farrall et al 1997 measuring fear of crime

22
BRTT.J. CRIMINOL VOL 37 NO. 4 AUTUMN 1997 QUESTIONING THE MEASUREMENT OF THE 'FEAR OF CRIME' Findings from a Major Methodological Study STEPHEN FARRALL.JON BANNISTER, JASON DITTON and ELIZABETH GILCHRIST* Research upon the fear of crime has grown substantially in recent years. From its very inception, this field has relied almost exclusively upon quantitative surveys, which have suggested that the fear of crime is a prevalent social problem. However, doubts about the nature of the instruments used to investigate this phenomenon have cumulatively raised the possibility that the fear of crime has been significantly misrepresented. Dealing with the epistemological, conceptual, operational and technical critiques of quantitative surveys in general and of fear ofcrime surveys in particular, this article suggests that our understanding of the fear of crime is a product of the way it has been researched rather than the way itis.Asthe aim of the research project under which this data was collected was to develop and design new quantitative questions, the article ends with some possible solutions to the epistemological, conceptual, operational and technical problems discussed which may improve future quantitative research in this field. The validity of a measuring instrument may be defined as the extent to which differences in scores on it reflect the true differences among individuals ... radier than constant or random error. (Selltiz et aL 1976: 169, quoted in Brewer and Hunter 1989: 129) Recent years have witnessed an increased interest in the fear of crime from both academics and policy makers. A plethora of studies, including several sweeps of the British Crime Survey (see, inter alia, Hough and Mayhew 1983; Chambers and Tombs 1984; Hough and Mayhew 1985; Mayhew et al. 1989; Maxfield 1987; Skogan 1990; Payne 1992; Kinsey and Anderson 1992) have concluded that the fear of crime impinges upon the well-being of a large proportion of the population. For example, Chambers and Tombs (1984: 29) reviewing the 1982 British Crime Survey Scotland, found that 'more than half of the respondents (58 per cent) said that at some time in the past they had been concerned about the possibility of being a victim of crime'. More recently, Hough (1995: 25) found that, when asked how safe they felt when walking alone in their area after dark, some 36 per cent of those surveyed said that they felt 'a bit unsafe' or * Stephen Farrall, Research Officer, Centre for Criminological Research, Oxford University, Jon Bannister, Lecturer in Social Policy, University of Glasgow, Jason Ditton, Professor of Criminology, Faculty of Law, Sheffield University, and Director, Scottish Centre for Criminology, and Elizabeth Gilchrist, Lecturer, School of Psychology, University of Birmingham. The research reported in this paper was supported by an E5RC grant (no. L210 25 2007) under the council's Crime and Social Order research programme. Versions of this paper were presented at the Faculty of Law, Sheffield University-, and it the British Criminology Conference, Loughborough University, July 1995. The authors are grateful to the Editor, two anonymous reviewers and Tony Jefferson for their invaluable comments on an earlier draft of this paper. We also extend our appreciation to those who made comments on the paper while attending the above seminars and conferences. 658 by guest on February 29, 2012 http://bjc.oxfordjournals.org/ Downloaded from

Upload: diana-l-panfiloiu

Post on 14-Apr-2015

78 views

Category:

Documents


0 download

DESCRIPTION

report on fear of crime

TRANSCRIPT

Page 1: Farrall Et Al 1997 Measuring Fear of Crime

BRTT.J. CRIMINOL VOL 37 NO. 4 AUTUMN 1997

QUESTIONING THE MEASUREMENT OF THE 'FEAR OFCRIME'

Findings from a Major Methodological Study

STEPHEN FARRALL.JON BANNISTER, JASON DITTON and ELIZABETH GILCHRIST*

Research upon the fear of crime has grown substantially in recent years. From its very inception,this field has relied almost exclusively upon quantitative surveys, which have suggested that thefear of crime is a prevalent social problem. However, doubts about the nature of the instrumentsused to investigate this phenomenon have cumulatively raised the possibility that the fear of crimehas been significantly misrepresented. Dealing with the epistemological, conceptual, operationaland technical critiques of quantitative surveys in general and of fear of crime surveys in particular,this article suggests that our understanding of the fear of crime is a product of the way it has beenresearched rather than the way itis.Asthe aim of the research project under which this data wascollected was to develop and design new quantitative questions, the article ends with some possiblesolutions to the epistemological, conceptual, operational and technical problems discussed whichmay improve future quantitative research in this field.

The validity of a measuring instrument may be defined as the extent to which differences in scoreson it reflect the true differences among individuals . . . radier than constant or random error. (Selltizet aL 1976: 169, quoted in Brewer and Hunter 1989: 129)

Recent years have witnessed an increased interest in the fear of crime from bothacademics and policy makers. A plethora of studies, including several sweeps of theBritish Crime Survey (see, inter alia, Hough and Mayhew 1983; Chambers and Tombs1984; Hough and Mayhew 1985; Mayhew et al. 1989; Maxfield 1987; Skogan 1990;Payne 1992; Kinsey and Anderson 1992) have concluded that the fear of crime impingesupon the well-being of a large proportion of the population. For example, Chambersand Tombs (1984: 29) reviewing the 1982 British Crime Survey Scotland, found that'more than half of the respondents (58 per cent) said that at some time in the past theyhad been concerned about the possibility of being a victim of crime'. More recently,Hough (1995: 25) found that, when asked how safe they felt when walking alone in theirarea after dark, some 36 per cent of those surveyed said that they felt 'a bit unsafe' or

* Stephen Farrall, Research Officer, Centre for Criminological Research, Oxford University, Jon Bannister, Lecturer in SocialPolicy, University of Glasgow, Jason Ditton, Professor of Criminology, Faculty of Law, Sheffield University, and Director, ScottishCentre for Criminology, and Elizabeth Gilchrist, Lecturer, School of Psychology, University of Birmingham.

The research reported in this paper was supported by an E5RC grant (no. L210 25 2007) under the council's Crime and SocialOrder research programme. Versions of this paper were presented at the Faculty of Law, Sheffield University-, and it the BritishCriminology Conference, Loughborough University, July 1995. The authors are grateful to the Editor, two anonymous reviewersand Tony Jefferson for their invaluable comments on an earlier draft of this paper. We also extend our appreciation to those whomade comments on the paper while attending the above seminars and conferences.

658

by guest on February 29, 2012http://bjc.oxfordjournals.org/

Dow

nloaded from

Page 2: Farrall Et Al 1997 Measuring Fear of Crime

STEPHEN FARRALL, JON BANNISTER, JASON DITTON AND ELIZABETH GILCHR1ST

'very unsafe'. It appears that the fear of crime is a social phenomenon of strikingdimensions.

However, several commentators have raised doubts surrounding the validity of theinstruments used to generate these findings (see, inter alia, Bernard 1992; Bowling1993; Fattah 1993; Schneider 1981; Skogan 1981; and Zauberman 1985). A range ofmethodological problems have been identified which cumulatively raise the possibilitythat the incidence of the fear of crime has been significantly misrepresented. Buildingupon these methodological insights, this article employs both quantitative andqualitative data sets to show that the reported incidence of the fear of crime is partlydependent upon the nature of the measurement instrument rather than a truereflection of'social reality'.

This article is constructed in the following fashion. In the first section, we reviewmethodological problems associated with the quantitative investigation of the fear ofcrime. These observations are drawn from critiques of quantitative surveys in generaland fear of crime surveys as they currently exist. Building upon these observations, inthe second section we review the arguments and evidence that suggests thatmethodological triangulation is an appropriate response to the identified problems.We shall utilize this technique (methodological triangulation), being sensitive toepistemological issues, in order to identify the methodological problems and to assessthe extent to which the reported incidence of the fear of crime is dependent upon theinstruments currently used to measure it. In the third section, we oudine thehypothetical methodological problems that are encountered with a quantitative crimesurvey, and discuss at length those actually observed in our data set. We conclude thatthe fear of crime is over-counted, and end with suggestions for future crime and fearof crime surveys in the light of our data. We start, however, with an outline of theresearch methodology.

