Immigration and the use of public maternity services in England

Transcription

1 Immigration and the use of public maternity services in England George Stoye PRELIMINARY - PLEASE DO NOT CITE 29th September 2015 Abstract Immigration has a number of potentially signicant eects on the economies of host countries. The current literature largely focuses on the impact on labour market outcomes, while relatively little attention is paid to understanding the pressures on demand for public services. This paper investigates the impact of immigration on the use and quality of publicly funded maternity services in England. It shows that areas with a greater concentration of immigrants have higher demands for maternity care, with signicantly higher birth rates, greater rates of emergency caesarean sections and an increased cost per maternity patient even after controlling for local demographic and economic characteristics. Estimates suggest that immigration increased the total number of births by 16,000 in 2013/14 relative to 2003/04, at an additional annual cost of 24 million. However, despite no changes in hospital budgets to reect these rapid and unexpected population changes, the increases in demand did not result in the deterioration of the observable quality of maternity services, with no increases in 30 day emergency readmissions for mothers or children. Results also hold when examining outcomes for a subset of native mothers only. These results raise questions as to how hospitals have been able to respond to increased demand pressures without quality reductions. Possible channels include eciency improvements or the transfer of resources from other types of care. Keywords: health care; migration; public services JEL Classication: J180; I110; H400 I am grateful to Eric French, Elaine Kelly, Ian Preston, Jim Smith, Gemma Tetlow and Marcos Vera-Hernandez for helpful guidance, and to participants of an IFS work-in-progress seminar and the 2015 EEA annual conference. I thank the NHS information Centre for providing access to the Hospital Episode Statistics data under data sharing agreement IC Z4K2F. This paper has been screened to ensure no condential information is revealed. I thank the Health Foundation and the ESRC, through The Centre for the Microeconomic Analysis of Public Policy (CPP) (ES/H021221/1), for nancial support. Any errors are my own. Author aliation and contact: George Stoye (Institute for Fiscal Studies and University College London, george_s@ifs.org.uk). 1

2 1 Introduction There is an established literature on the impact of immigration on the labour market outcomes of natives (see for example: Card [1990]; Borjas [2003]; Dustmann et al. [2013]). However, migrants and natives share more than just the same labour market. In particular, all residents share a common set of publicly provided goods and services, which may include roads, policing, schools and healthcare services. Immigration therefore has the potential to alter the demand for these services through changes in both the size and the characteristics of the population [ Preston, 2014]. Despite this, little evidence exists on how immigration aects the level of demand for these services, and the impacts of these changes on the quality of services available to existing residents. This paper provides evidence of the impact of immigration on the use and quality of public maternity services in England. 1 Since 2000, the foreign born population in England doubled in size, with large increases in immigration following the expansions of the European Union in 2004 and These migrants were typically younger and more likely to be female than the existing population, and has resulted in a large increase in the population of childbearing age. Moreover, migrants have, on average, higher fertility rates than natives [Waller et al., 2012]. Maternity services are therefore where one might expect the greatest eect of immigration on public service demand in the short term. In order to identify the eect of immigration on maternity services, I use variation in immigrant shares across 315 local authorities in England between 2003/04 and 2013/14. I use hospital xed eects and measures of local economic conditions and the socio-demographic composition to control for regional dierences. Endogeneity concerns, arising from the non-random settlement of immigrants and the potential for reverse causality, are addressed through the use of instrumental variables. I use an approach commonly employed in the labour market literature [ Bartel, 1989; Altonji and Card, 1991; Dustmann et al., 2003], instrumenting current immigrant concentration with lagged values of immigrant concentration from the 1991 census in the same local authority. This paper makes use of administrative hospital records data from the Hospital Episodes Statistics. These contain the census of all publicly funded hospital admissions to hospitals operated by the English National Health Services (NHS), and provide detailed information on the procedures undergone by patients and the cost to hospitals of providing treatment. Using these administrative data, I construct several measures of the demand for maternity services. This includes the local authority birth rate, the percentage of births that are medically induced or result in an emergency cesarean section, and the average cost per maternity patient. These measures aim to capture the extent to which maternity services are used, and the cost of provision. I also construct measures of 1 Healthcare accounts for a signicant proportion of government spending in England, and made up 33% of all service spending in 2014/15 [Crawford and Stoye, 2015]. Maternity services are one of the most frequently used health services, with more than 600,000 births taking place in public hospitals each year at a cost of more than 2 billion [National Audit Oce, 2013]. Care is provided free at the point of use for all individuals who have been resident in the UK for 6 months or more. 2

3 the quality of maternity care, calculating the rate of 30 day readmission rate for newborn children and mothers [Royal College of Obstetricians and Gynaecologists, 2013; Young et al., 2013]. These indicators are combined with population data from the UK Labour Force Survey, namely the percentage of local authority populations that are born outside of the UK, to examine the impact of changes in the immigrant concentration on the use and quality of maternity services. I present two main results. First, an increase in the immigrant concentration is associated with an increased cost of providing maternity care, both as a result of a higher birth rate and the increased use of more costly procedures. My estimates indicate that a percentage point increase in the immigrant concentration is associated with an increase of 0.09 live births per 1,000 individuals. This compares to a baseline birth rate of 10.9 per 1,000 individuals in 2003/4. This suggests that immigration increased the total number of births by 16,000 in 2013/14 relative to 2003/04, at an additional annual cost of 24 million. Areas with a greater concentration of immigrants also typically experience more expensive delivery episodes, with a higher rate of emergency cesareans and marginally longer hospital stays, resulting in an increased cost per maternity patient. This may potentially be explained by delayed use of antenatal care and a rise in maternal age. Second, the estimates indicate that immigration had little eect on the overall quality of maternity care, as measured by the local authority volume of 30 day readmissions for either mothers or children. These results hold when examining only the outcomes of white British mothers, suggesting that these results are not driven by the better underlying health of migrants. This means that despite the increase in demand for maternity services, increases in immigrant concentration are not associated with an observable decrease in the quality of maternity services. However, given that hospital funding over this period did not reect unexpected population changes, it is unclear in which way hospitals responded to the increased demand for maternity services without an observable reduction in the quality of services. This paper contributes to three sets of literature. First, it complements a small literature on the use of healthcare services by immigrants. This body of work typically focuses on whether immigrants use more or less services than natives. Under the assumption that immigrants pay similar taxes to natives in the long run, this work examines whether it is likely that there are any long run impacts on the public nances through the dierential use of public healthcare. Wadsworth [2013] uses individual-level panel data to examine whether immigrants are more or less likely to visit a GP and to use hospital services in the UK and Germany. He nds relatively little dierences in service use in either country. Laroche [2000] conducts a similar exercise in Canada, and nds no signicant dierences in the service use of immigrants and natives. Using administrative data, I add to this literature by providing quantitative estimates of the impact of immigration on the demand and cost of a specic, but widely used healthcare service. Second, I provide the rst evidence of the impact of immigration on the quality of public funded healthcare services. This builds upon an existing literature that until now has focused almost exclusively upon the impact of immigration on education, and in particular, the eect 3

