Abstract
During the past few years, there has been renewed interest in the issues and problems over using administrative and statistical spatial units as proxies for residential and workplace neighbourhoods in academic and policy research. This paper investigates the relationships between neighbourhoods defined by 693 people living or working in Cardiff. Differences between resident- and workplace-defined neighbourhoods were identified and the degree to which these varied was analysed with respect to personal characteristics such as gender, age, ethnic group and time spent in the neighbourhood. Young and retired people both have the largest neighbourhoods, possibly because both have more time to spend in their local area than other age groups. Gender differences were evident in the neighbourhoods of respondents who both live and work in an area, with females having larger neighbourhoods than males; this may reflect the types of employment undertaken by females who live and work in the same place. People living or working in close spatial proximity did not necessarily share the same neighbourhood definitions despite having similar personal characteristics.
1. Introduction
During the past few years there has been renewed interest in, and debates around, the issues and problems over the conceptualisation of neighbourhoods in academic and urban policy research (for example, de Marco and de Marco, 2010; Galster, 2001; Gauvin et al., 2007; Schaefer-Daniel et al., 2010; Weden et al., 2008). These debates have been of a particular concern in research that is trying to capture the effects of neighbourhood and place upon socioeconomic processes, such as in the fields of crime (for example, Brunton-Smith and Sturgis, 2011; Hipp, 2007; Innes et al., 2009). One of the main concerns is the recognition that the spatial units adopted in research can shape and influence the research outcomes, both in in-depth area studies and in comparative spatial analyses (Coombes, 2000; Hipp, 2007). Although these debates are not in themselves new (for example, Openshaw and Taylor, 1979), there remains a fundamental disagreement as to how to define neighbourhood and this is despite a long tradition of research concerned with analysing a neighbourhood’s effect on residents (Roosa et al., 2009; Campbell et al., 2009). Indeed, the task of conceptualising and operationalising neighbourhoods remains a daunting challenge even though many researchers approach this methodological task in ways that can be described as naïve with the definition of neighbourhood frequently buried in the methodological detail (Gauvin et al., 2007).
Substantial research has focused on the degree to which pre-defined spatial units such as administrative areas or census and postal geographies capture the underlying socioeconomic phenomenon under investigation and whether the particular geographical unit adopted is actually appropriate for the outcome of interest (for example, Flowerdew et al., 2008; Gauvin et al., 2007; Hipp, 2007; Schaefer et al. 2010; Stafford et al., 2008; Weden et al., 2008). There has been a lot less research investigating self-reported neighbourhood definitions and how these vary by both respondents’ characteristics and where they live in relation to one other. In this case it can be argued that understanding the variations in self-reported definitions of neighbourhood is an important first step in understanding how pre-defined neighbourhoods based on existing geographies affect research outcomes. The paucity of research is arguably because resident-defined neighbourhoods are hard to create or locate (Campbell et al., 2009) and those studies that have been undertaken have tended to use ethnographic approaches with a small sample size making generalisation difficult. Also, research in this area has concentrated on residential neighbourhoods with no mention of workplace neighbourhoods, despite the growing interest in workplace geographies and the fact that neighbourhoods are often the site of both living and working (Cockings et al., 2011).