The selection of areas

Four sites for sampling were selected along two dichotomies (outlying/inner city andpoor/affluent). The two inner city areas chosen both fall within two miles of Glasgowcity centre (Scotland). The outlying areas were both approximately five miles out fromthe centre of the city. Both of the areas selected as 'poor' have been identified byStrathclyde Regional Council as being Priority One Areas (the most deprived) undertheir 'Social Strategy for the Nineties'. These types of areas the Regional Councildescribe as being '. . . large areas of severe multiple deprivation . . .' (SRC 1994: 2).Unemployment rates for these two areas stood at (respectively) 39 per cent and 41 percent; overcrowded households at 6 per cent and 8 per cent; non-elderly illnesses at 15per cent and 21 per cent; long-term illnesses at 19 per cent and 27 per cent andhouseholds lacking amenities at 0.2 per cent and 0.3 per cent (all statistics from SRC1994). Data for the two areas selected as 'affluent' are generally harder to find, and thedata reported here refer to only one of these areas. The unemployment rate for thisarea was 5 per cent, with overcrowding at 0.5 per cent, non-elderly illnesses at 3 percent and long-term illnesses at 8 per cent (Muir 1994). Data for the other 'affluent' areaare even harder to find as it is neither a priority area nor does it form a distinctgeographical area.

659

by guest on February 29, 2012http://bjc.oxfordjournals.org/

Dow

nloaded from

Page 3: Farrall Et Al 1997 Measuring Fear of Crime

QUESTIONING THE MEASUREMENT OF THE 'FEAR OF CRIME1

The selection of individuals

The purpose of the first sweep of interviews (carried out from October to November1994) was to identify individuals in each area in terms of their fear of crime andperceived risk of victimization. Hence these interviews were very short.

Starting points for sampling were chosen and researchers used the 'random walk'method, calling at every fifth house. At each house, the first person over the age of16 years who answered the door was interviewed, except where some key groups werehard to find (such as young people). In one area, the caf6 in a local community centrewas also used to sample from. The times of day at which we were actively recruiting ineach area were varied so as to facilitate access to key groups which may otherwise behard to find (such as those working or at school). Approximately 40 people wereinterviewed in each area (N= 167) and were asked if they would be prepared to takepart in a longer interview if selected. The response rate for agreeing to be recontactedfor the longer interview was 90 per cent. Names, addresses and telephone numberswere collected for the purposes of recontacting.

Each individual interviewed at the first stage was classified according to their fearand risk ratings. Four groups were produced: high fear/low risk; low fear/low risk; highfear/high risk and low fear/high risk. In each area a total of 16 people was re-interviewed(four from each fear/risk group). Within each fear/risk group there was an equal gendersplit and, where possible, a range of ages. Thus the qualitative data set contains 64interviews.

Some Methodological Concerns

In this section we outline briefly some methodological concerns that have been levelledat quantitative surveys designed to explore the fear of crime. The criticisms raisedaddress issues of epistemology (e.g., the difficulties encountered when attempting toturn social processes into quantifiable events), conceptualization (e.g., what is meant bythe term the 'fear of crime'), operationalization (e.g., the design and wording of surveyquestions) and technique (e.g., factors governing the quality of data generated). Whilespecifically associated with crime surveys, some of these remarks resonate with moregeneral debates within the methodological literature.

Epistemological and conceptual concerns

At the epistemological level, Bryman (1984: 78) notes that surveys have been criticizedfor their tendency to 'view events from the outside', and for imposing empiricalconcerns 'upon social reality with little reference to the meaning of the observations tothe subject of investigation'. Similarly, Bowling (1993: 241-43) accuses crime surveymethodology of attempting to convert a social process into a series of quantifiable'moments' which do not adequately reflect the experiences or feelings of thoseinterviewed. Surveys are thus described as being 'static' and of reducing experience toa 'decontextualized snapshot' (1993: 232). This means that ongoing experiences orvariation in the range and strengths of emotions experienced are rarely adequatelycaptured by the survey tool.

660

by guest on February 29, 2012http://bjc.oxfordjournals.org/

Dow

nloaded from

Page 4: Farrall Et Al 1997 Measuring Fear of Crime

STEPHEN FARRALL, JON BANNISTER, JASON DITTON AND ELIZABETH CILCHRIST

Conceptual concerns about the fear of crime cannot be wholly divorced from theoperational concerns (discussed below). These concerns include defining what fear is,its nature and its occurrence in individuals' everyday lives. Many of these conceptualissues have not been successfully resolved as yet. As Ferraro and LaGrange note (1987:71) 'the phrase "fear of crime" has acquired so many divergent meanings that its currentutility is negligible'.

Operational concerns

A further set of methodological problems concerns the operationalization of concepts.Ferraro and LaGrange (1987) make a distinction between 'formless' and 'concrete'fears. 'Concrete' fears relate to particular offences (e.g. burglaries). 'Formless' fears donot refer to any particular crime, but instead ask, for example, more generally 'Howsafe do you feel walking around alone in this area after dark?'.

Even where specific mention is made of a particular crime (e.g. vandalism), theoperationalization is still incomplete because temporal, spatial and social contexts arenot addressed. 'Where' the respondent is, 'when' the respondent is there, and 'withwhom' they are with are rarely addressed. Clearly, with time, space and social contextplaying such key roles in the realization of specific fears of crime (Bannister 1993), toleave aside these issues is seriously to damage our understanding of the fear of crime.The operationalization of fear of crime questions in quantitative tools, which requireeasy to follow, short questions with precise predefined reference points, and hence donot take account of temporal, spatial and social contexts, has seriously limited both theconceptual development of this field and the quality of the data generated.

One of the most common tools in this field is the closed question using a Likert scaleto measure the extent of the fear of crime (e.g. see any one of the questions in thequantitative interview tool in the Appendix). With such a tool it is hard for respondentsto make truly qualitative distinctions between their feelings about particular crimes atparticular times and places. Respondents may well simply report generalized levels offear of crime, which may not adequately represent their actual emotions on any oneoccasion. Even where just one crime is concerned, a respondent's feelings about thatcrime may vary greatly with numerous other variables (e.g. social, geographical andtemporal dimensions), none of which is addressed.

In short, a simplistic, numerical, answer to a general dosed question cannot hope torepresent the breadth of experience and feelings about crime experienced by mostpeople.

Technical concerns

Several technical problems have been identified with the quantitative exploration ofthe fear of crime. Fattah (1993: 53) and Bernard (1987: 66) both report differences inresults when survey questions are asked in 'dosed' as opposed to 'open' forms. Closedquestions, it is argued, sensitize and direct respondents toward the set of answersoffered, and the empirical evidence supports this. A Harris poll (Harris and Assodates1975), for example, employed a 'dosed' question to evaluate whether an elderlypopulation was concerned about crime. They found that 23 per cent of the sample

661

by guest on February 29, 2012http://bjc.oxfordjournals.org/

Dow

nloaded from

Page 5: Farrall Et Al 1997 Measuring Fear of Crime

QUESTIONING THE MEASUREMENT OF THE 'FEAR OF CRIME'

considered crime to be a serious personal problem. Yin (1982), employing an 'open'question, found that only 1 per cent of a comparable elderly population consideredcrime to be a serious personal problem (reported in Fattah 1993:53-4). Itwould appearthat attempts to measure the extent of the fear of crime are grossly sensitive to thenature of the question asked. Closed questions, the most common tool in this field,greatly overestimate the incidence of the fear of crime. Closed questions are the stapleof many crime surveys. Considering just these two studies (Harris and Associates andYin), one can conclude that differences in the nature of the dependent variable (in thiscase open or closed), may result in differences in fear levels which could be as great asa factor of 23. It almost seems that any fear of crime survey will tell us little about thefear of crime, but a lot about the nature of the questions used.

