Immigration and Work Schedules. Theory and Evidence

Immigration and Work Schedules. Theory and Evidence Timothy N. Bond Purdue University Osea Giuntella University of Oxford, IZA October 9, 2016 Preliminary and Incomplete Draft Abstract Economists have long been interested in analyzing the effects of immigration on native wages and employment. Yet, there is little evidence on the effects of immigration on work conditions. Previous studies have shown that because of task complementarities immigration pushes natives towards more communication intensive jobs. This reallocation of tasks has important effects on native work conditions. We provide a theoretical framework to analyze the effects of immigration work schedules. The model allows for immigrants to have a comparative advantage in either the production or provision (i.e. lower disamenity costs) of night-time tasks, which leads them to disproportionately choose night-time employment. Because day-time and night-time tasks are imperfect substitutes, the relative wage of day-time tasks increases as their supply becomes relative more scarce. Consistent with our hypotheses we show that immigration decreases the likelihood of natives of working nightly shifts. Similarly, we find that immigration decreases natives likelihood of working in riskier jobs. By analyzing the differences in relative supplies and relative wages between natives and immigrants, we disentangle whether production or provision is more important for this selection effect. Keywords: Immigration, Comparative Advantage, Night Shifts, Occupational Risk JEL Classification Numbers: F22, J61, J31, R13 Purdue University, Department of Economics, Purdue University Krannert Building 100 S. Grant Street West Lafayette, Indiana 47907-2056 University of Oxford, Blavatnik School of Government and Nuffield College. 1 New Road, OX11NF, Oxford, Oxfordshire, UK. Email: osea.giuntella@nuffield.ox.ac.uk. 1

1 Introduction A popular argument in favor of immigration is that immigrants accept jobs that natives would never accept. There is a growing literature analyzing immigrant-native differences in occupational risk across several developed economies and recent empirical evidence indicates that immigrants are more likely to hold jobs involving worse working conditions (Orrenius and Zavodny, 2012, 2009). However, quite surprisingly, there has been little theoretical and empirical investigation of the relationship between immigration and natives non-pecuniary job characteristics. This paper attempts to fill this void, specifically focusing on work schedules. We build a model which allows immigrants to sort into night-time jobs for two reasons. First, these jobs may be a better match for their skills, that is they have a comparative advantage in production. Second, they may be more willing to work under such conditions, that is they have a lower utility from job amenities. Our model predicts that, because of these factors, an increase in the fraction of immigrants in an economy leads to an increase in the proportion of jobs worked at night. However, due to immigrants advantages at night, this increase causes a decrease in the fraction of natives who work night shifts. Further, because night and day production are imperfect substitutes, the relative wages of day-time workers increase. We confirm our predictions empirically using the American Community Survey. Using timeand spacial-variation of the concentration of immigrants, we find that a 10% increase the fraction of foreign workers leads to a 2 percentage point increase in the fraction of jobs worked at night.results go in the same direction when considering the an index of manual intensity of the job and injury rates. Economists have long been interested in understanding the effects of immigration on the labor market (Card, 1990; Hunt, 1992; Friedberg and Hunt, 1995; Borjas, 1995; Carrington and Lima, 1996; Dustmann et al., 2005; Borjas et al., 2011, 2008; Ottaviano and Peri, 2012; Glitz, 2012). Most of the studies found evidence of little or no negative effects of immigration on native wages and employment. The debate becomes more controversial when one focuses on individuals who are more likely to suffer immigrant competition on the labor market: low-skilled workers, ethnic minorities, and previous cohorts of immigrants. Yet, to this point we know very little about how 2

immigration affects other important labor market characteristics such as the occupational risk, physical intensity, and the type of schedule associated with a given job. One exceptional paper is Peri and Sparber (2009). Using a general equilibrium model with multiple tasks, Peri and Sparber (2009) shows that an increase in immigration leads to an increase in the relative provision of jobs with manual tasks, and an increase in the relative wages of jobs that require a high-degree of communication tasks. Our model builds on their framework. There is evidence that work conditions can have long-lasting effects on workers physical health (Ravesteijn et al., 2013; Fletcher and Sindelar, 2009; Case and Deaton, 2005) and cognitive abilities (Mazzonna and Peracchi, 2014). Working in physically demanding jobs accelerates aging, increasing stress and the risk of injury. Similarly, working irregular shifts or nightly schedules increases the risk of negative health outcomes, and reduces time spent with family and friends affecting the consumption of relational goods, marital stability, childrens well-being and family well-being Costa (1996); Presser (2000); Davis et al. (2008); Strazdins et al. (2006); Enchautegui et al. (2013); Vyas et al. (2012). Medical evidence suggests that working non-standard hours increases the risk of obesity, ischemic heart disease, and breast cancer. More generally, working schedules has important effects on the likelihood of reporting feelings of chronic fatigue, anxiety, and depression. One benefit to accepting worse work conditions is earning a compensating wage differential. However there is little empirical evidence on risk premiums. Overall, research indicates that immigrants earn risk premiums that are similar to natives, but some groups (e.g. Mexicans in the US) earn smaller or no risk premiums (Hersch and Viscusi, 2010). The wage premium for irregular shifts is also relatively small. In the United States, only a small fraction of workers reported to work non-standard hours because of the compensating wage differential (McMenamin, 2007). This evidence suggests that, for most workers, non-standard schedules are the result of limited labor market opportunities. Furthermore, the distribution of job-disamenties among workers is highly unequal. As reported by Enchautegui (2013), 60 percent of workers with non standard schedules have earnings below the median of the typical American worker, and 40 percent have earnings that are lower than those of 75 percent of all workers. Thus, there is a growing attention in increasing workers awareness of the risks associated with particular working conditions and improving the job quality of immigrants has become an important policy issue (Enchautegui, 3

