smooth (er) landing? the dynamic role of networks in the location and occupational choice of...
TRANSCRIPT
Smooth(er) Landing? The Dynamic Role of Networks in the
Location and Occupational Choice of Immigrants∗
Jeanne Lafortune † José Tessada‡
February 23, 2011
Abstract
How do networks influence the location and occupation decisions of immigrants? If so,is this influence long-lasting? This paper addresses these questions by analyzing immigrantflows to the United States between 1900 and 1930. We compare the distributions of immigrantsboth by intended and actual state of residence to counterfactual distributions constructed byallocating the national-level flows using network proxies. The distribution of immigrants byintended state of residence is most closely approximated by a distribution that allocates themwhere migrants of their own ethnicity, irrespective of their occupation, had settled. Mean-while, the actual distribution of immigrants is better approximated by the location of previousimmigrants of the same ethnicity and occupation, but only for the first 5 years of a migrant’sstay. These results are consistent with migrants using networks as a transitory mechanismwhile they learn about their new labor markets and not consistent with alternatives.
JEL Codes: F22, J61, N31
∗We thank Carol Graham, Sabrina Pabilonia, Peter Temin, Madeline Zavodny, and seminar participants at Yale Uni-versity, American University, Clemson University, PUC-Chile, University of British Columbia, UCL, the VIII Workshopof the Regional Integration Network (RIN) of LACEA, the Population Association of America 2010 Annual Meetings,and the Maryland Population Research Center for their comments, and particularly Judy Hellerstein for insightfuldiscussions and detailed comments on this project. We wish to acknowledge Carolina Gonzalez-Velosa for excellentresearch assistance, Gérard Lafortune for careful data entry, and Eileen Gallagher for editorial assistance. The usualdisclaimer applies.
†University of Maryland and Pontificia Universidad Católica de Chile. Corresponding author. Phone: (56-2)354-7101, Fax: (56-2) 553-2377. Email: [email protected].
‡Pontificia Universidad Católica de Chile. Email: [email protected].
It is the most natural thing in the world for an immigrant to want to settle where there arenumbers of others of his immediate kind.–Henry Pratt Fairchild, 1925
1 Introduction
The fact that immigrants of a particular country or ethnic group tend to locate in simi-lar regions once in their new adoptive country has been observed for a long time. The anti-immigration discourses of the early 20th century used this pattern to argue against (mostlyEuropean) immigration over that period. A similar pattern has been observed in the most recentwave of migration to the United States, this time coming mostly from Mexico and Central Amer-ica. This pattern of concentration has not only been used in policy discussion: various studieshave used it as a way to generate (plausibly exogenous) variation in the number of migrants to aparticular location (see for example Altonji and Card 1991, Card 2001, and more recently Cortés2008 and Peri and Sparber 2009, among many others). Munshi (2003) argues that the size ofthe ethnic network (in his case, at the level of the village of origin) increases the probability thata new migrant will find employment in the destination, suggesting that this clustering may belinked to the capacity of immigrants to help each other find work through referrals.1 Patel andVella (2007) emphasize that referrals, through the networks, may help explain the occupationalconcentration of different ethnicities in different cities that are observed in the current data.
However, the dynamic process through which networks may be providing labor market as-sistance has not received significant attention previously. Do migrants select their destination bymatching their set of skills to that of the main occupations within their network? How does therole of networks appear to change as the migrant spends more time in her adopted country andlearns about labor markets in the United States? Is the reliance of migrants on their network forlabor market purposes permanent or transitory? This paper attempts to answer these questionsby first setting up a simple theoretical framework and then empirically evaluating some of theimplications using the location and occupation choices of immigrants implicit in the gross andnet immigration flows to the United States between 1900 and 1930 by ethnicity, occupation andstate. The data allow us to measure concentrations across states and occupations, measured bothat arrival to the United States and after immigrants have spent time in the country.
Overview. This study first sets up a simple location-occupation decision framework to motivateour empirical results. We assume that migrants who have just arrived in their new country areunable to find employment unless they are being referred by another individual. Once they have
1As a matter of fact, part of the literature on labor market networks has used immigrants and/or ethnic groupsas a study group to analyze the role of networks in labor markets, focusing in the way information is shared amongthem.
1
acquired better knowledge of local labor markets and/or have obtained credentials that can beused to search for jobs without a direct referral they may become capable of finding employmentin the sector of their choice. Given this structure, immigrants who are deciding where to settlemay pick a location which provides a combination of two factors: a network of individuals whocan provide them referral/employment for the first period, and a labor market for the set of skillsthey possess, as it may become valuable later on.
This simple framework has some clear implications for the location and occupation patternsamong immigrants, both in the first years after their arrival and once they have spent a longerperiod in their destination. If ethnic networks only provide ethnic-specific amenities or consump-tion goods, then location choices should be related to the strength of pure ethnic networks andthe overall attractiveness of a state for one’s skills but should have little or no relation to the oc-cupations of the members of these networks beyond some ethnic-specific skill set.2 Furthermore,these patterns should not change as the migrant spends more time in the destination country.Similarly, if frictions or other features of the destination’s labor market force immigrants to relypermanently on their network for job seeking we would observe that their location decisionswould be driven by the presence of individuals of the same ethnicity and occupation and thatthis relationship will not change over time. In contrast, if individuals assimilate in the local labormarket, then we should observe that the occupations in which immigrants are employed in thefirst period after arrival are highly correlated with the most important occupations in their ethnicgroup in that particular destination. In addition, and contrary to the other two alternatives, thiseffect should change over time. If these jobs represent mostly a transition or a stepping stone,this relation should decrease as immigrants spend more time in the United States, consistentwith networks providing a smoother landing rather than a permanent “crutch”. Finally, ourhypothesized channel also suggests that, based on the skills immigrants possess before their ar-rival to the United States, we would not observe that migrants ex-ante select destinations wheretheir ethnic networks often practice their occupation. This is particularly the case when migrantsanticipate their reliance on the network to be short-lived and the loss associated with the occupa-tional switch is relatively small, thus reducing the benefit from choosing the best match in termsof occupation/skills set and opting for a larger concentration of compatriots.
We explore these implications using two sources of data that cover the period of the largewave of migration to the United States between 1900 and 1930, a period that spans part of theGreat Migration era and the years after World War I when major immigration reforms wereenacted. At that time, when immigrants arrived to the United States (and when they boardedthe ship), they were asked a series of questions including their ethnicity, the state where theyintended to live and their occupation in their country of origin. Their answers were tabulated and
2For example, Italians are more likely to work in pizza restaurants because they have specific knowledge aboutthe tasks (i.e. they have the skills), but that is not a reflection of the role of networks neither does it depends on thespecific destination.
2
presented in the annual reports of the Commissioner of Immigration from 1899 to 1930, a datasource previously unexplored to the best of our knowledge.3 As they settled in the United States,they were also surveyed during the decennial Censuses where they reported the year in whichthey first arrived, their country of birth, the state where they actually live and the occupation theyare currently performing. We match the cohorts of immigrants from these two data sources byyear of immigration and ethnicity to form a panel at the cohort-level.
Using these sources, we study whether the patterns of location and occupation choice implicitin the gross and net flows of immigration are compatible with the hypothesis that ethnic networksprovide temporary support in the labor market. To do this, we allocate the national flow ofimmigrants by ethnicity and occupation for every year of our analysis in three different ways,which proxy for functions networks may serve. All proxies are based on shares constructed usingthe stock of previous immigrants, measured in 1900 for this study, as is relatively standard inthis literature (see for example Altonji and Card 1991).4 First, we assume that all newly arrivedmigrants will locate in states where past migrants of their own ethnic group (regardless of theiroccupations) had settled. This would capture the role of ethnic networks as providing ethnic-specific benefits unrelated to labor markets. Second, we allocate the flow of migrants accordingto the geographic distribution of past migrants who shared the same occupation (regardless oftheir ethnicities). This should capture the labor market attractiveness of a location. Finally, weconstruct a third distribution where all current migrants are allocated using the distribution ofprevious migrants who shared both their ethnicity and occupation, which should correspond tothe role of labor market networks within ethnic groups. We simultaneously compare the fit ofthese three counterfactual distributions to the actual distributions of immigrants at two points:ex-ante, that is at their arrival to the United States (using their reported occupations in theircountry of origin and their intended state of residence) and once they have settled, using theiractual occupations and states of residence.5
For example, let us imagine that a state in 1900 had 10 percent of all Italians, 20 percent ofall bakers and 30 percent of all Italian bakers in the United States but also only 5 percent of allcarpenters and 2 percent of all Italian carpenters. Further assume that 100 more Italian bakersand 100 more Italian carpenters arrive to the United States in 1915 compared to 1910 and that inthe 1920 Census, 150 more Italian bakers and 50 more Italian carpenters declare to have arrivedto the United States in 1915 compared to 1910. The empirical exercise performed in this paper
3See section 3 for a detailed description of the original data contained in the annual Reports of the Commissionerof Immigration and the processed data used in this paper.
4We argue that our measure of past stocks represents a measure of a network and try to construct it in a way asto diminish the potential reflection problem, as emphasized by Manski (1993). We also control for a large set of fixedeffects to capture other confounding factors.
5One could be worried about the potential collinearity between these three proxies. In order to verify that theresults are not corrupted by this problem, specifications where the ethnic share was built excluding individuals of thesame occupation and the occupation share was built excluding individuals of the same ethnicity were also run andthe results are qualitatively and quantitatively very similar.
3
measures to which of the three proposed counterfactual is the actual allocation of migrants bystate, both ex-ante and ex-post, closest to.6 If individuals select their ex-ante location based onlyon ethnic networks, we would expect that 10 more Italian bakers and carpenters would declareto be heading to this state at their arrival to the United States. Were we to see 20 bakers but only5 carpenters, labor market factors would appear to be determining ex-ante choices. Similarly, forex-post location decisions, were we to observe 45 more Italian bakers but only 1 Italian carpentersin the flow of immigrants measured in the decennial Census who arrived in 1915 compared to1910, the role of ethnic-occupation networks would appear particularly influential.
As a first step to this analysis we perform a simpler exercise where we allocate newly arrivedmigrants using only two different proxies: based on the location of past migrants of the sameethnicity and of migrants of the same occupation. We find that individuals’ choice of intendedstate of residence, which were reported at their arrival to the United States, appears to be mostlyrelated to the existence of a network of individuals from the same ethnic group in their selectedstate of residence although there appears to be some role, albeit smaller, for decisions to bedriven by the existence of a labor market suited for the skills of the new migrant. However, theactual location of migrants appears to be much more influenced by the past prevalence of theiroccupation in the local labor market. This appears to be robust across various ways to computethe two network proxies.
We then make full use of the three proxy variables we described and try to separate morecarefully the effects of ethnic networks and labor market factors. In this case, our results showthat ex-ante decisions are particularly driven by the presence of individuals of the immigrant’sethnic group irrespective of their occupations. In contrast, ex-post decisions, as observed in theyears shortly after arrival, are strongly influenced by the presence of individuals of one’s ownethnic group who also appear to have the same set of skills as the migrant. These results arequalitatively similar whether we attempt to match the changes in the flow of immigrants overtime in each state by ethnicity, by occupation or by the combination of both. Results are alsorobust to alternative definitions and sample selections.
The difference between the factors influencing the distribution of immigrants observed beforeand after their arrival to the US are consistent with the hypothesized channel presented above:immigrants select their destination by looking for “strong” ethnic networks and, to a lesser de-gree, a good match with their skill set, and then finding temporary employment in the sameoccupations that are most prevalent within their specific ethnic network. These results also im-ply that immigrants do not appear to select their intended location looking for the places withan ethnic network where the most popular occupation is most similar to their own set of skills.
We explore further implications of this hypothesis. First, in our data, we can also measurehow the relative influence of these three factors, namely ethnic network, ethnic-occupation spe-
6We perform the exercise for all states, ethnicities and occupations simultaneously within the limits of the data asexplained in Section 3.
4
cific network, and occupation-specific labor markets, change as immigrants spend more time inthe United States. When we separate our regressions according to the time since their reportedarrival to the United States, we observe that the difference between the ex-ante and ex-post dis-tribution is more marked for immigrants recently arrived to the United States than for thosewho have been residing for longer periods. Moreover, the pattern highlighted above is especiallyvisible for immigrants with low occupational scores, which is logical because these individualswould have the least to lose by changing their occupation at arrival.7 Moreover, the pattern weobserve is most marked for occupations that produce “tradeable” goods, which again is logicalas occupations producing non-traded goods would have fewer incentives to provide employmentto individuals in the same occupation since this might be more costly to them in terms of wages.8
Finally, we also find that the results are strongest for non-English speaking groups, which is con-sistent with poor language proficiency reducing opportunities to find employment outside one’sethnic network upon arrival to the United States.9
Two other alternative hypotheses are also explored. First, the difference between ex-anteand ex-post distributions could be driven by selective return migration through which migrantswithout the support of their networks return to their country of origin. This appears to beunlikely as the difference between ex-ante and ex-post decisions is similar across ethnic groupswith high and low rates of return migration. Second, the pattern we highlight could also bedriven by immigrants relocating within the United States at their arrival if their networks cannotprovide them with the support needed. The evidence we present, although it does not rule outinternal migration playing a role, does not seem to support it as the main or sole mechanismthrough which this pattern arises. First, for internal migration to be the only factor explainingour results, it would have to be that immigrants migrate internally at their arrival to the UnitedStates and then relocate again after a few years, in order to match the decreasing relevance ofthe ethnic-occupation specific network. Second, we obtain a very poor match if we allocate theflow of immigrants by their occupation in their country of origin using our proxies to try toexplain the actual distribution of immigrants. If immigrants were just changing locations (andnot occupation), such an exercise should offer similar results as the ones discussed above; the factthat it does not would seem to indicate that immigrants do change their occupations in responseto the presence of networks.
Overall, we believe the evidence is consistent with ethnic networks playing a role upon arrival
7Using occupations at origin to predict occupation distribution of immigrants at destination has been used in theliterature. For example, Friedberg (2001) uses this approach in her study of Russian immigration to Israel. Our resultsimply that such an strategy will be more successful for more highly skilled immigrants, especially if used to predictoccupations long after arrival to the host country.
8Alternatively, it could be costlier or harder for previous immigrants working in non-tradeable occupations toprovide employment if their total output is determined by local demand for the services and goods. This is true evenif wages are equalized across states for equivalent workers.
9Related results have been observed in the context of the recent immigration wave to the United States. Periand Sparber (2009) show that differences in communication skills are related to the sorting of native and immigrantworkers into different occupations.
5
and during the first years following migration, leading migrants to adopt new occupations to fa-cilitate the transition process, but also with them being able to revert to their original occupationsonce they have learned about local labor market opportunities in subsequent years.
Related Literature. The results presented in this paper contribute to our understanding of therole networks have in the decisions of migrants and in immigration in general. First, the studiesexploiting local variation at the city level to study the impact of immigration (see for example(Card 2001; Card and Lewis 2005; Cortés 2008; Cortés and Tessada 2011; Peri and Sparber 2009)among several others) have used distributions of past waves of immigrants to generate variationin the flow of immigrants to different areas; because of the existence of ethnic networks, thisvariation is argued to be unrelated to contemporaneous local labor market conditions. Ourresults suggest that while the ex-post location of immigrants may be more endogenous to locallabor market conditions than the ex-ante one, this occurs also because immigrants change theiroccupation in response to the relative advantage of their networks and not only because theyre-optimize their location decision based on current labor market conditions. Our results alsoimply that occupation distributions among immigrants might be more persistent because of theethnic-occupation network component, but that this effect subsides as immigrants spend moretime in the host country.