Both Fattah (1993: 52) and Hale (1993: 9, 1996) question the assumption thatanybody is actually in a position to be able to make an accurate self assessment of theirown crime fear level. In a similar vein, Fattah (1993: 58), Schneider (1981: 831),Zauberman (1985: 35) and Block and Block (1984: 144) among others, have questionedthe ability of respondents to recall accurately their experiences of crime andvictimization. Yet, accurate recordings of respondents' fear levels and recall of pastvictimizations are central to this field of study. We are, after all, talking about thedependent variable and a major independent variable. Policy assumes this to beunproblematic, although research shows it to be highly problematic. This is, therefore,a major problem.

Taken together, these criticisms suggest that crime surveys ignore the meaningof events for respondents; turn 'processes' into 'events'; neglect that the fear of crimecan be a multi-faceted phenomenon; poorly conceptualize the fear of crime; ignoreimportant contextual variables (such as time and space); greatly influence thereported incidence of the fear of crime and rely too heavily on respondents' recall.From our review of the literature it is evident that the fear of crime is only partiallyunderstood at best. This has led to a 'vicious circle' where poor conceptualizationof the fear of crime (which has portrayed this emotion as static and unidimensional)has created methodological insensitivity, which has perpetuated poor operationaliza-tion. This poor operationalization is beset by technical inaccuracies which only serveultimately to muddy the conceptual waters. Where technical innovations have beenmade (e.g. referring directly to a type of crime rather than just crime in general orto general feelings of unsafety when at home alone at night), these have notimproved matters greatly. Overall, this situation has arisen from a limited under-standing of the phenomena at hand and from the use of a blunt instrument. Theseobservations are particularly significant because, as Hale (1993: 5, 1996) and Fattahand Sacco (1989: 210-11) note, the survey is the dominant instrument in researchon the fear of crime.

Assessing Validity through Methodological Triangidation

Given the predominantly quantitative nature of research in this field, some researchershave made calls for methodological triangulation in an attempt to rely less upon thesurvey tool (in particular, Bowling 1993: 246). However, this questions the nature andvalidity of the comparative results generated by quantitative and qualitative research

662

by guest on February 29, 2012http://bjc.oxfordjournals.org/

Dow

nloaded from

Page 6: Farrall Et Al 1997 Measuring Fear of Crime

STEPHEN FARRALL, JON BANNISTER, JASON DITTON AND ELIZABETH GILCHRIST

methodologies, and the extent to which they are compatible on epistemologicalgrounds.

Bryman (1984: 77) notes that quantitative and qualitative methodologies havedifferent views about what passes as warrantable knowledge. Quantitative surveysexhibit a 'preoccupation with operational definitions, objectivity, replicability, causality,and the like' (Bryman 1984: 77). Unstructured interviews and qualitative research ingeneral, is committed to 'seeing the social world from the point of view of the actor'(Bryman 1984: 77). Hence, the two epistemologies, broadly positivism and naturalism,produce quite different forms of knowledge, one measuring, and the other interpreting,reality.

Assessing compatibility, when undertaken, rarely elevates above the level of assessingwhether or not both yield similar research findings. One of the most thorough reportsof such an undertaking is by Belson (1986), who attempted to assess the validity ofquantitative tools by re-interviewing qualitatively respondents who had completed aquantitative interview. In one instance, respondents were asked about their chocolateconsumption during the previous week using a quantitative tool, and were thenimmediately re-interviewed in depth by a second researcher. Belson (1986: 64) reportsthat:

... the number of bars, etc. claimed in the first interview was about a fifth larger than the total numberfinally agreed in the intensive interview (which is interpreted as being nearer the truth).

Belson, (1986: 226), reporting another study (aimed at refining the use of semanticdifferential scales), states that:

About half the first [quantitative] interview ratings (51 per cent) did not stand up to the intensive[qualitative] challenge to them.

In a similar vein, Fielding and Fielding (1986: 74-8) also report discovering differentfindings as the result of using different methodologies from research on police attitudesto recruiting black police officers. They show (1986:76-8) that police officers gave quitedifferent answers at the quantitative stage of the research, to those given at thequalitative interviews stage. In short, the meaning of (and therefore the response to) aparticular research question is altered by the nature of the interview being undertaken.

Research directly relating to the fear of crime and methodological considerations ispresented by Maguire and Corbett (1987: 42-4). They show that self reported ratingsof the impact of offences are affected by the nature of the interview undertaken. Theyconclude that in-depth interviews produce three times as many 'badly affected'respondents as do survey instruments. Similar research findings are reported by King(1992), who shows that men, traditionally not seen as affected by crime in quantitativesurveys, report assault as a traumatic event when interviewed qualitatively.

Thus, from Belson's validity checks, the research reported by Fielding and Fielding,that reported by Maguire and Corbett and the evidence from King's research on malevictims of assault, it is clear that surveys and in-depth interviews do not producecompatible results. Indeed, it seems that a quantitative tool produces results at a quitegeneral level, while a qualitative tool produces responses which are more contextsensitive and specific in their focus.

The comments above pose a serious problem for those working in the fear of crimefield. To develop improved quantitative research methodologies and hence start to

663

by guest on February 29, 2012http://bjc.oxfordjournals.org/

Dow

nloaded from

Page 7: Farrall Et Al 1997 Measuring Fear of Crime

QUESTIONING THE MEASUREMENT OF THE 'FEAR OF CRIME'

answer some of the criticisms from Hale, Bowling, Fattah and Sacco et al., researchersworking in this field must do more than just triangulate their data by incorporatingqualitative techniques alongside quantitative research. In short, before triangulation isundertaken, researchers must first understand where (and why) the discrepanciesbetween quantitative and qualitative data lie, and then develop ways of reconciling thesedifferences. Identifying why discrepancies between quantitative and qualitative data setsexist, and the extent to which methodological triangulation may help to assess the validityof measuring tools and aid in efforts to improve them, is the task of the rest of this article.

Methodological triangulation: our position

At an epistemological level, we are concerned with the validity of the knowledge ofincidence of the fear of crime generated through survey research. At the technical,conceptual and operational levels, we shall employ qualitative interview data, in orderto evaluate the validity of quantitative research on the incidence of the fear of crime.Our epistemological stance is therefore primarily positivistic in that we are concernedwith the empirical measurement of the incidence of the fear of crime (rather than howthe fear of crime is socially constructed). We are concerned with the operationalizationof concepts and other technical issues, such as (for example) recall, and therefore willnot take a critical stance as to whether or not this knowledge is warrantable, but abouthow valid this approach is in a methodological sense.

Our data are the answers given by 64 respondents about their fear of crime atinterviews of both a quantitative and a qualitative nature (the instruments are describedin full in the Appendix). The quantitative interviews were completed before thequalitative interviews.' The discovery that respondents gave different answers to verysimilar (or identical) questions depending upon the nature of the interview beingundertaken was not made until after the interviewing was completed. As such thisobservation was not actively sought by the researchers, but occurred 'naturally' duringthe individual interviews themselves. We are writing about this because we were notexpecting it to occur, rather than because we were.

A Typology of Methodological Problems: Hypothetical and Observed

From the literature, it is dear that quantitative and qualitative research methodologiesdo not produce the same 'findings'. Where this has occurred in our data set we haveadopted the term 'mismatch'. In this section we offer a typology of possible explanationsfor these mismatches, noting for each the source of the mismatch (i.e. epistemological,conceptual, operational or technical).

