PLoS ONE plos plosone PLOS ONE 1932-6203 Public Library of Science San Francisco, CA USA 10.1371/journal.pone.0289423 PONE-D-23-08904 Research Article Social sciencesSociologyCriminologyCrime Medicine and health sciencesEpidemiologyMedical risk factorsTraumatic injury risk factorsViolent crime Medicine and health sciencesPublic and occupational healthTraumatic injury risk factorsViolent crime Social sciencesSociologyCriminologyCrimeViolent crime Social sciencesLaw and legal sciencesCriminal justice systemLaw enforcementPolice People and placesPopulation groupingsProfessionsPolice People and placesGeographical locationsEuropeEuropean UnionUnited KingdomEngland Earth sciencesGeographyHuman geographySocial geographyNeighborhoods Social sciencesHuman geographySocial geographyNeighborhoods Research and analysis methodsMathematical and statistical techniquesMathematical models Research and analysis methodsMathematical and statistical techniquesStatistical methodsRegression analysisLinear regression analysis Physical sciencesMathematicsStatisticsStatistical methodsRegression analysisLinear regression analysis Research and analysis methodsMathematical and statistical techniquesStatistical methodsForecasting Physical sciencesMathematicsStatisticsStatistical methodsForecasting Local deprivation predicts right-wing hate crime in England Local deprivation predicts right-wing hate crime in England Belgioioso Margherita Conceptualization Project administration Writing – original draft 1 * Dworschak Christoph Conceptualization Formal analysis Software Validation Visualization Writing – review & editing 2 https://orcid.org/0000-0003-4149-3211 Gleditsch Kristian Skrede Conceptualization Supervision Writing – review & editing 3 4 School of Politics and International Studies, University of Leeds, Leeds, United Kingdom Department of Politics, University of York, York, United Kingdom Department of Government, University of Essex, Colchester, United Kingdom Peace Research Institute Oslo, Oslo, Norway Braha Dan Editor University of Massachusetts, UNITED STATES

The authors have declared that no competing interests exist.

 * E-mail: M.Belgioioso@leeds.ac.uk 6 9 2023 2023 18 9 e0289423 3 4 2023 18 7 2023 2023 Belgioioso et al This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

We argue that community deprivation can increase the risk of right-wing radicalization and violent attacks and that measures of local deprivation can help improve forecasting local hate crime rates. A large body of research stresses how experiences of deprivation can erode the perceived legitimacy of political leaders and institutions, increase alienation, and encourage right-wing radicalization and hate crime. Existing analyses have found limited support for a close relationship between deprivation and radicalization among individuals. We provide an alternative approach using highly disaggregated data for England and show that information on local deprivation can improve predictions of the location of right-wing hate crime attacks. Beyond the ability to predict where right-wing hate crime is likely, our results suggest that efforts to decrease deprivation can have important consequences for political violence, and that targeting structural facilitators to prevent far-right violence ex ante can be an alternative or complement to ex post measures.

 Research Council of Norway 302445 https://orcid.org/0000-0003-4149-3211 Gleditsch Kristian Skrede Gleditsch is grateful for funding under the project "The Crime-Reducing Effect of Education", grant number 302445, from the Research Council of Norway. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Data Availability Data available at the Harvard Dataverse at https://doi.org/10.7910/DVN/GSRPQY. Introduction

Far-right violence is a major security concern in many high-income democracies, and the number of right-wing terrorist attacks has by one estimate increased by 250% over the period 2015–2019 [1]. Many have argued that social and economic deprivation can motivate political violence such as civil war and terrorist attacks [2], and the link between deprivation, radicalization, and violence seems particularly relevant to many nativist right-wing ideologies. Existing research has suggested limited evidence for a link between deprivation and radicalization or violence at the individual level [3, 4]. We argue that existing studies have often ignored the role of local community characteristics in fostering radicalization and increased risk of violent attacks, and that it may be possible to leverage information on local deprivation to better predict the risk of right-wing hate crime attacks.

In this article, we examine whether higher neighborhood-level deprivation is associated with a greater risk of far-right hate crime. Observing local deprivation can make individuals feel that their communities have been left behind, perceive political leaders and existing institutions as unresponsive, unhelpful, or illegitimate, and foster greater susceptibility to far-right ideas and frames of responsibility [5]. We provide a first study to systematically examine this relationship at the level of neighborhoods, using disaggregated data on deprivation across neighborhoods in England as well as the location of right-wing hate crime attacks. We train our models on data for 2015–2018 and find substantially more far-right attacks in more deprived local communities. We then show that information on local deprivation substantially improves forecasts of right-wing attacks over the period 2019–21 over a baseline model limited to other demographic and socio-economic information. This study demonstrates how it is possible to develop better models to forecast right-wing hate crime attacks by using local measures of deprivation. Our arguments on the mechanisms underlying our findings also suggest that policies aimed at decreasing deprivation could have important benefits for reducing political violence, and that efforts to target structural local drivers of radicalization beyond ideological beliefs could improve preventive counter-terrorist measures and help complement ex post interventions targeting people who are already radicalized or have already carried out attacks.

 Local deprivation and extremist political violence

Many theories of grievances and political violence argue that political and economic inequalities following group lines can generate resentment and political mobilization that increase the likelihood of resort to violence [2]. We believe that a similar process also applies to right-wing radicalization and violence at the local level. We focus on local socio-economic deprivation, by which we mean how average outcomes on relevant social and economic indicators for a local community compare to the average for other local communities. A local community is more deprived when it is characterized by low or worse outcomes compared to other local communities. High local deprivation can erode the perceived legitimacy of political leaders and central institutions, undermine trust, increase political and social alienation, and even lead to support for the use of violence [6, 7]. We posit that the local neighborhood is the main reference point for individual comparisons, shaping individual perceptions and beliefs about outcomes and distributions. The local neighborhood often forms the basis for community identification and the evaluations people make about whether outcomes or distributions are “fair” or not [8, 9]. Individuals who observe high deprivation in their proximate local environments are more likely to consider their communities as neglected by the state and other political authorities [10]. This form of political dissatisfaction is likely to be accompanied by declining confidence in established political channels and the ability to change outcomes or overcome injustices through conventional political participation.

Once individuals are politically disaffected after experiencing local deprivation, they can also become more susceptible to extremist right-wing ideologies. These often portray deprivation as a result of social and economic injustices where the central political establishments either actively conspire to undermine the interests of native communities or alternatively fail to adequately respond to their needs and interests [11]. Far-right narratives tend to present as a common theme “a collective sense of persecution [of the “native” population], presenting themselves as victims of societal oppression” [12, 13]. These narratives offer a sense of belonging to a group—encouraging the self-identification to a broader group of victimized native citizens. They provide a clear identification of responsibility and culpable agents, and offer narratives that dehumanize “outsiders” in ways that generate more polarized and hardened group identities and fuel fears of future victimization [14]. This can, in turn, increase the perceived legitimacy of political violence.

There is empirical evidence that perpetrators of hate crimes tend to carry out violence within their local neighborhood [15]. Perpetrators often seek out victims within their immediate vicinity, as 1) it is easier to identify targets they hold prejudice against in the local community [16], 2) access, planning, and execution might be easier for local targets [17], 3) local targets often have symbolic significance for both perpetrators and targeted communities [18], and 4) local hate crimes often receive more attention in the immediate community, which in turn can be helpful for generating fear and exerting control over targets and recruiting activists to the cause [19].