Second, our results also shed some light on the dynamics of the location and occupationchoice of immigrants. Bauer, Epstein, and Gang (2005) explore the role of networks on locationchoices of migrants but does so statically. The dynamic effect of networks, in particular in thecontext of immigration, is still a fertile area for research. Theoretical models of such dynamicshave been offered (see for example Calvo-Armengol and Jackson 2004, 2007), but the availableempirical evidence has focused on the resulting segregation and workplace concentration. Thereare some papers though that have explored dynamic effects of networks in different contexts;Munshi (2009) argues that networks may be able to break occupation-based traps and presentsevidence from the Mumbai diamond industry supporting his hypothesis. In another recent pa-per, Beaman (2009) exploits the allocation process of refugees resettled in the United States tostudy these effects on employment and other labor market outcomes. She finds evidence thatethnic networks influence the access to local labor markets of newly arrived refugees, and thatthe effect depends on the “age” of the network: while the size of the recently arrived cohortsof refugees negatively affects the outcomes of the current arrivals, the size of older cohorts (twoor more years before the current refugees) improves their outcomes. Our contribution is com-plimentary to Beaman (2009) as we argue that the role of networks itself changes as a migrantspends more time in the United States, thus looking at the dynamics as cohorts “age” rather thanthe differential effect of the cohorts already present at the moment of arrival. Our focus is alsoon the occupational and geographical distribution rather than on more immediate labor marketoutcomes. Finally, Aslund (2005) look at the initial and subsequent location choices of migrants
6
in Sweden and finds little variation over time in the factors influencing the decisions, in contrastto our findings.
Third, this paper is also related to the literature that studies the performance of immigrants inthe United States around the turn of the 20th century. Our results complement existing evidenceabout earnings and other labor market outcomes by linking the location and occupation choiceof immigrants to the strength of ethnic networks and labor market factors. In particular, ourestimates on how the ethnic-occupation network becomes less relevant as the immigrants spendmore time in the United States are consistent with the evidence presented in Minns (2000), whofinds evidence of positive wage growth within cohorts of immigrants arriving during the sameperiod we study. Furthermore, Minns (2000) also reports that immigrants showed high mobilityinto well-paid occupations which could be because migrants return to the occupation they wereperforming before their arrival.10
Finally, our results improve our understanding of the occupational concentration of immi-grants. Patel and Vella (2007) build on previous evidence on the role of ethnic networks, andnetworks in general, in the labor market performance of immigrants to study the occupationalchoices of immigrants from the same country in different cities. They start from the observationthat in many cities immigrants from particular countries have developed specific occupationalniches.11 Using data from the 1980, 1990 and 2000 Census in the United States, they find thatnew immigrants are more likely to choose the same occupation previous immigrants from thesame country have chosen, and that those who choose these occupations perceive a benefit intheir earnings too.12 Similar evidence is provided in the case of China by Chen, Jin, and Yue(2010). However, all these studies are constrained to look at occupations once the individual hasmigrated and thus cannot distinguish whether this pattern is driven by individuals of similarskills selecting the same location or by immigrants selecting the same occupation once they havemigrated.
Layout. The remainder of this paper is organized as follows. Section 2 sketches a framework tohelp us better understand the factors that could influence the network dynamics of immigration.Section 3 describes the data and section 4 explains the empirical methodology and the resultsare presented in section 5. Section 6 then explores what could be the reason behind the patternidentified in the previous section. Finally, in the last section we summarize the results and offersome conclusions.
10See also Hatton and Williamson (2005); O’Rourke and Williamson (1999) for a more comprehensive view ofglobalization, in general, and migration, in particular, during this period.
11See also Federman, Harrington, and Krynski (2006).12Munshi and Wilson (2008) study the connection between ethnic networks when the American Midwest was first
settled and occupational choice today.
7
2 Motivating Theory
Assume a migrant from ethnic group j and skills o, i.e. she worked in occupation o in hercountry of origin and thus has that exact set of skills, has just arrived to location s. The immigrantlives for two periods and discounts the future using a discount factor β. We assume migrantsare risk-neutral, although risk-averse behavior does not change the main implications of ourframework.
The labor market for immigrants at destination s functions as follows. The wage (wsqo) offeredto individuals in occupation q with skills o in location s is a function of the match between theoccupation and the skills, reaching a maximum when q = o. Furthermore, we assume that thewage is higher for occupations that are in high demand in s, which will be reflected by a higherpast concentration of individuals in that occupation Nsq. We also assume that frictions in the labormarket are more important for immigrants than for natives (or immigrants that have stayed inthe country for one period already), and that these frictions make it difficult for an immigrantto find employment unless they have some referral from their own network. In particular, let’sassume that the probability for a migrant of ethnicity j of finding employment in occupation qin state s upon arrival is given by P
(Njsq
), where Njsq denotes a measure of the relative strength
of the migrant’s ethnic network in occupation q in state s, such that ∑q Pjqst = 1 for a given (j, s)pair.13 This dependence captures the fact that job referrals and opportunities are more likely tocome in occupations that are already more popular within the immigrant’s ethnic community.
As the migrant lives longer in her new environment, her knowledge of the labor market in-creases and her capacity to find employment given her skills may become less and less dependenton her network. If such learning occurs, which we will denote by the parameter α = 1, she willbe able to pick the occupation in her state which will give her the highest income, most probablythe one corresponding to her own skills. If there is no learning (α = 0), then she will remain inthe occupation she obtained in the first period.
The migrant’s utility for a given s is
U(j, s, o) = γ(Njs) + ∑q
P(Njqs)
(wsqo + β
[α(max
q′wsq′o) + (1− α)wsqo
])(1)
where γ(·) represents benefits of having individuals of one’s ethnicity which are unrelated tolabor markets, such as having restaurants, temples, etc.
The migrant’s location decision s∗ can then be written as
s∗ = arg maxs
U(j, s, o).
13In our empirical work we will approximate it with the stock of migrants of one’s ethnicity and occupation in thatstate.
8
From this simple framework, the following general conclusions can be drawn. First, if therole of networks in the labor market is limited (P′(Njqs) ≈ 0), we should observe that migrantswould locate where they have fellow countrymen (regardless of their occupation and becauseof the non-labor market benefits of ethnic networks) and where there is demand for their skills(captured by the stock of past migrants in the same occupation). The same would be true if wewere looking at the location of immigrants based on their actual occupation q.14
Secondly, if α = 0, migrants would elect a location where they have a large number of theircountrymen exercising their own occupation as this will be their only way of working with theirskills from their own country in the United States. Once more, this conclusion should be thesame whether we consider the location choice of migrants based on their skills o or their actualoccupation once they have migrated q. If γ(·) is large enough, pure ethnic networks will alsoplay a role in the location decisions. Furthermore, neither of these two channels would changeover time and thus the patterns would be similar regardless of the time spent by the immigrantat her destination.
However, if α > 0 and P′(Njqs) > 0, a different location pattern could emerge. Conditional onthe migrant’s initial skill set o and ethnicity j, the probability that a given state will be selectedwill depend on the size of the ethnic network in that state, the wage that one could earn in theoccupations that are already taken by individuals of the network with one’s skills, and the gen-eral attractiveness of that state to migrants with similar skills sets (given the anticipation that inthe second period, such employment will become available).15 Conditional on the migrant’s firstoccupation in the United States q̃, more migrants will be found (in the first period) where indi-viduals of one’s ethnic network who perform that occupation were already settled, i.e. the newmigrants’ occupation distribution will be influenced by the occupations that are more popularamong previous migrants from the same ethnicity, this it what we call the ethnic-occupation spe-cific effect. Without that network, a migrant would be unable to find employment in that sector.More migrants would also be found in locations where this occupation q̃ is well rewarded as themigrant would have been unlikely to select a state where the wage in the first period would be,in expectations, low.
Finally, conditional on the migrant’s occupation in the second period (which will be either oor q̃), migrants should be found in states where the wages for that occupation are highest.
Assuming that the overall popularity of a location for past migrants in a given occupationgroup is an acceptable proxy for the wage that a migrant can earn in that occupation, this sim-ple model would suggest that we would observe the following pattern if the mechanism we’vedescribed is at play:
1. Based on their ex-ante skills, we would find that migrants select locations where many mi-14In our data we can study the distribution of immigrants across states by occupation and the distribution of
immigrants across states by ethnicity.15If the agents are too optimistic about the chances of obtaining a job in their own occupation this will imply a bias
towards picking states based on the demand for their own occupation.
9
grants of their ethnic network (which provides with help, information and assistance duringshortly after arrival), and of their skills have previously located (which we take as a proxy ofthe demand for their set of skills).
2. In the first period after arrival, however, we would find that migrants would be located instates where many migrants of their ethnic network (and less importantly, other previousimmigrants) also share their current occupations.
3. In the second period, however, we would find that migrants would be located in states wheremigrants in the past shared their current occupations.
4. This pattern should be more marked for groups where α will be large or where the loss inhuman capital associated with switching occupations may not be large.
These elements will thus be the key insights we will be looking for in the empirical sectionsthat follow.
3 Data Description
In order to study the dynamic role of networks in the location and occupational choices ofimmigrants to the United States in the framework presented in the previous section, we need tomeasure the immigrants’ “skill set” at entry and their subsequent occupational choices as wellas their location choices over time. Without detailed individual data, we collect data on grossflows constructed with information recorded at entry or arrival to the United States, and netflows recorded ex-post. All this information is provided by the combination of two main datasources: the United States Census for the ex-post data and the Report of the Commissioner ofImmigration (henceforth RCI) for the information at entry to the United Sates. Data from theUnited States Census is taken both from the Public Use Micro Sample (PUMS), as compiled byRuggles et al. (2010), and from the original published summary tables, which contain data forthe first year of each decade.16
3.1 Administrative Data from the Commissioner of Immigration
The RCI was published annually from 1899 to 1932 (except for 1931) and presented sum-mary tables constructed using micro data from the questionnaires each immigrant coming to theUnited States had to answer.17 For each year, immigrants are classified according to their eth-nicity, self-reported occupation in the country of departure and their intended state of residence;
16The summary tables include tabulations based on the full sample of each Decennial Census.17We have tried locating these tables for years 1932-1940 without success. Furthermore, while part of the micro-data
may be available at the National Archives or a similar location in paper format, it would have been prohibitivelyexpensive and time consuming to enter the information it contains. Hopefully, this information will be available inthe future.
10
this information was originally taken from the individual data each of them had to report whenboarding at origin and/or when arriving to the United States.18 Using the tables presented in theRCI, we can create an annual series of gross entry flows for each ethnic group according to theirintended state of residence, for each ethnic group according to their occupation in the country ofdeparture, and for each intended state of residence according to their occupation in the countryof departure. The three-way categorization, i.e., flows by ethnicity and occupation for each state,is not published in the RCI. Overall, we have data available for all states and territories in theUnited States for all years between 1899 and 1930 and for 1932, with a total of 75 occupationcategories, grouped in 3 major categories, and 42 ethnic groups. All the information taken fromthese reports is labeled as “administrative data” in the rest of the paper and was digitized for thepurpose of this paper.19
3.2 Census Data
The Decennial Census collects data every 10 years, and includes data on country of birthand year of first arrival in the United States (until 1930). From the 1900, 1910, 1920 and 1930Census, we obtain information on country of birth, labor market status, occupation and state ofresidence for all those who are surveyed. Our main sources for the Decennial Census data arethe 1% samples, plus the 5% sample for 1900, publicly available through the Public Use MicroSamples.20 Variables measured from any of these two sources are referred to as “Census data”or “IPUMS data” in later sections.
We define an immigrant as anyone who reports being born outside the United States ac-cording to the value recorded in the birthplace variable in the Census. Using this definition ofimmigrant, for each Census we compute the stock of immigrants in each state and group themaccording to their occupation and country of birth. Also, with the information on year of firstarrival to the US, we are able to create net flows by country of birth and actual state of residenceand occupation. In order to improve accuracy in the calculation of shares and to avoid small-cellmeasurement problems, whenever possible we use the publicly available summary tables whichwere constructed using the original data set of the 1900 Decennial Census.21
We want to emphasize one important difference between the administrative data and theCensus, and that is the timing of the information with respect to the moment of arrival to the
18It is our understanding that there was little incentive for a migrant to misreport their answers as this had nobearing on their acceptance to the United States, except in the case where their answers when boarding the vessel andat the port of entry were different. It is only in the early 1920s that entry restrictions affecting most of the potentialimmigrants were effectively imposed in the United States.
19To the best of our knowledge the disaggregated data contained in the RCI has not been used before at this levelof detail.
20Available online at http://usa.ipums.org/usa/.21The 1890 Decennial Census micro samples are not available yet. While the summary tables are available, there are
significant differences between the classification of occupations and countries of birth used in them and those used inthe samples for subsequent years.
11
United States. In the administrative data the information on occupations is collected beforearrival, “ex-ante”, while that in the Census reflects the occupations after agents have interactedupon and after arrival, hence we label it “ex-post”. This is also true for the state distribution.Furthermore, while the Census data measures immigrants who have settled in the United States(or have at least stayed until the Census date), the administrative data measures entry flows andthus could differ from the Census measures if there is return or internal migration.22
3.3 Matching both Data Sources
The RCI and the Census use different classifications for occupations and for the origin of theimmigrants. The RCI uses ethnic groups as classification, but the Census data records birthplace.Also, while the large occupation groups are relatively similar, specific occupations are groupedin different ways and some specific categories are used only by one of the two sources.
The Census data in IPUMS provides two different occupations codes, the 1900 (for 1900and 1910) and the 1950 Census occupation codes (for all years). Administrative data comesin a different classification; with four main groups –professional, skilled, unskilled, and no-occupation, being divided into very detailed lists of subgroups (for each of the first three groups).We match groups of administrative data occupations to groups from the Census classifications.We match the administrative data to the IPUMS samples using the 1950 Census occupationcodes to preserve comparability across years. When matching to the 1900 Census publishedtables, however, occupations were paired using the 1900 Census codes because this was the onlyclassification available. The exact groups are in Table A-1 together with the corresponding Censuscodes. The appendix table also shows our classification of occupations as large or small, definedas being above or below the median size for all immigrant flow over the period.
The Commissioner of Immigration classified immigrants by their “ethnicity” rather than theircountry of origin.23 In order to reach a matching set of ethnicities, we grouped both countriesof birth and ethnicities from the administrative data. Our final classification includes 28 “ethnic-ities” used in the regression.24 Most of the pairings are fairly intuitive but we needed to makea few adjustments in order to represent the definitions used by both data sets. For example, theRCI classified all Blacks as Africans. However, over the period studied, most Black immigrantsare from the Caribbean and not from Africa. This explains why they are paired with West Indiansrather than with Africans. Similarly, Jewish immigrants were classified as “Hebrews”. We allo-cate Jewish immigrants by their country of birth using the available information from the RCI,which presents a table with the distribution of individuals by their ethnicity and the country oflast residence, although this pairing is definitely a gross approximation. We also collated our
22Mortality rates is another reason why these two flows could differ.23For some years, a table highlights the distribution of individuals by their ethnicity and the country of last residence
but does not indicate the intended state of residence nor the occupations by countries. We use the information in thesetables to allocate certain ethnic groups to different countries as explained later in this section.
24Those groups are described in Appendix Table A-2.
12
various “ethnicities” by groups to present more easily understandable summary statistics and torun some of the regressions separately for different ethnic groups.25
Finally, we also classify each ethnic group by whether its flow was above or below the me-dian over this period and classified as “large” those that were above the median. Similarly, wecomputed the difference between the national flow as measured in the administrative data andin the Census data and classify as those that have a difference in percentage term larger than themedian as high return migration ethnicities. Most of those correspond closely to the ones iden-tified in the existing literature as having large circular migration flows (Hatton and Williamson2005).