We presume that observed mismatches are the direct result of the different methodolo-gies employed, rather than changes in respondent experiences or attitudes betweeninterviews. On average the time that elapsed between the two interviews with eachrespondent was about one month. For the purposes of this paper, the length of interval

1 This allow* us to make a substantial case about quantitative-qualitative mismatches, but because in all of the interviews thequantitative data collection preceded tlie qualitative data collection, we cannot address mismatches lliat may have been generatedliad the interview order been reversed, this would, of course, represent that next step in the investigation of die issues raised here.

664

by guest on February 29, 2012http://bjc.oxfordjournals.org/

Dow

nloaded from

Page 8: Farrall Et Al 1997 Measuring Fear of Crime

STEPHEN FARRALL, JON BANNISTER, JASON DITTON AND ELIZABETH GILCHRIST

between the quantitative and qualitative interviews is unimportant, as, within quantitativeepistemology, a one-hour interval is the same as a one-year interval. Further, eachrespondent was asked, on both occasions, to report their victimizations and fear of crime,for the same prior time period. Some of these problems are immanent in research intoother areas of social life,2 to which the following explanations will be pertinent.

The different epistemological focus of the interviews

Quantitative instruments appear to measure feelings on a very general level, whilequalitative interviews allow for these feelings to be expressed more discursively. Hencesocial and geographical contexts may be invoked in qualitative interviews which mayproduce mismatching with quantitative data in certain cases. For example, during onequantitative interview the respondent said that he was particularly worried about beingrobbed or assaulted (4 on a 1-5 scale where 1 represented not being worried at all ever,and 5 meant worrying a lot, all the time). In the later qualitative interview, when askedif he worried about robbery all the time, he said 'No, it's when I go out at night and seea group of people, that's when it starts coming in to your head, but if you're walkingaround during the day you don't think about it really at all, or when you're sitting athome you don't think "Oh, if I go out tonight, what'll happen?", you know'. In this(qualitative) sense he was a 3 at night, a 5 when he saw a group of people, and a 1 athome during die day. This was found to be the second most common type of mismatch(19 per cent, n=22) and contained a high number of'serious' mismatches (see belowfor issues surrounding the rating of mismatch seriousness).

The measurement of formless or concrete fears

This is an operational problem. 'Formless' (as opposed to 'concrete') fears may be beingmeasured in some instances (Ferraro and LaGrange 1987). This particularly applies toquestions such as 'How safe do you feel walking around alone in your neighbourhoodafter dark?' and 'How much do you worry about being a victim of any crime?'. Theseproduce different results to 'concrete' questions. For example, after the section of thequantitative interview concerned with concrete fears, one respondent was rated asexpressing a low level of fear. Later she was (quantitatively) asked about how much sheworried about a range of urban or life worries (including 'crime'). At this point shedescribed herself as worrying 'a lot' about crime, thereby contradicting her earlier reply.So, she was a 1 when asked quantitatively about particular crimes, but a 5 when askedqualitatively about 'crime' in general. This type was identified as being the cause of11 per cent (n= 13) of mismatches.

The nature of open/closed questions

Recall that open/dosed questions are alleged to produce different results (Yin 1982:242; Fattah 1993: 53; Bernard 1992: 66). Hitherto this explanation has been offeredto explain differences in results between one quantitative data set and another. For us,

1 See for example the McKeganey-Powcr debate in Addiction, McKeganey (1995) and Power (1006).

665

by guest on February 29, 2012http://bjc.oxfordjournals.org/

Dow

nloaded from

Page 9: Farrall Et Al 1997 Measuring Fear of Crime

QUESTIONING THE MEASUREMENT OF THE 'FEAR OF CRIME'

this explanation has implications for the comparison of our quantitative and qualitativedata sets. For example, when one respondent was asked how much he worried aboutbeing robbed or assaulted at the quantitative interview, he replied by placing himselfin the middle of the 1-5 scale. At the qualitative interview, when asked again about howmuch he worried about this type of offence, he said 'No, no, 'cause as I say, since I gotdone the first time [during the late 1970s], I'm very careful'. So he was a 5, is now a 1,and personally averages himself as a 3. This appears to be a particularly important typeof mismatch. It accounts for a large number of them (n=46, 40 per cent) and containsa very high number of 'catastrophic' and 'serious' mismatches.

Other possible explanations point to the respondent as being a source of mismatch.

A genuine change in fear level

An epistemological problem. Given that the data were collected for each respondentduring two time separated interviews, it could be that the changes observed are as aresult of a genuine change in fear levels. For example, one respondent who had recentlysuffered a vicious assault when first interviewed, reported being very fearful about thistype of offence. At the qualitative interview he said that these feelings were due then tothe recency of the event and that he now felt less worried. This type accounted for 8per cent (n=9) of the mismatches observed.

The meaning of the word worry' is variously interpreted by the respondent

This is a conceptual problem. During the quantitative interview, each respondent wasprobably aware that the word 'worry' in the question could be used as a surrogate forother words or feelings, and may have given a response for one or more of these otherfeelings, rather than for 'worry'. In other words, 'worry' means different things todifferent people. At the qualitative interview this 'worry' was sometimes redefined bythe respondent, hence creating a mismatch. The following extract comes from aninterview with one respondent who had appeared quite worried at the quantitativeinterview, but who gave answers which directly countered this interpretation duringthe later qualitative interview. When asked to expand in his own words about how hefelt about crime he said 'I think I have an awareness of it and I don't think I invite itbut at the same time I do not worry about it'. This type was found to account for14 per cent (n=16) of mismatches and contained a large number of 'serious'mismatches.

The interpretation of the question by the respondent

Bryman (1984: 81) recalls Herbert Gans's observation that surveys can 'only report whatpeople say they do and feel and not what a researcher has seen them say, do and feel'(emphasis added). In some cases, questions posed as being about actualities may beinterpreted as being hypothetical by the respondent This applies particularly toquantitative interviews because, due to their very nature, the interviewer is unable toprobe discursively whether or not the respondent actually does X or just thinks thatunder certain conditions they would do X. One respondent clearly interpreted the

666

by guest on February 29, 2012http://bjc.oxfordjournals.org/

Dow

nloaded from

Page 10: Farrall Et Al 1997 Measuring Fear of Crime

STEPHEN FARRALL, JON BANNISTER, JASON DITTON AND ELIZABETH GILCHRIST

following question during the quantitative interview as being hypothetical. When askedhow much she worried about somebody 'vandalizing her home or something outsideit' she placed herself at 5, the most worried end of the scale yet, at the qualitativeinterview she said, when asked about it again, 'It would worry me if it happened*. Here,in reality, is a 1 who might be a 5. This type accounted for 4 per cent (n=4) of themismatches observed.

Memory decay

This technical problem refers to forgetting victimizations, or when they occurred, atthe time of the interview. The following example of memory decay refers to forgettingan incident of vandalism. At the qualitative interview the respondent, when remindedthat he had earlier claimed to have had two experiences of this sort, said: 'I can'tremember the other time, I don't know why I said two.' A single victimizationmasquerading innocently as a multiple one (or vice versa}). This type accounted for4 per cent (n=4) of the mismatches observed.

'Careless' replies

A further technical problem. Some respondents may answer questions without a greatdeal of attention in order to get the interview over and done with (Belson 1986: 342).This also includes respondents giving an opinion rather than their opinion (p. 233). Inthis data set there were no mismatches which could be explained this way, but bothsources are obviously difficult to trace.

Concealment by the respondent

This refers to a general unwillingness on the part of the respondent to admit to beinga victim or to being fearful of crime. No examples of this were discovered. However,this is not unexpected, and would be untraceable unless respondents actually say 'Ididn't tell you that earlier because men don't cry' (for example).