Our concept of local deprivation focuses on the characteristic outcomes for local communities rather than individual outcomes or distributions within the community population. Note that if local community deprivation is the most salient focus for relevant influence for radicalization and attacks, then community deprivation does not necessarily translate to a systematic relationship between indicators of deprivation and radicalization at the individual level. The individuals who carry out attacks need not necessarily be relatively more deprived, even if their local community is deprived. A similar claim in the context of terrorism notes that one would not expect a clear relationship between deprivation and attacks if groups can select the most able among the pool of their potential supporters [20]. Therefore, neither highly geographically aggregated approaches nor individual level studies can conclusively evaluate the potential relationship between local deprivation and radicalization or the risk of hate crime.

Clearly not all individuals holding extremist attitudes will go on to endorse or commit violent attacks. The UK Prevent scheme, for example, has suggested a pyramid model of individual radicalization, outlined in Fig 1 [21, 22]. At the top of the pyramid is the subset of the fully radicalized individuals who actually carry out acts of terrorism political violence. These are a relatively small share of a larger set of individuals who hold some degree of sympathy for, or are receptive to, radical ideological beliefs. This larger, active group of “sympathizers” can provide a pool of potential recruits or facilitate attacks, although it may be difficult to identify which sympathizers go on to carry out violent attacks. Below the active sympathizers are a larger group of uncommitted individuals who may be potentially susceptible to extreme ideological messages, i.e., the set considered as “vulnerable populations” in the UK prevent scheme. Our interest here is in seeing whether the notion of such a subset can be used to identify local contexts where there is a higher probability of radicalization and higher risk of right-wing attacks.

10.1371/journal.pone.0289423.g001 Pyramid of radicalization.

The UK Prevent Scheme explicitly “aims to stop people being drawn into terrorist-related activity” [21]. Existing work often takes a negative view on the prospects for forecasting right-wing hate crime on the grounds that it is very difficult to predict precisely which individuals among the potentially susceptible will actually become fully radicalized or carry out violent attacks [23]. But even if one cannot reliably predict which individuals will go on to engage in right-wing hate crime, we may still be able to identify what socio-economic characteristics make local contexts more prone to far-right attacks, if neighborhood-level deprivation plays an important role in inducing vulnerability to right-wing radicalization and a higher risk of local attacks.

Following this discussion, we propose that higher neighborhood-level social and economic deprivation increases the risk of right-wing violence, and that measures of local deprivation can help improve predictions of right-wing hate crime.

 Empirical data

We examine local deprivation and right-wing hate crime across England using data for Lower Layer Super Output Areas (LSOAs), the highest geographical resolution used in the UK census data. Each LSOA comprises between 400–1,200 households, with a resident population between 1,000 and 3,000 persons [23].

To capture hate crime, we obtained disaggregated Hate Crime Statistics, recorded by the UK Police for 2015–2021, and provided by the Crime Analysis Unit of the Home Office following a Freedom of Information request by the authors (FOI 69240). Hate crimes differ from other criminal offences in that they “express a number of socio-political objectives by targeting individuals based on their perceived group membership”, including race, religion, disability, or sexual orientation and gender identity [24]. Hate crimes often resemble the more general concept of terrorist attacks, where the identity of the direct victims may be subordinate to the intention of indirectly imposing “terror” or cost on a different specific target such as the government or the general population. Hate crimes are often intended to convey a message to a broader audience beyond the direct victim, and often aim to instill fear, force a change in government policy, or help attract followers and recruits to the cause. Hate crimes also reflect key right-wing beliefs such as national chauvinism, xenophobia, and racism, where certain individuals are acceptable targets by virtue of their identity [25]. The perpetrators often seek out suitable victims within their immediate vicinity, as it is easier to identify targets and plan attacks within more familiar local environments, to gain access to these targets, and local targets often have a clear symbolic significance for perpetrators and audiences. Existing surveys find that hate crimes tend to be perpetrated locally, reflected in clear hate-crime hotspots [26]. In light of the above, we posit that the location of hate crimes is likely to reflect local right-wing radicalization and can be used to consider possible factors contributing to radicalization and the risk of hate crime attacks.

To capture local deprivation–which we conceptualize as and the extent of deviation from equality in the national distributions of goods, services, or negative outcomes (i.e., burdens),–we rely on indicators from the 2015 English Indices of Deprivation (IoD). These scores measure relative local deprivation through a series of indices for 32,844 neighborhoods in England. We use the indices for deprivation in income, education, and environment as complimentary measures of local deprivation. We refer to the S1 Appendix for more detailed information on the construction of these indices and underlying data as well as the correlations between the different indices. As we only have access to the reported aggregate indices, we are unable to directly explore the inputs to the distinct indicators into a single local deprivation index in more detail, or to combine the input indicators in one general deprivation index. Fig 2 displays the distribution of hate crime and deprivation across LSOAs in England. The data underlying these deprivation indicators were gathered between 2012–2013. We believe the lag between our local deprivation measures and the post 2015 hate crime outcomes ensures that the measures are clearly prior to the response. Deprivation tends to be sticky, and comparisons to data measured at least two years before should be a reasonable window to consider the possible impact of past exposure to deprivation on outcomes. However, we acknowledge that the specific time lag used here is partly driven by available data. We do not have adequate coverage of hate crime data prior to 2015 or more recent deprivation indices. As such, we are restricted to a cross-sectional setup and are unable to consider empirically the temporal dimension of the exposure-violence relationship. Doing so would be a valuable contribution for future research to pursue.

10.1371/journal.pone.0289423.g002 Maps and histograms of the distributions of hate crime and deprivation indices.

Source for maps: Office for National Statistics licensed under the Open Government Licence v.3.0.

Local deprivation and right-wing extremism may have common causes, and we condition on several other covariates to reduce confounding and better identify plausible partial relationships. We include data on total LSOA population size (logged), average age, percent of rural/urban population, level of unemployment, the Black and Minority Ethnic (BAME) population share, the share of people reporting to have Christian religion, and the share of people with Arab ethnicity. We also include spatial lags of our deprivation measures, capturing the mean values of all bordering neighborhoods. A key concern here is that many forms of political violence often diffuse, and attacks in one area may inspire similar efforts in nearby communities even in the absence of similar characteristics [27, 28]. However, if neighboring areas are likely to be similar to one another, then we overstate the effects of local deprivation without accounting for spatial clustering [28]. Finally, the reliability of hate crime reporting likely differs between police force districts, which also coincide with administrative districts. Therefore, we also include police force district fixed effects to partial out differences between districts. In the S1 Appendix we further explore the role of missingness and report additional analyses in which we omit neighborhoods without any hate crime reports from our sample (S2 Table in S1 Appendix) and model the zeros as a separate data generating process (S3 Table in S1 Appendix). We also report our main results with cluster-robust standard errors at the police district level (S5 Table in S1 Appendix). Our inferences remain the substantively unchanged. A list of all police force districts is included in S1 Table in S1 Appendix.

In Table 1 (below) we report the result for linear regression models of the log number of hate crimes per 100,000 between 2015–2018 against a baseline model with covariates only (1), models with deprivation measures only (2–4), and models combining deprivation measures and covariates (5–7). We find positive conditional associations for all the deprivation indicators, reflecting that we see more hate crime events in areas with more deprivation. These effects are visualized in Fig 3, based on bootstrapped observed values. Moreover, the full models 5–7 exhibit higher R2s and lower RMSEs and AICs than the baseline model 1, illustrating that information on deprivation contributes notably to accounting for variance in hate crimes.

10.1371/journal.pone.0289423.g003 Marginal effect visualizations of full models (5–7). 10.1371/journal.pone.0289423.t001 Linear regression of logged hate crime 2015–2018 on deprivation.
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7
Income deprivation 7.226 4.048
(0.095) (0.181)
Living envir. deprivation 0.029 0.020
(0.001) (0.001)
Education deprivation 0.028 0.008
(0.001) (0.001)
Income depr. (neighb.) 1.704
(0.171)
Living envir. depr. (neighb.) −0.002
(0.001)
Education depr. (neighb.) 0.008
(0.001)
Covariates x x x x
Police district FE x x x x
Num.Obs. 32201 32201 32201 32201 32201 32201 32201
R2 0.424 0.151 0.057 0.075 0.439 0.440 0.430
RMSE 1.46 1.77 1.86 1.84 1.44 1.44 1.45
AIC 115661 128048 131444 130822 115332 114758 114802

(Standard errors in parentheses).

