Framing migration and the process of crimmigration: A systematic analysis of the media representation of unauthorized immigrants in the Netherlands

Agenda-setting, framing and moral panics: Why media representations matter

Crimmigration legislation does not arise in a vacuum, but is the result of inter-relational discursive processes in which migrants are constructed as social threats. The social construction of the migratory threat essentially rests on reinforcing interactions between political, public and media discourses, with direct causal relationships within this triangle being generally quite difficult to establish (Duffy and Frere-Smith, 2014). The mass media are assigned a central position in these processes, through the selection of topics and issues and through processes of labelling and attributing qualities to groups and individuals, and inferring causes and meaning (Helbling, 2013; Maneri and ter Wal, 2005). The organization and selection of topics relates to the agenda-setting theory, which suggests that, by paying considerable attention to certain issues, the media have the ability to influence what people think about and as such can set the public’s agenda (Dunaway et al., 2010; McCombs and Shaw, 1972). Expanding on this notion, and related to processes of labelling and attribution, is the concept of framing (Goffman, 1974): it is not only relevant what issues the media write about, it is equally important how they write about these topics (Boomgaarden and Vliegenthart, 2009). According to Entman (1993: 52), this means to ‘select some aspects of a perceived reality and make them more salient in a communicating context’, which can be identified through ‘the presence or absence of certain keywords, stock phrases, stereotyped images, sources of information and sentences that provide thematically reinforcing clusters of facts or judgments’.

Caviedes (2015: 900) has argued that ‘the more often the press mentions a particular issue and links it to a social ill, the more likely that issue is to be considered a “crisis” meriting political action and resolution’. This resonates closely with Cohen’s (1972: 9) concept of moral panics, which entails that a ‘condition, episode, person or group of persons emerges to become defined as a threat to societal values and interests’. A defining characteristic of moral panics is that the concern itself and the actions taken are highly disproportionate; they exaggerate concerns when the actual threat itself does not justify criminalization or the curtailing of rights (Goode and Ben-Yehuda, 2009; Hall et al., 1978). The concept has previously been used to explain the criminalization of illegal stay in Italy (Maneri, 2011), the criminalization of immigration in post-9/11 United States (Hauptman, 2013), the criminalization of asylum seekers in the United Kingdom and the United States (Welch and Schuster, 2005; Welch, 2004) and crimmigration processes in Australia (Welch, 2012).

Terminology

Describing unauthorized immigrants as criminals can happen in direct ways, but more subtle forms are apparent too. In this regard, considerable attention has been drawn to the use of the term ‘illegal’, which is criticized not only because it ‘stresses criminality’ and ‘defines immigrants as criminals’ (Lakoff and Ferguson, 2006: 1), but also for being inaccurate: although a migratory act might be illegal, people themselves cannot be illegal. Moreover, it does not do justice to the complex question of legal and illegal stay, which includes numerous ‘in-between’ situations described as ‘semi-compliance’ (Anderson, 2013; Düvell, 2011; Guild, 2004) and the dynamic nature of the phenomenon (Van Meeteren, 2010). As such, it is what Maneri (2011: 80) defines as ‘“collective categories” that lack any descriptive coherence or precision, but are nevertheless replete with connotations and implicit associations’. Many academics, human rights organizations and EU institutions, but also various international news associations, ¹ therefore use terms that are perceived to be more neutral, most notably ‘irregular’ and ‘undocumented’, although some academics also explicitly use the term ‘illegal migrant’ in order to emphasize the role of the state in creating processes that render individuals illegal (Schuster, 2011; Van Eijl, 2008). For the same reason, Bauder (2014) recently argued in favour of the term ‘illegalized’. Others also defend use of the term ‘illegal’ for being clear and accurate, arguing that the alternatives are merely politically correct euphemisms. ² The debate surrounding terminology has taken on a distinct political character, perhaps most notably in the United States, where liberal and conservative advocates have both sought to get their terminological frames to dominate in the media (Merolla et al., 2013). The preferred word choice then roughly reflects whether someone is in favour of a more restrictive approach or a more rights-based approach. For example, Anderson and Ruhs (2010: 175) argue that ‘“illegality” is also the term that is often used by those elements of the mass media that promote and reinforce negative public attitudes to immigration, and illegal immigration in particular’. ³