The few recent studies that have investigated self-reported measures of neighbourhood in any detail with large sample sizes have had a US focus. Such studies have shown that neighbourhood definitions can vary significantly by gender, age, ethnic group and the length of time a person has resided in a neighbourhood (for example, Coulton et al., 2001; Campbell et al., 2009; Roosa et al., 2009). Many researchers have reported gender differences (Kitchen and Blades, 2002) with men generally having larger defined neighbourhoods than women and this has usually been attributed to women being less mobile than men with women’s patterns of movement being limited by factors such as fear of crime (Roosa et al., 2009). However, Kitchen and Blades (2002) have postulated whether the increase in female mobility in recent years has reduced these gender differences and whether men and women display similar-sized self-reported neighbourhoods. In terms of age differences, studies have highlighted the differences between young people and older people in general and teens and adults in particular. So Campbell et al., (2009) showed that adults tended to agree more than teens about where to draw neighbourhood boundaries, even if the respondents live in the same household. Other studies (for example, Roosa et al., 2009; Coulton et al., 2001) have shown that, generally, neighbourhood size tends to be a reflection of age and gender, with young and middle-aged women drawing smaller neighbourhoods than men, and older people drawing smaller neighbourhoods than younger people. Differences in race and ethnicity can serve to separate people and affect the definition of perceived neighbourhood boundaries. Generally, immigrant populations tend to have smaller and more defined neighbourhood boundaries with a higher degree of consensus than the dominant non-migrant population as illustrated by Campbell et al. (2009) and Roosa et al. (2009). In addition, size of self-reported neighbourhoods tends to be positively related to length of residence with highly mobile people tending to report smaller neighbourhoods. These studies have also raised questions of whether people living close to each other share similar neighbourhood definitions. Coulton et al. (2001) analysed maps drawn by residents and showed that residents who lived in close proximity differed markedly from one another in how they defined their neighbourhood and this disagreement existed even amongst residents with similar personal characteristics. In particular, the size of the neighbourhood had the most variation amongst respondents. Similar conclusions were reported in the studies by Campbell et al. (2009) and Roosa et al. (2009).
The aim of this paper is to therefore investigate how people’s definitions of the neighbourhood within which they live or work vary by their personal characteristics and the time spent in the neighbourhood. In contrast to much previous North American focused research, the context of this study is the British city of Cardiff, which was chosen due to the availability of a large dataset of resident- and workplace-defined neighbourhoods. The definition and operationalisation of neighbourhood used in this research is discussed in detail in the methodology but in brief respondents were interviewed and asked to identify the location of their home or workplace on a large-scale paper map and asked to show the areas that they frequently used for daily activities that they might think of as their neighbourhood and were asked to consider places where they would walk, shop, attend a religious building or frequent on a regular basis. Glaster (2001) has developed three concepts of neighbourhood that can be used as a theoretical framework to help investigate the agreement amongst self-reported neighbourhood maps: generality, accordance and congruence. Generality is the degree to which individuals’ neighbourhoods vary by different attributes, such as personal characteristics. Accordance is the correspondence between neighbourhoods for different individuals living or working in close proximity. Congruence is the degree to which an individual’s neighbourhood maps onto pre-determined boundaries, such as administrative areas. Together, the concepts of generality and accordance can be used to unpack the relationships between self-reported definitions of neighbourhood, and the concept of congruency their relationship with other geographies. In this paper, the focus will be on the concepts of generality and accordance only in relation to self-reported neighbourhoods, with the concept of congruency with other neighbourhoods being the focus of future research. A set of spatial metrics for each neighbourhood map are constructed to allow comparative analysis. These relate to area, perimeter and shape of the neighbourhoods and are described in the next section.
The two objectives of the research are to investigate the differences between resident- and workplace-defined neighbourhoods and how these vary based on personal characteristics such as gender, age, ethnic group and time spent in the neighbourhood (which corresponds to Glaster’s definition of generality) and to investigate whether people living or working in spatial proximity to one another share similar neighbourhood definitions (or Glaster’s definition of accordance). Given past research on self-reported neighbourhoods, we posit the following six hypotheses. First, male respondents will, on average, have larger defined neighbourhoods than female respondents. Second, the size of neighbourhoods will have a negative relationship with age such that younger respondents will have larger neighbourhoods than older respondents. Third, respondents of a non-White ethnic origin will have smaller and more contained neighbourhoods than those of a White ethnic origin. Fourth, the size of the neighbourhood will be affected by whether a person lives, works or both lives and works in the area—it is hypothesised that respondents who live and work in the same area will have a larger defined neighbourhood than those who just live in the area and in turn these respondents will have a larger defined neighbourhood that those who just work in the area. This is based on the assumption that neighbourhood size is positively associated with the time an individual spends in that neighbourhood on a daily basis. Fifth, the length of time that a respondent has lived or worked in the area will have a positive relationship on the size of the neighbourhood. The sixth and final hypothesis is that respondents living in close spatial proximity to one another will share similar neighbourhood definitions compared with those living further away, although these may vary by the personal characteristics outlined earlier.