2008). We investigate how immigration affects the allocation of work schedules and in particular the likelihood of working nightly shifts. We use the American Community Survey data (2005-2014) and study how changes in the spatial concentration of immigrants over time affect the likelihood of working nightly shifts and the allocation of risky jobs. As immigration may be endogenous to unobservable local trends in the labor market and because natives may respond to immigration by moving to other areas we use the traditional shift-share instrument proposed by Altonji and Card (1991) and Card (2001). We argue that pull factors that attract more immigration, such as economic growth, should lead to a downward bias in the effect of interest because of the positive correlation between economic cycle and the incidence of nigthly shifts and because of the well-known negative (shortrun) correlation between the economic cycle and work-related injuries. Ruhm (2000, 2013); Haaland and Telle (2014) showed that these types of injuries follow the economic cycle while Borjas (1994) shows the pro-cyclicality of migration. Therefore, although we expect a positive correlation between immigration and wages, we may expect a negative (if any) correlation between economic conditions and natives likelihood of working nightly shifts or working in riskier jobs. In other words, the effects of immigration on our main variable of interest would be downward biased if we do not properly control for the economic cycle. Thus, we control for local labor market fixed effects and a large set of time-varying local labor market characteristics (GDP, unemployment, etc.) that should account for the omitted variable bias associated with permanent and time-varying local area characteristics. We argue that by including state fixed effects, regionyear fixed effects and time-varying local labor market characteristics, we can reasonably assume that past immigrant concentrations are uncorrelated with current unobserved local trends that may be correlated with the demand for nightly shifts and riskier jobs. This paper is organized as follows. Section 2 reviews previous literature on immigration, job risk, and work schedules. Section 3 introduces our theoretical framework, In Section 4, we illustrate the data. Section 5 discusses the main results. We provide concluding remarks in Section 6. 4

2 Previous Work 2.1 Why immigrants hold riskier jobs Media reports have contributed to popularize the idea that immigrant workers are in jobs that native workers would never accept. But beside anecdotal evidence, there are several reasons we might expect immigrants to hold riskier jobs. Coming from countries that are, on average, characterized by worse working conditions, immigrants may have different perception of job risks than natives. Differences in risk knowledge and perception may be also explained by differences in socio-economic status and language proficiency (Dávila et al., 2011). Because of language barriers, the cost to provide safety training to immigrant workers may be higher (Hersch and Viscusi, 2010). Furthermore, immigrants, who took the risk of migration, may have lower risk aversion than natives (Berger and Gabriel, 1991). This may explain the self-selection of immigrants in riskier jobs, but also a lower productive within jobs as immigrants take higher risks than expected or necessary. In addition, as most immigrants in the developed world arrive with lower human capital and less financial assets than natives, they have higher incentives than natives to accept worse working conditions for higher life-time earnings (Grossman, 1972). These incentives are reinforced by the fact that immigrants are usually young and relatively healthy (healthy immigrant effect-see for instance Antecol and Bedard (2006) and Kennedy et al. (2006)- and might therefore be willing to trade-off some of their health capital for better wages at worse economic conditions. Newly arrived immigrants may face language barriers, and, therefore, may have a comparative advantage in working in more manual-intensive jobs, than in occupations requiring communication and social interaction skills. Furthermore, as the exit rates from these jobs are higher (Martin et al., 2012), there may be more opportunities and lower search costs for recent immigrants. One may therefore expect immigrants to self-select in occupations involving higher physical intensity and worse schedules. 2.2 Previous evidence on immigrant-native differences in risky jobs Despite all these arguments and the anecdotal evidence on immigrants injured or killed in dangerous jobs, Berger and Gabriel (1991) and Hamermesh (1998) found little evidence that immigrants work in riskier jobs than natives. However, more recent studies have found that 5

immigrants are in fact more likely to work in riskier jobs (Loh and Richardson, 2004), (Giuntella, 2012; Orrenius and Zavodny, 2012, 2009). The differences in the results with respect to earlier research are explained by differences in the way of measuring risk, but also by the different samples analyzed. In particular,orrenius and Zavodny (2009) argue that the increase in immigrants job risk in the US may be explained by a decline in the average human capital among immigrants, and by the fact that immigrants were crowded into riskier jobs because of the increase in the immigrant population over time. However, it is important to stress that most of previous studies have focused on occupational risk analyzing the different likelihood of natives and immigrants to work in jobs with high injury and fatality rates. To the best of our knowledge this is the first study analyzing the effects of immigration on work schedules in the US and one of the very few studies on the effects of immigration on non-pecuniary working conditions. 3 Model 3.1 Primitives We consider an open economy similar to Peri and Sparber (2009). Workers produce two inputs, D and N, which are combined through a CES production function to produce a final good Y Y = [βd θ 1 θ + (1 β)n θ 1 θ ] 1 θ θ (1) where θ (0, ) measures the elasticity of substitution between the two inputs and β captures their relative productivities. We think of these inputs primarily as day-time and night-time services, though they could equivalently be thought of safe and risky tasks, or normal and distasteful work. We treat Y as the numeraire good so that all wages are in terms of the price of Y. Workers are endowed with a stock of day skills (d) and night skills (n) which determines their output at day and night tasks, respectively. These stocks are determined by the worker s type which is either foreign ( f ) or native (n). The set of workers in the economy is of Lebesgue measure 1, a fraction (1 f ) of which are foreign. Foreign workers have a comparative advantage in nighttime production so that n f d f n n d n. This could be, for instance, because foreign workers posses 6