Methodology

In general there is no lack of scholarly attention for the representations of migrants and other minorities in Western media, although Thorbjornsrud (2015) notes that the number of studies that focus specifically on unauthorized immigrants is relatively small. Most of this research has been more or less a critique of the media, focusing on negative depictions of immigrants and minorities that demonstrate an underlying structural bias in the media – or even society at large (Bleich et al., 2015a). Many of these studies have adopted a qualitative approach and often focused on a relatively short time period or a single event, analysing only a small number of texts (Sciortino and Colombo, 2004). Because a large number of these researchers also take an explicit stand, such studies have sometimes been criticized for lacking representativeness and being subject to ‘cherry-picking’, where researchers select only those articles that confirm their existing beliefs and initial hypotheses (Hier, 2009; Koller and Mautner, 2004; Orpin, 2005; Stubbs, 1994). Several more recent studies have therefore employed more systematic and comparative methods, resulting in various articles demonstrating the complexity of immigration coverage and challenging the idea that the media are intrinsically biased and consistently engage in negative framing of immigrants and minorities (Bleich et al., 2015b; Caviedes, 2015; Hallin, 2015; Lawlor, 2015; Thorbjornsrud, 2015; Tolley, 2015). Yet, despite these important methodological advances, none of these studies employs what we believe to be one of the most promising approaches towards media analyses: a corpus linguistics approach.

Corpus linguistics (CL) refers to the study of (often very large bodies of) real-life textual data (the corpus) with the aid of computer software (Baker, 2006; Mautner, 2009; McEnery and Wilson, 1996). Corpora are large, representative (or even comprehensive) bodies of naturally occurring language, and, because they are stored electronically, it becomes possible to carry out statistical analyses that can reveal – possibly counter-intuitive – linguistic patterns and frequency information (Baker, 2006). Until now, corpus techniques have only rarely been used for discourse analyses, although the advantages of using CL in media studies have repeatedly been demonstrated (Allen and Blinder, 2013; Baker, 2012; Baker et al., 2008; Gabrielatos and Baker, 2008; Koller and Mautner, 2004; Mautner, 2009). One of these advantages is the fact that a researcher can work with very large amounts of data, which is particularly interesting when studying media content: as Fairclough (1989: 54) has rightly argued, ‘a single text on its own is quite insignificant: the effects of media power are cumulative, working through the repetition of particular ways of handling causality and agency’. Computer-assisted analyses of big data sets also greatly reduce a researcher’s bias and increase the internal validity of a study, by employing a comprehensive rather than a selective approach (Baker, 2006; Mautner, 2009). Of course, as Tolley (2015: 968) also acknowledges, the distinction between manual and automated analyses is far from clear-cut, with computers amplifying rather than replacing human work (see also Grimmer and Stewart, 2013). But, although much still depends on the individual researcher’s choices and his/her interpretation of the findings, ‘it becomes less easy to be selective about a single newspaper article when we are looking at hundreds of articles’ (Baker, 2006: 12).

For the current study we used the software program ‘Wordsmith Tools’ to carry out four types of analysis. First, frequency lists are one of the most basic applications but a good starting point for the analysis of a corpus (Baker, 2006). They reveal the most frequently occurring words in a corpus, or show how often specific words appear. Through the analysis of collocates a more discursive analysis of the way immigrants are described becomes possible on a quantitative level. Collocates are words that appear near another word more often than could be expected by chance only (Blinder and Allen, 2015), and thus they provide ‘a way of understanding meanings and associations between words which are otherwise difficult to ascertain from a small-scale analysis of a single text’ (Baker, 2006: 96). Wordsmith Tools can identify statistically significant collocates, based on their co-occurrence, the relative frequency of both words in the corpus and the full size of the corpus. Following previous studies, we used a window of 10 words for our analysis, 5 to the left and 5 to the right, using both log likelihood (LL, required minimum score of 6.63) and mutual information (MI, required minimum score of 5.0) to test for statistical significance and strength of the relationship between two words (Allen and Blinder, 2013; Baker et al., 2013; Blinder and Allen, 2015). We focused specifically on the L1 collocate, meaning the word that appears directly before the keyword, because this will often be an adjective directly describing the word of interest (Blinder and Allen, 2015). Following Gabrielatos and Baker (2008), we also use the notion of consistent collocates (c-collocates) for words that are collocates in at least two-thirds of the annual sub-corpora. The results of a collocation analysis go further than a mere content analysis, providing ‘the most salient and obvious lexical patterns surrounding a subject, from which a number of discourses can be obtained’ (Baker, 2006: 114). Keyness allows the researcher to compare two data sets and see which words occur significantly more often in one of them, making it a particularly valuable tool for comparisons. Finally, Concordances combine quantitative and qualitative methods of content analysis. Baker (2006: 71) describes concordances as ‘simply a list of all of the occurrences of a particular search term in a corpus, presented within the context that they occur in’. This context usually means a number of words, to be decided by the researcher, to the left and the right of a search term – for example, ‘illegal migrants’. Concordance analyses thus allow for a closer and more in-depth examination and manual reading of a selection of relevant articles, and can act as a bridge towards an informed critical discourse analysis of a smaller number of relevant articles. Here we used it mainly to provide a ‘vital validity check’ of our quantitative results (Blinder and Allen, 2015: 12).

