Residential turnover and crime—Evidence from administrative data for

England and Wales

Br J Criminol (2023) 63(6): 1460–1481
1 November 2023

The British Journal of Criminology > 2023 - Volume 63 > Issue 6, 1 November > Articles

The British Journal of Criminology

Nils Braakmann

Newcastle University, Business School 5 Barrack Road, Newcastle upon Tyne NE1 4SE, UK;
[email protected]
© Centre for Crime & Justice Studies (formerly ISTD) 2023

This article combines administrative data on housing transactions with monthly crime data for the years 2014–18 to
investigate the link between residential turnover and crime. Using a panel of 34,753 small neighbourhoods in
England and Wales as well as street-level data for London, estimates from models with multiple fixed effects
suggest that higher turnover is associated with higher overall crime rates. These effects are small and driven by
increases in property and violent crime. Using information on price growth, I document differential effects for high-
turnover, high price growth ('attractive') and high turnover, low price growth ('declining') areas.
KEY WORDS: social disorganization, routine activity theory, residential turnover, gentrification, neighbourhoods,
crime
Crime

Introduction

How does residential turnover and the resulting disruption to local communities affect crime? In this article, I
empirically investigate this relationship in two different contexts—small neighbourhoods covering the entirety of
England and Wales and the streets of London.

A link between crime and residential turnover can be made based on several theoretical arguments. The first is
social disorganization theory (Shaw and McKay 1942), which links neighbourhood factors such as low economic
status, ethnic heterogeneity, residential mobility and family disruption to community social disorganization, e.g.,
lower social networks, which in turn leads to higher crime rates. Supporting evidence for this mechanism can be
found, for example, in Sampson and Groves (1989) for the United Kingdom and Sampson et al. (1997) using data
from Chicago. The second is routine activity theory (Cohen and Felson 1979), which explains crime as a
combination of likely offenders, suitable targets and the absence of capable guardians against crime coming
together in the same space (see, e.g., Sherman et al. 1989, for some evidence in support). Residential turnover in
this setting could influence both the availability of

Br J Criminol (2023) 63(6): 1460–1481 at 1461

 Residential turnover and crime—Evidence from administrative data for England and Wales

suitable targets, for example, due to higher income residents arriving in a gentrifying neighbourhood, as well as the
availability of capable guardians, for example, if residential turnover makes it more difficult to identify strangers in a
neighbourhood or if new arrivals to a community are initially less invested in its safety. Finally, a related argument
can be made based on economists' workhorse model when investigating crime, the Becker model (Becker 1968): A
prospective criminal weights the utility from a life in legality against the utility from committing a crime. The latter is a
combination of the utility of a successful crime (occurring with probability 1 – p, where p is the probability of getting
caught) and the disutility from punishment when caught. Most empirical work in economics that has focused on
changes to p has looked at police deployment or tactics (e.g., Di Tella and Schargrodsky 2004; Draca et al. 2011;
MacDonald et al. 2016; Cheng and Long, 2018; Mastrobuoni 2019; McMillen et al. 2019; Braakmann 2022).
However, it is easy to imagine that residential turnover might influence p, e.g., by making it harder to decide
whether strangers to the neighbourhood are simply new neighbours or whether people carrying crates out of a
house and loading these into a truck are burglars or a removal company.

The general idea that residential turnover and crime are linked has been tested in multiple contexts. The first are
typically longitudinal studies investigating the link between residential mobility and crime as well as the contextual
factors, such as the ethnic makeup of neighbourhoods (e.g., Boggess and Hipp 2010). A second and important
piece of quantitative evidence comes from studies looking at foreclosures. Focusing on these helps to overcome
concerns around the endogeneity of residential turnover and possible reverse causality with respect to crime rates,
but also arguably provides estimates based on a highly specific context—residential moves that coincide with
financial distress of existing residents. Foreclosures have been found to cause spill overs to nearby properties thus
influencing residential mobility beyond the foreclosed homes (e.g., Harding et al. 2008; Schuetz et al. 2008; Rogers
and Winter 2009). In terms of their impact on crime the evidence suggests a positive association between
foreclosures and crime rates with some papers arguing for a causal link (e.g., Ellen et al. 2013; Katz et al. 2013)
and some arguing for a spurious correlation caused by underlying community characteristics (Kirk and Hyra 2012).
Finally, several—often qualitative—papers investigate the effects of neighbourhood restructuring and gentrification
on crime as well as intermediate outcomes such as neighbourhood destabilization and disruption of social ties (e.g.,
Thompson et al. 2013; Urbanik et al. 2017).

This article uses UK administrative data on housing transactions and recorded crime for the years 2014–18
aggregated at two levels. The first is a monthly panel of 34,753 small neighbourhoods covering the entirety of
England and Wales. The second is a monthly panel of 196,604 streets/street segments covering London,
specifically the Metropolitan and City of London Police Force Areas. Econometrically I rely on flexible fixed effects
models to adjust for possible confounders, containing at the most comprehensive level neighbourhood fixed effects
and municipality * month-by-year fixed effects. Across all models and both datasets I find evidence for a link
between high housing turnover and overall crime rates, although effects are generally small. Effects are strongest in
high-turnover, high-price growth neighbourhoods, followed by high-turnover areas with 'normal' price growth and
high-turnover, low-price growth areas. Delving deeper into crime categories I find that these effects are driven by
changes to property crime and as well as changes to weapons offences, violent and sexual crime. Property crime
increases in turn are driven by increases across almost all categories—burglary, robbery, shoplifting, vehicle crime,
theft from the person and other theft—with the exception of bicycle theft. However, the pattern of these effects
differs across areas.

The article makes several contributions to the literature. First, this article is—to the best of my knowledge—the first
to use small neighbourhood data covering a whole country, including rural and urban areas, instead of, e.g., a
single city. Indeed, the results show that the link between

Br J Criminol (2023) 63(6): 1460–1481 at 1462

residential turnover and crime is driven by urban areas and particular urban neighbourhoods in conurbations.
Secondly, using the fact that my data also contains property transaction prices allows me to investigate an
important contextual factor, namely whether a neighbourhood is experiencing high residential turnover because it is
 Residential turnover and crime—Evidence from administrative data for England and Wales

becoming more attractive, for example, because of gentrification (in this context thought of as the inflow of higher
socio-economic status households into previously poorer areas, accompanied by increasing prices and changes to
local amenities), or because it is in decline. The former would generally be associated with a combination of high
turnover and high price growth, while the latter would be a combination of high turnover and low (or negative) price
growth. I find that this distinction indeed matters and that crime increases are often stronger in high-turnover, high-
price growth neighbourhoods. Some additional heterogeneity analysis additionally suggests that these effects are
strongest in relatively affluent and homogeneous neighbourhoods.

This paper is organized as follows. Section 2 discusses data and the empirical strategy, section 3 contains results
and section 4 concludes.

Data and empirical strategy

To create a panel of neighbourhoods for England and Wales, I combine data on recorded offences and housing
transaction from two sources at the level of small neighbourhoods, specifically lower-layer super output areas
(LSOAs). LSOAs are relatively small spatial units with a minimum population of 1,000 (with a mean of 1,500), equal
to approximately 650 households. For comparison, the US census tracts used for example by Boggess and Hipp
(2010) have an average population of about 5,000. LSOAs are designed for the publication of census data and are
stable over time. At present, there are 34,753 LSOAs in England and Wales. LSOAs are nested within local
authorities (LA), which form the basic level of local government in the United Kingdom and have responsibility for
public service provision within their boundary. They usually consist of a city or amalgamations of smaller towns and
rural areas. London as a special case is split into multiple boroughs each designated as a Local Authority. There
are a total of 348 local authorities in England and Wales. The average local authority contains 100 LSOAs with a
range from 1 (Isles of Scilly) to 639 (Birmingham). Figure 1 provides a map of all LSOAs in England and Wales.

