Improving the Explanatory Content of Analysis Products using Hypothesis Testing

Transcription

1 108 Article Improving the Explanatory Content of Analysis Products using Hypothesis Testing Spencer Chainey Abstract Analysis is an integral part of police and public safety decision making if a crime problem is clearly understood, it can help identify the solutions that will most likely be effective. Although the profile of analysis has been raised in recent years, its routine production has often resulted in many analysis products often offering only a descriptive presentation of the problem that is being examined, rather than the one that is explanatory in its tone. In this article, we propose the use of a hypothesis testing methodology to improve the explanatory content of crime and intelligence analysis, and illustrate its use with an example of residential burglary in Oldham, Greater Manchester. We argue that this approach produces analytical products that are richer in explanatory and interpretative substance, helps to improve commissioning dialogue, and generates results that help to more specifically identify how a crime problem can be tackled. Introduction In the last 20 years, the growth of the intelligenceled paradigm in policing has placed a greater emphasis on the need to conduct analysis. The gathering of information and its interpretation is a key principle that underpins intelligence-led policing, be it for supporting the daily tactical and operational targeting of police patrols, assisting an investigation, or for identifying persistent issues that require a strategic response. The generation of good quality analysis is also at the heart of the problem-oriented policing approach introduced by Goldstein (1979, 1990). The production of analysis has been formalized in several countries with the introduction of more routinized management processes. For example, CompStat introduced in New York in 1994 (see Walsh, 2001 and McDonald, 2002 for more details on CompStat) and adopted in many other police departments in the World (e.g. UK (Home Office, 2005), Brazil (Beato, 2008), and Australia (Mazerolle et al., 2011) requires the systematic production of crime statistics and other analytical material to support this operational management process. In addition, the adoption of police management frameworks such as the UK s National Intelligence Model (NIM) (NCIS, 2000) are designed to better integrate analysis and intelligence into the core of all police business and decision making (Ratcliffe, 2008). These more Department of Security and Crime Science, University College London, 35 Tavistock Square, London WC1H 9EZ, UK. s.chainey@ucl.ac.uk Advance Access publication: 14 March 2012 Policing, Volume 6, Number 2, pp doi: /police/pas007 ß The Authors Published by Oxford University Press. All rights reserved. For permissions please journals.permissions@oup.com

2 Hypothesis Testing Approach for Analytical Products Article Policing 109 formal, systematic processes have resulted in the creation of a more standardized approach to analysis report production. In UK NIM terms, these include the strategic assessment (a document produced annually for identifying the key strategic priorities that require attention), tactical assessments (produced for regular meetings where decisions on the tasking and coordination of resources are discussed by the local police command team) and problem profiles (analytical products that aim to help better understand particular crime problems) (ACPO, 2005). With this increasing demand for analysis and the subsequent standardization of intelligence products has come the production of templates that aim to determine a consistent structure and content to these materials. Over time, the routine production of these analytical materials has led to many reports being constrained in their analytical creativity and most often only providing a general descriptive narrative of a crime or public safety problem, rather than a thorough understanding of why the problem exists or has recently emerged. In this article, we explore the use of hypothesis testing as a methodology for the production of analytical products. This approach is used for framing the direction and content of the analysis. The research relating to this work draws from 8 years of empirical study with police forces and community safety partnerships in the UK. In particular, it draws from a UK Economic and Social Research Council knowledge transfer secondment that placed the author for 8 working days of each month, for 6 months (November 2010 to April 2011) in Oldham (Greater Manchester Police) and Sunderland (Northumbria Police). We show that results from field testing this approach produces analytical materials that are richer in their explanatory content and help to more specifically identify the types of responses that are most likely to be effective. This is achieved without compromising the length of time it should take to produce useful analysis materials, and helps to ensure that written reports are concise and fit for the purpose of the audience for whom they are designed. We argue that the hypothesis testing methodology is easy to adopt, encourages more effective commissioning of analysis products, and provides a practical means of assisting in developing more effective analysis. The next section begins by determining the role that analysis should play in modern policing and community safety. We then explore the principles of the hypothesis testing approach and how it can be applied to assist the production of crime and intelligence analysis materials. We then describe a structure for the creation of analysis products using this hypothesis testing approach, with particular focus on the creation of problem profiles. Our principal focus is directed at the generation of analytical products that aim to more fully understand a persistent or emerging problem, rather than the production of tactical analysis that requires a quicker turnaround. That is not to say the proposed methodology does not apply to tactical analysis, but rather its testing to date has initially been applied to analysis that requires a deeper consideration of the underlying causes of a crime problem. We illustrate the use of this approach with an example of an analysis of burglary to residential properties in Oldham. We then discuss the practical use of the hypothesis testing approach, including the difficulties that users experienced in its adoption and the benefits they have identified. The role of analysis Analysis involves generating intelligence that helps to understand and explain the problem that is being studied (Chainey, 2008). Crucially, it involves an interpretation of what is causing the problem. A useful means for illustrating the role of crime analysis is Ratcliffe s (2004) 3i Model (Fig. 1). The role of the intelligence or analysis unit is to interpret the criminal environment or the specific crime problem that is of concern. This unit should then produce analytical materials that influence the decision-making on which responses to implement,