The corpus

For this study we have created our own specialized corpus, consisting of all newspaper items on unauthorized migrants that appeared in Dutch national newspapers between 1 January 1999 and 31 December 2013. ⁴ This period was chosen because it also covers a period from before the watershed events at the beginning of the century that have been identified as causing a negative change in Dutch public discourse on immigration (Pakes, 2006); the years 1999–2000 can be used as a period where we would expect more ‘neutral’ media content. ⁵ We chose newspapers because their archives are readily available online and, although numbers have been declining, newspaper readership is still relatively high in the Netherlands, with dailies reaching about 70 percent of the population. Of the five main newspapers, De Volkskrant, Trouw and NRC Handelsblad are considered as broadsheets, or ‘quality newspapers’, with the first two being more left-wing and NRC Handelsblad liberal-conservative. De Telegraaf has the highest circulation in the Netherlands and, together with Algemeen Dagblad, is considered to be a right-wing, ‘popular’ newspaper. However, unlike in for example the United Kingdom, the distinction between tabloid and broadsheet newspapers holds minor relevance in the Netherlands: the ‘popular’ newspapers are generally far less populist and sensationalist than, for example, the Sun or the Daily Express (Bakker and Vasterman, 2007; Broersma and Graham, 2013). The Dutch media landscape further consists of various free dailies mainly distributed on public transport (Metro, Spits, De Pers and Dag, with the latter two existing only briefly) and two conservative Christian newspapers (Nederlands Dagblad and Reformatorisch Dagblad). To avoid too much repetition we excluded regional newspapers, which, following a number of mergers, often have the same sections on national and foreign news as national newspapers.

Before starting our analyses we manually read a small random sample of newspaper articles. Besides further familiarizing a researcher with its corpus by already pointing towards interesting language patterns that require attention further in the analysis, this procedure can also help either to find specific terms or synonyms that have been missed in the original search string or to reveal that a high number of irrelevant articles is found with the search string. The final search string we used was the result of various of these trial and error steps and, although one can never be entirely sure to have captured all relevant articles, we believe that in this way we have created a comprehensive ‘unauthorized migrants’ corpus without a high number of irrelevant articles. Our final corpus contains 28,274 articles from 12 national newspapers and consists of just over 10 million words. ⁶ All these articles were sorted by publication month and newspaper, allowing us to carry out comparisons between newspapers and over time and identify trends.

Results

Figure 1 shows the number of articles on unauthorized migrants per annum. Because not all newspapers appeared throughout the complete research period – some went out of business, and others started – and results would therefore be skewed towards years with more newspapers, this figure is based on the five main newspapers only, ⁷ with additional checks confirming that other newspapers showed similar trend lines. Whereas newspaper attention for unauthorized migrants was quite stable until 2006 – fluctuating only slightly each year – there was a sharp decrease in the number of articles per year during the period 2006–9, after which attention steadily increased again. However, the annual number of newspaper articles did not ever reach the same numbers as in the 1999–2006 period. If we look at the distribution of newspaper articles per month in 2010, November and December have by far the most articles. The proposal to criminalize unauthorized stay was introduced in October 2010, practically the lowest point of media attention. Frequency lists and concordance analyses confirmed that the rise in attention for unauthorized migrants during the last three years of our study is primarily due to the political and public controversy over the proposal.