The paper is structured into five sections. The next section describes the dataset and the methodology used to compare respondent-defined neighbourhoods. Section 3 is an analysis of respondent defined neighbourhoods with respect to differences in respondent characteristics; section 4 is an analysis of the consensus of the spatial metrics pertaining to neighbourhoods of respondents located in close spatial proximity. The final section presents a summary of the analysis and a discussion of the findings.
2. Methodology
The data for the research were provided by the Universities’ Police Science Institute (UPSI) at Cardiff University which has been assembled as part of a research project into neighbourhood security and reassurance in Cardiff. The data consist of a representative sample of the population of Cardiff and were collected using a method called intelligence-orientated neighbourhood security interviews (i-nsi) that integrates the principles of cognitive interviewing (Geiselman et al., 1986; Fisher and Geiselman, 1992) into a qualitative GIS approach (Elwood and Cope, 2009). The interviews are administered using a bespoke map-based computerised personal interview software and analysis package. It has been designed for use by front-line practitioners (Lowe and Innes, 2011) and was initially developed during an extensive programme of research conducted as part of the National Reassurance Policing Programme (Innes et al., 2004). It has since been refined and developed through extensive field testing (Innes, 2005; Innes et al., 2008; Innes and Roberts, 2008) and designed to provide police with a ‘rich’ community intelligence-led picture of insecurity in neighbourhoods.
To ensure spatial coverage, interviewees were sampled by census output area (OA). Each OA contains a relatively consistent population count of approximately 250 to 300 people. The interviewees were selected based on whether they either lived or worked in the OA; that they had cross-cutting, overlapping ties to people in the local area; that they tend to spend comparatively large amounts of time traversing the local area—for instance, walking children to school; and that they tend to invest time in the local area. The i-nsi methodology aims for one interview to be undertaken per OA but due to the large size of Cardiff and resource constraints, only 75 per cent of OAs were included, representing 693 people. This is a large sample size compared with previous studies, however, and will allow statistical analytical tests to be undertaken on the data.
Prior to data collection, each interviewer undertook a two-day training course that was provided by UPSI to ensure that they understood the i-nsi methodology, had the necessary interview skills and were proficient in using the bespoke GIS software. As part of the training course, all interviewers undertook an accompanied interview with a respondent that was overseen and quality assured by an UPSI trainer. The interviews took place during a 10 week period between January and April 2008 and each interview lasted an average of 34 minutes. The interviews were administered in a face-to-face setting, either at the respondent’s home or at a location of their choice. The interviewer used a laptop running a custom-written software programme structured around a participatory GIS, displaying pre-loaded Ordnance Survey 1:10,000 scale basemaps of the respondent’s local area. The interview process was a hybrid methodology that combined aspects of semi-structured interviews with structured approaches.
The interview process was guided by the bespoke GIS package that was loaded on the interviewers’ laptops. Each interview commenced with the interviewer explaining to the respondent that they will be asked some questions about their local area, starting with questions about their neighbourhood and leading on to identifying any crime or social disorder that is making them feel unsafe both inside and beyond their neighbourhood. The interviewee was then handed a paper map (Ordnance Survey 1:10,000) of the area that replicated the on-screen map that was displayed on the interviewer’s laptop. The first questions that the interviewee was asked were basic demographic questions such as age, gender and ethnicity. The interviewer also recorded the length of time that the respondent has lived and/or worked in the area.
The interviewer asked the respondent to look at the paper map of their area and in order to orientate themselves geographically they were asked to identify their home or work location on the map. The software prompts the interviewer using a pop-up message box to ask specifically, “Looking at the map, can you tell me where you live/work?”. The interviewee would identify the location of their home or workplace on the paper map and the interviewer would digitise the location on-screen.