a comparative advantage in manual skills, as shown by Peri and Sparber (2009), and manual tasks are more prevalent at night. 1 What is important is that these are factors which influence the relative productivities of, and thus the demand for, workers. Without loss of generality, define n n d n η and n f d f a comparative advantage at night tasks. = νη where ν 1 represents the degree to which foreign workers possess Workers supply one unit of labor inelastically to the market. They are hired by perfectly competitive firms to either a day or night task, which pays w D and w N respectively per unit of production. That is a foreign worker assigned to a day task earns w D d f, a foreign worker assigned to a night task earns w N n f, and similarly for native workers. Night-time tasks are less desirable than day-time tasks, thus workers incur a fixed disamenity cost when producing n which varies throughout the population. Native worker i incurs cost c i, while a similar foreign worker incurs cost λc i with λ 1. The parameter c is distributed over (0, ) throughout the population by the continuous, twice-differentiable probability distribution G(c). Thus, we allow for foreign workers to incur a lower disamenity cost for night tasks. This could be, for instance, because foreign workers have a higher health stock, which allows them to work under less desirable conditions, or because they have a lower marginal utility from leisure and other job amenities due to differences in wealth or family composition. What is important is that this encapsulates any factors that influence the relative supply curve of foreign workers for night-time tasks. Workers have concave utility with respect for wages. Specifically a worker who earns income x receives utility U = ln(x). This functional form allows for a tractable characterization of the equilibrium, but is otherwise inconsequential. 1 In an earlier version of this paper, we modeled a possible difference in the productivity of manual and communicative tasks at night directly. However, to our knowledge there does not exist a dataset which allows us to measure differences in the tasks of jobs during different times of the day. Thus, while interesting, this model did not yield any additional predictions relative to the ones we present here. 7

3.2 Labor Demand Perfectly competitive firms maximize profits by choosing the amount of D and N to purchase from workers given wages w D and w N, π = [βd θ 1 θ + (1 β)n θ 1 θ ] 1 θ θ wd D w N N (2) From the two first order conditions, then, we obtain a relative demand function D N = ( ) 1 β θ ( ) θ wn (3) β w D 3.3 Labor Supply Workers take wages as given and choose whether to work a day or night task. Turning first to natives, the utility for such a worker who works during the day is U i = ln(d n w D ) (4) while at night, U i = ln(n n w N ) c i (5) By equating (4) and (5), we see that a native worker will choose to work in a night job so long as e c i ω N η (6) where ω N w N w D, the relative wage for night workers, and recall that η is the relative productivity of native workers at night. Then, denoting h(c) = exp g(c), we can define the total output of night services from native workers as N n = (1 f )H(ω N η)n n (7) The first two terms represent the fraction of the population which is native multiplied by the fraction of natives who receive higher utility at market wages from night-time tasks than day- 8

time tasks. The final term is simply the per native output at night-time tasks. Likewise for day time tasks, D n = (1 f )[1 H(ω N η)]d n (8) Foreign workers experience the same utility as native workers for daytime employment, U i = ln(d f w D ) (9) However, for work at nights they receive utility, U i = ln(d n w N ) λc i (10) as they have a weakly lower utility cost for supplying night-time labor. They are thus willing to work nights provided, e c i (ω N νη) 1 λ (11) We can then use the distribution of c to find the total output of night services from foreign workers, ( ) N f = f H [νω N η] λ 1 n f (12) The first two terms again represent the fraction of the population which is foreign multiplied by the fraction of foreigners who receive higher utility at market wages from night-time tasks than day-time tasks. The final term is the per foreign worker output at night-time tasks. The same applies for the total output of day-time tasks, D f = f [ ( )] 1 H [νω N η] λ 1 d f (13) Combining the expressions for the labor supply of foreign and domestic workers, we can arrive 9

at the total relative labor supply in the economy, D N = D n + D f (14) N n + N f [ ( )] D N = (1 f )[1 H(ω Nη)]d n + f 1 H (1 f )H(ω N η)n n + f H ( [νω N η] 1 λ d f [νω N η] λ 1 ) (15) n f 3.4 Equilibrium The labor market will be at equilibrium when the relative supply of labor is equal to the relative demand for labor. By equating (3) to (15) we find, [ ( )] ω N = (1 f )[1 H(ω Nη)]d n + f 1 H [νω N η] λ 1 1 θ d f ( ) 1 β (1 f )H(ω N η)n n + f H [νω N η] λ 1 n β f (16) Equation (16) implicitly defines the equilibrium relative wages in the economy. Given these relative wages, the equilibrium relative task output is determined through equation (3). Note that we can close the model for general equilibrium by assuming that workers spend all of their income on Y. 3.5 Comparative Statics: Immigration The main interest of our model is analyzing the effect of an in increase in immigration. We first show that immigrants always have stronger preferences for night-time jobs. Lemma 1. The proportion of foreign workers in night-jobs is weakly higher than the proportion of native workers in night-jobs provided ν 1 and λ 1. If either of these inequalities is strict, then the proportion of foreign workers in night jobs is strictly higher than the proportion of native workers in night jobs. ( ) Proof. The proportion of foreign workers in night jobs is simply H [νω N η] λ 1, while that same proportion for native workers is H(νω N ). As H is a strictly increasing function, inspection of the arguments proves the lemma. Foreign workers can differ from natives in two ways. They can be relatively more productive at night-tasks (ν > 1), and they can provide night-tasks at lower personal cost (λ < 1). Both 10