Finally, a number of potential problems exist with the interviewers themselves. Thisincludes the impact of the interviewer's age, gender or class upon reporting ofparticularly sensitive topics, or unintentional changes in question wording. Again, noinstances of this were discovered in the data set and again tracing them would be difficult.

Discussion

In this section, we will present analysis of the data as it relates to the typology above.We shall show that mismatches identified cannot be explained by reference to anyparticular demographic variable.3 We shall focus in greater depth upon the nature of

3 The mismatch data have been analysed controlling Tor gender, age, area (wliidi can be seen as standing for socio-economicgroup within U>e confines of this dau set); self reported Tear, and risk of victimization scaling. None oTtliese appear to influencedie incidence of mismatching (see Tables 6-8, p. 073). In oilier words, mismatches would appear lo be evenly spread acroa all ofdiese groups and cannot be explained by reference to any particular demographic variable.

667

by guest on February 29, 2012http://bjc.oxfordjournals.org/

Dow

nloaded from

Page 11: Farrall Et Al 1997 Measuring Fear of Crime

QUESTIONING THE MEASUREMENT OF THE 'FEAR OF CRIME'

the mismatches (in terms of 'where' they are located and 'what' was being discussedwhen they were produced). We will then discuss the level of severity and the extent ofmismatch.

Table 1 shows the overall numbers of mismatches. It can be seen that a large numberof them occur as 'one offs' (n=20) or with just one other mismatch (n= 14). Thus, some34 of the 64 cases exhibit a low frequency of mismatch. However, some 15 cases exhibitthree or more mismatches. Taken together, the whole data set contains within it 114instances of mismatch.

The nature of the 'mismatches'

Given that the research had a number of questions on the fear of crime in bothquantitative and qualitative interviews, the mismatches that occurred could have doneso at any one of a number of points, both within either quantitative and qualitativeinterviews with a respondent, or between the quantitative and qualitative interviews.

Table 2 indicates that of the 114 mismatches identified, the vast majority (n=98,86 per cent) were located between the quantitative and qualitative interviews. Forexample, one respondent when asked 'How safe do you feel walking alone in this areaafter dark?' chose the answer 'Very unsafe'. However, at the qualitative interview shesaid 'I know exactly the areas where I could go for a walk and know reasonably well I'dbe reasonably safe, not one hundred per cent, but reasonably safe'.

A smaller number of mismatches (n= 13) occurred internally during the quantitativeinterview. For example, one respondent was rated as being high on an aggregate worry

TABLE 1 The overall incidence of mismatches

Number of mismatcheswithin a case Number of cases Total

01234367

15201444511

0202812162567

Total casesTotal mismatches

TABLE 2 The

Location of mismatch

64

location' ofthe

114

mismatches

Total

Quantitative and qualitative interviews 98Internal to the quantitative interviews 13Internal to the qualitative interviews 2New worry introduced at qualitative interview 1

Total 114

668

by guest on February 29, 2012http://bjc.oxfordjournals.org/

Dow

nloaded from

Page 12: Farrall Et Al 1997 Measuring Fear of Crime

STEPHEN FARRALL.JON BANNISTER, JASON DITTON AND ELIZABETH CrLCHRIST

level, but answered 'very little' when asked 'How much do you worry about becominga victim of crime?' during the same interview. Two further cases of internal mismatchesinvolved respondents contradicting themselves during the qualitative interview.

Quite clearly the vast majority of the mismatches occur between the quantitativeinterview and the qualitative interview. This indicates, together with the fact that mostmismatches are complex qualifications of earlier precise quantifications, that it ismethodology, rather than respondent change, which is the source of most mismatches.4

This is confirmed when we consider the subject of mismatches. Looking at Table 3,there is clear evidence that the 'worry about crimes' category is by far the largest, whichis disproportionate as both interviews were concerned equally (at dependent variablelevel) with crime and with worry about crime.

Examples of worry about crime mismatches are readily found. For example, onerespondent placed himself at the mid-point on the five-point scale concerning worryabout vandalism during the quantitative interview. However at the qualitative interviewhe claimed that he was 'not worried about that at all'. When the interviewer (who hadconducted both interviews) probed him further on this discrepancy, the respondentreplied 'I think I might have thought you meant would I be worried if somebody haddone that'. This clearly draws out the interpretative process which is performed by therespondent during the quantitative interview, and shows how questions based on actualworry levels (as designed and worded) are interpreted as being hypothetical by somerespondents.

A surprising number of previous victimizations were not recalled, despite the obvioussalience of the event for the respondent. One respondent recalled at the qualitativeinterview a particularly vicious burglary during which the burglars 'took a lot of ourstuff and wrecked the place, ripped the mattress and ripped the settee and urinatedand took stuff out of the cupboard and smashed it'. This incident was not reportedduring the earlier quantitative interview, despite the researcher asking at that timedirectly about the number of times the respondent had been burgled.

The non-'fear of crime' mismatches refer in the main to a question in the quantitativeinterview which asked respondents to indicate how much diey worried about a rangeof urban or life worries. The existence of such mismatches confirms that such problemscan exist in other areas of social research.

TABLE 3 The 'reference' of the mismatches

Reference of mismatch Total

Worry about crimes 89Risk of victimization 14Recall of past victimizations 6Non-'fear of crime1 5

Total 114

* And, as such, legitimate! our earlier daim dial 'the length of interval between the quantitative and qualitative interview! isunimportant, as a one-hour interval is epistemologkally the same as a one-year interval'.

669

by guest on February 29, 2012http://bjc.oxfordjournals.org/

Dow

nloaded from

Page 13: Farrall Et Al 1997 Measuring Fear of Crime

QUESTIONING THE MEASUREMENT OF THE 'FEAR OF CRIME'

Levels of 'mismatch' seriousness

As mentioned above, not all mismatches were rated the same in terms of their severity.For example, a respondent who claims not to worry 'at all' about burglary at thequantitative interview but at the qualitative interview indicates that they worry'occasionally', is vastly different to a respondent who makes the same claim at thequantitative interview but during the qualitative interview indicates that they worry 'allthe time' and 'cannot go out leaving the home unattended' for fear of burglary. Bearingthis, and other possibilities in mind, a three-point scale of mismatch seriousness wasdeveloped (ranging from 'catastrophic' to 'mild') in order to rate these differences.5

The 'catastrophic' level includes: responses which were direct contradictions, such asreporting very high levels of fear during quantitative interviews, but low levels duringthe qualitative interview (or vice versa); illogical statements (such as worrying 'a lot' abouthaving one's car stolen but not actually having a car); inabilities to make an estimationof fear or risk at one interview but not at the other and memory decay/recall mismatchesof significant victimizations.

The 'serious' level includes: qualifications and expansions of previous answers (suchas new words used to describe fear or worry, or social and geographical contexts beingbrought into play); less severe cases of inabilities to estimate worry or risk between thequantitative interview and the qualitative interview, and memory decay/recall of the lesssevere victimizations.

The 'mild' level includes: minor discrepancies or slight shifts in emphasis, andmemory decay/recall problems of a minor nature.

Table 4 shows the distribution of the mismatch scales. Some 36 per cent (n=41) ofthe mismatches were scaled at the 'catastrophic' level. The second level (n=56) accountsfor some 49 per cent of the mismatches with the remaining 15 per cent (n= 17) fallinginto the 'mild' level. Clearly, then, mismatch is a serious problem.

Mismatch seriousness and mismatch explanation

Each mismatch was coded not just for its seriousness but also for possible explanationof cause. This entailed a great deal of interpretation on the part of the researchers.Each real mismatch was recorded and notes made about preceding comments thatcould have explained it.