3 110 Policing Article S. Chainey Figure 1: The 3i Model, adapted from Ratcliffe (2004) to show bi-directional arrows. with these responses being effective in impacting upon the crime problem. In Fig. 1, bi-directional arrows have been added to Ratcliffe s original 3i Model due to the need for: the interpretation stage to be recognized as an iterative process; the need for decision-makers to be clear to the intelligence unit on what their requirements are (i.e. influence the content of the analytical materials); responses are monitored for their impact, and are revised or replaced with other solutions that are considered to be more effective (Chainey, 2012). In practice, rarely do these exact conditions exist. Instead, analysts are offered little guidance with the objectives for the analysis often being unclear (Weisel, 2005); have to face the challenge of a hierarchy and culture that may take little notice of civilian staff views (Cope, 2004); and often produce results that offer little more than a descriptive narrative of the problem, rather than an explanation on why it is present (Chainey, 2012). This failure to ensure that analysis products are explanatory in their content can in turn undermine the influence they then have on decision-making (Ratcliffe, 2008). If a decision-maker is faced with addressing a chronic issue with street robbery, a general descriptive problem profile is unlikely to tell them anything they did not already know. Importantly, the lack of specificity and interpretation in the analysis often makes it difficult to identify the mechanisms that are likely to be most effective in tackling the problem (Pawson and Tilley, 1997), and who is competent and most responsible for doing something about it (Laycock, 1996) (i.e. local agencies other than the police). That is, if the problem is poorly understood, deciding how to effectively tackle it is difficult. This then calls for the need to consider alternative methods to the design and content of analysis products that can help provide greater detail and more explanatory content. This though needs to be weighed against the requirement to ensure that the analysis can still be produced in a timely manner. The principles of the hypothesis testing approach Hypothesis testing is an approach routinely applied in science to help establish knowledge. It requires a true (or false) statement to be made that offers a plausible explanation about the problem. Testing the hypothesis results in coming to some conclusions. Hypothesis testing is a key principle in the crime science approach. The adoption of the scientific method using data and information, generating hypotheses, and building testable models to explain observations is aimed at helping to generate intelligence and evidence that can be used to determine how best to deal with crime problems (Laycock, 2005). Indeed the notion of hypothesis testing is not strange to policing. Criminal investigations rely upon developing plausible ideas for directing the gathering of intelligence that may help identify likely suspects, the motivation behind crime commission, how the crime was carried out and how the offender escaped from the scene of the crime. The use of hypothesis testing as a matter of routine in the production of intelligence products is not so evident. It is argued that its adoption can

4 Hypothesis Testing Approach for Analytical Products Article Policing 111 result in the production of better quality analysis products that are more explanatory in content because the natural outcome from testing a hypothesis is to come to a conclusion a conclusion that states whether there is or is not evidence to support the reason given for the crime problem. Another benefit of the hypothesis testing approach is that it identifies the data that are required. For example, if a hypothesis for an increase in metal theft stated that it was to do with the increased availability of metal on building sites, this would direct the analyst to source data that identified metal thefts from building sites in relation to other sites. This helps to avoid gathering all possible types of data relating to metal theft and instead focuses on gathering the data necessary to test the hypothesis. The concept of hypothesis development and testing for crime analysis has been previously discussed by Chainey and Ratcliffe (2005) and Boba (2009) who stress at the outset of an analysis project the importance of determining the questions that it should attempt to answer. This is also reflected in the structure of the problem-solving guides developed by the Center for Problem Oriented Policing ( These guides, covering over 60 different crime problems, stress the importance of considering certain questions that need answering at each stage of the SARA problem solving process (Eck and Spelman, 1987). The value of applying hypothesis testing to crime analysis has also recently been illustrated by Townsley et al. (2011). They argue that the Analysis of Competing Hypothesis can be used to make the reasoning that underpins the hypotheses more transparent, auditable, and address many of the cognitive biases that crime analysis can be prone to (Heuer Jr, 1999; Johnston, 2005; Groopman, 2007; Kebbell et al., 2010). In this article, we complement the previous research by describing a practical methodology for using hypothesis testing in a policing and public safety environment, and illustrate its adoption with an example. Structuring the content of analysis products using hypothesis testing The production of analysis using the hypothesis testing approach involves four stages: an overview of the problem deciding on hypotheses analysis discussion of conclusions and recommendations The overview involves clearly defining the crime problem. It is where, after an initial request for analysis, we should determine the magnitude and scale of the problem, trends (i.e. increasing trend, seasonal trends), and any specific information that helps to more clearly identify what the problem involves. For example, if the problem was initially voiced as being an issue with a recent increase in violent crime, the overview should show evidence that there has been an increase that should warrant concern, and if this increase is concentrated to just a few areas in a police district, rather than the whole district (see Clarke and Eck, 2003 for further illustration on how problems can be specified). The overview should lead to a clearer and more specific definition of the problem, for instance, from we have a problem with an increase in violent crime to identifying that the problem is specific to an increase in minor assaults and anti-social behaviour, most noticeably in the town centre. The overview should be concise (i.e. a maximum of three pages in length), but should provide enough detail for Stage 2: deciding on hypotheses. In this second stage, the key stakeholders of the problem should review the overview, state if they agree with the problem definition (and if not what else the overview should consider), and determine reasons for why they think the problem exists. The key stakeholders are those officers who have a keen interest in the problem, and/or are in positions of responsibility to potentially do something about the problem. Ideally, these key stakeholders should come from a number of different local