of these characteristics work in the same direction, causing the relative supply of night tasks by foreign workers to always be higher than for native workers. Proposition 1 then immediately follows. Proposition 1. An increase in the proportion of the population that is foreign f leads to a decrease in the relative provision of day-time tasks D N and a decrease in the relative wages of night-time jobs ω N, provided ν 1 and λ 1, with at least one inequality holding strict Proof. Implicitly differentiating (16), ω N f = 1 θ ω1 θ N ( 1 β β ) 1 1 θ D ( 1 D N ω N ) 1 where D = [ ( 1 H [νω N η] 1 λ )] ( H(ω N η)n n d f [1 H(ω N η)]h ) ] 2 n f [ (1 f )H(ω N η)n n + f H ( [νω N η] 1 λ [νω N η] 1 λ ) n f d n the change in the relative supply function, D N, due to a change in f holding w N fixed. Wages and the elasticity of substitution are always positive. The relative supply of day-time services due to ( ) a decrease in the relative wage of day-time services, is always negative, so 1 d N D 1 dω N is also positive. 2 The sign of this derivative thus depends on D. From lemma we know that for ( ) ( ) ν 1 and λ 1, 1 H [νω N η] λ 1 [1 H(ω N η)] and H(ω N η) H [νω N η] λ 1 and that these latter inequalities are strict provided any of the former inequalities are also strict. Comparing the last two terms, n n d f n f d n follows from the comparative advantage of foreign workers to night tasks (ν 1). Thus dω N d f < 0. Using the labor supply equation [equation (15)], d D N dω N D N = ( 1 β β ) θ ω θ N Since dω N d f < 0, it follows that d D N d f < 0. Since foreign workers always work more night-tasks than native workers, increasing the relative proportion of foreign workers leads to a decrease in the relative provision of night-tasks. In 2 It is easy to show this derivative is strictly negative in our model. The exact analytical expression is long but is available upon request. 11

equilibrium then, since day-time and night-time tasks are imperfect substitutes, the relative wage of day-time tasks increases as their supply becomes relatively more scarce. 3.6 Comparative Statics: Mechanism The main results of our model were driven by the increased willigness of foreign workers to accept employment at night. This came from two possible channels. First, they may have a comparative advantage in production (ν 1), that is firms are willing to pay them relatively higher wages for night-time tasks. Second they may have a comparative advantage in provision (λ 1), that is foreign workers can work in night-time tasks at relatively lower cost. We can disentangle between the supply and demand factors by looking at the relationship between relative wages and relative supplies. Proposition 2. If foreign workers have a comparative advantage in production at night-tasks, then in equilibrium the ratio of foreign relative wages to native relative wages will be greater than 1. Proof. The equilibrium relative wages for native workers, ω n, are simply ω n = ωη (17) while for foreign workers, ω f = ωνη (18) Taking the ratio between the two relative wages, ω f ω n = ν (19) which is greater than 1 whenever there is a comparative advantage for foreign workers in production. Proposition 2 simply states that when foreign workers have a comparative advantage in nighttime work, it should be reflected in their wages. Because the market is competitive, workers are paid their marginal product at their tasks, regardless of when it is produced. Since foreign 12

workers produce relatively more at night, their wages should be relatively higher. 3 Disamenity costs do not effect the wages an employer is willing to offer in a competitive environment. Instead, they effect the willingness to supply, and the amount of labor supplied given the wages offered for a task. This intuition leads to Proposition 3. Proposition 3. If foreign workers have a lower cost in supply night-tasks, then in equilibrium, conditional on their relative wages, the fraction of foreign workers in night-tasks will be strictly higher Proof. The equilibrium fraction of native workers in night-tasks is simply H(ω n ) 1 H(ω n ) (20) while for foreign workers, H(ω 1 λ f ) 1 H(ω 1 λ f ) (21) Inspection of the arguments proves the proposition. The worker with marginal distaste for night-employment is indifferent between night and day tasks at the equilibrium relative wages. The observed relative wages will be different between foreign and native workers, as the equilibrium determines a price for output, and foreign and native workers differ in their relative outputs. However, in data we do not observe output, but wages. Thus, conditional on the relative wages for each group, we will observe more foreign workers working in night-time tasks if they in fact are able to supply it at lower cost. 4 Data We use the American Community Survey (ACS) data for the years 2005-2011. The ACS public use files use the same geographic coding as the 2000 Census. We restrict the sample to the 3 It is possible that the relationship between output and wages of foreign workers could be distorted by employer or co-worker prejudice. However, as our proposition relies on relative wages, in order for prejudice to effect our results it would have to be that employer s hold disproportionate animus towards foreign workers during day-time shifts relative to evening shifts. We note that in a competitive framework prejudice should not have long-run implications for wages (Becker, 1971), though it is possible in the short-run (Charles and Guryan, 2008) or when the market has frictions (Black, 1995; Bond and Lehmann, 2015). As customer prejudice directly affects the output of workers, customers holding disproportionate animus towards foreign workers in day-time tasks would lead to foreign workers having a comparative advantage in production at night-time tasks. 13

working age population (16-64). We calculated worker s experience as age - education - 6. We define immigrants based on the country of birth and calculate the share of immigrants living in each metropolitan statistical area (MSA) and Public Use Microdata Area (PUMA) in a given year. To construct our instrument we use the geographical distribution of immigrants in the US across MSAs as of 1990 as reported by the 1990 US Census. Our main dependent variable is the likelihood of working a nightly shift. The American Community Survey contains information on commuting time, and time of departure and arrival from home to the workplace. Using information on arrival time at work we construct an indicator for nightly shifts which takes values equal to one for all workers who started their shift between 6pm and 6am. 4 While our main contribution is the study of the effects of immigration on the reallocation of working schedules, we do also analyze the effects on other working conditions that may affect workers health and examine the job risks associated with a given occupation. To this goal we merge data on accidents and injury by industry with information on occupation and industry of workers in our ACS samples. Finally, we replicate the analysis by Peri and Sparber (2009) and Ottaviano et al. (2010) and confirm that immigration pushes natives out of manual tasks into more communication-intensive jobs. Table 1 illustrates the summary statistics for our main variable of interest (night schedule, occupational injury rate and index of manual intensity of an occupation by immigration status, ethnicity and gender. In particular, column 3 reports the summary statistics of immigrants who reported a Hispanic ethnicity and column 4 reports the summary statistics for immigrants of Mexican origin. Panel B and C present these statistics separately for men and women. Immigrants are in general more likely than natives to work in occupations involving less standard schedules, a higher injury rate and that are generally more physically demanding (manually intensive). The differences are larger when we focus on immigrants of Hispanic origin, particularly among Mexican immigrants who are by far the largest immigrant group in the United States. Though women are in general less likely to work nightly shifts and are employed in jobs involving lower injury risk and less physical intensity than men, we do observe important differences between native and foreign-born women (see Panel C). 4 We test the robustness of our results to alternative metrics for late hours work. 14