TABLE 4 The seriousness of the mismatches

Mismatch seriousness Total

'Catastrophic' 41'Serious' 56'Mild' 17Total 114

8 A Uircc-poiiu scale fcatastropliic'-'serious'-'niild') was developed rather tlian a more crude dkhotomous 'high-low' split. Thescaling of muinatdtes was validated by a member of the research team wlto had not initially classified them and who classified asub-sample 'blind'. Ninety per cent oftlie blind classifications were to the same point on die scale.

670

by guest on February 29, 2012http://bjc.oxfordjournals.org/

Dow

nloaded from

Page 14: Farrall Et Al 1997 Measuring Fear of Crime

STEPHEN FARRALL.JON BANNISTER,JASON DITTON AND ELIZABETH CILCHRIST

Table 5 crosstabulates the seriousness of mismatches by the explanation given foreach. As can be seen from the table, the four most common mismatches were judgedto be the results of using an 'open' as opposed to a 'dosed' question (n=46,40 per cent);the differences in the epistemological focus of the interview (n=22, 19 per cent); themeasuring of'formless' as opposed to 'concrete' fears (n= 13, 11 per cent) and the useof the word 'worry' as a surrogate for other words (n= 16, 14 per cent).

The majority of the 'catastrophic' mismatches are between 'open' and 'dosed'questions (n=21). Of the 46 mismatches that were explained by this explanation, 38were in the two highest levels ('catastrophic' and 'serious'), suggesting not just a commonproblem but also an important one. Note that the explanations which locate the sourceof error with the respondents in some way (interpretation, genuine change and memorydecay), taken together, only account for 15 per cent (n= 17) of the mismatches.

Itwould appear that the majority of the mismatches are located in explanations whichpoint to methodology as their source (the epistemological focus of the interviews,open/closed questions and the distinction between 'formless' and 'concrete' fears,71 percent, n=81). If one takes poor conceptualization as the root cause of the variousinterpretations of worry (conceptualization being logically prior to design andimplementation of the research instrument) then this Figure rises to 85 per cent (n=97)of mismatches.

At anotfier level, Yin (1982) cites evidence to suggest that not only are there differencesbetween 'open' and 'closed' questions in terms of their answers, but also that thesediscrepandes follow a uniform direction. 'Open' questions produce lower reported ratesof fear of crime; 'dosed' questions, higher rates. The data under consideration here werealso analysed for this. Not all mismatches applied to 'worry' about crime,6 but, of the 40'open' and 'dosed' mismatches that did relate to worry about crime, 37 confirmed Yin'sfinding that 'dosed' questions generate higher levels of fear. Clearly it is the methodwhich produces these mismatches rather than the respondents.

The importance of these findings

We feel that there are good reasons to view these findings as important. They areimportant because they highlight the range of the epistemological, conceptual,

TABLE 5 Mismatch scale by mismatch type

Type of mismatch 'Catastrophic' 'Serious' 'Mild' Total

Epistemological interview focus'Formless' or 'concrete' fears'Open' or 'dosed' questions"Worry' variously interpretedGenuine changeInterpretation of question by RMemory decay

Total

35

215322

41

166

1710421

56

32812-1

17

22134616944

114

8 Five niuiiutdies applied to risk assessments and one to worries about having a fire at home.

671

by guest on February 29, 2012http://bjc.oxfordjournals.org/

Dow

nloaded from

Page 15: Farrall Et Al 1997 Measuring Fear of Crime

QUESTIONING THE MEASUREMENT OF THE 'FEAR OF CRIME1

operational and technical problems which plague the quantitative investigation of theincidence of the fear of crime and which indicate cumulatively that the fear of crime issignificantly misrepresented. There are four of these.

First, as noted above, the vast majority of the mismatches were located betweenquantitative and qualitative instruments. Without wishing to suggest that one measureis any more valid than another (as Kaplan 1964: 202 notes, this is a fallacy), it suggeststhat problems exist with measurement somewhere, even if we are not exactly sure where.However, the fact that the vast majority of mismatches with the 'open' and 'dosed'question distinction follow the same pattern as noted previously suggests thatquantitative measures consistently overestimate levels of worry. If this is correct, thenquestions asked during quantitative research need to be re-thought. Given also that themajority of these mismatches were 'serious' or 'catastrophic', then this would suggestnot just that these mismatches exist (and therefore overestimate worry levels), but alsothat they may be exaggerating these worry levels substantially.

Secondly, we must consider the extent of the problem (in terms of how validly it wouldmeasure the fear of crime) if this study were to be conducted on a larger scale. Takingonly those mismatches relating to worry about crime (n=89, Table 3), we estimate thatsome 17 per cent of the responses do not concur with one another.7 This is not, however,a particularly high figure, as it must be remembered that the mismatching was notrecognized until after the field work had been completed. Had the respondents beenmore systematically interviewed with this methodological consideration in mind, theextent of mismatch could be much larger.

Thirdly, a number of the mismatches concerned the various interpretations of theword 'worry'. All of the following were offered as word substitutes for 'worry' byrespondents (either in direct response to being asked to offer new words, or unasked):distress; anger; shock; annoyance; thinking about crime, and consciousness orawareness of the possibility of crime. Interestingly, one respondent was rated duringthe quantitative interview as being consistently a very low worrier. At the qualitativeinterview she made a distinction between vxrrrying about and thinking about crime. Theresearcher then invited her to answer the same question (as used at the quantitativeinterview) but with the word 'think' in place of 'worry'. The respondent changed heroriginal score of 1 to 'a 4 or a 5'. While this is by no means a mismatch, it shows howthe use of language impacts upon the results gained.

While 'think' and 'worry' clearly meant different things to this respondent, thegeneral fact that some respondents were able to offer different words to describe howthey felt about crime, reinforces the assertion that the 'fear of crime' field may beplagued by poor conceptualization and subsequent poor operationalization. If thesewords are used interchangeably by both the academic community (Fattah 1993: 45-6),and by respondents, it suggests that the 'fear of crime' can mean many different thingsto different people, which introduces great uncertainty exactly where certainty has beentaken for granted.

7 This figure was calculated as follow: [64 interview] x [4 Offence Types] x [2 Dimensions (fear and risk)]»512.00. 89 as apercentage of 512-17.38.

It is, unfortunately, impossible to calculate similar rates for just the qualitative data, as they are, by definition, non-numerical.

672

by guest on February 29, 2012http://bjc.oxfordjournals.org/

Dow

nloaded from

Page 16: Farrall Et Al 1997 Measuring Fear of Crime

STEPHEN FARRALL, JON BANNISTER, JASON DITTON AND ELIZABETH CILCHRIST

Finally, another issue concerns people's worries across time and space. Worries orfears about crime have been conceptualized as enduring for individuals over time, withindividuals' worry levels holding constant. Because the interviews in this research wereundertaken with a time gap (albeit short) between them, we are able to make some briefcomments upon changes in worry levels as they were accounted for by the respondentsthemselves. The subtle shifts of feelings over time are missed when purely quantitativemethods are used to measure worry about crime. This has two effects. First, we cannotadequately relate previous experiences of crime to current feelings about crime, andsecond, our conceptualization is thus weakened. We need to understand changes inworry levels as much as we need to understand worry levels themselves.

In short, the fear of crime has been conceptualized without reference to time, spaceand social context. Hence it is seen as a static and enduring feature of an individual'slife. Even if the fear of crime had not been conceptualized in such a manner, thequantitative investigation of the fear of crime would still have been hampered by itsbeing epistemologically locked into measuring the fear of crime in a manner that isunlikely to elucidate variations in fear of crime levels across time and space.

General Statements

From the analysis presented, a number of general statements can be made. The vastmajority of cases exhibit instances of mismatching (n=49 out of 64), but these are mainlyrestricted to one or two instances per case (Table 1). Mismatching cases are spreadevenly across each geographical area (Table 6) and gender (Table 7). Age would alsoappear to be unrelated (Table 8).

