Immigration and the Reallocation of Work Health Risks

This paper studies the effects of immigration on the allocation of occupational physical burden and work injury risks. Using data for England and Wales from the Labour Force Survey (2003-2013), we find that, on average, immigration leads to a reallocation of UK-born workers towards jobs characterized by lower physical burden and injury risk. The results also show important differences across skill groups. Immigration reduces the average physical burden of UK-born workers with medium levels of education, but has no significant effect on those with low levels. These findings, together with the evidence that immigrants report lower injury rates than natives, suggest that the reallocation of tasks could reduce overall health care costs and the human and financial costs typically associated with workplace injuries.


Introduction
There is a large literature exploring the impacts of immigration on different factors such as labour markets, public finances, delivery of public services, housing market and criminality, among others (Dustmann et al., 2013(Dustmann et al., , 2010Dustmann and Frattini, 2014;Sá, 2015;Bell et al., 2013;Giuntella et al., 2016). However, there is less evidence about the impact of immigration on health care costs. This is an important gap in the evidence as immigrants are often blamed for high levels of health care expenditure in host countries, particularly in countries that have publicly funded health care systems (Giuntella et al., 2016). The existing evidence has mainly focused on exploring the health trajectories of immigrants and suggests that immigrants are often healthier upon arrival in the host country but that their health outcomes converge to those of natives over time (Kennedy et al., 2015). However, just a few studies have explored the impact of immigration on the health outcomes of natives (Giuntella and Mazzonna, 2015), a major factor in the determination of the overall impact of immigration on health care expenditure.
The classical model of labor demand and supply suggests that immigration has a negative effect on the wages and employment of the residents of the host country (Borjas, 2014). However, most studies have found little empirical support for this effect. Previous research suggests that this lack of evidence could be explained by differences in comparative advantage between immigrant and native workers. Immigrants have a comparative advantage in manual-intensive jobs, while native workers have an advantage in communication-intensive jobs due to better language skills. An expansion in the supply of immigrants increases the relative returns to communicationintensive jobs pushing native workers towards those jobs (Peri, 2016(Peri, , 2012D'Amuri and Peri, 2014;Ottaviano et al., 2013;Peri and Sparber, 2009). This paper contributes to this literature by exploring if these labor market adjustments lead to a reallocation of natives occupational physical burden (e.g. lifting and carrying heavy loads) and occupational health risks (i.e. injury risk) to immigrants. The analysis also tests if the effects of immigration are similar for natives with different levels of education, occupations and gender.
In order to provide this evidence we use 2003-2013 data for England and Wales for the analysis. The consequences of immigration are at the centre of the political discussion in the UK and analysis suggests that immigration was one of the key drivers of the British vote to leave the EU 2 (Vargas-Silva, 2016). According to the 2011 Census there were 7.5 million foreign-born persons living in England and Wales, corresponding to 13.4% of the population. Close to 40% of these immigrants arrived from 2004 onwards and, many of them are citizens of the new European Union (EU) member states who found jobs in the low-wage sector (Drinkwater et al., 2009). There is widespread geographic dispersion on the level and change in immigration (Figure 1). In fact, in 2011, immigrants represented over 10 percent of the population in a quarter of local authorities in England and Wales.
The increase in immigration in the UK over the last decade has been accompanied by a decrease in UK-born workers' average physical burden and injury rates ( Figure 2) and share of high-physically demanding jobs held by UK-born workers ( Figure 3). This paper explores the connection between these trends.
We exploit spatial and temporal variation in the share of immigrants residing across local authorities. To address the concern that immigration may be endogenous to labor market demand and correlated with unobserved determinants of working conditions and work health risks, we used an instrumental variable approach exploiting the correlation between immigrant inflows and historical concentration of immigrants across local authorities in England and Wales (Bell et al., 2013;Sá, 2015). Furthermore, using retrospective information on worker's occupational characteristics, we analyse the effects of immigration on occupational changes at the individual level. Examining individual labor market transitions allows controlling for individual time invariant characteristics. This exercise strengthens the causal interpretation of our results mitigating the concern that our identification strategy may be confounded by spillover effects and internal mobility (Borjas et al., 1996;Borjas, 2003).