Table 2 and 3 report unconditional and conditional differences in the likelihood of working nightly shifts (columns 1-3), in the average occupational injury rate (columns 4-6), and in the the index of manual intensity associated with a given occupation (columns 7-9) between natives and foreign-born by ethnicity and gender. 5 Empirical Specification To identify the effect of immigration on natives likelihood of working nightly shifts, we exploit variation over time in the share of immigrants living in a given PUMA between 2005 and 2011. Formally, we estimate the following equation: N ipt = α + βs pt + X ipt γ + Z ptλ + δ s + λ RY + ɛ irt, (22) where N ipt is an indicator for whether individual i at time t in PUMA p worked at night; S pt is the share of immigrants in PUMA p at time t; X is a vector of time-varying individual characteristics (such as age, education, marital status and number of children); Z MSA,t is a vector of time-varying labour market and economic conditions (at the MSA level); δ s are state year fixed effects; λ RY are region-year fixed effects; and ɛ ipt captures the residual variation in the likelihood of working nightly shifts. Because we are interested in estimating the average effect of immigration on the population, a linear probability model might represent a good approximation of this effect (Angrist and Pischke, 2008). While the spatial correlation approach has been widely used in the immigration literature, it is subject to two main criticisms (e.g., Borjas et al., 1996; Borjas, 2003). Immigration may have effects on internal mobility as natives may leave their residence as a response to the labor market effects induced by immigration in a local area. Furthermore, immigration is a non-random process and economic immigrants will cluster in areas with better economic opportunities. In this setting, we argue that pull factors that attract more immigration, such as economic growth, should lead to a downward bias in the effect of interest based on the fact that in expansion periods (particularly in the short-run) employers may prefer increasing production by extending work hours rather than hiring and this may suggest a positive correlation between economic trends and the incidence of 15

nightly shifts.furthermore, there is an extensive literature on the negative (short-run) correlation between the economic cycle and health (Ruhm, 2000) finding in particular that work-related accidents tend to be pro-cyclical. 5 All our estimates control both for state fixed effects, region-year fixed effects and a key set of economic characteristics at the MSA level. We argue that even if the large number of controls for market and economic conditions at the MSA level would not account for all the time-varying unobservables that may confound our analysis, it is unlikely that they might confound the relationship of interest in the opposite direction with respect to the large number of observable MSA characteristics that we control for. However, to address the concern of a local unobserved shock affecting both native and immigrant labour demand, we follow Altonji and Card (1991) and Card (2001) and use an instrumental variable based on a shift-share of national levels of immigration into MSAs to impute the supply-driven increase in immigrants in each PUMA. In practice, we exploit the fact that immigrants tend to locate in areas that have higher densities of immigrants from their own country of origin, and we distribute the annual national inflow of immigrants from a given source country across the PUMAs using the US Census 1990 distribution of immigrants from a given country of origin. Specifically, let us define F ct as the total population of immigrants from country c residing in the US in year t and s cr,1990 as the share of that population residing in MSA m as of year 1990. We then construct ˆF cmt, the imputed population from country c in MSA m in year t, as follows: ˆF cmt = s cm,1990 F ct (23) and the imputed total share of immigrants as: Ŝ mt = c ˆF cmt /P m,1990 (24) where P r,1990 is the total population in MSA m as of 1990. The variation of Ŝ mt is only driven by the changes in the imputed foreign population (the denominator is held fixed at its 1990 5 Though recent papers (Ruhm, 2013; Tekin et al., 2013) show that healthy recession effects are not present when using more recent cuts of the data-including the Great Recession, Ruhm (2013) confirms that deaths due to cardiovascular disease, transport accidents and deaths resulting from falls, drowning or fires continue to be pro-cyclical. 16

value) and is used as an instrument for the actual share of immigrants in MSA m at time t (S mt ). Using the distribution of immigrants in 1990, we reduce the risk of endogeneity because annual immigration inflows across MSAs might be driven by time-varying characteristics of the MSA that are associated with the likelihood of working nightly shifts. A typical criticsm to the validity of this instrument is that it cannot credibly address the endogeneity of immigration in the presence of persistent local economic trends correlated both with immigration and labor market conditions. However, we argue that this criticism is mitigated by the inclusion of state and region-year fixed effects and a large set of economic controls. Our identification assumes that the imputed inflow of immigrants is orthogonal to the MSA-specific shocks and trends in labour market conditions after controlling for state and region-year fixed effects and local economic trends at the MSA level. 6 Results Effects on Nightly Shifts Table 4 reports the main results of our analysis. We focus on employed men aged between 15 and 64 years old and use the share of male immigrants in a PUMA to identify the effects of immigration on native likelihood to work nightly shifts as gender-specific immigration shares may better proxy for actual exposure to immigrants in the labour market. However, Table A.1 replicates the main estimates for the entire sample and by gender using the share of immigrants in the working age population. Our baseline specification includes controls for education (4 groups), marital status, experience, a quartic in age, dummies for race, MSA time-varying characteristics (gdp, unemployment rate,median wage, population), state fixed effects, and region-year fixed effects. In column 1, we illustrate the first-stage regression. The instrument is relevant and the F statistic is well above the conventional levels. We report the OLS estimate in column 2. The coefficient suggests that a one standard deviation in the immigration share (.147) would decrease natives likelihood of working nightly shifts by 4% with respect with to the mean of the dependent variable (.14). In column 3, we instrument the share of male immigrants using the shift-share IV. Consistent with our prior that if anything the endogeneity would downward bias our result, we find that 17

