SEARCH STRING FOR CHAPTERS 5 AND 6........................................ 155
ANNEX 6. SEARCH STRINGS AND CONDITIONS FOR CHAPTERS 5 AND 6 . 156
ANNEX 7. ADDITIONAL ANALYSES CHAPTER 6................................................. 164
NOTES ............................................................................................................................. 167
BIBLIOGRAPHY............................................................................................................ 173
INDEX .............................................................................................................................. 185
NEDERLANDSTALIGE SAMENVATTING .............................................................. 189
viii
List of Figures
F
IGURE
1.1
T
HEORETICAL FRAMEWORK
................................................................................. 5
F
IGURE
2.1
S
TEP
-
WISE APPROACH IN
ARIMA
AND
VAR-
MODELLING
................................. 22
F
IGURE
2.2
ACF
S AND
PACF
S FOR
ARIMA
(1,0,0)
AND
ARIMA
(0,0,1)
MODELS
.............. 25
F
IGURE
3.1
A
TTENTION TO THE ISSUE OF IMMIGRATION AND INTEGRATION OF MINORITIES IN
PARLIAMENT
(
NUMBER OF OFFICIAL DOCUMENTS
)
AND NEWSPAPERS
(
NUMBER OF
ARTICLES
) ................................................................................................................... 39
F
IGURE
3.2
Y
EARLY USE OF EACH FRAME IN OFFICIAL PARLIAMENTARY DOCUMENTS
........ 40
F
IGURE
3.3
Y
EARLY USE OF EACH FRAME IN NEWSPAPER ARTICLES
.................................... 42
F
IGURE
4.1
S
IMULATION OF THE RESPONSE TO A ONE STANDARD DEVIATION CHANGE IN
STANDARDIZED ATTENTION AND FRAME SCORES FOR PARLIAMENT AND MEDIA
......... 66
F
IGURE
5.1
D
IFFERENT TYPES OF INTERVENTIONS
............................................................... 84
F
IGURE
5.2
I
MMIGRATION AND NUMBER OF ASYLUM APPLICATIONS IN THE
N
ETHERLANDS
85
F
IGURE
5.3
A
TTENTION FOR IMMIGRATION AND INTEGRATION IN
D
UTCH NEWSPAPERS
...... 85
F
IGURE
6.1
U
NEMPLOYMENT AND SUPPORT FOR ANTI
-
IMMIGRANT POPULISM IN THE
N
ETHERLANDS
,
PERIOD
1990-2002
(
MONTHLY DATA
).............................................. 100
F
IGURE
7.1
F
RAMING FAVORABILITY IN PARLIAMENT AND MEDIA
..................................... 117
F
IGURE
7.2
A
NTI
-
IMMIGRANT PARTY SUPPORT AND NUMBER OF ASYLUM APPLICATIONS
.. 117
F
IGURE
7.3
S
IMULATION OF THE
R
ESPONSE TO A
O
NE
S
TANDARD
D
EVIATION
C
HANGE IN
V
ARIOUS
V
ARIABLES
(A
TTENTION
-
MODEL
) ............................................................. 121
F
IGURE
7.4
S
IMULATION OF THE
R
ESPONSE TO A
O
NE
S
TANDARD
D
EVIATION
S
HOCK IN
V
ARIOUS
V
ARIABLES
(F
RAMING
-M
ODEL
) ............................................................... 125
F
IGURE
7.5
S
IMULATION OF THE
R
ESPONSE TO A
O
NE
S
TANDARD
D
EVIATION
S
HOCK IN
P
ARLIAMENT AND
M
EDIA
(
WEEKLY LEVEL
) ............................................................. 127
F
IGURE
8.1
S
UMMARY OF RESULTS
.................................................................................... 132
F
IGURE
8.2
P
ROTOTYPES OF THE DEMOCRATIC PROCESS
.................................................... 136
ix
List of Tables
T
ABLE
3.1
P
ERCENTAGE USE OF EACH FRAME IN QUESTIONS ASKED IN PARLIAMENT
BETWEEN
1995
AND
2004 ........................................................................................... 41
T
ABLE
3.2
C
ORRELATION BETWEEN TOTAL ATTENTION AND PROPORTIONAL USE OF EACH
FRAME IN PARLIAMENT AND MEDIA
(
MONTHLY VALUES
)............................................ 45
T
ABLE
3.3
D
IFFERENT USE OF FRAMES DURING DIFFERENT PERIODS IN PARLIAMENT AND
MEDIA
(
MONTHLY VALUES
) ........................................................................................ 46
T
ABLE
3.4
F
RAME DIVERSITY IN PARLIAMENT AND MEDIA
(
MONTHLY VALUES
).................. 48
T
ABLE
3.5
F
RAMING USE BY SEPARATE NEWSPAPERS
.......................................................... 49
T
ABLE
4.1
A
UGMENTED
D
ICKEY
-F
ULLER TEST FOR PRESENCE OF UNIT ROOT
..................... 60
T
ABLE
4.2
VAR
A
NALYSIS
(4
LAGS
)
FOR THE ATTENTION FOR IMMIGRATION AND
INTEGRATION IN MEDIA AND PARLIAMENT
(
PERIOD
1995-2004)................................. 63
T
ABLE
4.3
D
ECOMPOSITION OF
F
ORECAST
E
RROR
V
ARIANCE FOR
A
TTENTION AND FRAMING
IN PARLIAMENT AND MEDIA
........................................................................................ 67
T
ABLE
4.4
VAR
A
NALYSIS FOR THE USE OF SIX IMMIGRATION
-
AND INTEGRATION
-
FRAMES
IN MEDIA AND POLITICS
(
PERIOD
1995-2004).............................................................. 68
T
ABLE
5.1
D
IAGNOSTIC CHARACTERISTICS OF THE VARIABLES
,
PERIOD
M
AY
1991-
D
ECEMBER
2002......................................................................................................... 82
T
ABLE
5.2
I
NFLUENCE OF SOCIAL
-
ECONOMIC REAL WORLD DEVELOPMENTS AND KEY
-
EVENTS ON MEDIA ATTENTION FOR IMMIGRATION AND INTEGRATION
;
PERIOD
M
AY
1991-
D
ECEMBER
2002............................................................................................... 86
T
ABLE
5.3
A
CLASSIFICATION OF THE EFFECTS OF VARIOUS KEY
-
EVENTS
............................ 88
T
ABLE
6.1
I
NFLUENCE OF
F
ORTUYN
,
IMMIGRATION AND MEDIA ON ANTI
-
IMMIGRANT
POPULIST PARTY SUPPORT
,
PERIOD
1990-2002 ......................................................... 102
T
ABLE
6.2
I
NFLUENCE OF ECONOMY
,
IMMIGRATION AND MEDIA ON ANTI
-
IMMIGRANT
POPULIST PARTY SUPPORT
,
PERIOD
1990-
J
UNE
2001................................................ 104
T
ABLE
7.1
P
OLITICAL ORIENTATION
-
FRAME COMBINATIONS IN
D
UTCH NEWSPAPERS
....... 114
T
ABLE
7.2
P
OLITICAL ORIENTATION
-
FRAME COMBINATIONS IN
D
UTCH PARLIAMENT
....... 114
T
ABLE
7.3
V
ECTOR
A
UTOREGRESSIVE
A
NALYSIS FOR THE ATTENTION
-
MODEL
(
FIVE LAGS
,
PERIOD
1995-2004)................................................................................................... 120
T
ABLE
7.4
V
ECTOR
A
UTOREGRESSIVE
A
NALYSIS FOR THE
F
RAMING
-M
ODEL
(
FIVE LAGS
,
PERIOD
1995-2004)................................................................................................... 124
T
ABLE
7.5
C
ORRELATION BETWEEN RESIDUALS FOR BOTH ATTENTION
-
AND FRAMING
-
MODEL
...................................................................................................................... 127
x
List of Abbreviations
ACF Autocorrelation Function
ADF Augmented Dickey-Fuller (test)
AIC Akaike Information Criterion
AR Autoregressive
ARCH Autoregressive Conditional Heteroscedasticity
ARIMA Auto Regressive Integrated Moving Average
BIC Bayesian Information Criterion
CBS Dutch Statistical Office (Governmental organization)
CCF Cross-correlation Function
CD Centrumdemocraten (Dutch Extreme-Right Party)
CDA Christen-Democratisch Appèl (Dutch Christian-Democratic Party)
CP Centrumpartij (Dutch Extreme-Right Party)
CU Christen-Unie (Dutch Christian Party)
D66 Democraten '66 (Dutch Progressive Liberal Party)
Eq Equation
EU European Union
FPÖ Freiheitlich Partei Österreich (Austrian Extreme-Right Party)
GARCH Generalized Autoregressive Conditional Heteroscedasticity
GL GroenLinks (Dutch Greens)
HTK Handelingen Tweede Kamer / Proceedings of the Second Chamber
LN Leefbaar Nederland (Dutch Traditionalist Moderate-Right Party)
LPF Lijst Pim Fortuyn (Party founded by Pim Fortuyn)
MA Moving Average
MAR Moving Average Representation
MP Member of Parliament
MRG Manifesto Research Group
PACF Partial Autocorrelations Function
PvdA Partij van de Arbeid (Dutch Social-Democratic Party)
PVV Partij Voor de Vrijheid (Dutch Anti-Immigrant Party)
RMS Residuals Means Square
SEM Simultaneous Equation Modelling
SGP Staatkundig Gereformeerde Partij (Dutch Conservative Christian Party)
SP Socialistische Partij (Dutch Socialist Party)
VAR Vector Autoregression
VECM Vector Error Correction Models
VVD Volkspartij voor Vrijheid en Democratie (Dutch Conservative-Liberal Party)
1
Chapter 1. Introduction
1.1 Setting the scene
Over the past few years, the integration of ethnic minorities has been high on the political
and public agenda in almost all Western countries. Especially since 9/11 and the subsequent
‘war on terrorism’, the position in society of immigrants in general and Islamic immigrants
in specific has been the issue of intensive political debate (see also Norris et al., 2003).
In the Netherlands, an extra dimension was added to the debate on immigrants and
their position in society (‘integration’), because the debate has been sparked – apart from
international events such as 9/11 – by several domestic events; first and foremost the rise of
new-politician Pim Fortuyn in 2001-2002, the assassination of Theo van Gogh and the
controversy around politician and Somalian refugee Ayaan Hirsi Ali, which also attracted
considerable international attention (Koopmans and Vliegenthart, 2006). This highly
sensitive political situation makes it likely that debates in the Netherlands about the issue of
immigration and integration are magnifications of trends found in other Western countries.
From a policy point of view, the Netherlands is described as one of the clearest examples of
an actual paradigm shift, where multicultural policy goals have been abandoned for a more
restrictive, assimilationist approach towards immigrants (Modood, 2003; Joppke, 2004;
Kofman, 2005).
All these circumstances and the alleged change in paradigm in debates and policy
make the Netherlands an interesting case to study and indeed, the possible causes of these
changes are heavily debated, both in society and among social science scholars (Koopmans
et al., 2005; Ghorashi, 2003). Numerous explanations are offered, but three –by no means
mutually exclusive – of these seem to be dominant. First, changes are foremost interpreted
as a consequence of the changed (inter)national situation after 9/11 and other terrorist
attacks. These events are said to have caused changes in the intensity and content of debates
in politics, media and society and have resulted in a more hostile approach towards the
Islamic part of the population.
Second, a mediation of feelings that have already been present in society by political
entrepreneurs (mainly Pim Fortuyn) is argued to be at the heart of these changes. These
entrepreneurs were able to place an issue on the political and societal agenda that had
previously been neglected (consciously or unconsciously) by the political elite, while
cherishing the ideal of the ‘multicultural society’ (Koopmans et al., 2005). Multicultural
policies led, to paraphrase Koopmans and colleagues (2005: 245): ‘to the fragmentation of
migrant communities along ever smaller ethnic and religious group lines, resulting in an
institutionalization of inequality’. Their empirical analyses demonstrate the negative
consequences of the multicultural policy paradigm for the integration of minorities. Pim
Fortuyn made this issue debatable and his anti-immigrant rhetoric resulted in new
Framing Integration and Immigration
2
opportunities for all kind of political actors to use racist and anti-immigrant sentiments to
obtain political support. In the past few years, several new far-right political parties have
capitalized these opportunities and gained seats in Dutch national and local elections.
Third, the paradigm shift is argued to be merely a media phenomenon; the problem is
constructed within the media realm and picked up by both politicians and public. According
to this thesis the Dutch society is to be regarded as a mediacracy. This explanation has been
especially popular among politicians, who in several cases have pointed to the media,
which they hold responsible for creating a tense and negative context around immigrants
and the issue of immigration and integration. This media-criticism has not been limited to
the specific issue of immigration and integration. In 2003, two governmental advisory
committees published reports on the relation between the media and politics (RMO, 2003;
ROB, 2003), both having a general critical attitude towards politicians, but especially
towards journalists. Recently, a former Dutch reporter in the Middle East, Joris Luyendijk,
published the book ‘Het zijn net mensen’ [They are like ordinary people] (Luyendijk,
2006), in which he discusses the journalistic practices leading to a one-sided image of
Arabic people in Western media. In short, there seems to be a lot of ‘debate about the
debate’i on the issue of integration of minorities in the Netherlands (see also Koopmans and
Vliegenthart, 2006). It is perfectly clear that the debate on the issue has not been limited to
a purely political and policy-related matter, but has led to discussions in many other parts of
the society as well, most visibly in the mass media. A lot of media attention has been
devoted to questions such as whether Muslims were provoked, and whether freedom of
speech is incompatible with religious’ sensitivity towards harsh criticism.ii In general,
immigration and integration of minorities is the center of attention and subject of intense
discussions. In other words, it is a contested issue. Among the actors participating in these
debates are politicians, journalists, prominent opinion makers, civil organizations, social
movements, but also individual citizens that express their attitudes, opinions and feelings.
Although dramatic changes in the tone and intensity of the debate have been widely
recognized (Entzinger, 2003; Luedtke, 2005), the causes of these changes have not been
investigated thoroughly.
In sum, in these three explanations mentioned above, four possible sources of change
can be identified: politics, media, public and real world developments (both national and
international). Exactly the relationships between these factors are the main concern of this
dissertation. The research presented in this dissertation finds its roots in the desire to
describe the changes that have occurred and to shed light on the controversy that exists
about how to explain these dramatic changes in the immigration debate.
1.2 This research
Research questions
The central research questions that this dissertation seeks to answer are what changes have
occurred in the presentation of the immigration and minorities’ integration issue in the
Introduction
3
political and media realm, how did these realms influence each other, how they are
influenced by real world developments and the public – i.e. anti-immigrant party support –
and how politics and media influenced the public on their own right.
To outline the variation in political and media debates, first a set of more specific
descriptive questions needs to be answered. How prominent has this issue been in the
various realms over the past few years? Has the debate been as negative as some politicians
suggest? How prominent were various actors? How did these actors ‘frame’ the issue? Are
there indeed large differences between the presentation of the issue in politics and in the
media and if so, what are these differences composed of?
Second, to understand the causes and consequences of these variations, a set of
explanatory questions arises. To what extent do changes in the intensity and presentation of
the issue in parliament alter intensity and presentation in the media and vice versa? To what
extent have all kind of key-events (e.g. 9/11, the murder on Pim Fortuyn and Theo van
Gogh) influenced characteristics of the debates in both arenas? Do media and politics exert
a strong influence on the public with respect to this issue? Does the public in return affect
politicians and media as well?
Research design
The relevance of focusing on the Netherlands, lies not only in the sequence of events
mentioned in the first paragraph, but also in the large variation in all factors (real world
cues, politics, media and public) can be expected. This makes the Netherlands descriptively
and explanatory an interesting case. The focus is on a single country and thus it is not
comparative in the strict sense of the word (Peters, 1998). However, comparison does play
an important role, since this dissertation looks at debates in various realms, making a very
explicit comparison between them. Furthermore, cross-time comparisons are central in all
the studies. This approach can be regarded as complementary to many cross-national
comparisons (e.g. Koopmans et al., 2005) that are well able to explain differences across
units, but largely fail to grasp changes within one unit. Therefore, an in-depth analysis of
one country needs to precede a similar over-time analysis of several countries. The various
chapters and especially the conclusion pay ample attention to the possibilities for
generalization and for extending the research cross-nationally.
In the various empirical studies included in this dissertation, I will look at the period
between 1990 and 2004. Computer-assisted content analysis is used to describe the
presentation of the issue in the political and media realm. The outcomes of these analyses
are related to available data regarding real world developments and public (i.e. anti-
immigrant party support). Causal relationships between the four factors are analyzed using
advanced econometric time series techniques.
Framing Integration and Immigration
4
Relevance
This dissertation provides specific answers that help to understand the current political and
societal situation in the Netherlands and therefore has a clear societal relevance. Moreover,
in several ways it seeks to advance our more general (social-) scientific knowledge of
relations between media, parliament, public and the larger social-economic context.
Overall, a detailed picture of these relations is offered.
For my theoretical framework, I base myself upon insights from communication
science (mainly agenda setting and framing literature), sociology (mainly social
movements: framing and claims-making) and political science (on explaining shifts in party
support and effectiveness of government policies). Using this combination of theoretical
approaches contributes to an integral understanding of the interaction between real world
developments, politics, media and public.
The various studies presented in this dissertation extend our knowledge of the specific
relationships between these four factors. First, when considering relationships between
media and politics, research has been limited to agenda building processes, dealing with the
level of attention for issues. This research also takes into account how the issue is presented
(framed). Second, media-effects on support for anti-immigration parties have never been
assessed in a time series design controlling for a whole range of other possible
explanations. Nevertheless, in this dissertation some recent findings (mainly issue-voting
hypothesis, see Walgrave and De Swert, 2004) can be put to a rigid empirical test. Third, a
typology of context-effects on media attention for issues is developed, which enhances the
broader understanding of how media coverage comes about.