The software then prompts the interviewer to ask the respondent to look at the paper map and to define their neighbourhood boundary. It is suggested by a message box that the interviewer may ask “Looking at the map, can you show me the areas that you use on a regular basis that you might think of as your neighbourhood?”. The interviewer may prompt the respondent to think about the areas on the map that they frequently use for daily activities and define the boundary of their home or work neighbourhood using their finger on the paper map. Most interviewees found this an intuitive task and would instantly start tracing their perceived neighbourhood boundary. If the interviewee was still unsure regarding what was meant by ‘neighbourhood’, the interviewer would suggest that the interviewee considers places where they would walk, shop, attend a religious building or frequent on a regular basis. It is at this point that some interviewees would mention ‘additional neighbourhoods’ or areas that they definitely would not visit, this would help them to identify their own neighbourhood boundary. When interviewees mentioned areas/neighbourhoods that they would not frequent, this would often be because they would feel less safe or that they considered the residents different from themselves. At each juncture that the interviewer digitised—either the home/work location or perceived neighbourhood boundary—the respondent would be asked to verify that the digitised information on-screen was correct.
Figure 1 shows all of the 693 respondent-drawn maps together with the location of their home or work and an example of maps for three of the respondents living in central Cardiff. This shows the variation in size and shape of the neighbourhoods as well as the fact that many of them overlap and in some areas they share distinctive common boundaries. Personal characteristics of the interviewee such as their gender, age, ethnicity, whether they lived or worked in the area, and the length of time they had lived or worked in the area were recorded and are summarised in Table 1. An equal number of males and females were interviewed with 10 per cent of respondents belonging to a non-White ethnic group—reflecting the ethnic composition of Cardiff. Around half of respondents were between 36 and 65 years old. Over two-thirds of respondents who were residents in the neighbourhood and just over half of respondents who worked in the neighbourhood had done so for over five years. The data allowed us to characterise respondents into three different neighbourhood types: respondents who were residents (Lives Only); respondents who worked in the neighbourhood (Works Only); and respondents who both lived and work in the neighbourhood (Lives and Works). Over half of respondents (394) were resident in the neighbourhood with almost a sixth (112) working there and just over a quarter (187) living and working in the neighbourhood.
Self-reported neighbourhoods with residential/work location and Cardiff LSOAs.
Descriptive summary of respondents’ characteristics (percentages)
An analytical framework for comparing the respondents’ neighbourhood maps was developed based the theoretical concepts developed by Galster (2001) and on methodological techniques developed by Coulton et al. (2001) and Campbell et al. (2009). In terms of measuring spatial proximity, respondents who lived in the same lower-layer super output area (LSOA) were defined as being close neighbours for comparison purposes. These are small areas of around 1500 people and 400 households with an average area in Cardiff of 0.73 square km. LSOAs have been designed so that they are as similar as possible in terms of population size with a secondary set of criteria based on similarity in the characteristics of housing tenure and dwelling type and that they do not cross topographical barriers such as rivers, railway lines or main roads (Cockings et al., 2011). Using LSOAs to capture the spatial proximity of respondents has the advantage of reflecting general similarities in the housing stock where respondents live and not crossing barriers that may impede social activity (Campbell et al., 2009). On average, four to six respondents were assigned to a single LSOA.
GIS analysis was undertaken to generate spatial metrics of each respondent map including area, perimeter and shape. Respondents’ maps are compared without reference to spatial proximity to gauge the degree of similarity in respondents across the city differentiating by neighbourhood type and length of residence and by characteristics such as gender, age and ethnicity as studies have shown that these can alter the perceptions of neighbourhood boundaries (Roosa et al., 2009). This analysis relates to Galster’s (2001) concept of generality. The shape of each neighbourhood is measured by the Cox index (Niemi et al., 1990), which is the ratio of the area of the neighbourhood to the area of a circle with the same perimeter
where, A is the area and P is the perimeter.
This has the advantage that it is bounded by 0 (as the neighbourhood becomes more elongated) and 1 (as the neighbourhood becomes more circular). In terms of measuring the accordance of maps of respondents who live or work in close spatial proximity, the maps of respondents in the same LSOA are compared and the degree of consensus or accordance with the spatial metrics will be measured using the coefficient of variation (CV). A CV <1.0 suggests low variance in the variables being measured and hence greater similarity in the spatial metrics of respondents’ maps. In addition, inferential statistics have been used to determine the significance of the differences between the spatial metric measures where appropriate. These are typically mean difference (t-statistic) and analysis of variance (F-statistic) tests and are reported in the text.