The second dataset used in the analysis is a monthly panel of streets/street segments in London. To arrive at this
data, I start from a list of all unit postcodes in the Metropolitan and City of London Police force areas. UK unit
postcodes are relatively small spatial units, roughly corresponding to streets or parts of streets and covering on
average 15 delivery points (addresses). To arrive at the final sample, I eliminate all duplicate postcodes at the same
latitude and longitude combination—for example, where a large tower has multiple postcodes—which leaves me
with 196,604 unique points.

Administrative data on recorded offences comes from data.police.uk and originates from crime recorded by each of
the Police Forces in England and Wales.1 Offence data is available in a geocoded format by offence category and
month from December 2010. Offences are coded into 14 categories, namely Anti-social behaviour, Bicycle theft,
Burglary, Criminal damage and arson, Drugs, Other crime, Other theft, Possession of weapons, Public order,
Robbery, Shoplifting, Theft from the person, Vehicle crime, Violence and sexual offences. For most analyses, I
aggregate these to four categories: Anti-social behaviour, criminal damage and public order; property

Br J Criminol (2023) 63(6): 1460–1481 at 1463

Figure 1. Neighbourhoods in England and Wales.

crime; drugs and weapons offences and violent crime. I also look at total recorded crime counts. The raw data are
measured at the offence level. For each offence the data notes the offence category, the date (coarsened to month
and year as part of the anonymization process) and the latitude and longitude of the place where the offence was
committed. Locations are anonymized slightly to preserve anonymity of the victims; however, the respective
neighbourhood for each offence is precisely measured. Specifically, each crime is placed on the closest map point
from a master list. These anonymized map points are chosen to correspond to the centre point of a street, a public
place such as a park or airport or a commercial premise like a shopping centre or nightclub. Each map point
contains at least eight postal addresses or no postal addresses at

Br J Criminol (2023) 63(6): 1460–1481 at 1464

 Residential turnover and crime—Evidence from administrative data for England and Wales

all.2 The vast majority of map points across England and Wales are streets (count of 679,089), followed by sports
and recreation areas (24,510), parking areas (17,797), parks and open spaces (14,051) and supermarkets and
petrol stations (5,703 and 5,501), respectively. In the case of the London panel, I match each crime to the closest
location based on the geodedic distance (see Braakmann 2022, for a similar approach). Locations without a
matched crime in a given month get assigned a count of zero for that period. For the LSOA level data, I convert
crime counts to rates per 100 population using population counts from the 2011 census. Due to the small spatial
scale of the London panel population counts are not available and the outcomes are left as counts.

Residential turnover is measured using data from HM Land Registry, which is the central registry for all landowners
in England (and Wales). The specific data used here is the 'price paid data' recording final transaction prices and
other details, which is publicly available from 1995 onwards.3 The data contain information on the full address of
each property, the price paid, the date of the transaction, the property type (flat, terraced house, semi-detached
house or detached house), whether the property is newly built and whether the property is freehold or leasehold.
The transaction date in the data is the 'completion date', i.e., the date that keys are exchanged and usually the date
the physical move happens.

An advantage of this data is full coverage of all transactions in England and Wales at a very low spatial level, down
to the level of the specific address. A possible disadvantage is that the only covers residential turnover by
homeowners and not by tenants. Data on the latter is generally not available at the neighbourhood level. However,
there are good reasons to focus on homeowners when looking at the link between residential turnover and crime
(see Boggess and Hipp 2010, for a similar approach). There is evidence that homeowners have closer ties to their
local area (e.g., Bolan 1997; Oh 2004) and invest more in social capital (e.g., DiPasquale and Glaeser 1999) as
well as more into their properties and their general environment (see Dietz and Haurin 2003, for a review). This in
turn will make them more likely to serve as 'capable guardians' (Cohen and Felson 1979) against criminal activity
and to look out for crimes in their neighbourhood. Previous research has indeed suggested a link between
homeownership rates and crime (Ross 1977; Roncek 1981; Krivo and Peterson 1996; Smith et al. 2000; Hipp
2007).

I use the property transaction data to generate multiple indicators of residential turnover. The first is simply the
number of sales in each neighbourhood/street and month—on average there are 2.4 sales in each neighbourhood
but with a substantial range from 0 to 253. For the street level data, the average is 0.031 sales per month with a
range from 0 to 167. Additionally, for the neighbourhood data, I generate two dummy variables indicating whether a
neighbourhood is within the top 10% or top 25% in terms of housing turnover in each local authority. For the street
level data, I instead generate a dummy variable indicating whether there were any property sales at the location in a
given month.

Finally, the availability of property prices allows me to investigate an important contextual factor, namely whether a
neighbourhood is experiencing high residential turnover because it is becoming more attractive, for example,
because of gentrification, or because it is in decline. Specifically, I define a high turnover, high price growth
neighbourhood as being in the top 10% or top 25% in the respective local authority in terms of (a) sales, (b) prices
in the current month and (c) price growth over the preceding 12 months. Conversely, a high turnover, low price
growth neighbourhood is defined as one in the top 10% or top 25% in terms of sales, but in the bottom 10%/25% for
prices and price growth. This analysis is impossible to replicate for the street level data as only about 1% of street-
month combinations have property transactions (and hence price information).

Br J Criminol (2023) 63(6): 1460–1481 at 1465

Table 1 presents descriptive statistics for the estimation sample. In total there are 2,085,180 observations covering
34,753 neighbourhoods over 60 months for the neighbourhood sample and 11,796,240 observations covering
196,604 locations over 60 months for the London sample.
 Residential turnover and crime—Evidence from administrative data for England and Wales

Econometrically, I estimate variations of the following estimating equation for the neighbourhood sample

(1)
where i indexes neighbourhoods, nested within local authorities l, and t time. yi(l)t is the outcome, either the general
crime rate per 100 population or a crime rate for a specific offence or offence group and D(l)it marks the variable(s)
of interest. At the street level I estimate

(2)
where s indexes streets, nested within neighbourhoods i and local authorities l, and t time. ys(il)t is the crime count
in a given street and month, while Ds(il)t are measures of residential

Table 1. Descriptive statistics

Variable Mean Std. dev. Min. Max.