The reported results from these regression models are all in-sample, and intended to demonstrate the conditional associations between the local deprivation indicators and hate crime in a manner that is simple to evaluate. However, in-sample estimation is prone to problems of overfitting. To see whether the apparent improvement in difference in performance also extends to improved forecasts further into the future, we conduct a prediction exercise using data for 2019–2021 as a hold-out period. Moreover, since a linear regression framework imposes a number of potentially strong assumptions about functional form, we use more flexible random forests, which also allow us to better consider the plausible importance of the different deprivation indicators simultaneously in the same model.

 Prediction

Turning from identification to prediction, we employ random forest models to benchmark the importance of local deprivation features for predicting hate crimes in England based on out-of-bag (OOB) node purity, comparing both the null model and full model feature space. Fig 4 suggests that structural information on the local BAME share and distribution of faith contribute the most to the prediction of hate crimes. In line with the model statistics of the in-sample analyses above, income deprivation visibly outperforms both education deprivation and living environment deprivation. The performance of the living environment and education deprivation indices for predicting hate crimes is comparable to that of other relevant covariates. One factor contributing to the performance of the income deprivation index may be that it takes the local number of asylum seekers into account as one of its components. In the S1 Appendix we further explore the role of ethnic diversity by interacting BAME share with our deprivation indicators, finding that the effect of income deprivation is indeed slightly moderated by ethnic diversity (S4 Table in S1 Appendix).

10.1371/journal.pone.0289423.g004 Feature importance.

Turning to the substantive predictive performance of the measures, we then compare observed hate crime rates to those predicted by our models. Stratifying hate crime rate into eight bins, Fig 5 suggests that our model performs reasonably well in predicting the spatial distribution of hate crimes. In addition, comparing performance metrics of the full model to those of the null model that does not include information on deprivation, we find that the RMSE increases from 1.64 to 1.67, and the R2 decreases from 0.128 to 0.088. In other words, using demographic and structural information predicts about 9% of variance in the number of future hate crimes, while including information on deprivation increases this predicted variance to about 13%. Fig 5 suggests that deprivation provides important information to anticipate the predicted outcomes, in line with our argument that local deprivation breeds radicalization. For example, drawing on the deprivation measures attenuates the predicted number of hate crimes for many locations in the Cumberland unitary authority district in North-West England, but also increases numbers for several LSOAs in Durham in the North East, thereby moving the predictions closer to the observed rate of hate crimes.

10.1371/journal.pone.0289423.g005 Observed and predicted hate crimes rates, 2019–2021.

Source for maps: Office for National Statistics licensed under the Open Government Licence v.3.0.

Fig 6 shows how the full and null model compare to the observed data and to each other. The dotted line shows the line of perfect fit and the continuous line shows the regression lines. The null model clusters marginally more at low values, while the full model shows more spread across the range of observed hate crime. As may be expected, the plot also suggests that the model’s prediction error follows a regression-to-the-mean pattern, in which values at the lower end of the scale tend to be overestimated and very high hate crime rates are underestimated. The high degree of dispersion, with most neighborhoods having low rates of hate crime and some having very high rates, leads both models to underestimate the rate of hate crime for more extreme cases.

10.1371/journal.pone.0289423.g006 Comparing predictions with and without deprivation measures. Discussion and conclusion

We have examined the relationship between local deprivation and right-wing hate crime. Unlike previous studies that examine this relationship either at the individual level or at a highly aggregated level, our theory and analysis focus on local communities. We argue that neighborhood-level analyses allow more inferential power on the underlined links between deprivation and radicalization. Our results are consistent with the expectation that local deprivation is an important driver of right-wing radicalization and risk of far-right attacks. We also show that measures of local deprivation help improve forecasts of where right-wing hate crime can be expected.

Beyond their importance for theories of deprivation and radicalization and for forecasting where hate crimes are likely to occur, our results also suggest a potential scope for policies to prevent future radicalization through reducing deprivation. Efforts to identify potential perpetrators of violent attacks to date appear to have low predictive ability and more general efforts to counter ideologies that may encourage attacks have been criticized for problematic profiling and demonization of specific communities [29]. The most common responses to countering hate crime focus on ex-post efforts to deradicalize, and target people who are already radicalized and either have carried out terrorist attacks or have attempted to do so. Our findings suggest that efforts to reduce local deprivation, beyond leading to other socially desired positive outcomes, may have important additional consequences for reducing hate crime. For example, policies focused on “levelling up”, or improving the most deprived communities [30], could in addition to improving local living standards and decreasing inequalities, also help decrease the risk of far-right violence. However, we must stress that the ability to change deprivation through measures and intervention is a separate issue altogether, not directly examined in our analyses here. Direct plans for interventions must consider potential barriers for implementation and the possible unintended consequences as well as strategies to evaluate effectiveness. However, even though inequalities tend to be sticky and may be difficult to reduce, the efforts to reduce specific forms of inequality such as child poverty and deprivation among pensioners under the New Labour government over the period 1997–2010 are generally considered to have been remarkably effective, and provide a proof of concept for interventions to reduce deprivation and inequalities [31]. Above all, our work suggests that predicting right-wing violence in geographical hotspots based on local deprivation may be considerably easier than predicting the actions of individual perpetrators. It provides at least an intellectual rationale for exploring how neighborhoods can be a fruitful context to explore possible new preventive counter-terrorist measures.

 Supporting information Additional supporting information is provided in the document.

(DOCX)