5 112 Policing Article S. Chainey agencies rather than a single agency. This helps to ensure a better balance on the reasons why the problem exists, and includes them at the outset if later they are considered to be well-placed to help tackle the problem. It is the role of the analyst to collate the hypotheses, and either at a meeting or via electronic means determine with the key stakeholders the hypotheses that should be tested. Initially, all hypotheses should be accepted. Each should then be qualified to help produce a short list of hypotheses. Two tests should be conducted to qualify each hypothesis: Can it be attributed to the problem that is being defined over the timeframe that the problem has emerged/exists: the hypothesis needs to be stated in terms that link it to the crime problem. From this, we can then determine if the hypothesis offers a sensible explanation. For example, an increase in burglaries in residential properties could be explained as being linked to insecure properties. However, for this hypothesis to hold we must be confident that during the time when there has been an increase in burglaries, this has coincided with an increase in houses becoming less secure. This helps us to qualify that housing design is unlikely to have contributed to the recent increase because the types of houses have not changed, and neither has there been any new construction projects in the area where burglary has increased. The reason for the recent increase in burglary could though be, to do with residents leaving windows open due to a recent prolonged spell of high temperatures. We can do something about it: it is better to test hypotheses that are more likely to come to conclusions that enable the key stakeholders to then respond with solutions that they can implement. For example, one reason that could be given for a recent increase in residential burglaries is due to a rise in unemployment. Although findings from testing this hypothesis could be interesting, it is unlikely that the key stakeholders involved in helping to address the problem can have a direct impact on boosting the local employment market. When determining a hypothesis to test, it is important to arrive at an explanation that is specific and clearly articulated. For example, if a reason for the increase in residential burglaries is to do with an increase in offenders, it is better to specifically describe why there has been a recent increase in offenders: is it to do with a recent increase in prison releases into the area of individuals known to have previously been prolific burglars, or is it due to an increase in offenders travelling from neighbouring areas to offend in the area where burglaries have increased? Hypotheses that are stated could also be location or time-specific. Indeed, the hypothesis testing approach should help identify that the problem in one area may be due to completely different reasons to those in another area of the same district. It is most unlikely that any crime problem can be easily explained in simple terms. The hypothesis testing approach should help to recognize that the problem may exist for several different reasons, and these vary by time and location. This qualification process will help to produce a short list of hypotheses for testing. These should be narrowed down to three to five hypotheses, simply because there is unlikely to be analytical capacity to test any more. To narrow down the short list, the key stakeholders should determine which hypothesis they believe to be the main reasons for the problem. It is also best to consider posing hypotheses that offer a balance between offender-based explanations (e.g. an increase in prison releases) and those that can help to determine if there has been a change in the opportunity structure, i.e. test if more opportunities to commit a certain type of crime have become available. The hypotheses should frame the direction and content of the analysis. The primary objective is to come to some conclusions that provide evidence