Our results suggest that immigration pushes UK-born workers towards jobs characterized by lower physical burden and injury risk. The effects are particularly large for UK-born males with medium levels of education holding physically demanding jobs. These workers have lower search and training costs for new jobs and can take advantage of the increased demand for communication-intensive jobs induced by the inflow of immigrants. Consistent with these findings, immigration also reduces the average occupational risk for natives with medium levels of education. The reallocation of tasks, together with the evidence that immigrants report lower injury rates than natives, suggests that immigration reduces health care, productivity and financial 3 costs associated with work-related injuries in the UK. 1 This paper is organized as follows. Section 2 provides the theoretical intuition behind the analysis. Section 3 provides a discussion of the data, the empirical specification, and the identification strategy. Section 4 presents the main results of the paper. Section 5 presents the robustness checks. Concluding remarks are given in Section 6.

Theoretical framework
There are several theoretical frameworks that are relevant for our analysis. These frameworks relate to the willingness of immigrants to accept occupational risk, the discrimination of employers towards immigrants and the possible substitution of the occupational risk of natives.
Concerning the willingness of immigrants to accept occupational risk, there is a substantial literature which suggests that immigrant workers tend to be overrepresented in risky jobs (Orrenius and Zavodny, 2012;Bauer et al., 1998). Several factors might explain the overrepresentation of immigrants in risky jobs. First, immigrants may just be less risk adverse than natives. There is evidence that even within the household the less risk adverse are more likely to migrate (Dustmann et al., 2017). Second, immigrants may underestimate occupational risk because employers deliberately misinform them about it. This is more likely for those immigrants with less familiarity with the host country, low proficiency in the language and smaller ethnic networks (Dávila et al., 2011). Third, immigrants tend to have a health advantage over natives (Antecol and Bedard, 2006;Kennedy et al., 2015;Giuntella, 2017). This could give immigrants greater incentives to self-select into more strenuous jobs for which they could have a comparative advantage.
The theoretical frameworks dealing with the level and type of complementarity between immigrants and natives can also provide important insights. In the classical model of labor demand and supply, immigrants and natives are assumed to be perfect substitutes which implies that immigration should have a negative effect on the wage and employment levels of natives (Borjas, 2014). However, most research has found little empirical support for this effect. Some previous studies suggest that this lack of evidence could be explained by differences in comparative ad-1 A recent report from the UK Health and Safety Executive suggests that health care costs are only a small portion of the overall costs associated with work-related injuries. See http://www.hse.gov.uk/statistics/pdf/cost-tobritain.pdf 4 vantage between immigrant and native workers. In this setting, immigrants have a comparative advantage in manual-intensive jobs, while native workers have an advantage in communicationintensive jobs due to better language skills. An expansion in the supply of immigrant workers increases the relative returns to communication-intensive jobs pushing native workers towards those jobs (Peri, 2016(Peri, , 2012D'Amuri and Peri, 2014;Ottaviano et al., 2013;Peri and Sparber, 2009). We would expect an overall positive correlation between the manual content of a job and its risk of injury. This could be one of the channels by which immigration leads to a reallocation of work risk from natives to immigrants. However, this correlation is not one to one. Two similar jobs in terms of their manual content can have very different injury rates. For example, in our analysis we find that the injury rate risk for medical doctors is much lower than for veterinarians. 2 3 Data and empirical specification

Data
The main dataset is the special license version of the LFS from 2003 to 2013. The special license version of the LFS is only available since 2003. The sample is limited to employed individuals between 20 and 59 years of age. The information on country of birth and location is used to construct an indicator of the immigrant (i.e. foreign-born) share of the population by local authority.
The ISCO-88 classification and the General Index for Job Demands in Occupations constructed by Kroll (2011) is used to a create variable (1 to 10 metric) for the average physical burden of a given job. The factors determining the physical burden of a job include considerations such as: having to lift and/or carry heavy loads, bend, kneel or lye, working in the presence of smoke, dust, gases, vapours, working in cold, heat, wet conditions, etc. We also created two indicators for jobs with high physical burden (above median) and very high physical burden (highest quartile).
Workers are also classified according to occupations (1-digit) and blue-and white-collar status following standard OECD classifications.