the 2SLS estimate is larger than the OLS. The coefficients indicates that a one standard deviation in the immigration share (.147) would decrease natives likelihood of working nightly shifts by 10% with respect with to the mean of the dependent variable (.14). Table 7 replicates the main analysis on immigrants. The effect is similar to the one observed in the natives sample. A one standard deviation increase in immigration decreases immigrants likelihood of working night shifts by 12% with respect to the mean of the dependent variable in the sample (15.2%). If anything the effect is larger when we exclude recent immigrant cohorts (column 5), but the differences are not statistically significant. 6 7 Conclusion This paper develops a simple theoretical framework to analyze the effects of immigration on the allocation of work schedules. Because immigrants have a comparative advantage in production or provision at night, immigrants specialize in non-standard shifts. Thus, immigration increases the supply of nightly shifts, pushing natives towards daily schedules. The empirical findings support the main implication of the model. We find that immigration reduces natives likelihood of working nightly shifts. More generally we observe an improvement in other non-pecuniary working conditions. As immigrants push natives towards more communicationintensive jobs, we observe a reduction in the average injury rate faced by native workers. The reallocation of tasks and schedules may therefore have non-trivial implication on the health of native workers. Our findings suggest that policy-makers should not neglect the effects of immigration on work schedules and other working condition that are known to affect individual health and well-being. References Altonji, J. G., Card, D., 1991. The effects of immigration on the labor market outcomes of lessskilled natives. In: Immigration, trade, and the labor market. University of Chicago Press, pp. 201 234. 6 We exclude immigrants who have been in the US for less than 5 years. 18

Angrist, J. D., Pischke, J.-S., 2008. Mostly Harmless Econometrics: An empiricist s companion. Princeton University Press, Princeton, NJ. Antecol, H., Bedard, K., May 2006. Unhealthy assimilation: Why do immigrants converge to American health status levels? Demography 43 (2), 337 360. Becker, G. S., 1971. The Economics of Discrimination, 2nd Edition. University of Chicago Press. Berger, M. C., Gabriel, P. E., 1991. Risk aversion and the earnings of US immigrants and natives. Applied Economics 23 (2), 311 318. Black, D. A., April 1995. Discrimination in an equilibrium search model. Journal of Labor Economics 13 (2), 309 334. Bond, T. N., Lehmann, J.-Y. K., 2015. Prejudice and racial matches in employment. Borjas, G., 1994. The economics of immigration. Journal of economic literature, 1667 1717. Borjas, G. J., 1995. The economic benefits from immigration. Journal of Economic Perspectives 9 (2), 3 22. Borjas, G. J., 2003. The labor demand curve is downward sloping: Reexamining the impact of immigration on the labor market. The Quarterly Journal of Economics 118 (4), 1335 1374. Borjas, G. J., Freeman, R. B., Katz, L. F., 1996. Searching for the effect of immigration on the labor market. The American Economic Review, 246 251. Borjas, G. J., Grogger, J., Hanson, G. H., March 2008. Imperfect substitution between immigrants and natives: A reappraisal. NBER Working Paper (13887). Borjas, G. J., Grogger, J., Hanson, G. H., September 2011. Substitution between immigrants, natives, and skill groups. NBER Working Paper (17461). Card, D., 1990. The impact of the Mariel boatlift on the Miami labor market. Industrial and Labor Relations Review 43 (2), 245 257. Card, D., 2001. Immigrant inflows, native outflows, and the local labor market impacts of higher immigration. Journal of Labor Economics 19 (1), 22 64. 19

Carrington, W. J., Lima, P., 1996. The impact of 1970s repatriates from Africa on the Portuguese labor market. Industrial and Labor Relations Review 42 (9), 330 347. Case, A., Deaton, A., 2005. Analyses in the economics of aging. NBER Chapters (7588), 185 212. Charles, K. K., Guryan, J., October 2008. Prejudice and wages: An empirical assessment of becker s the economics of discrimination. Journal of Political Economy 116 (5), 773 809. Costa, G., 1996. The impact of shift and night work on health. Applied Ergonomics 27 (1), 9 16. Dávila, A., Mora, M. T., González, R., 2011. English-language proficiency and occupational risk among hispanic immigrant men in the united states. Industrial Relations: A Journal of Economy and Society 50 (2), 263 296. Davis, K., Goodman, W. B., Pirretti, A. E., Almeida, D. M., 2008. Nonstandard Work Schedules, Perceived Family Well-being, and Daily Stressors. Dustmann, C., Fabbri, F., Preston, I., 2005. The impact of immigration on the British labour market. The Economic Journal 115 (507), F324 F341. Enchautegui, M. E., et al., 2013. Nonstandard work schedules and the well-being of low-income families. Fletcher, J., Sindelar, J., 2009. Estimating causal effects of early occupational choice on later health: Evidence using the psid. NBER Working Paper 15256. Friedberg, R., Hunt, J., 1995. The impact of immigrants on host country wages, employment and growth. Journal of Economic Perspectives 9 (2), 23 44. Giuntella, O., 2012. Do immigrants squeeze natives out of bad schedules? Evidence from italy. IZA Journal of Migration 1 (7), 1 21. Glitz, A., 2012. The labor market impact of immigration: A quasi-experiment exploiting immigrant location rules in Germany. Journal of Labor Economics 30 (1), 175 213. Grossman, M., 1972. On the concept of health capital and demand for health. Journal of Political Economy 80 (2), 223 255. 20