6 Hypothesis Testing Approach for Analytical Products Article Policing 113 that do or do not support each hypothesis. The type of analysis to conduct is determined by the hypothesis to be tested. For example, if a hypothesis states that the reason for a recent increase in residential burglaries is related to an increase in homes being left insecure (due to residents leaving windows open because of a recent prolonged period of hot weather), the analysis will need to identify if the volume and proportion of insecure burglaries has increased in line with the burglary increase, and if this increase is evident in those areas where burglary has increased the most. A framework common to crime analysis is to answer a compendium question (Ekblom, 1988, p. 11) that determines what offences have occurred, where and when they occurred, how they occurred, who was involved or what was the target, and why it has occurred. In practice, little is often recorded on why. Using the hypothesis testing approach, the focus is on testing why, using whichever of the techniques relating to What, Where, When, Who, and How are most suitable for testing each hypothesis (see NPIA, 2008; Clarke and Eck, 2003; Boba, 2009; and Osborne and Wernicke, 2003 for more details on suitable analytical techniques). It is vital that the analysis includes some interpretation, rather than just a description of the problem. The hypothesis testing approach naturally leads to the discovery of evidence that can help explain why the problem exists, and in turn makes it easier to interpret these conclusions. This leads to the fourth stage. The analyst alone should not be responsible for interpreting the findings and determining the conclusions. Instead, the key stakeholders should be involved in this process, informed by the results of the analysis, and agree on the evidence. The evidence may be questioned for its validity, with the requirement to then find other evidence that does or does not support the initial conclusions. This stage is best conducted by holding a meeting involving the key stakeholders and the analyst. It is at this stage that the key stakeholders should use the evidence from the analysis to help decide how the problem can be addressed. Applying the hypothesis testing approach to crime problems an analysis of residential burglary increases in Oldham In this section an example of the hypothesis testing approach is illustrated. This analysis was conducted by the analytical unit in the Oldham Community Safety Partnership, with supporting guidance from the author. This unit includes a police intelligence analyst from Greater Manchester Police and a strategic partnership analyst. Previous problem profiles on burglary (and other crime types) were rich in statistical content, but offered little explanatory substance. Space restrictions preclude the inclusion of all the content from the analysis, but the example includes sufficient information to illustrate the use of the hypothesis testing methodology to readers. Between November 2010 and February 2011, the district of Oldham in Greater Manchester, experienced an 18% increase in burglaries in residential properties compared with the previous 4-month period. This was an increase of 91 burglaries. Further analysis revealed that this increase was not consistent across the district but was particularly concentrated in one area Q2 (Table 1), and that another area had actually experienced a reduction in burglaries (Q5). Further analysis presented in an overview of the burglary problem identified a very small number of emerging problem areas that were most responsible for the 18% increase (using an analytical method described by Ratcliffe, 2010). As a result of this overview, a number of hypotheses to explain the increase were posed by Oldham s Neighbourhood Policing Teams, a police Superintendent, Crime Prevention Officers and analysts from the Oldham Community Safety Partnership. These were short-listed to four: one was associated with there being more offenders, the second was explained as a decrease in the effectiveness of a crime prevention scheme, the third was associated with the targeting of gold jewellery, and the fourth was a seasonal explanation associated with darker evenings.

7 114 Policing Article S. Chainey Table 1: Residential burglaries in Oldham from November 2010 to February 2011 compared with previous periods Neighbourhood Number of residential burglaries Percentage change in residential burglaries (with volume change in brackets) November 2009 to February 2010 July October 2010 November 2010 to February 2011 November 2010 to February 2011 same period previous year, % (n) November 2010 to February 2011 previous 4 months, %(n) Q (22) +4 (5) Q (63) +65 (53) Q (19) +18 (14) Q (8) +27 (19) Q ( 5) 22 ( 23) Q (8) +53 (23) Hypothesis 1: The increase in burglary is attributable to an increase in burglary offenders, mainly as a direct result of an increase in prison releases. The first hypothesis required an analysis on whether there had been an increase in the number of active burglars in the area, and whether their offending had contributed to the increase. The analysis was split into three stages: to explore whether there had been an increase in burglary offenders; identify if there had been an increase in prison releases of known burglary offenders; and identify the amount of offending that known burglary offenders who had recently been released from prison had made to the increase in burglary. Between November 2010 and February 2011, there was a 6% increase in the number of known residential burglary offenders in Oldham (seven more offenders). While the only data available to test this hypothesis related to burglaries that had been detected, the detection level during this period compared with previous periods had remained unchanged, initially suggesting that this small increase in offenders could have contributed in part to the increase in burglaries, i.e. if there are more burglars active, there is likely to be an increase in the number of burglaries. An analysis of prison releases identified that between November 2010 and February 2011, there had been a 52% increase in the number of offenders released from prison into Oldham who had been convicted for burglary. However, only one of these prison releases was known to have committed a single burglary between November 2010 and February Oldham employs an integrated offender management programme that involves a number of local agencies. Ex-offenders receive intensive supervision immediately after their release from prison, with the aim to minimize re-offending and maximize their rehabilitation. The scheme in Oldham, known locally as Spotlight, is seen as an example of good practice (Home Office, 2010). Whereas certain individuals released from prison may go undetected for burglaries they have committed soon after their release, it would be expected that due to the very close monitoring of these individuals, many more would have been known to have committed burglaries during the period from November 2010 to February Further analysis revealed that of the top nine most prolific (known) offenders during this period, six were not subject to Oldham s integrated offender management programme, and only five offences were linked to four offenders who had been released

