 |
Data sources, methods and methodology
|
|
|
|
DATA SOURCES, METHODS AND METHODOLOGY
Victimization surveys
All data are more or less dirty ( clean/dirty data ). As with hygiene, we may aspire to perfect cleanliness, but there always remain germs. Official crime records are, how- ever, notoriously dirty as a source of data on relative crime rates and on crime trends. The definition of crime changes and varies; reporting rates for crime events vary and fluctuate; and recording practices likewise differ from place to place and from time to time (Tseloni and Tilley, 2016). We have drawn on quite a wide range of available data, trying in all cases to use the best that we could find. We have drawn especially heavily on victimization surveys.
In response to the well-known and widely acknowledged limitations in official crime records, victimization surveys have been developed which attempt to estimate crime rates by asking detailed questions of those who may have experienced crimes to uncover whether they have suffered crimes of different sorts and, if so, how many. The questions relate to a reference period, most often a year, to enable annual
prevalence rates (proportion of the population experiencing a given crime) and inci- dence rates (typically the number of crimes per 1, 000 or 100, 000 population or potential targets) to be calculated, within specified confidence limits. Incidence rates are normally (and controversially) calculated with a ‘cap’ on the number of incidents that count for any victim experiencing multiple incidents (see Farrell and Pease, 2007). Applying a ‘cap’ is controversial because it affects repeat crime rates (the proportion of crimes which occurred against the same victim or target) and crime concentration (the number of crimes experienced by each victim on average). For this reason, criminologists often study the ‘uncapped’ crime figures captured in vic- timization surveys (Chenery et al., 1996; Osborn and Tseloni, 1998). Victimization surveys also ask follow-up questions on details of at least some of the criminal inci- dents victims have suffered. In addition, surveys often include additional ‘modules’ for subsets of respondents, asking a raft of supplementary questions, for example on security measures in place, cars owned, lifestyles and attitudes to the police and criminal justice system.
Many countries, such as England and Wales, the USA, Mexico and Chile, admin-
ister their own victimization surveys. In addition, there is an International Crime Victimization Survey (ICVS) that asks standard questions to respondents in partici- pant countries, to enable valid comparisons to be made of rates in different jurisdictions as well as the identification of international trends (van Dijk and Tseloni, 2012).
Victimization surveys are imperfect, of course, as sources on crime trends. Limited sample sizes (especially in relation to relatively rare crimes), non-response, omitted victim populations (such as the very young, the homeless and many collective entities such as schools or hospitals) and memory failure all put limits on what can be said and with what confidence. Victimization surveys, nevertheless, by common consent, provide more robust estimates of relative rates and trends, especially for high-volume crimes, than recorded crime figures. It is for this reason that the data we have used to track trends in crime and to try to test many of our security hypotheses have come from high-quality victimization surveys. We have, however, as will become clear, drawn on other sources – the best we can find – where necessary.
|
|
|
We made use of the five sweeps of the ICVS (in 1989, 1992, 1996, 2000 and 2005) to determine whether there had, indeed, been an international crime drop. This task is more complex than might at first sight be apparent. Not all countries participate in each sweep of the ICVS because of the monetary cost of conducting the survey. We were able to identify a subset of countries that had participated in enough sweeps to analyse overall trends. We found 26 that had taken part in at least three sweeps and used these to model the crime trajectories. We were able to extend existing analysis by the ICVS team (van Dijk et al., 2007) to confirm that there had been a rise followed by a fall in the main crime categories included in the ICVS (Tseloni et al., 2010). The most important crimes in individual countries followed similar trends: they tended to go up at around the same time and down at around the same time.
In addition, we were also able to identify the timing of the fall across different crime types: as shown in Figure 10. 1, not all offence types fell at the same time. Burglary and car theft fell from the late 1980s onwards. Then it was theft from cars and theft from people, which began to fall in the mid-1990s. And, finally, assaults started declining around the turn of the century. Another finding was that, since 1995, theft from cars and thefts from people fell the most. Theft from cars fell massively – by over three-quarters! Personal theft also fell by 60%. The smallest declines were in assaults but even these went down by one in five.
10
8
1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
 |
Years
FIgure 10. 1 International Trends in Crime Incidence, ICVS 1988–2004
Source: Tseloni et al. (2010): 383.
We made extensive use of the Crime Survey of England and Wales (CSEW, previ- ously known as the British Crime Survey or ‘BCS’) to begin to test our security hypotheses about the cause of the crime drop. We chose the CSEW because it was familiar to members of the research team (we had used it in previous studies), because it is recognized as a high-quality survey, because its supplementary questions spoke to various aspects of the security hypothesis and because it spans quite a long time – 35 years at the time of writing.
|
|
|