3.4 Stylized Facts and Summary Statistics
The late 19th and early 20th centuries were periods of massive migration, with the U.S. andother countries in the New World receiving significant inflows. For example at the turn of the 20thcentury around 15% of the United States population was foreign born, a number that remainedvery high until 1930.26 Annual gross immigration flows of more than a million people per yearare observed between 1900 and 1914, a trend that was broken with the onset of the war inEurope. After the war, a significant recovery in the flows is observed but they decline again afterthe introduction of changes in the immigration: the 1917 Immigration Act, the 1921 EmergencyQuota Act and the 1924 Johnson-Reid Act. By the end of the 1920 decade and the beginningof the Great Depression era the gross immigration flows are small. While we can consider thelimits on migration before 1917 to be almost non-existent, they increased progressively over thisperiod.27 These restrictions particularly limited the entry of “unskilled” workers and favoredfamily reunification. Furthermore, while a majority of the immigration over this period wasEuropean (or from countries with European ties such as Canada and Australia), an increase inthe number of Hispanics migrants was observed after 1920. One important dimension in whichimmigrants from different countries differed was in the return and seasonality patterns; whilefor some groups return migration and seasonal migration was not particularly prevalent, othergroups had important return rates, which were close to 30% for Spaniards and Italians between1890 and 1914 (see O’Rourke and Williamson 1999).
Geographic clustering is definitely visible in this period: we find in both data sets, annualimmigrations reports and decennial Census, an extremely high level of geographical clusteringby ethnic group. States popular among immigrants today were also popular in the early 20th
25Lafortune (2010) presents a similar classification of countries into groups using Census data from the same period.26As a reference, using data from the 2000 Census micro sample we calculate that the fraction of foreign born in the
United States lies around 11%; according to the 2007 and 2008 American Community Survey the same fraction liesslightly above 12%.
27See O’Rourke and Williamson (1999), Hatton and Williamson (2005) and the references therein for a more detaileddescription of the patterns of immigration for this period, including data for other countries besides the United States,and of the possible causes behind the protectionist backslash. Also see Bandiera, Rasul, and Viarengo (2010) for ananalysis of the validity of the immigration and emigration data for this period based on flows recorded at Ellis Island.
13
century: New York, Texas, California, etc. In Table 1 we present the top 3 states and the percent-age going to the 10 most popular states for each ethnic group included in the previous figures.We can observe that for all groups in all periods at least 3 out of 4 immigrants went to one of thetop 10 destinations of the group; furthermore, it is also evident that there is a significant degreeof persistence in the top 3 destinations in the Census and in the administrative data althoughthere is no perfect coincidence between the top 3 intended states and the top 3 actual states ofresidence for each of the ethnic groups. Furthermore, this appears to be fairly constant over time:the geographical concentration of the stock of immigrants in 1900 generally reflects that of theex-ante decisions made by subsequent migrants as well as their ex-post decisions. Geographicalclustering appears to be larger in ex-ante than ex-post decisions, suggesting that more individualsturn to less traditional destinations once they set foot in their country of adoption or that thosewho eventually decide to stay in the United States are not necessarily those who picked the mostpopular destinations. However, this difference is also decreasing with time as the last migrationcohorts of this period become more and more concentrated ex-post in the United States.28
We also document something that has often gone unnoticed in some previous studies which isthat there is also a very important occupational clustering by ethnicity. Table 2 shows the patternof occupational choice according to the declared occupation in the United States and accordingto the occupation in the country of origin. We observe there is also high concentration, with the10 most popular occupations for each ethnic accounting for 65% or more of the total membersof each group. While some occupations are particularly important for all immigrants (such aslaborers and farm laborers), there is also evidence of ethnic-specific clustering on different skills.This specialization in a given skill set appears also to be relatively constant over time. Oncemore, ex-ante occupational clustering is larger than ex-post but the difference remains more orless constant throughout the period. When we compare the Census and the administrative datawe can observe that in this case we see that unskilled occupations such as laborers, farm laborersand servants are among the most popular ones in both data sets. We also observe that for mostof the groups there is a certain level of agreement between the top three categories in each ofthe data sets, we also see that certain occupations are more popular ex-post than ex-ante, e.g.,miners, hotel keepers, manufacturers, etc.
Finally, the United States tend to demand different occupations in different states, leading togeographical clustering among individuals of a given occupation. This concentration is slightlyless marked than the two previous ones as the top ten states by occupation in 1900 capturedaround 70 per cent of all workers in that occupation, but it is still noticeable.
28This change could be related to the modifications to the immigration laws introduced after the end of World WarI which not only reduced immigration overall but also change the composition towards skilled workers and familyreunification.
14
4 Empirical strategy
We explore whether we can find any evidence of the empirical regularities mentioned insection 2 by attempting to contrast the “fit” of counterfactual distributions built using differentmeasures of networks to the actual ex-ante and ex-post distribution of immigrants. We areconstrained in our empirical strategy by the fact that we do not have access to micro-level datawith the individual records of migrants entering the US during this period. However, we are,as we explained in section 3, capable of constructing flows by ethnicity (j), occupation (o), state(s) and time (t), which we will denote as njost, measured at the moment of arrival to and aftersettling in the United States.
We will approximate the (relative) strength of networks in the same way traditionally used inthe immigration literature, i.e. using the distribution by states of a given group of immigrants.However, given that the stock and the flow of immigrants at a given period of time could bedetermined by other factors such as specific demand shocks, we will include the stock of immi-grants with given characteristics who are residents in a specific location in a determined periodin the past as proxies of the strength of networks.29 Specifically, we measure our networks mostlyin 1900, and include in the regressions only the immigration flows from 1905 to 1930 to avoidtoo much proximity between our flow measure and the shares employed in the prediction. Thisargument is further strengthened by the fact that immigration had significantly dampened be-tween 1890 and 1900 and thus that our measure of stocks probably reflects the location choices ofimmigrants who arrived mostly before 1890.30 Finally, our interest will lie more in the compari-son of various proxies for networks than in the significance or magnitude of a given coefficient.If all proxies suffer from the same bias, the comparison would still be valid.
In order to control for other factors such as a particular match between an ethnicity and astate (because of a particular endowment of natural resources for example), we include a numberof fixed effects to attempt to capture this and other elements.
The framework presented in section 2 suggests an empirical regression of the following type:
njost = αAjosnjot + βBjosnjot + γCjosnjot + µjos + νjot + ωjst + ηost + ε jost (2)
where n represents the immigrant flow and the shares A, B and C represent different ways ofallocating the national flow njot to different states, related to distinct types of networks.
29We recognize that this could be a biased measure of networks if past immigrants simply responded to long-lastingdemand shocks to which current migrants are also responding at their moment of arrival.
30To the best of our knowledge there is no available digitalized sample from the 1890 US Census.
15
Specifically, we define
Ajos ≡(
Njs
Nj
)(3a)
Bjos ≡(
Nos
No
)(3b)
Cjos ≡(
Njos
Njo
)(3c)
where N represents the stock of immigrants with certain characteristics already in the UnitedStates. Thus, Ajos refers to the share of individuals from one’s ethnic group who elected to livein state s in the past. The second share, Bjos refers to the geographical distribution of immigrantswho share one’s occupation. Finally, Cjos represents the share of individuals from the sameethnicity and the same occupation who lived in a particular state in the past. We will refer to Aas measuring an ethnic effect, to B as a labor/occupational effect and to C as an ethnic-specificoccupational component. 31
The regression equation includes fixed effects for all triple interactions between ethnicity,occupation, state and time. This allows to control for all possible confounding effects that areaffecting the immigration rates of a particular sub-group or the overall effects in a particularstate. We have used fewer fixed effects and the results are similar. Also, the standard errorsare clustered by ethnicity-occupation-state cells although very similar results were obtained withmuch more aggressive clustering.
Intuitively, this regression attempts to test whether a state has a differential growth in aparticular occupation between two different ethnic groups compared to a different state for oneof the three following reasons. Is it because that state was popular for a particular ethnic groupand that ethnic group has a growth in the number of people of that occupation coming fromthat ethnic group? Or because that state was popular for a particular occupation and there’snow an increase among that ethnic group of that occupation? Or because that state was popularfor individuals of that ethnic group and that occupation and there’s a growth in the number ofindividuals that satisfy both categories? Each one of the three components defined in equation(3) attempt to capture each one of these reasons separately, thus enabling us to compare themwith the actual observed distributions.
The theoretical framework presented in section 2 would suggest that when s represents theintended state of residence and o the skills with which an individual arrives in the United States,we should obtain estimates of α and β to be positive and significant while γ should not be. Onthe other hand, when s represents the actual state of residence and o the occupation in which an
31One could be worried about the potential collinearity between these three proxies. In order to verify that theresults are not corrupted by this problem, specifications where the ethnic share was built excluding individuals of thesame occupation and the occupation share was built excluding individuals of the same ethnicity were also run andthe results are qualitatively and quantitatively very similar.
16
immigrant is working in the United States, one would expect to find the parameter γ to be largeand significant with little importance of either α or β, and that the magnitude of the estimatedcoefficient should be smaller when we consider the case of immigrants that have spent more timein the United States.
Unfortunately, njost is only measured ex-post in our data, that is where s represents the actualstate of residence and o one’s occupation in the United States. However, because the flows nost
and njst are observed both ex-ante and ex-post, it will be useful to estimate a simpler version ofequation (2) by summing over all ethnicities/occupations in a given cell. Denoting the variableover which one is actually summing by k′ and the other by k, where k, k′ = j, o, one obtains
nkst = α ∑k′
Ajosnjot + β ∑k′
Bjosnjot + γ ∑k′
Cjosnjot + µks + νkt + ωst + εkst (4)
where the left-hand side variable nkst represents the flow of immigrants arriving in period t,either intending on residing or living in state s, depending on whether we use the administrativeor the census data, respectively, and from ethnicity (when k = j) or occupation (when k = o).
Regression equation (4) again contrasts the strength of the three different networks and it doesso by comparing the correlation between the predicted distributions across states of individualsof characteristic k to the actual distribution observed in the data. Exactly as before the variablesA, B and C are our measures of the strength of the pure ethnic, the occupation/labor market,and the ethnic-occupation specific networks, respectively.
In this case, all the double-interactions between characteristic k, time t and state s are includedand standard errors are clustered at the ks level. We will estimate these equations using onlycells where k = j and where k = o separately as well as jointly. In the latter case, the estimationconstrains the parameters α, β and γ to be the same but allows for a more efficient estimation asthe parameters ωst are common to both sets of regressions and thus estimated using more datapoints.
One limit to the last regression equation is that the network measure C can only be obtainedfrom one data source (IPUMS) because we require information on the distribution according tothe jos cells over time. To verify that this is not driving our results, another simplification can bemade and the following equation is estimated:
nkst = α ∑k′
(Ajos
)njot + β ∑
k′
(Bjos
)njot + µks + νkt + ωst + εkst (5)
once more for k = j, o. In this regression equation, we simply exclude the role of ethnic-occupation network. The impact of these will be bundled up with that of ethnic effects (inAjos) or occupational factors(in Bjos). While the results obtained here are less informative, theygive us an opportunity of verifying the robustness of our results to the use of different measuresof networks.
17
The above regression includes a full set of fixed effects to control for any ethnic/occupation-state, ethnic/occupation-time and state-time confounding factors. However, we also used asmaller set of fixed effects and found similar results. The standard errors are clustered by ksbut similar results were obtained when clustering by state only. All regressions are un-weighted.
Table 3 presents the summary statistics for all variables used in the analysis, for each of theaggregations that will be employed in the empirical sections. Two elements are particularly im-portant to notice from this table. First, there is substantial amount of variation in both the actualand predicted flows to exploit in our analysis. Secondly, the mean and standard deviations ofour predicted flows measures are all roughly similar which implies that we can simply comparethe coefficients to understand the relative strength of the match from each of the counterfactualdistributions.
5 Results
5.1 Ethnic vs. Labor networks
We first begin our empirical exploration by contrasting the role of two general types of net-works, ethnic and labor markets, as presented in equation (5). This strategy, although less infor-mative in terms of the factors driving the location and occupational choices of migrants, has theadvantage of allowing us to compare the results across a variety of alternative proxies for therelative strength of networks.
The results of this strategy are presented in Table 4 when estimating the flow of immigrantsselecting a particular state and of a given ethnic group (in Panel A) and occupation (in PanelB). These tables include all individuals who report an occupation but very similar results wereobtained when all individuals, including those reporting no occupation, were incorporated inthe sample.32 Similarly, in these regressions a period t corresponds to a 5-year period, with nooverlap between two consecutive periods; we also explored the same specification using annualvariation and obtained similar results.
In the first three columns, we attempt to explain the ex-ante location choices (as measured inthe administrative data) while the last three use as the left-hand side variable ex-post decisionsfrom IPUMS. In the first three columns, the measures of predicted flows are constructed byallocating the national flow (nkt) from the administrative data using the two types of shares asdetailed in equation (5). The last three columns repeat the same exercise but this time allocatingthe national flow as measured from the IPUMS. Finally, columns also differ by how the networkmeasures are computed. In columns (1) and (4) we report results using the average share ofindividuals with a characteristic k or k′ who declare state s as their intended state of residencefor the first 6 years of the administrative data, that is from 1899 to 1904. In columns (2) and
32Results that also include individuals without an occupation in the sample are available upon request from theauthors.
18
(5) we use the published Census summary tables to compute the number of individuals from agiven ethnicity or from a given occupation residing in each state in 1900. This has the advantageof offering large sample sizes and reducing the probability of small sample bias. On the otherhand, it also assumes that immigrants who are active in a given set of occupations are locatingin the same states as their natives counterpart, because we do not have summary tables withemployment levels by state and birthplace. Finally, Columns (3) and (6) compute the share ofimmigrants living in a particular state from a given ethnic group or a given occupation in the1900 Census Public Use Micro-Sample (IPUMS).
The first three columns of Table 4 indicate that when trying to explain the ex-ante “intended”location decisions of migrants, the presence of individuals of their own ethnic group is particu-larly important. When we attempt to match the distribution of immigrants by ethnicity, we findthat the past location of immigrants by occupation also contribute to increase the match withthe actual distribution, although the magnitudes of the coefficient on the ethnic component arealways larger than those on the occupational element.
On the other hand, when we explore the factors behind the ex-post location decisions ofindividuals in the IPUMS, we find quite a distinct pattern, as observed in the last three columnsof Table 4. In this case, the importance of having individuals of one’s occupation who previouslylocated in that state becomes much more important and that of ethnic networks is reduced inboth panels.
These results would thus suggest that when one attempts to explain why a particular statenow has a larger set of individuals with a given set of skills claiming to be heading there, the factthat this state was particularly attractive to individuals with those skills in the past and that moreof them are now landing in the United States is of little relevance. Instead, our results suggestthat it is because a larger number of individuals with those particular sets of skills are comingfrom ethnic groups that had a large fraction of their previous migrants already living in thatgiven state by 1900.33 On the other hand, the increase in the number of immigrants performinga particular occupation in a particular state ex-post appears to be driven by the fact that moreimmigrants from that occupation are now within the United States and this particular state wasattractive to that occupation in the past. We find little evidence that this is linked to the fact thatthis state was attractive to a given ethnic group in the past and that this ethnic group now hasmore individuals reporting working as that occupation after their arrival.
The results appear to be fairly robust across the shares as measured from the IPUMS or theCensus tables, even if the occupational distributions from the Census table include all individualsin the United States, natives or immigrants. The counterfactual distributions built with sharesfrom the administrative data appear to be more weakly related to the actual flows than the ones
33Notice that our fixed effects allows us to remove the variation that comes from permanent characteristics thatcreate an intrinsic match between states and occupations or ethnicities. For example, we rule out that our results aredriven by a state having during all this period a dominant industry that offers a perfect match for a country of originthat happens to have the same dominant industry.