Panel (a):
Neighbourhood data
for England and
Wales

# Sales 2.387 2.723 0 253

Area definition based

on 90th percentile

High turnover area 0.102 0.302 0 1

High price growth, 0.003 0.055 0 1

high turnover area

Low price growth, 0.001 0.023 0 1

high turnover area

Area definition based

on 75th percentile

High turnover area 0.263 0.440 0 1

High price growth, 0.034 0.182 0 1

high turnover area
Low price growth,
high turnover area

Low price growth, 0.009 0.093 0 1

high turnover area

Crime rates (per 100 90.468

population) Total
crime

Total crime 0.867 1.265 0 90.468

Anti-social behaviour, 0.374 0.521 0 24.040

criminal damage,
arson, public order

Property crime 0.278 0.561 0 69.277

 Residential turnover and crime—Evidence from administrative data for England and Wales

Drug offences 0.020 0.071 0 12.658

Weapons, violent and 0.184 0.280 0 26.229

sexual offences

Observations (N*T) 2,085,180

Neighbourhoods (N) 34,753

Months (T) 60

Panel (b): Street-level

data for London

# Sales 0.031 0.376 0 167

Any sales 0.025 0.157 0 1

Crime counts

Total crime 0.373 1.411 0 218

Anti-social behaviour, 0.142 0.691 0 55

criminal damage,
arson, public order

Property crime 0.125 0.672 0 153

Drug offences 0.014 0.172 0 122

Weapons, violent and 0.086 0.448 0 46

sexual offences

Observations (N*T) 11,796,240

Streets/street 196,604
segments (N)

Months (T) 60

Br J Criminol (2023) 63(6): 1460–1481 at 1466

turnover. The most comprehensive version contains neighbourhood fixed effects αi, that control for characteristics
influencing crime and residential turnover, such as the general attractiveness of an area, social deprivation or a
neighbourhood's dependence on the night time economy—all of which will likely be relatively constant over a four
year period. It also contains local authority-month-by-year fixed effects θlt that non-parametrically control for any
factor varying at the local authority-time level, such as local labour market conditions. Less comprehensive
specifications replace the neighbourhood fixed effects with local authority or local authority-by-time fixed effects or
constrain time effects to be identical across local authorities. Standard errors are adjusted for clustering at the level
at which the variables of interest vary (neighbourhood level for the neighbourhood panel, street level for the street
panel), which is usually the appropriate level due to the reasons explained in Moulton (1990, also see Angrist and
Pischke 2009, section 8.2). Furthermore, clustering on what is essentially the cross-sectional unit in a panel is the
appropriate way to adjust for heteroskedasticity and autocorrelation (Stock and Watson 2009).

An important concern is the possible misreporting of crime. The analysis relies on police-recorded crime, which is a
 Residential turnover and crime—Evidence from administrative data for England and Wales

subset of actual crime. An important remediation of these concerns are the included fixed effects: The
neighbourhood fixed effects αi control for any time-constant underreporting of crime rates—the variation used to
identify τ in equation (1) comes from with-in-neighbourhood changes in reported crime, i.e., any bias would need to
come from changes to reporting practices over time. Additionally, the local authority-month-by-year fixed effects θlt
control for any time-varying changes to reporting practices at the local authority level. These would include changes
to policing practices that affect all neighbourhoods or any general trends in attitudes towards the police that are
common to all neighbourhoods in a local authority. Against this background, underreporting could be problematic to
the extent that it is time-varying, specific to neighbourhoods and correlated with residential turnover. An example
would be if newcomers to a neighbourhood have initially lower trust in the police and are thus less likely to report
crimes. If this was the case, estimates for the effect of residential turnover would be downward biased.

While the fixed effects adjust for a range of unobserved confounders, possible remaining threats to identification are
unobserved random shocks specific to a neighbourhood that are correlated with both crime rates and residential
turnover or direct reverse causality where an increase in crime triggers an outflow of residents. There are, however,
good reasons to believe that these are unlikely to be major problems. As mentioned earlier, the data is measured
on the month level and sales are measured in the month that the actual move occurs. Given time frames for
property sales in the United Kingdom, the decision to sell a property would have occurred several months before:
The seller would, after making the decision to sell, need to find an estate agent who would advertise the property
and arrange viewings with prospective buyers, while buyers—after deciding to the buy the property—would need to
obtain a mortgage, which usually requires the property to be inspected by a surveyor, and then arrange for the legal
requirements and the actual move. To give some perspective on timelines, consumers' advice company Which?
(2020) suggests that it takes on average 22 weeks between the start of house hunting and the exchange of
contracts (the point at which the purchase become legally binding) and then usually another two weeks between the
exchange of contracts and completion. This in turn means that to cause bias any random shock would need to
affect crime in t, but the decision to sell properties in t-k, which—based on the above rough timelines would usually
be about 6 months earlier. The same holds for direct reverse causality. Overall, even though it is possible that sales
triggered by spikes in crime move on a faster timeline than the one sketched above, it seems likely that there would
still be several months between the decision to sell/buy and the actual transaction date.

While the United Kingdom has a comparatively high proportion of owner-occupiers—64% of all households in the
2011 census owned their accommodation—and while—as outlined

Br J Criminol (2023) 63(6): 1460–1481 at 1467

earlier—there are good reasons to focus on these when looking at the relationship between residential turnover and
crime, it is useful to investigate how this relationship might depend on the renting population. Unfortunately, there is
no data analogous to the land registry that would allow me to measure turnover of renting households. However,
the 2011 census contains information on the proportion of the usual residential population in owned or rented
accommodation at the neighbourhood level. Figure 2 plots the relationship between the proportion of households
living in owned accommodation and the number of property sales in panel (a) as well as the probability that an area
is classed as high turnover in panel (b).

In both cases, the data suggests that areas with a higher proportion of the population in owned accommodation are
more likely to see high turnover. In a sense this result is hardly surprising— any move by an owner-occupier will be
accompanied by a sale, while the same is not true for mobility of renters. It is, however, useful to think through the
consequences for our main estimates, given that turnover of renters is unobserved. Given the inclusion of
neighbourhood fixed effects in equation (1)—which would capture any time-constant correlation between residential
turnover of renters and owners at the neighbourhood level—our main concern is correlated shocks on both
variables, i.e., are periods when residential turnover of owners is high also periods when renters' turnover is high
and what would this imply for the turnover-crime relationship?

Standard omitted variable bias arguments suggests that this depends on two parameters: The relationship between
 Residential turnover and crime—Evidence from administrative data for England and Wales

renter turnover and crime and the correlation between renter and owner turnover. It seems reasonable to assume
that the former will be positive—all the theoretical arguments linking residential turnover and crime outlined in the
introduction would equally apply to turnover of renters. Additionally, evidence by Farrall et al. (2016) suggests that
while (social) renters and owner-occupiers experience slightly different levels of crime, their general trends over
time were not too dissimilar during the 1980s and 1990s. The latter is more difficult to pin down: One the one hand,
it could be positive if unobserved shocks to neighbourhoods triggered outflows of both renters and owner occupiers.
On the other hand, the more owner-occupiers there are in a neighbourhood, the fewer renters exist: 2011 census
data suggests a correlation of −0.64 between the number of households owning and those renting at the
neighbourhood level. Given the strength of the latter correlation it seems reasonable to assume that the overall
correlation would also be negative.

In terms of overall bias, these arguments suggest that τ in equation (1) is likely downward biased. Intuitively, this
bias occurs because areas with high residential turnover of owner-occupiers (and consequently high sales) would
have a low (unobserved) turnover of renters and vice versa. This in turn diminishes differences in actual turnover
between 'treated' and 'untreated' neighbourhoods.

Results

Residential turnover and crime

Figure 3 presents a first look at the raw data. Panel (a) is a simple binned scatter plot of the number of crimes per
100 population against the number of property sales in a neighbourhood, while panel (b) presents the same plot,
but with neighbourhood fixed effects netted out. Both panels suggest a clear positive link between housing turnover
and crime that—in the model with fixed effects—also looks clearly linear.

Table 2 looks at the same relationship using a variety of specifications. Column (1) is a simple pooled OLS
regression that controls for time fixed effects. Column (2) adds local authority fixed effects, while column (4)
replaces these with neighbourhood fixed effects. Column (3) combines time and local authority fixed effects into
local authority-time fixed effects, while column (5) adds neighbourhood fixed effects to the specification in column
(3). Panel (a) uses the neighbourhood data and panel (b) covers estimates using the street level data for London.