8 Hypothesis Testing Approach for Analytical Products Article Policing 115 from prison into Oldham between July and October This analysis helped to conclude that as only one of the recent prison releases was linked to a burglary during the period from November 2010 to February 2011, it was unlikely that an increase in individuals released from prison was the main explanation for the increase in burglaries in Oldham. Although there had been a small increase in the number of known burglary offenders, it was unlikely that this reason was solely attributable to the recent increase in burglaries in residential properties in Oldham. It is possible that offenders have become more prolific in their activities, but at this stage, as this was not one of the initial hypotheses to test, analysis at some later stage could explore this. Six of the nine most prolific offenders (based on offences detected) were not on the integrated offender management programme. This led to the suggestion that some attention to these individuals, and attention to the process for updating who should be included on the integrated offender management programme, may be required. Hypothesis 2: The increase in burglary is attributable to a decrease in the effectiveness of cocooning in these areas, leading to a higher level of repeat and nearrepeat victimization. Over the last 3 years, the Oldham Community Safety Partnership has placed particular attention on minimizing the risk of burglary victimization, following an initial incident. Repeat victimization of burglaries is most likely to occur soon after an initial incident (Polvi et al., 1990; 1991; Farrell and Pease, 1993; Johnson et al., 1997). In recognition of this, Crime Prevention Officers in Oldham visit a property that has been burgled within 18 hours of the incident, and immediately implement solutions to minimize the risk of future victimization. This scheme has helped to reduce burglary repeats to 3.1% (July October 2010), when in most areas, repeat victimization levels are more of the order of 7 16% (Weisel, 2005). Research also indicates that properties near to a recently burgled property are also at a heightened risk of burglary victimization, but that this risk decays rapidly over time, similar to patterns shown for repeat victimization (Bowers et al., 2004; Johnson et al., 2007). In an effort to prevent neighbouring properties being burgled, a cocooning scheme operates in Oldham that involves the provision of crime prevention advice to these neighbouring properties. As a result of recent budget cuts, it was considered that resources for targeting crime prevention to recently burgled and neighbouring properties were being stretched and this was having a detrimental impact on the effectiveness of the scheme. Between November 2010 and February 2011, only 1.2% of all burglaries in residential properties were repeats. This was a reduction on the previous 4-month period, and at a time when burglaries had increased by 18% in Oldham. This extremely low level of repeat victimization was indicative of the effectiveness of the continual activities of Crime Prevention Officers targeting recently burgled properties. In the same period, 9% of all burglaries were near-repeats, an increase from 7.2% on the previous 4-month period. Further analysis revealed near-repeats were now taking place with a higher level of intensity, within 2 days and within 200 m around an initial event (compared with within 2 days and within 300 m previously). This increase in near-repeats was responsible for 20% (an additional 17 offences) of the recent burglary dwelling increase. The results from this analysis helped to conclude that repeat victimization was not a contributing factor to the recent burglary increase. Only 1 in 100 properties burgled were subsequently burgled again, encouraging the continuation of the active work of Crime Prevention Officers to minimize burglary repeats. However, near-repeat victimization levels had increased, and on further scrutiny, opportunities to engage with neighbouring residents to help minimize their future burglary risks were not being taken advantage of. If near-repeat victimization could be reduced to the levels

9 116 Policing Article S. Chainey experienced in July October 2010, this alone could contribute towards a 4% overall reduction in burglary dwellings. On average, there were four burglaries per day in Oldham, suggesting that it would be practical to implement a tactic that utilized Neighbourhood Policing Teams to more effectively engage with neighbours close to the recently burgled properties and provide tailored crime prevention advice that could minimize their future risk of burglary. Hypothesis 3: The increase in burglary has been driven by an increase in gold jewellery thefts, particularly in Asian neighbourhoods, as a result of an increased market for second hand jewellery in the cash for gold retail outlets. Of the total Oldham population, 12% is Asian (Indian, Pakistani, Bangladeshi), compared with an England average of 4.6%. Gold plays a symbolic cultural role in Asian communities and is therefore a popular type of jewellery. In 2010, the price of gold rose by 37% (Goldprice.org, 2010), and a number of new cash for gold retail outlets opened in Oldham. It was considered that this increase in the price of gold meant that gold jewellery was now increasingly being targeted by burglars and that this targeting was specifically being directed to Asian households. Local cash for gold stores now made it easier than ever to sell gold stolen from house burglaries. To test this hypothesis, the analysis involved identifying if there had been an increase in the proportion of residential burglaries where gold had been stolen, if these burglaries were being targeted at Asian households, and the contribution that this increase in thefts of gold had made to the district-wide and local increases in burglary. Prior to the increase in burglaries in Oldham, 11% of all burglaries included the theft of gold jewellery. Between November 2010 and February 2011, this proportion increased to 13%. This increase in burglaries where gold jewellery was stolen was measured to account for 15% of the total increase in residential burglaries. Further analysis revealed that there was no evidence for the specific targeting of Asian households, and that there were no specific hotspots for jewellery theft the proportion of burglaries where jewellery was stolen was similar across Oldham. This allowed the analyst to conclude that whereas gold jewellery had become a more common target, the overall increase was quite small. There was no evidence to suggest that any particular ethnic group or locality was being targeted, but rather, areas where there had been an increase in jewellery burglaries were areas that had experienced an increase in burglary generally. Hypothesis 4: The increase in burglary is attributable to an increased opportunity for burglars to offend in the early evening due to the extended hours of darkness over the winter. It is widely recognized across Greater Manchester that the risk of residential burglaries increases during the winter months. This is due to darker late afternoon and early evening hours providing more opportunities for burglars to observe whether houses are unoccupied, and to break in unnoticed. In response, Greater Manchester Police operate an autumn and winter crime prevention programme to help minimize this risk. In Oldham, while this risk in increased burglaries is recognized, there was little evidence of specific strategic crime prevention responses, other than general messages to residents about keeping their homes secure. Operationally, during November 2010 to February 2011, the police response was most focused at targeting possible suspects and detecting offences, rather than exploiting supporting opportunities to minimize the increased vulnerability of residents to burglary victimization. Indeed, a hypothesis relating to darker evening hours was not suggested by any of the senior or neighbourhood police officers in Oldham, but instead was posed by one of the analysts. If Greater Manchester Police were aware of this increased risk of burglary in the winter months, then the programmes that were designed