 References Global Terrorism Index: Measuring the Impact of Terrorism. Sydney: Institute for Economics and Peace; 2020. Cederman L.-E., Gleditsch K.S. Buhaug H. Inequality, Grievances, and Civil War. New York: Cambridge University Press; 2013. Krueger A.B., Malečková J. Education, poverty and terrorism: Is there a causal connection? Journal of Economic Perspectives. 2002; 17: 119–144. Bhui K., Warfa N., Jones E. Is violent radicalisation associated with poverty, migration, poor self-reported health and common mental disorders? PLOS ONE. 2014; 9: 90718–90718. doi: 10.1371/journal.pone.0090718 24599058 Feola M. “You will not replace us”: The melancholic nationalism of whiteness. Political Theory. 2020; 49: 528–553. Schnaudt C. Political Confidence and Democracy in Europe: Antecedents and Consequences of Citizens’ Confidence in Representative and Regulative Institutions and Authorities. New York: Springer; 2019. Marien S, Werenr H. Fair treatment, fair play? The relationship between fair treatment perceptions, political trust and compliant and cooperative attitudes cross-nationally. European Journal of Political Research. 2019; 58: 72–95. Davies W., Herbert D. Communities within Cities: An Urban Social Geography. London: Belhaven Press; 1993. Galster G. On the nature of the neighbourhood. Urban Studies 2021; 38: 2111–2124. Kunst J.R., Milan O. Understanding violent extremism in the 21st century: The (re)emerging role of relative deprivation. Current Opinion in Psychology. 2020; 35: 55–59. doi: 10.1016/j.copsyc.2020.03.010 32344297 Hofstadter, R. The paranoid style in American politics. Harper’s Magazine 1967; November: 77–86. Lee, B. Narratives of the far right. Crest Security Review. 2021; Summer: 21–22. Gest J, Reny T., Mayer J. Roots of the radical right: Nostalgic deprivation in the United States and Britain. Comparative Political Studies. 2018; 51: 1694–1719. Homolar A., Löfflmann G. Populism and the affective politics of humiliation narratives. Global Studies Quarterly. 2021; 1: 1–11. Mason Gail. Hate crime and the image of the stranger. British Journal of Criminology. 2005; 45; 6: 837–859. Kleg M. Hate Prejudice and Racism. Albany, NY: State University of New York Press; 1993. Sullaway M. Psychological perspectives on hate crime laws. Psychology, Public Policy, and Law. 2004; 10; 3: 250–292. Wand Lu-in. Reframing hate. Journal of Criminal Law and Criminology. 2022; 112; 4; 847–73. Yahagi, Ken. The effects of hate groups on hate crimes, WINPEC Working Paper Series E1718; 2017. https://www.waseda.jp/fpse/winpec/assets/uploads/2014/05/E1718.pdf. Bueno de Mesquita E. The quality of terror. American Journal of Political Science. 2005; 49: 515–530. Prevent Strategy. Presented to Parliament by the Secretary of State for the Home Department 2011; https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/97976/prevent-strategy-review.pdf. Moghaddam F. M. The staircase to terrorism: A psychological exploration. In Bongar B., Brown L. M., Beutler L. E., Breckenridge J. N., & Zimbardo P. G. (Eds.). Psychology of Terrorism. New York: Oxford University Press. 2007; 69–80. Crone M. Radicalization revisited: Violence, politics and the skills of the body. International Affairs. 2016; 92: 587–604. Mills C.E., Freilich J.D., Chermak S.M. Extreme hatred: Revisiting the hate crime and terrorism relationship to determine whether they are “close cousins” or “distant relatives”. Crime and Delinquency. 2017; 63: 1191–1223. Jaki, S., Smedt, T.D. Right-wing German hate speech on twitter: Analysis and automatic detection. 2019; https://arxiv.org/abs/1910.07518. Roberts, C., Innes, M., Williams, M, Tregidga, J., Gadd, D. Understanding who commits hate crime and why they do it. Welsh Government Social Research 38/2013; 28–30. Braha D. Global civil unrest: contagion, self-organization, and prediction. PLOS ONE. 2012; 7(10), e48596. doi: 10.1371/journal.pone.0048596 23119067 Buhaug H., Gleditsch K.S. Contagion or confusion? Why conflicts cluster in space. International Studies Quarterly. 2008; 52; 2: 215–233 Prevent scheme: Why the government’s programme is so difficult. BBC News 20 October 2021. https://www.bbc.co.uk/news/uk-58975385. White Paper on “Levelling Up the United Kingdom”. UK Department of Levelling Up, Housing and Communities. 2022. https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/1052706/Levelling_Up_WP_HRES.pdf. Robert, J., Sibieta, L. Labour’s record on poverty and inequality. Institute for Fiscal Studies. 2013. https://ifs.org.uk/articles/labours-record-poverty-and-inequality. 10.1371/journal.pone.0289423.r001 Decision Letter 0 Braha Dan Academic Editor 2023 Dan Braha This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Submission Version 0

12 Jun 2023

PONE-D-23-08904Local deprivation predicts right-wing hate crime in EnglandPLOS ONE

Dear Dr. Gleditsch,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

==============================

Editor Comments:

The paper provides an intriguing and informative analysis. The referees have highlighted several concerns that require thorough attention and addressing from the authors. In addition, to further enhance its robustness, I would suggest that the authors incorporate lagged dependent variables in the regression model. This inclusion can account for the likelihood that previous instances of hate crimes contribute to subsequent hate crimes, potentially overshadowing the effect of local deprivation. Furthermore, it would be valuable for the authors to explore the possibility that the local deprivation effect can be understood within the context of micro-dynamic models that incorporate social networks as a mechanism for the diffusion of civil unrest behavior. For instance, Braha, D. (2012) examines this topic in their work "Global civil unrest: contagion, self-organization, and prediction" published in PloS One, 7(10), e48596. Discussing and referencing this work would provide additional insight into the dynamics of local deprivation and hate crime behavior.

==============================

Please submit your revised manuscript by Jul 15 2023 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols.

We look forward to receiving your revised manuscript.

Best,

Dan

-----------------------------

Prof. Dan Braha

Academic Editor

PLOS ONE

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at 

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and 

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. Thank you for stating the following financial disclosure: 

"Gleditsch is grateful for funding under the project "The Crime-Reducing Effect of Education", funded under the FINNUT Programme of the Research Council of Norway."

Please state what role the funders took in the study. If the funders had no role, please state: ""The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript."" 

If this statement is not correct you must amend it as needed. 

Please include this amended Role of Funder statement in your cover letter; we will change the online submission form on your behalf.

3. In your Data Availability statement, you have not specified where the minimal data set underlying the results described in your manuscript can be found. PLOS defines a study's minimal data set as the underlying data used to reach the conclusions drawn in the manuscript and any additional data required to replicate the reported study findings in their entirety. All PLOS journals require that the minimal data set be made fully available. For more information about our data policy, please see http://journals.plos.org/plosone/s/data-availability.

Upon re-submitting your revised manuscript, please upload your study’s minimal underlying data set as either Supporting Information files or to a stable, public repository and include the relevant URLs, DOIs, or accession numbers within your revised cover letter. For a list of acceptable repositories, please see http://journals.plos.org/plosone/s/data-availability#loc-recommended-repositories. Any potentially identifying patient information must be fully anonymized.

Important: If there are ethical or legal restrictions to sharing your data publicly, please explain these restrictions in detail. Please see our guidelines for more information on what we consider unacceptable restrictions to publicly sharing data: http://journals.plos.org/plosone/s/data-availability#loc-unacceptable-data-access-restrictions. Note that it is not acceptable for the authors to be the sole named individuals responsible for ensuring data access.

We will update your Data Availability statement to reflect the information you provide in your cover letter.

4. We note that Figures 2 and 5  in your submission contain map/satellite images which may be copyrighted. All PLOS content is published under the Creative Commons Attribution License (CC BY 4.0), which means that the manuscript, images, and Supporting Information files will be freely available online, and any third party is permitted to access, download, copy, distribute, and use these materials in any way, even commercially, with proper attribution. For these reasons, we cannot publish previously copyrighted maps or satellite images created using proprietary data, such as Google software (Google Maps, Street View, and Earth). For more information, see our copyright guidelines: http://journals.plos.org/plosone/s/licenses-and-copyright.

We require you to either (1) present written permission from the copyright holder to publish these figures specifically under the CC BY 4.0 license, or (2) remove the figures from your submission:

a. You may seek permission from the original copyright holder of Figures 2 and 5 to publish the content specifically under the CC BY 4.0 license.  

We recommend that you contact the original copyright holder with the Content Permission Form (http://journals.plos.org/plosone/s/file?id=7c09/content-permission-form.pdf) and the following text:

“I request permission for the open-access journal PLOS ONE to publish XXX under the Creative Commons Attribution License (CCAL) CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). Please be aware that this license allows unrestricted use and distribution, even commercially, by third parties. Please reply and provide explicit written permission to publish XXX under a CC BY license and complete the attached form.”

Please upload the completed Content Permission Form or other proof of granted permissions as an ""Other"" file with your submission.

In the figure caption of the copyrighted figure, please include the following text: “Reprinted from [ref] under a CC BY license, with permission from [name of publisher], original copyright [original copyright year].”

b. If you are unable to obtain permission from the original copyright holder to publish these figures under the CC BY 4.0 license or if the copyright holder’s requirements are incompatible with the CC BY 4.0 license, please either i) remove the figure or ii) supply a replacement figure that complies with the CC BY 4.0 license. Please check copyright information on all replacement figures and update the figure caption with source information. If applicable, please specify in the figure caption text when a figure is similar but not identical to the original image and is therefore for illustrative purposes only.