Methodologically, this dissertation sets out to demonstrate the usefulness of both
computer-assisted content analysis to analyze written texts and more advanced econometric
time series techniques to statistically analyze relationships between social phenomena at a
macro level. Computer-assisted content analysis offers opportunities to analyze large
amounts of texts consistently, which would be an immense task when manual coding was
used. The use of econometric time series techniques helps to assess the causal relationships
between the variables in a reliable manner.
1.3 Theoretical framework
Public debate is carried out in different forums: in the mass media, in parliament, in public
gatherings, in the streets and other places. Each forum has its own distinct characteristics
and differs in terms of participating actors and ‘rules of the game’. In this dissertation, two
of these forums will play a central role: the political forum and the media forum. However,
it does not limit itself to these two and also takes into account general public’s reactions by
considering anti-immigrant party support. Furthermore, it relates the forums to the larger
socio-economic and international political context by taking into account trends like the
level of immigration and events like 9/11. Figure 1.1 summarizes the main concepts and
Introduction
5
their relations (see Kleinnijenhuis and Rietberg, 1995 for a study that takes into account
equivalent relationships).
Figure 1.1 Theoretical framework
Identifying the main concepts
P
OLITICAL ARENA
:
OFFICIAL PARLIAMENTARY DOCUMENTS
Talking about ‘the’ political arena can be problematic, since it simplifies too much the
political institutional context. Polities in Western countries consist of various levels of
government and authority (e.g. local, regional, national). Furthermore, in recent decades we
have witnessed the coming into existence of another level of authority with increasing
policy capabilities: the European Community and later the European Union. This has
resulted in a multi-level governance structure with a variety decision capabilities and
relations between the levels (Hooghe and Marks, 2001). Whether the main authority lies at
the sub-national, national or supra-national (EU) level differs significantly across issues.
The immigration issue has remained a strongly nationally determined issue: the regulation
of immigration and integration is still the territory of the state and is even argued to
reinforce nation-statism (Kofman, 2005).
So, if we position the main authority regarding this issue at the national level, it makes
sense to look at the parliament (in the Netherlands the Second Chamber) as the most
prominent political arena. This is the place where parliamentarians, who have been directly
elected by the people, interact with government and where most decision-making takes
place. Moreover, it is a forum that is documented consistently and therefore it is most
accessible for analysis. Consequently, in this dissertation, I will focus on proposals by
Larger context: Social-economic developments and key events in the ‘real world’
Political arena
Attention for and
framing of immigration
and integration issue in
parliament
Media arena
Attention for and
framing of immigration
and integration issue in
newspapers
Chapter 7
Chapter 3,4,7
Chapter 4,5,7
Chapter 4,7
Public
Support for anti-
immigrant populist
parties
Chapter 6,7
Chapter 6,7
Framing Integration and Immigration
6
parliamentarians and members of the cabinet that are discussed in parliament and questions
posed by parliamentarians to government officials.
M
EDIA ARENA
:
NEWSPAPER COVERAGE
The broader societal debate largely takes place in the mass media. To reach a broader
audience or to be heard by (other) politicians, it is almost a necessary condition to receive
media attention. But merely receiving attention is not enough; the way actors and their
opinions are covered makes a huge difference. As Koopmans (2004: 368) puts it with
regard to social protest: ‘Authorities will not react to – and will often not even know about
– protests that are not reported in the media and if they are reported, they will not react to
the protests as they “really” were, but as they appeared in the media’. Theoretically, one
might expect the mass media to be a more open place compared to the political realm,
allowing all different kinds of actors to participate in discussions and offering a wide range
of problem definitions to gain ground and prevail. Though scholars have argued - both
normatively and empirically - for mass media (or a Habermasian ‘public sphere’) to have
strong pluralistic characteristics (e.g. Norris, 2000), many others have convincingly shown
that this realm is largely institutionalized as well and that selection processes (e.g. based on
news values, Galtung and Ruge, 1965) and gate keeping processes make it far less diverse
than often presumed (see also Hilgartner and Bosk, 1988; Koopmans, 2004 for more
encompassing frameworks to understand dynamics in the public sphere). Only certain
(elite) actors and certain opinions are likely to receive considerable attention and resonance
in the mass media (Koopmans, 2004). Due to these selection processes and the limited
space available, the ‘ordinary’ citizen is not likely to participate intensively in the debate
via the mass media.
P
UBLIC
:
ANTI
-
IMMIGRANT PARTY SUPPORT
If one wants to gain a complete overview of the discussion around an issue, one can not
limit oneself to studying the political and media realm, both highly elitist, although in
different ways. Additional insight can be gained by looking at the general public opinion
about minorities but also about which political solutions to existing problems are favored.
This last opinion is all the more relevant since in certain situations it can become an
important factor when deciding which political party to vote for, especially when this issue
is highly salient in the mass media (issue-voting hypothesis, see Walgrave and De Swert,
2004). Therefore, changes in public opinion can directly alter the political landscape by
shifting party support but can do so indirectly as well by prospective vote-seeking behavior
by politicians.
The public agenda is most straightforwardly measured when regarding the question as
to what the public considers being the most important problems the country faces.
However, this measure does not tell us much about how the overall public thinks about the
problem and what the preferred solutions might be. Furthermore, for the Netherlands these
data are not available throughout the whole research period. Therefore, I regard support for
Introduction
7
anti-immigrant parties as being the crucial variable when looking at the public. Next to the
pragmatic reason of data availability, there are two substantial reasons to use this measure.
First, anti-immigrant party support is directly relevant on its own and an important object of
study in political science. Second, derived from the issue-voting hypothesis (Walgrave and
De Swert, 2004), it indirectly tells us something about whether the public regard
immigration as an important issue and in fact, whether indeed it intends to act politically in
accordance with this importance as well. Extreme-right populist political parties provide
one way of channeling anti-immigrant sentiments (Koopmans, 1996b), with social
movements and non-institutional action like violence being the other. For the Netherlands,
Koopmans et al. (2005) demonstrate that both extreme-right party support and violence take
moderate overall values compared to other European countries, where either party support
or violence reach consistently higher levels. The past few years, however, witness an
increase in anti-immigrant party support in the Netherlands, while extreme-right violence
has decreased (Van Donselaar and Rodrigues, 2006).
R
EAL WORLD
:
DEVELOPMENTS AND KEY
-
EVENTS
Debates in various forums and public opinion formation do not take place in a vacuum. All
kind of larger societal developments are relevant and are likely to affect debates and public
opinion. These should be taken into account to get a complete picture of the debate.
Regarding this larger ‘real world’ (Kleinnijenhuis and Rietberg, 1995), I make a distinction
between developments and key-events, which can be both expected and unexpected, such
as elections, disasters or terrorism attacks.
For real world developments, the selection is pretty straightforward: first and foremost,
directly related tendencies should be considered. With regard to the issue of immigration
and integration of minorities, these include immigration figures and number of asylum
applicants (Lubbers et al., 1998). Second, economic developments like unemployment and
economic growth are argued to influence political preferences in general and anti-
immigrant party support in particular (Golder, 2003a, 2003b).
The selection of key-events is less straightforward. After all, the question as to which
events are crucial in understanding changes in the parliamentary, media and public agendas
and in framing of issues is a difficult one to answer ex ante. One seems to be inclined to
‘sample on the dependent variable’, meaning that one looks at changes in the ‘outcome’
(e.g. changes in media coverage) and tries to find events that ‘fit’ the data and are co-
occurring with significant shifts in this outcome over-time. To avoid this, I base myself
upon existing literature in which crucial events are distinguished (see for a more elaborate
discussion Chapters 4 and 5). In most cases, they intuitively make a lot of sense (especially
those related to the ‘war on terrorism’), in other cases, an outsider might be less aware of
their impact.
Framing Integration and Immigration
8
Issue attention and framing in parliament and media
To be able to systematically compare the political and media arena, one needs to identify
the characteristics along which they are to be compared. Furthermore, to be able to identify
similarities and differences in these characteristics, it is necessary to measure them
similarly for both arenas. To describe the characteristics of parliamentary documents and
newspaper articles, I make a distinction between the attention (how often and how
prominent is it presented?) and the framing of the issue (on which aspects does the
presentation focus?) (see also Rein and Schön, 1996; Snow and Benford, 1988).
Attention for issues has been regarded as extremely important in the study of media
coverage. At least since McCombs and Shaw’s agenda setting hypothesis (1972) it has been
one of the central concepts in media research and the crucial element of one of the
theoretical paradigms that has dominated media research and more specifically political
communication during the past decade (for an overview, see Dearing and Rogers, 1996). It
is probably the theory most frequently employed to investigate the effects of media on
public opinion. Attention can also be used to study political debates and compare various
agendas (Baumgartner and Jones, 1993; Walgrave and Van Aelst, 2006).
The concept of framing is often employed in the study of policy documents (Hajer and
Wagenaar, 2003) and a wide variety of approaches exist. Alike, framing has gained ground
in media-research over the past years (see for overviews De Vreese, 2005; Scheufele, 1999)
and also in this field numerous ways of defining and conceptualizing frames exist. In this
dissertation, I largely follow Snow and Benford’s (1988) approach. Though developed in
the context of social movements’ communication towards potential participants, their
analytical distinction turns out also to be of use in the analysis of more general media
content (see also Snow et al., 2007). Their approach is suitable for analyzing parliamentary
documents in a similar vein, allowing a systematic comparison between media and
parliament. Different representations of the problem are articulated in these parliamentary
and media documents. These representations include a diagnosis (what is the problem,
where is it located and what/who causes this problem?), connected to a prognosis (how
should the problem be resolved, what ends and means should be used and who is
responsible for the solution?) and a motivation, a rationale or call to action (what courses of
action are recommended, and who is responsible for this action?) (Snow and Benford,
1988). Theoretically, one can argue that a frame can exist of each possible combination of
existing diagnoses, prognoses and motivations, resulting in huge amounts of possible
frames. However, qualitative analyses, as reported in Chapter 3, suggest that only a limited
number of logically connected diagnoses and prognoses are indeed more than sporadically
present, while motivational elements are typically part of social movement communication
and are hardly present at all in especially media coverage.
Thus, based on Snow and Benford’s classification, issue-specific frames are identified
that are more suitable for comparing framing in media and parliament than more generic
news frames proposed by many communication scientists, which are strongly based upon
characteristics of news media (e.g. Semetko and Valkenburg, 2000; Iyengar, 1991; Price et
Introduction
9
al., 1997). The approach followed here differs in another way from that of many
communication scientists. While social movement scholars pay a lot of attention to the
sources of framing, communication scientists regard framing rather as a media
characteristic and are to a large extent ‘actor-blind’ (see for a more general argument about
the neglect of power in framing research: Carragee and Roefs, 2004). However, frames are
not actor-independent and involved actors often propagate their unique and largely
diverging views of the issue (Snow et al., 2007; Terkildsen et al., 1998). Therefore, taking
into account the actor that is given the right to speak and that is actually using a certain
frame (being a politician in parliament or basically any actor in the media) helps to
understand the characteristics of the coverage. Here, I distinguish the source of framing as
well, though the possibilities to do so are bounded by the use of computer assisted content
analysis tools (see also Chapter 2). These limitations result in the use of the framing-
concept mainly as a tool to understand the origins, content and consequences of
parliamentary and media debates, that do not necessarily correspond with the intentions of
sources or predispositions of recipients.
Relationships between the variables
After having introduced the four main concepts of my theoretical framework, I now turn to
the relationships between these concepts, represented by the arrows in Figure 1.1 in this
introduction. I do not strive for a complete overview, since each chapter includes a more
elaborate discussion of the relationships that are part of the empirical research presented in
that chapter. Therefore, I limit myself here to highlighting some main points on each of the
investigated relationships and indicate which elements are parts of the following chapters.
P
ARLIAMENT AND MEDIA
The relation between the political and media realm has been subject of several studies. The
main focus has almost solely been on the issue attention, rather than the way the issue is
presented (i.e. issue-framing). Extending the classical agenda setting thesis (McCombs and
Shaw, 1972), the primary focus has been on the agenda building process, in which the
arenas affect each other in terms of the attention for issues. The results of studies
investigating this relationship do not give a conclusive answer to the question ‘who
influences whom?’ Whether media influence politics is largely dependent upon the context
in which the study takes place. Researchers find significant media-effects when regarding
‘symbolic’ politics (interviews, speeches) rather than substantial (laws) policies and routine
politics rather than election times (see Walgrave and Van Aelst, 2006 for an overview).
Research on the reverse relationship does not find consistent results either: in some cases
rather strong effects from politics on media are found, especially in the Dutch context
(Kleinnijenhuis, 2003; Kleinnijenhuis and Rietberg, 1995; see for the United Kingdom
Brandenburg, 2002), while in other cases, this effect is non-present (Wood and Peake,
1998) or dependent upon issues’ characteristics (Soroka, 2002). My focus is on substantial,
middle- and long-term relationships between media and parliament and therefore I mainly
Framing Integration and Immigration
10
focus on official parliamentary documents discussed in parliament (laws, policy proposals
and written parliamentary questions to members of government). Additionally, in Chapter 7
I look at transcripts of parliamentary debates, for which Walgrave and Van Aelst (2006)
argue that it is more heavily influenced by media coverage.
Furthermore, I do not only test the agenda building hypothesis, but also regard the way
the issue is presented (framing) as an object of study, something that has not been done
until now. In Chapter 3, I hypothesize about the substantial difference in framing between
the two arenas, looking at which frames are used, how fast changes over-time occur and
how much variation we find in the two arenas. In Chapter 4, I develop a theoretical
argument that conceives for certain frames an influence from politics on media and for
other frames an influence from media and politics, depending upon the previous use of the
frame in both arenas.
M
EDIA AND ANTI
-
IMMIGRANT PARTY SUPPORT
The relation between media and public is explored in Chapter 6, which looks into media-
effects of issue attention on anti-immigration party support. Relying on the issue-voting
hypothesis, I expect more media attention to result in more anti-immigrant support, even
when controlling for real world developments. This hypothesis combines the ideas of
agenda setting and issue ownership. The basic idea of the agenda setting hypothesis is that
issues that are high on the media agenda will become high on the public agenda as well. In
other words, the media do not so much determine what people think, but where they think
about (Cohen, 1963; Dearing and Rogers, 1996; McCombs and Shaw, 1972). Issues high
on the media agenda are then likely to be the issues the public takes into account when
deciding which political party to vote for. The issue ownership hypothesis assumes that for
every political issue there is a certain political party that people feel drawn to and from
which they believe that it can deal with the issue at stake more effectively than other parties
(Budge and Farlie, 1983; Petrocik et al., 2002). Walgrave and De Swert (2004) find
considerable support for their issue-voting hypothesis in the case of the Flemish extreme-
right party Vlaams Blok. In Chapter 7, I extend the analysis and also look into framing-
effects. I expect framing that is more congruent with positions of anti-immigrant parties to
have a positive impact on support for these parties (for a similar argument, see 'valenced
news frames', De Vreese and Boomgaarden, 2003).
Most empirical studies regard the causal relation between media and public to go from
media to public and scholars tend to focus on media-effects on the public. However, one
can could investigate whether the public’s perceptions have an effect on media coverage as
well. From a more normative point of view, those supporting the idea of ‘public journalism’
emphasize the importance of this relationship. In public journalism, the expression function
of media is crucial (Eksterowicz et al., 1998). According to this idea, the main task of the
journalist is to represent feelings and ideas among the population. If journalists take this
function seriously, one would expect that the overall coverage reflect (at least partly) issues
and feelings that are present among the population. Then, through media coverage
Introduction
11
politicians get information about what citizens think. On the one hand, empirical research
has found little evidence for news media working in correspondence with this function (but
see Oegema et al., 2007; Soroka, 2002). On the other hand, one could argue that the issue
of immigration and integration of minorities has some ‘populist’ characteristics that might
make journalists more sensible to changes in public opinion. In Chapter 7, I will investigate
whether the public’s influence on media coverage is present in the case of the issue of
immigration and integration.
P
ARLIAMENT AND ANTI
-
IMMIGRANT PARTY SUPPORT
Combining the expectations for Chapters 4 and 6 suggests an indirect relation from politics
via media to public. However, there might also exist a direct interaction between the two
that is not mediated by the media (see Kleinnijenhuis, 2003). Does the public react upon
what politicians do or express or is it the other way around? The latter is referred to as the
‘bottom-up’ agenda setting process (Kleinnijenhuis and Rietberg, 1995). According to Page
and Shapiro (1983), this influence indeed exists and more recent work also points to the
presence of this relationship as well (see Wlezien, 2004 for an overview). Geer (1996)
argues that nowadays this influence is even stronger than it used to be, since the increased
possibilities of polling public opinion can give politicians and policy makers direct and
plentiful information about the public’s preferences. The direct influence of politics on
public is not as often investigated as the influence of the public on politics, which is
probably due to the fact that political information reaches citizens first and foremost via
media coverage (Kleinnijenhuis and Rietberg, 1995). In Chapter 7, I will test for the
presence of these relations in the case under study.
R
EAL
W
ORLD AND PARLIAMENT
,
MEDIA AND ANTI
-
IMMIGRANT PARTY SUPPORT
Including real world developments as explanatory variables is worthwhile for two reasons.
First, they function as control variables, decreasing the chances for spurious correlations
between parliament, media and public. Second, from a substantive point of view, the larger
social-economic context might also have an impact on all three arenas and this impact can
be theoretically interpreted as well. In times of high levels of immigration, for example,
there might be a higher awareness of the problems associated with immigration among
politicians, media and public (see for example Lubbers et al., 1998). In Chapter 5, I
especially pay attention to the question as to what determines media attention for the issue.