10 Hypothesis Testing Approach for Analytical Products Article Policing 117 to address this should show evidence that in recent years there had not been a significant seasonal difference between the time of the day when burglary occurs. To test this hypothesis, the analysis involved exploring if the time of the day patterns for when burglary occurred using an aoristic technique to counter for the time and date reporting range for burglaries (Ratcliffe, 2002) were different during the period burglaries increased, compared with those in the previous 4 months. Figure 2 shows that temporal patterns for when burglary occurred between November 2010 and February 2011 and between July and October 2010 were similar between the hours of and 14.59, but between and 20.29, there were many more burglaries during the winter. This difference was equivalent to 90 more burglaries, and an 18% increase in Oldham. Further analysis identified that it was likely that offenders in Oldham were taking the opportunity of darker evenings to commit more burglaries during the winter months, i.e. effectively extending their hours of business. The pattern had repeated in each of the previous 3 years, and therefore it was highly likely that it would happen again in the winter of 2011/12. The analysis also showed that the areas affected most by this increase in offending during afternoon and early evening hours were the areas that had experienced an increase in near-repeat victimization. The analysis conducted in Oldham has provided fresh evidence to inform and influence discussions on the most suitable responses to implement. In particular, the analysis revealed two highly predictable characteristics about residential burglary in Oldham: in the Winter of 2011/12, it is likely that burglaries will increase; properties that have been burgled or are near to those that have recently been burgled are at a heightened risk of being burgled. Crime prevention responses in Oldham for tackling repeat victimization appear to be extremely effective, with key principles from this approach now being applied to help address near-repeat victimization. It was critiqued that the responses Figure 2: Residential burglary in Oldham by the time of the day of the offence, comparing November 2010 to February 2011 with July October 2010.

11 118 Policing Article S. Chainey that had previously been implemented to address the winter increase in burglaries may not have had the desired effect, therefore needed re-examining. This helped to highlight the need for targeting clear messages to residents in burglary hotspots about the practical actions they could take to minimize their burglary risk. The integrated offender management programme in Oldham appeared to be effective in reducing the re-offending of those on the programme, but a review of the existing process for monitoring active and prolific burglary offenders would be useful to ensure that the best people for current disruption and management were identified and registered on the programme. The analysis also helped to dispel some theories, such as it being unlikely that the rise in burglaries was due to an increase in the number of burglary offenders being released from prison. Discussion The production of analysis for providing intelligence and evidence to support the targeting of enforcement and prevention responses has become a standard process in many police and law enforcement agencies. However, with the requirement to inform operational management processes, the production of analysis has become routinized, often to the point of it lacking creativity in its production, and failing to provide materials that are explanatory in content. In turn, this is seen to restrict the influence that analysis products can have on deciding which responses are most likely to be effective and who is best placed to implement them. If the crime problem is poorly understood, it makes it difficult to determine which responses will have most impact. The hypothesis testing methodology is designed to produce evidence-based analytical conclusions. The approach helps to frame the direction and content of the analysis, and overcome the production of information that is merely general and descriptive. Instead, it helps to produce analytical materials that are more explanatory in content, encourage interpretation, and help to explore specific features about the crime problem. The approach of generating analysis using hypothesis testing does not provide all the answers as to why the crime problem exists, but can help produce intelligence documents that are richer in their analytical substance, and offer conclusions that help explain several of the main likely causes of the crime problem. The field testing of this approach in Sunderland and Oldham has helped to examine the practicality of the approach (further examples of analysis using the hypothesis testing approach are available in Chainey, 2012). In most cases, the overview provided enough information to help the key stakeholders pose hypotheses. Where they required more information, convening a meeting to initially discuss the crime problem and the requirements for the analysis helped these stakeholders to offer their explanations. The challenge then was to articulate these more clearly, more specifically (i.e. avoiding general statements), qualify them, and determine which ones should be tested in the analysis. Using the qualification process described in this article helped to determine the hypothesis that would be tested. This process of encouraging key stakeholders to pose hypotheses also led to the better commissioning of the analysis product. Rather than an analyst just being briefed on the need to produce a problem profile on a crime problem, and the objectives being ambiguous, the requirement for key stakeholders to suggest hypotheses helped to improve the dialogue with the analyst on exactly what was required. If a number of different agencies were involved in the hypothesis posing stage, this was also found to lead to a richer range of hypotheses being suggested. Encouraging key stakeholders to pose hypotheses also helped to better involve them in the analysis production process, and in some cases, certain stakeholders assisted in interpreting the results of the analysis while it was being developed. This helped lead to the production of analytical documents that were more fit for purpose of immediate