The following resources for replacing copyrighted map figures may be helpful:

USGS National Map Viewer (public domain): http://viewer.nationalmap.gov/viewer/

The Gateway to Astronaut Photography of Earth (public domain): http://eol.jsc.nasa.gov/sseop/clickmap/

Maps at the CIA (public domain): https://www.cia.gov/library/publications/the-world-factbook/index.html and https://www.cia.gov/library/publications/cia-maps-publications/index.html

NASA Earth Observatory (public domain): http://earthobservatory.nasa.gov/

Landsat: http://landsat.visibleearth.nasa.gov/

USGS EROS (Earth Resources Observatory and Science (EROS) Center) (public domain): http://eros.usgs.gov/#

Natural Earth (public domain): http://www.naturalearthdata.com/

5. Please include captions for your Supporting Information files at the end of your manuscript, and update any in-text citations to match accordingly. Please see our Supporting Information guidelines for more information: http://journals.plos.org/plosone/s/supporting-information. 

Additional Editor Comments:

Editor Comment:

The paper provides an intriguing and informative analysis. The referees have highlighted several concerns that require thorough attention and addressing from the authors. In addition, to further enhance its robustness, I would suggest that the authors incorporate lagged dependent variables in the regression model. This inclusion can account for the likelihood that previous instances of hate crimes contribute to subsequent hate crimes, potentially overshadowing the effect of local deprivation. Furthermore, it would be valuable for the authors to explore the possibility that the local deprivation effect can be understood within the context of micro-dynamic models that incorporate social networks as a mechanism for the diffusion of civil unrest behavior. For instance, Braha, D. (2012) examines this topic in their work "Global civil unrest: contagion, self-organization, and prediction" published in PloS One, 7(10), e48596. Discussing and referencing this work would provide additional insight into the dynamics of local deprivation and hate crime behavior.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Partly

Reviewer #2: Partly

Reviewer #3: Partly

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: No

Reviewer #3: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: No

Reviewer #2: No

Reviewer #3: No

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: This paper posits that community deprivation increases the risk of right-wing radicalization. It argues that previous literature has focused on the wrong level of analysis, by analyzing the deprivation and radicalization at the individual level. The analysis is based on disaggregated data for neighborhoods in England between 2015-2019.

The paper deals with an interesting question that merits empirical investigation. However, I believe that the theoretical arguments and empirical analysis could be made stronger. Below I list some comments, suggestions, and questions that I hope the authors may find useful.

One of the central arguments of the paper is that the local neighborhood is the main reference point for individuals and acts as a basis for community identification. Individuals who observe high deprivation in their local environment are then expected to consider their own community as neglected by the state and through their political dissatisfaction they may become more susceptible to extremist right-wing ideologies.

If individuals indeed strongly identify with their local community, it is unclear why one would expect that observing deprivation within one’s own community would lead to an increase in far-right hate crime within that same community. Especially since the communities included in the analysis are very small, comprising only between 400 – 1,200 households. Currently, this assumption is not explicitly discussed, except for one sentence on p.7 in which the authors refer to a policy document that argues hate crimes tend to be perpetrated locally. I believe the paper would benefit from making this assumption more explicit and from strengthening the theoretical arguments and references to empirical work to back it up.

The authors create three indices of local deprivation. When analyzing the results, they find that the income deprivation index outperforms education deprivation and living environment deprivation. The Appendix provides details on the indicators that are included in these indices. One of the indicators included in the income deprivation index is “Asylum seekers in England in receipt of subsistence support, accommodation support, or both”. I wonder to what extent this indicator is driving the estimated coefficients on the income deprivation index, i.e., it makes sense to expect more right-wing hate crime in local neighborhoods that host a higher share of asylum seekers. It would be interesting to see a robustness check in which this indicator is dropped from the income deprivation index (and perhaps included as a covariate).

The extent to which individuals identify with their local neighborhood likely strongly varies across individuals and neighborhoods. One element that might influence such identification could be the level of ethnic diversity. The authors currently control for the Black and Minority Ethnic Population share. Would there be a way to leverage the heterogeneity in this variable (or other variables the authors may think of), e.g., looking at an interaction with the deprivation indices?

Figure 2a shows that a substantial part of the local neighborhoods in the sample report zero hate crimes for the entire study period. In the appendix the authors say they think it is likely that some areas fail to report hate crime statistics to the Home Office. Would it be possible to provide more detailed information on the number of neighborhoods that reported zero hate crimes, and provide descriptive statistics of the covariates and deprivation indices for these neighborhoods compared to the others that do report hate crimes?

If the authors suspect that hate crime statistics are incomplete, would it be possible to conduct a robustness check relying on ACLED data – which is collected in a similar manner for the whole country and possible to link to local neighborhoods? In addition, this would allow to look at different types of conflict. It is not entirely clear why you would expect exposure to deprivation to lead to right-wing hate crimes in particular? One might expect a positive correlation between neighborhood deprivation and crime/conflict in general. ACLED data would allow to look at this. And, since ACLED records the actors that are involved in conflict events and provides some qualitative information on what happened, it might be possible to identify right wing hate crime and look at its relative importance as compared to overall crime in a neighborhood.

Data underlying the deprivation indices were collected between 2012-2013. The analysis looks at hate crimes committed between 2015-2019. I am wondering if it would be possible to say something about the timing between witnessing deprivation and committing hate crimes? E.g., is there any theoretical work suggesting what time frame matters, how long one should be exposed to deprivation before moving to committing hate crimes? Is there anything that can be done empirically? E.g., by collecting info on deprivation from earlier periods as well & looking at longer historical deprivation?

Could you clarify for what time period the covariates were collected?

The analysis includes fixed effects at the level of police force districts. It would be interesting to see more information on the police force districts. How large are they, how many of them are there, and how do they relate to the local neighborhoods / LSOAs? Perhaps you could show this on a map?

Could you add explanatory notes to the tables? I wonder for instance if the standard errors are clustered at the level of the LSOA, or at the level of the police force district?

It would be interesting to see full results reported (maybe in appendix), i.e., including estimated coefficients on the covariates.

In terms of predicting future hate crimes there is a relatively small difference between a model that only includes demographic and structural information compared to a model that additionally includes information on deprivation (a move from explaining 9% of variance to 13%). Also visually, in Figure 5, it is hard to see a difference between the full model (b) and the null model (c). Then, in terms of policy implications, to what extent could one expect that efforts to reduce local deprivation may reduce hate crime?

While replication data is not yet made available, the authors do note that it will be made available on Harvard Dataverse upon publication.

The resolution of the Figures in the current PDF version could be improved.

Reviewer #2: Dear author(s),

I enjoyed reading your manuscript. Ultimately, however, I felt that some of your arguments need more development, and I could not understand with clarity some of your measures and methodological choices. Consequently, I was not sure what to learn from your findings, beyond the fact that neighborhoods are important and that deprivation is statistically associated with deprivation.

In this review I document my concerns, as well as suggestions on how to address some of them. Please take my suggestions just as recommendations.

GENERAL COMMENTS:

1. (p. 3, para 1) Personally, I would avoid framing individual-level research as the “wrong unit of analysis.” At least I would defend that statement very carefully. In any case, I find that these kinds of dichotomies can often be a fallacy. Individual and community-level research are arguably both important, perhaps for different reasons, but that importance can co-exist.

a. In the case of your research study, some individual-characteristics are important correlates and/or causes of right-wing extremism (e.g., certain political attitudes), while certain places may also draw that type of violence from its characteristics.

b. You could argue convincingly that one level is more important, but that any level is “wrong” is a strong statement that would require more support.