4 of immigration on the educational attainment of natives [Hunt, 2012; Brunello and Rocco, 2013; Geay et al., 2013]. My ndings complement this literature by extending this work to examine the impacts of immigration on another important public service. Finally, there is a large literature on the consequences of neonatal health on later life outcomes. Existing work suggests that poor standards of maternity care can have long lasting eects on education, labour market outcomes and health [Almond and Doyle, 2011; Bharadwaj et al., 2013; Fitzsimons and Vera-Hernandez, 2013]. I contribute to this literature by examining one possible driver of the quality of these services, and provide evidence on the short-run eects of congestion of maternity services on maternal and neonatal health. The remainder of the paper is structured as follows. Section 2 describes the data and the main outcomes of interest. It then sets out the empirical strategy and discusses identication issues. Section 3 presents descriptive statistics, and describes how the volume and characteristics of immigrants have changed over the period. Section 4 presents the results and discusses the potential mechanisms through which these impacts operate. Section 5 concludes. 2 Data and empirical method This paper uses data from three main sources: the British Quarterly Labour Force Survey (LFS), the UK census and the admitted care inpatient Hospital Episodes Statistics (HES). The LFS is a quarterly dataset, available at the local authority level between 2003 and It provides a large representative sample of all UK households, including approximately 120,000 individuals in each nancial quarter. It contains information on the labour market status and educational achievements of all respondents, in addition to a range of other socio-demographic variables. This includes country of birth and, in the case of respondents who are born outside of the UK, the year of arrival in the UK. The data are pooled across calendar years to produce a dataset at the nancial year level, combining quarters two, three and four from one calendar year with the rst quarter of the next calendar year. I dene an immigrant as an individual who is born outside of the UK. I use this to count the number of natives and immigrants who are resident in each local authority for each year, applying national sampling weights to scale up population size to the correct national total. I also calculate the proportion of the local authority population who are dened as immigrants in any given period. Identical information on population size and immigrant status are contained within the 1991 UK census data. This information is used to instrument contemporary immigrant concentrations in the empirical approach detailed below. These population data are augmented with a range of economic and socio-demographic controls derived from the LFS. This includes: the demographic composition of the local authority, measuring the proportion of natives and immigrants who fall into each of nine age-sex specic bands; the proportion of the local authority who completed A levels, GCSEs or O levels, or have no qualica- 4

5 tions, and; employment and unemployment rates. Local authority deprivation measures are also included, using the 2004 level of the ONS Index of Multiple Deprivation (IMD). 2 HES contain the record of all NHS-funded care in England. Data are available between 2003/04 and 2014/15. Inpatient data contain information on patients, including their age, sex, local authority of residence, the dates of admission and discharge, admission type (elective, emergency or maternity), and up to 20 ICD-10 diagnoses codes. In addition, detailed information on maternity care is included, such as the specic delivery method used and antenatal care received by mothers. This includes the week of pregnancy at which mothers rst attended an hospital outpatient appointment for an antenatal scan. I use this information to calculate a number of measures relating to maternity care, in line with clinical recommendations from the Royal College of Obstetricians and Gynaecologists [ Royal College of Obstetricians and Gynaecologists, 2013] and an existing medical literature on the quality of maternity care. These indicators serve two distinct purposes. First, I create four indicators that capture overall demand for maternity services. These indicate whether the demand for these services, and the public cost of providing them, increased as a result of immigration. Second, I calculate two measures of the short-run quality of maternity care in each local authority. These indicate whether the additional demand pressures from immigration have resulted in a deterioration in the quality of care for patients. I construct the following demand measures. First, the live birth rate in each local authority, which indicates the overall demand pressure on maternity services within the local area. I calculate birth rates in each local area by calculating the total number of women resident in each area who are admitted to an NHS hospital for a delivery episode in any year, and divide by total population. 3 Second, I calculate the percentage of labours that use particular delivery methods. This includes the percentage of labours that are medically or surgically induced, and the percentage of labours that result in an emergency cesarean section. These measures indicate demand for births that are more-resource intensive than natural births, and therefore represent a higher cost to the NHS. In addition, the overuse of these procedures can have serious health implications for children, and for mothers in both current and future pregnancies [Bragg et al., 2010]. Third, I calculate the local authority mean length of stay for both children and mothers in each nancial year. More complex births will require longer hospital spells, and this is likely to result in an increased cost of care. Finally, I calculate the average cost per maternity patient in order to explicitly test whether the cost of care has changed. 4 2 IMD scores provide an index of nine dierent dimensions of deprivation at the Lower Layer Super Output Area (LSOA). Local authority level measures are created using appropriate population weights. 3 96% of all births in England took place in NHS hospital in 2011/12 [Royal College of Obstetricians and Gynaecologists, 2013]. As a result, the vast majority of births are contained in the data. Rates are presented as per 1,000 residents. 4 Costs are calculated through the use of Healthcare Resource Groups (HRGs). Nationally xed prices are paid to hospitals for providing specic procedures, and adjusted for local variation in costs. Costs only include those associated with mothers, and directly incurred through delivery episodes and the subsequent spells in hospital 5

6 Measures of the quality of care include the percentage of newborn children who are readmitted to hospital via emergency readmission within 30 days of initial discharge and an analogous measure for maternal readmission. A higher rate of child readmission is indicative of poorer maternity care [Young et al., 2013], while increased maternal readmissions indicate variation from the usual postnatal recovery process. Poor maternity care can also signicant long-lasting implications on health and educational outcomes for children [Almond and Doyle, 2011; Fitzsimons and Vera-Hernandez, 2013]. Understanding potential factors in the variation of the quality of care is therefore important. To estimate the impact of changes in the local area immigrant concentration on the use of health services, I divide England into 315 local authorities. In the baseline specication, the birth rate of residents in local authority i, within NHS hospital area h, in year t (as measured by the number of residents admitted to hospitals for delivery episodes) is given by the following 5 : Y iht = α iht + β 1 M iht + β 2 X iht + λ h + τ t + ε iht (1) where M iht is the percentage of non-uk born residents in local authority i, in year t. The hospital and time xed eects, λ h and τ t respectively, control for permanent dierences across hospitals, and for the national time trend, in admissions for delivery episodes. 6 The coecient of interest is β 1, the eect of changing the percentage of immigrants living in the local authority, on the local authority level birth rate. The same specication is used to explore the impact on a range of dierent demand measures (use of dierent delivery methods, length of stay and cost), the characteristics of mothers (maternal age and pregnancy week of rst antenatal appointment), and the quality of maternity services (percentage of children and mothers readmitted within 30 days of discharge). The model has two endogeneity concerns. First, areas with a greater growth in admittances for the use of healthcare services, for reasons other than changes in the immigrant concentration, may nevertheless be correlated with a larger growth in immigrant concentrations. In an attempt to control for these factors, X iht includes a number of time-varying measures of local area characteristics which may be correlated with both hospital admissions and the immigrant concentration. This includes the demographic composition of the local authority, which measures the proportion of the population who fall into 18 distinct age-sex categories, the proportion of the population who attained A levels, GCSEs (or equivalent), and those who have no ocial qualications, and the local authority unemployment and employment rates. I also include a time invariant measures of before an initial discharge. They do not capture costs associated with general running of the hospitals, neonatal care, antenatal care or follow up appointments. For more detail on costing hospital episodes, see Kelly et al. [2015]. 5 Local authorities, i, are assigned to the nearest hospital area, h. Patients do, in theory, have the option to attend any NHS hospital. However, HES indicates that 90% of patients attend their local hospital. In 2003/04, 140 hospitals recorded delivery episodes. 6 There is substantial variation in maternity care practices across dierent hospitals over time. When examining the impact of immigration on the use of dierent birth processes, readmissions and length of stay, the introduction of hospital xed eects controls for permanent dierences in `best maternity practices' as viewed by dierent providers and clinicians. 6