12 Hypothesis Testing Approach for Analytical Products Article Policing 119 use by decision-makers. Feedback on the problem profiles that followed the hypothesis testing approach has also indicated that the documents were much more useable in helping to determine the types of responses that were likely to be most effective in tackling the crime problem that had been analysed. To date, many analysts have followed the What, When, Where, Who, Why, and How structure to their production of analysis, with a tendency to feel compelled to record something under each item. The content of these analysis products often tend to be general and lack specificity, with little actually on Why. The focus of the hypothesis testing approach is on examining Why, using whichever of the What, When, Where, Who, and How lines of enquiry are most suitable. This means that analysts do not necessarily need to learn new techniques, but instead they apply techniques they are most likely to have used in the past or have become aware of through research. This helps to avoid performing unnecessary analysis if it does not contribute to helping test a hypothesis. A weakness in the methodology proposed in this article is the risk of the analysis being influenced by cognitive biases. This is where prior beliefs and assumptions, judgements on the likelihood of events, a bias towards evidence that is vivid and personal rather than abstract and statistical, and the failure to distinguish between correlation and causation may lead to incorrect judgements being made (Heuer Jr, 1999). Townsley et al. (2011) propose a possible solution to this by employing an Analysis of Competing Hypothesis matrix that weighs each of the alternative hypotheses in terms of the evidence and conclusions they offer. A key principle in this approach is to seek evidence that falsifies the hypothesis, rather than seek evidence that supports it. This concept of falsification is also considered to be a more rigorous method for testing hypotheses because the strongest explanations are those that will have the least amount of evidence against them (Townsley et al., 2011). The example used in this article was on residential burglary. Analysis that follows the hypothesis testing approach need not just be for crime problems, but can also be applied to other policing and community safety issues, e.g. re-offending by young people, fall in confidence in the police. Hypotheses can be stated for crime, disorder, anti-social behaviour, and other community safety problems; with the testing of them leading to helping explain why the problem exists. The integration of the hypothesis testing approach with the problem-solving guides provided by the Center for Problem-Oriented Policing offers a means of assisting practitioners in improving their understanding of a wide range of local crime and public safety problems. Conclusions The production of analysis is a fundamental component of intelligence-led and problem-oriented policing, and the need to be evidence-based in deciding how crime and other public safety issues can be addressed. Operational management processes such as CompStat and the UK s National Intelligence Model have importantly helped to raise the profile of analysis, and its integration into the framework of police business and decision-making. However, with these more routinized processes has come the tendency for more routinized analysis. This has meant that analysis products have become constrained in their creative content, often offer little other than a descriptive, general narrative on the problem and lack explanatory substance. Hypothesis testing naturally leads to the formation of conclusions. The use of a hypothesis testing approach in the production of police and community safety intelligence materials helps to frame the direction and content of the analysis, helps to identify the types of analysis to conduct (and the data to use), avoids conducting analysis that is unnecessary,