Are real world developments or key-events important factors in determining media
attention and how do they affect this attention? In Chapter 6, when trying to explain anti-
immigrant party support I consider several real world indicators, which previous research
has found to be of importance. In Chapter 4 and 7, I take into account these real world
developments when looking at relations between media, parliament and public. In Chapter
5, I investigate in more detail their impact on issue attention in the media.
Finally, is there a relation to be expected from any of the arenas to larger social-
economic trends? Though it is not the main focus of this dissertation, I do control for these
Framing Integration and Immigration
12
relationships and one might indeed expect some influence – be it moderately – to be
present. For example, if one regards action in parliament to be more than ‘speech acts’ and
assumes that policy making is to a certain extent effective, changes in the political realm,
both in terms of changes in attention for an issue and in the way the issue is discussed
should be reflected in social-economic trends. So, on the one hand in the long run a
decrease in immigration levels can be expected when this topic is discussed intensively and
negatively in parliament and policy is made accordingly. On the other hand, as we know
from the study of policy, the effectiveness can be rather limited, due to all kind of long-
term, institutional constraints (Czada et al., 1998), other factors that are not easy to control
(Green and Green, 1995) or unanticipated negative consequences of proposed measures
(Enders and Sandler, 1993). Similarly, a negative nationwide public or media attitude
might, over-time and again probably only to a limited extent, affect the image a country has
for possible immigrants and thereby decrease the level of immigrants. This hypothesis is
presuming the existence of strong transnational links between diaspora and their home
country and a certain level of integration of these diaspora in the host country, since some
knowledge about media and public attitudes has to be obtained and transmitted. These
relationships are not taken into account very often (see the neglect of this relationship by
Kleinnijenhuis and Rietberg, 1995), but are considered in Chapter 7 of this dissertation.
1.4 Outline of the dissertation
The remainder of the dissertation includes a methodological chapter, five empirical studies
and a conclusion. In the methodological chapter (Chapter 2) I elaborate on several crucial
points regarding data collection and data analysis.
The first study (Chapter 3), Divergent Framing: The Public Debate on Migration in
the Dutch Parliament and Media gives an overview of the framing of immigration and
integration in both parliament and media and systematically compares their characteristics.
The second study (Chapter 4), Press and Parliament: Relationships Between the
Debate in Media and Politics looks into the causal relationships between the two arenas in
more detail: do changes in one arena lead to changes in the other arena?
In the third study (Chapter 5), Attention in the Media: Facts and Issue-Coverage in
Dutch Newspapers, issue attention is the dependent variable and different socio-economic
trends and key-events are used to explain this attention. Explaining the Rise of Anti-
Immigrant Parties: The Role of News Media Content (Chapter 6) looks at the consequences
of media attention for anti-immigration party support, while also taking into account other
possible explanations (immigration, unemployment).
In the final study (Chapter 7), Fitting the Pieces Together: An Anatomy of a Political
Process, I link the four factors that are central in this dissertation and establish their
relationships.
Introduction
13
In the concluding chapter (Chapter 8), I will highlight the main findings, formulate an
answer to the central research question and discuss the practical and scientific implications
of this dissertation, as well as its limitations.
15
Chapter 2. Methodological Considerations
The nature of the research as described in the introduction requires certain methodological
choices to be made. In each of the next chapters the specific data sources,
operationalizations and analyses are explained. In this chapter, I will discuss two issues that
recur in all the studies. First, I will introduce the method to analyze newspaper articles and
parliamentary documents. Second, the statistical techniques that are employed are
discussed. The nature of the data (aggregated level time series) and the research questions
(the interest in causality) require sophisticated statistical techniques.
2.1 Measuring the concepts
For my data on the public I rely on existing sources and use monthly (aggregated) level
data on anti-immigrant party support as has been measured in existing public opinion polls.
Similarly, for real world cues I use available data from statistical offices to establish
monthly figures on immigration, asylum seekers and unemployment.
The real challenge lies in assessing the political and media debates. Here I use
quantitative content analysis. To quote Shapiro and Markoff (1998), it can be defined as a
technique of measurement, trying to ‘capture text and meaning in numbers and figures’. It
differs from qualitative content analysis (or discourse analysis, see for example Wester,
1995), in that it tries to reduce complexity and makes data suitable for statistical inferences
(Krippendorff, 2003). Since I am dealing with a large time scale and have numerous
parliamentary documents and newspaper articles to analyze, I use computer-assisted
content analysis, which makes it possible to analyze large amounts of texts with relative
ease. Here, numerous strategies exist, varying from simple document count to all kind of
more complicated analyses (e.g. Latent Semantic Analysis) or even full grammatical
parsing (Collins, 2003), yielding very rich data structures. However, these data are non-
trivial to interpret and more complicated methods are still suffering from reliability
problems (West, 2000). Therefore, my strategy is based upon a rather simple, but efficient
method. Using manually constructed wordlists, I am able to measure attention of actors and
issues and even frames accurately. Regarding the distinction of frames, I loosely follow
Gamson and Mogdiliani (1989), who mention ‘catchphrases’ as one of the devices that
signify the presence of a frame (see also Pan and Kosicki, 1993: 61-62).
To be able to pursue a project like this it is worth noting that an infrastructure that
makes the management and actual content analysis of large amounts of documents is an
absolute requirement. The ANOKO database, as developed under supervision of Wouter
van Atteveldt at the Vrije Universiteit Amsterdam, provides such an infrastructure and is
used for both data management and content analysis. For more elaborate description of the
Framing Integration and Immigration
16
ANOKO database, please see Kleinnijenhuis and Van Atteveldt (Kleinnijenhuis and Van
Atteveldt, 2006).
Attention
With the choice of computer-assisted content analysis, I limit myself to a certain type of
media, namely the written press. More specifically, I analyze newspaper articles. Though it
might be theoretically possible to analyze for example television coverage using computer-
assisted techniques, this was in this specific case problematic, mainly because transcripts
were not freely available. From a more substantive point of view, one can also argue that if
the interest lies in the debate around the issue of immigration and integration of minorities,
newspapers are more suitable to consider, since they contain more substantive news than
television news bulletins (Kleinnijenhuis, 2003). And, contrary to popular belief television
and newspapers are fairly close in content – at least in the Netherlands (Kleinnijenhuis,
2003). Furthermore, I take into account the whole range of national newspapers, thereby
capturing the largest part of the variation that is present within the Dutch newspaper
landscape. These newspapers represent mainstream politics in the Netherlands (Van der
Eijk, 2000) and reach one third of the Dutch population (Cebuco, 2006).
As has been mentioned earlier, I am first of all interested in attention for the issue. In
most studies, operationalization has been straightforward: attention is usually measured as
the frequency of mentioning of the issue, combined with a measure of prominence. In order
to establish attention-scores for newspaper coverage, I follow a two-step approach. First,
articles are selected from the digital newspaper archive LexisNexis with a very broad
search string. Second, the selected articles are used to search for a list of keywords, which
frequencies and position determine the attention for the issue. This list of keywords is
constructed based upon manually coding of a (relatively small) sample of the articles.
However, the meaning of keywords is not in all cases unambiguous and words may have
different meanings depending on the context. The computer cannot grasp the semantic
meaning of words. To disambiguate the meaning of the keywords it may be necessary to
add a condition consisting of one or more words that need to occur within the article or
within a certain distance (e.g. ten words) from the keyword (see for an example of those
lists Annex 6). Hence, the keyword ‘counts’ and is included in the calculation of the
attention score when this condition is met. The attention score per article is not just the sum
of the occurrence of all keywords, but is regarded as being sublineary dependent upon these
occurrences. The underlying idea is that for example the tenth occurrence of a keyword
contributes less to the saliency of the issue within the article than the first does (see also
Manning and Schütze, 2002). Furthermore, occurrences in the headline are regarded as
more important than occurrences in the body of the text.
The prominence of the article can be taken into account as well. The assumption here
is that more prominent articles can be regarded as more visible and will reach a larger
audience. This prominence is based upon two criteria: the position of the article in the
newspaper (front page or not) and the circulation number of the newspaper. In Chapter 5
Methodological Considerations
17
and 6, prominence is fully taken into account. More specifically, as soon as any of the
keywords was present at least one time, the following formula is used to calculate the
attention for the issue:
Eq 2-1
(
)
fpacircffissuevissuev
bodyhead
articlesa articlesa
a
⋅⋅+==
∑
∑
∈ ∈
)(28log)()(
2
Where v(issue) is the attention for the issue within a certain period in time. This attention is
dependent upon the attention for the issue in all articles (a) that appeared during that period
in Dutch national newspapers. The article-level attention score is dependent upon the
number of key words in the headline (
head
f
) and the number of key words in the body of
the article (
body
f
). The multipliers for head and body occurrences are chosen in such a way
that the first occurrence in the headline is weighed as 3 (2log 8*1) and in the body of the
text as 1 (2log 2*1). With the presence of more key words, the score increases sublineary,
with each multiplication of the weighted number of hits by two resulting in a 1 point higher
visibility score. This indicates that the first occurrences of keywords contribute most to the
visibility score, while later keywords make less of a difference. Circ(a) is the circulation
figure of the newspaper in which the article is published divided by the mean circulation of
the newspapers included in the analyses and fp the article salience, which has a value of 2 if
the article is on the front page and 1 otherwise. In Chapters 3, 4 and 7, attention scores are
more straightforwardly based upon summing up the number of parliamentary documents
and newspaper articles in which each of the pre-defined frames occurs. Prominence is only
taken into account by counting newspaper articles that appear on the front page twice.
Though search strings and weighting-procedures differ there is high correlation between the
monthly media attention scores used in Chapter 5 and 6 and Chapter 3,4 and 7 (for the
overlapping period 1995-2002: r= .75).
Framing
When regarding the presentation of an issue several characteristics of the communicating
text under study (or even more specific elements, like sentences) can be taken into account
offering a wide range of characteristics: e.g. general tone, issue stances, attributions
(second-level of agenda setting) and framing. Regarding the latter, scientific debate exists
about conceptualization (e.g. using generic versus issue-specific frames) and
operationalization (how to measure it?) (De Vreese, 2005). As has been mentioned
previously, in this dissertation, frame definition is mainly based upon the classical
distinction between diagnostic, prognostic and motivational framing, as has been made by
Snow and Benford (1988) for the study of social movements. In a qualitative study, as will
be described in Chapter 3, five issue-specific frames are distinguished.
For both newspaper articles (from LexisNexis) and parliamentary documents (from
Parlando) a collection of documents is selected using a broad search string. Based on Snow
Framing Integration and Immigration
18
and Benford’s classification, a sample of these possible relevant documents is manually
coded. During this procedure, word-combinations (using disambiguity rules like ‘AND’,
‘OR’, ‘NOT’, within x number of words, see Annex 4) are constructed that identify each of
the five frames. In the next step, the presence of these frames is established in all
newspaper articles and parliamentary documents.
For both the attention and frames and both for newspapers and parliament, document
level scores are aggregated to a monthly level. For this time interval, data for real world
developments and anti-immigrant party support are also available.
Finally, the presence of a combination of an actor and a frame within newspaper
articles can be used to establish which actor is more often associated with a certain frame
than other actors. Though this co-occurrence does not directly measure the actual frame-use
of actors, supplemented with more specific information on what parliamentary questions
are posed by which political parties (explicitly listed in Parlando), it provides insight into
which actors employ what frames (see Chapter 7).
2.2 Applying time series techniques on aggregate level data
In this paragraph, I will discuss some particularities of the data and appropriate techniques
that are required to analyze them.
Aggregate level data
For all studies presented in this dissertation, I use aggregated monthly measures of the
various factors taken into account. Applying time series techniques on these data allow for
strong claims on causal relationships between real world cues, parliamentary debates,
media coverage and anti-immigrant party support. From a political communication point of
view, it is worth mentioning that some skepticism towards the use of aggregate level data
sets exists. Critics point to the famous ecological fallacy problem (Robinson, 1950).
Ecological fallacy refers to the incorrect assumption that relationships between variables at
a group level imply the same relationships to be present at the individual level. This makes
inferences about individual behavior from aggregate level data problematic. One should
indeed be very cautious to do so: a lot of theoretical claims and empirical findings on a
macro level assume certain mechanisms that take place on the individual level, but can not
be explicitly tested. In the context of this study, two comments should be made with regard
to this issue. First, whether this issue is of any relevance depends on the research questions
posed. In this dissertation, the main interest lays in general relationships rather than in –for
example - individual-level media-effects as in many studies in the field of political
communication. For example, hypothesizing that media attention for the immigration issue
increases overall anti-immigrant party support (see Chapter 6) is not to say that one certain
individual being exposed to this media attention will vote for the anti-immigrant party. It
does result in the expectation that overall more individuals will be inclined to do so. This
does not mean that individual-level studies are worthless in this context. On the contrary,
Methodological Considerations
19
they can offer a lot of additional insight and help to understand certain processes more
specifically (see for example Boomgaarden and De Vreese, 2007). Neither does this mean
that I do not – in some instances - rely on theories that have been mainly applied in
individual-level studies, especially on media-effects on public opinion, to formulate
expectations about relationships on the aggregate level. Some caution, however, is indeed
required: one might not be able to establish causality on the individual level, but one can
establish this causality at the aggregate level and make plausible that certain mechanisms
are present at the individual level. These mechanisms might include individual choices by
politicians and journalists and changes in electoral preferences of individual citizens.
A second point that is worth making is that how to deal with ecological fallacy
problem is still not widely agreed upon. Scholars argue that similar and even worse
problems can arise with individual level data (Kramer, 1983) or that applying the right
statistical techniques on aggregate level data can overcome the problems (King, 1997).
From a substantive point of view, aggregate level analyses offer several advantages. They
allow sorting out time order which leads to stronger claims about causality. A second
advantage compared to individual level approaches in general is that they allow for
investigation of long-term influence on larger social systems (Wu et al., 2002). As Slater
(2004: 178) puts it: ‘When such [time series] studies are an option […] they permit
exceptionally robust inference concerning real world effects of media exposure on national
populations.’
In this dissertation, the units of analysis are months. Practically, for real world
developments and anti-immigrant party support, data for shorter periods are often not
available. From a substantive point of view, monthly level data is appropriate to profit from
the potential advantages of time series analysis: the period is long enough to take into
account substantial amounts of parliamentary documents and newspaper coverage per unit
of analysis and it makes possible to capture both middle-range and long-term influences
(see for similar monthly-level analyses for example Wu et al., 2002; Goidel and Langley,
1995; Sanders and Gavin, 2004).
Time series analysis
To get reliable results from an analysis with time series data it is a requirement that these
data are treated in the appropriate statistical manner. The research questions posed in this
dissertation require the establishment of causality between two or more series such as
monthly issue attention in parliament and media. To be able to do so, it is important to first
fully understand the dynamics that each of the series contains. In this dissertation,
econometric time series techniques are employed, which make it possible to take into
account the specific characteristics of time-related data and treat them with care.
Furthermore, such techniques offer the opportunity to deal with one of the requirements of
causality: the cause (independent variable) has to precede the consequence (dependent
variable). Sometimes this effect from the independent on the dependent variable sets in
immediately. In other cases this might take several months. On the one hand this is a
Framing Integration and Immigration
20
theoretical question: how much time does one expect before the influence from the
independent on the dependent variable becomes apparent. Dearing and Rogers (1996) for
example argue that the occurrence of agenda setting effects can range from a couple of days
to several months. On the other hand, within the range of theoretically plausible options,
econometric techniques help to select the statistically most appropriate time lag.
Though numerous other techniques exist to analyze monthly level data with, the
discussion is limited to the ones that are used in the following chapters: ARIMA and VAR
analyses. Both are often-employed in econometrics: ARIMA is suitable when it comes to
establishing the size and delay of effects of pre-established independent variable on a
dependent variable, VAR when establishing what causal relationships between two or more
variables exist. Before discussing these two techniques in more detail, I will first discuss
potential pitfalls when analyzing time series data. In general, three potential problems with
time series are to be taken into account: problems with non-stationarity, problems with
correctly accounting for the series’ past and problems with autoregressive conditional
heteroscedasticity.
First, problems arise the series contains an upward (or downward) trend, which is
called non-stationarity. When a series contains an upward trend, its values are steadily
increasing over time. If this is the case for both the dependent series and (one of the)
independent series, it is likely that statistically a very large influence from the latter on the
first is found that does not make sense from a theoretical point of view and is (largely)
spurious. This point is nicely illustrated by David Hendry’s (1980) classical example in
which he almost perfectly explains the UK consumer’s price index by the cumulative
rainfall in the UK. This very strong effect is of course theoretically nonsense, but shows up
in the analyses because both series show a similar upward trend over time. If a series is
non-stationary, a solution that works most of the times is taking differences. This means
that instead of using the absolute values in the analysis, the difference between the current
and previous value is used. Usually, this removes the trend that is present in the series.
Using differenced series still offers the opportunity to determine the effects of an
independent variable on a dependent variable. In the next section, I discuss more in detail
how to establish whether a series is non-stationary.
The second problem arises when the series’ own past is not correctly taken into
account. Again, if one fails to deal in a proper manner with the series’ past and one leaves
information from this past unused, significant effects from other variables on the series can
be spurious. In time series values at a certain point in time do not arise out of the blue, but
largely depend on previous values. A clear example is ‘news being news, partly because it
was news yesterday’. The value one time period earlier is a very good predictor of the
current value. Taking into account the previous value of the series and use it as an
independent variable to explain the series is often a good starting point and sometimes even
a sufficient solution. However, in some cases more complicated dynamics are present and
these need to be dealt with as well. For example, media attention at certain times peaks.
After this peak, it usually moves back quickly in the direction of previous values. This
Methodological Considerations
21
specific dynamic can be modeled within the ARIMA-framework (see the next section for
more detail on how to do so).
The third potential problem arises when the series displays periods of relative
tranquility - showing up in steady values with little variance - followed by more turbulent
periods, when scores fluctuate more heavily. In this case, the series’ variance depends upon
the time of measurement, which is called autoregressive conditional heteroscedasticity.