3. Measuring Generality: Substantive Differences among Respondents’ Maps
Comparing generality amongst respondents addresses the first five hypotheses: that male respondents will have larger defined neighbourhoods than female respondents; that the size of neighbourhoods will have a negative relationship with age; that respondents of a non-White ethnic origin will have smaller and more contained neighbourhoods; that the size of the neighbourhood will be affected by whether a person lives, works or both lives and works in the area; and the length of time that a respondent has lived or work in the area will have a positive relationship on the size of the neighbourhood. Table 2 reports the summary of the spatial metrics of respondents by neighbourhood type and length of residence. The size of respondents’ neighbourhoods varied depending upon whether they were residents or worked in the area. As hypothesised, on average, workplace neighbourhoods were the smallest with an area of 0.48 square km and a perimeter of 2.72 km respectively, with residential neighbourhoods being 1.6 times larger. Respondents who both live and work in an area had the largest neighbourhoods on average at almost 1 square km—twice as big as workplace neighbourhoods and over a quarter times bigger than residential neighbourhoods. T-tests (independent means) showed that the differences in average area and perimeter between the three groups of respondents were statistically significant at the 5 per cent level.
Summary of spatial metrics by neighbourhood type and length of residence
Average areas are also larger the longer someone has lived or worked in a neighbourhood. Neighbourhoods are 1.5 times larger if respondents have lived there for five or more years and this difference is nearly twice as large for respondents who both live and work in a neighbourhood. T-tests (independent means) show that these differences are statistically significant at the 5 per cent level. However, there is no statistically significant difference in the size of workplace neighbourhoods regardless of the length of time a respondent has worked there. The time a respondent has spent living or working in an area also makes no statistically significant difference to the average length of their neighbourhood perimeter. The average shapes of neighbourhoods are very similar amongst all respondents regardless of whether they live or work in the neighbourhood or the length of time they have been there. Therefore, length of time only has a statistical impact on the size of residential, but not workplace, neighbourhoods.
There are also important differences between genders, as reported in Table 3. For respondents who either live or work in an area, males have slightly larger neighbourhoods in terms of their average size and perimeter compared with females although the differences in these means are not statistically significant at the 5 per cent level. However, for respondents who both live and work in the same neighbourhood, the area and perimeter of these neighbourhoods for female respondents are 1.7 and 1.2 times bigger than those for male respondents and these differences are statistically significant. Similar to previous findings, male and female respondents living or working longer in an area have larger neighbourhoods than more recent arrivals. There are no discernible differences in the shapes of the neighbourhoods between genders for the three neighbourhood types or the length of time a respondent has been in a neighbourhood. With the exception of the statistically significant difference in the size of neighbourhoods between males and females respondents who live and work in the same area, these findings negate the stated hypothesis and confirms what Kitchen and Blades (2002) have stated about the differences between males’ and females’ neighbourhood definitions becoming generally more similar in recent years possibly due to increases in female activity patterns.
Summary of spatial metrics by neighbourhood type and gender
The small sample size of respondents from different ethnic groups makes it difficult to say anything substantive or statistical about the differences in neighbourhoods between White and non-White respondents although tentative observations can be made. As hypothesised, White respondents tend to have larger residential neighbourhoods than non-White respondents and non-White females have much smaller neighbourhoods than all other respondents. However, non-White female respondents tend to have larger-sized workplace neighbourhoods than other respondents, although none these differences is statistically significant at the 5 per cent level.
There is an interesting non-linear relationship between age of respondents and their neighbourhoods shown in Figure 2 which challenges the stated hypothesis. In terms of respondents who only live in the neighbourhood, those who are 21 years or younger have, on average, the largest neighbourhoods in terms of area and perimeter and these are larger for males. The general trend is then of a decline in neighbourhood size by age group until the age of 65 years old. Then between the ages of 66 to 70, average neighbourhood size increases so that it is similar to those aged 21 years or younger (although the increase for female respondents is much smaller). Neighbourhood size then declines again with age, although male respondents’ neighbourhoods continue to remain larger. In terms of the shape of neighbourhoods, however, there are very little differences in responses between genders and age groups. Analysis of variance (F-statistic) shows that the overall differences in size and shape by age categories are not significant at the 5 per cent level, although pair-wise t-tests (independent mean) show that there are significant differences in size between respondents in the 21 years and under and the 66 to 70 year age categories and the remaining age categories for male respondents but not for female respondents.