7 the deprivation level of the local authority, using the ONS Index of Multiple Deprivation (IMD). The identifying assumption is that conditional on X iht, the other unobserved and time varying determinants of hospital admissions, ε iht, are uncorrelated with the changes in the immigrant population. A second potential source of endogeneity is that immigration is unlikely to be random across regions, and this could be correlated with the quality of health services. For example, immigrants may choose to settle in areas with better health services and better health outcomes. This would therefore cause an upward bias on the estimate of the positive eect of immigration on maternity outcomes. More generally, any negative impacts of immigration on the quantity or quality of maternity services received by residents could be hidden by the fact that immigrants are more likely to settle in areas which are economically more successful, and are therefore likely to have better health outcomes and services. I address this concern through the use of instrumental variables, using lagged values of the immigrant concentration for current concentration values. One determinant of immigrant location is the existence of immigrant communities within an area [Bartel, 1989; Altonji and Card, 1991; Dustmann et al., 2003]. As a result, immigration tends to follow historical settlement patterns. In an approach common to the migration and labour market outcomes literature, I use data from the 1991 UK census to instrument immigrant concentration in local authority i, in year t, with the immigrant concentration in the same local authority, in This assumes both that current immigrant concentrations are correlated to past immigrant concentrations, and that the lagged value of the immigration concentration is unrelated to the health service outcomes other than through its eect on the current immigrant concentration. 3 Descriptive statistics Before examining the impacts of immigration on the demand for public healthcare services, it is important to understand the changes in the size and the composition of the immigrant population in recent years. Figure 1 shows the number of immigrants resident in England in each nancial year between 2003/04 and 2013/14. There are two points of note. First, the size of the immigrant population has increased substantially. In 2003/04, there were 4.6 million immigrants living in England. In 2013/14, this population had grown to 7.4 million immigrants. This represents an increase of 61% in just ten years. Second, immigrants accounted for an increasingly larger share of the national population over time. In comparison to the population growth in the immigrant population displayed in Figure 1, the native population grew by only 4.7% over the same period. As a result, immigrants accounted for 13.9% of the national population in 2013/14, as compared to a share of 9.6% in 2003/04. It is important to note that there has been substantial geographical heterogeneity in these changes in population size and composition. Table 1 presents descriptive statistics for local authority populations in 2003/04 and 2013/14, which highlight the variation across regions. The 7

8 table makes three main points. First, it shows that the mean local authority population grew by 9.7% over the period, from 144,000 in 2003/04 to 158,000 in 2013/14. The size of these populations vary substantially in both years. These dierences also grew over time. In 2003/04, the largest local authority had a population 29 times the size of the smallest local authority. This increased to 32 times in 2013/14. Second, mean local authority immigrant concentration grew over this period, increasing from 8.14% in 2003/04 to 11.54% in 2013/14. Again, there are large dierences in the concentration between regions in any given year. In 2013/14, one region had an immigrant concentration of only 0.35%. This compares to the most concentrated local authority of 63.31%. Finally, the nal row reveals that there was substantial geographical variation in the change in immigrant concentrations over the period. The average change in immigrant concentration was 3.49 percentage points. However, these changes were far greater in magnitude in other areas. One area experienced a fall in the immigrant concentration of 7.02 percentage points, while the concentration increased by percentage points in another local authority. These trends clearly show that the size of the immigrant population in England has grown in recent years. This raises a number of further questions: what are the characteristics of these immigrants, how do these characteristics compare to natives, and how has this changed over time? If immigrants are very dierent to natives (in both observable and unobservable ways), it is likely that they will place dierent demands on health services when living in England. In order to understand the impacts that immigration may have on the demand for, and the quality of, services it is important to examine these dierences. Figure 2 shows the age distribution of the native and immigrant population in 2013/14. This highlights stark dierences in the demographic structure of these populations, with a much greater share of immigrants aged between 25 and 40, and a much smaller share at older ages, relative to natives. The dierences are even greater if examining the age distribution of immigrants who have arrived in the UK in the past ve years, as shown by Figure 3. These dierent distributions suggest that the types of health services demanded by immigrants are likely to be very dierent than the native population, both in the overall level of healthcare demanded and the specic types of services, with a shift away from procedures required by older individuals towards services such as maternity care. Table 2 compares a wider set of characteristics for natives and immigrants in 2003/04 and 2013/14. It also separately distinguishes the characteristics of recent immigrants who have moved to England in the previous ve years. This allows a comparison of characteristics both between immigrants and natives over time, and also to examine if and how the types of recent immigrants have changed. The table presents four main points. First, the mean age of natives increased over the period, from in 2003/04 to in 2013/14. In contrast, immigrant age has fallen slightly, from in 2003/04 to in 2013/14. As shown in Figure 3, recent immigrants are substantially 8

9 younger than both the typical native and pre-resident immigrant. Second, recent immigrants are more likely to be female than in the past, with the percentage of male recent immigrants falling from 49.56% in 2003/04 to 47.89% in 2013/14. As a result, the proportion of male immigrants has fallen slightly over time. This trend is dierent to the native population, where the proportion of male natives has increased slightly over time. Third, immigrant education levels have increased substantially over the period. The average age at which an immigrant left full time education increased from in 2003/04 to in 2013/14. This is signicantly above the equivalent age for the native population (17.72 in 2013/14), and is driven through a sharp increase in the the age at which recent immigrants left full time education (20.48). Existing literature suggests that there is a strong socio-demographic gradient in health [Smith, 2005]. Educational levels can therefore be viewed as a good indicator for the levels of services required by dierent populations. Finally, focusing only on immigrants, the average number of years since arrival has fallen from years in 2003/04 to in 2013/14. This is driven by the increased number of recent immigrants arriving after the expansions of the EU in 2004 and 2007, with the share of immigrants originating from Europe (32.54% in 2013/14) and the A8 countries (11.48% in 2013/14) increasing sharply over the period. 7 Taken together, these trends suggest that the immigrant population in England has grown over the past decade, and that this population is increasingly younger and better educated. This is likely to result in a change in the composition of services demanded, with the use of some services decreasing while others increase. increase as the female population of child-bearing age grows. In particular, the demand for maternity services is likely to Table 3 shows indicators of local authority demand for maternity services in 2003/4 and 2013/14. The average birth rate remained unchanged at the beginning and end of the period at births per 1,000 individuals. However, this masks substantial variation in the birth rate over the period, rising to a maximum mean of births per 1,000 individuals in 2010/11 before falling back to in 2013/14. In addition, the use of dierent delivery methods has also changed over this period. The percentage of induced labours and cesareans have both increased. The increase in induced births is in line with international experiences, but is concerning given clinical evidence of worse outcomes for mothers and children [Royal College of Obstetricians and Gynaecologists, 2013]. The incidence of cesarean section has also increased over this period. In some aspects, this may represent increased patient choice [Bragg et al., 2010]. However, the percentage of emergency cesareans has also increased, increasing from 13.27% of all births in 2003/04 to 14.64% in 2013/14. There is also substantial regional variation in the use of this procedure, with the maximum regional rate of emergency cesareans more than 2.5 times greater than the region with the lowest rate of the 7 The A8 countries refer to 8 countries which joined the European Union in They are: Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Slovakia and Slovenia. 9