When this is the case ARIMA and VAR-models do not suffice and other statistical
techniques – more specifically ARCH and GARCH models - are necessary to obtain
reliable results. Since this problem does not occur with the data used in the analyses
presented in Chapter 3 till 7, it is not discussed in further detail.
In the next sections, ARIMA and VAR-techniques will be presented in more
(technical) detail. It becomes clear in what situations which technique is chosen and what
steps need to be taken to assess the power and nature of relationships between the variables
included in the analysis.
Figure 2.1 summarizes these steps for both approaches. For both, the first step deals
with the first mentioned problem of non-stationarity and the second step with the second
(modeling the series’ past) and third problem (over-time instability of variance). Step three
involves the actual analysis of causal relationships between two or more series. The tests
that are used to answer substantial research questions are marked grey, while the other tests
deal with proper handling of the data as summarized above.
O
NE
-
DIRECTIONAL RELATIONSHIPS
:
ARIMA
What is it and when to use it?
ARIMA is the abbreviation for Auto Regressive Integrated Moving Average. The method
was initially developed by econometrists George Box and Gwilym Jenkins (1970) and is
therefore also called Box-Jenkins modeling. Its main assumption is that a time series own
past contains a lot of information that can help to explain its current value and therefore,
before endogenous explanatory variables can even be considered, it is first necessary to a
build a model that is based on the series’ own past. Only in a second phase explanatory
variables are added. Accounting for the series’ past in a time series design offers similar
advantages as the use of an experimental, individual-level study does. To determine the
consequences of certain stimuli for individuals’ attitudes in an experimental setting, checks
for causality are built in by measuring attitudes twice (both before and after the stimuli)
and/or by the use of control groups. In that way, one controls for the past. This is what
happens within the ARIMA framework as well: by modeling the series’ past, one controls
for it and likewise within an experimental setting, claims on causality are more persuasive.
ARIMA models are especially suitable for time series that are measured over a longer
period in time without missing values. Furthermore, as with ordinary regression analysis,
one can analyze causal relationships only in one direction. In other words, it is necessary to
establish the dependent variable and the independent variable(s) in advance.
Framing Integration and Immigration
22
Figure 2.1 Step-wise approach in ARIMA and VAR-modelling
Note: The tests that are used to answer substantial research questions are marked grey.
Research question:
What causal relationships are involved?
ARIMA-framework
Used for one-directional relationships
Step 1: Test for stationarity (using ADF) and
establish order of integration (I)
Step 2: Add autoregressive and moving average
terms (using ACF and PACF-functions) and
determine complete ARIMA(p,I,q)-model
Test
Are residuals white noise? (Q-statistic)
Is the model the most parsimonious one? (AIC
and BIC)
Is there no autoregressive conditional
heteroscedasticity? (Lagrange Multplier test)
Step 3: Add explanatory variables, establish
appropriate lag length using CCF
Test
Is there a significant influence from the
explanatory variable? (coefficient and standard
error)
Has the explanatory power of the model
improved? (AIC and BIC)
Are residuals white noise? (Q-statistic)
VAR-framework
Used for multi-directional relationships with
less certainty about nature of relationships
Step 1: Test for stationarity (using ADF) and/or
cointegration (Engle-Granger test)
Step 2: Model specification: determine lag
length (likelihood test for dropping lags)
Test
Are residuals white noise? (Q-statistic)
Is there no autoregressive conditional
heteroscedasticity? (Lagrange Multiplier test)
Step 3: Interpretation of results
Test
Is there an influence from variable x on
variable y? (F-test for Granger causality)
What is the direction and size of the influence
of variable x on variable y? (coefficients of
individual lags / MAR-representation)
To what extent does variable x account for
variation of variable y in the long run?
(Decomposition of Forecast Error Variance)
Is there no contemporaneous correlation
between the residuals? (Pearson’s Correlation)
Methodological Considerations
23
In Chapter 5, explaining media attention for immigration and integration is the main
aim and media attention for the issue is therefore used as the dependent variable in the
analysis, while real world developments and key events are the independent variables. In
Chapter 6, anti-immigrant party support is the dependent variable, while media attention
and other contextual factors are the independent variables. Finally, the ARIMA-framework
allows one to easily establish the appropriate lag length on which an independent variable
influences the dependent variable.
According to the ARIMA-framework a series can be described as a (p,I,q) process,
with p autoregressive terms, I orders of integration (number of times a series needs to be
differenced before it is stationary) and q lagged moving-average terms. In the following
sections, I will explain the issues related to each of these elements. I will start discussing
the integration part that is closely related to the concept of stationarity.
Step one: testing for stationarity
The first step in the ARIMA-model building process is to establish whether the dependent
time series is stationary. The assumption of stationarity is one of the basic requirements for
properly modeling time series. Stationarity is necessary to be able to make a reliable
estimation of parameters within the ARIMA-model. A series is stationary if its mean is
unaffected by a change of time origin and thus the expected value is the same for all time
points.i Most often, the augmented Dickey-Fuller (ADF) test that accounts for the
complicated structure of most time series is used to test whether a series is stationary. If the
ADF test indicates that the series is non-stationary, the series is said to contain a ‘unit root’
and to be integrated of order one or I(1). More specifically, a unit root means that including
the lagged dependent variable (t-1) as an explanatory variable for a series results in a
parameter that is equal to 1. It is then necessary to take differences and continue the model
building process with the differenced series. For the differenced series, it is again required
to test whether it is stationary. If not, the series is integrated of order two or I(2) and one
needs to difference the differenced series. Series that are integrated of order two or higher
are quite rare and in the subsequent studies and none of the dependent series has had to be
differenced more than once.
Step two: determining the autoregressive and moving average terms
If a stationary series is obtained, one can continue with the second step in the model
building process, determining the autoregressive (AR) and moving average (MA) terms.
Here, the earlier mentioned second problem is treated: correctly accounting for the series’
past. Autoregressive terms indicate the lagged endogenous variables that are added to the
model and resemble the effects of previous values of the series on the current value.
Moving averages represent the influence of residuals from specific previous values on the
current value. They resemble shocks (resulting in larger unexplained residuals or error
terms) in the series that affect later values of the same series at specific points in time.
Negative moving averages, for example, indicate that the error terms of specific previous
Framing Integration and Immigration
24
lags have a negative impact on the series’ current value. These negative MAs often show up
when modeling for example media attention for specific issues, where an increase in
attention caused by a newsworthy event results in a high value and a relatively high
unexplained error term. This increase in attention is often followed by a quick return to
initial values, resulting in a situation where higher error terms in the past result in sharper
decreases afterwards, which can be modeled with negative MAs in the ARIMA-framework.
Concretely, this means that the error terms of previous values of specific lags are added to
the model and that the coefficient for these MA-terms is taking a negative value.
Substantially, these results can be explained by essential features of the news gathering
process (see also Chapter 5).
Formally, an Autoregressive Moving Average (p, q) process with p autoregressive
terms and q lagged moving-average terms is defined as:
Eq 2-2
yt =
This formula formalizes the basic idea that the time series y is a function of a constant, its
lagged values, the influence of previous error terms and an unexplained part (the error term
or residual). The ultimate goal is to find a specification of the model that properly models
the series’ own past, without leaving any information from this past unused. Unused
information shows up as autocorrelation in the residuals, meaning that the residual at a
certain point in time is significantly correlated with residuals from previous values. In that
case, the residuals are not mimicking white noise, a term borrowed from natural sciences
indicating a signal or process that does not correlate in time. When a series resembles white
noise, it takes independent random values with a normal distribution. Therefore, the most
important criterion to establish whether a certain model is well-specified and the
information from the series past is properly included in the model is to test whether the
residual is indeed white noise and therefore autocorrelation in the residuals is absent.
In the model building process, one should strive for parsimony. Usually, different
ARIMA-specifications of a series may result in residuals that resemble white noise. Model
building is a trial and error process: fitting different ARIMA-models and getting the
residuals white noise, without including too many AR and MA-terms. Usually, model
building is done by comparing the pattern of the autocorrelation function (ACF; giving the
correlation of the series at different lag lengths) and the partial autocorrelations (PACF;
which measures the correlation between yt and yt-s controlling for the effect of intervening
values of y between t and t-s) of the original series and compare them to patterns typical for
different AR and MA-models (see Figure 2.2). This is indicative for the appropriate choice
of p and q and constitutes the crucial step in the estimation procedure. The autocorrelation
function of an AR(1)-process exhibits a geometrically declining pattern, falling down to
zero as the lag length becomes large. For an MA(1)-process the autocorrelation drops to
a0 +
∑
=
p
i1
ai*yt-i +
∑
=
q
i1
β
ββ
βi*ε
εε
εt-i +
ε
εε
εt
Constant Autoregressive
term (AR)
Moving Average
term (MA)
Error
term
Methodological Considerations
25
zero for all lag lengths larger than 1, and, more generally, drops to zero for lag length larger
than q for a MA(q)-process. The partial autocorrelation function mirrors the pattern of the
ACF. It is declining for an MA(1)-process and drops to zero at lag length bigger than 1 for
AR(1)-processes; the intuition here is that these past values do not influence the current
value directly, as they are not a part of the model, but only indirectly, through the dynamic
structure of the model (yt-2 influences yt-1, which in turn influences yt).
Figure 2.2 ACFs and PACFs for ARIMA (1,0,0) and ARIMA (0,0,1) models
ACF for ARIMA(1,0,0)
1 2 3 4 5 6 7
time
PACF for ARIMA(1,0,0)
1 2 3 4 5 6 7
time
ACF for ARIMA(0,0,1)
1 2 3 4 5 6 7
time
PACF for ARIMA(0,0,1)
1 2 3 4 5 6 7
time
Framing Integration and Immigration
26
The estimation procedure is fully empirically driven; an a priori theoretical reason is not
required to include the AR and MA-terms. Basically then, one fits models until the
residuals do not show any autocorrelation. In practical research, values of p and q of over 3
are rare with seasonality being the important exception: for monthly data the lag of 12
months should be carefully examined because many time series – and parliamentary and
media data are no exception - contain yearly cycles.
To test the absence of autocorrelation in the residuals, statistics like Ljung-Box Q are
preferred over the widely used Durbin-Watson statistic. Despite its popularity among social
scientists, the latter test has severe shortcomings: it only tests for first-order autocorrelation
and is not valid when the model includes autoregressive (AR) terms.
Additionally, the magnitude of change in a series might be time-dependent and the
variance serially correlated – the third potential problem mentioned above. In media
attention series, for example, this volatility occurs when periods of relative stable levels of
attention alternate with periods in which the attention is less stable and fluctuates sharply.
This is called autoregressive conditional heteroscedasticity (ARCH) and results in residual
variance that is not stable over-time. When this autoregressive conditional
heteroscedasticity is present in the series, it is necessary to model this ARCH-error process.
However, also data transformations like using logarithmic values of the original series can
be applied to deal with heteroscedasticity. To test whether heteroscedasticity is present in
the series, the Lagrange Multiplier test (Engle, 1982) can be used and autocorrelation in the
squared residuals can be tested using the Ljung-Box Q test. In the series used for this
dissertation, little evidence for autoregressive conditional heteroscedasticity is found and no
models from the ARCH-family are used. Therefore, I will not discuss them here in further
detail.
As mentioned before, more than one model might be suitable to capture the dynamic
structure of the series. Choosing between models -the model selection problem- can be
done by comparing the Akaike Information Criterion (AIC) and Schwarz Bayesian
Information Criterion (BIC) for the various models. These criteria can be seen as a measure
of goodness-of-fit, penalizing the inclusion of more parameters.
In the empirical analyses, I find most media series best presented by either (1,0,q) or
(0,1,q) models with one or two negative MA-terms at short lags. This exactly represents the
dynamic character of media data: it is inherently an autoregressive process (occasionally
even with a clear upward trend), where previous values determine the current value (AR(1)
process). But it is also a process in which after large shocks (resulting in large error terms)
the series tend to move back in the direction of initial values (negative MA-terms).
Step three: Adding explanatory variables
After a suitable specification of the univariate ARIMA-model is obtained, one can expand
the model and add exogenous, explanatory variables to the right-hand side of Eq 2-2. To
meet one of the basic criteria for causality, i.e. temporal consistency, a change in the
independent variable should precede a change in the dependent variable. Therefore, it is
Methodological Considerations
27
usually necessary to include lagged values of the independent variable in the equation. To
select the appropriate lag length in which this independent variable influences the
dependent variable one first uses theory: is a short-term (e.g. one month) influence to be
expected, or can this influence also occur at longer lags? The second step is then to choose
the best fitting lag length within the range that is theoretically expected based upon an
analysis of the cross-correlation function (CCF) of the independent and dependent variable.
Again, the residuals of the new model should mimic white noise. Furthermore, as with the
dependent variable, the independent variable should be stationary as well. Coefficients and
standard errors of the independent variables show whether the effect of the independent
variable on the dependent variable is significant, while the AIC and BIC indicate whether
the model has improved compared to the univariate models. For an even more elaborate and
technical discussion of ARIMA-modeling I refer to Hollanders and Vliegenthart. (2007).
M
ULTIDIRECTIONAL RELATIONSHIPS
:
VAR
What is it and when to use it?
ARIMA-models are suitable for assessing relationships between variables in one causal
direction. When expectations about this direction are not trivial or recursive and theory is
less informative about the characteristics of the relationships (e.g. delay, duration,
direction), other techniques, like VAR, are more appropriate (Brandt and Williams, 2007).
VAR stands for Vector Autoregression. In a VAR analysis, several equations are estimated,
each including a different dependent variable. The right hand side of each equation includes
lags of both the independent variable and lags of other variables. A VAR-model, including
two dependent variables y and z and constants c1 and c2, can be written as:
Eq 2-3
yt = c1 + a11*yt-1 + a12*zt-1… + a1k*yt-k + a1k*zt-k + ε1t
Eq 2-4
zt = c2 + a21*yt-1 + a22*zt-1… + a2k*yt-k + a2k*zt-k + ε2t
VAR-analyses are suitable for situations where both variables y and z are regarded as
endogenous. To put it differently, within a VAR-model both variables serve as independent
and dependent variables in separate equations. This method is especially suitable when
there are no strong a priori theoretical expectations about the causal direction of the
relationship between the two variables or when a mutual influence is expected. In Chapter
4, the relations between issue attention and framing in media and parliament are estimated
and in Chapter 7, the model is extended by also including the number of asylum
applications and anti-immigrant party support.
VAR-analyses offer several advantages over Simultaneous Equation Modelling
(SEM), a strategy probably most often applied by social science researchers dealing with
models that include more than one dependent variable. In fact, SEM can be regarded as a
form of a VAR-model with restrictions (Brandt and Williams, 2007). It is exactly towards
Framing Integration and Immigration
28
these restrictions that VAR-modelers direct their criticism. More specifically, in SEM,
choices about the inclusion and exclusion of different variables and lagged values have to
be made before estimation. Similarly, assumptions about exogeneity and endogeneity of
variables have to be made. Often, these assumptions are based on ‘vague appeals […] to
theory’ and/or are not well supported by empirical analysis’ (Brandt and Williams, 2007: 5-
6; Sims, 1980). Furthermore, these restrictions frequently do not allow for testing
alternative theories about causal relationships in one model.
A model with few restrictions is required for the case under study: based upon theory
and previous research, there are some expectations on the causal direction of the
relationships between for example media and parliament, but these expectations are in
many instances not so strong and empirical validation in encompassing models has been
lacking. A possible disadvantage of VAR compared to SEM is that the latter – if the
imposed restrictions are valid – gives more precise estimations of effect sizes and results
that are more convenient to interpret (Freeman et al., 1989). Another advantage of SEM is
that it allows for the estimation of contemporaneous relations between the variables, while
VAR only works with lagged values. More specifically, with SEM it is also possible to
assess the direction and size of relationships within time unit (i.e. months). Within the
VAR-framework, it is possible to assess whether an additional significant contemporaneous
relationship is present by looking at the correlations between the residuals ε1t and ε2t from
equations Eq 2-3 and Eq 2-4. To assess the direction of this relationship, however, it is
necessary to take a lower time unit (e.g. weeks, see also Chapter 7).
Step one: testing for stationarity
The model specification process is very similar to the one applied with ARIMA-modeling.
First, stationarity of the series has to be established using the (augmented) Dickey-Fuller
test. However, contrary to the ARIMA-model building process, if any of the series is non-
stationary, several scholars advise against differencing because it removes long-term
dynamics. They suggest the use of a different technique, Vector Error Correction Models
(VECM) (Brandt and Williams, 2007; Hamilton, 1994). A VAR-analysis can also be
processed if the series are non-stationary but co-integrated according to the Engle-Granger
test, indicating that the linear combination of two non-stationary variables is stationary.
This co-integration shows up in the observation that the series have a similar trend and
substantially means that they develop in a similar way over time. In the studies included in
this dissertation, the series are stationary and therefore, I will not go into detail about co-
integration and VECM.
Step two: Model specification
The next step is to establish the number of lags of the various variables to be included in the
equations. This can be determined using theory and various test-statistics. The number of
lags has to be equal for each endogenous variable. First, based on theory about the time lag
in which causal influences take place, one can limit the maximum number of lags. Second,
Methodological Considerations
29
within the boundaries established by this maximum number of lags, one can choose the best
model using fit-statistics like AIC and BIC or a Likelihood Test for Dropping Lags (Enders,
1996). As with ARIMA, one has to check whether the series’ past is adequately modeled
and thus for the absence of autocorrelation in the residuals. If autocorrelation in the
residuals of any of the series is present, additional lags for all variables need to be added to
the equations. Furthermore, also the Engle-test for ARCH-errors can be conducted. Finally,
residuals can be checked for the presence of contemporaneous correlation.