19
built from Census (table or micro-sample) data. This is comforting because it indicates thatimmigrants were less likely to follow the location decisions of recent migrants (as measured byflows) than the ones of established migrants (as measured by stock). This would suggest that ourresults are unlikely to be driven by contemporary shocks since in that case, one would expect theflows to have a higher predictive power than the stocks. Furthermore, this is consistent with theempirical results of Beaman (2009).
We also confirmed that the effect we find is not simply driven by similar patterns amongimmigrants, regardless of their ethnicity. First, all regressions include fixed effects for the in-teraction of state and time, thus capturing all characteristics specific to a particular location at agiven moment. A more stringent robustness test is provide in Table 5, where we add as additionalregressors, the counterfactual distribution one would obtain by allocating each period’s flow us-ing similar ethnicities. Specifically, we include all ethnicities within a broader ethnic group asdefined in Table A-2 except one’s own. For example, Irish immigrants are allocated using the ge-ographical distribution from 1900 IPUMS of other British Isles, Australia and Canada. In almostno case are these variables entering significantly in the regression equation and in all cases, themain results remain extremely similar. The results are fairly similar across samples and suggestthat it is the location choices of one’s particular ethnicity which influences locations decisionsand not that of fairly close ethnic groups.
5.2 More Disaggregated Measures of Networks
Having shown in the previous section that the results are more or less robust to the way thenetwork measures are constructed, we now turn to the estimation of the impact of our threedistinct measures of networks: ethnic, occupational and ethnic-specific occupational networks.In this case, all shares will be computed from the IPUMS as this is the only source that allows usto do such a distinction.
Table 6 presents our baseline results for four distinct samples. The first three panels corre-spond to the results of estimating equation (4) for the location choice of individuals by ethnicity(when k = j), by occupation (when k = o) and jointly (when k = j, o). Finally, panel D shows theresults of estimating equation (2). This table presents once more the regressions where ex-antedecisions are explored in the first two columns and those about ex-post decisions are in the fol-lowing two. Since the flow by ethnicity, occupation and state is only available in the IPUMS data,the last panel only includes results about ex-post location decisions.
The results from column (1) are remarkably similar across samples and all emphasize that thekey factor driving the ex-ante location decisions of individuals lies in the presence of individualsin 1900 that were of the same ethnicity than the newly arrived migrants. Allocating newly arrivedmigrants based on the past distribution of their occupation appears to only contribute to matchthe actual distribution by ethnicity, but not by occupation or in the joint estimation. Finally, in allregressions, the ethnic-specific occupational factor is small and insignificant.
20
Our theoretical framework suggests that while an individual would pick ex-ante a locationbased on the presence of an ethnic network, that location would be particularly attractive toher when the occupations within this network are more similar to hers. This is explored incolumn (2) where an interaction term is added which captures the absolute difference betweenthe average occupational score of the network and that of the occupation of the migrant. As isexpected, this is negative in all specifications and significantly so in the first panel, suggestingthat while individuals select ex-ante locations based on the presence of a network of individualsof their ethnicity, they are particularly likely to pick one where the occupations of that networkare similar to theirs (although not necessarily identical).
The pattern is strikingly different in columns (3) and (4) where the ex-post distributionsare explored. The role of the ethnic factors disappears in this case and turns negative in allregressions. The counterfactual distributions based on the past location choices of immigrantsof one’s ethnicity are thus not related to the actual distribution, conditional on the other twocounterfactual distributions we constructed. The results presented in this table emphasize thatthe ex-post distributions appear to be much more closely correlated with our predicted flowsbased on occupational factors and more specifically, on the past location choices of immigrantssharing one’s ethnicity and one’s occupation.
Taken together, these results suggest that an increase in the number of individuals that intendto reside in a particular state, compared to the average for all states, is due mostly to the fact thatthis state was attractive in the past to ethnicities which are now receiving larger national flows.At the same, the results also imply that if we observe an increase in the number of migrantsactually practicing a given occupation in a particular state, compared to the average for all states,this is due mostly because that state in the past was popular among individuals who practicedthat occupation and who were of ethnicities that are currently experiencing larger flows. Themagnitudes of the coefficients should be interpreted as follows. Taking the example of Italianbakers, these results would suggest that a state which had received 10 percent more immigrantsof that ethnic group (irrespective of their occupation) in the past would attract 12 percent moreimmigrants of those newly arrived immigrants ex-ante. However, it is in a state where one found,in 1900, 10 percent more Italian bakers that one would eventually find 6-11 percent more of theseimmigrants ex-post, i.e. after they have settled in the United States.
Robustness Checks. We explore the robustness of these results in various ways. We first es-timate these regressions for each of our nine broader ethnic categories separately. While theresults are fairly noisy, they do not indicate that one ethnic group is driving the results morethan another. Table 7 further explores whether the results depend on the sample over whichthese estimates are obtained. One may first be worried that our results may be sensitive to prob-lems related to small cells. Columns (1) and (4) thus restrict the sample to groups whose nationalflows over the period were above the median. In Panel A, this corresponds to ethnicities with
21
large inflows, in Panel B, to occupations with large inflows, and in the last panel, to ethnicitiesand occupations that were observed in large numbers (ex-post). The results from these regres-sions are extremely similar to the ones presented in the previous table, suggesting that our mainempirical conclusions are not driven by small cells. Although not presented, we find that forsmall ethnicities, occupational groups are more relevant in driving the location choices and forrare occupations, ethnic networks appear to matter more. This is logical if one needs a networkof a significant size to be able to derive any benefits from it or if the returns to a network areincreasing with its size.
One could also be worried that after the introduction of limits on immigration, migrantsmay have had more of an incentive to provide false information to the immigration officer atarrival, thus contaminating our ex-ante measures. Furthermore, migrants arriving after the 1921Act were more skilled and more likely to come for the purpose of family reunification, thuspotentially explaining the importance of ethnic networks in that group. Columns (2) and (4)verify whether these concerns are founded by limiting the sample to immigrants arriving before1920, that is before any immigration restrictions were imposed. Once more, these results areextremely similar to those presented in Table 6.
Finally, the state of New York was the largest recipient of immigrants and, given that ourestimating equation is in levels and not in a logarithmic form, one could be worried that thisparticular state is driving our results. The results presented in columns (3) and (6) find littleindication of this: excluding this particular state has very limited impact on the overall results.
6 Exploring the difference between ex-ante and ex-post decisions
The previous section explored the factors influencing ex-ante and ex-post location decisionsof immigrants to the United States in the first three decades of the twentieth century. We findstriking differences between the factors that are determining location decisions at their arrivalin the United States and once they have settled. In the first case, ethnic networks and labormarket effects appear to drive the ex-ante location decisions of immigrants. Ex-post, however,we find a very important role to be played by the presence of individuals of the same ethnicityand occupation, which we interpret as evidence of the role played by ethnic networks in theincorporation of immigrants to the local labor markets.
In the theoretical framework presented in section 2 we suggest that this result is compatiblewith newly arrived immigrants taking on the same occupations of individuals in their ethnicnetworks at their arrival because they are more likely to receive job offers, or to obtain referralsfor positions, through them. In this sense, networks among immigrants become a tool to smooththe transition into the new labor market.
22
6.1 Differences Across Occupations and Ethnicities
If our hypothesis that networks are the reason behind this pattern is correct, then, as weexplained in our motivating framework, we should be able to find some additional evidenceconsistent with it. For example, networks are more likely to play an important role among indi-viduals who arrive in the United States with low levels of skills.34 A doctor is much less likely tochange his occupation than a general laborer, for example. Furthermore, higher-skilled individ-uals may only offer a low chance of referrals in the same area of specialization. This hypothesisis explored in the first two columns of Table 8 where we divide the sample between occupationsthat had an occupational score (based on the average wage in 1950) above or below the median,which we take to be a proxy for the skill level of the occupation. On the one hand, for highlypaid occupations, ex-ante decisions are very closely related to the presence of an ethnic networkbut the ex-post decisions are driven by pure labor market considerations.35 On the other hand,individuals of lower paid occupations display the reliance on one’s ethnic-specific occupationalnetwork as presented above. Although not shown here, very similar results were obtained ifone divides the occupations based on the RCI classification of occupations into “professionals”,“skilled” and “unskilled”. In that case, it is for individuals with occupations identified as “un-skilled” that the pattern shown above is particularly marked, with both skilled tradesmen and,particularly, professionals relying ex-post more heavily on labor market considerations.
Columns (3) and (4) of Table 8 explore another distinction between occupations. We at-tempted to classify occupations based on whether their main product is a tradable or a non-tradable (see Table A-1 for which occupations were placed into each group). This is naturallya very gross matching as some occupations may produce a variety of goods, some which aretraded and some that are not, depending on the industry in which they perform. Neverthe-less, we would expect that the use of one’s ethnic network as a transition tool may be muchmore likely to occur in the case of occupations where the output is traded, as the wage of theseindividuals would be less likely to be affected by the number of individuals performing thatoccupation in a given location. Referrals, for these individuals, would thus be less “costly”. Col-umn (3) clearly shows that the pattern we identified above is particularly visible for occupationsproducing traded products. Column (4) show that ex-post decisions of individuals working in“non-traded” sectors are more influenced by labor market concerns than by the presence of in-dividuals sharing both one’s ethnicity and occupation. Somewhat surprisingly, however, ex-antedecisions among that group appear to be somewhat correlated with the presence of an ethnic-specific occupational network. The difference in the coefficients between the two columns issignificant at 10% level in Panel B and at 1% in Panel A.
34Hellerstein, McInerney, and Neumark (2008) use matched employer-employee data and find that networks gen-erated by residential proximity are relatively more important for minorities and less-skilled workers, particularlyHispanics.
35The coefficients of interest are statistically different for both groups at 5% for Panel A (although not for anycoefficient separately) and at 10% for Panel B.
23
Furthermore, learning about local labor markets might be particularly important for individ-uals who do not speak the same language as US natives. The first columns of 9 explore thisissue by separating ethnic groups between those who spoke English (British Isles, Canada andAustralia) and those who did not. The first group has limited variation that can be exploitedand the results are much noisier in that case. Nevertheless, it seems that the pattern we iden-tified in the overall sample is most closely related to that of non-English speaking individuals.English-speaking ethnicities do rely on the ethnic-specific occupational networks but they appearto do so consistently both at arrival and once settled, which is what would be expected if thereis no learning involved in their adaptation process. The hypothesis of equal coefficients for bothgroups is not rejected in the ex-ante data but clearly rejected in the ex-post regressions.
6.2 Time since arrival
Our framework does suggest that if the lack of credentials or verifiable experience was oneof the reasons why ethnic networks were important, then ethnic-specific occupational networksshould become less relevant (i.e. reduce their correlation with the observed location and occupa-tion choices) as immigrants stay for a longer time in their new country. Thus for a given cohort ofimmigrants the estimated effect of the ethnic-specific occupation network should be decreasingin the time since arrival to the United States.
The last two columns of Tables 8 and 9 explore this issue. The difference between ex-anteand ex-post decisions is particularly marked for individuals who have just arrived in the UnitedStates. It is for that group that we observe a very strong preference for electing a state of in-tended residence based on ethnicity and labor market considerations but a very strong preferencefor eventually residing in a state where individuals who share one’s occupation and ethnicitycan be found. For more established immigrants, we find a different pattern where all factorsappear to determine almost equally ex-ante decisions but ex-post decisions depend much moreon factors related to labor markets, although not particularly the ones related to one’s ethnicity.An extremely similar pattern (if only maybe more marked) can be found in the correspondingcolumns of Table 8. The fact that the ex-ante decisions appear to be driven by similar factors is re-assuring, suggesting that this is not because these two groups were somehow different at entry.36
Furthermore, we have repeated this exercise adding data for the groups with even longer staysand found similar results to the ones we present; the effect of the ethnic-occupation network isdiminished as the migrants spend more time in the United States, while the other two, the pureethnic and the labor market effects, are strengthened.37
36The hypothesis of equal coefficients cannot be rejected for either sets of parameters because of the noisiness ofsome results. However, the test statistic is much larger for ex-post regressions than ex-ante ones.
37Results are available upon request.
24
6.3 Alternative Channels
Two other factors could explain the difference between ex-ante and ex-post decisions. First, itmay be that individuals without a group of past migrants sharing their ethnicity and occupationare more likely to leave the United States. As we discussed before, this is a period where returnmigration was a fairly important phenomena and thus could explain the patterns we observe.Since our measure of ex-ante flows include all immigrant entries into the United States but ourex-post measure involves only those who eventually stay, the difference in our results couldbe driven by selective return migration. In order to evaluate this alternative hypothesis, thefirst two columns of Table 9 estimates the same regressions but dividing the sample betweenethnicities with high and low levels of return migration. We define a group as high returnmigration if the national (gross) flow of immigrants over the entire period of analysis obtainedfrom the administrative data is, in percentage term, above the median compared to the net flowobtained from the Census data.38 The results presented in Table 9 do not, however, suggestthat the location choices of these two groups are extremely different. In both cases, the ex-antedistribution appears to be best approximated when the national flow is allocated according tothe presence of individuals of one’s ethnic group while the ex-post distribution appears to bebest approximated when using the ethnic-specific occupational networks.39 This does not seemto suggest that the selective return migration can explain the patterns we have highlighted above.
Also, the importance of the pure labor markets do appear to be more important for groupsthat have low levels of return migration, particularly for the ex-ante distribution, which is con-sistent with our framework. When individuals are forecasting staying in their new location for along period of time, the quality of the labor market for their skills becomes more crucial.
Secondly, it may also be that individuals first locate based on general considerations butthen relocate within the United States to a location where they have someone from their ethnicgroup who also share their occupation. The difference in the pattern observed would then notbe driven by the fact that immigrants change their occupation in response to the composition oftheir networks but rather that they re-optimize their location decision to bring themselves closerto a network that can provide them with referrals in their given occupation, for example. Toexplore this, the ideal data would have included information regarding internal migration, as theone currently compiled by the Census. Unfortunately, for this period, no question on whetheran individual has moved recently is available. However, the results presented above suggest thatthe pattern we observe is particularly strong for individuals recently arrived to the United States.If internal migration is an important explanation between the phenomena we just described,we would expect that individuals who have been in the United States for a longer time period
38This definition is in agreement with Bandiera, Rasul, and Viarengo’s (2010) assessment of the emigration patternsfor this period. However, we also conducted the same analysis using the emigration flows from the administrativedata to classify the ethnicities into the high and low return groups, and we obtained similar results.
39The equality of coefficients can be rejected in Panel A because of the difference in the second and third coefficient;however, we cannot reject equality of the coefficients in Panel B.
25
would have had more time to re-optimize and thus their ex-post location decisions may respondto different factors. Thus, internal migration would only explain this pattern if individuals, attheir arrival to the United States, moved temporarily to a location where their ethnic-specificoccupational network is more important to then return to one where their skills are more valuedlater on.
Finally, we can further explore the role of internal migration by exploring the differencesbetween IPUMS and administrative data. If immigrants were mostly changing their state of resi-dence but not their occupation in response to the presence of networks, the results of regressionswhere we attempt to predict the ex-post distribution by allocating the ex-ante national flows andvice versa should be identical. This is because the national flow njot should be the same whenmeasured in the administrative data than in the Census data if individuals do not change theiroccupation upon arrival. Thus, the same pattern should be observed when estimating equations(4) or (5) whether or not the right-hand side variable is built using administrative or Census data.The result of this exercise is presented in Table 10. Each column here represents one of our sam-ple. In the first panel, the left-hand side variable is measured ex-ante from the administrative databut the predicted flows are built using the national flows from the IPUMS dataset. The reverse istrue in Panel B. The results of that table are very different from those presented in Table 6. In thefirst panel, the results are fairly noisy but those using the sample of occupation-state-time cellspresented in Column (2) suggest that contrary to our previous results, the counterfactual distri-bution built from allocating the immigrants by their actual occupation in the US according to thepast location of immigrants who shared the same occupation has significant predictive powerwhen trying to identify ex-ante location decisions, something that was absent of the previousresults. Similarly, in Panel B, the importance of ethnic-specific occupational factors are almostalways absent and the ethnic component becomes much more significant. This is not consistentwith a situation where individuals simply relocate after their arrival to the United States.