10 procedure. Overuse of these procedures may have important impacts on the health of both mothers and children, with a signicant increase in the risk of complications, and in the most extreme cases death, in both current and future births [Shorten, 2007; Landon et al., 2004]. The increased incidence of these procedures also represents the more common use of resource-intensive maternity procedures and is likely to be reected in an increased cost of providing maternity services regardless of the health implications. Table 4 presents summary statistics for post-delivery outcomes for both mothers and children in 2003/04 and 2013/14. The rate of readmissions for mothers and children has increased over time. In 2003/04, the mean local authority readmission rate for mothers was 0.84%. This increased to 0.94% in 2013/14. Similarly, the readmission rate rose for children, increasing from an average of 4.07% in 2003/04 to 6.52% in 2013/14. There is also substantial regional variation, although this has decreased over time. When examining length of stay, large declines for both mothers and children are observed. Average mother length of stay decreased from 2.73 days in 2003/04 to 2.22 days in 2013/14. Similarly, child length of stay decreased from 2.37 at the start of the period to 1.90 in 2013/14. Taken together, Tables 3 and 4 suggest that there is an increasing use of hospital care by mothers, both for delivery episodes and subsequent readmissions. I now examine whether this increase in demand is related to immigration over this period, and what impact this has on the cost and quality of maternity care. 4 Results 4.1 The demand for maternity services Table 5 presents estimates of the relationship between immigrant concentration and the live birth rate (births per 1,000 individuals) between 2003/04 and 2013/14. In all cases the dependent variable is the number of live births recorded in NHS hospitals per 1,000 local authority residents. Column one regresses the local birth rate upon the percentage of the population who are immigrants, and controls only for the national time trend in the birth rate over the period. The estimated coecient on immigrant concentration is positive and signicant at the 1% level, and indicates that an increase of one percentage point in the immigrant concentration is associated with an additional 0.17 live births per 1,000 individuals in each local authority in each year. This compares to a baseline mean birth rate of 10.9 per 1,000 individuals in 2003/04. Column two introduces hospital xed eects to control for permanent dierences in birth rate across areas which are predominately served by the dierent hospitals. The estimated coecient on immigrant coecient remains signicant at the 1% level, and increases in magnitude from to This suggests that immigrants have typically settled in areas with lower birth rates. Column three includes additional controls for the demographic composition and the economic 10

11 conditions of the local authority. This substantially reduces the magnitude of the eect of the immigrant concentration, although the estimated coecient remains positive and signicant at the 1% level. As one would anticipate, areas with a larger proportion of females aged between have higher birth rates on average. In addition, areas with lower levels of education are associated with higher births rates. For example, a percentage point increase in the percentage of individuals with no qualications (instead of standard school-leaving qualications such as GCSEs or O-levels) is associated with an additional 0.17 births per 1,000 individuals. More deprived areas also have signicantly higher birth rates, as indicated by the positive coecient on the 2004 IMD score. An important distinction should be drawn between the set of results provided in columns two and three. The specication in columns two controls only for permanent dierences across hospitals. It does not account for the changes to the characteristics of the population that occur as a result of immigration. As a result, such results should be considered `unconditional' or `net' estimates of the impact of immigration on the use of these health services. In contrast, column three controls for the changes in the characteristics of the local population which are (at least in part) driven by immigration. For example, this would account for changes in demographic composition over the period as a result of the inow of (predominantly) young men and women. In this way, the estimates presented in column three abstract from the changes to the characteristics of the population, and therefore estimate the eect of changes in the immigrant concentration holding all other relevant characteristics constant. These results therefore provide answers to two distinct questions. The rst estimates (column two) indicate the overall eect on the demand for maternity services as a result of immigration. This includes both the direct impact on demand for healthcare services as a result of an increased share of the population being immigrants, and the indirect changes to the demand for healthcare services as the characteristics (such as age) of the population change. The second set of results (column three) focuses solely upon the rst of these channels, abstracting from the secondary impacts of the related demographic changes. This provides evidence on how a particular set of the population (i.e. immigrants) use maternity services, and how the overall demand for the service changes as this share of the population grows. These coecients can be used to estimate the number of additional births that took place in NHS hospitals each year as a result of immigration. Table 1 indicates an average increase in the immigrant population of 3.5 percentage points over the period, and an average population size of 158,000 over 315 local authorities. Using the unconditional estimates from column two, these estimates indicate an additional 0.78 births per 1,000 individuals, or 39,000 births across the national population, in each year. An approximate cost can also be calculated by combining these estimates with the tari received by NHS hospitals for each delivery. In 2013/14, this tari was 1,500. As a result, the estimates indicate an additional cost to NHS hospitals of 58.8 million in 2013/14. Using the estimates from column three, these estimates reduce to 13,300 births at a cost 11

12 of 19.9 million. In columns four and ve, I instrument current immigrant concentration with the value of immigrant concentration in When controlling only for permanent dierences across hospital areas in column four, the estimated coecient on immigrant concentration is and is statistically signicant at the 1% level. In column ve, I introduce additional controls for the local economic conditions and the demographic composition of the local area. The magnitude of the estimated coecient is substantially reduced at 0.091, and remains signicant at the 1% level. This suggests that an increase in the immigrant concentration by one percentage point is associated with an additional 0.09 live births per 1,000 individuals. Using the coecients from column four, I estimate that immigration led to an additional 44,600 births (at a cost of 67.0 million) in 2013/14. Using the estimates from the nal specication, these estimates reduce to 15,900 additional births (and a cost of 23.8 million). Noticeably, the magnitude of the estimates is similar in the specications where IV is used and when it is not. This suggests that the immigration location decision is likely to be exogenous to health service outcomes (once controlling for other relevant factors). The results in Table 5 indicate a strong positive relationship between immigrant concentration and the birth rate. This represents a large increase in the demand for maternity services in areas where immigration has been strong in recent years. In addition, sudden population changes may also have altered the types of procedures required by maternity patients. Any changes in the types of procedures used will have further cost implications and may also have important impacts on the health of mothers and children. Table 6 therefore examines the relationship between immigrant concentration and the use of specic delivery procedures, including the percentage of births which required an emergency cesarean (columns one and two) and the percentage of labours that were medically or surgically induced (columns three and four). The estimates in columns one and three present the unconditional specication (column four above), while columns two and four present results from the full specication (column ve above). In all cases, I instrument contemporary immigrant concentration with 1991 values. Columns one and two indicate a positive relationship between immigrant concentration and the percentage of births which resulted in an emergency cesarean. In both cases, the estimate is positive and signicant at the 1% level. In the full specication, the estimates indicate that a percentage point increase in the immigrant concentration is associated with a 0.05 percentage point increase in emergency cesareans. This compares to a baseline level of 13.27% in 2003/04. Taken together with the results in Table 5, these estimates indicate a substantial increase in the demand for maternity services, both in terms of a greater use of services, and also in the use of more resource-intensive or expensive procedures. In 2013/14, the tari paid to hospitals for an emergency cesarean was 48% higher than that of a natural delivery [Department of Health, 2013]. As a result, the average cost of birth is likely to have increased. In addition, as highlighted above, there are potential negative health implications for patients as a result of the overuse of these procedures. 12