Estimation is done by ordinary least squares regression (OLS) and the system can
include more than two variables. Also additional exogenous variables can be added to the
equations, but again similarity for each of the equations is required.
Step three: Interpretation of results
Single coefficients of the effects of lagged values of the endogenous variables should be
interpreted with the highest caution because of the high levels of collinearity due to the
multiple lags included for each variable. In line with Freeman et al. (1989; see also Brandt
and Williams, 2007), the following outcomes of the analyses are reported to obtain a good
picture of the relationship between the various endogenous variables:
(1) F-tests for the whole blocks of lags of each variable, which indicates whether –overall-
any significant causal relationships between the independent variables and dependent
variable exists. Establishing causal relationships in this manner is based upon the notion of
‘Granger-causality’, where a variable y Granger-causes another variable z if the behavior of
y in the past improves the prediction of the behavior of z compared to z’s past alone
(Brandt and Williams, 2007: 32).
(2) Every finite autoregressive process can be rewritten as a infinite lagged moving average
model. A VAR can be rewritten in a Moving Average Representation (MAR). A moving
average representation of the variables in the system can help to acquire insight in the
consequences of a shock (impulse) in one variable at time 0 on the following values of the
other variables (see e.g. Figure 4.1). This helps to obtain a more specific picture and
graphical representation of the direction and size of the over-time effects of one series on
the other series and therefore contributes to the understanding in the dynamic interaction
between the variables. To obtain an impulse response analysis of the moving average
representation of the variables included in the system, it is necessary to construct a causal
pathway for which the variables are put in the most likely causal order (based on theoretical
considerations) and/or to test whether different causal orderings lead to similar results.
Framing Integration and Immigration
30
(3) A presentation of the decomposition of the Forecast Error Variance, indicating for each
variable over-time what portion of the movement in a series can be attributed to its own
shocks versus shocks from the other variables. In other words, this method estimates over-
time the amount of variation in each of the endogenous variables that can be attributed to its
own past and to the past of each of the other endogenous variables. As with the impulse
response analysis of the moving average representation, a causal ordering of the variables is
necessary.
31
Chapter 3. Divergent Framing: The Public Debate on
Migration in the Dutch Parliament and Media
1
3.1 Abstract
In this chapter we reconstruct how the issues of migration and integration have been framed
in the Dutch public debate over the last decade. We examine the patterns in both the
parliamentary and the media arena and look at similarities and differences between them.
On the basis of two contradictory theories, we formulate hypotheses about overlap and
differences between the two arenas and diversity within them. Our results reveal
incongruence in framing between the two arenas. After 9/11, however, the framing in
which Islam is perceived as a threat to Western society becomes dominant in both arenas.
Furthermore, we do not find any proof for the idea of the media as a civic forum, with a
high diversity of framing. Framing in parliament, instead, is far more diverse. In contrast to
the hegemonic framing in the media, the Islam-as-threat frame is actively contested in the
political realm.
3.2 Introduction
In the early 1990s, Dutch politician Frits Bolkestein caused controversy through his claims
about the negative impact of immigrants on the Dutch society. The Liberal-Conservative
leader emphasized the incompatibility between Islamic and Western values and argued for
more demanding policies to pressure immigrants to integrate into Dutch society and accept
its norms and values (de Volkskrant, September 12 1991). Bolkestein’s statements caused
some shock since until then, these views had been exclusively associated with the radical
right. Also, it was the first time a member of a major political party criticized what was
widely regarded as a successful policy approach.
Now, 15 years later, Bolkestein’s views have become mainstream. Many politicians,
both right- and left-wing, have publicly declared that Dutch integration policies have failed
and that urgent policy change is therefore required. The ideal of multiculturalism as a
policy goal appears to have been abandoned, and the multicultural society is perceived as a
failure (Entzinger, 2003). Moreover, this negative vision is accompanied by a shift in public
opinion on migrants. As a recent survey has revealed, Islam is increasingly perceived as a
threat to Dutch values (survey de Volkskrant, June 2004). These changes in the public
discourse are not unique to the Netherlands. The last decade could be labeled a ‘decade of
1
This chapter is based on: Roggeband, C. M., & Vliegenthart, R. (2007). Divergent Framing: The
Public Debate on Migration in the Dutch Parliament and Media, 1995–2004 West European Politics,
30(3), 524-548.
Framing Integration and Immigration
32
change’ (Luedtke, 2005) in relation the immigration and integration issues. Across Europe
a retreat of multiculturalism and a turn to civic integration is visible (Joppke, 2004;
Kofman, 2005). In the past few years more pressure is put on immigrants to conform and
integrate. This in particular affects the Muslim population, which after 9/11 and other
events is viewed as a group reluctant to integrate and accept political values (Modood,
2003).
In the Netherlands a number of recent events, like the rise of right wing politician Pim
Fortuyn, the provocative claims of MP Ayaan Hirsi Ali, and the murder of film director
Theo van Gogh, sparked off an active debate about the integration of minorities. Opinion
makers and politicians appear to agree that a dramatic change in the public debate has
occurred, but no research has yet been undertaken to trace exactly when and how this
change happened. We will try to fill this lacuna by reconstructing how the issues of
migration and integration have been framed in the public debate over the last decade.
Public debate is carried out in different forums: in the mass media, in parliament, in
public gatherings, in the streets and other places. We focus here on the two most central
forums: the political arena and the media arena. Within these arenas, the different actors
involved articulate particular ways of understanding issues. They define not only what
issues and problems need our attention, but also how we should look at them. The question
as to whether framing in the two arenas largely overlaps or differs is central in this paper.
Here we look at whether a similar shift or reversal occurred in both arenas. Furthermore, we
compare more substantial characteristics of the debate in the arenas, in terms of diversity
and consistency. From different views on the relation between media and politics we derive
contradictory hypotheses about the correlation between media framing and parliamentary
framing, about the diversity of the framing in both arenas and about the congruence of
ideologically different newspapers and use our data to test these hypotheses both
qualitatively and quantitatively.
The outline of the chapter is as follows. We start with a theoretical discussion about
the framing characteristics of, and possible relationships between, the two arenas. Second,
we introduce our methods and data sets. Third, we reconstruct the frames on migration and
integration presented in the political and media arena between 1995 and 2005. Finally, we
discuss our findings on the differences and similarities between the two debates and
examine possible explanations for these outcomes.
3.3 Media and policy framing
In this chapter, we compare how the integration of minorities is framed in the political and
in the media arena. The relationship between both arenas is a much-debated issue in social
sciences. In chapter 4, we demonstrate that a mutual influence between framing in
newspaper coverage and official parliamentary documents exists, but that this influence is
rather weak and dependent upon previous use of a frame: an increase in the use of a frame
in one arena leads to an increase in the other arena only if this frame has previously been
Divergent Framing
33
used regularly in the latter arena. In this chapter, we look more in depth at the nature of the
differences and similarities between the two arenas and try to explain these differences and
similarities. Additionally, we look at framing changes over-time in both arenas and what
can account for these changes.
One of the main questions when discussing the relationship between media and
politics is a normative one: to what extent should the media reflect the political debate in a
democratic society (e.g. Gunther and Mughan, 2000) (McQuail and Van Cuilenburg, 1983).
Drawing upon Schumpeterian notions of democracy and closely associated with the
Habermas’ ideal notion of the public sphere, political scientist Norris (2000) promotes the
idea of media as a civic forum. The media should allow for pluralistic competition among
parties and individuals for all positions in power. This specific conception of the role of the
media leads to the requirement of availability and balance of the news in terms of both the
amount of coverage of issues and the way these issues are covered. This diversity can be
reached in different ways, either by treating external diversity between various media
outlets that have strong political leanings or internal diversity, in which case the outlet does
not take a stance, but presents multiple and contrasting perspectives on a certain issue.
Norris presents a sequential model of political communication in which a source
communicates its message to the public through a channel (the media). The media act as
instrumental intermediaries between politics and public, and media content, in its
relationship to politics, is viewed as information. Though her theory of the media as a civic
forum is largely informed by normative notions, in her empirical cases Norris finds
considerable evidence for this view.
A contrasting vision is presented by Meyer (2002) who focuses rather on the media as
an active political player. Meyer argues that ‘the media diminish the opportunities that civil
society might have to exert influence on political inputs’ (p. 108). Politics and media are
two separate realms with different logics. The media logic, according to Meyer, is
contained in two filters. The first involves the selection of the news (what counts as
newsworthy) and the second the presentation (the means to trigger the audience). As a
result of the logic of the media, politics is increasingly ‘colonized’ by the media. According
to Meyer, ‘politics becomes dependent on the media’s rules, but without completely losing
its separate identity’ (2002: 57). Koopmans (2004) advances a similar idea and focuses on
the selection mechanisms of the media. Contrary to Meyer, Koopmans does not focus on
the media as player, but rather sees the media as a bounded space for political
communication. The available communication space is limited, while the number of
potential actors that want to occupy this space is disproportionately high. It is not only
formal political actors such as governments and political parties that insert messages in the
public sphere; a large number of other speakers such as interest groups, social movements
and opinion makers also try to get their messages across. This results in a high level of
competition between actors. The selection mechanisms that media use, according to
Koopmans, are similar to Meyer’s filters: novelty, prominence and prestige of speaker,
level of conflict and possibilities for dramatization. These filters, according to Koopmans et
Framing Integration and Immigration
34
al. (2005: 262), do not vary strongly within a country: ‘news values (e.g. which actors are
considered prominent and which issues relevant) are strongly affected by the institutional
and discursive opportunity structures in a country on a given issue’. Newspapers, however,
express their views and make their own contribution to political agenda setting and opinion
formation through their editorials (Koopmans et al., 2005).
3.4 Hypotheses
From these competing theoretical approaches to the relationship between the political arena
and the media arena, we can deduce different expectations about the relationship between
parliamentary framing and media framing of the issue of the integration of minorities.
Norris’ idea of media as civic forum informs the following three hypotheses:
I. Media reflect pluralist framing in the political realm and we expect a considerable
overlap between parliamentary framing and media framing of the issue.
II. Media framing is more diverse, because the media arena provides as a civic forum
space for a broader range of actors with diverging views on the issue to present their
opinions.
III. In view of the differences in political leaning of the different newspapers, we will
find some external diversity and differences between the various newspapers in the
framing of the issue according to their political position, with more right-wing
newspapers framing the issue more in accordance with right-wing politicians’
framing of the issue, whereas more left-wing newspapers will frame the issue more
in the way left-wing parties frame the issue.
Meyer and Koopmans’ idea that the media act as filters or selectors informs our next three
hypotheses:
IV. Politics and the media are two separate spheres, each with its own logic, and will
therefore significantly differ in their framing of the issue. The media will present the
issue more in terms of frames that have newsworthy characteristics (e.g. conflict and
dramatization), whereas in the political arena, where decisions about ‘what needs to
be done’ have to be made, frames with an emphasis on policy will be used.
V. Because media framing is characterized by more competition due to a larger number
of competing actors and limited space, only frames that resonate and are used by a
wide variety of actors will be covered in the media. This results in less variety in
media framing compared to parliamentary framing.
VI. In view of the active function of the media as political actors (Meyer) and the
freedom of the media to come up with their own position in political issues, we
expect considerable variation between newspapers. We expect that variation between
newspapers is greater than internal diversity and that dominant framing in the
different newspapers does not necessarily coincide with the framing of political
parties that have a similar leaning.
Divergent Framing
35
3.5 Data
Our research findings are based on two datasets, one consisting of parliamentary data and
the other of media data. These are gathered through both qualitative and computer-assisted,
quantitative content analysis.
For the political realm, we analyzed all official documents presented to and discussed
in parliament between January 1995 and the end of 2004. In this period, four different
coalitions governed. From 1995 to 1998 the first so-called ‘Purple’ cabinet took office,
consisting of the social-democratic party (PvdA), the conservative-liberal party (VVD) and
the progressive liberal party (D66). This was the first Dutch cabinet since World War II that
did not include the Christian democrats (CDA). The Purple coalition continued for a second
period from 1998 to 2002. In 2002, a new right-wing populist party (Lijst Pim Fortuyn
[LPF], named after its murdered charismatic leader) entered the Dutch electoral arena and
had a major victory in the parliamentary elections. This party formed a coalition with the
Christian Democrats and the Liberal Party, headed by Prime Minister Balkenende. This
coalition lasted only 87 days. In the ensuing elections, the LPF lost votes and a new centre-
right coalition of CDA, VVD and D66 took office in May 2003.
The policy texts that are presented to parliament can be found through the search
engine Parlando, which gives access to all documents presented to, and discussed in,
parliament and senate. Additionally we analyze the official written parliamentary questions
asked by parliamentarians to members of government, that are also available via Parlando.
To obtain data on the media debate on migration and integration, we analysed news
reporting of the five most-read Dutch national newspapers throughout the research period
(Bakker and Scholten, 2003): NRC Handelsblad, Algemeen Dagblad, De Volkskrant,
Trouw and De Telegraaf. Although completely independent of political parties, these
newspapers do have some different political leanings. De Telegraaf and Algemeen Dagblad
are both regarded as right of centre, popular newspapers that are read by a less educated
part of the population than the other newspapers (Bakker and Scholten, 2003). NRC
Handelsblad is a conservative, neo-liberal right-leaning quality paper, and de Volkskrant,
having originally a Catholic background, now has a more left-leaning, quality signature.
Trouw is a newspaper that can be regarded as Protestant-progressive, paying special
attention to Christian-religious and Third World issues (Van der Eijk, 2000). In the result
section, we make a distinction between popular (De Telegraaf, Algemeen Dagblad) and
quality (Trouw, Volkskrant, NRC Handelsblad) and between right-leaning (De Telegraaf,
Algemeen Dagblad, NRC Handelsblad) and centre/left-leaning (Trouw, Volkskrant)
newspapers.
All these newspapers are electronically available via LexisNexis for the period we
have studied, except for De Telegraaf, which is only digitally available from 1999 onwards.
Using framing in later years compared to the framing by the other newspapers in this period
and the data available for these other newspapers before 1998, we controlled for the
absence of De Telegraaf between 1995-1998.
Framing Integration and Immigration
36
3.6 Methods
To analyze the different frames that prevail in both arenas, we draw from social movement
theory (particularly Ferree et al., 2002; Snow and Benford, 1988) and policy (frame)
analysis (Rein and Schön, 1996; Van de Graaf and Hoppe, 1992; Verloo, 2005).
We start from the assumption that in both official parliamentary documents and
newspaper coverage different representations of the problem are articulated. These
representations include a diagnosis (what is the problem, where is it located and what/who
causes this problem?), connected to a prognosis (how should the problem be resolved, what
ends and what means should be used?) and a rationale or call for action (what courses of
action are suggested, and who is responsible for this?) (Snow and Benford, 1988). We
added the element of standing (Ferree et al., 2002) to this basic structure to locate the actor
associated with the frame. Ferree et al. (2002: 86) borrow this concept from legal discourse,
‘where it refers to the right of a person or group to challenge in a judicial forum the conduct
of another’. Standing, or having a voice, is a contested terrain and both in the media arena
and the political arena different actors compete to bring forward their interpretations and
visions. Standing refers to agency, who is given space to speak. Within policy analysis, Van
de Graaf and Hoppe’s breakdown of policy texts into their respective causal chains, final
chains and normative chains is particularly useful. Causal chains are the connections
between what is seen as a cause and what is seen as an effect of a certain policy problem.
Final chains are the connections between what is defined as goal and what the means are to
reach this goal. Normative chains are the connections between various levels of what is
seen as right and what is seen as wrong. For the reconstruction of all these chains, both
explicit and implicit interpretations are important.
The above-mentioned elements of diagnosis, prognosis, roles, causality, normativity
and finality have been translated into an analytic framework that consists of a coding
scheme with a set of sensitizing questions for each element (Annex 1). The first category is
about voice or standing and asks which actor speaks, on which occasion, to what audience
or forum, and in what form (interview, policy document, letter, essay, etc.). The second
category of diagnosis asks what is represented as the problem, why it is seen as a problem,
and what is mentioned as causes of the problem. It also analyzes the attribution of roles in
the diagnosis, such as who is seen to have caused the problem, and who is the problem
holder, who are possible victims and perpetrators. The third category concerns the
prognosis that contains what is represented as the solution to the problem, what goals are
formulated, how these goals should be achieved. It also analyzes the relationship between
ends and means. The fourth category is the call for action (or non-action), who has a voice
in suggesting the course of actions, who should be acting and who is acted upon?
These sensitizing questions are used to manually code the different positions of the
dimensions mentioned above in both arenas, starting from the assumption that different
frames may be presented within one policy document or newspaper article. This analytic
Divergent Framing
37
tool helps us to group ideas into frames that typically differ in what is presented as the
central problem or as the solution to this problem. Again, this is not to say that we expect to
find numerous combinations of framing elements resulting in huge amounts of different
frames, but these categories and questions help us to make our search for the main existing
frames more systematic.
This instrument was used, in a first inductive phase, for a qualitative analysis on the
major policy documents in the field of migration and integration (see Annex 2). We
identified five major frames used within the political realm between 1995 and 2005:
Multicultural frame
The multicultural frame, which is often presented as the prototypical Dutch cultural frame,
sees (cultural) diversity as an asset that enhances the quality of society. Policy interventions
to reduce inequalities in participation in education, labour and politics should respect and
foster diversity. Dialogue and civil participation are viewed as a central means to create
more understanding and respect towards cultural diversity. In this frame, the diagnosis is
little elaborated, it is mainly argued that there is insufficient space for diversity and that
discrimination and ‘glass ceilings’ may prevent migrants from entering the higher echelons
in society. Most emphasis is on the prognosis, promoting respect for cultural diversity and
creating more space for difference.