7 Conclusions
We have thus presented, through various methods, evidence regarding the role played bynetworks and labor market characteristics in the determination of the location and occupationchoices of immigrants in the United States in the first three decades of the twentieth century. Wehave shown empirical evidence consistent with the hypothesis that the presence of individuals ofone’s ethnicity, but not necessarily in the same occupation, is strongly influential in the intendedlocation reported by an immigrant at her arrival in the United States. However, the actual locationand occupation choices appear to be driven much more by considerations linked to the labormarkets, in particular to the presence of individuals of one’s ethnicity who also share one’soccupation, a measure that we call ethnic-specific occupation network. We have also shownsuggestive evidence that the reason behind this change in behavior lies in the fact that unskilled
26
immigrants are likely to change their occupation once they arrive into the United States to benefitfrom the labor market connections of their ethnic networks.
The findings in this paper help us understand the role networks play in the location choiceof immigrants. While most studies have looked separately at location or occupation choices,our study combines both decisions and seems to suggest important interactions between the twochoices. It also emphasizes the dynamic aspect of the process as immigrants spend more time intheir destination country, a fact that has not been carefully documented before. It even seems toindicate that ethnic networks are particularly useful as a smoothing mechanism for newly arrivedimmigrants and are not something that migrants need to permanently rely upon.
These conclusions might also be relevant in shaping optimal immigration policy. If immi-grants of different skill levels select their location and occupation in their new country usingnetworks differently, this has implications for the impact that these immigrants can have on na-tive and previous migrant workers. Understanding the mechanics of the effects and its connec-tion to immigrant geographical and occupational concentration can thus shed light on potentialmeasures aimed at improving immigrants incorporation into labor markets and to benefit nativepopulation.
While our specific data allows us to look at a historical period, our results might be relevantfor today’s immigration debate. The immigration wave of the early 20th century resembles thatof today in some important aspects: migrants were less skilled than natives (or at least wereperceived to be less skilled), they were perceived to be culturally different than the remainder ofthe population, and their arrival generated controversy over their potential adverse effects. Moreimportantly for our results and analysis, immigrants today also share ethnic and occupationalnetworks as defined in this paper. Whether the patterns we have highlighted here are also presentin the current migration wave is a subject of future research.
References
Altonji, Joseph G. and David Card. 1991. “The Effects of Immigration on the Labor MarketOutcomes of Less-Skilled Natives.” In Immigration, Trade and Labor, edited by John M. Abowdand Richard B. Freeman. Chicago, IL: University of Chicago Press.
Aslund, O. 2005. “Now and Forever? Initial and Subsequent Location Choices of Immigrants.”Regional Science and Urban Economics 35:141–165.
Bandiera, Oriana, Imran Rasul, and Martina Viarengo. 2010. “The Making of Moden America:Migratory Flows in the Age of Mass Migration.” URL http://www.homepages.ucl.ac.uk/~uctpimr/research/ellis_accounting.pdf. Mimeo, Harvard University, LSE and UCL.
Bauer, T., G. S. Epstein, and I. N. Gang. 2005. “Enclaves, Language, and the Location Choice ofMigrants.” Journal of Population Economics 18:649–662.
27
Beaman, Lori. 2009. “Social Networks and the Dynamics of Labor Market Outcomes: Evidencefrom Refugees Resettled in the U.S.” URL http://faculty.wcas.northwestern.edu/~lab823/beaman_ethnicnetworks.pdf. Mimeo, Northwestern University.
Calvo-Armengol, Antoni and Matthew O. Jackson. 2004. “The Effects of Social Networks onEmployment and Inequality.” American Economic Review 94 (3):426–454.
———. 2007. “Networks in labor markets: Wage and employment dynamics and inequal-ity.” Journal of Economic Theory 132 (1):27–46. URL http://ideas.repec.org/a/eee/jetheo/v132y2007i1p27-46.html.
Card, David. 2001. “Immigrant Inflows, Native Outflows, and the Local Labor Market Impactsof Higher Immigration.” Journal of Labor Economics 19:22–64.
Card, David and Ethan Lewis. 2005. “The Diffusion of Mexican Immigrants During the 1990s:Explanations and Impacts.” NBER Working Paper 11552, National Bureau of Economic Re-search.
Chen, Yuyu, Ginger Jin, and Yang Yue. 2010. “Peer Migration in China.” Mimeo, University ofMaryland–College Park.
Cortés, Patricia. 2008. “The Effect of Low-Skilled Immigration on U.S. Prices: Evidence from CPIData.” Journal of Political Economy 116 (3):381–422. URL http://www.journals.uchicago.edu/doi/abs/10.1086/589756.
Cortés, Patricia and José A. Tessada. 2011. “Low-Skilled Immigration and the Labor Supply ofHighly Skilled Women.” American Economic Journal: Applied Economics.
Fairchild, Henry Pratt. 1925. Immigration: a world movement and its American significance. Macmil-lan Co., rev. edition ed.
Federman, May, David Harrington, and Kathy Krynski. 2006. “Vietnamese Manicurists: Displac-ing Natives or Finding New Nails to Polish?” Industrial and Labor Relations Review 59 (2):302–318.
Friedberg, Rachel M. 2001. “The Impact Of Mass Migration On The Israeli Labor Market.”Quarterly Journal of Economics 116 (4):1373–1408.
Hatton, Timothy J. and Jeffrey G. Williamson. 2005. Global Migration and the World Economy: TwoCenturies of Policy and Performance. Cambridge: MA: MIT Press.
Hellerstein, Judith, Melissa McInerney, and David Neumark. 2008. “Measuring the Importanceof Labor Market Networks.” Mimeo, University of Maryland.
28
Lafortune, Jeanne. 2010. “Making Yourself Attractive: Pre-Marital Investments and the Returnsto Education in the Marriage Market.” URL http://sites.google.com/site/jeannelafortune/research. Mimeo, University of Maryland.
Manski, Charles F. 1993. “Identification of Endogenous Social Effects: The Reflection Problem.”Review of Economic Studies 60 (3):531–542. URL http://www.jstor.org/stable/2298123.
Minns, Chris. 2000. “Income, Cohort Effects, and Occupational Mobility: A New Look at Immi-gration to the United States at the Turn of the 20th Century.” Explorations in Economic History37:326–350.
Munshi, Kaivan. 2003. “Networks in the Modern Economy: Mexican Immigrants in the US LaborMarket.” Quarterly Journal of Economics 118 (2):549–599.
———. 2009. “Strength in Numbers: Networks as a Solution to Occupational Traps.” Mimeo,Brown University.
Munshi, Kaivan and Nicholas Wilson. 2008. “Identity, Parochial Institutions, and Career Deci-sions: Linking the Past to the Present in the American Midwest.” URL http://www.econ.brown.edu/fac/Kaivan_Munshi/midwest7.pdf. Mimeo, Brown University.
O’Rourke, Kevin H. and Jeffrey G. Williamson. 1999. Globalization and History: The Evolution of aNineteenth-Century Atlantic Economy. Cambridge: MA: MIT Press.
Patel, Krishna and Francis Vella. 2007. “Immigrant Networks and Their Implications for Occu-pational Choice and Wages.” Discussion Paper 3217, IZA.
Peri, Giovanni and Chad Sparber. 2009. “Task Specialization, Immigration, and Wages.” AmericanEconomic Journal: Applied Economics 1 (3):135–69. URL http://ideas.repec.org/a/aea/aejapp/v1y2009i3p135-69.html.
Ruggles, Steven, J. Trent Alexander, Katie Genadek, Matthew B. Schroeder, and Matthew Sobek.2010. Integrated Public Use Microdata Series: Version 5.0 [Machine-readable database]. Minneapolis,MN: Minnesota Population Center.
29
Tabl
e1.
Spat
iald
istr
ibut
ion
ofim
mig
rant
sby
ethn
icgr
oup,
over
tim
e,C
ensu
san
dad
min
istr
ativ
eda
taco
ntra
sted
Stoc
ks19
00Fl
ows
1901
-191
0Fl
ows
1911
-192
0Fl
ows
1921
-193
01s
t2n
d3r
dTO
P10
1st
2nd
3rd
TOP
101s
t2n
d3r
dTO
P10
1st
2nd
3rd
TOP
10
Pane
lA
:Fro
mC
ensu
sda
ta
Brit
ish
ance
stry
NY
MA
PA75
.4PA
NY
MA
78.1
MI
MA
NY
78.4
MA
MI
NY
90.6
Fren
chM
AM
IN
Y75
.6C
AM
IN
Y76
.5C
AM
IN
Y83
.9M
IM
AN
Y86
.2So
uth
Euro
pean
sN
YPA
MA
86.3
MA
PAN
Y84
.5M
APA
NY
87.5
PAN
JN
Y94
.3H
ispa
nics
TXA
ZFL
93.9
AZ
CA
TX89
.7N
YC
AT
X89
.8N
YT
XC
A88
.0G
erm
anic
NY
ILPA
77.1
ILN
YPA
80.0
OH
PAN
Y82
.6N
JIL
NY
89.7
Scan
dina
vian
sM
NIL
WI
76.3
WI
ILM
N74
.2IL
MN
NY
77.9
MN
ILN
Y83
.9R
ussi
ans
and
othe
rsN
YPA
IL83
.6IL
PAN
Y86
.8IL
PAN
Y87
.9PA
ILN
Y92
.1O
ther
Euro
pean
sN
YPA
IL85
.9N
YO
HPA
77.5
NY
OH
PA78
.9IL
PAN
Y91
.7O
ther
coun
trie
sH
IC
AM
A86
.4N
YW
AC
A72
.2N
YH
IC
A85
.5W
AN
YC
A90
.3
Pane
lB
:Fro
mad
min
istr
ativ
eda
ta
Brit
ish
ance
stry
NY
MA
PA84
.2N
YM
AM
I80
.3N
YM
IM
A87
.3Fr
ench
NY
PAM
A89
.9N
YPA
MA
88.7
NY
PAM
A91
.8So
uth
Euro
pean
sN
YIL
MI
77.0
NY
MA
MI
78.0
MA
NY
MI
84.9
His
pani
csFL
NY
TX96
.1TX
NY
AZ
95.1
TX
CA
NY
94.8
Ger
man
icN
YPA
IL81
.3N
YIL
PA77
.7N
YIL
NJ
84.9
Scan
dina
vian
sN
YM
NIL
80.4
NY
MN
IL77
.9N
YIL
MN
80.1
Rus
sian
san
dot
hers
NY
PAIL
93.1
NY
PAIL
90.6
NY
PAIL
90.9
Oth
erEu
rope
ans
PAN
YIL
88.1
NY
PAO
H83
.6N
YPA
OH
87.5
Oth
erco
untr
ies
HI
CA
NY
88.9
CA
HI
NY
85.8
CA
NY
HI
83.9
30
Tabl
e2.