The special license of the LFS is combined with the standard version to measure work-related 2 See Table A.1 for further details.
risks. There is no information on work-related injuries in the special license of the LFS. 3 This information is available in the standard version, but this version does not include information on the individual's local authority of residence. In order to analyse the relation between immigration and actual injury rates, we constructed an index of occupational risk based on injury rates by occupation and year. Injury rates are calculated as the share of individuals in a given occupation which reported accidents resulting in injury at work or in the course of work in the last 12 months. Those occupations with an injury rate above the median are categorised as risky.
Examples of occupations with high/low physical burden and injury rate are reported in Table   A.1.
We also explore the impact of immigration on natives with different levels of education.
Natives are divided in three educational groups. The "high education" group refers to those with a university degree or equivalent. The "medium education" group refers to those with a high school degree or equivalent, including GCE, A-level and GCSE grades A*-C. Finally, the "low education" category refers to those natives with no qualifications or qualifications below the ones included in other categories.
Descriptive statistics for the outcomes and covariates are reported in Table 1. On average immigrants are more likely to work in jobs with a higher physical burden, but the injury rate is similar across the two groups. Immigrants are also younger than natives and more likely to be concentrated in the higher or lower educational groups..
We also present evidence exploiting retrospective information on worker's occupational characteristics. Since 2003 the first quarter of the standard LFS collects information on respondents' occupation in the previous year. This allows us to analyse the effects of immigration on occupational changes at the individual level. By removing any individual time invariant characteristics and following the worker wherever he/she moves we can address the concern about the potential spillovers on other labor markets due to spatial arbitrage (Borjas, 2003). Table 2 reports immigrant-native differences in the likelihood of working in physically intensive jobs (1 to 10 metric) by gender. All estimates include standard demographic controls (a quartic in age, marital status, and number of children), year and local authority fixed effects.
Previous studies suggest that as immigrants are often positively selected on health they have 3 There is no firm level publicly available information on work related injuries in the UK. 6 incentives to self-select into more strenuous jobs (Giuntella and Mazzonna, 2015) and are more likely to hold risky jobs (Orrenius and Zavodny, 2012). The estimates in Table 2 support this dynamic. Immigrants are significantly more likely to hold jobs characterised by higher physical burden (column 1). With respect to the mean, immigrants are 11% more likely to hold jobs in the upper quartile of the physical burden index distribution (physical burden > 7, see column 3). The coefficients are smaller, but the differences remain significant when controlling for sociodemographic characteristics (columns 2 and 4). With respect to the mean, immigrants are 5% more likely to hold high physical burden jobs than natives with similar characteristics.
The native-immigrant difference is also present for women. With respect to the mean of the dependent variable, foreign born women are 53% more likely to be employed in high physically intensive occupations. However, it is worth noting that in general women are less likely to work in physically demanding jobs (only 12% of native women work in high physical demanding jobs vs. 30% of native men). For this reason, in our analysis we focus primarily on native men. Table 3 shows differences in occupational risk and individual likelihood of experiencing an injury between natives and immigrants. The sample is smaller as the information on occupational injury rate is not available for all the occupations in every year. 4 In the first two columns, we estimate the native-immigrants difference in occupational risk (continuous variable and above median indicator). Given the higher share of immigrants in physical demanding jobs (see Table   1), it is unsurprising that we find that immigrants are 10% more likely to work in occupations with a higher injury risk (column 2). At the same time, using information on self-reported injuries, we show that immigrants are 5% less likely to report an injury (column 3) and that this result holds when we compare immigrants and natives in the same occupational category (column 4). 5

Empirical Specification
To identify the effect of immigration on job physical burden and occupational risk we exploit variation over time in the share of immigrants living in each local authority between 2003 and 2013. The estimated empirical model is as follows: where Y ilt is a metric of job physical burden or occupational risk of individual i, in local authority l at time t; S lt is the share of immigrants in local authority l at time t; X ilt is a vector of individual characteristics; Z lt is a vector of time-varying characteristics at the local authority level (share of White, Asian, and Black population, share of individuals with low, medium, and high education, share of female population, log of average gross income, local-authority employment rate, and share of individuals claiming unemployment benefits) and µ l and η t are local authority and year fixed effects, respectively; and ilt captures the residual variation.