OPEN IN VIEWER

Figure 1.

Number of articles on unauthorized migrants per year, 1999–2013.

OPEN IN VIEWER

Figure 2.

Number of articles on unauthorized migrants per month, 1999–2013.

One possible reason for the decline in newspaper articles on unauthorized migrants is the enlargement of the European Union with 10 (mainly East European) countries in 2004, and Bulgaria and Romania in 2007. Overnight this legalized unauthorized immigrants from those countries and it is one of the most important explanations for the decreasing number of estimated unauthorized migrants in the Netherlands (Van der Heijden et al., 2011). At the same time there appears to have been a shift in the public debate, where populist resistance against Islam and immigration has been replaced, or at least supplemented, by a fierce resistance against European integration (Wansink, 2007). It therefore seemed plausible that part of the media attention for unauthorized migrants has shifted to migrants from East European countries – and that they are no longer referred to as unauthorized migrants. But, whereas a LexisNexis analysis indeed showed a much higher number of annual articles about East European migrants since 2004, a content analysis of various nationalities and terms such as ‘East European’ did not show a declining trend in our corpus – either absolute or relative. This means that, although there has been a general increase in newspaper attention on East European migrants, this cannot explain the decrease in articles on unauthorized migrants. Instead, there does not seem to be one, clear explanation for the decline in newspaper attention for unauthorized migrants.

Looking at the distribution of newspaper articles per month, it becomes immediately clear that there are a few spikes in newspaper attention for unauthorized migrants. Closer examination of these months reveals that these can be explained by specific events that attracted strong media attention for a brief period of time. In June 2000, most articles are about 58 Chinese migrants who died of suffocation in the back of a truck while on their way to the United Kingdom. The high points in February 2004 and September and December 2006 are primarily due to political controversies surrounding the former Dutch Minister for Aliens Affairs and Integration Rita Verdonk, who was known for her tough stance towards unauthorized migrants. The most recent peak in May 2013 is due to serious debates within the governing Labour Party about the proposed criminalization of illegal stay. It therefore seems that political debates and developments are the main cause of intense media attention on unauthorized migrants. Gabrielatos and Baker (2008: 18) noted a similar effect in the United Kingdom, where immigrants ‘were thus functionalised as part of a struggle for political hegemony, being discursively constructed as a people who merely constitute the topic of political debate, somewhat dehumanised as an “issue”’. It is also noteworthy that other high-profile events that involve or could potentially be linked to unauthorized migrants – a fire at the airport immigrant detention centre that killed 11 unauthorized migrants, various (international) terrorist attacks – have not caused the same increase in newspaper articles on unauthorized migrants.

Discourse: Terminology

In Dutch newspapers, ‘illegal’ is by far the most often used term to refer to unauthorized migrants – over 95 percent of instances. ‘Irregular’ and ‘undocumented’ are very rarely used, and sometimes even in different contexts, such as ‘irregular rebel groups’. Moreover, this finding applies, without exception, to newspapers across the political spectrum. Similar figures were recently found in the United States, where. in over 95 percent of cases newspapers used the term ‘illegal’ (Merolla et al., 2013). If we look at the distribution of cases where the term ‘illegal’ is used, it becomes evident that it is much more often used as a noun (‘illegals’) than as an adjective (‘illegal migrant’ or ‘illegal immigrant’). When ‘illegal’ is used as an adjective, the most common noun is ‘immigrant’, followed by ‘migrant’ and ‘alien’. Instances of ‘illegal asylum seeker’ and ‘illegal refugee’ – legally speaking impossible constructions – also appear several dozen times per year. The term ‘rejected asylum seekers’ (‘uitgeprocedeerde asielzoekers’), which does not convey the exact same meaning, is the second most often used term, but the saliency of this term varies widely between years. Use of ‘illegal’ as a noun does show a steady decline throughout the period, both absolutely and relatively, but this cannot be explained by a replacing term. Rather, it seems to be the result of an increasing focus on the condition of ‘illegality’ in later years, as a place in which people can ‘disappear’ or ‘end up in’.