Occ
upat
iona
ldis
trib
utio
nof
imm
igra
nts
byet
hnic
grou
p,ov
erti
me,
Cen
sus
and
adm
inis
trat
ive
data
cont
rast
ed
Stoc
ks19
00Fl
ows
1901
-191
0Fl
ows
1921
-193
01s
t2n
d3r
dTO
P10
1st
2nd
3rd
TOP
101s
t2n
d3r
dTO
P10
Pane
lA
:Fro
mC
ensu
sda
ta
Brit
ish
ance
stry
Labo
rers
Farm
ers
Serv
ants
68.1
Labo
rers
Serv
ants
Cle
rks
69.4
Labo
rers
Serv
ants
Cle
rks
72.3
Fren
chLa
bore
rsTe
xtile
wrk
ers
Farm
ers
75.9
Labo
rers
Farm
labo
rers
Serv
ants
73.0
Labo
rers
Serv
ants
Farm
labo
rers
68.9
Sout
hEu
rope
ans
Labo
rers
Min
ers
Man
ufac
ture
rs82
.5La
bore
rsM
iner
sFa
rmla
bore
rs84
.8La
bore
rsM
anuf
actu
rers
Min
ers
72.7
His
pani
csLa
bore
rsFa
rmla
bore
rsSe
rvan
ts84
.0La
bore
rsFa
rmla
bore
rsM
iner
s90
.6La
bore
rsFa
rmla
bore
rsSe
rvan
ts87
.2G
erm
anic
Labo
rers
Farm
ers
Man
ufac
ture
rs71
.0La
bore
rsM
iner
sSe
rvan
ts79
.7La
bore
rsSe
rvan
tsM
achi
nist
s64
.1Sc
andi
navi
ans
Farm
ers
Labo
rers
Serv
ants
77.4
Labo
rers
Serv
ants
Farm
labo
rers
79.6
Labo
rers
Serv
ants
Cle
rks
70.7
Rus
sian
san
dot
hers
Labo
rers
Tailo
rsM
anuf
actu
rers
82.8
Labo
rers
Min
ers
Tailo
rs81
.2La
bore
rsM
erch
ants
Man
ufac
ture
rs72
.3O
ther
Euro
pean
sLa
bore
rsFa
rmer
sM
iner
s83
.0La
bore
rsM
iner
sSe
rvan
ts85
.5La
bore
rsSe
rvan
tsH
otel
keep
ers
73.0
Oth
erco
untr
ies
Labo
rers
Farm
labo
rers
Oth
erun
skill
ed94
.3La
bore
rsFa
rmla
bore
rsSe
rvan
ts89
.3Fa
rmla
bore
rsH
otel
keep
ers
Oth
erun
skill
ed89
.1
Pane
lB
:Fro
mad
min
istr
ativ
eda
ta
Brit
ish
ance
stry
Serv
ants
Labo
rers
Cle
rks
75.3
Serv
ants
Labo
rers
Cle
rks
72.3
Fren
chLa
bore
rsSe
rvan
tsFa
rmla
bore
rs71
.9La
bore
rsSe
rvan
tsFa
rmer
s75
.6So
uth
Euro
pean
sLa
bore
rsFa
rmla
bore
rsSe
rvan
ts91
.7La
bore
rsSe
rvan
tsFa
rmla
bore
rs84
.8H
ispa
nics
Labo
rers
Tobb
aco
wke
rsSe
rvan
ts82
.6La
bore
rsSe
rvan
tsC
lerk
s89
.1G
erm
anic
Serv
ants
Labo
rers
Farm
labo
rers
75.3
Serv
ants
Farm
ers
Cle
rks
72.8
Scan
dina
vian
sLa
bore
rsSe
rvan
tsFa
rmla
bore
rs91
.0Se
rvan
tsFa
rmer
sFa
rmla
bore
rs80
.8R
ussi
ans
and
othe
rsLa
bore
rsFa
rmla
bore
rsSe
rvan
ts90
.3Se
rvan
tsLa
bore
rsTa
ilors
78.1
Oth
erEu
rope
ans
Labo
rers
Farm
labo
rers
Serv
ants
95.7
Labo
rers
Serv
ants
Farm
labo
rers
87.5
Oth
erco
untr
ies
Farm
labo
rers
Labo
rers
Mer
chan
ts88
.2M
erch
ants
Labo
rers
Serv
ants
77.1
31
Table 3. Summary Statistics
Variable Mean Standard deviation
Data set by ethnicity-state-period (N=7140)
Flow from administrative data 2071.3 11806.9Flow from IPUMS 729.5 3910.9Pred. flow based on ethn. (Admin shares, admin flow) 2071.3 13511.3Pred. flow based on ethn. (Census shares, admin flow) 1998.4 10636.9Pred. flow based on ethn. (IPUMS shares, admin flow) 2071.3 10637.6Pred. flow based on ethn. (Admin shares, IPUMS flow) 729.5 5047.0Pred. flow based on ethn. (Census shares, IPUMS flow) 716.5 3913.4Pred. flow based on ethn. (IPUMS shares, IPUMS flow) 729.5 3877.3Pred. flow based on occ. (Admin shares, admin flow) 1423.1 7796.4Pred. flow based on occ. (Census shares, admin flow) 1426.2 3722.1Pred. flow based on occ. (IPUMS shares, admin flow) 1426.2 5008.0Pred. flow based on occ. (Admin shares, IPUMS flow) 717.3 4095.3Pred. flow based on occ. (Census shares, IPUMS flow) 719.4 2242.0Pred. flow based on occ. (IPUMS shares, IPUMS flow) 719.4 2908.4Pred. flow based on ethn.-occ. (admin flow) 1410.0 7139.4Pred. flow based on ethn.-occ. (IPUMS flow) 714.4 3931.5Pred. flow based on ethn.*occ. score diff. (admin flow) 17186.1 98064.1Pred. flow based on ethn.*occ. score diff. (IPUMS flow) 6166.7 31332.0
Data set by occupation-state-period(N=15810)
Flow from administrative data 643.4 6534.4Flow from IPUMS 324.9 2940.5Pred. flow based on ethn. (Admin shares, admin flow) 644.1 7422.1Pred. flow based on ethn. (Census shares, admin flow) 616.3 5377.0Pred. flow based on ethn. (IPUMS shares, admin flow) 644.1 5664.0Pred. flow based on ethn. (Admin shares, IPUMS flow) 324.9 3889.7Pred. flow based on ethn. (Census shares, IPUMS flow) 319.1 2889.4Pred. flow based on ethn. (IPUMS shares, IPUMS flow) 324.9 2911.3Pred. flow based on occ. (Admin shares, admin flow) 642.0 7107.9Pred. flow based on occ. (Census shares, admin flow) 643.4 3739.7Pred. flow based on occ. (IPUMS shares, admin flow) 643.4 4942.8Pred. flow based on occ. (Admin shares, IPUMS flow) 324.0 3788.0Pred. flow based on occ. (Census shares, IPUMS flow) 324.9 2068.7Pred. flow based on occ. (IPUMS shares, IPUMS flow) 324.9 2778.5Pred. flow based on ethn.-occ. (admin flow) 636.8 5598.7Pred. flow based on ethn.-occ. (IPUMS flow) 322.6 3126.3Pred. flow based on ethn.*occ. score difference (admin flow) 7761.5 91167.0Pred. flow based on ethn.*occ. score difference (IPUMS flow) 2784.9 23561.5
32
Variable Mean Standard deviation
Data set by ethnicity-occupation-state-period (N=442680)
Flow from IPUMS 11.6 262.3Pred. flow based on occ. (admin flow) 23.0 349.8Pred. flow based on ethn. (admin flow) 23.0 516.7Pred. flow based on ethn.-occ. (admin flow) 22.7 587.8Pred. flow based on occ. (IPUMS flow) 11.6 196.6Pred. flow based on ethn. (IPUMS flow) 11.6 255.0Pred. flow based on ethn.-occ. (IPUMS flow) 11.5 278.9Pred. flow based on ethn.*occ. score difference (admin flow) 277.2 7925.4Pred. flow based on ethn.*occ. score difference (IPUMS flow) 99.5 2142.7
33
Table 4. Explaining changes in location choices of immigrants (excluding individuals with nooccupation)
From administrative data From IPUMS dataAdmin Census IPUMS Admin Census IPUMS
(1) (2) (3) (4) (5) (6)
Panel A: By ethnicity (N=7140)
Pred. flow by ethnicity 0.623*** 0.916*** 0.838*** 0.273* 0.590*** 0.539***(0.085) (0.048) (0.089) (0.108) (0.076) (0.098)
Pred. flow by occupation 0.390** 0.641*** 0.537*** 0.540*** 0.794*** 0.626**(0.130) (0.132) (0.127) (0.146) (0.229) (0.192)
R-square 0.963 0.947 0.936 0.907 0.929 0.930
Panel B: By occupation (N=15810)
Pred. flow by ethnicity 0.549*** 1.124*** 1.163*** 0.121 0.437*** 0.102(0.150) (0.170) (0.238) (0.108) (0.092) (0.104)
Pred. flow by occupation 0.344* 0.109 -0.166 0.564** 0.835*** 0.918***(0.150) (0.141) (0.226) (0.185) (0.174) (0.141)
R-square 0.969 0.937 0.930 0.931 0.944 0.960The left-hand side variable of the regressions presented in this table is the flow of immigrantsfrom a particular ethnicity/occupation electing a particular state in a particular period of migra-tion and the right-hand side variables are predicted flows as detailed in the left-hand column.Actual and predicted flows are measured from the administrative data (ex-ante) in the firstthree columns but from the IPUMS (ex-post) in the last three. In columns (1) and (4), thesepredicted flows are built using location shares from the administrative data, in columns (2) and(5), from the 1900 Census tables and in columns (3) and (6), from the 1900 IPUMS. All regres-sions include fixed effects for the double interactions of state, ethnicity/occupation and periodof immigration.Standard errors are clustered at the group-state level.*: 5% significance, **: 1% significance, ***: 0.1% significance
34
Table 5. Robustness checks: using close ethnic groups as an alternative
From administrative data From IPUMS dataAdmin Census IPUMS Admin Census IPUMS
(1) (2) (3) (4) (5) (6)
Panel A: By ethnicity (N=6885)
Pred. flow by ethnicity 0.625*** 0.898*** 0.839*** 0.274* 0.577*** 0.536***(0.086) (0.057) (0.088) (0.110) (0.075) (0.097)
Pred. flow by occupation 0.378** 0.604*** 0.506*** 0.522** 0.775** 0.620**(0.133) (0.131) (0.123) (0.166) (0.253) (0.215)
Pred. flow by close eth. grp 0.015 0.078 0.025 0.035 0.059 0.018(0.008) (0.050) (0.052) (0.051) (0.040) (0.052)
R-square 0.964 0.949 0.937 0.909 0.930 0.930
Panel B: By occupation (N=15810)
Pred. flow by ethnicity 0.541*** 1.063*** 1.309*** 0.110 0.370*** 0.120(0.150) (0.163) (0.234) (0.100) (0.086) (0.092)
Pred. flow by occupation 0.324* 0.116 -0.124 0.501** 0.813*** 0.983***(0.139) (0.139) (0.227) (0.184) (0.180) (0.119)
Pred. flow by close eth. grp 0.034 0.083 -0.223* 0.122* 0.159 -0.136(0.029) (0.090) (0.099) (0.051) (0.111) (0.095)
R-square 0.969 0.937 0.933 0.933 0.946 0.962The left-hand side variable of the regressions presented in this table is the flow of immigrantsfrom a particular ethnicity (Panels A) or a particular occupation (Panels C) electing a particularstate in a particular period of migration and the right-hand side variables are predicted flows asdetailed by the list. Actual and predicted flows are measured from the administrative data (ex-ante) in the first three columns but from the IPUMS (ex-post) in the last three. In columns (1)and (4), predicted flows are built using location shares from the administrative data, in columns(2) and (5), from the 1900 Census tables and in columns (3) and (6), from the 1900 IPUMS. Allregressions include fixed effects for the double interactions of state, ethnicity/occupation andperiod of immigration.Standard errors are clustered at the group-state level.*: 5% significance, **: 1% significance, ***: 0.1% significance
35
Table 6. Explaining changes in location choices of immigrants, networks more finely defined
From administrative data From IPUMS data
(1) (2) (3) (4)
Panel A: By ethnicity (N=7140)Pred. flow by ethnicity 1.226*** 2.541*** -0.609** -0.782**
(0.174) (0.381) (0.231) (0.275)Pred. flow by ethnicity*occ. score diff. -0.097*** 0.023
(0.028) (0.014)Pred. flow by occupation 0.613*** 0.615*** 0.275*** 0.278***
(0.165) (0.152) (0.072) (0.072)Pred. flow by ethn.-occupation -0.225 -0.264 1.346*** 1.338***
(0.148) (0.139) (0.240) (0.239)R-square 0.929 0.938 0.953 0.953
Panel B: By occupation (N=15810)Pred. flow by ethnicity 1.193*** 1.409*** -0.182 -0.196
(0.245) (0.217) (0.103) (0.125)Pred. flow by ethnicity*occ. score diff. -0.012 0.002
(0.014) (0.006)Pred. flow by occupation -0.072 -0.076 0.561*** 0.561***
(0.155) (0.155) (0.099) (0.100)Pred. flow by ethn.-occupation -0.104 -0.143 0.536*** 0.636***
(0.120) (0.103) (0.115) (0.116)R-square 0.937 0.939 0.966 0.966
Panel C: Joint estimation (N=22950)Pred. flow by ethnicity 1.249*** 1.682*** -0.236 -0.315
(0.210) (0.270) (0.145) (0.193)Pred. flow by ethnicity*occ. score diff. -0.026 0.011
(0.018) (0.008)Pred. flow by occupation 0.136 0.129 0.323*** 0.323***
(0.118) (0.116) (0.061) (0.062)Pred. flow by ethn.-occupation -0.179 -0.249* 0.839*** 0.836***
(0.139) (0.103) (0.126) (0.127)R-square 0.924 0.927 0.954 0.954
Panel D: By ethnicity and occupation (N=442680)Pred. flow by ethnicity -0.212 -0.200
(0.116) (0.125)Pred. flow by ethnicity*occ. score diff. -0.002
(0.007)Pred. flow by occupation 0.159* 0.159*
(0.065) (0.065)Pred. flow by ethn.-occupation 0.891*** 0.891***
(0.101) (0.101)R-square 0.747 0.745The left-hand side variable of the regressions presented in this table is the flow of immigrants from a particularethnicity/occupation electing a particular state in a particular period of migration and the right hand side variablesare predicted flows constructed as detailed by the left-hand column using shares from 1900 IPUMS. Actual andpredicted flows are measured from the administrative data (first two columns) and from IPUMS (last two). Allregressions include fixed effects for the double interactions of state, group and period of immigration for the firstthree panels and for all triple interactions of state, ethnicity, occupation and period of immigration for panel D.Standard errors are clustered at the group-state level.*: 5% significance, **: 1% significance, ***: 0.1% significance
36
Table 7. Robustness checks: networks more finely defined
From administrative data From IPUMS dataLarge Bef. 1920 Excl. NY Large Bef. 1920 Excl. NY
(1) (2) (3) (4) (5) (6)
Panel A: By ethnicityPred. flow by ethnicity 1.195*** 1.188*** 0.948*** -0.645** -0.484* -0.529*
(0.178) (0.206) (0.220) (0.228) (0.250) (0.246)Pred. flow by occupation 0.651*** 0.740*** 0.461*** 0.254** 0.295*** 0.290***
(0.173) (0.221) (0.139) (0.077) (0.079) (0.074)Pred. flow by ethn.-occupation -0.231 -0.236 -0.237 1.409*** 1.218*** 1.255***
(0.138) (0.181) (0.128) (0.242) (0.250) (0.255)R-square 0.938 0.939 0.846 0.958 0.965 0.927N 3570 4284 7000 3570 4284 7000
Panel B: By occupationPred. flow by ethnicity 1.185*** 1.200*** 1.525*** -0.178 -0.181 -0.032
(0.244) (0.279) (0.252) (0.102) (0.110) (0.081)Pred. flow by occupation -0.070 -0.053 -0.225 0.557*** 0.451*** 0.407***
(0.154) (0.195) (0.138) (0.099) (0.123) (0.093)Pred. flow by ethn.-occupation -0.103 -0.093 -0.209 0.535*** 0.664*** 0.627***
(0.117) (0.140) (0.112) (0.115) (0.139) (0.102)R-square 0.938 0.951 0.920 0.967 0.974 0.954N 7905 9486 15500 7905 9486 15500
Panel C: By ethnicity and occupationPred. flow by ethnicity -0.215 -0.278* -0.265*
(0.118) (0.121) (0.131)Pred. flow by occupation 0.157* 0.110 0.214***
(0.065) (0.067) (0.070)Pred. flow by ethn.-occupation 0.894*** 0.971*** 0.902***
(0.101) (0.104) (0.123)R-square 0.746 0.744 0.658N 244290 265608 434000The left-hand side variable of the regressions presented in this table is the flow of immigrants from a partic-ular ethnicity/occupation electing a particular state in a particular period of migration and the right handside variables are predicted flows constructed as detailed by the left-hand column using shares from 1900IPUMS. Actual and predicted flows are measured from the administrative data (first three columns) andfrom IPUMS (last three). All regressions include fixed effects for the double interactions of state, group andperiod of immigration for the first two panels and for all triple interactions of state, ethnicity, occupationand period of immigration for panel C.Standard errors are clustered at the group-state level.*: 5% significance, **: 1% significance, ***: 0.1% significance
37
Table 8. Exploring the differences between ex-ante and ex-post regressions, by occupation.
High occ. Low occ. Traded Non- Within Betweenscore score traded 5 years 5-10 years
(1) (2) (3) (4) (5) (6)
Panel A: Flows from administrative data
Pred. flow by ethnicity 1.066** 1.186*** 1.205*** 0.428 1.239*** 1.132***(0.342) (0.243) (0.242) (0.397) (0.303) (0.242)
Pred. flow by occupation 0.130 -0.070 -0.135 0.224 -0.097 -0.102(0.194) (0.154) (0.154) (0.179) (0.206) (0.147)
Pred. flow by ethn.-occupation 0.119 -0.102 -0.101 0.405* -0.103 -0.108(0.200) (0.118) (0.110) (0.205) (0.154) (0.101)
R-square 0.922 0.938 0.933 0.983 0.905 0.954N 7905 7905 8925 6885 9486 6324
Panel B: Flows from IPUMS data
Pred. flow by ethnicity 0.165 -0.182 -0.186 0.072 -0.175 -0.186(0.146) (0.102) (0.103) (0.153) (0.121) (0.147)
Pred. flow by occupation 0.764*** 0.558*** 0.553*** 0.764*** 0.516*** 1.007**(0.213) (0.099) (0.099) (0.112) (0.122) (0.348)
Pred. flow by ethn.-occupation 0.215 0.536*** 0.542*** 0.250* 0.573*** 0.156(0.158) (0.116) (0.117) (0.126) (0.154) (0.252)
R-square 0.885 0.967 0.966 0.962 0.966 0.967N 7905 7905 8925 6885 9486 6324The left-hand side variable of the regressions presented in this table is the flow of immigrants from aparticular occupation electing a particular state in a particular period of migration and the right handside variables are predicted flows constructed as detailed by the left-hand column using shares from1900 IPUMS. Actual and predicted flows are measured from the administrative data (Panel A) and fromIPUMS (Panel B). The first 2 columns divide the sample by the occupational score of each occupation.Columns (3) and (4) compare occupations classified as traded and non-traded as detailed in Table A-1.Column (5) restricts the sample to immigrants observed in the Census within 5 years of their arrivalto the United States while the last includes individuals observed in the Census 5-10 years after theirarrival. All regressions include fixed effects for the double interactions of state, occupation and periodof immigration.Standard errors are clustered at the occupation-state level.*: 5% significance, **: 1% significance, ***: 0.1% significance
38
Table 9. Exploring the differences between ex-ante and ex-post regressions, by ethnicity.