Variation in spatial metrics by age of respondents, gender and neighbourhood type.
The increase in neighbourhood size in later life coincides with retirement age in males and therefore could reflect the fact that they spend more time at home. The decline after 70 years is probably due to mobility problems associated with old age and perhaps perceptions of neighbourhood safety. The only other feature of note is that, between the ages of 22 and 35 years, female respondents’ neighbourhoods are much larger on average, and this is the only age group within which this occurs. This could be due to females in this age group having young children and therefore being more active in their neighbourhood in terms of walking to nursery, school, etc.
A different relationship exists between age and size for workplace neighbourhoods. Here neighbourhood size progressively increases with age and reaches a peak with the 36–50-year age group. Male respondents tend to have larger sized neighbourhoods, particularly in the 36–50-year age group. There is very little difference between gender and age groups and the shape of workplace neighbourhoods, although the male respondents’ neighbourhoods become slightly more elongated as they get older. None of the mean differences was statistically significant at the 5 per cent level.
For those respondents who both live and work in the same area, the relationship between age and neighbourhood size in terms of area and perimeter is again non-linear but follows a distinctive inverted-U-shaped distribution. Here, neighbourhood size increases and peaks at the 36–50-year age group and then declines to retirement age. The other feature of note is that female respondents tend to have larger neighbourhoods on average, contrasting with the discussions earlier. Analysis of variance (F-statistic) revealed that there were no significant differences between neighbourhood size and age categories for either gender but pair-wise t-tests (independent mean) show a statistically significant difference in the size of neighbourhood between males and females aged 36–50 years old, with females having the larger neighbourhoods supporting the previous findings. In terms of shape of neighbourhood, there is a trend that neighbourhoods become more circular with age and this is true for both genders although these differences were not statistically significant.
Therefore, statistically significant age differences appear to only occur for males aged 20 or under and between 66 and 70 years old in respect to the size of their residential neighbourhood definitions. This is probably a reflection of the activity patterns of males in these two age groups which may be more spatially extensive than those of females of similar ages. In comparison, females aged 36–50 who live and work in the same area have larger defined neighbourhoods than males in the same age group. This again may reflect the activity patterns of this cohort of respondents and may be indicative of the work they undertake in the neighbourhood.
4. Measuring Accordance: Consensus amongst Respondents in Close Spatial Proximity
Accordance relates to the final hypothesis—that respondents living in close spatial proximity to one another will share similar neighbourhood definitions compared with those living further away, although these may vary by the personal characteristics. Consensus amongst respondents living or working in close spatial proximity (i.e. in the same LSOA) was calculated using the CV for the area, perimeter and shape of neighbourhoods. The CV was calculated for all respondents in close spatial proximity differentiated by gender and their neighbourhood type. Due to small sample sizes and the general lack of statistical significance in previous findings, CVs of respondents differentiated by age group and ethnicity were not calculated. Summaries of the CVs by gender and neighbourhood type are reported in Table 4 and in Figure 3. The average CV for neighbourhood area was marginally larger and hence more varied for male compared with female respondents living in the same LSOA. This was not the case for the perimeters of the neighbourhoods which had similar average CVs and were smaller and hence showed less variation. The shape of neighbourhoods also showed very little variation by gender and the distribution of CVs confirms this. For the perimeter and shape of neighbourhoods this was around 90 per cent of LSOAs for both male and female respondents. For neighbourhood area, the CV was less than 0.5 in a third of LSOAs for female respondents compared with a quarter of LSOAs for male respondents. This was over 50 per cent of LSOAs for male and female respondents with respect to variation in perimeters of neighbourhoods and around 80 per cent of LSOAs for the variation in shapes of neighbourhoods. Hence males living in close proximity tend to have larger variation in the size of their residential neighbourhoods than female respondents.
Summary of coefficient of variation in respondents’ spatial metrics by gender and neighbourhood type
Coefficient of variation in respondents’ spatial metrics by gender and neighbourhood type.