Immigrants might endogenously cluster in areas with better economic conditions and have an impact on natives' internal mobility (e.g., Borjas et al., 1996;Borjas, 2003). We adopt the traditional "shift share" instrumental variable approach (Altonji and Card, 1991;Card, 2001;Bell et al., 2013;Sá, 2015) to address this endogeneity. This approach exploits the fact that immigrants tend to locate in areas with higher densities of individuals from their same country of origin.
The annual national inflow of immigrants from each country across local authorities is distributed according to the concentration of foreign-born individuals in the 1991 UK Census, reducing the bias from endogeneity.
Define F ct as the total population of immigrants from country c residing in England and Wales in year t and s cl1991 as the share of that population residing in local authority l in year 1991. We then constructF clt , the imputed population from country c in local authority l in year t, as follows:F and the imputed total share of immigrantsŜ lt in local authority l in year t will be: where P l,1991 is the total population in local authority l in 1991. Thus, the predicted number of new immigrants from a given country c in year t in local authority l is obtained by redistributing 8 the national inflow of immigrants from country c based on the distribution of immigrants across local authorities in 1991. Adding data for all countries of origin, it is possible to obtain a measure of the predicted total immigrant inflow in each local authority and use it as an instrument for the actual share of immigrants. We consider nine foreign regions of origin: Africa, Americas and Caribbean, Bangladesh and Pakistan, India, Ireland, EU-15, Poland, and other countries.
One potential threat to the validity of this approach is that the instrument cannot credibly address the resulting endogeneity problem if the local economic shocks that attracted immigrants persist over time. However, this problem is substantially mitigated by including local authority fixed effects and by controlling for time-varying characteristics at the local authority. Thus, it is reasonable to assume that past immigrant concentrations are not correlated with current unobserved local shocks that might be correlated with job physical burden and occupational risk. In other words, the exclusion restriction holds under the assumption that-after controlling for local authority and year fixed effects, and local authority time-varying characteristics-the imputed inflow of immigrants is orthogonal to the local specific shocks and trends in labor market conditions.
We test the robustness of our results to a change in the geographical unit using a higher level of aggregation to address the concern that our results may be biased by the effects of immigration on native internal mobility (Borjas et al., 1996). We also show that our results are robust to the inclusion of local authority specific time trends. Finally, a placebo test is conducted to analyse the effects of immigration on past trends in occupational physical burden and injury risk and find there is no evidence of significant correlations. Table 4 reports on the relationship between immigration and the physical burden associated with a given occupation. In Panel A, we restrict the analysis to UK-born male workers. The OLS estimates show that there is a negative association between the share of immigrants living in a local authority and average physical burden. A 10 percentage point increase in share of immigrants in a local authority (one standard deviation) is associated with a 0.10 points decrease in average physical burden of native males (column 1, OLS). 2SLS estimates are larger than the OLS ones suggesting that immigrants tend to locate in areas where occupations are characterized by a higher physical burden. A 10 percentage point increase in share of immigrants in a local authority (one standard deviation) reduces the average physical burden of native males by 0.25 points (column 2), which corresponds to a 0.09 standard deviation. This is a reduction of 5% with respect to the mean of the dependent variable. These effects are larger when we focus on the likelihood of being employed in a highly physically intensive job. A 10 percentage point increase in share of immigrants reduces the likelihood of male natives to work in a job in the upper quartile of the physical burden distribution by a 15% effect with respect to the mean (column 4).

Physical Burden
The effects are smaller when focusing on women (Panel B). A 10 percentage point increase in share of immigrants in a local authority (one standard deviation) reduces the average physical burden of native females by 0.13 points (column 2), which corresponds to a 0.06 standard deviation. Again, these results are not surprising given the low number of native women working in these jobs. For this reason, henceforth we focus on the results on UK-born men, but we report results for UK-born women in the Appendix. Table 5 shows that the effects are largely concentrated among men with medium levels of education. 6 For male native workers with a medium level of education, a 10 percentage point increase in the share of immigrants (one standard deviation) would lead to a 0.14 standard deviations reduction in physical burden(column 3).
We also find some evidence of a reduction in physical burden (0.06 standard deviations) for men with high levels of education (column 2). On the other hand, there is no effect for those with low levels of education.
These results indicate that immigration reduces the physical burden of those with medium level of education who may be overqualified for a physically intensive job. Individual with low re-training costs are those who are more likely to be pushed towards less physically intensive jobs as a response to immigration (Orrenius and Zavodny, 2010).