13 In contrast, the estimates in columns three and four indicate a reduction in the percentage of medically or surgically induced births. In the nal specication, a percentage point increase in the immigrant concentration is associated with a reduction of 0.15 percentage points in the percentage of induced births. This is relative to a baseline of 16.1% in 2003/04. This would suggest a reduction in the demand for another resource-intensive procedure, and should result in a lower average cost per birth. An additional factor which determines the cost of providing maternity care is patient length of stay. Complex cases are typically associated with longer periods of recovery in hospital, and this requires a greater use of hospital resources. In 2013/14, hospitals were compensated 377 per day for mothers who stayed in hospital for more than a week after giving birth [Department of Health, 2013]. As a result, an increased length of stay for mothers and children may again indicate more resource-intensive care. Table 7 tests whether there are observable changes in the length of stay for children and mothers in areas with dierent immigrant concentrations. Columns one and two examine the impact on average length of stay for children, and indicate a positive eect. In both cases, the estimates are signicant at the 5% level. However, the impacts are relatively small in magnitude with a percentage point increase in the immigrant concentration associated with an additional 0.01 day in hospital (relative to a baseline of 2.4 days in 2003/04). Columns three and four repeat this analysis for mothers. The coecients are positive, but are again small in magnitude. Taken together, the results in Tables 5, 6 and 7 suggest that there has been an increase in the use of maternity services in areas with higher immigrant concentrations. This is reected in the higher birth rates and marginally longer length of stay. For some costly procedures, such as emergency cesarean sections, demand has increased. For others, such as induced births, demand has fallen. As a result, although total cost has increased as a result of the increase in the total number of births, the impact on the average cost per patient is ambiguous. Table 8 presents estimates of the relationship between immigrant concentration and the average cost of maternity patients. Column one controls only for the national time trend in the birth rate over the period. The estimated coecient on immigrant concentration is 0.9 and is signicant at the 1% level. This indicates that an increase of one percentage point in the immigrant concentration is associated with an additional cost of 0.90 per maternity patient. When including demographic and economic controls, the estimated coecient increases in magnitude to 1.4, and remains signicant at the 1% level. In column three, I instrument local area immigrant concentrations with their 1991 values. Relative to the estimates in column one, the magnitude of the coecient is reduced and is no longer statistically signicant, even at the 10% level. This suggests that migrants have settled in areas where the cost of maternity patients has increased regardless of immigration. However, in column four, when the instrument is again used, the results are similar to those in column two, with a positive and signicant impact of the immigrant concentration on the average cost of a 13

14 maternity patient. These estimates indicate that a percentage point increase in the immigration concentration is associated with an increased cost of 1.50 per maternity patient. These estimates can be combined with the total number of maternity patients in 2013/14 to produce additional cost estimates (relative to the cost of treating patients in 2003/04). These estimates indicate an additional public cost of 2.9 million for maternity patients in 2013/14. The above results indicate that increased immigrant concentrations are associated with an increased per patient cost of providing maternity care. Table 9 examines two mechanisms which may explain these ndings. Antenatal care is important in ensuring that pregnancy proceeds as normal, and may identify potential birth complications prior to delivery, particularly for mothers with chronic conditions or other risk factors [Carroli et al., 2001]. Reduced or delayed use of such care could therefore translate in unexpected complications at birth, leading to an higher incidence of emergency cesarean and a longer length of stay. Maternal age is also an important determinant in the probability of undergoing a cesarean section, with older mothers signicantly more likely to undergo such a procedure [Bragg et al., 2010]. Columns one and two present the estimated eect of the immigration concentration on the mean week of pregnancy in which patients rst attended an hospital outpatient appointment for antenatal care. I include demographic and economic controls in all specications, and nd a positive and statistically signicant relationship between the immigrant concentration and the timing of the rst antenatal appointment. When instrumenting concentration in the second specication, a percentage point increase in the immigrant concentration is associated with a delay of 0.18 weeks in the rst antenatal appointment, relative to the 2003/04 mean of 14.7 weeks. Columns three and four examine the impact of immigration on the mean age of mothers in the local authority. In column three, the estimated coecient on the immigrant concentration is and is signicant at the 1% level. In column four, when instrumenting for immigrant concentration, the estimated impact doubles in magnitude, and remains signicant at the 1% level. Overall, Tables 5-8 indicate that there is a positive relationship between immigration and the use of maternity publicly funded maternity services. The estimates suggest that there were an additional 16,000 births in England in 2013/14, at a cost of 24 million, as a result of immigration. In addition, there has been an increase in the average cost of around 5 for each maternity patient as a result of longer hospital stays and the use of more costly procedures such as emergency cesareans. It is important to note however that this does not indicate the overall cost of immigration on the health service. As a result of a younger, or healthier population, per capita demand for other health services are likely to have decreased. Instead, these estimates indicate a shift in the demand for particular services. I now turn to examining the impacts of these demand increases on the quality of care received by patients. 14

15 4.2 The quality of maternity services Given the large increases in the demand for maternity services, a natural extension is to ask what happens to the quality of maternity services. The example of maternity services provide a useful case study for studying the eects of immigration on the quality of public care given the availability of administrative panel data. Emergency readmissions in the month after birth are a strong indicator of adverse complications for mothers and children [Young et al., 2013] and signify a signicant deviation from the usual post-natal recovery process for mothers. Investigating potential changes in quality is also particularly important given the slow speed at which regional public funding formulas adapt to slow population change in the UK. During the period of interest, funds were allocated to regional health authorities (known as Primary Care Trusts), based on the size of the local population and their characteristics, including the demographic composition, local economic conditions and indicators of healthcare need. Crucially however, much of the allocation formula explicitly depends upon historical allocations, with any population changes only feeding into changes in funding over a number of years [ Department of Health, 2012]. As a result, resources did not immediately adjust to unexpected changes in the population due to migration, and hospitals would have had to meet additional demands with largely unchanged total resources, making reductions in the quality of services relatively likely. Table 10 displays the relationship between the immigrant concentration and the percentage of patients who are readmitted to hospital for emergency treatment within 30 days of initial discharge. Columns one and two examine readmissions for mothers, and columns three and four examine readmissions for children only. 8 In column one, where I only control for the national time trend and hospital xed eects, the estimates indicate no signicant correlation between immigrant concentration and the percentage of mothers re-admitted within 30 days of initial discharge. In the second column, I include local area demographic and economic controls. The estimated coecient on immigrant concentration now becomes negative, indicating that an increase in the immigrant concentration is associated with a decrease in emergency readmissions for mothers. This suggests that, at the very least, quality of care for mothers has not decreased as a result of additional demand on maternity services. These ndings are perhaps surprising given that one might expect to observe declines in quality (and therefore more admissions) in areas where demand for maternal care was greater. However, a lower use of hospital care in the post-natal period may also be consistent with the results found in Table 9, which indicated lower hospital service use by mothers in areas where the immigrant concentration is higher. As a result, the fall in emergency readmissions may therefore reect a genuine reduction in the need for such services, or may again represent a lower propensity to use 8 It is important to note that some complications associated with worse maternity care may be captured by the primary care system rather than through readmission to NHS hospitals. Unfortunately, the available data do not capture such episodes, and therefore visits to GPs, practice nurses or outpatient midwife appointments will not be included in the analysis. 15