A criminal discourse?

Considering the fact that ‘illegals’ was the most common way to describe unauthorized migrants, and we were interested in how unauthorized migrants as a group were described, we carried out a collocation analysis for this term. Table 1 shows an overview of the statistically significant collocates of ‘illegals’, sorted on the L1 position (the position immediately before ‘illegals’) for the period 1999–2013, without words such as ‘the’, ‘or’, ‘and’.

OPEN IN VIEWER

Table 1.

Collocates of ‘illegals’, 1999–2013.

N	Word	Total	Total left	Total right	L1
3	White	1936	1879	57	1842
7	000	715	648	67	518
11	Number/amount	646	603	43	477
12	Criminal (adj.)	541	489	52	450
14	Many/a lot	765	629	136	408
17	Million	394	375	19	336
18	All	451	392	59	264
20	More	852	543	309	229
21	Against	696	503	193	219
22	Group	388	353	35	195
23	Most	214	203	11	171
25	Thousands	179	165	14	139
27	Chinese	159	148	11	138
28	Stream	181	178	3	133
29	Other	377	242	135	124
30	Thousand	172	154	18	124
34	Hundred	160	147	13	105
36	None	738	334	404	104
37	Arrested	106	106	0	100
38	Eleven	198	189	9	89
40	How many	89	88	1	82
41	African	100	86	14	79
42	Tens	128	117	11	79
46	Groups	118	110	8	74
48	Tens of thousands	85	75	10	69
49	Less	170	105	65	68
50	Young	85	73	12	64

The most important L1 collocate for ‘illegals’ is ‘white’, referring to a group of unauthorized migrants who had long resided in the Netherlands but, owing to the enactment of the 1998 Linking Act, suddenly lost most of their rights. The word ‘white’ is used here with the connotation of a semi-documented position in society and does not refer to physical appearance. The focus on this group is entirely clustered in the initial years, particularly in 1999. Our results also show that numerical terms are often used to describe ‘illegals’, indicating that an important characteristic is how many there are. Except for absolute numbers (‘000’, ‘million’, etc.), these include more vague numerical descriptions (‘thousands’, ‘tens of thousands’, even ‘millions’) and group descriptors (‘a lot’, ‘groups of’). Vollmer (2011: 330) noted that ‘number games’ play a crucial role in policy discourses on unauthorized migration throughout the EU, where ‘higher numbers justify control and enforcement policies, whereas lower numbers ease the political landscape’. ‘Stream’ is also a strong L1 collocate, which confirms earlier findings that aquatic terms are frequently used in conjunction with migratory movements. Tsoukala (2005) argues that these terms have quickly become standardized in discourses on immigration and create a notion of uncontrollability and threat. What is, however, particularly notable for our main hypothesis is that ‘criminal’ is one of the most important adjectives used with the term ‘illegals’, showing that one of the most common ways to describe ‘illegals’ is as criminals.

Trends and developments

Although ‘criminal’ was one of the most important L1 collocates for ‘illegals’, we were also interested in identifying patterns over time. Therefore we tested whether it was a c-collocate by carrying out collocation analyses for the word ‘illegals’ in each separate year. This showed that ‘criminal’ as an adjective ⁸ scored significant LL scores in all 15 years of our corpus and significant MI scores in 13 years (in 2000 and 2012 the scores fell just short of the 5.0 requirement). Turning it around, the term ‘criminals’ is a significant c-collocate of the adjective ‘illegal’, falling short of the required statistical scores only in 2006 and 2011. The term ‘criminals’ was also a significant collocate of ‘illegals’ in all years after 2002, except for 2004 and 2009, thus being a c-collocate for the period 2002–13. This shows that ‘illegals’ often appear in the same sentence as ‘criminal’ and ‘criminals’.

Both ‘criminal illegals’ and ‘illegal criminals’ are thus constructions that can be regularly found, demonstrating that in newspaper articles on unauthorized migrants crime is one of the main topics discussed, and this has been the case throughout the 15 years of our corpus. Similar analyses with the terms ‘asylum seekers’, ‘refugees’, ‘migrants’ and ‘immigrants’ showed fewer or no references to crime. Although our corpus is not a representative reflection of all newspaper content on these groups – it contains only newspaper articles that also mention unauthorized migrants – this suggests that, in Dutch newspapers, unauthorized migrants in particular are strongly associated with crime.