Haaland, V. F., Telle, K., 2014. Pro-cyclical mortality across socioeconomic groups and health status. Journal of health economics. Hamermesh, D. S., 1998. Immigration and the quality of jobs. In: Hamermesh, D., F.D. Bean, e. (Eds.), Help or Hidrance? The Economic Implications of Immigration for African Americans. The Russel Sage Foundation,New York, NY. Hersch, J., Viscusi, W. K., 2010. Immigrant status and the value of statistical life. Journal of Human Resources 45 (3), 749 771. Hunt, J., 1992. The impact of the 1962 repatriates from Algeria on the French labor market. Industrial and Labor Relations Review 45 (3), 556 572. Kennedy, S., McDonald, J. T., Biddle, N., Dec. 2006. The healthy immigrant effect and immigrant selection: Evidence from four countries. Social and Economic Dimensions of an Aging Population Research Papers 164, McMaster University. URL http://ideas.repec.org/p/mcm/sedapp/164.html Loh, K., Richardson, S., 2004. Foreign-born workers: trends in fatal occupational injuries, 1996-2001. Monthly Lab. Rev. 127, 42. Mazzonna, F., Peracchi, F., 2014. Unhealthy retirement? evidence of occupation heterogeneity. Orrenius, P. M., Zavodny, M., 2009. Do immigrants work in riskier jobs? Demography 46 (3), 531 551. Orrenius, P. M., Zavodny, M., 2012. Immigrants in risky occupations. IZA Discussion Paper 6693. Ottaviano, G. I., Peri, G., Wright, G. C., 2010. Immigration, offshoring and american jobs. Tech. rep., National Bureau of Economic Research. Ottaviano, G. I. P., Peri, G., 2012. Rethinking the effects of immigration on wages. Journal of the European Economic Association 10 (1), 152 197. Peri, G., Sparber, C., July 2009. Task specialization, immigration, and wages. American Economic Journal: Applied Economics 1 (3), pp. 135 169. 21

Presser, H. B., 2000. Nonstandard Work Schedules and Marital Instability. Ravesteijn, B., van Kippersluis, J., Doorslaer, E., 2013. The wear and tear on health: what is the role of occupation? Tech. rep., Tinbergen Institute Discussion Paper Series. Ruhm, C. J., 2000. Are recessions good for your health? The Quarterly Journal of Economics 115 (2), 617 650. Ruhm, C. J., August 2013. Recessions, healthy no more? Working Paper 19287, National Bureau of Economic Research. URL http://www.nber.org/papers/w19287 Strazdins, L., Clements, M. S., Korda, R. J., Broom, D. H., DSouza, R. M., 2006. Unsociable work? nonstandard work schedules, family relationships, and childrens well-being. Journal of Marriage and Family 68 (2), 394 410. Tekin, E., McClellan, C., Minyard, K. J., July 2013. Health and health behaviors during the worst of times: Evidence from the great recession. Working Paper 19234, National Bureau of Economic Research. URL http://www.nber.org/papers/w19234 Vyas, M. V., Garg, A. X., Iansavichus, A. V., Costella, J., Donner, A., Laugsand, L. E., Janszky, I., Mrkobrada, M., Parraga, G., Hackam, D. G., 2012. Shift work and vascular events: systematic review and meta-analysis. Bmj 345, e4800. 22

Table 1: Working Conditions by Immigrant Status, Ethnicity and Gender Natives Immigrants Hispanic Mexican Immigrants Immigrants Mean and Std.Dev N Mean and Std.Dev N Mean and Std.Dev N Mean and Std.Dev N Overall Sample working at night 0.11 4,966,368 0.13 1,162,413 0.17 507789 0.20 290,860 (0.32) (0.34) (0.37) (0.40) occupational injury rate 3.26 4,721,827 3.41 1,132,137 3.60 498417 3.71 289,517 (1.74) (1.68) (1.42) (1.31) manual occupation 0.51 5,257,302 0.57 1,228,211 0.65 534081 0.69 305,186 (0.20) (0.21) (0.19) (0.18) Panel B: Men working at night 0.14 2,554,120 0.15 655,491 0.19 310046 0.22 191,859 (0.35) (0.36) (0.40) (0.41) occupational injury rate 3.34 2,386,561 3.43 642,163 3.71 309224 3.80 193,474 (1.64) (1.55) (1.28) (1.19) manual occupation 0.55 2,663,919 0.61 683,406 0.70 321658 0.74 198,763 (0.22) (0.22) (0.19) (0.16) Panel C: Women working at night 0.08 2,412,248 0.11 506,922 0.13 197743 0.16 99,001 (0.27) (0.31) (0.33) (0.36) occupational injury rate 3.17 2,335,266 3.38 489,974 3.43 189193 3.54 96,043 (1.83) (1.83) (1.59) (1.51) manual occupation 0.47 2,593,383 0.52 544,805 0.56 212423 0.60 106,423 (0.17) (0.18) (0.17) (0.17) Notes - Source: ACS (2005-2011). The sample is restricted to 15-64 years old employed workers. 23