16 hospital services in areas where more migrants are resident. The results look slightly dierent when examining the readmissions of newborn children. In column three, where again I control only for time and hospital xed eects, there is some evidence of a higher rate of readmission for children. The estimates indicate that a percentage point increase in the immigrant concentration is associated with an increase in emergency admission of 0.04 percentage points. This compares to a baseline of 4.07% in 2003/04. However, when controlling for the demographic structure and local economic conditions, the coecient reduces in magnitude and is no longer statistically signicant. These results therefore indicate that despite the increased demand for maternity services in areas where immigrant concentrations have increased, there is little evidence of negative impacts on the quality of maternity care received by patients. However, it might be possible to nd such results due to a dierence in the underlying health status of natives and immigrants, even if genuine congestion costs exist. For example, if immigrants are on average more healthy than natives, immigrant mothers may require less care than natives. 9 As a result, even if they receive less care as a result of congestion in maternity care, the readmission rates of immigrant mothers and children would be below the overall readmission rates for all patients. A rise in the number of immigrant patients may therefore lead to a reduction in the readmission rate, all things being equal. However, native patients with a greater underlying need for maternity care may experience worse outcomes and a higher readmission rate as a result of care congestion. These opposing eects may not be reected in changes in the average readmission rate, therefore masking any negative consequences on the quality of care in the previous analysis. This possibility means that one would like to explicitly examine the impact of immigration on the quality of care received by natives only. Unfortunately, HES data do not contain information on the immigrant status of mothers. Instead, patient ethnicity can be used to proxy for immigrant status. Table 2 showed that more than 90 per cent of natives are white. In Table 11, I therefore examine the outcomes for mothers and children whose ethnicity is recorded as `White British' in HES, and compare the results with Table 10 to examine whether the impacts of immigration on the quality of care diers across patient group. 10 Columns one and two examine readmissions for mothers, and columns three and four examine 9 The LFS contains little information on the health of respondents, making it infeasible to directly test whether there are relative dierences in migrant and native health. Wadsworth [2013] uses data from the British Household Panel Survey to test this, and nds no signicant dierences in the self-reported health status of immigrants and natives of childbearing ages. 10 HES records 16 ethnicity categories: British (White); Irish (White); Any other white background; White and Black Caribbean (Mixed); White and Black African (Mixed); White and Asian (Mixed); Any other mixed background; Indian (Asian or Asian British); Pakistani (Asian or Asian British); Bangladeshi (Asian or Asian British); Any other Asian background; Caribbean (Black or Black British); African (Black or Black British); Any other black background; Chinese; Other. White British mothers and children can therefore be distinguished (with only small measurement error) from non-british white patients in the data. However, the data do not distinguish between UK and non-uk born individuals in other ethnicity categories. In these categories, signicant proportions of patients are native born and therefore do not proxy immigrant status well. As a result, I do not report separate results for non-white British patients. 16

17 readmissions for children only. The results in columns one and two suggest that, if anything, an increased immigrant concentration is associated with a fall in maternal readmissions. Although the magnitude of the results changes slightly from the estimates in Table 10, these dierences are statistically insignicant, and suggest that the impacts are broadly similar across mothers of all immigrant status. For children there is again some evidence of worse outcomes. In column three, where I control only for time and hospital xed eects, the estimated coecient is positive and signicant at the 1% level. The magnitude of the coecient is larger than in Table 10, and indicates that a percentage point increase in the immigrant concentration is associated with a 0.09 percentage point increase in 30 day emergency readmissions of newborn children. This is equivalent to an increase of approximately 2% of the baseline readmission rate in 2003/04. However, when controlling for local demographic and economic conditions the coecient reduces in magnitude and becomes statistically insignicant even at the 10% level. Taken together, the estimates in Tables 10 and 11 indicate that despite the increased demand for maternity services in areas where immigrant concentrations have increased, there is little evidence of negative impacts on the quality of care received by children or mothers. There is some evidence that changes in the local population over the period of interest have resulted in a small increase in emergency readmissions for children. However, the impact of immigration, over and above these changes to the demographic and economic structure of the local area, had no signicant impact. When restricting attention to mothers only, the rate of readmission has, if anything, declined. These results hold when examining white British patients only, and suggest that the shift in demand for maternity care has not resulted in large, observable congestion costs. 5 Discussion This paper exploits spatial variation in the change in hospital outcomes and in the local concentration of immigrants to examine the relationship between immigration and the use of publicly funded maternity services in England. I combine administrative hospital data from the Hospital Episodes Statistics with population information from the British Labour Force Survey to estimate how the demand for, and quality of, public-funded maternity care varies with the percentage of the local population born outside of the UK. Endogeneity concerns arising from the non-random settlement of immigrants are addressed through the use of lagged immigrant concentrations as an instrument for contemporary concentrations. The main results of the paper can be summarised as follows. First, there is a signicant and positive relationship between immigrant concentration and the local authority birth rate. When including time and hospital xed eects, and using a full set of controls for the socio-demographic and economic characteristics of the local authority, the instrumental variable estimates indicate that an increase in the immigrant concentration of one percentage point is associated with an 17

18 additional 0.09 births per 1,000 residents. This compares to a baseline of 10.9 births in 2003/04, and is estimated to increase the public cost of providing maternity care by at least 24 million per year. Second, the rate of emergency cesareans is greater in areas with greater immigrant concentrations. This may have long run impacts on the health of mothers and children, and combined with longer hospital stays, represents the use of more costly procedures. Third, the average cost of a maternity patient increases with immigrant concentration. Taken together, these results suggest that areas with greater shares of immigrant population demand more, and more costly, maternity services. Finally, despite these increases in demand pressures and relatively xed hospital budgets, I nd no signicant relationship between immigrant concentration and the shortrun quality of maternity care, as measured by the immediate health outcomes for mothers and children, with little evidence of an increase in 30 day readmissions for newborn children or mothers. These ndings add to a sparse literature on the impacts of immigration on the use and quality of public services, providing new evidence on the eects on healthcare and extending a literature that has previously focused only on education. These ndings suggest that increased immigration introduces additional demand for certain types of publicly funded services and represents a shift in the composition of health services demanded by immigrants. However, it is important to bear in mind that per capita demand for wider health services is likely to fall, particularly when the age distribution of immigrants is taken into account. Further work should therefore examine the overall nancial impacts on the public health system as a result of immigration. The nding that the quality of maternity services does not deteriorate, despite the additional demand pressures on these services, is also signicant. Future work should examine the mechanisms through which the system adapts to this additional demand pressure: either through switching resources from other sectors, in which case negative consequences may exist for the quality of other types of care; eciency gains; or through pre-existing spare capacity in providing these services. Such work would provide an important understanding of how hospitals react to demand shocks and could potentially make an important contribution to the literature on healthcare production. 18