Emancipation frame
The emancipation frame problematizes the position and orientation of migrants: they are
viewed as backward in participation, customs and views. The prognosis views participation
in the labor market and in decision making as key to the process of emancipation. Obstacles
to participation should be removed by the government and state policies and arrangements
should help migrants in their integration and emancipation process. Migrants are expected
to actively shape their own emancipation process.
Restriction frame
The restriction frame sees the entrance of new immigrants as a problem, especially when
they are not economically independent. Implicitly, the emphasis is on new immigrants who
come to the Netherlands because of marriage with non-native Dutch residents. These
migrants are labeled as ‘newcomers’ or ‘import brides’. They constitute a problem not only
because of their economic dependence, but also because they often have a traditional, non-
emancipated orientation. It is an implicitly gendered frame, as in the political debate
politicians principally speak about female immigrants. Solutions presented are restrictions
on incoming migrants, and not granting independent permits to these new immigrants.
Framing Integration and Immigration
38
Victimization frame
The victimization frame is mainly applied in relation to women. Women are presented as
victims of a misogynous religious culture that forces them to be obedient and cover
themselves to be respected. The headscarf is presented as either a symbol of subordination
or something women can be, directly or indirectly, forced to wear. Violence against women
is the central mechanism used to reinforce gender inequality, and examples mentioned as
‘cultural violence’ are domestic violence, honour killings and genital mutilation. The
solution to this problem is mainly protection and more secondary emancipation of the
women (which is perceived as a way to reduce their vulnerability).
Islam-as-threat frame
In this definition of the problem, the culture and religion of migrants, particularly Islam, is
seen as a problem. Islam is presented as a threat to the values of Western civilization, such
as church-state relations, freedom of expression, equal gender relations and tolerance
towards homosexuality. Proponents of this frame fear the ‘Islamization’ of Dutch society,
which according to them is manifested by the growing number of civil servants wearing
headscarves. The proposed solution is to stress the neutrality of the state and to ban religion
to the private sphere. Also, the residues of Dutch polarizationi, especially in the school
system, are questioned. The creation of Islamic schools is seen as reinforcing segregation in
society.
A second phase in our analysis was to translate these frames into search strings to
make possible an electronic analysis of all parliamentary documents (including official
questions from parliamentarians to members of government) and newspaper articles
between 1995 and the end of 2004. Frames are distinguished using search strings to indicate
the presence of a given frame.ii This approach is based on one of the devices suggested by
Gamson and Mogdiliani (1989) that signify the presence of a frame, namely, catchphrases,
which Pan and Kosicki (1993) mention as an element of the rhetorical structure of a text.
The search strings were composed on the basis of the coding of the documents included in
our qualitative analysis, complemented with a random sample of both parliamentary
documents and newspaper articles (see Annex 4). For each of these documents and articles,
the coder established which frames were present and which word combinations were
indicative of these frames (see Annex 4). In the final search strings words are combined in
both additive (word X or word Y) and conjunctoral ways (word X AND word Y, or word X
within certain distance of word Y). Although a complete frame contains diagnostic,
prognostic and motivational components, we did not take that as an absolute requirement
for the presence of a frame. Especially in newspaper articles, framing is often incomplete
for several reasons, amongst others due to the limited length of the text.
We used our search strings to search for all official documents presented to parliament
between 1995 and 2005 that included one or more of our frames. A total of 5,376 frames
Divergent Framing
39
were found in the sample. Our use of the same search strings to analyze the media debate
on migration and integration resulted in a total of 14,850 articles that contained a mean of
1.09 frames per article.
To check the reliability of the search strings, a coder coded a sample of 30 articles
manually. Coder and computer agreed in 91% of the coding decisions, resulting in a
Krippendorf’s alpha of .72, which is more than acceptable (Krippendorff, 2003).
3.7 Results
Framing in two arenas
When we compare the overall attention given to the issue of immigration and integration of
minorities in Dutch newspapers and parliament, we can see that attention given to these
issues fluctuates strongly over-time. Figure 3.1 presents the total yearly number of
newspaper articles and parliamentary documents using the five frames.
Figure 3.1 Attention to the issue of immigration and integration of minorities in
parliament (number of official documents) and newspapers (number of articles)
Media framing seems especially vulnerable to specific events and shows clear peaks during
2001 and 2002, probably caused by the attacks on the World Trade Centre in New York.
After a decrease in 2003, the murder of film director Van Gogh (November 2004) leads to
increased attention to the issue of immigration and integration of minorities in 2004. In
parliament we can see a gradual increase in the attention paid to immigration and
integration issues, with two slight declines in 1998 and 2002. In general, on a monthly
level, for both media (r = .55) and parliament (r = .44) a positive correlation exists between
0
500
1000
1500
2000
2500
3000
3500
4000
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
media
0
50
100
150
200
250
300
350
400
450
500
parliament
media parliament
Framing Integration and Immigration
40
the point of measurement and the number of documents, which indicates an overall increase
overtime. Also in Figure 3.1 a general increase in overall attention to the issue is visible.
Framing in the political realm
Our findings demonstrate some important shifts in policy framing during the period 1995-
2004. As Figure 3.2 shows, different frames have dominated the political realm over-time.
Figure 3.2 Yearly use of each frame in official parliamentary documents
During the beginning of our research period (1995-1997), emancipation and multicultural
framing dominated in parliament, and both multiculturalism and emancipation through
participation were the official policy goals. The multicultural frame dropped after the right-
wing coalition of Balkenende I took office in 2002. The emancipation frame was the
dominant Dutch policy frame until 2002, and, although it lost its dominance, it remains a
very important policy frame. In 1997, the restriction frame gained ground in parliamentary
documents. Discussions about the limited integration of considerable parts of the migrant
population led to new legislation for gaining citizenship. Finally, the victimization frame
strongly fluctuated in official parliamentary documents throughout the whole period, with
peaks in 1996, 1999, 2001 and a continuous ascent since 2003. After 2002, the Islam-as-
threat frame, the frame that emerged in the political debate after the 9/11 attacks, became
the dominant frame in policy documents and parliamentary debates.
0
20
40
60
80
100
120
140
160
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
multicultural emancipation restriction
victimization Islam-as-threat
Divergent Framing
41
These different frames are promoted by different actors. In the mid 1990s, there was
significant political consensus about the policy goals of emancipation and multiculturalism.
In 1996, the Minister responsible for Minority Policies, Dijkstal (VVD), even noted that
presumably ‘We all agree that minorities should be given the perspective of full
participation in society and that pluriformity, mutual respect and maintenance of cultural
identity, solidarity, tolerance and integration should be the core concepts of our policy’
(Dutch Parliamentary Records, HTK 1995-1996, 24401, no. 15: 7).
From 1997 onwards, a growing left-right divide became visible over the issues of
migration and integration. The second Purple cabinet decided to change the direction of its
integration policies and appointed a special Minister for Integration Policies. Although new
Minister Van Boxtel (D66) defended the policy logic of multiculturalism, he also criticized
earlier policies for not sufficiently shaping the ‘active citizenship’ of migrants (Dutch
Parliamentary Records, HTK 1999-2002, 26333, no. 2). One important new instrument was
the law on the integration of newcomers (Wet Inburgering Nieuwkomers), which aimed to
actively integrate new migrants as citizens. Whereas the Liberal coalition partner, VVD,
promoted a more restrictive immigrant and integration policy, the other two coalition
partners, the Social Democrats and the Democrats 66 opted for a less stringent policy. The
restriction frame was supported by the Christian Democrats (CDA) and some other more
right-wing parties.
The right-wing coalitions headed by Prime Minister Balkenende dismissed the
multicultural frame and instead promoted a more assimilatory emancipation frame
(Entzinger, 2003; Kofman, 2005). Also, a new Coordinating Directorate for the Integration
of Minorities was created inside the Ministry of Justice, and this resulted in a more legal
framing in integration issues (Pakes, 2004).
Table 3.1 shows which frames dominate in the formal questions members of
parliament posed to the government. It reveals a clear left-right divide in the framing of
political parties. Emancipation and multiculturalism are frames mainly used by left-wing
parties, whereas Islam-as-treat predominates among right-wing parties. The framing of the
political parties in the middle is less clear-cut, the emancipation frame slightly
predominates, but other frames such as the restriction frame and the Islam-as-threat frame
are frequently used as well.iii
Table 3.1 Percentage use of each frame in questions asked in parliament between 1995
and 2004
Left (SP,
GroenLinks, Pvda)
Centre (CDA, D66)
Right (VVD, CU,
SGP, LPF)
Emancipation 47% 33% 22%
Multiculturalism 17% 13% 3%
Restriction 15% 21% 2%
Victimization 5% 10% 6%
Islam-as-threat 16% 23% 67%
Total number of questions (units) 126 67 90
Framing Integration and Immigration
42
Although the Islam-as-threat frame has become the dominant frame in the political realm
since 2001, it is a highly contested one. Since 2002, all frames have been rising in absolute
terms in the political realm. The emancipation frame, in particular, competes with the
Islam-as-treat frame. Also, the multicultural frame is rising but has not reached its pre 9/11
peak.
Framing in the media
The evolution of frames in the media arena presents a different picture compared to that in
political debate. Overall, the attention in the media to the issue of immigration and
integration of minorities has increased dramatically during past decade, reaching a peak in
2002 (see Figure 3.3). Remarkably, and in contrast to the image of the Netherlands as a
country that cherishes the ideal of multiculturalism, our findings demonstrate that the frame
that focuses on Islam as a threat has been dominant in the media debate throughout the
whole period under study, except for the period between 1998-2000. Partly this is due to
inclusion of reporting on foreign events, especially in the Middle East.
Figure 3.3 Yearly use of each frame in newspaper articles
However, the media frame the issues of integration and migration in Western societies - and
more specifically the Dutch society - mainly as issues concerning Islam and Muslim
migrants. It is important to note that this is not a new trend, but rather a tendency that has
been present since the early 1990s (Baumann, 1998; Snel, 2003). According to Baumann,
the European media framed the Rushdie controversy in the early 1990s as a polarization
between Islam and the Western world. In these representations, Muslims were forced to
choose between Islam or freedom of expression, which was defined as a central value of the
Western world (Baumann, 1998). This disjunction between Islam and the Western world
0
200
400
600
800
1000
1200
1400
1600
1800
2000
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
multi-cultural emancipation restriction
victimization Islam-as-threat
Divergent Framing
43
has also been actively constructed within the specific Dutch media debate since the early
1990s (Prins, 2005; Snel, 2003). VVD leader Frits Bolkestein and publicist Pim Fortuyn,
who later became a key figure in the Dutch political landscape, used the media to express
their discontent with the Dutch multicultural policy model. Both claimed that the values of
the Islam are largely incompatible with the values promoted by the Western Enlightenment.
According to Bolkestein and Fortuyn, the Dutch policy model that promoted the
preservation of cultural identity and values threatened the integration process of migrants in
order to conserve their own cultures. Muslims living in the Netherlands should be forced to
assimilate to Dutch norms and values (Prins, 2005).
The second most frequently used frame is the multicultural frame. This frame
dominated in the newspapers between 1998 and 2000. Remarkably, in this period several
well-known PvdA (social-democratic) members such as Paul Scheffer, Arie van der Zwan
and Ayaan Hirsi Ali started to criticize the multicultural model in the press. In 2000, Paul
Scheffer, a publicist and prominent PvdA member, became a core figure in the media
because of his article ‘The multicultural drama’ (NRC Handelsblad, 29 January 2000).
Scheffer argued that the policy ideal of multiculturalism had failed. Politicians had ignored
the formation of an underclass of migrants with little prospect of integration. Immigrants
themselves had been too passive in their efforts to integrate into Dutch society. Contrary to
Bolkestein and Fortuyn, Scheffer did not stress Islam as a central factor. Scheffer’s essay
had a direct impact on the political debate. In April 2000, a plenary debate in parliament
was held to discuss Scheffer’s statement that integration policies had failed (Dutch
Parliamentary Records, HTK 70, 18 April 2000). This debate was requested by PvdA
leader Melkert after VVD leader Dijkstal had criticised coalition-partner PvdA for not
backing certain policy measures to improve the position of minorities. In the plenary
debate, the leaders of most political parties praised Scheffer for pushing the issue of
integration higher up the political agenda. However, PvdA leader Melkert also criticized
Scheffer’s negative vision. Rather than calling integration a drama or failure, he preferred
to speak of a ‘serial full of struggle’ (NRC Handelsblad, 19 April 2000).
The terrorist attacks of 9/11 caused an enormous peak in newspaper reports that
focused on Islam as a threat to the Western world. In trying to reconstruct the causes of
these attacks, the media often attributed the blame to radical Islamists and their spiritual
leadership. On the one hand it was stressed that radicalization should be seen as a deviation
of a small group; on the other hand Islam was presented as a monolithic culture
incompatible with Western cultural values. This presentation of a cultural gap reinforced
the image of the Muslim as a potentially dangerous ‘other’. In this debate, Ayaan Hirsi Ali,
a young Dutch Somali woman, played a central role. She was first a member of the Social-
Democratic Party, but switched to the Conservative-Liberal Party because she no longer
agreed with the Social Democratic stance on immigration issues. She castigated Islam for
being an authoritarian religion and especially backward when it came to the position of
women (Ghorashi, 2003; Prins, 2005; Roggeband and Verloo, 2007).
Framing Integration and Immigration
44
Other frames important in the political realm, such as the restriction frame that
dominated in 1997, remained marginal in the media. The victimization frame, which
strongly fluctuated in the political debate throughout the whole period, with peaks in 1996,
1999, 2001 and a continuous ascent since 2003, has been very gradually rising in the media.
Since 2003, this frame has become the second most dominant after the Islam-as-threat
frame.
It is important to note that the actors that gain standing in the media are not necessarily
formal political actors. Newspapers appear to provide a forum for some critical outsiders.
However, these outsiders are closely related to the formal political arena (party members or
new candidate politicians). So, where in the political realm the consensus on
multiculturalism as a policy ideal still prevailed between 1995 and 1998, some right-wing
politicians and publicists used the media to criticize this dominant policy frame. For critical
outsiders who seek to contest the dominant news frame, it appears to be far more difficult to
gain standing. According to some, it has become almost impossible to present more
moderate viewpoints (PvdA MP Albayrak in de Volkskrant, 3 April 2004).
More pluriformity in the civic forum or media as selector of frames?
According to our first hypothesis, the media reflect the pluralist framing in the political
realm, and therefore a considerable overlap exists between parliamentary framing and
media framing of the issue. As has become clear from Figure 3.2 and Figure 3.3, the frames
used in both arenas differ to a large extent. However, do they follow the trends? In other
words, if one frame is going up in one arena, does it also go up in the other? Table 3.2
shows the correlation of the total attention to the issue and the proportional use of each of
the frames between the media arena and the political arena. Since the institutionalized
character of the parliamentary arena might make it slower in reacting to all kind of external
developments causing framing change than the media arena, we allow the media to have a
maximum lag of six months. This is in line with outcomes of classical agenda setting
research indicating influences up to a couple of months between various agendas (Dearing
and Rogers, 1996). The largest coefficients for each frame within this time span are
reported. In contrast to our first hypothesis, we do not find evidence here either for a large
overlap between the two arenas. Only the total attention and the proportional use of the
Islam-as-threat frame are correlated to a considerable extent at the same time lag. This
correlation is likely to be caused by external events such as 9/11 and, in general, there does
not seem to be that much overlap between the two arenas. Other frames show smaller
correlations on longer lags or even no significant correlations at all (emancipation frame). If
we take into account the fact that both arenas are subject to the same external events and
might influence each other as well (chapter 4), it is legitimate to call the overlap fairly
small.
Divergent Framing
45
Table 3.2 Correlation between total attention and proportional use of each frame in
parliament and media (monthly values)
Total attention (no lags) .287***
Multicultural Frame (media lagged 4 months) .204**
Emancipation (media lagged 1 month) .020
Restriction (media lagged 1 month) .145*
Victimization (media lagged 1 month) .179**
Islam-as-threat (no lags) .358***
Note: N=120 months; *** p < 0.01; ** p < 0.05; * p < 0.10 (one tailed test); Lag selection based on
cross-correlation functions (CCF)
A final way to look at the congruency of both arenas is to test whether earlier-mentioned
events and persons, that from our qualitative analysis turned out to be important in shaping
the debate, changed the framing of the issue in both arenas in similar ways. More
specifically, we look at how the debate changed after Scheffer’s newspaper article on the
‘multicultural fiasco’ (January 2000), after 9/11 (September 2001), after the appearance of
Fortuyn (February 2002)iv and his assassination (May 2002), and the election of Hirsi Ali in
parliament (January 2003). To do this, we split our research period in six sub-periods:
(1) Purple era (January 1995-January 2000)
(2) Post-Scheffer (February 2000-August 2001)
(3) Post-9/11 (September 2001-January 2002)
(4) Fortuyn (February 2002-May 2002)
(5) Post-Fortuyn with his party LPF in government (June 2002-December 2002)
(6) Hirsi Ali (January 2003-December 2004)
Though we acknowledge that these events and actors were far from the only ones
determining the framing during a certain period in time, they are regarded as crucial and, if
parliament and media overlap in their framing trends, they should have caused at least some
similar changes in both arenas. Table 3.3 presents the framing in both arenas during each of
the six periods.