Br J Criminol (2023) 63(6): 1460–1481 at 1468

Figure 2. Property transactions and proportion of resident households in owned accommodation in 2011. Panel (a):
Number of sales. Panel (b): Probability to be a high-turnover neighbourhood (90th percentile). Notes: The
underlying number of observations is 2,085,180 (34,753 neighbourhoods over 60 months).

Br J Criminol (2023) 63(6): 1460–1481 at 1469

Figure 3. Total crime and property transactions, binned scatter plot. Panel (a): Raw data. Panel (b): Residualized on
LSOA fixed effects. Notes: Panel (a) plots the mean of the crime rate against the number of sales per month and
neighbourhood (8 and more sales are grouped together, representing 3% of observations). Panel (b) does the
same after residualizing both crime rate and sales on neighbourhood fixed effects. Bins in panel (b) are 20 quantiles
of the residualized number of sales. The underlying number of observations is 2,085,180 (34,753 neighbourhoods
over 60 months).

Br J Criminol (2023) 63(6): 1460–1481 at 1470

Table 2. Residential turnover and overall crime

 Residential turnover and crime—Evidence from administrative data for England and Wales

(1) (2) (3) (4) (5)

Panel (a): Neighbourhoods in England and Wales

Linear specification

# sales 0.029*** 0.036*** 0.037*** 0.001*** 0.001***

(0.003) (0.003) (0.003) (0.000) (0.000)

Dummy specification - # sales >= 75th percentile

High turnover area 0.133*** 0.159*** 0.161*** 0.003*** 0.003***

(0.012) (0.012) (0.012) (0.001) (0.001)

Dummy specification - # sales >= 90th percentile

High turnover area 0.232*** 0.268*** 0.272*** 0.006*** 0.006***

(0.018) (0.018) (0.018) (0.001) (0.001)

Observations (N*T) 2,085,180 2,085,180 2,085,180 2,085,180 2,085,180

Neighbourhoods (N) 34,753 34,753 34,753 34,753 34,753

Months (T) 60 60 60 60 60

Panel (b): London Streets

Linear specification

# Sales −0.002 −0.001 −0.001 −0.001 −0.001

(0.002) (0.002) (0.002) (0.002) (0.002)

Dummy specification—any sales

Any sales 0.008 0.015*** 0.015*** 0.022*** 0.022***

(0.005) (0.005) (0.005) (0.004) (0.004)

Observations (N*T) 11,796,240 11,796,240 11,796,240 11,796,240 11,796,240

Streets/street segments (N) 196,604 196,604 196,604 196,604 196,604

Months (T) 60 60 60 60 60

Month-by-year Fes Yes Yes Yes

Local authority FE Yes

Local authority * month-by-year Yes Yes

Fes
 Residential turnover and crime—Evidence from administrative data for England and Wales

Neighbourhood Fes Yes Yes

Notes: Coefficients, standard errors in parentheses. Standard errors adjusted for clustering on the neighbourhood
level (neighbourhood panel) or the street level (London panel).

*/**/*** denote statistical significance at the 10%, 5% and 1% level. Outcomes are rates per100 population for the
neighbourhood panel and counts for the London panel.

Focusing first on panel (a), estimates across all columns suggest a positive link between residential turnover and
crime. The magnitude, however, differs significantly between specifications with (columns (4) and (5)) and without
neighbourhood fixed effects (columns (1) to (3)). Estimates with neighbourhood fixed effects suggest an increase in
the crime rate by 0.001 for each additional sale, while an area being a high turnover area increases crime by 0.003
(75th percentile) or 0.006 (90th percentile). Estimates without neighbourhood fixed effects—essentially comparing
high and low turnover neighbourhoods in the same local authority and month—suggest larger increases by 0.037
for each sale and 0.161 (75th percentile) or 0.272 (90th percentile) for being in a high turnover neighbourhood. In
panel (b), there does not appear to be any significant link between the number of sales and crime, however, only
about 2.5% of streets experience a sale in any given month. This mass point at zero is unlikely to be captured well
by a linear specification. Indeed, the specification using a dummy for having any sale in a street/month combination
again gives very similar results to the neighbourhood panel—there is again a positive link between residential
turnover and crime rates. Taken together these results suggest that (a) across neighbourhoods those with higher
residential turnover will have higher crime rates, (b) within neighbourhoods, crime

Br J Criminol (2023) 63(6): 1460–1481 at 1471

increases in times of high residential turnover and (c) within London neighbourhoods, streets with any residential
turnover will experience higher crime than those without.

The large difference in magnitude between the estimates with and without fixed effects can be explained by multiple
mechanisms. First, the fixed effects absorb any possible unobserved confounder at the neighbourhood level—of
which there are many, as there are few other controls included. These would include, for example, whether a
neighbourhood is affluent or socially deprived, whether there is generally a high trust in the police during the
observation period or indeed other aspects of an area's social capital. Secondly, the neighbourhood fixed effects
also control for possible time-constant differences in crime reporting, for example, the possibility that areas with
higher turnover also have characteristics that affect the probability that a given crime is recorded to the police.

How large are these effects? Focusing on the most comprehensive specifications, each sale increases the overall
crime rate by 0.01 in the neighbourhood panel, while having any sale in a street increases crime counts by 0.022 in
the London data. Compared to means of 0.87 (neighbourhood panel) and 0.37 (London panel), these effects range
from small—about 0.1% of the mean in the neighbourhood data—to a fairly substantial 6% in the London data.
However, it is important to remember that while residential turnover might contribute to crime, it will unlikely be
among the most important causes of criminal activity. Hence the magnitude of these effects is ultimately not that
surprising. That being said, a simple back of the envelope calculation suggests that residential turnover appears to
be responsible for an additional 108 crimes across London in each month—there are 196,604 locations across
London, of these 2.5% experience some turnover, which in turn increases crime counts by 0.022 per location and
month.

Attractive and declining neighbourhoods

 Residential turnover and crime—Evidence from administrative data for England and Wales

An important contextual aspect is that some neighbourhoods will experience residential turnover because they are
becoming more attractive, for example, because of gentrification, while others will experience turnover because
they are in decline. While some of the underpinning mechanism linking residential turnover to crime—such as
weaking community ties or difficulties to identify strangers in an area—would apply to both situations, others will
likely differ—for example, gentrification usually brings wealthier potential targets to an area. Using the
neighbourhood panel—which has price information for each neighbourhood and area—I refine the definition of a
high-turnover neighbourhood by separating it into high turnover, high price growth ('attractive') areas (top 10%/25%
of residential turnover, prices in current month and 12-month price growth in the top 10%/25% of the respective
local authority), high turnover, low price growth ('declining') areas (top 10%/25% of residential turnover, prices in
current month and 12-month price growth in the bottom 10%/25% of the respective local authority) and all
remaining high turnover areas.

Table 3 presents results. As with the earlier estimates, the inclusion of neighbourhood fixed effects in columns (4)
and (5) makes a major difference to both the magnitude and the significance of the results. Focusing on the results
including these fixed effects and a treatment area definition based on the 90th percentile, we see that crime
increases appear to be most pronounced for 'attractive' high turnover areas, where crime increases by 0.041,
followed by 'average' high turnover areas with a much smaller increase of 0.005 and finally declining areas where
the effect is statistically insignificant. An important aspect of these results is that they appear unlikely to be caused
by reverse causality—while it is easy to imagine higher crime rates causing an outflow of people, we would also
expect them to cause a decline in house prices (see, e.g., Gibbons 2004, or Braakmann 2017, for UK evidence). A
pattern where increases in crime go hand in hand with residential turnover and increasing prices is much more
difficult to explain through either unobserved shocks or direct reverse causality. It is, however, consistent with
wealthier targets entering a neighbourhood and causing an increase in property crimes.