Yet, although the adjective ‘criminal’ is a c-collocate of ‘illegals’, a content analysis of the phrase ‘criminal illegals’ showed a steady decrease in use of the term between 1999 and 2013.

Figure 3 shows that, over time, not only did usage of the word ‘illegal’ show a steady decrease, but Dutch newspapers also described unauthorized migrants less often as criminals. In the last two years of the period studied, only four instances of ‘criminal illegals’ per year occurred, of which two did not even refer to the Netherlands. Although some other variations appear in the corpus – especially ‘criminal aliens’, albeit still only one-third of the times ‘criminal illegals’ – there is no contrary trend visible in the use of these terms. Although crime remains an important topic in newspaper articles throughout our corpus, it is far less of an issue in the period around the proposed criminalization of illegal stay than it was around 2000, and unauthorized migrants are gradually less often directly described as criminal.

OPEN IN VIEWER

Figure 3.

Content analyses of the phrase ‘criminal illegals’.

Differences between newspapers

Whereas there are thus indications of a criminalizing discourse in Dutch newspapers regarding unauthorized migrants, we found considerable differences between newspapers, as illustrated by a keyness comparison between the left-wing ‘quality’ newspaper De Volkskrant and the right-wing ‘popular’ De Telegraaf.

Figure 4 shows how often various crime-related terms appeared per 10,000 words in both newspapers. The words ‘criminal’ and ‘criminals’ appeared three to four times more often in De Telegraaf than in De Volkskrant. Related terms – ‘police’ and ‘arrested’ – were also found significantly more often in De Telegraaf, thus lending empirical support to the assertion that right-wing newspapers more often link unauthorized migrants to issues of crime.

OPEN IN VIEWER

Figure 4.

Keyness of crime-related terms in De Telegraaf and De Volkskrant.

Conclusion and discussion

In this study we have analysed 15 years of Dutch newspaper content on unauthorized migrants. Employing a corpus linguistics methodology, we have been able to identify various patterns in both the frequency of newspaper articles and the discursive elements of how unauthorized migrants are described, thus testing the hypothesis that media content fuels the crimmigration process. Our analysis revealed some familiar ways in which unauthorized migrants are consistently depicted by all Dutch newspapers. The first and foremost is the constant way in which these newspapers refer to unauthorized migrants: ‘illegals’, or at best ‘illegal (im)migrants’. The term ‘illegal’ is also consistently used in the Dutch political discourse, and it seems to have relevance for arguments regarding actual legislation: in the debate about the formal criminalization of unauthorized stay at least one politician publicly argued that ‘it is of course very strange that something is illegal, but has no punishment to it’. ⁹ At the EU level there have been attempts to stop use of the term ‘illegal’ and instead employ the words ‘irregular’ and ‘undocumented’. ¹⁰ The Platform for International Cooperation on Undocumented Migrants (PICUM) recently launched a campaign to ban use of the term ‘illegal migrant’, including a leaflet with preferable equivalents in all major European languages. For the Dutch language the NGO suggests the terms ‘mensen zonder papieren’ and ‘mensen zonder wettig verblijf’ (literally: ‘people without papers’ and ‘people without lawful stay’). These descriptions seem rather long for everyday usage and therefore unlikely to become part of popular discourse; although the second option is to some extent an accepted term in Flanders, ¹¹ it does not appear a single time in 15 years of Dutch newspaper articles. Whereas in the United States recent changes in the stylebooks of major media outlets suggest a greater prevalence of alternative terms in the near future (Merolla et al., 2013), in the Netherlands there does not appear to be any meaningful discussion going on about terminology. At the same time, one can wonder if a terminology replacement might not merely lead to what has been dubbed ‘the euphemism treadmill’ (Pinker, 2007), whereby the replacement terms themselves become offensive over time. Moreover, two recent studies in the United States showed that terminology frames – ‘illegal’ versus ‘unauthorized’/‘irregular’/‘undocumented’ – did not influence public opinion regarding the provision of certain rights for unauthorized migrants, leading the authors to suggest that it might be more useful for migrant rights activists to focus on actual policies and how they are framed, rather than on terminology frames (Knoll et al., 2012; Merolla et al., 2013). In this regard it is interesting to note that the rise in newspaper articles about unauthorized migrants in the last years of our corpus can be solely attributed to the political and public discussion this bill triggered, but newspaper content during this time was not necessarily characterized by the framing of unauthorized migrants as criminal. Some concordance analyses even indicate that various newspapers offered a platform for substantial critique of the proposal, in this way possibly even playing a role in its eventual withdrawal in 2014.