English Non Low High Within BetweenEnglish return return 5 years 5-10 years
(1) (2) (3) (4) (5) (6)
Panel A: Flows from administrative data
Pred. flow by ethnicity -0.424 1.222*** 1.264*** 0.943*** 1.243*** 1.203***(1.156) (0.177) (0.209) (0.182) (0.224) (0.194)
Pred. flow by occupation 0.844 0.618*** 0.710** 0.154 0.686** 0.404*(0.509) (0.170) (0.228) (0.130) (0.213) (0.187)
Pred. flow by ethn.-occupation 1.263 -0.223 -0.311 0.216 -0.240 -0.214(1.152) (0.148) (0.172) (0.175) (0.211) (0.125)
R-square 0.964 0.928 0.935 0.925 0.886 0.949N 1020 6120 3570 3570 4284 2856
Panel B: Flows from IPUMS
Pred. flow by ethnicity -0.724* -0.593* -0.688** -0.230 -0.683** -0.460(0.362) (0.244) (0.235) (0.397) (0.256) (0.436)
Pred. flow by occupation -0.317** 0.291*** 0.256*** 0.234 0.165* 0.634***(0.096) (0.073) (0.070) (0.176) (0.069) (0.142)
Pred. flow by ethn.-occupation 2.043*** 1.315*** 1.445*** 0.967* 1.503*** 0.903*(0.358) (0.253) (0.246) (0.425) (0.264) (0.450)
R-square 0.949 0.956 0.970 0.893 0.954 0.949N 1020 6120 3570 3570 4284 2856The left-hand side variable of the regressions presented in this table is the flow of immigrants from aparticular ethnicity electing a particular state in a particular period of migration and the right handside variables are predicted flows constructed as detailed by the left-hand column using shares from1900 IPUMS. Actual and predicted flows are measured from the administrative data (Panel A) and fromIPUMS (Panel B). The first 2 columns divide the sample by whether the ethnicity is English speakingor not. Columns (3) and (4) compare ethnicities by the share of return migration observed over thisentire period. Column (5) restricts the sample to immigrants observed in the Census within 5 years oftheir arrival to the United States while the last includes individuals observed in the Census 5-10 yearsafter their arrival. All regressions include fixed effects for the double interactions of state, ethnicity andperiod of immigration.Standard errors are clustered at the ethnic-state level.*: 5% significance, **: 1% significance, ***: 0.1% significance
39
Table 10. Explaining location choice using the “wrong” predicted flows
Ethnicity Occupation Joint Ethn. and Occ.
(1) (2) (3) (4)
Panel A: Flows from administrative data
Pred. flow by ethnicity 1.786 1.579* 1.885*(1.380) (0.722) (0.770)
Pred. flow by occupation -0.015 0.780** -0.024(0.538) (0.266) (0.525)
Pred. flow by ethn.-occupation 0.435 -0.850 -0.154(1.074) (0.778) (0.656)
R-square 0.830 0.681 0.756N 7140 15810 22950
Panel B: Flows from IPUMS data
Pred. flow by ethnicity 0.205*** 0.041 0.090 0.050(0.052) (0.096) (0.088) (0.046)
Pred. flow by occupation 0.255** 0.219 0.232** 0.142***(0.083) (0.115) (0.078) (0.038)
Pred. flow by ethn.-occupation -0.004 0.113 0.086 0.049(0.049) (0.132) (0.083) (0.029)
R-square 0.758 0.772 0.763 0.111N 7140 15810 22950 442680The left-hand side variable of the regressions presented in this table is the flow of im-migrants from a particular occupation electing a particular state in a particular period ofmigration and the right hand side variables are predicted flows constructed as detailed bythe left-hand column using shares from 1900 IPUMS. Actual flows are from the admin-istrative data and the predicted flows from IPUMS in Panel A and vice versa in Panel B.Column (1) represents the regression computed on the sample by ethnicity, column (2) onthe one by occupation. Column (3) corresponds to the regressions on both ethnicity andoccupation cells are estimated simultaneously. The final column includes all ethnicity-occupation combinations as observations. All regressions include fixed effects for thedouble interactions of state, group and period of immigration, Column (4) also include alltriple interactions between state, ethnicity, occupation and period of immigration.Standard errors are clustered at the group-state level.*: 5% significance, **: 1% significance, ***: 0.1% significance
40
Tabl
eA
-1.M
atch
ing
ofoc
cupa
tion
sbe
twee
nad
min
istr
ativ
ean
dC
ensu
sda
ta
Adm
inis
trat
ive
Dat
a19
50O
ccup
atio
nC
lass
ifica
tion
1900
Occ
upat
ion
Cla
ssifi
cati
onLa
rge?
Trad
ed?
Cat
egor
ies
Cod
eD
escr
ipti
on
TO
TAL
PRO
FESS
ION
AL
Elec
tric
ians
515
Elec
tric
ians
Elec
tric
ians
No
No
603
App
rent
ice
elec
tric
ians
Engi
neer
s(p
rofe
ssio
nal)
41En
gine
ers,
aero
naut
ical
Engi
neer
s(c
ivil,
etc)
and
surv
eyor
sYe
sN
o42
Engi
neer
s,ch
emic
alD
esig
ners
,dra
ught
smen
and
inve
ntor
s43
Engi
neer
s,ci
vil
44En
gine
ers,
elec
tric
al45
Engi
neer
s,in
dust
rial
46En
gine
ers,
mec
hani
cal
47En
gine
ers,
met
allu
rgic
al,m
etal
lurg
ists
48En
gine
ers,
min
ing
49En
gine
ers
(n.e
.c.)
35D
raft
smen
92Su
rvey
ors
Scul
ptor
san
dar
tist
s4
Art
ists
and
art
teac
hers
Art
ists
and
teac
hers
ofar
tN
oN
o31
Dan
cers
and
danc
ing
teac
hers
Lite
rary
and
scie
ntifi
cpe
rson
s6
Aut
hors
Lite
rary
and
scie
ntifi
cpe
rson
sN
oN
o56
Libr
aria
ns30
1A
tten
dant
san
das
sist
ants
,lib
rary
7C
hem
ists
61A
gric
ultu
rals
cien
tist
s62
Biol
ogic
alsc
ient
ists
63G
eolo
gist
san
dge
ophy
sici
sts
67M
athe
mat
icia
ns68
Phys
icis
ts69
Mis
cella
neou
sna
tura
lsci
enti
sts
83St
atis
tici
ans
and
actu
arie
s
Act
ors
1A
ctor
san
dac
tres
ses
Act
ors
No
No
Thea
tret
ical
man
ager
s,et
c
Mus
icia
ns51
Ente
rtai
ners
(n.e
.c.)
Mus
icia
nsan
dte
ache
rsof
mus
icYe
sN
o57
Mus
icia
nsan
dm
usic
teac
hers
Prof
essi
onal
show
men
Teac
hers
12A
gric
ultu
rals
cien
ces
Teac
hers
and
prof
esso
rsin
colle
ges,
etc
Yes
No
13Bi
olog
ical
scie
nces
14C
hem
istr
y15
Econ
omic
s16
Engi
neer
ing
17G
eolo
gyan
dge
ophy
sics
18M
athe
mat
ics
19M
edic
alsc
ienc
es23
Phys
ics
24Ps
ycho
logy
25St
atis
tics
26N
atur
alsc
ienc
e(n
.e.c
.)27
Soci
alsc
ienc
es(n
.e.c
.)28
Non
scie
ntifi
csu
bjec
ts
41
Adm
inis
trat
ive
Dat
a19
50O
ccup
atio
nC
lass
ifica
tion
1900
Occ
upat
ion
Cla
ssifi
cati
onLa
rge?
Trad
ed?
Cat
egor
ies
Cod
eD
escr
ipti
on
29Su
bjec
tno
tsp
ecifi
ed93
Teac
hers
(n.e
.c.)
Cle
rgy
9C
lerg
ymen
Cle
rgym
enN
oN
o
Offi
cial
s(G
over
nmen
t)21
0In
spec
tors
,pub
licad
min
istr
atio
nO
ffici
als
(gov
ernm
ent)
No
No
250
Offi
cial
san
dad
min
istr
ator
s(n
.e.c
.),pu
blic
adm
inis
trat
ion
Sold
iers
,sai
lors
and
mar
ines
(U.S
.)27
0Po
stm
aste
rs59
5M
embe
rsof
the
arm
edse
rvic
es77
1M
arsh
als
and
cons
tabl
es78
2Sh
eriff
san
dba
iliff
s
Phys
icia
ns75
Phys
icia
nsan
dsu
rgeo
nsPh
ysic
ians
and
surg
eons
No
No
Arc
hite
cts
3A
rchi
tect
sA
rchi
tect
sN
oN
o
Edit
ors
36Ed
itor
san
dre
port
ers
Jour
nalis
tsN
oN
o
Law
yers
55La
wye
rsan
dju
dges
Law
yers
No
No
Oth
erpr
ofes
sion
als
2A
irpl
ane
pilo
tsan
dna
viga
tors
Den
tist
sYe
sN
o5
Ath
lete
sO
ther
prof
essi
onal
serv
ice
8C
hiro
prac
tors
Nuu
rses
(tra
ined
)32
Den
tist
sN
urse
s(n
otsp
ecifi
ed)
70O
ptom
etri
sts
Und
erta
kers
71O
steo
path
s10
Col
lege
pres
iden
tsan
dde
ans
33D
esig
ners
34D
ieti
cian
san
dnu
trit
ioni
sts
52Fa
rman
dho
me
man
agem
ent
advi
sors
53Fo
rest
ers
and
cons
erva
tion
ists
54Fu
nera
ldir
ecto
rsan
dem
balm
ers
58N
urse
s,pr
ofes
sion
al59
Nur
ses,
stud
ent
prof
essi
onal
72Pe
rson
nela
ndla
bor
rela
tion
sw
orke
rs73
Phar
mac
ists
76R
adio
oper
ator
s77
Rec
reat
ion
and
grou
pw
orke
rs78
Rel
igio
usw
orke
rs79
Soci
alan
dw
elfa
rew
orke
rs,e
xcep
tgr
oup
81Ec
onom
ists
82Ps
ycho
logi
sts
84M
isce
llane
ous
soci
alsc
ient
ists
91Sp
orts
inst
ruct
ors
and
offic
ials
94Te
chni
cian
s,m
edic
alan
dde
ntal
95Te
chni
cian
s,te
stin
g96
Tech
nici
ans
(n.e
.c.)
97Th
erap
ists
and
heal
ers
(n.e
.c.)
98Ve
teri
nari
ans
99Pr
ofes
sion
al,t
echn
ical
and
kind
red
wor
kers
(n.e
.c.)
260
Offi
cial
s,lo
dge,
soci
ety,
unio
n,et
c.53
2In
spec
tors
,sca
lers
,and
grad
ers,
log
and
lum
ber
533
Insp
ecto
rs(n
.e.c
.)78
1Pr
acti
caln
urse
s
42
Adm
inis
trat
ive
Dat
a19
50O
ccup
atio
nC
lass
ifica
tion
1900
Occ
upat
ion
Cla
ssifi
cati
onLa
rge?
Trad
ed?
Cat
egor
ies
Cod
eD
escr
ipti
on
TO
TAL
SKIL
LED
Barb
ers
and
hair
dres
sers
740
Barb
ers,
beau
tici
ans,
and
man
icur
ists
Barb
ers
and
hair
dres
sers
Yes
No
Cle
rks
and
acco
unta
nts
0A
ccou
ntan
tsan
dau
dito
rsBo
okee
pers
and
acco
unta
nts
Yes
No
302
Att
enda
nts,
phys
icia
n’s
and
dent
ist’s
offic
eC
lerk
san
dco
ypis
ts31
0Bo
okke
eper
sM
esse
nger
san
der
rand
and
offic
ebo
ys32
0C
ashi
ers
Sten
ogra
pher
san
dty
pew
rite
rs32
1C
olle
ctor
s,bi
llan
dac
coun
t32
5Ex
pres
sm
esse
nger
san
dra
ilway
mai
lcle
rks
335
Mai
lcar
rier
s34
0M
esse
nger
san
dof
fice
boys
341
Offi
cem
achi
neop
erat
ors
342
Ship
ping
and
rece
ivin
gcl
erks
350
Sten
ogra
pher
s,ty
pist
s,an
dse
cret
arie
s36
0Te
legr
aph
mes
seng
ers
365
Tele
grap
hop
erat
ors
370
Tele
phon
eop
erat
ors
390
Cle
rica
land
kind
red
wor
kers
(n.e
.c.)
450
Insu
ranc
eag
ents
and
brok
ers
520
Elec
trot
yper
san
dst
ereo
type
rs
Bank
ers
204
Cre
dit
men
Bank
ers
and
brok
ers
No
No
305
Bank
telle
rsO
ffici
als
ofba
nks
and
com
pani
es48
0St
ock
and
bond
sale
smen
Gar
dene
rs93
0G
arde
ners
,exc
ept
farm
,and
grou
ndsk
eepe
rsG
arde
ners
,flor
ists
,nur
sery
men
,etc
Yes
No
Gar
den
and
nurs
ery
labo
rers
Wat
chan
dcl
ock
mak
ers
534
Jew
eler
s,w
atch
mak
ers,
gold
smit
hs,a
ndsi
lver
smit
hsC
lock
and
wat
chm
aker
san
dre
pair
ers
Yes
Yes
Jew
eler
sG
old
and
silv
erw
orke
rs
Car
pent
ers
and
join
ers
510
Car
pent
ers
Car
pent
ers
and
join
ers
Yes
Yes
Ship
wri
ghts
602
App
rent
ice
carp
ente
rs
Cab
inet
mak
ers
505
Cab
inet
mak
ers
Cab
inet
mak
ers
No
Yes
Woo
dwor
kers
(not
spec
ified
)67
4Sa
wye
rsC
oope
rsN
oYe
sSa
wan
dpl
anin
gm
illem
ploy
ees
Oth
erw
ood
wor
kers
Plum
bers
574
Plum
bers
and
pipe
fitte
rsPl
umbe
rsan
dga
san
dst
eam
fitte
rsN
oN
o61
0A
ppre
ntic
epl
umbe
rsan
dpi
pefit
ters
Pain
ters
and
glaz
iers
530
Gla
zier
sPa
inte
rs,g
lazi
ers
and
varn
ishe
rsYe
sYe
s56
4Pa
inte
rs,c
onst
ruct
ion
and
mai
nten
ance
670
Pain
ters
,exc
ept
cons
truc
tion
orm
aint
enan
ce
Mas
ons
504
Bric
kmas
ons,
ston
emas
ons,
and
tile
sett
ers
Mas
ons
Yes
Yes
601
App
rent
ice
bric
klay
ers
and
mas
ons
Ston
ecut
ters
584
Ston
ecu
tter
san
dst
one
carv
ers
Mar
ble
and
ston
ecu
tter
sYe
sYe
s63
5Fi
lers
,gri
nder
s,an
dpo
lishe
rs,m
etal
Blac
ksm
iths
501
Blac
ksm
iths
Blac
ksm
iths
Yes
Yes
524
Forg
emen
and
ham
mer
men
43
Adm
inis
trat
ive
Dat
a19
50O
ccup
atio
nC
lass
ifica
tion
1900
Occ
upat
ion
Cla
ssifi
cati
onLa
rge?
Trad
ed?
Cat
egor
ies
Cod
eD
escr
ipti
on
Engi
neer
s20
3C
ondu
ctor
s,ra
ilroa
dC
ondu
ctor
s(s
team
railr
oad)
Yes
No
541
Loco
mot
ive
engi
neer
sEn
gine
ers
and
firem
en58
3St
atio
nary
engi
neer
sEn
gine
ers
and
firem
en(n
otra
ilroa
d)
Iron
and
stee
lwor
kers
503
Boile
rmak
ers
Iron
and
stee
lwor
kers
Yes
Yes
Met
alw
orke
rs53
1H
eat
trea
ters
,ann
eale
rs,t
empe
rers
Stea
mbo
iler
mak
ers
535
Job
sett
ers,
met
alSt
ove,
furn
ace
and
grat
em
aker
s56
1M
olde
rs,m
etal
Wir
ew
orke
rs58
0R
olle
rsan
dro
llha
nds,
met
alBr
ass
wor
kers
585
Stru
ctur
alm
etal
wor
kers
Oth
erm
etal
wor
kers
612
App
rent
ices
,met
alw
orki
ngtr
ades
(n.e
.c.)