19 References Almond, D., and J. J. Doyle, `After Midnight: A Regression Discontinuity Design in Length of Postpartum Hospital Stays', American Economic Journal: Economic Policy, 3 (3), 134, Altonji, J. G., and D. Card, The Eects of Immigration on the Labor Market Outcomes of Lessskilled Natives, pp , University of Chicago Press, Bartel, A. P., `Where Do the New U.S. Immigrants Live?', Journal of Labor Economics, 7 (4), , Bharadwaj, P., K. V. Løken, and C. Neilson, `Early Life Health Interventions and Academic Achievement', American Economic Review, 103 (5), , Borjas, G. J., `The Labor Demand Curve is Downward Sloping: Reexamining the Impact of Immigration on the Labor Market', The Quarterly Journal of Economics, 118 (4), , Bragg, F., D. A. Cromwell, L. C. Edozien, I. Gurol-Urganci, T. A. Mahmood, A. Templeton, and J. H. van der Meulen, `Variation in rates of caesarean section among English NHS trusts after accounting for maternal and clinical risk: cross sectional study', BMJ, 341, Brunello, G., and L. Rocco, `The Eect of Immigration on the School Performance of Natives: Cross Country Evidence Using PISA Test Scores', Economics of Education Review, 32, , Card, D., `The Impact of Mariel Boatlift on the Miami Labor Market', Industrial and Labor Relations Review, 43 (2), , Carroli, G., C. Rooney, and J. Villar, `How eective is antenatal care in preventing maternal mortality and serious morbidity? An overview of the evidence', Pediatric and Perinatal Epidemiology, 15, 142, Crawford, R., and G. Stoye, IFS Green Budget 2015, chap. Challengs for health spending, pp , Institute for Fiscal Studies, Department of Health, Resource allocation: Weighted capitation formula, 7th edition, Department of Health: London, Department of Health, A guide to Payment by Results: National Taris, Department of Health: London, Dustmann, C., F. Fabbri, I. Preston, and J. Wadsworth, The local labour market eects of immigration in the uk, Home Oce Report,

20 Dustmann, C., T. Frattini, and I. Preston, `The Eect of Immigration Along the Distribution of Wages', Review of Economic Studies, 80 (1), , Fitzsimons, E., and M. Vera-Hernandez, `Food for Thought? Breastfeeding and Child Development', IFS Working Papers W13/31, Institute for Fiscal Studies, Geay, C., S. McNally, and S. Telhaj, `Non-native Speakers of English in the Classroom: What Are the Eects on Pupil Performance?', The Economic Journal, 123 (570), F281F307, Hunt, J., `The Impact of Immigration on the Educational Attainment of Natives', Working Paper 18047, National Bureau of Economic Research, Kelly, E., G. Stoye, and M. Vera-Hernandez, `Public hospital spending in England: Evidence from National Health Service administrative records', IFS Working Paper, W15/21, Landon, M. B., J. C. Hauth, K. J. Leveno, C. Y. Spong, S. Leindecker, M. W. Varner, A. H. Moawad, S. N. Caritis, M. Harper, R. J. Wapner, Y. Sorokin, M. Miodovnik, M. Carpenter, A. M. Peaceman, M. J. O'Sullivan, B. Sibai, O. Langer, J. M. Thorp, S. M. Ramin, B. M. Mercer, and S. G. Gabbe, `Maternal and Perinatal Outcomes Associated with a Trial of Labor after Prior Cesarean Delivery', New England Journal of Medicine, 351 (25), , pmid: , Laroche, M., `Health Status and Health Services Utilization of Canada's Immigrant and Non- Immigrant Populations', Canadian Public Policy, 26 (1), 5175, National Audit Oce, Maternity services in england: Report by the comptroller and auditor general, National Audit Oce: London, Preston, I., `The Eect of Immigration on Public Finances', The Economic Journal, 124 (580), F569F592, Royal College of Obstetricians and Gynaecologists, Patterns of Maternity Care in English NHS Hospitals, RCOG: London, Shorten, A., `Maternal and neonatal eects of caesarean section', BMJ, 335 (7628), , Smith, J. P., `Unraveling the SES-Health Connection', Labor and Demography , EconWPA, Wadsworth, J., `Mustn't Grumble: Immigration, Health and Health Service Use in the UK and Germany', Fiscal Studies, 34 (1), 5582,

21 Waller, L., A. Berrington, and J. Raymer, `Understanding recent migrant fertility in the United Kingdom', Tech. Rep. 27, ESRC Centre for Population Change Working Paper, Young, P., K. Korgenski, and K. Buci, `Early readmission of newborns in a large health care system', Pediatrics, 131 (5), ,

22 Table 1: Local authority populations, 2003/4-2013/14 Mean S.D. Min Max Total population (000s) 2003/ / ,087 Immigrant (%) concentration 2003/ / Change in concentration (Percentage points 2003/ /14) Notes: Authors' calculations using LFS data with national probability weights. Immigrants are dened as non-uk born individuals. Table 2: The characteristics of immigrants, 2003/4-2013/14 Natives Immigrants Recent Immigrants 2003/ / / / / /14 Mean age Mean gender (% male) Ethnicity (% white) % Age when leaving FTE Mean years since arrival % Europe % A8 countries Notes: Author's calculations using LFS data. Recent immigrants are dened as arrivals within the last ve years. 22

23 Table 3: Local authority maternity indicators, 2003/4 and 2013/14 Mean S.D. Min Max Live births per 1,000 individuals 2003/ / % induced births 2003/ / % cesareans 2003/ / % emergency cesareans 2003/ / Notes: Author's calculations using HES data. Procedures are dened using delivery onset and Oce of Population Censuses and Surveys (OPCS) Classication of Interventions and prcoedures version 4 codes contained in HES. Induced births are dened by the appropriate delivery onset code, elective cesareans by OPCS code R17 and emergency cesareans by OPCS code R18. Table 4: Local authority post-delivery outcomes, 2003/4 and 2013/14 Mothers % readmitted within 30 days Mean S.D. Min Max 2003/ / Length of stay (days) 2003/ / Children % readmitted within 30 days 2003/ / Length of stay (days) 2003/ / Notes: Author's calculations using HES data. The 30 day periods covers the period following the initial hospital discharge date. Length of stay measures the number of days between admission (birth date in the case of children) and initial discharge. 23

24 Table 5: Immigrant concentration and local authority live birth rate per 1,000 individuals, 2003/4-2013/14 LA live birth rate (per 1,000 individuals) (1) (2) (3) (4) (5) OLS OLS OLS IV IV Population Immigrant concentration 0.167*** 0.224*** *** 0.257*** *** ( ) ( ) ( ) ( ) (0.0109) Local Authority Economic Controls IMD score (2004) 0.177*** 0.174*** ( ) ( ) % no qualication 1.856* 1.745* (1.044) (1.047) Year Fixed Eects Yes Yes Yes Yes Yes Hospital Fixed Eects No Yes Yes Yes Yes Demographic Controls No No Yes No Yes Observations 3,465 3,465 3,465 3,465 3,465 First stage F-Stat Pseudo R-squared Number of hospitals Notes: *** denotes signicance at 1%, ** at 5%, and * at 10% level. The dependent variable is the number of live births recorded in NHS hospitals for 1,000 individuals in the local authority. Immigrant concentration measures the percentage of the local authority population born outside of the UK. Columns 4 and 5 uses the percentage of the local authority population who are born outside of the UK from in 1991 as an instrument for immigrant concentration in year t. Demographic controls account for the size and composition of the local authority population (age-sex adjusted). 24