2. Your front-end was well-written and clear, but I also found it underdeveloped. For example, consider the UK Prevent scheme. To be clear, I understand that the authors’ have a different perspective of the scheme, or that my interpretation could be misguided, though I am sharing my perspective for reflection. It begins with the name, Prevent, suggesting that the framework could help curtain terrorism. Second, it suggests that being a “vulnerable population” is a condition for being a “sympathizer,” which is a condition for being a “terrorist.” The very terminology portraits terrorists as naïve individuals who are passively enticed to a narrative of violence because of their vulnerability. That might be the case for many terrorists, but is it for all of them? Aren’t there any active terrorists? Those who are not vulnerable at all, or who are cynical about the very ideologies they purportedly endorse.

Most importantly, the scheme implies that a productive path for prevent terrorism is to scrutinize vulnerable groups and, most specifically, individuals who sympathize with a certain ideology. I understand that vulnerability could be statistically associated with terrorism, but is it a condition for it? How often violent extremists do not fit under our definition of “vulnerable?” How often are sympathizers not terrorists? What would be the outcome of leveraging a group’s vulnerability and political attitudes to identify terrorism? Wouldn’t we simply run the risk of add burden to these vulnerabilities and attitudes?

This scheme looks to me as an under simplification of a complex issue. It implies conditionality out of possibly theoretical associations. I would be very careful here, particularly because this kind of rationale shares the outline of an ecological fallacy. Ultimately you did not convince of the value of using neighbors for predicting violent extremists. It is true that you show an association, but that association is nowhere near as strong or deterministic to warrant a targeted intervention. In fact, the types of targeted interventions that your framework seems to endorse could very well increase the deprivation and marginalization of the communities you investigate.

To be clear, I am not arguing that these issues are major flaws of your study. Of course it can be productive to explore the links between neighborhoods and extremists, but how? Under which circumstances? You tough on those subjects, but you offer no depth or response.

a. You had a range of other statements which, I felt, were overstated. I listed some in my specific comments.

3. I had some methodological concerns, most of which were minor.

a. (Minor concern) You use LSOAs as a proxy for neighborhoods (or communities). Is that reasonable? I am fine with that decision, though I feel you should defend it a little.

b. (Minor concern) There are time mismatches between your data sources, which could introduce issues given the time-sensitivity of the issues you explore. Data on deprivation is from 2012-2013, on hate crimes is from 2015-2021. That could introduce bias, and the authors should make it clear if they find this bias is negligent.

c. (Major concern) Throughout your front end you speak about local deprivation, which read to me as poverty or marginalization. However, you describe your measure as one about relative deprivation. Relative and absolute deprivation are very different measures and concepts, with very different implications for your findings and results. In fact, from reading your Supporting Information you seemed to include some measures of relative deprivation (i.e., inequality), and some of absolute deprivation (i.e., poverty). That distinction should be clear, and should be in the text itself, as opposed to a SI. It is one thing for a neighborhood to be very poor and isolated, and it is another thing for a neighborhood to be very unequal, meaning it has both rich and poor individuals (relative to one-another). I suggest you clarify what you mean by deprivation.

d. (Minor) The fond size of your figure 2 is too small. The figure was hard to read, and hard to compare across indicators.

e. (Minor) Unclear to me what was the benefit of splitting your sample and to use estimates from earlier years to predict future years. I recommend you clarify the benefit, or that you avoid that part of your analysis. Why not simply use all cases to develop your model?

f. (Minor) In Figure 3 it was unclear why you separated Income Deprivation from Loving environment and education deprivation. I assume it was to avoid clutter in the figure. If that was the case, I would point out that, in my view, have population data instead of data about a representative sample from which you are trying the extract inferences. In other words, your coefficients are not sample estimates of an unknown population parameter, but instead reflect the population parameters themselves. Though your parameters are still subject to several kinds of errors, they are not subject to sampling error, which is the only kind of error that is informed by the p-value. Therefore, your p-values and significances carry relatively little importance. For this reason, consider removing the confidence intervals from Figure 3.

SPECIFIC COMMENTS:

4. I appreciated the simplicity and objectivity of your research question, which makes it appealing to the broader scientific audient of PLOS ONE.

a. In fact, all aspects of your study are clear and parsimonious, including your analysis and text. Overall a very enjoyable article to read.

5. (p. 3) About the passage: “We provide a first study to systematically examine this relationship at the level of neighborhoods,” I suggest you rephrase. I am not entirely sure what you mean by “systematic” here, but it is very possible that your study is not the first—which does not take from your merit.

6. (p. 4) About: “We believe that a similar process also applies,” you include citations, which suggests the argument is not originally yours. I suggest you either clarify or rephrase to “A similar process could also apply.”

7. (p. 4) About: “We posit that the local neighborhood is the main reference point for individuals’ comparisons, shaping their perceptions and beliefs about outcomes and distributions.” Why not their street? Or their family? I understand that neighborhoods are important, but are you confident that they are the “most” important? That is a strong statement. Consider toning it down, (e.g., “is a key reference) or supporting it more in terms of it being the “most important.”

8. (p. 13) Which “theoretical mechanism” are you referring to specifically? Unclear. Your discussion and conclusion is generally underdeveloped.

Reviewer #3: I enjoyed reading this paper on neighbourhood deprivation and hate crime. I am convinced by the central argument and I think it should be published. I have just some comments that might change the analysis.

First though, I note that the authors have NOT made data and code available at time of submission. This means I have to answer NO to the data availability question above. Also, it makes for less good peer review, since I can't scrutinise the code or run the analyses in writing my report. I would strongly urge the authors to always prepare the data and code archive to go along with the manuscript for evaluation. In my discipline, and in many journals, this would actually be a condition of submission. If the authors are worried about precedence, they should publish a preprint version at the same time.

Now, for my main comment, which is that I don't understand the status of the three separate deprivation measures. The authors do not report the intercorrelations between these. If the intercorrelations are very high, and the authors consider them multiple measures of the same thing, then they should just average them. (Or use the overall IMD score, which effectively does this). If on the other hand they consider them to be potentially important independent components, then they should also evaluate models containing more than one of them at a time (e.g. in table 1). The way I would do this is to run AIC-based model selection (such as R package MuMIn) on the set of the covariate-only model, the null model, then all models including the covariates plus every combination of the three deprivation measures. It may be that the one with just the income deprivation measure wins, but this would be interesting and also worthy of substantive discussion (i.e. income is the prime factor in the kind of deprivation that matters for hate crime).

More generally, reporting AIC and AIC change is a much better approach than comparing R2 to show that variables improve model fit.

Specific comments

p. 5. "The individuals who carry out attacks are necessarily relatively more deprived, even if their local community is deprived." - is this sentence missing a 'not'?

p. 6. "Hate crimes differ from other criminal offences in that they express a number of socio-political objectives by...." Do the authors have any concerns about classification bias by police? i.e. that the same offence committed in a deprived neighbourhood is more likely to be classified as a hate crime than when committed in an affluent neighbourhood? (there could be a host of reasons why this is true, including the ethnic composition of the people there, etc.). Is it worth discussing this possibility, or at least saying something about how hate crime status is determined. It must be a judgement call. The authors already consider the possibility that different police forces might classify differently, which suggests a degree of classification latitude that could cause endogeneity problems.

p. 7. "We use the indices for deprivation in income, education, and environment as complimentary measures of local deprivation" - How well correlated are these measures? Are they so well correlated that they should just be considered as the same information, or are you trying to argue that they capture something distinct? Their inter-correlations should be reported.

p. 9. Table 1. I would be much happier with the analysis if the authors reported AIC instead of R2. The critical question is how much AIC goes down by inclusion of deprivation indices.