This intuition is confirmed by the evidence reported in Table 6, which considers information 6 The heterogeneity of results by educational groups is consistent with recent findings on the effects of immigration on wages showing that the impact of immigration can be different along the wage distribution (Dustmann et al., 2013). Consistent with previous literature, we find no evidence of significant effects on wages (Table A.4). While not precisely estimated the coefficient on wages is negative and (larger) in absolute value when focusing on the low-skilled who are more likely to suffer immigrant competition.
on previous year occupation (available for the second quarter of each year in the LFS). In this Table we compare occupation one year ago with current occupation and determined whether the current job has a higher or lower physical burden. 7 Panel A examines the effect of immigration on the likelihood that a native man will switch to a less physically intensive job. As expected there is a large and statistically significant effect among individuals with medium levels of education previously working in blue collar jobs (column 5). A 10 percentage point increase in the share of immigrants increases the likelihood of moving to an occupation with lower physical burden by a 0.1 standard deviation (approximately a 30% effect with respect to the mean). On the contrary, the same change in the immigrant share would reduce the likelihood of moving to a less physically intensive job by a 0.09 standard deviation (a 40% reduction with respect to the mean of the dependent variable) for those with low levels of education. Panel B reports similar effects when we use the absolute change in the physical burden measure between the previous and current year as the dependent variable.

Occupational Risk
We now turn to investigate whether the reallocation of physical burden induced by immigration affects occupational risk. Table 7 shows that an increase in the share of immigrants living in a local authority is associated with a reduction in the likelihood of being employed in a riskier occupation. A 10 percentage point increase in the share of immigrants is associated with a 0.5 standard deviation reduction in the likelihood of native men working in an occupation with an injury rate higher than the median (a 40% effect with respect to the sample mean). Again, the effect is only significant for those with medium levels of education. 8

Welfare Implications
The Health and Safety Executive (HSE) statistics suggests that there were approximately 629,000 non-fatal injuries in the UK during 2014-2015. The HSE estimate the average cost of a non-fatal injury to be around GBP 7,500. 9 7 Note that those who leave employment are and not in the sample and this could lead to some selection issues. 8 As shown in the appendix, we obtained similar results for the impact of immigration on the likelihood of working in occupations in the highest tercile of injury risk (Panel A, Table A.5). 9 http://www.hse.gov.uk/statistics/pdf/cost-to-britain.pdf Immigration reduced the average physical burden and injury risk among UK-born workers and immigrants exhibit a lower likelihood of reporting any injury in a given occupation (see Table 3). These two factors suggest that immigration could lead to a reduction in the overall injury rate. We estimate that, on average, a 10 percentage point increase in the share immigration in a local authority reduces the total injury rate by 1% with respect to the mean (using the full sample including both natives and immigrants). Using the HSE statistics we can conduct a backof-the envelope calculation of the reduction in costs associated with the lower injury rate which results from a 10 percentage point increase in immigration. A 1% reduction in non-fatal injuries (6,290 fewer injuries per year) is the equivalent to a reduction in costs of roughly GBP 47,175,000.
It is worth noting that health care costs are only a minor component of total costs to society of workplace injuries. According to the Health and Safety Executive estimates 10 in 2014/15 the major costs to employers arise from productivity costs associated with workplace injuries. 11 Another key aspect for the welfare implication is the change in working conditions of immigrants with respect to the pre-migration situation. Is there a Pareto-improvement? It is possible for immigrants to have lower injury rates in the UK than in their home countries, even if they work at riskier jobs than UK natives. This would imply an improvement in welfare for both natives and immigrants as a result of immigration. To gauge whether this is the case we use the 2007 European Labour Force Survey which contains the Work Related Accidents, Health Problems and Hazardous Exposure ad-hoc module. We compare the likelihood of reporting non-fatal injuries in the UK and in the Eastern Europeanseveral new EU member states which represented the mainkey countries of immigration to the UK in the period under study 12 . As shown in Table   8, we find that the likelihood of reporting any injury is lower in the UK (-60% with respect to the mean) than in the new EU member states (columns 1). This difference remains significant (-20% with respect to the mean) when including occupation fixed effects (see column 2). In columns 3 and 4 we focus on the differences in the likelihood of injuries between the UK and Poland which is by far the major country of origin of immigrants for the period considered in the paper (Rienzo and Vargas-Silva, 2012). This suggests that immigration could lead to "pareto-improvement" in working conditions. 10 http://www.hse.gov.uk/statistics/pdf/cost-to-britain.pdf 11 http://www.hse.gov.uk/statistics/pdf/cost-to-britain.pdf 12 Bulgaria, Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Romania, Slovakia and Slovenia.