Br J Criminol (2023) 63(6): 1460–1481 at 1472

Table 3. Attractive and unattractive neighbourhoods, outcome = all crime per 100 population

(1) (2) (3) (4) (5)

Panel (a): Dummy specification - # sales >= 75th percentile with 12-month price growth

High turnover area 0.093*** 0.126*** 0.127*** 0.002*** 0.002***

(0.009) (0.009) (0.009) (0.001) (0.001)

High price growth, high turnover area 0.281*** 0.215*** 0.224*** 0.003* 0.004*

(0.024) (0.020) (0.021) (0.002) (0.002)

Low price growth, high turnover area 0.089*** 0.122*** 0.122*** 0.002 0.002

(0.017) (0.017) (0.017) (0.004) (0.004)

Panel (b): Dummy specification - # sales >= 90th percentile with 12-month price growth

High turnover area 0.200*** 0.238*** 0.241*** 0.005*** 0.005***

(0.015) (0.015) (0.016) (0.001) (0.001)

High price growth, high turnover area 0.932*** 0.863*** 0.876*** 0.042*** 0.041***

(0.098) (0.093) (0.094) (0.011) (0.010)

 Residential turnover and crime—Evidence from administrative data for England and Wales

Low price growth, high turnover area 0.783*** 0.772*** 0.783*** 0.034 0.035

(0.121) (0.118) (0.121) (0.023) (0.023)

Month-by-year Fes Yes Yes Yes

Local authority FE Yes

Local authority * month-by-year Fes Yes Yes

Neighbourhood Fes Yes Yes

Observations (N*T) 2,085,180 2,085,180 2,085,180 2,085,180 2,085,180

Neighbourhoods (N) 34,753 34,753 34,753 34,753 34,753

Months (T) 60 60 60 60 60

Notes: Coefficients, standard errors in parentheses. Standard errors adjusted for clustering on the neighbourhood
level.

*/**/*** denote statistical significance at the 10%, 5% and 1% level. Outcomes are rates per 100 population.

Effects by type of crime

Following on from this argument, I now turn to an investigation of—first—broad crime categories, namely (a) anti-
social behaviour4 , criminal damage, arson, and public order offences5 , (b) property crime, (c) drug offences and (d)
weapon, violent and sexual offences and—later— more detailed categories of property crime, namely burglary,
robbery, shoplifting, vehicle crime, bicycle theft and theft from the person.

Table 4 contains results for four different specifications for the broad categories. Column (1) and (2) use the
neighbourhood data—column (1) with just an indicator for a high-turnover areas, column (2) additionally
distinguishing between attractive, average and declining areas. Columns (3) and (4) use the London street data,
either with the number of property sales in column (3) or an indicator for any sales in a street and month in column
(4).

Focusing first on columns (1) and (2) it appears that increases in anti-social behaviour and related offences as well
as property crime are driven by 'attractive' high-turnover areas, where effect sizes are much larger than for average
high-turnover areas. There appears to be no increase in drug

Br J Criminol (2023) 63(6): 1460–1481 at 1473

Table 4. Residential turnover and crime by category

(1) (2) (3) (4)

Neighbourhoods in London Streets

England and Wales
 Residential turnover and crime—Evidence from administrative data for England and Wales

Panel (a): Outcome = ASB, criminal damage, arson, public order

High turnover area 0.000 0.000

(0.001) (0.001)

High price growth, high turnover area 0.009**

(0.004)

Low price growth, high turnover area 0.015

(0.011)

# Sales 0.001

(0.001)

Any sales 0.013***

(0.002)

Panel (b): Outcome = Property crime

High turnover area 0.004*** 0.003***

(0.001) (0.001)

High price growth, high turnover area 0.020***

(0.007)

Low price growth, high turnover area 0.006

(0.010)

# Sales −0.001

(0.001)

Any sales 0.005***

(0.002)

Panel (c): Outcome = Drug offences High turnover area 0.000 0.000

(0.000) (0.000)

High price growth, high turnover area 0.001

(0.001)

Low price growth, high turnover area −0.003

(0.002)
 Residential turnover and crime—Evidence from administrative data for England and Wales

# Sales −0.000***

(0.000)

Any sales −0.000

(0.000)

Panel (d): Outcome = Weapons, violent and sexual offences

High turnover area 0.002*** 0.002***

(0.000) (0.000)

High price growth, high turnover area 0.011***

(0.004)

Low price growth, high turnover area 0.017*

(0.009)

# Sales −0.000

(0.000)

Any sales 0.005***

(0.001)

Local authority * month-by-year FEs Yes Yes Yes Yes

Br J Criminol (2023) 63(6): 1460–1481 at 1474

Table 4. Continued

(1) (2) (3) (4)

Neighbourhoods in England and Wales London Streets

Neighbourhood FEs Yes Yes Yes Yes

Observations (N*T) 2,085,180 2,085,180 11,796,240 11,796,240

Neighbourhoods/streets (N) 34,753 34,753 196,604 196,604

Months (T) 60 60 60 60

Notes: Coefficients, standard errors in parentheses. Standard errors adjusted for clustering on the neighbourhood
level (neighbourhood panel) or the street level (London panel).
 Residential turnover and crime—Evidence from administrative data for England and Wales

*/**/*** denote statistical significance at the 10%, 5% and 1% level. Outcomes are rates per 100 population for the
neighbourhood panel and counts for the London panel.

offences. For weapon, violent, and sexual offences the picture is slightly different—both 'attractive' and 'declining'
high-turnover areas experience larger increases than average high-turnover areas, which in turn are higher than the
base category with turnover below the 90th percentile. Comparing these to the London data, we again see that the
specification using an indicator for any sales gives results that are in line with the neighbourhood data—streets with
any property sales experience increases in anti-social behaviour and related offences, property crime as well as
weapon, violent and sexual offences relative to other streets in the same neighbourhood.

Table 5 looks more deeply into property crime. Starting again with the neighbourhood data, we can see increases in
burglary, shoplifting, vehicle crime and theft from the person across all high-turnover areas. The increases in
burglary and vehicle crime can also be seen in the London street data. There is, however, considerable
heterogeneity once we split high-turnover neighbourhoods into attractive, declining and average areas: Only
burglaries increase across all high-turnover areas, while increases in robberies and theft from the person are
concentrated in attractive high-turnover areas and increases in vehicle crime are found outside of average high-
turnover areas.

Heterogeneity

In this section, I explore some basic heterogeneity in effects across neighbourhoods with different compositions. As
the contextual data is either not available at the street level or—in the case of urban and rural splits—does not
make sense for London, I rely purely on the neighbourhood panel. Table 6 splits the sample by urbanization status
into urban in conurbation, urban city and town and rural. The results essentially suggest that the link between
residential turnover and crime is largely an urban phenomenon, in particular one that occurs in larger urban areas
and less in towns.