Also, as above, I fail to see the logic of models in which one at a time of the three deprivation measures (but not the others) is included. The authors should establish the intercorrelations between the three deprivation measures a priori. If this is very high, their argument would be better served by making one measure out of the three. (I believe in fact that there is an overall IMD measure already provided). If it is not very high and they make a unique contribution, then they need to try models including more than one of the three. You could also using AIC-based model selection to create the optimally predictive model (e.g. using the MuMIn package in R). This would give a sense of which is most important, income, education or environment.

Figure 6. I like this figure a lot, but I wonder if it would be more legible with two side-by-side panels, one for the without deprivation model, and one for the with deprivation. And, it would be helpful to have the line of best fit as well as the y=x line in each case; it should be appreciably flatter without deprivation. Are all three deprivation measures used for the with-deprivation predictions here?

p. 11. I heartily agree with the authors' suggestion that 'upstream' relief of deprivation is likely to be effective at reducing hate crime (among other things). But, given the scatter in figure 6, it is a very blunt instrument. That is, if your objective were ONLY to reduce hate crime, levelling up would probably not be a cost-effective way of doing it, since the relationship between deprivation and hate crime is messy. The way I tend to think about these things is that we should be reducing deprivation for many reasons; reducing hate crime is, plausibly, one of the many benefits, but not by itself a justification for doing so.

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: Yes: Nik Stoop

Reviewer #2: No

Reviewer #3: No

**********

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

 10.1371/journal.pone.0289423.r002 Author response to Decision Letter 0 Submission Version 1

14 Jul 2023

[Note: We have also uploaded a revision memo as a separate file with the manuscript. We reproduce the text below, but would encourage you to look at the file since some formating does not transfer correctly as plain text.]

Revision Memo: PONE-D-23-08904 "Local deprivation predicts right-wing hate crime in England"

We refer to your email of 12 June regarding our submission PONE-D-23-08904 "Local deprivation predicts right-wing hate crime in England". We are very grateful for the invitation to resubmit a revised version, and we appreciate the efforts of the editor and reviewers for engaging so thoroughly with the manuscript.

We respond below to the main concerns raised, ordered by topic, with our responses and comments in italics. The reviewers also made a number of smaller comments and suggestions that we have tried to incorporate to the best of our ability in the revisions. We believe the revised version has been notably strengthened as a result of the revisions made in response to the comments.

Please do not hesitate to contact us if we can help with any additional information.

Sincerely yours,

The authors

A. Theory

A1) Clarification of relationship between community deprivation and attacks

R1: “If individuals indeed strongly identify with their local community, it is unclear why one would expect that observing deprivation within one’s own community would lead to an increase in far-right hate crime within that same community. …. Currently, this assumption is not explicitly discussed …. I believe the paper would benefit from making this assumption more explicit and from strengthening the theoretical arguments and references to empirical work to back it up”. On a related issue, R2 warns against “framing individual-level research as the ‘wrong unit of analysis’ … and [how] these kinds of dichotomies can often be a fallacy

We have expanded our discussion of why we would expect to find a relationship between local deprivation and attacks. In brief, we argue that hate crimes typically target specific communities or individuals based on their characteristics. There is empirical evidence demonstrating that perpetrators of hate crimes tend to carry out violence within their local neighborhood. There are a number of reasons why perpetrators often seek out victims within their immediate vicinity, notably 1) it is easier to identify targets they hold prejudice against in the local community, 2) access, planning and execution is typically easier for local targets, 3) local targets often have symbolic significance for both perpetrators and targeted communities, and 4) local hate crimes will often receive more attention in the immediate community, which in turn can be helpful for generating fear, exerting control, and recruiting to the cause. We agree with R2 that it is unhelpful to draw a false dichotomy between individual level analysis and other levels of aggregation. We have revised the text to emphasize more clearly the added and complementary value of local level analysis.

A2) Relationship to relative and absolute deprivation.

R2 asked us to clarify the relationship between relative and absolute deprivation in the theoretical framework. We appreciate that our previous discussion may have been confusing on this point, and we have clarified our conceptualization and actual measures of deprivation in the manuscript. We emphasize that what we think drives the key relevant comparisons here are differences between neighbourhoods, not within community differences.

A3) Discussion of UK prevent scheme

R2 notes that … “This scheme looks to me as an under simplification of a complex issue … the scheme implies that a productive path for prevent terrorism is to scrutinize vulnerable groups and, most specifically, individuals who sympathize with a certain ideology. … What would be the outcome of leveraging a group’s vulnerability and political attitudes to identify terrorism? Wouldn’t we simply run the risk of add burden to these vulnerabilities and attitudes?”

We appreciate the comments on the UK Prevent scheme. We also share many of the concerns expressed over the scheme, and we try to be more upfront about this in the manuscript. At the same time, this scheme has clearly guided policy and discussion, and as such it is useful to discuss our approach in comparison to this. We did not intend to apply the term “vulnerable” as a value judgement or to underplay individual agency, rather we simply seek to examine if features discussed in this context can help identify local contexts with a higher probability of radicalization and observed attacks. We primarily discuss the potential influence of deprivation in our manuscript rather than measures against potentially radicalized individuals or communities. We have revised the discussion so as to be much clearer on this and avoid potential misunderstanding.

A4) Diffusion and network effects

The editor noted that “it would be valuable … to explore the possibility that the local deprivation effect can be understood within the context of micro-dynamic models that incorporate social networks as a mechanism for the diffusion of civil unrest behavior … [discussing and referencing] Braha (2012) … would provide additional insight into the dynamics of local deprivation and hate crime behavior.

We entirely agree that networks and diffusion are important components in accounting for observed unrest and hate crime. We have added a brief discussion and referenced work on the diffusion of political violence. However, a full examination of this would require a much more complex analyses of the role of “local fundamentals” vs “position”, and ideally consider detailed analyses of individual connections using actual data rather than proxies based on geographical distance. We also note that very strong diffusion or emulation effects could plausibly attenuate the direct local relationships, and this appears to be the case in many studies of organized political violence.

A5) Time ordering in exposure and response

R1 and R2 noted that gap in time between deprivation indices for 2012-2013 and hate crimes committed 2015-2019. “… would [it] be possible to say something about the timing between witnessing deprivation and committing hate crimes? E.g., is there any theoretical work suggesting what time frame matters, how long one should be exposed to deprivation before moving to committing hate crimes?” (R1)

We use prior deprivation data since we want to ensure that the deprivation data is clearly prior and is not likely to arise from as a result of or influenced by our outcome. There is little direct guidance in previous work on the likely timing of exposure and potential impact on attacks, but we believe the lag length in our case is a reasonable length to reflect the impact of past exposure to deprivation. In practice, there exists considerable persistence in deprivation over time, and this is supported by the data on deprivation published by Ministry of Housing (see for example Indices of Deprivation: 2019-2015 at https://dclgapps.communities.gov.uk/imd/iod_index.html). We have revised our discussion so as to make clearer the assumptions guiding our analysis, the limits related to actually considering sensitivity to timing in the current data and our analysis, and the value of future more detailed research on this.

A6) Conditional effects

R1 notes that “The extent to which individuals identify with their local neighborhood [might vary by] … ethnic diversity. … Would there be a way to leverage the heterogeneity in this variable (or other variables the authors may think of), e.g., looking at an interaction with the deprivation indices?”

We think it is plausible that such heterogeneity may be present in the observed data, and we have conducted analyses where we interact ethnic diversity, proxied as the local population share with BAME background, with our indicators of deprivation. We find that ethnic diversity does not substantially moderate the effects of living environment deprivation and education deprivation, but that high levels of diversity may mitigate the effect of income deprivation. However, we ultimately feel that extending the analysis further along these lines is somewhat removed from our main focus in the manuscript, and we thus report in an appendix without an extended discussion in the main manuscript text. We think a fuller investigation of this would amount to a different project and better pursued in a separate manuscript. We also note that measuring conditional interactions or moderators reliably demands a great deal more of the data than main effects, and could benefit from more extended data than we have at hand now.