In terms of the nature of mismatch, most would appear to occur between thequantitative and qualitative interviews (Table 2). Most of the mismatches refer to'worries' about crime (Table 3). In terms of the seriousness of mismatch, there is a biastowards the more serious end of the scale given that 36 per cent (n=41) of themismatches are 'catastrophic' and a mere 15 per cent (n=17) 'mild' (Table 4).Comparing the seriousness of mismatch to the explanation for it, it is apparent thatmost explanations relate to methodological concerns (rather than to errors associatedwith the respondent), and that the 'open' and 'closed' question distinction is the mostfrequent, and is skewed heavily towards the serious end of the scale (Table 5). Thissuggests that poor question design may be at the root of the problem. The use of theword 'worry' as a surrogate for other emotions or words was another of the most frequenttypes, and was also markedly skewed towards the more serious end of the scale (Table 5).

In short, mismatches stem from methodological issues, are more likely to be focusedover issues concerning worry, are more likely to be serious than minor and appear tobe best explained by reference to methodological explanations.

Methodological suggestions for future research

We conclude by noting that there are problems with both the measurement techniquesused, and the conceptual tools employed by researchers. Our words echo Hale, whowrites that 'theoretical and empirical chaos has been the order of the day in the studies

673

by guest on February 29, 2012http://bjc.oxfordjournals.org/

Dow

nloaded from

Page 17: Farrall Et Al 1997 Measuring Fear of Crime

QUESTIONING THE MEASUREMENT OF THE 'FEAR OF CRIME'

TABLE 6 Non-mismatches and mismatches by area

Area

Outer-city poorInner-city poorInner-city richOuter-city rich

Total

Area

Outer-city poorInner-city poorInner-city richOuter-city rich

Total

Age range

16-3940-64+65

Total

Mismatches

10141213

49

Non-mismatches

6243

15

TABLE 7 Mismatches by gender and area

Males

6677

26

Females

4856

23

TABLE 8 Mismatches by age range

Mismatching

2420

5

49

Non-mismatching

735

15

Total

16161616

64

Total

10141213

49

Total

312310

64

of the fear of crime' (1993: 13). However, we are also in a position to suggest a numberof issues to be considered in the future.8

1. First, more attempts at validating measures of the fear of crime need to beundertaken. This means not just research aimed solely at this task, but also thatsuch validation techniques should be incorporated into future crime surveys.

2. The fear of crime could fruitfully be measured as a multi-faceted phenomenon.This will allow for a greater conceptual understanding of the fear of crime. Thiswill mean asking questions that in future in addition measure emotional,cognitive and affective elements of it. From our initial research, we suggest thatone avenue researchers may explore is the extent to which a respondent thinksabout a certain crime, how afraid of it they are, and how angry they feel whenthey think of it happening to them.

3. Given that levels of fear of crime are unstable over time, change in levels of fear,and in emotional responses to particular events (such as victimization) must be

* This artide and a further article in Urn journal (Gildirist H aL, forthcoming) are the lira of what u anticipated to be a series ofpublications from this research project. Later ones will deal with many of the issues raised herein.

674

by guest on February 29, 2012http://bjc.oxfordjournals.org/

Dow

nloaded from

Page 18: Farrall Et Al 1997 Measuring Fear of Crime

STEPHEN FARRALL, JON BANNISTER, JASON DITTON AND ELIZABETH GILCHRIST

addressed. These could be measured using questions such as 'How did you feelimmediately you knew you had been victimized?' and 'How do you feel aboutthat victimization now?'. Time elapsed between victimization and stated feelingsshould be used to scale the salience of the latter. Similarly, respondents can beasked questions about how they felt before a nominated victimization as a way ofmeasuring the relationship between fear and victimization more accurately.

4. In order to measure on-going as opposed to intermittent fears of victimizationwe suggest using questions which ask about crime experiences in ways other thanthe usual 'last 12 months'. Questions can be phrased in a number of ways, but itis important that the respondent understands that they are being asked about asingle event or about a series of events that may have taken place at any time, andare being asked to explain what actually took place, and how this affected theirfeelings about crime, both then and later. These statements can subsequently becoded for quantitative analysis thus: 'on-going racist graffiti in neighbourhood,R worries for children's safety', 'assault, more than 5 years ago, made R lessfearful', and so on.

5. 'Formless' fear of crime questions could attempt to measure why an individualfeels unsafe while walking around in their neighbourhood after dark, after askingthe general question. This can be achieved by seeking levels of agreement to abattery of statements such as 'I feel unsafe walking around here after darkbecause I may be attacked'. This principle could even be extended to 'concrete'fear questions.

6. The social, temporal and geographical aspects of the fear of crime could bemeasured by asking respondents if any of the specific locales in the area in whichthey live are 'unsafe' during the day or night and why this is so. For example,'Are the shops nearby unsafe during the day?' with (for example) dosed responsessuch as 'Yes, because there is never anyone there' and 'Yes, because of the peoplethat go there'. Such offered responses have been developed from the analysis ofour qualitative data.

7. Questions should be developed which describe a fictitious everyday event (notnecessarily crime-related) and which subsequently ask the respondent to expresstheir feelings and actions if they were to be in that situation. This would helpcontextualize people's feelings about crime in their everyday lives. Thisapproach was first adopted by van der Wurff et al. (1989) and in the light of theirquantitative data and our own qualitative data this appears to be a fruitfulapproach.

Conclusion.

At the start of this article we reviewed the calls made by some authors formethodological triangulation in an attempt to rely less heavily upon the survey tool.While this is an admirable ideal, it perhaps assumes that the combination of qualitativeand quantitative data is straightforward. Making sense of data is never straightforward, making sense of quantitative and qualitative data which contradict oneanother is considerably harder. Given that crime surveys will always require some level

675

by guest on February 29, 2012http://bjc.oxfordjournals.org/

Dow

nloaded from

Page 19: Farrall Et Al 1997 Measuring Fear of Crime

QUESTIONING THE MEASUREMENT OF THE 'FEAR OF CRIME'

of quantitative data, we have made suggestions as to how qualitative data can be usedto improve the design of survey questions. These suggestions (made immediately above)incorporate qualitative data through open ended questions, more thorough codes andthe use of vignettes. To this end, our use of qualitative data to triangulate survey datahas proceeded at a stage earlier to that proposed by (for example) Ben Bowling (1993).In the light of the data we have to hand, this would appear to be a better solution to avery thorny problem.

We are fortunate that we have been in a position to examine in great depth the fearof crime at a methodological level. At least one of us (Jason Ditton) has extensiveexperience of being involved in crime surveys that fall foul of all of the criticisms levelledabove. Our interim conclusion is that the results of fear of crime surveys appear to bea function of the way the topic is researched, rather than the way it is. The traditionalmethods used are methods which seem consistently to over-emphasize the levels andextent of the fear of crime. It seems that levels of fear of crime, and, to a lesser extent,of victimization itself, have been hugely overestimated. The policy implications of thisshould escape no one. The political utility of the fear of crime is entirely dependentupon its being measurable. If, as we suggest, the fear of crime is not as easily measurableas was hitherto thought, then the remarkably rapid ascent of the fear of crime in a veryshort time during the 1980s might well in the future be surpassed only by its swifterdemise.

APPENDIX

Instruments Used

Quantitative interview

On a scale of 1 to 5 (with 1 representing not being worried at all ever; and 5 meaning worryinga lot all the time) how worried are you that somebody might:

— Break into your house (or try to break in) and steal things, or try to, or damage things?— Steal your vehicle, or things from it or off it, or do damage to it?—Vandalize your house or something outside it?— Rob you or assault you or threaten to do either?