Table 7 investigates effects by the social composition of neighbourhoods based on data from the 2011 Census,
specifically the UK NRS social grades commonly used in the UK. The classification is based on occupations. I look
at the proportion of the population falling into the two highest social classes in the UK (professional and managerial,
A and B) as well as the bottom two classes (D semiskilled and unskilled manual workers and E casual and lowest
grade workers, benefit recipients) and compare neighbourhoods in the top quartile of each measure with those in
the bottom quartile. I also create a measure of neighbourhood homogeneity using a commonly employed measure
of concentration—the neighbourhood's Hirschman-Herfindahl/Simpson diversity index across the six social classes6
and compare the 25% most socially homogeneous neighbourhoods with the 25%

Br J Criminol (2023) 63(6): 1460–1481 at 1475

Table 5. Unpacking property crime effects, crime per 100 population

(1) (2) (3) (4)

Neighbourhoods in England and London streets

Wales

Panel (a): Outcome = Burglary

High turnover area 0.013*** 0.011***

 Residential turnover and crime—Evidence from administrative data for England and Wales

(0.003) (0.003)

High price growth, high turnover area 0.048**

(0.021)

Low price growth, high turnover area 0.119**

(0.057)

# Sales 0.001***

(0.000)

Any sales 0.005***

(0.001)

Panel (b): Outcome = Robbery

High turnover area 0.002 0.001

(0.001) (0.001)

High price growth, high turnover area 0.027**

(0.013)

Low price growth, high turnover area 0.005

(0.023)

# Sales −0.000

(0.000)

Any sales −0.000

(0.000)

Panel (c): Outcome = Shoplifting

High turnover area 0.019*** 0.017***

(0.005) (0.005)

High price growth, high turnover area 0.066

(0.042)

Low price growth, high turnover area 0.053

(0.097)

# Sales −0.001***
 Residential turnover and crime—Evidence from administrative data for England and Wales

(0.000)

Any sales −0.005***

(0.001)

Panel (d): Outcome = Vehicle crime

High turnover area 0.007* 0.004

(0.004) (0.004)

High price growth, high turnover area 0.060***

(0.023)

Low price growth, high turnover area 0.122**

(0.054)

# Sales 0.001***

(0.000)

Any sales 0.007***

(0.001)

Br J Criminol (2023) 63(6): 1460–1481 at 1476

Table 5. Continued

(1) (2) (3) (4)

Neighbourhoods in England and London streets

Wales

Panel (e): Outcome = Bicycle theft

High turnover area 0.003 0.003

(0.002) (0.002)

High price growth, high turnover area −0.002

(0.017)

Low price growth, high turnover area −0.030

(0.033)

# Sales −0.000**
 Residential turnover and crime—Evidence from administrative data for England and Wales

(0.000)

Any sales −0.000

(0.000)

Panel (f): Outcome = Theftfrom the person

High turnover area 0.008** 0.006*

(0.004) (0.004)

High price growth, high turnover area 0.069**

(0.031)

Low price growth, high turnover area −0.066*

(0.035)

# Sales −0.001***

(0.000)

Any sales −0.002***

(0.000)

Local authority * month-by-year FEs Yes Yes Yes Yes

Neighbourhood FEs Yes Yes Yes Yes

Observations (N*T) 2,085,180 2,085,180 11,796,240 11,796,240

Neighbourhoods/streets (N) 34,753 34,753 196,604 196,604

Months (T) 60 60 60 60

Notes: Coefficients, standard errors in parentheses. Standard errors adjusted for clustering on the neighbourhood
level (neighbourhood panel) or the street level (London panel).

*/**/*** denote statistical significance at the 10%, 5% and 1% level. Outcomes are rates per 100 population for the
neighbourhood panel and counts for the London panel.

most heterogeneous. The results suggest that the observed relationship between turnover and crime is largely
driven by areas with a high proportion of social classes A and B, a low proportion of social classes D and E and by
areas that are comparatively homogeneous in terms of the class structure.

Conclusion

This paper investigated the link between residential turnover and crime in two contexts—a neighbourhood panel
covering the whole of England and Wales and a street-level panel for London. In both cases, the results suggest a
positive link between residential turnover and crime. Specifically, (a) across neighbourhoods those with higher
 Residential turnover and crime—Evidence from administrative data for England and Wales

residential turnover will have higher crime rates, (b) within neighbourhoods, crime increases in times of high
residential turnover and (c) within London neighbourhoods, streets with any residential turnover will experience

Br J Criminol (2023) 63(6): 1460–1481 at 1477

Table 6. Residential turnover and crime, urban vs. rural neighbourhoods

(1) (2) (3) (4)

All areas Urban in conurbation Urban city and town Rural

90th percentile

Panel (a): Outcome = ASB, criminal damage, arson, public order

High turnover area 0.000 −0.001 0.000 0.001

(0.001) (0.001) (0.001) (0.001)

High price growth, 0.009** 0.008 0.013* 0.001

high turnover area (0.004) (0.006) (0.007) (0.006)

Low price growth, 0.015 0.023 0.019 −0.017*

high turnover area (0.011) (0.020) (0.019) (0.009)

Panel (b): Outcome = Property crime

High turnover area 0.003*** 0.006*** 0.002*** 0.001*

(0.001) (0.002) (0.001) (0.001)

High price growth, 0.020*** 0.041*** 0.006 0.005

high turnover area (0.007) (0.014) (0.007) (0.005)

Low price growth, 0.006 −0.004 0.017 −0.006

high turnover area (0.010) (0.019) (0.015) (0.009)

Panel (c): Outcome = Drug offences

High turnover area 0.000 −0.000 0.000 −0.000

(0.000) (0.000) (0.000) (0.000)

High price growth, 0.001 0.003 −0.002 0.001

high turnover area (0.001) (0.003) (0.002) (0.001)

Low price growth, -0.003 −0.005 −0.005 0.001

high turnover area (0.002) (0.004) (0.004) (0.003)

 Residential turnover and crime—Evidence from administrative data for England and Wales

Panel (d): Outcome = Weapons, violent and sexual offences

High turnover area 0.002*** 0.002** 0.002*** 0.001

(0.000) (0.001) (0.001) (0.001)

High price growth, 0.011*** 0.016** 0.010** 0.003

high turnover area (0.004) (0.006) (0.005) (0.006)

Low price growth, 0.017* 0.030* 0.010 0.001

high turnover area (0.009) (0.018) (0.013) (0.008)

Local authority * Yes Yes Yes Yes

month-by-year FEs

Neighbourhood FEs Yes Yes Yes Yes

Observations (N*T) 2,085,120 763,740 948,660 371,160

Neighbourhoods 34,753 12,731 15,818 6,204

(N)

Months (T) 60 60 60 60

Notes: Coefficients, standard errors in parentheses. Standard errors adjusted for clustering on the neighbourhood
level.

*/**/*** denote statistical significance at the 10%, 5% and 1% level. Outcomes are rates per 100 population.

higher crime than those without. Effects are also stronger for neighbourhoods where residential turnover goes hand
in hand with rising property prices and for more affluent and more socially neighbourhoods in urban settings.
Finally, looking at detailed crime categories a consistent finding is an increase in property crimes, specifically
burglaries and vehicle crime, that span all neigh-bourhoods with high turnover. Other crime categories show a
somewhat more mixed pattern, although violent crime also shows fairly consistent increases in high turnover
neighbourhoods.