B. Data and empirical analysis

B1) Lagged dependent variables and autoregressive trends

The editor noted that “to further enhance its robustness, I would suggest that the authors incorporate lagged dependent variables in the regression model …[to] account for the likelihood that previous instances of hate crimes contribute to subsequent hate crimes, potentially overshadowing the effect of local deprivation”.

We appreciate the point made, and we entirely agree that hate crime data are likely to show considerable persistence and autoregressive trends. Including a lagged dependent variable would be natural if we had panel data, but unfortunately, we do not – the statistics for hate crimes that we rely on, and the time gaps between releases of the Indicators of Deprivation, do not allow for a time-series setup. Therefore, we are essentially only able to conduct a cross-sectional analysis, and it is not possible to include a lagged dependent variable in the regression model. We now explicitly clarify this in the revised version of the manuscript. We also do not propose to measure feature/variable importance by contribution to fit in-sample, which is prone to overfitting, but rather contribution to out-of-sample prediction (see B3 below).

B2) Measures and specific components

R1 notes that the income deprivation index [includes] ‘Asylum seekers in England in receipt of subsistence support, accommodation support, or both’. I wonder to what extent this indicator is driving the estimated coefficients on the income deprivation index … It would be interesting to see a robustness check in which this indicator is dropped from the income deprivation index (and perhaps included as a covariate).”

We have looked at the available data and the Home Office statistics to verify if we could find any plausible proxies. There is unfortunately no publicly available sub-national data, so we are unable to probe how sensitive the findings may be to this. We believe it is likely that the IoD obtained these data directly from the Home Office at LSOA level. The only way which we might be able to obtain these data would be to submit a FOI request, which may or may not be successful. This is not feasible within the deadline for the resubmission, and we feel that this goes beyond the main priorities for the resubmission. The fact that we consider alternative indicators for deprivation other than income, which do not include asylum seekers in receipt of support, makes it unlikely that this component alone would dominate the results. However, we now acknowledge this limitation in the main text and explicitly flag this as an area for future study.

On a similar note, R3 argues that the status of the deprivation measures must be clarified, and that we should “consider … average[ing] them … or evaluate models containing more than one of them at a time”

We now report the correlations in the Supplementary Appendix. We understand the case for trying to aggregate the information into a combined deprivation index, but to do this systematically should ideally be based on the raw inputs of the indices, rather than averaging across aggregate indices with different scales/metrics. Ultimately the regression models are all in-sample, and they are intended to evaluate the conditional associations/causal effect approximations in a simple manner. For this particular purpose, including all deprivation measures in one model would make the interpretation more difficult. For prediction, on the other hand, we actually use a random forest model that includes all the deprivation scores in the same models for computing the predictions of future levels of hate crime. We have now clarified this issue in the discussion in the manuscript.

B3) Role of prediction in model evaluation

Rs 1 and 2 raised some question about the value of splitting the data, emphasis on prediction, and added value of the model with deprivation. We have now revised the text so as to clarify why cross validation is an important check on overfitting to observed data. It also helps emulate the potential to use measures for out of sample forecasts. We have also clarified that even if the variance accounted for may seem small in absolute terms, there is a near double increase for the deprivation model. We think this demonstrates the added value, although we try to be explicit on not overstating the absolute predictive power.

B4) Clarify data

Rs1, 2, and 3 asked for clarification for many aspects of the data, including the timing of the covariates, and the size/number of police force districts used for fixed effects and relationship to local LSOAs, and the district reporting no hate crimes. We have revised our discussion so as to provide more details explicitly on the issues raised. We also discuss in more detail the potential bias from underreporting of hate crimes (which under plausible condition would likely attenuate the reported relationship with deprivation). The Home Office informed us about the police districts that did not submit any reports in specific years (where the “zero hate crimes” is the result of a reporting issue), and these districts were removed from all analyses to begin with. In addition, we report results with all units with zero hate crimes removed from the analyses in the Supplementary Appendix, referenced in the main manuscript.

B4) Clarify the analyses reported

Rs 1,2, and 3 called for more details in explanatory notes (e.g., types of standard errors), and we have implemented these in the revised manuscript. We also report AICs in addition to R2, as requested.

B5) Figures/image resolution

R3 suggested to revise Figure 6 to include two side-by-side panels, one for the without deprivation model, and one for the with deprivation, with a best fitting regression line. We have also improved the resolution of the images in the manuscript. The small font sizes are the result of downscaling in MS Word and potentially only PDF conversion. This will not be an issue in the published version of the manuscript using the separate images.

B6) Supplementary analyses

R1 suggested a robustness check relying on the ACLED data. We appreciate this suggestion, but we are unable to use ACLED, since the ACLED data for the United Kingdom are limited to the period between 2020 and 2022, cutting significantly our sample. (Moreover, ACLED has a restrictive terms of use license, which does not allow for comparisons to other data sources, see https://acleddata.com/acleddatanew/wp-content/uploads/2022/06/ACLED-Terms-of-Use-Attribution-Policy_V2_8June2022.pdf). In general, with news-based event data (like ACLED and GTD) we can expect the coverage of the events to be more limited and arguably thus inferior to our administrative data. We believe the selection problems from media reports are likely to be worse than the potential problems arising from differences in reporting by administrative agencies. We discuss in more detail how to best attempt to mitigate issues arising from underreporting.

C. Broader implications/policy

Some of the reviewers raised issues about the potential broader implications and actionable policies that our manuscript could be seen as supporting. We have tried to be careful to be modest in our discussion of policy implications, in particular since we have not studied intervention to reduce deprivation or inequality. However, we note the reductions in child poverty under New Labour as a possible example of how reforms seeking to reduce inequality could have wider impacts or dividends on violence, even if reducing violent attacks by itself is the main target of policy proposals.

D. Journal Requirements:

D1) Compliance with style requirements, including file names.

We have carefully reviewed the guidelines to ensure that the manuscript and file names follow the PLOS One guidelines.

D2) Role of funders

Please state what role the funders took in the study… Please include this amended Role of Funder statement in your cover letter; we will change the online submission form on your behalf.

We have included the suggested text in our cover letter: "The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript."

D3. Data Availability statement

We will update your Data Availability statement to reflect the information you provide in your cover letter.

We have made our data available on the Harvard Dataverse at https://doi.org/10.7910/DVN/GSRPQY.

D4) Re license/copyright for maps used for Figures 2 and 5

These images have been generated by shape files that are publicly available from the UK Office for National Statistics. They are supplied under the Open Government Licence. We have added the source and license as “Office for National Statistics licensed under the Open Government Licence v.3.0”.

D5). … include captions for your Supporting Information files at the end of your manuscript, and update any in-text citations to match accordingly.

We have now implemented this.

Submitted filename: 2023-07-14_PLOSONE_resubmission_memo.docx

 10.1371/journal.pone.0289423.r003 Decision Letter 1 Braha Dan Academic Editor 2023 Dan Braha This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Submission Version 1

19 Jul 2023

Local deprivation predicts right-wing hate crime in England

PONE-D-23-08904R1

Dear Dr. Gleditsch,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Best,

Dan Braha

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

 10.1371/journal.pone.0289423.r004 Acceptance letter Braha Dan Academic Editor 2023 Dan Braha This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

24 Jul 2023

PONE-D-23-08904R1

Local deprivation predicts right-wing hate crime in England

Dear Dr. Gleditsch:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Professor Dan Braha

Academic Editor

PLOS ONE