Framing Integration and Immigration
46
Table 3.3 Different use of frames during different periods in parliament and media
(monthly values)
pre-
Scheffer
post-
Scheffer post-9/11 Fortuyn
post
Fortuyn Hirsi Ali
N (months) 61 19 5 4 7 24
Parliament
F-score (one
way Anova)
Multicultural 23% 22% 25% 21% 19% 16% 3.179**
Emancipation
33% 30% 24% 27% 28% 23% 3.862***
Restriction 22% 29% 13% 22% 29% 24% 3.365***
Victimization 14% 14% 10% 12% 8% 14% 1.377
Islam-as-
threat 9% 5% 29% 18% 17% 24% 42.333***
Attention 20 24 33 18 21 31 3.749***
Media
Multicultural 32% 44% 16% 35% 24% 21% 15.989***
Emancipation
4% 5% 1% 5% 9% 6% 4.747***
Restriction 4% 6% 3% 7% 5% 6% 1.505
Victimization 14% 19% 6% 11% 11% 17% 4.675***
Islam-as-
threat 47% 27% 74% 42% 52% 49% 14.108***
Attention 72 101 397 182 337 209 33.238***
Note: N=120 months; *** p < 0.01; ** p < 0.05 (two-tailed test); for each arena columns sum up to
100% per period, F-scores test for the use of each frame and total attention whether significant
changes across periods exist. Attention is expressed in mean number of frames per month.
The F-scores in Table 3.3, comparing means across periods, show that all frames, except
for victimization in media and restriction in parliament, differed significantly across the
various periods under study. Also, we find that the eventual changes in framing in each
period are not similar in both realms (with the exception of the post-9/11 period).
In period 1 (January 1995-January 2000), there is a predominance of the
emancipation and multicultural frame in parliament, whereas in the media the
Islam-as-threat frame is the most frequently used, followed by the multicultural
frame.
In period 2, after the publication of Scheffer’s essay on the ‘multicultural fiasco’
(February 2000-August 2001), we see few changes in parliamentary framing (the
restriction frame is going up slightly), whereas some changes in media framing
can be observed. Scheffer’s essay appears to have had some effects on the media
framing as we can see an increase in the number of articles discussing the
Divergent Framing
47
multicultural frame. Also, there is a slight increase of the victimization frame. The
Islam-as-threat frame is less prevalent during this period.
In period 3, after the events of 9/11 (September 2001-January 2002), there is an
increase of the Islam-as-threat frame in both the media and parliament, although
this change is considerably larger in the media.
Period 4 is the period when Pim Fortuyn launched his election campaign
(February 2002-May 2002). The framing in both parliament and media in this
period is very similar to the pre-9/11 period, with exception of the Islam-as-threat
frame being more prevalent than prior to 9/11.
Period 5 is the period after Fortuyn was murdered and his party LPF formed part
of the new right-wing government Balkenende I (June 2002-December 2002). In
this period, we can see a decline of the multicultural frame in the political realm.
In the media, the emancipation frame is slightly increasing, also the Islam-as-
threat frame (probably due to the retrospection on 9/11).
In period 6 (January 2003-December 2004), Ayaan Hirsi Ali became a member of
parliament for the Conservative-Liberal party (VVD). Her voice appears to have
affected the framing in the political realm. Our findings show an increase in the
frames promoted by Hirsi Ali (victimization and Islam-as-threat) in parliament
(Roggeband and Verloo, 2007), but this increase is only weakly reflected in the
media framing.
Our findings thus suggest few similarities in framing trends in both realms. Some events or
persons have had more impact on the framing in the political realm as is the case with Hirsi
Ali and the Lijst Pim Fortuyn (these are mainly political events). The publication of
Scheffer’s essay, which can be considered a media event, appears to have caused more
discussions in the media than in the political realm. We therefore refute the first hypothesis.
The fourth hypothesis, on the contrary, is largely confirmed by our findings: the
incongruence between framing in both arenas (in number of frames and in timing) supports
Meyer’s idea that politics and the media are two separate realms, each operating according
to its own logic. We indeed found that in the media arena newsworthy frames dominated
most of the time, whereas in the political arena more policy-related frames circulated. This,
however, changed after the 9/11 attacks. From 2002 onwards, the debate in the political
arena also focuses principally on Islam as a threat to Dutch identity and values.
We now turn to our second and fifth hypotheses. The second hypothesis predicted
more diversity in the media framing compared to the political arena, because, as a civic
forum, it provides space for other actors with diverging views on the issue. We do not find
support for this thesis either. The framing in the political realm is more diverse, whereas the
framing in the media appears to have an almost hegemonic character. Table 3.4 shows for
the parliamentary and the media arena the mean of the monthly frames necessary to
encompass 50 per cent and 75 per cent of the monthly total coverage of the issue.
Furthermore, it presents the same statistics for the five separate newspapers.
Framing Integration and Immigration
48
Table 3.4 Frame diversity in parliament and media (monthly values)
Source No. of frames 50% of
coverage
No. of frames 75% of
coverage
Parliament 2.05 3.23
Media total 1.57 2.35
T-test for equality of means 8.277*** 12.729***
NRC Handelsblad 1.44 2.25
Algemeen Dagblad 1.25 2.02
Volkskrant 1.48 2.34
De Telegraaf 1.21 1.87
Trouw 1.45 2.02
F-test (one way Anova) 6.692*** 8.963***
Note: N=120 months, except De Telegraaf (N=72); *** p < 0.01 (two tailed test)
The most resonant frame is the Islam-as-threat frame and is adopted across a wide range of
different newspapers. For all newspapers, it takes on average less than 1.5 frames to reach
50 per cent of the coverage and around 2 frames to reach 75 per cent. There are, however,
significant differences between quality newspapers on the one hand and popular ones on the
other, with the quality newspapers being more diverse in their framing. In particular, De
Volkskrant scores considerably higher than, for example, De Telegraaf (1.48 and 2.34
compared to 1.21 and 1.87). Although the media show some frame variation, selection
mechanisms indeed appear to be at work.
If we compare the total of the media arena to the parliament, we see that the framing in
parliament is considerable more diverse: t-tests show that the difference in means between
both arenas is highly significant. This difference is even more striking if we compare
individual newspapers with parliamentary framing. The results confirm the view of the
media arena as being highly competitive and selective in the use of (hegemonic) frames
(hypothesis 5). This suggests that not all actors have equal access to the media to express
their views, but that actors that promoting hegemonic frame have more standing.
Our third hypothesis predicts external diversity and differences between the various
newspapers in framing of the issue according to their political position, with more right-
wing newspapers (mainly NRC Handelsblad and De Telegraaf) framing the issue more in
accordance with right-wing politicians’ framing of the issue (e.g. more use of Islam-as-
threat frame), whereas more left-wing newspapers (mainly de Volkskrant) will frame the
issue more in the way left-wing parties frame the issue (e.g. multicultural framing and
emancipation framing). Table 3.5 presents the proportional use of each of the five frames
by each of the five newspapers, while Table 3.4 indicates the frame diversity of each
newspaper.
Divergent Framing
49
Table 3.5 Framing use by separate newspapers
Newspaper Multicultural Emancipation Restriction Victimization Islam-as-
threat
NRC Handelsblad 29% 4% 5% 15% 47%
Algemeen Dagblad
24% 3% 7% 14% 51%
Volkskrant 34% 5% 4% 15% 43%
De Telegraaf 30% 4% 7% 17% 42%
Trouw 33% 5% 5% 15% 42%
F-value 5.32*** 2.16* 3.60*** 0.72 3.96***
Note: N=120 months, except De Telegraaf (N=72); *** p < 0.01; * p < 0.10 (two tailed test); rows
sum op to 100% for each newspaper
The results of this analysis offer some support for the third hypothesis: there is indeed a
difference between the more right-wing newspapers (NRC Handelsblad, Algemeen
Dagblad and De Telegraaf) compared to the mainstream and left-leaning newspapers (de
Volkskrant and Trouw). The more right-wing newspapers make in general more use of
restriction and Islam-as-threat framing, framings more often associated with right-wing
politics: Algemeen Dagblad scores highest on both: 0.07 and 0.51 respectively, while NRC
Handelsblad scores 0.05 and 0.47. Both Trouw and de Volkskrant score lower, especially
on the latter: 0.43 and 0.42. They make more use of the multicultural and emancipation
framing. De Telegraaf is an exception and shows relatively high multicultural coverage and
low Islam-as-threat coverage. This is largely due to the fact that De Telegraaf is only
included from 1999 onwards. If we conduct an analysis for the various newspapers since
1999, we find that the differences between Islam-as-threat are not significant (De Telegraaf
at the same level as NRC and Volkskrant [0.43], Trouw lower [0.38], Algemeen Dagblad
higher [0.47]) and the differences in multicultural framing decrease, with De Telegraaf
becoming the lowest multicultural framer after Algemeen Dagblad. For four of the five
frames, the differences between newspapers are significant. However, if we correlate the
proportional use of each of the frames by each of the newspapers (not presented here), we
find high correlations, especially for the frames that are used regularly; this means that the
various newspapers largely follow the same patterns in frame use over-time and differences
are rather limited.
Another interesting difference between the various newspapers is in frame diversity.
Table 3.4 shows that the popular newspapers (Algemeen Dagblad and De Telegraaf) are
less diverse than the quality newspapers (NRC Handelsblad, de Volkskrant and Trouw). So,
quality newspapers appear to come closer to the ideal of a civic forum than more popular
newspapers. The total media arena is slightly more diverse than the separate newspapers;
this, again, indicates some level of variation across these newspapers as well.
So we find some support for our third hypothesis and for the idea that newspaper
differences should not be neglected. However, we cannot confirm our sixth hypothesis:
newspapers follow largely the same framing trends and the differences between the various
newspapers seem to be rather small, compared to the differences that exist within
Framing Integration and Immigration
50
newspapers. Furthermore, differences in newspapers framing are in line with their political
leaning.
3.8 Conclusions
In this paper we have reconstructed the framing of the issues of migration and integration in
the Dutch political arena and the media during the last decade. We have found quantitative
(issue attention) and qualitative (framing) shifts in both realms. While the attention for the
issues of migration and integration in parliament is more stable throughout the whole period
(with yearly cycles) and has risen somewhat since 9/11, the attention in the media has risen
enormously since 2001. In both domains, since 2001 there has been an emphasis on
immigrant religious culture, equated with Islam that is perceived as a threat to Dutch
culture and values and as a principle obstacle to the integration of Muslim migrants.
More remarkable are the differences in framing between the political realm and the
media. We found that (a) there is relatively little correlation between both spheres, (b) we
see divergent shifts in framing in both realms during different periods in time (following
important events in society) and (c) there are differences in frame variation between the two
arenas. Although the Islam-as-threat frame has dominated the media debate throughout
most of the period we have studied, this frame was rather marginal in the political realm
before the events of 9/11. In the political realm, there is a larger diversity in the frames
used, and different frames dominated over-time. During most of the time, the issues of
migration and integration were framed in terms of emancipation. In 1996/1997 a restriction
frame dominated, and since 2002 a frame that points to Islam as a threat and an obstacle to
integration has become dominant. However, our findings indicate that this frame is heavily
contested in the political realm. It are mainly right-wing parties that promote the Islam-as-
threat frame, whereas left-wing parties continue to support emancipation and multicultural
frames.
The incongruence between framing in both spheres supports Meyer’s claim that the
political arena and the media are to an important extent separate spheres, with different
agendas. Parliamentary framing overlaps only to a limited extent with media framing. We
did not find support for the idea that the media act as a pluralist forum. In contrast, our
findings rather indicate that media framing is highly selective, resulting in hegemonic rather
than pluralist framing. This suggests that not all actors have equal access to the media to
express their view, but that actors that promote or support the hegemonic frame have more
standing. It is rather that the selection mechanism of the media prevents deviant actors from
taking part in the discussion than that the media are the open arena that Norris (2000)
claims them to be. However, we found that some quality newspapers allow for more
diversity than more popular newspapers.
Pluriformity in framing in the political arena appears to be greater than in the media,
with each of the frames being used on a considerable number of occasions. Our results
show that this frame completeness in the political arena is not reflected in the media. This
Divergent Framing
51
result might lead to a non-representative picture in the media that might affect public
perceptions of the issue. Also, as Guiraudon (1998: 292) argues, the ‘negative dynamic
between the press, public opinion, and the electioneering politician’ may lead to more
restrictive policies. Similarly, Givens and Luedtke (2004) demonstrate that an increase in
media attention to the issue of immigration is related to more restrictive immigration
policies. Our findings demonstrate that the turn to more restrictive policies in the
Netherlands occurred well before the boom of media attention on migration issues. The
dominant framing in the media of Islam as a threat, however, appears to have diffused to
the political realm since 2001. In contrast to the hegemonic position of this frame in the
media, the Islam-as-threat frame is highly contested in the political realm.
Meyer’s (2002) claim of separate spheres with different dynamics appears to apply
particularly well to the issue of immigration and integration of minorities, which
simultaneously and paradoxically are becoming more international and national issues.
Whereas there is a continuing international debate about the ‘clash of civilizations’ that
Western nation states are facing, the regulation of immigration and integration remains the
territory of the state, the level of Europeanization of immigration policy remains low
(Luedtke, 2005) and the recent turn to civic integration may even reinforce nation-statism
(Kofman, 2005).
53
Chapter 4. Press and Parliament: Relationships Between the
Debate in Media and Politics
2
4.1 Abstract
In this article, we examine how the attention and framing of political issues in the press and
in parliament influence each other and how this attention and framing is influenced by key-
events outside the media and parliamentary realms. Our case is the debate on immigration
and integration in the Netherlands between 1995-2004. Our empirical analyses are based on
a computer-assisted content analysis of both parliamentary documents and newspaper
articles. Results show bi-directional causal relationships between media and parliament. In
the case of attention only long-term influence relationships are found, while framing
influences follow an interesting pattern: an increase in the use of a frame in one arena leads
to an increase in the other arena only if this frame has already been used regularly in the
latter arena. External events have more considerable and consistent impacts on issue
attention and framing in both arenas.
4.2 Introduction
In recent years, the issue of immigration and integration of minorities have been central to
political and public debates in most Western European countries. International
developments such as 9/11, the war in Iraq and recent attacks in Madrid and London not
only changed the amount of attention on the issue, but influenced the tone and focus of
these debates as well. In the Netherlands, heightened attention for the immigration and
integration issue and shifts in the framing of this issue are also due to internal political and
societal developments – mainly the rise and assassination of new politician Pim Fortuyn
and the murder on film maker and criticaster Theo van Gogh – making this country a
compelling case to study the complexity of framing processes.
In this relatively turbulent environment, multiple actors try to define events by trying
to bring forward their diagnoses and prognoses of the problems involved. In this article, we
focus on two central arenas in which these framing processes take place, the parliamentary
and the media arena. In the first, the actors involved are elected representatives who interact
with the executive power. In the latter, elected officials play a role, but also various other
actors (e.g. opinion leaders, journalists, social movements) strive for a place. Due to this
partial overlap in actors, but also because these arenas are vulnerable to all kind of external
influences, it is likely that causal relationships exist between both arenas. Here, we ask in
2
This chapter is based on Vliegenthart, R., & Roggeband, C. M. (2007). Framing Immigration and
Integration: Relationships between Press and Parliament in the Netherlands. International
Communication Gazette, 69(3), 295-319.
Framing Integration and Immigration
54
which arena which framing is propagated and how each arena influences the other. We
track the development of five frames and how frequently they are used in each arena. We
also explore how key national and international events influence the media and
parliamentary framing.
We collected both newspaper articles on immigration and integration in the five
national newspapers and official documents from the Dutch parliament between 1995 and
2004 and analyzed these using computer-assisted content analysis. First, we established the
occurrence of different frames over-time in each arena. Then we used VAR-modeling to
look at causal relationships between the two arenas, while controlling for major external
events. Results offer insight into frame-formation processes for this specific case, but could
potentially be generalized to other countries and other politically contested issues.
4.3 Relationships between parliamentary and media framing
Although there are several studies available on the relations between political agendas and
media coverage, this research has mainly focused on the question of which issue(s)
(attention) are presented and not on how these issues are presented. This might be the case
for several reasons. First, convincingly assessing relationships between both arenas requires
labor-intensive coding that spans a relatively long period of research. Second, the dominant
theoretical concepts and approaches have directed researchers into a certain direction,
which has left some questions unaddressed. For instance, one of the central concepts in
political communication during the past decades has been agenda setting (1972). Stemming
from this theory is the idea of agenda building, which deals with how the media agenda is
created (Dearing and Rogers, 1996). The focus is on which issues make it on the agenda
and how salient they are, rather than how they are presented.
During recent years the concept of framing has become increasingly present in the
social sciences. Framing helps us to understand how salient issues are presented. Scholarly
attention to this concept has grown rapidly, not only in the field of political communication,
but in various other social science sub-fields as well (e.g. Benford and Snow, 2000;
Entman, 1993; Valkenburg et al., 1999). No consensus has been reached about how
framing should be defined and operationalized in a media setting. Current
conceptualizations of framing vary in complexity (for an overview see De Vreese, 2005).
Most of these conceptualizations, however, are at least partly in line with the classical
definition of the framing concept offered by Entman (1993: 53), who speaks of ‘[selecting]
some aspects of perceived reality and make them more salient in a communicating text, in
such a way as to promote a particular definition of a problem, causal interpretation, moral
evaluation, and/or treatment recommendation for the item described’. Little is yet known
about the relations between framing in the media and in the political realm. The agenda
setting literature shed a light on this question, since it deals with the same overarching
question: Who follows whom? Though agenda setting and framing are inherently different
concepts, the underlying mechanisms for influence relations between media and parliament
Press and Parliament
55
might be similar. On the one hand, agenda setting deals with which issues are on the
agendas and which are not and can be seen as an ‘inter-issue’ contest between various
actors promoting different issues. On the other hand, framing deals with the different way
these issues are presented, which may be called an ‘intra-issue’ contest. However, in both
instances actors try to influence agendas and make them as favorable as possible, being
either in terms of the problem salience or in problem ‘definition’. In their overview of the
literature concerning the mass media’s agenda setting influence on the political realm,
Walgrave and Van Aelst (2006) conclude that different studies have led to contradicting
results. This is due to the divergent definitions and operationalizations of the political
agenda that have been employed on the one hand and to a focus on either election
campaigns or routine times politics on the other hand. Walgrave and Van Aelst argue that
findings indicating the media agenda’s influence on the political agenda depend on whether
the researcher looks at institutional/substantial measures (e.g. laws, budgetary spending) or
more symbolic policies (speeches, interviews, hearings and debates). Though the use of the
term ‘symbolic’ might be somewhat confusing, since it reduces political debates to little
more than speech-acts, they convincingly demonstrate the differential impact of the media.