Br J Criminol (2023) 63(6): 1460–1481 at 1478

Table 7. Residential turnover and crime by social structure of neighbourhoods

(1) (2) (3) (4) (5) (6) (7)

All areas Share professional Share semi-/unskilled, Herfindahl index of social classes
and managerial lowest grade workers,
(social classes A and benefit recipients
B) (social classes D and
E)
 Residential turnover and crime—Evidence from administrative data for England and Wales

Top Bottom Top Bottom 25% most 25% least

quartile quartile quartile quartile homogenous homogenous

Panel (a): Outcome = ASB, criminal damage, arson,

public order

High turnover 0.000 0.001 0.001 0.000 0.001 0.000 0.001

area

(0.001) (0.001) (0.002) (0.002) (0.001) (0.002) (0.001)

High price 0.009** 0.002 0.013 0.009 −0.000 0.010 0.007

growth,

high turnover (0.004) (0.007) (0.014) (0.014) (0.007) (0.009) (0.007)

area

Low price 0.015 0.010 0.007 0.006 0.019 0.053 −0.003

growth,

high turnover (0.011) (0.016) (0.032) (0.031) (0.022) (0.034) (0.019)

area

Panel (b): Outcome =

Property crime

High turnover 0.003** 0.003* 0.007*** 0.009*** 0.004** 0.009*** 0.002**

area *

(0.001) (0.002) (0.002) (0.002) (0.002) (0.003) (0.001)

High price 0.020** 0.037** 0.003 0.002 0.044** 0.049** 0.013**

growth, *

high turnover (0.007) (0.017) (0.009) (0.010) (0.018) (0.021) (0.006)

area

Low price 0.006 −0.016 0.027 0.017 −0.017 0.014 0.018

growth,

high turnover (0.010) (0.017) (0.023) (0.025) (0.021) (0.029) (0.016)

area

Panel (c): Outcome = Drug

offences

High turnover 0.000 −0.000 −0.000 −0.000 0.000 0.000 0.000

area

(0.000) (0.000) (0.000) (0.000) (0.000) (0.001) (0.000)

Br J Criminol (2023) 63(6): 1460–1481 at 1479

Table 7. Continued

(1) (2) (3) (4) (5) (6) (7)

 Residential turnover and crime—Evidence from administrative data for England and Wales

All areas Share professional Share semi-/unskilled, Herfindahl index of social classes
and managerial lowest grade workers,
(social classes A and benefit recipients
B) (social classes D and
E)

Top Bottom Top Bottom 25% most 25% least

quartile quartile quartile quartile homogenous homogenous

High price 0.001 0.004 −0.001 −0.001 0.005* 0.007* −0.001

growth,

high turnover (0.001) (0.003) (0.002) (0.002) (0.003) (0.004) (0.002)

area

Low price −0.003 −0.005 −0.005 −0.010* −0.005 −0.010 0.000

growth,

high turnover (0.002) (0.003) (0.005) (0.005) (0.004) (0.006) (0.002)

area

Panel (d): Outcome = Weapons, violent

and sexual offences

High 0.002*** 0.001 0.003** 0.005*** 0.001* 0.004*** 0.001

turnover area

(0.000) (0.001) (0.002) (0.002) (0.001) (0.001) (0.001)

High price 0.011*** 0.018** 0.004 0.009 0.021*** 0.025** 0.002

growth,

high turnover (0.004) (0.007) (0.009) (0.009) (0.008) (0.010) (0.005)

area

Low price 0.017* 0.004 0.029 0.020 0.013 0.014 0.016

growth,

high turnover (0.009) (0.010) (0.025) (0.024) (0.012) (0.028) (0.014)

area

Local Yes Yes Yes Yes Yes Yes Yes

authority *

month-by-
year Fes

Neighbourho Yes Yes Yes Yes Yes Yes Yes

od

FEs

Observations 2,085,120 520,560 520,200 520,260 521,160 520,560 521,280

(N*T)

Notes: Coefficients, standard errors in parentheses. Standard errors adjusted for clustering on the neighbourhood
level.
 Residential turnover and crime—Evidence from administrative data for England and Wales

*/**/*** denote statistical significance at the 10%, 5% and 1% level. Outcomes are rates per 100 population.

Br J Criminol (2023) 63(6): 1460–1481 at 1480

What do these results imply? Broadly spoken, the predictions from the various theoretical arguments appear to be
supported by the data—higher turnover does seem to increase crime in a range of different contexts. Increases in
violent crime appear to support all possible mechanisms from social disorganization to lower detection probabilities
for crime. However, the fact that anti-social behaviour, criminal damage, arson and public order offences, i.e.,
offences closely linked to social disorder, appear to be unchanged by residential turnover appears to speak against
the mechanism underpinning social disorganization as an explanation. The fact that effects are particularly
pronounced for property crimes supports the same mechanisms as well as an additional mechanism—more affluent
targets arriving in the neighbourhood. In light of routine activity theory this inflow would increase the availability of
attractive targets for property crime—consistent with the pronounced increases in burglary and theft—but also
change the availability of capable guardians, for example, if residential turnover makes it more difficult to identify
strangers in a neighbourhood. Finally, the increase in burglaries could also be explained by a mechanism akin to
the Becker-model: Are people carrying crates out of a house and loading these into a truck burglars or a removal
company? However, the fact that there is no uniform increase in crime suggests that the mechanism is not an
across-the-board shift in the probability of getting caught when committing a crime.

On a practical level, the results also suggest that additional police resources devoted to areas that experience
residential turnover might be able to counteract the additional crimes that would otherwise occur. However, it is
important to be clear that effect sizes suggest a fairly modest proportion of crime can be linked to residential
turnover.

References

Angrist, Joshua and Stephen Pischke (2009). Mostly Harmless Econometrics: An Empiricist's Companion.
Princeton University Press.

Becker, G. S. (1968), 'Crime and Punishment: An Economic Approach', Journal of Political Economy, 76: 169–217.
doi:10.1086/259394.

Boggess, L. N. and Hipp, J. R. (2010), 'Violent Crime, Residential Instability and Mobility: Does the Relationship
Differ in Minority Neighborhoods', Journal of Quantitative Criminology, 26: 351–70. doi:10.1007/s10940-010-9093-
7.

Bolan, M. (1997), 'The Mobility Experience And Neighborhood Attachment', Demography, 34: 225–37.
doi:10.2307/2061701.

Braakmann, N. (2017), 'The Link Between Crime Risk and Property Prices in England and Wales: Evidence from
Street-Level Data', Urban Studies, 54: 1990–2007.

Braakmann, N. (2022), 'Does Stop and Search Reduce Crime? Evidence from Street-Level Data and a Surge in
Operations Following a High-Profile Crime', Forthcoming: Journal of the Royal Statistical Society, Series A, 185(3):
1370–1397.

Cheng, C. and Long, W. (2018), 'Improving Police Services: Evidence from the French Quarter Task Force', Journal
of Public Economics, 164: 1–18. doi:10.1016/j.jpubeco.2018.05.002.

Cohen, L. E. and Felson, M. (1979), 'Social Change and Crime Rate Trends: A Routine Activity Approach',
American Sociological Review, 44: 588–608. doi:10.2307/2094589.
 Residential turnover and crime—Evidence from administrative data for England and Wales

Di Tella, R. and Schargrodsky, E. (2004), 'Do Police Reduce Crime? Estimates Using the Allocation of Police
Forces after a Terrorist Attack', American Economic Review, 94: 115–33. doi:10.1257/000282804322970733.

Dietz, R. D. and Haurin, D. R. (2003), 'The Social and Private Micro-Level Consequences Of Homeownership',
Journal of Urban Economics, 54: 401–50. doi:10.1016/s0094-1190(03)00080-9.

DiPasquale, D. and Glaeser, E. L. (1999), 'Incentives and Social Capital: Are Homeowners Better Citizens?',
Journal of Urban Economics, 45: 354–84. doi:10.1006/juec.1998.2098.

Draca, M., Machin, S. and Witt, R. (2011), 'Panic on the Streets of London: Police, Crime, and the July 2005 Terror
Attacks', American Economic Review, 101: 2157–81. doi:10.1257/aer.101.5.2157.

Ellen, I. G., Lacoe, J. and Sharygin, C. A. (2013), 'Do Foreclosures Cause Crime?', Journal of Urban Economics,
74: 59–70. doi:10.1016/j.jue.2012.09.003.