Table 2: Unconditional Differences in Working Conditions by Immigrant Status, Ethnicity and Gender (1) (2) (3) (4) (5) (6) (7) (8) (9) VARIABLES Work Work Work Occupational Occupational Occupational Manual Manual manual at night at night at night Injury Rate Injury Rate Injury Rate Occupation Occupation Occupation Panel A: All Immigrant 0.0187*** 0.1547*** 0.0571*** (0.000) (0.002) (0.000) Hispanic Immigrant 0.0542*** 0.3447*** 0.1361*** (0.001) (0.002) (0.000) Mexican Immigrant 0.0841*** 0.4561*** 0.1761*** (0.001) (0.003) (0.000) Observations 6,128,781 5,474,157 5,257,228 5,853,964 5,220,244 5,011,344 6,485,513 5,791,383 5,562,488 R-squared 0.001 0.002 0.004 0.001 0.003 0.004 0.012 0.037 0.039 Panel B: Men Immigrant 0.0086*** 0.0960*** 0.0544*** (0.000) (0.002) (0.000) Hispanic Immigrant 0.0505*** 0.3705*** 0.1503*** (0.001) (0.003) (0.000) Mexican Immigrant 0.0742*** 0.4614*** 0.1836*** (0.001) (0.003) (0.000) Observations 3,209,611 2,864,166 2,745,979 3,028,724 2,695,785 2,580,035 3,347,325 2,985,577 2,862,682 R-squared 0.000 0.002 0.003 0.001 0.005 0.006 0.010 0.044 0.044 Panel C: Women Immigrant 0.0246*** 0.2084*** 0.0511*** (0.000) (0.003) (0.000) Hispanic Immigrant 0.0446*** 0.2528*** 0.0945*** (0.001) (0.004) (0.000) Mexican Immigrant 0.0765*** 0.3647*** 0.1272*** (0.001) (0.005) (0.001) Observations 2,919,170 2,609,991 2,511,249 2,825,240 2,524,459 2,431,309 3,138,188 2,805,806 2,699,806 R-squared 0.001 0.002 0.003 0.002 0.001 0.002 0.013 0.021 0.021 Notes - Source: ACS (2005-2011). The sample is restricted to 15-64 years old employed workers. Robust standard errors are in parentheses. 24

Table 3: Conditional Differences in Working Conditions by Immigrant Status, Ethnicity and Gender (1) (2) (3) (4) (5) (6) (7) (8) (9) VARIABLES Work Work Work Occupational Occupational Occupational Manual Manual manual at night at night at night Injury Rate Injury Rate Injury Rate Occupation Occupation Occupation Panel A: All Immigrant 0.0050*** 0.1516*** 0.0376*** (0.000) (0.002) (0.000) Hispanic Immigrant 0.0132*** 0.1480*** 0.0610*** (0.001) (0.003) (0.000) Mexican Immigrant 0.0218*** 0.1738*** 0.0785*** (0.001) (0.003) (0.000) Observations 5,990,853 5,372,537 5,164,944 5,715,609 5,118,230 4,918,465 6,343,225 5,685,102 5,465,967 R-squared 0.037 0.039 0.039 0.064 0.064 0.065 0.235 0.240 0.237 Panel B: Men Immigrant -0.0030*** 0.1039*** 0.0340*** (0.001) (0.003) (0.000) Hispanic Immigrant 0.0075*** 0.1412*** 0.0605*** (0.001) (0.003) (0.000) Mexican Immigrant 0.0114*** 0.1466*** 0.0748*** (0.001) (0.004) (0.001) Observations 3,123,597 2,801,148 2,688,759 2,942,899 2,632,821 2,522,916 3,259,993 2,920,676 2,803,773 R-squared 0.038 0.038 0.039 0.094 0.093 0.095 0.295 0.296 0.296 Panel B: Women Immigrant 0.0149*** 0.2068*** 0.0417*** (0.001) (0.003) (0.000) Hispanic Immigrant 0.0193*** 0.1395*** 0.0573*** (0.001) (0.005) (0.000) Mexican Immigrant 0.0351*** 0.1840*** 0.0745*** (0.001) (0.006) (0.001) Observations 2,867,256 2,571,389 2,476,185 2,772,710 2,485,409 2,395,549 3,083,232 2,764,426 2,662,194 R-squared 0.021 0.022 0.022 0.044 0.043 0.044 0.117 0.116 0.113 Notes - Source: ACS (2005-2011). The sample is restricted to 15-64 years old employed workers. All estimates include controls for education (4 groups), marital status, experience, a quartic in age, dummies for race, MSA time-varying characteristics (gdp, unemployment rate,median wage, population), state fixed effects, and region-year fixed effects. Robust standard errors are in parentheses. 25

Table 4: Effects of Immigration on Native Night Shifts (Men, 25-54) FIRST STAGE OLS 2SLS FIRST STAGE OLS 2SLS (1) (2) (3) (4) (5) (6) Dep. Var: Share of immigrants Night Night Share of immigrants Night Night MSA-level Work Work MSA-level Work Work IV 0.7261*** 0.6494*** Predicted Share of Immigrants (0.059) (0.069) Share of immigrants -0.0017*** -0.0015*** -0.0017*** -0.0014** (MSA-level) (0.000) (0.000) (0.000) (0.001) Observations 795,310 827,329 795,310 795,310 827,329 795,310 Mean of Dep. Var. Standard Deviation Sociodemographic controls YES YES YES YES YES YES State fixed effects YES YES YES YES YES YES Region-year fixed effects NO NO NO YES YES YES Notes - Source: ACS (2005-2011). The sample is restricted to 15-64 years old employed men. All estimates include controls for education (4 groups), marital status, experience, a quartic in age, dummies for race, MSA time-varying characteristics (gdp, unemployment rate,median wage, population), state fixed effects, and region-year fixed effects. Standard errors are clustered at the MSA level. 26