With regard to substantial measures little or no influence is found, whereas a strong
influence is often established in case of symbolic policies (e.g. Edwards and Wood, 1999).
Here, we focus on official parliamentary documents, which according to Walgrave and Van
Aelst’s distinction are more institutional (the parliament as legislative power) and on
written parliamentary questions, that are more symbolic in nature (controlling the
executive, or ‘grilling the government’ to use Walgrave and Van Aelst’s terminology).
Combining both elements, a moderate influence from media on parliament can be
expected.i
The second element that determines the influence of media on politics is the time-
period under study: either election time or routine politics. Different studies have
demonstrated that during election times there is limited or no influence from the media
agenda on the political agenda, while in routine times more mixed results are found (e.g.
Brandenburg, 2002; Kleinnijenhuis et al., 2003). According to Walgrave and Van Aelst
(2006), there might be several reasons for the limited influence of media on the political
agenda during election campaigns. First, studies of election campaigns deal with short
periods in time and are not able to grasp any long-term effects. Second, during election
periods politicians vigorously try to garner attention for their points of view and to
influence the media agenda, instead of the other way around. Third, in election times media
devote more attention to politicians, thereby providing them with the opportunities to get
their messages out. Fourth, particularly during election times certain rules, traditions and
practices regarding fairness and balance limit the autonomy of the media. In the ten-year
period we studied (1995-2004), three parliamentary national elections were held, but the
largest part of our research-period consists of ‘routinized’ politics, so there is no a-priori
expectation of limited or no media-influence.
Framing Integration and Immigration
56
The next question is what form this influence takes. In general, an increase in issue
attention in the media leads to an increase in issue attention on the parliamentary agenda.
This influence, however, is more likely when the amount of attention on the issue is
substantial (Eilders, 2000, 2002). Since we expect that the same mechanisms operate for
framing, we assume that increased use of a certain frame by the media leads to an increased
circulation of this frame in parliament, but this influence only occurs when the frame is
present in a substantial amount of the total media coverage of the issue.
We not only expect media framing to influence parliamentary framing, but the reverse
relationship as well. Several studies take both possible influence patterns into account and,
again, show mixed results. Soroka (2002), researching the relation between parliament,
media and public opinion for three issues in Canada in the period 1986-1995, finds different
relations for different issues. He argues that, in line with Zucker’s obtrusiveness hypothesis,
media effects on politics and public are limited for those issues that can be experienced
directly by public and politicians (e.g. inflation). Wood and Peake (1998) show that for
several foreign issues in the USA, the media determine presidential actions, while the
reverse relationship is not found. Brandenburg’s study (2002) relating the daily issue
agenda’s of the two main political parties in the 1997 UK parliamentary elections to various
media outlets contradicts these findings. Taking nine issues into account, he finds that the
content of all researched media outlets are to a certain extent guided by the political parties,
while no influence of media content on parties’ agendas is found. Finally, two studies of the
Dutch context are worth mentioning. Studying the salience of social-economic issues in
party programs, media coverage and among the public in the period 1980-1986
Kleinnijenhuis and Rietberg (1995) find a positive influence of the political agenda on the
media agenda and a boomerang effect (i.e. a negative influence) from the media on the
political agenda. Kleinnijenhuis’ study (2003) that takes into account a wide range of issues
during the parliamentary election campaigns of 1994, 1998 and 2002, indicates that opinion
making in the media is to a large extent influenced by the earlier presented opinions of
politicians. Even though this research does not compare the political agenda to that of the
media, but only looks at media coverage, it confirms Kleinnijenhuis and Rietberg’s (1995)
conclusion that ‘media follow politics’.
To summarize, a positive influence of the political agenda on the media agenda is
often found, but this finding is not robust. We assume a similar mechanism like the
influence of media framing on parliamentary framing is operating. We therefore
hypothesize that increased use of a frame within parliament leads to increased use of this
frame in media, but this influence only occurs when this frame is present in a substantial
amount of the parliamentary documents.
Press and Parliament
57
4.4 Events
Events that occur both outside the media arena and the parliament play a crucial role in the
agenda building and framing of issues. Shifts in framing might occur because of events
within institutionalized politics (e.g. elections that change the power relations within
parliament) or sudden (inter-)national events (e.g. attacks on the World Trade Center in
New York, murder on politician Pim Fortuyn) that dramatically alter the current political
climate.
The media arena is particularly vulnerable to the occurrence of events (Kepplinger and
Habermeier, 1995). News values determine why some but not other events make in into the
news. Events have to pass the news threshold, which consists of news values attributed to
the real world. Galtung and Ruge (1965) note that what happens in the real world gets
attributed some news factors (or not) that determine the chance that events make it into the
news. Again, making it into the news is something different from significantly shifting the
framing of an issue for a period of time. To accomplish this an event must (a) dominate the
news for a longer period in time or change the political power base in society permanently
and (b) be incongruent with the way the issue has been framed before the event or result in
political actors propagating other frames that gain dominance in society.
Though it is argued that the political is more routinized and less influenced by external
events than the media arena is (e.g. Wood and Peake, 1998), external ‘shocks’ are likely to
have an impact as well. The most obvious influence is from parliamentary elections that
change the composition of parliament and consequently the configuration of actors that are
involved in the framing process. Assuming that different political parties with different
ideologies propagate different frames, electoral gains for a certain party might help to gain
standing and legitimacy for its framing of a certain issue, consequently leading to a more
frequent use of this framing in both the parliamentary and media arena.
Even when an event meets previously mentioned requirements, differences may persist
regarding how the influence takes place. First, differences exist in what kinds of framing-
shifts are taking place. In some cases, this is quite predictable. For example, the attacks on
the World Trade Center on September 11, 2001 are likely to increase the frequency of a
frame that presents Muslims and Islam as a threat to Western societies. In other cases, this
is less obvious: the announcement of populist Pim Fortuyn to run for the parliamentary
elections in 2002 could have led to a lot of media attention to his opinions on the issue of
integration of minorities, which were a rather diverse palette of partly contradictory
statements (Kleinnijenhuis et al., 2003). The influence of this event on the parliamentary
arena, if any, is even less predictable. Second, there are differences in the duration of the
framing-shift. Some events change the framing of the issue (semi-)permanently, as might
be the case for events that cause a sustaining change in power relations within society,
while others only change the framing during the course of the event or for a limited time
period afterwards. Third, the size of the framing-shift can differ. In some cases, shifts are
subtle, with one frame gaining some ground at the expense of other frames. In other cases,
Framing Integration and Immigration
58
like 9/11, events may cause much larger changes (paradigm-shift), such as complete
abandonment of certain frames, while others become dominant for the first time.
While we have some a priori expectations about how certain events might influence
the parliamentary and media arenas, little theory-building and empirical research is
available on this issue, so we treat it here mainly as an empirical, exploratory question. In
the methods section, where we describe our events, we will formulate some more specific
expectations about the kind of influences each event may have. We now turn to our
methods, where we will address how we developed our frames and which key-events we
take into account.
4.5 Methods
Frames
De Vreese (2005: 53) argues that there is little consensus about how to identify frames in
the news. According to him, roughly two strategies can be applied. The first is inductive in
nature: no pre-defined frames are used and frames emerge during the course of analysis
(e.g. Gamson, 1992). The second is deductive in nature and texts are analyzed using pre-
defined and operationalized frames. Furthermore, De Vreese also makes a distinction
between generic and issue-specific frames. The former are of a more general nature and can
be applied to different issues in different contexts (Valkenburg et al., 1999), the latter are
tailored to the specific issue and context under study (e.g. Shah et al., 2002).
In our study we use pre-defined, issue-specific frames established through a qualitative
pre-study. We chose issue-specific frames, because we analyze both media content and
parliamentary documents. Since these documents differ in significant respects from
newspaper articles, proposed generic news frames (e.g. human interest, economic
consequences, see Valkenburg et al., 1999) are unlikely to be present in parliamentary
documents, so issue-specific frames are most appropriate for comparison purposes and to
trace influences between the two arenas more precisely. In chapter 3, the procedure, frames
and resulting data set are described in more detail.
Events
We use dummy-variables to capture the effects of key-events. Several events are expected
to be influential in determining the intensity of news coverage about immigration and
integration. The selection of events deserves special attention. One does not know in
advance which events might affect the intensity of immigration and integration news and
parliamentary coverage, and looking at this coverage itself for possibly relevant events can
be considered as selecting on the dependent variable, which is theoretically and
methodologically inappropriate. Therefore, we chose a somewhat common-sense procedure
to establish key-events. We distinguish three types of events that could potentially
determine the intensity of reporting about immigration and integration issues.
Press and Parliament
59
The first type is parliamentary elections, here classified as institutional events, where
we distinguish two different categories of events. First, as a result of election outcomes, the
framing of the issue of immigration and integration might fluctuate because of changing
power distribution among parties and changes in the composition of the government. In our
research period, we distinguish the parliamentary elections of 1998 and 2002. The Liberal-
Conservative Volkspartij voor Vrijheid en Democratie (VVD), booked electoral gains in
1998. In 2002, a week after party leader Pim Fortuyn was assassinated by an
environmentalist, the new populist anti-immigration party Lijst Pim Fortuyn (LPF) won
about 17 percent of the votes and entered government together with the VVD and the
Christian-Democrats (CDA). We did not include the advanced parliamentary elections of
January 2003, since the political attitude of the new government towards the issue remained
similar, even though it shifted the support for the various parties. In general, one might
expect that a shift in political power to more right-wing parties leads to a framing-shift
towards more immigrant-unfriendly frames (e.g. Islam-as-threat and restriction) at the
expense of the other frames. Second, we included the election campaigns of 1998 and 2003
as separate events, since these are likely to be times in which debate on political contested
issues intensifies and in which politicians are likely to promote frames that they expect to
be electorally successful.ii
The second type of key-events encompasses international incidents, here referred to as
unpremeditated events. We take into account three key-events: 9/11 needs to be considered,
since it dramatically heightened the public debate about the integration of Muslim
minorities in the Netherlands. Furthermore, the war in Iraq, starting in March 2003, and the
bombing of several trains in Madrid in March 2004 are likely to have changed media
coverage towards immigrants in general and Muslims in particular. In general, we expect
these events to have resulted in an increased framing of Islam-as-a-threat.
A third type of key-event relates to political and societal actors trying to influence
public debate with their opinions. Here, we focus on unconventional opinions about the
issue of immigration and integration of minorities and classify these attempts as deliberate
events. Koopmans (2002) and Prins (2005) suggest that two national actors and their
statements have been of central importance in the debate on immigration and integration.
First, the publication of Paul Scheffer’s newspaper article on the multicultural fiasco,
written in January 2000, which might have caused increased multicultural framing and
second, the political campaign of Pim Fortuyn, who was elected leader of new-party
Leefbaar Nederland in November 2001 and stirred up the public debate on the issue of
immigration and integration of minorities .
A final event that we included is the murder on movie director and social critic Theo
van Gogh in November 2004. He was assassinated by a Muslim-extremist because of his
controversial attitudes towards the Islam. This event resulted in increasing tensions between
Muslim minorities and other parts of the population and probably increased attention to the
Islam-as-threat frame as well.
Framing Integration and Immigration
60
4.6 Analysis
To analyze the influences of the media and parliamentary agenda on each other and
the effects of the key-events on both, we conduct a Vector Autoregressive (VAR) analysis
(see chapter 2). Our dependent variables are the monthly proportional uses of each of the
frames in each of the arenas, meaning that the sum of scores of the five frames every month
equals 1.00. We try to explain the proportional use of each frame in one arena with the
previous use of this frame in both arenas. Furthermore, we also conduct an analysis for the
total number of used frames, which indicates the influence of each arena on the total
substantial attention (agenda) for the issue. With the augmented Dickey Fuller test, we test
whether the series are stationary, which is a requirement for using VAR-analysis (see
Chapter 2).
Table 4.1 presents the results of this test. We used the AIC to establish the appropriate
lag length for the test.
Table 4.1 Augmented Dickey-Fuller test for presence of unit root
Parliament Media
Total attention
Multicultural
Emancipation
Restriction
Victimization
Islam-as-threat
-6.46 (1 lag)
-6.20 (0 lags)
-6.70 (1 lag)
-4.39 (5 lags)
-5.95 (0 lags)
-3.69 (1 lag)
-4.87 (0 lags)
-3.72 (1 lag)
-8.38 (0 lags)
-5.74 (0 lags)
-4.51 (4 lags)
-4.27 (1 lag)
Note: AIC is used to establish appropriate lag-length; All results indicate absence of unit-root (p<.01)
Results indicate stationarity for all our variables, which makes a VAR-analysis appropriate.
The number of lags that needs to be included for each variable can be established using a
Likelihood Test for Dropping Lags (Enders, 2004, see also chapter 2). We assume that
influences between both agendas do not occur over a longer period in time than 6 months,
which is in line with outcomes of classical agenda setting research indicating influences up
to a couple of months (Dearing and Rogers, 1996). We start with a model including six lags
and drop lag by lag, until the Likelihood Test indicates loss of significant explanatory
power of the model.
Press and Parliament
61
Following Freeman et al. (1989), we report the following:
(1) F-tests for the whole blocks of lags of each variable, which indicates whether –
overall- any significant causal relationship between the independent variables and
dependent variable exists (Granger-causality), supplemented with the coefficients of the
dummy-variables and the explained variance.
(2) A presentation of the decomposition of the Forecast Error Variance, indicating for
each variable over-time what portion of the movement in a series can be attributed to its
own shocks versus shocks from the other variables.
(3) An impulse response analysis of the moving average representation of the variables
in the system (see also chapter 2).
iii
In our analysis, key-events are added as independent variables to the various models.
Key-events are captured by including dummy-variables for each of the events. These
effects can be both temporarily (having values as …,0,1,0) or permanent (…,0,1,1).iv
Furthermore, effects occur directly, but it can also take some time before the event leads to
changes in framing. We refrain from formulating explicit hypotheses about how various
events affect the dependent variables, since we have no firm theoretical foundation on
which to base these hypotheses. For some of the introduced effects, however, we can think
of certain expectations. We anticipate, for example, that parliamentary elections have
permanent effects, since they change the political power base for a longer period in time,
but it might take several months, before this effect occurs due to for example coalition
formations. We expect international events to have direct effects because they occur during
a limited period and lead to (unexpected) changes in the political environment. In most of
the cases, this effect is likely to wear of quickly, though in the instance of 9/11 a more
permanent change in attitudes towards the Islamic world and also Muslims in Western
societies may have occurred. For the deliberative events the scholarly literature suggests
that these events have led to an enduring shift in the political debate on the issue of
immigration and integration of minorities (Kleinnijenhuis et al., 2003; Koopmans, 2002).
For the murder on Van Gogh we expect an abrupt effect. Whether this effect is permanent
or temporarily will probably be hard to establish using our data, since our research period
ends one month after the murder. For each of the events, the time lag and the duration of
the effect on framing will be assessed empirically by looking at the common model-test
statistics (R-squared, Akaike Info Criterion) for attention and framings, choosing for each
dummy the lag length and duration that overall captures the consequences of the external
event best. We allowed a maximum lag of three months for parliamentary elections and one
for other events, to ensure that changes in framing are indeed likely to be caused by the
specific event.
Framing Integration and Immigration
62
4.7 Results
Relationships in issue attention
We now turn to the bi-directional relationship between the two arenas. As
demonstrated in chapter 3 there is some overlap in the overall attention to the issue and the
proportional use of each of the frames in each arena, but differences exist as well. However,
causal relations between the two arenas may become visible if we account for timing and
external key-events. Table 4.2 presents the results of the analysis for total attention to the
issue of immigration and integration of minorities. The Likelihood Test suggests that a
model with four lags is the most appropriate. As expected there is indeed a strong yearly
trend in the parliamentary arena, with the 12-month coefficient being highly significant.
The recent past has a significant impact as well (F-test for Granger-causality highly
significant). Attention in the media arena also has an influence on parliamentary attention,
it Granger-causes parliamentary attention for the issue. This influence, however, is not
straightforward: the impulse response analysis demonstrates that effects are initially
negative and become temporarily positive after five months (Figure 4.1a). Overall, this
results in a rather limited impact: after 18 months media attention accounts for 5.7 percent
of changes in parliamentary attention (Table 4.3). Interestingly enough, the media coverage
one year earlier has a positive influence on the current parliamentary attention to the issue.
Apparently, there are long-term influential relationships where media coverage affects
actions in parliament in the long run. Furthermore, several events have a significant impact
on the overall parliamentary attention to the issue of immigration and integration of
minorities. For instance, 9/11 has a positive and permanent influence, while Fortuyn’s
entrance into the political arena has a negative influence. The latter is likely to be a
statistical artifact, because it occurs only two months after 9/11 and probably indicates
declining attention to 9/11 rather than the actual influence of Fortuyn. The third event that
has a significant and permanent influence on the parliamentary agenda is the war in Iraq.
The discussion about the Dutch government’s official position in relation to the Iraq war
and the subsequent debate afterwards has included elements of the immigration and
integration issue, perhaps the Islam-as-threat frame in particular. Remarkably enough, both
the elections of 1998 and 2002 have no impact. A change in government and shifts in
parliamentary standing of various parties does not change the attention to the issue in
parliament. However, as we will see further on, changes in parliamentary standing do
change the framing of the issue.
Press and Parliament
63
Table 4.2 VAR Analysis (4 lags) for the attention for immigration and integration in
media and parliament (period 1995-2004)
Dependent variables
Parliament Media