In comparison, for those respondents who only work within an LSOA, the average CV of neighbourhood area shows substantially less variance for male respondents compared with female respondents (0.47 compared with 0.73 respectively). A similar situation occurs with the variation in perimeters of neighbourhoods (0.33 compared with 0.46 for males and females respectively). However, female respondents showed less variation in the shape of their neighbourhoods within an LSOA than male respondents (0.20 compared with 0.39 respectively). Male respondents had a CV less than 0.5 for their neighbourhood area in over three-fifths of LSOAs compared with 30 per cent of LSOAs for female respondents. This is three-quarters of LSOAs for male respondents compared with just fewer than 60 per cent of LSOAs for female respondents with respect to variation in neighbourhood perimeters. In terms of shape, all LSOAs had a CV of under 0.5 for female respondents, but this was half of LSOAS for male respondents. Therefore, females working in close proximity tend to have more varied neighbourhood sizes compared with males and again perhaps this reflects the activity patterns of female employment.
In terms of respondents who both live and work in the same LSOA, the average CV of the area and shape of neighbourhoods was larger for male than female respondents but this was reversed for the perimeter of neighbourhoods. For neighbourhood area, CV was under 0.5 for half of LSOAs for female respondents compared with only 20 per cent for male respondents. This was around 60 per cent of LSOAs for the perimeter of neighbourhoods for both male and female respondents and around 80 per cent and 50 per cent of LSOAs respectively for the shape of neighbourhoods. Therefore there is much more variation in the size of neighbourhoods for males and females who live and work in close spatial proximity than those who solely live or work close by.
5. Discussion and Conclusions
The paper reports some initial findings on research comparing self-reported neighbourhoods, for a large number of respondents in a British urban setting. The most sensitive spatial metric substantively and statistically was the size of the neighbourhood measured by area and to a lesser extent by perimeter. Shape had little effect, as most of the neighbourhoods were regular-shaped convex polygons and therefore had similar Cox indexes. However, there were some nuanced differences with respect to the age and ethnic group of respondent.
The analysis has shown that the relationship between respondents’ self-reported neighbourhoods depends not only on respondent characteristics, but also on whether a respondent lives, works or both lives and works in a neighbourhood, reflecting the amount of time a person spends in each neighbourhood and the activities that they perform. There are differences between male and female respondents, although in some instances these were quite minor and statistically insignificant. There are indications of differences between respondents from White and non-White ethnic groups, with respondents from non-White ethnic groups having smaller self-reported neighbourhoods, especially females, but the small sample size makes it difficult to say anything conclusive. Age of respondent has an interesting and, in some cases, a statistically significant effect on neighbourhood size and gender differences are evident. Young and retired people both have the largest neighbourhoods, possibly because both have more time to spend in their local area than other age groups. Workplace neighbourhoods tend to increase in size with age of respondent, but the differences are not substantial or significant. Gender differences were evident in the neighbourhoods of respondents who both live and work in an area, with females having larger neighbourhoods than males and this may reflect the types of employment undertaken by females who live and work in the same place.
Consensus between the spatial metrics of the neighbourhoods of respondents living or working within close spatial proximity revealed a lot of variation. This was particularly true with neighbourhood area and also between male respondents. There was more accordance in the spatial metrics of respondents’ workplace neighbourhoods, although accordance was slightly worse for those respondents who both live and work in a neighbourhood. These findings tend to support the conclusions drawn by Coulton et al. (2001), Roosa et al. (2009) and Campbell et al. (2009). Self-reported definitions of neighbourhood tend to be quite divergent. Residents’ social interaction patterns within an area may not necessarily map onto their neighbours as they may socialise differently, rely on different modes of transport and use different neighbourhood amenities. People who live and work in close proximity can differ markedly from one another in how they define the physical space of their neighbourhood and this also varies by personal characteristics.
For future research, the next stage will be to start to investigate how these self-reported neighbourhoods map onto other geographies which may act as proxies for neighbourhoods, such as primary school catchment areas. This will use Galster’s (2001) concept of congruency and will investigate both the degree to which respondents’ self-reported neighbourhoods overlap with these areas and also the degree to which they share common boundaries. This will move towards bridging the gap between research on self-defined neighbourhoods and on the use of other neighbourhoods in social science urban research.
Footnotes
Funding
This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.