Robustness Checks
To address the concern that results may be biased by the effects of immigration on internal native mobility, we check the robustness of our results to changing the geographical unit of analysis to UK regions. 13 The coefficients on physical burden (column 2, Table 9) remain substantially unchanged compared to the local authority units (columns 1 and 3). Note that all the estimates include socio-demographic controls and year fixed effects. 14 In Table 10, we conduct a placebo test to check if the results are driven by pre-existing trends affecting immigration and occupational physical burden and injury risk. As in Foged and Peri and pre-trends in our outcomes of interest. Overall, these results provide support to a causal interpretation of our main results.
Finally, since the burden associated with each occupation might be multidimensional, we also consider the psycho-social burden of a given job (Kroll, 2011). However, the results reported in 13 The LFS contains information on region of usual residence. England and Wales are divided in 17 regions: Tine and Wear, South West, Rest of Northern Region, West Midlands (Metropolitan), South York Shire, Rest of West Midlands, West Yorkshire, Greater Manchester, Rest of Yorkshire and Humberside, Merseyside, East Midlands, Rest of North West, East Anglia, and Wales. 14 The regional estimations do not include regional fixed effects as there is not enough variation when using both year and regional fixed effects.
13 Table A.6 show that there is no evidence of significant effects on psychological burden.

Conclusions
This article contributes to the literature on the labor market effects of immigration by estimating its impact on the physical burden and work-related health risk of UK-born workers in England and Wales from 2003 to 2013. The results suggest that immigration reduces the average physical burden of native workers. However, the mean effects mask important differences along the skill distribution. Immigration significantly reduces the average physical burden of native workers with high or medium levels of education and has no significant impact on those with low levels of education.
Our results are consistent with the existence of imperfect substitution between immigrant and native workers and the observation that immigrants have a comparative advantage in selfselecting into more strenuous jobs. The inflow of workers with a comparative advantage in manual tasks increases the demand for and returns to communication-intensive ones. This increase in returns leads individuals with low re-training costs (medium and high-skilled) towards jobs that are less physically intensive and involve lower injury risks.
These findings, together with the evidence that immigrants report lower injury rates than natives, suggest that the reallocation of tasks may result in fewer total injuries and lower health care and productivity costs of workplace injuries.
14 Compliance with Ethical Standards:        (2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013). All the estimates include controls for education (dummies), a quartic in age, marital status, and number of children. Local authority time-varying characteristics include share of white, asian, black population, share of individuals with low, medium, high education, log of average gross income, local-authority employment rate, share of individuals claiming unemployment benefits, and share of female population. Standard errors are clustered at the local authority level and are reported in parentheses. Note that column (1) includes 18674 observations with missing information on education.      (2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013). All the estimates include controls for education (dummies), a quartic in age, marital status, and number of children. Local authority time-varying characteristics include share of white, asian, black population, share of individuals with low, medium, high education, log of average gross income, local-authority employment rate, share of individuals claiming unemployment benefits, and share of female population. Standard errors are clustered at the regional level and are reported in parentheses.       (2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013). All the estimates include controls for education (dummies), a quartic in age, marital status, and number of children. Local authority time-varying characteristics include share of white, asian, black population, share of individuals with low, medium, high education, log of average gross income, local-authority employment rate, share of individuals claiming unemployment benefits, and share of female population. Standard errors are clustered at the local authority level and are reported in parentheses. Standard errors are clustered at the local authority level and are reported in parentheses.  (2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013). All the estimates include controls for education (dummies), a quartic in age, marital status, and number of children. Local authority time-varying characteristics include share of white, asian, black population, share of individuals with low, medium, high education, log of average gross income, local-authority employment rate, share of individuals claiming unemployment benefits, and share of female population. Standard errors are clustered at the local authority level and are reported in parentheses. Standard errors are clustered at the local authority level and are reported in parentheses.