13 120 Policing Article S. Chainey and helps to draw out evidence that can explain why the problem exists. Its use has been tested in two UK police force areas (Oldham and Sunderland), and has provided valuable intelligence that has helped to explain crime problems that have recently emerged, or have persisted for some time. Police and community safety agencies are facing increasing pressure on their budgets, with the need to ensure more than ever that resources are effectively used. This has meant that the analytical function has come under a great degree of scrutiny, with many of these agencies reducing their analytical capacity. Responding to crime and public safety problems will obviously continue, but with smaller analytical capacity, it is vital that the analytical function provides intelligence that helps to decide specifically how these problems can be tackled. Analysis that is produced using a hypothesis testing approach can help to better understand why a problem exists, and in turn offer a more effective means of deciding more specifically the types of responses that are most likely to work. Funding This work was supported by the Economic and Social Research Council (grant number RES ). References Association of Chief Police Officers (ACPO). (2005). Guidance on the National Intelligence Model. Wyboston: ACPO NPIA. Beato, C. (2008). Comprehendo e Avaliando: Projetos de Segurança Publica. Belo Horizonte: Editora UFMG. Boba, R. (2009). Crime Analysis with Crime Mapping. Los Angeles: Sage. Bowers, K. J., Johnson, S. and Pease, K. (2004). Prospective Hotspotting: The Future of Crime Mapping?. British Journal of Criminology 44(5): Chainey, S. P. (2008). Crime Analysis. In Wakefield, A. and Fleming, J. (eds), The SAGE Dictionary of Policing. London: Sage. Chainey, S. P. (2012). Guidance and Practice Advice on Problem Profiles. London: University College London. Chainey, S. P. and Ratcliffe, J. H. (2005). GIS and Crime Mapping. Chichester: John Wiley and Sons. Clarke, R. and Eck, J. E. (2003). Become a Problem Solving Crime Analyst. London: University College London. Cope, N. (2004). Intelligence Led Policing or Policing Led Intelligence?: Integrating Volume Crime Analysis into Policing. British Journal of Criminology 44(2): Eck, J. E. and Spelman, W. (1987). Problem Solving: Problem-Oriented Policing. Newport News. Washington DC: Police Executive Research Forum. Ekblom, P. (1988). Getting the Best Out of Crime Analysis. London: Home OfEce. Farrell, G. and Pease, K. (1993). Once bitten, twice bitten: repeat victimisation and its implications for crime prevention. Police Research Group, Crime Prevention Unit Paper 46. London: Home Office. Goldprice.org. (2010). Goldprice percentage annual change Europe. (accessed 1 September 2011). Goldstein, H. (1979). Improving Policing: A Problem-Oriented Approach. Crime and Deliquency 25: Goldstein, H. (1990). Problem-Oriented Policing. New York: McGraw-Hill Inc. Groopman, J. (2007). How Doctors Think. New York: Houghton MifFin. Heuer, R. J. Jr (1999). Psychology of Intelligence Analysis. Langley, VA: Center for the Study of Intelligence, Central Intelligence Agency. Home Office. (2005). Crime Mapping: Improving Performance. London: Home Office. Home Office. (2010). Violent Crime Management: Case Studies. London: Home Office. Johnson, S., Bowers, K. and Hirschfield, A. (1997). New Insights into the Spatial and Temporal Distribution of Repeat Victimisation. British Journal of Criminology 37(2): Johnson, S. D., Bernasco, W., Bowers, K. J. et al. (2007). Near Repeats: A Cross National Assessment of Residential Burglary. Journal of Quantitative Criminology 23(3): Johnston, R. (2005). Analytic Culture in the US Intelligence Community: An Ethnographic Study. Washington, DC: Central Intelligence Agency. Kebbell, M. R., Muller, D. A. and Martin, K. (2010). Understanding and Managing Bias. In Bammer, G. (ed.), Dealing with Uncertainties in Policing Serious Crime. Canberra: ANU E Press, pp Laycock, G. (1996). Rights, Roles and Responsibilities in the Prevention of Crime. In Bennett, T. (ed.), Preventing

14 Hypothesis Testing Approach for Analytical Products Article Policing 121 Crime and Disorder. Cambridge: University of Cambridge Cropwood Series. Laycock, G. (2005). Defining Crime Science. In Smith, M. J. and Tilley, N. (eds), Crime Science: New Approaches to Preventing and Detecting Crime. Cullompton: Willan Publishing. Mazerolle, L., McBroom, J. and Rombouts, S. (2011). Compstat in Australia: An Analysis of the Spatial and Temporal Impact. Journal of Criminal Justice 39(2): McDonald, P. P. (2002). Managing Police Operations: Implementing the New York Crime Control Model CompStat. Belmont, CA: Wadsworth. National Criminal Intelligence Service (NCIS). (2000). The National Intelligence Model. London: NCIS. National Policing Improvement Agency (NPIA). (2008). Practice Advice on Analysis. Wyboston: ACPO NPIA. Osborne, D. A. and Wernicke, S. C. (2003). Introduction to Crime Analysis. New York, NY: The Haworth Press. Pawson, R. and Tilley, N. (1997). Realistic Evaluation. London: Sage. Polvi, N., Looman, T., Humphries, C. and Pease, K. (1990). Repeat Break-and-Enter Victimization: Time Course and Crime Prevention Opportunity. Journal of Police Science and Administration 17(1): Polvi, N., Looman, T., Humphries, C. and Pease, K. (1991). The Time Course of Repeat Burglary Victimisation. British Journal of Criminology 31(4): Ratcliffe, J. H. (2002). Aoristic Signatures and the Temporal Analysis of High Volume Crime Patterns. Journal of Quantitative Criminology 18(1): Ratcliffe, J. H. (2004). Strategic Thinking in Criminal Intelligence. Sydney: Federation Press. Ratcliffe, J. H. (2008). Intelligence-led Policing. Cullompton: Willan Publishing. Ratcliffe, J. H. (2010). The Spatial Dependency of Crime Increase Dispersion. Security Journal 23(1): Townsley, M., Mann, M. and Garrett, K. (2011). The Missing Link of Crime Analysis: A Systematic Approach to Testing Competing Hypotheses. Policing 5(2): Walsh, W. F. (2001). CompStat: An Analysis of an Emerging Police Managerial Paradigm. Policing 24(3): Weisel,D.L.(2005).Analyzingrepeatvictimization.Problem- Oriented Guides for Police Problem-Solving Tools Series No.4. Washington, DC: U.S. Department of Justice.