Br J Criminol (2023) 63(6): 1460–1481 at 1481

Farrall, S., Hay, C., Jennings, W. and Gray, E. (2016), 'Thatcherite Idegology, Housing Tenure and Crime: The
Socio-Spatial Consequences of the Right to buy for Domestic Property Crime', British Journal of Criminology, 56:
1235–52.

Gibbons, S. (2004), 'The Cost of Urban Property Crime', Economic Journal, 114: F441–63. doi:10.1111/j.1468-
0297.2004.00254.x.

Harding, J. P., Rosenblatt, E. and Yao, V. W. (2008), 'The Contagion Effect of Foreclosed Properties', Journal of
Urban Economics, 66: 164–78.

Herfindahl, Orris C. (1950) Concentration in the U.S. Steel Industry. Unpublished doctoral dissertation, Columbia
University.

Hipp, J. R. (2007), 'Income Inequality, Race, and Place: Does the Distribution of Race and Class Within
Neighborhoods Affect Crime Rates?', Criminology, 45: 665–97. doi:10.1111/j.1745-9125.2007.00088.x.

Katz, C. M., Wallace, D. and Hedberg, E. C. (2013), 'A Longitudinal Assessment of the Impact of Foreclosure on
Neighborhood Crime', Journal of Research in Crime and Delinquency, 50: 359–89.

Kirk, D. S. and Hyra, D. S. (2012), 'Home Foreclosures and Community Crime', Social Science Quarterly, 93: 648–
70. doi:10.1111/j.1540-6237.2012.00891.x.

Krivo, L. J. and Peterson, R. D. (1996), 'Extremely Disadvantaged Neighborhoods and Urban Crime', Social Forces,
75: 619–48. doi:10.2307/2580416.

MacDonald, J., Klick, J. and Grunwald, B. (2016), 'The Effect of Private Police on Crime: Evidence from a
Geographic Regression Discontinuity Design', Journal of the Royal Statistical Society, Series A, 179: 831–46.

Mastrobuoni, G. (2019), 'Police Disruption and Performance: Evidence from Recurrent Redeployments within a
City', Journal of Public Economics, 176: 18–31. doi:10.1016/j.jpubeco.2019.05.003.

McMillen, D., Sarmiento-Barbieri, I. and Singh, R. (2019), 'Do More Eyes on the Street Reduce Crime? Evidence
from Chicago's Safe Passage Program', Journal of Urban Economics, 110: 1–25. doi:10.1016/j. jue.2019.01.001.
 Residential turnover and crime—Evidence from administrative data for England and Wales

Moulton, B. (1990), 'An Illustration of a Pitfall in Estimating the Effects of Aggregate Variables on Micro Units',
Review of Economics and Statistics, 72: 334–8.

Oh, J. -H. (2004), 'Race/Ethnicity, Homeownership and Attachment', Race and Society, 7: 63–77.
doi:10.1016/j.racsoc.2005.05.002.

Ross, M. (1977), Economics, Opportunity and Crime. Montreal, Canada: Renouf.

Rogers, W. H. and Winter, W. (2009), 'The Impact of Foreclosures on Neighboring House Sales', Journal of Real
Estate Research, 31: 455–80.

Roncek, D. W. (1981), 'Dangerous Places: Crime and Residential Environment', Social Forces, 60: 74–96.
doi:10.2307/2577933.

Sampson, R. J. and Byron Groves, W. (1989), 'Community Structure and Crime: Testing Social-Disorganization
Theory', American Journal of Sociology, 94: 774–802.

Sampson, R. J., Raudenbush, S. W. and Earls, F. (1997), 'Neighbourhoods and Violent Crime: a Multilevel Study of
Collective Efficacy', Science, 277: 918–24. doi:10.1126/science.277.5328.918.

Schuetz, J., Been, V. and Ellen, I. G. (2008), 'Neighborhood Effects of Concentrated Mortgage Foreclosures',
Journal of Housing Economics, 17: 306–19. doi:10.1016/j.jhe.2008.09.004.

Shaw, Clifford, and Henry McKay (1942). Juvenile Delinquency and Urban Areas. Chicago: University of Chicago
Press.

Sherman, L. W., Gartin, P. R. and Buerger, M. E. (1989), 'Hot Spots of Predatory Crime: Routine Activities and the
Criminology of Place', Criminology, 27: 27–56. doi:10.1111/j.1745-9125.1989.tb00862.x.

Simpson, E. H. (1949), 'Measurement of Diversity', Nature, 163: 688–688. doi:10.1038/163688a0.

Smith, W. R., Frazee, S. G. and Davison, E. L. (2000), 'Furthering the Integration of Routine Activity and Social
Disorganization Theories: Small Units of Analysis and the Study of Street Robbery as a Diffusion Process',
Criminology, 38: 489–524. doi:10.1111/j.1745-9125.2000.tb00897.x.

Stock, J. and Watson, M. (2009), 'Heteroskedasticity-Robust Standard Errors for Fixed Effects Panel Data
Regression', Econometrica, 76: 155–74.

Thompson, S. K., Bucerius, S. M. and Luguya, M. (2013), 'Unintended Consequences of Neighbourhood

Restructuring—Uncertainty, Disrupted Social Networks and Increased Fear Of Victimization Among Young Adults',
British Journal of Criminology, 53: 924–41. doi:10.1093/bjc/azt032.

Urbanik, M. -M., Thompson, S. K. and Bucerius, S. M. (2017), '`Before there was Danger but there was Rules. And
Safety in those Rules': Effects of Neighbourhood Redevelopment On Criminal Structure', British Journal of
Criminology, 57: 422–40.

Which? (2020). Exchange and Completion. https://www.which.co.uk/money/mortgages-and-property/first-time-

buyers/buying-a-home/exchange-and-completion-a6n6m8p8czvh .
 Residential turnover and crime—Evidence from administrative data for England and Wales

1 Policing in England and Wales is organized through 43 Police Forces covering specific areas. In addition, Scotland
and Northern Ireland have their own police forces. There are three further police forces covering railway land (the
British Transport Police), civilian nuclear sites (Civil Nuclear Constabulary) and military installations and personnel
(Ministry of Defence Police).
2 Details on the location anonymization process and a description of map points can be found at
https://data.police.uk/about/#location-anonymisation .
3 See https://www.gov.uk/government/statistical-data-sets/price-paid-data-downloads . © Crown copyright and
database right 2020. These data are licensed under the Open Government Licence v3.0.
4 The notion of anti-social behaviour was introduced in the Crime and Disorder Act 1998 with some later changes
through the Anti-social Behaviour Act 2003. It basically describes acting 'in a manner that caused or was likely to cause
harassment, alarm or distress to one or more persons not of the same household as himself' (the perpetrator) (Part I,
Chapter 1, Section 1 of the Crime and Disorder Act 1998).
5 Broadly offences which cause fear, alarm or distress to the public that are more serious than anti-social behaviour,
such as affray, the provocation of violence, harassment, violent disorder, public nuisance, acts to stir up racial or
religious hatred or acts outraging public decency. Public order offences in the data also principally cover offences such
as treason, espionage, terrorism, attempting to injure or alarm the sovereign and the prohibition of quasi-military
organizations. However, these are rare.
6 The Hirschman-Herfindahl index (Herfindahl, 1950)/Simpson diversity index (Simpson, 1949) is a measure
concentration. They are defined as the sum of squared proportions/market shares of each type in an area of market. A
value of 1 suggest that there is only one type, while values close to zero suggests the existence of many different types
each making up a small proportion of the population.

End of Document