One of the findings that stood out in our analysis was that ‘criminal’ was one of the most important collocates of ‘illegals’, providing evidence for a criminal discourse and a merger of references to migration and crime. Dutch newspaper articles about unauthorized migrants deal significantly often with issues of crime; indeed, it is one of the most salient themes. Needless to say, this is the quintessential discourse on which the criminalization of unauthorized migrants rests. Framing unauthorized migrants as criminals, in combination with a strong focus on numbers, serves as a powerful catalyst for what Vollmer (2011: 331) calls ‘the demonstration of efficient governance’, resulting in repressive and punitive policies with an intense focus on control.

Although these findings are a reason for concern, our main hypothesis proved incorrect: media attention for unauthorized migrants strongly decreased after 2006 and newspaper articles about crime among unauthorized migrants featured mainly in the first years of our corpus. The proposal to criminalize illegal stay came at practically the lowest point of media attention for unauthorized migrants, and seems to have been the result of a change in government rather than of active agenda-setting by the media or another actor. This provides some modest support for Threadgold’s (2009: 1) statement that there is ‘a small but growing body of evidence that political and policy discourse concerning immigration actually fuel the media discourse, which in turn drives policy’. The present study has clearly demonstrated that the bill to criminalize illegal stay in the Netherlands was not preceded by strong and increasing media attention for criminal illegal migrants. Rather, the framing of migrants as criminals is a more diffuse process in which the media seem to follow rather than fuel politics and policy.

Whereas the initial years of our corpus support the idea of a moral panic episode about ‘criminal illegals’, this is not something we see in later years. However, we also see that the problem of the ‘criminal illegal’ keeps lingering on and surfacing every now and then. In this regard it is interesting to see that Cohen (2002) more recently noted that reactions to asylum seekers do not follow the ordinary temporary moral panic model, but are subject to continuous negative, excluding and hostile messages. In their study on UK media reporting around Bulgarian and Romanian EU accession, Mawby and Gisby (2009: 48) argue that this model is ‘now a more accurate way to model a cluster of issues, such as immigration in general’ and they note that ‘this open-ended form of moral panic has seemingly “stepping stoned” its way from asylum seekers to immigrants, to EU enlargement, to general anxieties about crime’. This is relevant, because although our corpus showed that newspaper articles about unauthorized migrants decreased, articles about migrants from new EU countries actually increased, and it is questionable whether the general public is always aware of such distinctions within the broad category of ‘foreigners’. If negative attention for unauthorized migrants has merely been replaced by negative attention for foreign EU citizens and other ethnic and racial minority groups, and in the public perception these various categories are assembled into one homogeneous problematic group of outsiders, the decrease in newspaper articles on (criminal) unauthorized migrants will have had hardly any effect on public opinion about this specific group. In light of increasing European integration and the negative backlash this seems to cause in primarily West European countries, it would be worthwhile to examine media content on new European citizens and what this means for the future of the European project.

While this study provided valuable insights into newspaper reporting on unauthorized migrants in the Netherlands, it simultaneously raised various questions that point to interesting areas for future research. Here we have focused only on print media, but analyses of web-based content could be equally interesting. Furthermore, linking the analysis of media content to a more systematic analysis of the political discourse could increase our understanding of who influences whom. Finally, and most importantly, although here we have focused on the Netherlands only, we believe the approach we have introduced is particularly suitable for large-scale comparisons and hope to have laid the foundation for future comparative work between various European countries. This could focus on unauthorized migrants, but it would also be particularly worthwhile to examine how newspapers in different European countries have reported on the recent refugee crisis. The corpus linguistics approach allows for quick replications of analyses with large sets of longitudinal data. Because the statistical tests offer easy to compare outcomes, it is eminently appropriate for a systematic comparison of discourses in various countries.