25 Table 6: Immigrant concentration and the use of maternity procedures, 2003/4-2013/14 % emergency cesarean % induced births (1) (2) (3) (4) IV IV IV IV Immigrant concentration *** *** ** *** ( ) (0.0362) (0.0319) (0.0544) Year Fixed Eects Yes Yes Yes Yes Hospital Fixed Eects Yes Yes Yes Yes Demographic Controls No Yes No Yes Economic Controls No Yes No Yes Observations 3,465 3,465 3,465 3,465 First stage F-Stat Pseudo R-squared Number of Hospitals Notes: *** denotes signicance at 1%, ** at 5%, and * at 10% level. Immigrant concentration measures the percentage of the local authority population born outside of the UK. All columns use the percentage of the local authority population who are born outside of the UK from in 1991 as an instrument for immigrant concentration in year t. Demographic controls account for the size and composition of the local authority population (age-sex adjusted). Economic controls include the percentage of the working age population with A-levels and some higher education, the median gross weekly wage, and the 2004 IMD score. Procedures are dened using delivery onset and Oce of Population Censuses and Surveys (OPCS) Classication of Interventions and prcoedures version 4 codes contained in HES. Induced births are dened by the appropriate delivery onset code and emergency cesareans by OPCS code R18. 25

26 Table 7: Immigrant concentration and length of hospital stay, 2003/4-2013/14 Child length of stay Mother length of stay (1) (2) (3) (4) IV IV IV IV Immigrant concentration *** ** *** *** ( ) ( ) ( ) ( ) Year Fixed Eects Yes Yes Yes Yes Hospital Fixed Eects Yes Yes Yes Yes Demographic Controls No Yes No Yes Economic Controls No Yes No Yes Observations 3,465 3,465 3,465 3,465 First stage F-Stat Pseudo R-squared Number of Hospitals Notes: *** denotes signicance at 1%, ** at 5%, and * at 10% level. Immigrant concentration measures the percentage of the local authority population born outside of the UK. All columns use the percentage of the local authority population who are born outside of the UK from in 1991 as an instrument for immigrant concentration in year t. Demographic controls account for the size and composition of the local authority population (age-sex adjusted). Economic controls include the percentage of the working age population with A-levels and some higher education, the median gross weekly wage, and the 2004 IMD score. Length of stay measures the number of days between admission (birth date in the case of children) and initial discharge. 26

27 Table 8: Immigrant concentration and patient cost, 2003/4-2013/14 Average cost per mother (1) (2) (3) (4) OLS OLS IV IV Immigrant concentration 0.908*** 1.422*** ** (0.330) (0.415) (0.472) (0.707) Year Fixed Eects Yes Yes Yes Yes Hospital Fixed Eects Yes Yes Yes Yes Demographic Controls No Yes No Yes Economic Controls No Yes No Yes Observations 3,465 3,465 3,465 3,465 First stage F-Stat Pseudo R-squared Number of Hospitals Notes: *** denotes signicance at 1%, ** at 5%, and * at 10% level. Immigrant concentration measures the percentage of the local authority population born outside of the UK. Columns three and four use the percentage of the local authority population who are born outside of the UK from in 1991 as an instrument for immigrant concentration in year t. Demographic controls account for the size and composition of the local authority population (age-sex adjusted). Economic controls include the percentage of the working age population with A-levels and some higher education, the median gross weekly wage, and the 2004 IMD score. Costs are calculated using 2014/15 tari prices and Version 4 Healthcare Resource Group (HRG) codes. Costs are adjusted for regional variation in provision costs, and reported in 2014 prices. Costs do not include care received by children, antenatal or postnatal admissions. 27

28 Table 9: Immigrant concentration, antenatal care and maternal age, 2003/4-2013/14 First antenatal appointment (weeks) Mother age (1) (2) (3) (4) OLS IV OLS IV Immigrant concentration *** 0.184*** *** *** (0.0209) (0.0357) ( ) ( ) Year Fixed Eects Yes Yes Yes Yes Hospital Fixed Eects Yes Yes Yes Yes Demographic Controls Yes Yes Yes Yes Economic Controls Yes Yes Yes Yes Observations 3,465 3,465 3,465 3,465 First stage F-Stat Pseudo R-squared Number of Hospitals Notes: *** denotes signicance at 1%, ** at 5%, and * at 10% level. Immigrant concentration measures the percentage of the local authority population born outside of the UK. Columns two and four use the percentage of the local authority population who are born outside of the UK from in 1991 as an instrument for immigrant concentration in year t. Demographic controls account for the size and composition of the local authority population (age-sex adjusted). Economic controls include the percentage of the working age population with A-levels and some higher education, the median gross weekly wage, and the 2004 IMD score. First antenatal appointment records the week of pregnancy at which patients rst received antenatal care in an NHS hospital. Mother age records the age of patients at the date of admission. Table 10: Immigrant concentration and the percentage of patients readmitted to hospital within 30 days of discharge 2003/4-2013/14 Mothers Children (1) (2) (3) (4) IV IV IV IV Immigrant concentration *** *** ( ) ( ) (0.0101) (0.0151) Year Fixed Eects Yes Yes Yes Yes Hospital Fixed Eects Yes Yes Yes Yes Demographic Controls No Yes No Yes Economic Controls No Yes No Yes Observations 3,465 3,465 3,465 3,465 First stage F-Stat Pseudo R-squared Number of Hospitals Notes: *** denotes signicance at 1%, ** at 5%, and * at 10% level. The dependent variable is the percentage of patients resident in the local authority who are re-admitted to hospital within 30 days of initial discharge. In columns one and two this includes only mothers, while in columns three and four this includes only newborn children. Immigrant concentration measures the percentage of the local authority population born outside of the UK. All columns use the percentage of the local authority population who are born outside of the UK from in 1991 as an instrument for immigrant concentration in year t. Demographic controls account for the size and composition of the local authority population (age-sex adjusted). Economic controls include the percentage of the working age population with A-levels and some higher education, the median gross weekly wage, and the 2004 IMD score. 28

29 Table 11: Immigrant concentration and the percentage of `White (British)' patients readmitted to hospital within 30 days of discharge, 2003/4-2013/14 Mothers Children (1) (2) (3) (4) IV IV IV IV Immigrant concentration * *** ( ) ( ) (0.0211) (0.0316) Year Fixed Eects Yes Yes Yes Yes Hospital Fixed Eects Yes Yes Yes Yes Demographic Controls No Yes No Yes Economic Controls No Yes No Yes Observations 3,465 3,465 3,465 3,465 First stage F-Stat Pseudo R-squared Number of Hospitals Notes: *** denotes signicance at 1%, ** at 5%, and * at 10% level. The dependent variable is the percentage of patients with ethnicity recorded as `White (British)' in HES resident in the local authority who are re-admitted to hospital within 30 days of initial discharge. In columns one and two this includes only mothers, while in columns three and four this includes only newborn children. All columns use the percentage of the local authority population who are born outside of the UK from in 1991 as an instrument for immigrant concentration in year t. Demographic controls account for the size and composition of the local authority population (age-sex adjusted). Economic controls include the percentage of the working age population with A-levels and some higher education, the median gross weekly wage, and the 2004 IMD score. 29

30 Immigrants (000s) Figure 1: The non-uk born population in England, 2003/ /14 8,000 7,000 6,000 5,000 4,000 3,000 2,000 1, Financial year Source: Author's calculation using UK labour force survey (LFS) data and national sampling weights. 30

31 Figure 2: The English population age distribution in 2013, by native-immigrant status Source: Author's calculation using UK labour force survey (LFS) data and national sampling weights. Data include individuals living in England only. 31

32 Figure 3: The English population age distribution in 2013, by native-recent immigrant status Source: Author's calculation using UK labour force survey (LFS) data and national sampling weights. Data include individuals living in England only. 32