How many times have you been a victim of any of these crimes in the last five years?How likely (with 1 representing very unlikely; and 5 almost certain to happen) is it, do you think,that you will be a victim of any of these crimes in the next year?How safe do you feel walking alone in this area after dark?1 =very safe; 2=fairly safe; 3=a bit unsafe; 4=very unsafeHow much do you worry about: Becoming seriously ill? Losing your job or being unable to finda job? Another member of your family losing their job? Becoming involved in a road accident?Your finances? Environmental pollution? Being a victim of any crime? Having a fire in yourhome? Noise from your neighbours?1 =a lot, 2=a bit, 3=not much, 4=very litde, 5=not at all

676

by guest on February 29, 2012http://bjc.oxfordjournals.org/

Dow

nloaded from

Page 20: Farrall Et Al 1997 Measuring Fear of Crime

STEPHEN FARRALL, JON BANNISTER, JASON DITTON AND ELIZABETH GILCHRIST

Qualitative interview

1: Worry about crimeYou said that you did/did not worry about-What do you mean by worry? [Intensity: mildly pissed off—terrified]Why worry about this?How often do you worry? Constant? Intermittent?

2. AreaWhat is it like?Safe or not? Why? [Relate to R and to specific crime]Darkness/time of day—any effect?

3. BehaviourWhat types of things do you do? Not do?Where do you go? How get there?Anywhere not go? Why?Clothing/dress—any effect?Avoidance strategies

4. VictimizationWhat happened?How did you feel before? and after?Who/ what helped? [Formal and informal forms of help]How feel now? (fear/risk)Why? (loss/damage/inconvenience/fear/discomfort)

5. OtherIdeas about other areasEffect of others' ideas—friends/ media etc.Culpability for victimization?Domestic violence location of dangerConcern—corporate crime? Health and safety issues—ferry disasters, Kings CrossUnderground fire? Large corporate fraud? Environmental crimes? Government deception?

REFERENCES

BANNISTER, J. (1993), 'Locating Fear: Environment and Ontological Security', in H.Jones, ed.,Crime And The Urban Environment. Aldershot: Avebury.

BELSON, W. A. (1986), Validity in Survey Research. Aldershot: Gower.BERNARD, Y. (1992), 'North American and European Research on Fear of Crime', Applied

Psychology: An International Review, 41/1: 65-75.BLOCK, C. R. and BLOCK, R. L. (1984), 'Crime Definition, Crime Measurement and Victim

Surveys'./ourna/ of Social Issues, 40/1: 137-60.BOWLING, B. (1993), 'Racial Harassment and the Process of Victimization', British Journal of

Criminology, 33/2: 231-50.BREWER, J. and HUNTER, A. (1989), Multimethod Research: A Synthesis of Styles. London: Sage.BRYMAN, A. (1984), The Debate About Quantitative and Qualitative Research: A Question of

Method or Epistemology?', British Journal of Sociology, 35/1: 75-92.CHAMBERS, G. and TOMBS, J. (1984), The British Crime Survey: Scotland. London: HMSO.

677

by guest on February 29, 2012http://bjc.oxfordjournals.org/

Dow

nloaded from

Page 21: Farrall Et Al 1997 Measuring Fear of Crime

QUESTIONING THE MEASUREMENT OF THE 'FEAR OF CRIME'

FATTAH, E. A- (1993), 'Research on Fear of Crime: Some Common Conceptual and MeasurementProblems', in Bilsky et al., eds., Fear of Crime and Criminal Victimisation. Stuttgart: FerdinandEnke Verlag.

FATTAH, E. A. and SACCO, V. F. (1989), Crime and Victimisation of the Elderly. New York:Springer-Verlag.

FERRARO, K. F. and LAGRANCE, R. (1987), The Measurement of Fear of Crime', SociologicalInquiry, 57/1: 70-101.

FIELDING, N. G. and FIELDING, J. L. (1986), Linking Data, Qualitative Research Methods, vol. 4.Sage: Newbury Park.

FIGGIE, H. E. (1980), The Figgie Report on The Fear of Crime: America Afraid. Part One: The GeneralPublic. Willoughby, OH: A-T-O Inc.

HALE, C. (1993), 'Fear of Crime: A Review OfThe Literature'. Report to the Metropolitan PoliceService Working Party on the Fear of Crime.

(1996), 'Fear of Crime: A Review OfThe Literature', International Review OfVictimology, 4:79-150.

HARRIS, L. AND ASSOCIATES (197 5), Myth and Reality of Aging m America. Washington, DC: NationalCouncil on Aging.

HOUGH, M. (1995), Anxiety About Crime: Findings From the 1994 British Crime Survey, Home OfficeResearch Study No. 147. London: Home Office.

HOUGH, M. and MAYHEW, P. (1983), The British Crime Survey: First Report, London: HMSO.(1985), Taking Account of Crime: Key Findings from the Second British Crime Survey. London:

HMSO.KAPLAN, A- (1964), The Conduct of Inquiry. San Francisco: Chandler Publishing Company.KING, M. B. (1992), 'Male Sexual Assault in the Community1, in G. C. Mezey and M. B. King,

eds., Male Victims of Sexual Assault. Oxford: OUP.KINSEY, R. and ANDERSON, S. (1992), 'Crime and the Quality Of life: Public Perceptions and

Experiences of Crime in Scotland', Central Research Unit Paper, Edinburgh: Scottish Office.MAGUIRE, M. and CORBETT, C. (1987), The Effects of Crime and the Work of Victim Support Schemes.

Aldershot: Gower.MAYHEW, P., ELLIOT, D. and DOWDS, L. (1989), The 1988 British Crime Survey. London: HMSO.MAXFIELD, M. (1987), Explaining the Fear of Crime: Evidence from the 1984 British Crime Survey,

Home Office Research Paper No. 41. London: Home Office.MCKEGANEY, N. (1995), 'Quantitative and Qualitative Research in the Addictions: An Unhelpful

Divide', Addiction, 90/6: 749-51.MUIR, C. (1994), Personal Communication, 1991 Census Data.PAYNE, D. (1992), Crime In Scotland: Findings from the 1988 British Crime Survey, Central Research

Unit Paper. Edinburgh: Scottish Office.POWER, R. (1996), 'Comment on McKeganey's Editorial", Addiction, 91/1: 146-7.SCHNEIDER, A. L. (1981), 'Methodological Problems in Victim Surveys and Their Implications

for Research in Victimology'./ourna/ of Criminal Law and Criminobgy, 72/2: 818-38.SELLTIZ, C , WRIGHTSMAN, L. S. and COOK, S. W. (1976), Research Methods m Social Relations, 3/e.

New York: Holt, Reinhart and Winston.SKOGAN, W. (1981), Issues in the Measurement of Victimisation, US Department of Justice,

Washington, DC: US Government Printing Office.(1990), The Police and Public in England and Wales: A British Crime Survey Report, Home Office

Research Study No. 117. London: HMSO.

678

by guest on February 29, 2012http://bjc.oxfordjournals.org/

Dow

nloaded from

Page 22: Farrall Et Al 1997 Measuring Fear of Crime

STEPHEN FARRAIX, JON BANNISTER, JASON DtTTON AND EUZABETH GILCHRIST

STRATHCLYDE REGIONAL COUNCIL (1994), Social Strategy for the Nineties: Priority Areas, July.Glasgow.

VAN DER WURFF, A., VAN STAALDUNIEN, L. andSTRiNCER, P. (1989), 'Fear of Crime in ResidentialEnvironments: Testing a Social Psychological Model', Journal of Social Psychology, 129/2:141-60.

YIN, P. (1982), 'Fear of Crime as a Problem for the Elderly', Social Problems, 30/2: 240-45.ZAUBERMAN, R. (1985), 'Sources of Information About Victims and Methodological Problems in

This Field', in Research on Victimisation, vol. 22. Council of Europe.

679

by guest on February 29, 2012http://bjc.oxfordjournals.org/

Dow

nloaded from