642
Hea
ters
,met
al68
5W
elde
rsan
dfla
me
cutt
ers
Stok
ers
641
Furn
acem
en,s
mel
term
enan
dpo
urer
sC
harc
oal,
coke
and
lime
burn
ers
No
Yes
Mac
hini
sts
544
Mac
hini
sts
Mac
hini
sts
Yes
Yes
604
App
rent
ice
mac
hini
sts
and
tool
mak
ers
Whe
elw
righ
ts54
5M
echa
nics
and
repa
irm
en,a
irpl
ane
Whe
elw
righ
tsYe
sYe
sM
echa
nics
550
Mec
hani
csan
dre
pair
men
,aut
omob
ileM
echa
nics
(nec
)Lo
cksm
iths
551
Mec
hani
csan
dre
pair
men
,offi
cem
achi
ne55
2M
echa
nics
and
repa
irm
en,r
adio
and
tele
visi
on55
3M
echa
nics
and
repa
irm
en,r
ailr
oad
and
car
shop
554
Mec
hani
csan
dre
pair
men
(n.e
.c.)
600
App
rent
ice
auto
mec
hani
cs60
5A
ppre
ntic
em
echa
nics
,exc
ept
auto
Prin
ters
512
Com
posi
tors
and
type
sett
ers
Prin
ters
,lit
hogr
aphe
rsan
dpr
essm
enN
oYe
sTi
nner
s57
5Pr
essm
enan
dpl
ate
prin
ters
,pri
ntin
g61
3A
ppre
ntic
es,p
rint
ing
trad
esTi
npla
tean
dti
nwar
em
aker
s59
1Ti
nsm
iths
,cop
pers
mit
hs,a
ndsh
eet
met
alw
orke
rs
Hat
and
Cap
mak
ers
645
Mill
iner
sH
atan
dca
pm
aker
sN
oYe
sM
illin
ers
Mill
iner
s
Seam
stre
sses
633
Dre
ssm
aker
san
dse
amst
ress
es,e
xcep
tfa
ctor
ySe
amst
ress
esYe
sYe
sD
ress
mak
ers
Dre
ssm
aker
s
Shoe
mak
ers
582
Shoe
mak
ers
and
repa
irer
s,ex
cept
fact
ory
Boot
and
shoe
mak
ers
and
repa
irer
sYe
sYe
s
Tailo
rs59
0Ta
ilors
and
tailo
ress
esTa
ilors
and
tailo
ress
esYe
sYe
s
Text
ilew
orke
rs(n
otsp
ecifi
ed)
543
Loom
fixer
sBl
each
ery
and
dye
wor
ksop
erat
ives
Yes
Yes
634
Dye
rsC
arpe
tfa
ctor
yop
erat
ives
Cot
ton
mill
oper
ativ
esH
osie
ryan
dkn
itti
gnm
illop
erat
ives
Woo
len
mill
oper
ativ
esSh
irt,
colla
r,an
dcu
ffm
aker
s
Wea
vers
and
spin
ners
675
Spin
ners
,tex
tile
Oth
erte
xtile
mill
oper
ativ
esYe
sYe
sFu
rrie
rsan
dfu
rw
orke
rs68
4W
eave
rs,t
exti
leO
ther
text
ilew
orke
rs52
5Fu
rrie
rs
Uph
olst
erer
s59
3U
phol
ster
ers
Uph
olst
erer
sN
oYe
s
Mar
iner
s24
0O
ffice
rs,p
ilots
,pur
sers
and
engi
neer
s,sh
ipBo
atm
enan
dsa
ilors
Yes
No
44
Adm
inis
trat
ive
Dat
a19
50O
ccup
atio
nC
lass
ifica
tion
1900
Occ
upat
ion
Cla
ssifi
cati
onLa
rge?
Trad
ed?
Cat
egor
ies
Cod
eD
escr
ipti
on
623
Boat
men
,can
alm
en,a
ndlo
ckke
eper
s67
3Sa
ilors
and
deck
hand
s
Bake
rs50
0Ba
kers
Bake
rsYe
sYe
s
Butc
hers
644
Mea
tcu
tter
s,ex
cept
slau
ghte
ran
dpa
ckin
gho
use
Butc
hers
Yes
Yes
Mill
ers
555
Mill
ers,
grai
n,flo
ur,f
eed,
etc.
Mill
ers
Yes
Yes
Min
ers
650
Min
eop
erat
ives
and
labo
rers
Min
ers
and
quar
rym
enYe
sYe
s
Phot
ogra
pher
s74
Phot
ogra
pher
sPh
otog
raph
ers
No
No
671
Phot
ogra
phic
proc
ess
wor
kers
Plas
tere
rs57
3Pl
aste
rers
Plas
tere
rsN
oYe
s
Book
bind
ers
502
Book
bind
ers
Book
bind
ers
No
Yes
Engr
aver
s52
1En
grav
ers,
exce
ptph
otoe
ngra
vers
Engr
aver
sN
oYe
s57
1Ph
otoe
ngra
vers
and
litho
grap
hers
Patt
ern
mak
ers
570
Patt
ern
and
mod
elm
aker
s,ex
cept
pape
rM
odel
and
patt
ern
mak
ers
No
Yes
Sadd
lers
and
harn
ess
mak
ers
511
Cem
ent
and
conc
rete
finis
hers
Har
ness
and
sadd
lem
aker
san
dre
pair
sYe
sYe
sTa
nner
san
dcu
rrie
rs51
3C
rane
men
,der
rick
men
,and
hois
tmen
Leat
her
curr
iers
and
tann
ers
Brew
ers
514
Dec
orat
ors
and
win
dow
dres
sers
Brew
ers
and
mal
ster
sC
igar
pack
ers
522
Exca
vati
ng,g
radi
ng,a
ndro
adm
achi
nery
oper
ator
sD
isti
llers
and
rect
ifier
sC
igar
ette
mak
ers
523
Fore
men
(n.e
.c.)
Toba
cco
and
ciga
rfa
ctor
yop
erat
ives
Toba
cco
wor
kers
540
Line
men
and
serv
icem
en,t
eleg
raph
,tel
epho
ne,a
ndpo
wer
Swit
chm
en,y
ardm
enan
dfla
gmen
Cig
arm
aker
s54
2Lo
com
otiv
efir
emen
Tele
grap
han
dte
leph
one
linem
enO
ther
skill
ed56
2M
otio
npi
ctur
epr
ojec
tion
ists
Dec
orat
ors
and
win
dow
dres
sers
563
Opt
icia
nsan
dle
nsgr
inde
rsan
dpo
lishe
rsW
eigh
ters
,gau
gers
,and
mea
sure
rs56
5Pa
perh
ange
rsPa
per
hang
ers
572
Pian
oan
dor
gan
tune
rsan
dre
pair
men
Roo
fers
and
slat
ers
581
Roo
fers
and
slat
ers
Bric
kan
dti
lem
aker
s59
2To
olm
aker
s,an
ddi
em
aker
san
dse
tter
sG
lass
woe
krs
594
Cra
ftsm
enan
dki
ndre
dw
orke
rs(n
.e.c
.)Po
tter
s61
1A
ppre
ntic
es,b
uild
ing
trad
es(n
.e.c
.)Bu
tter
and
chee
sem
aker
s61
4A
ppre
ntic
es,o
ther
spec
ified
trad
esC
onfe
ctio
nner
s61
5A
ppre
ntic
es,t
rade
not
spec
ified
Trun
kan
dle
athe
r-ca
sem
aker
s62
0A
sbes
tos
and
insu
lati
onw
orke
rsBo
ttle
rsan
dso
dam
aker
s62
2Bl
aste
rsan
dpo
wde
rmen
Box
mak
ers
624
Brak
emen
,rai
lroa
dBr
oom
and
brus
hm
aker
s63
0C
hain
men
,rod
men
,and
axm
en,s
urve
ying
Glo
vem
aker
s66
2O
ilers
and
grea
ser,
exce
ptau
toru
bber
fact
ory
oper
ativ
es67
2Po
wer
stat
ion
oper
ator
sTo
olan
dcu
tler
ym
aker
s68
1Sw
itch
men
,rai
lroa
d
TO
TAL
UN
SKIL
LED
Farm
labo
rers
640
Frui
t,nu
t,an
dve
geta
ble
grad
ers,
and
pack
ers
Farm
and
plan
tati
onla
bore
rsYe
sYe
s81
0Fa
rmfo
rem
enFa
rmla
bore
rs(m
embe
rsof
fam
ily)
820
Farm
labo
rers
,wag
ew
orke
rsD
airy
men
and
dair
ywom
en83
0Fa
rmla
bore
rs,u
npai
dfa
mily
wor
kers
Stoc
kra
iser
s,he
rder
san
ddr
over
s84
0Fa
rmse
rvic
ela
bore
rs,s
elf-
empl
oyed
Turp
enti
nefa
rmer
san
dla
bore
rs
Farm
ers
100
Farm
ers
(ow
ners
and
tena
nts)
Farm
ers,
plan
ters
,and
over
seer
sYe
sYe
s
45
Adm
inis
trat
ive
Dat
a19
50O
ccup
atio
nC
lass
ifica
tion
1900
Occ
upat
ion
Cla
ssifi
cati
onLa
rge?
Trad
ed?
Cat
egor
ies
Cod
eD
escr
ipti
on
123
Farm
man
ager
s
Fish
erm
en91
0Fi
sher
men
and
oyst
erm
enFi
sher
men
and
oyst
erm
enN
oYe
s
Labo
rers
690
Ope
rati
vean
dki
ndre
dw
orke
rs(n
.e.c
.)La
bore
rs(n
otsp
ecifi
ed)
Yes
Yes
920
Gar
age
labo
rers
and
car
was
hers
and
grea
sers
Labo
rers
(ste
amra
ilroa
d)95
0Lu
mbe
rmen
,raf
tsm
en,a
ndw
oodc
hopp
ers
Labo
rers
(str
eet
railw
ay)
970
Labo
rers
(n.e
.c.)
Oil
wel
land
oilw
orks
empl
oyee
s94
0Lo
ngsh
orem
enan
dst
eved
ores
Oth
erch
emic
alw
orke
rsO
ther
food
prep
arer
sO
ther
mis
cella
neou
sin
dust
ries
Serv
ants
700
Hou
seke
eper
s,pr
ivat
eho
useh
old
Hou
seke
eper
san
dst
ewar
dsYe
sN
o71
0La
undr
esss
es,p
riva
teho
useh
old
Serv
ants
720
Priv
ate
hous
ehol
dw
orke
rs(n
.e.c
.)75
3C
harw
omen
and
clea
ners
764
Hou
seke
eper
san
dst
ewar
ds,e
xcep
tpr
ivat
eho
useh
old
780
Port
ers
Age
nts
280
Purc
hasi
ngag
ents
and
buye
rs(n
.e.c
.)A
gent
sN
oN
o30
0A
gent
s(n
.e.c
.)St
atio
nag
ents
and
empl
oyee
s(s
team
railr
aod)
380
Tick
et,s
tati
on,a
ndex
pres
sag
ents
Stat
ion
agen
tsan
dem
ploy
ees
(str
eet
railr
oad)
470
Rea
lest
ate
agen
tsan
dbr
oker
s
Dra
ymen
,hac
kmen
and
team
ster
s30
4Ba
ggag
emen
,tra
nspo
rtat
ion
Dra
ymen
,hac
kmen
,tea
mst
ers
No
No
322
Dis
patc
hers
and
star
ters
,veh
icle
Bagg
agem
en62
5Bu
sdr
iver
sBr
akem
en63
1C
ondu
ctor
s,bu
san
dst
reet
railw
ayC
ondu
ctor
s(s
tree
tra
ilway
)63
2D
eliv
erym
enan
dro
utem
enD
rive
rs(s
tree
tra
ilway
)66
0M
otor
men
,min
e,fa
ctor
y,lo
ggin
gca
mp,
etc.
Mot
orm
en66
1M
otor
men
,str
eet,
subw
ay,a
ndel
evat
edra
ilway
682
Taxi
cab
driv
ers
and
chau
ffer
s68
3Tr
uck
and
trac
tor
driv
ers
960
Team
ster
s
Man
ufac
ture
rs29
0M
anag
ers,
offic
ials
,and
prop
riet
ors
(n.e
.c.)
Fore
men
and
over
seer
sN
oYe
s56
0M
illw
righ
tsM
anuf
actu
rers
and
offic
ials
Mer
chan
tsan
dde
aler
s20
0Bu
yers
and
depa
rtm
ent
head
s,st
ore
Com
mer
cial
trav
eler
sYe
sYe
s20
1Bu
yers
and
ship
pers
,far
mpr
oduc
tsH
ucks
ters
and
pedd
lers
205
Floo
rmen
and
floor
man
ager
s,st
ore
Mer
chan
tsan
dde
aler
s(e
xcep
tw
hole
sale
)40
0A
dver
tisi
ngag
ents
and
sale
smen
Mer
chan
tsan
dde
aler
s(w
hole
sale
)41
0A
ucti
onee
rsSa
lesm
enan
dsa
lesw
omen
420
Dem
onst
rato
rsA
ucti
onne
ers
430
Huc
kste
rsan
dpe
ddle
rsN
ewsp
aper
carr
iers
and
new
sboy
s46
0N
ewsb
oys
490
Sale
smen
and
sale
scl
erks
(n.e
.c.)
Hot
elke
eper
s62
1A
tten
dant
s,au
tose
rvic
ean
dpa
rkin
gBa
rten
ders
No
No
750
Bart
ende
rsBo
ardi
ngan
dlo
dgin
gho
use
keep
rs75
2Bo
ardi
ngan
dlo
dgin
gho
use
keep
ers
Hot
elke
eper
s75
4C
ooks
,exc
ept
priv
ate
hous
ehol
dR
esta
uran
tke
eper
s76
0C
ount
eran
dfo
unta
inw
orke
rsSa
loon
keep
ers
784
Wai
ters
and
wai
tres
ses
Wai
ters
Oth
erm
isce
llane
ous
230
Man
ager
san
dsu
peri
nten
dent
s,bu
ildin
gLu
mbe
rmen
and
raft
smen
Yes
No
46
Adm
inis
trat
ive
Dat
a19
50O
ccup
atio
nC
lass
ifica
tion
1900
Occ
upat
ion
Cla
ssifi
cati
onLa
rge?
Trad
ed?
Cat
egor
ies
Cod
eD
escr
ipti
on
643
Laun
dry
and
dry
clea
ning
oper
ativ
esW
ood
chop
pers
680
Stat
iona
ryfir
emen
Oth
erag
ricu
ltur
alpu
rsui
ts73
0A
tten
dant
s,ho
spit
alan
dot
her
inst
itut
ion
Jani
tors
and
sext
ons
731
Att
enda
nts,
prof
essi
onal
and
pers
onal
serv
ice
(n.e
.c.)
Laun
dere
rsan
dla
undr
esse
s73
2A
tten
dant
s,re
crea
tion
and
amus
emen
tM
idw
ives
751
Boot
blac
ksW
atch
men
,pol
icem
en,fi
rem
en,e
tc.
761
Elev
ator
oper
ator
sO
ther
dom
esti
can
dpe
rson
alse
rvic
es76
2Fi
rem
en,fi
repr
otec
tion
Hos
tler
s76
3G
uard
s,w
atch
men
,and
door
keep
ers
Live
ryst
able
keep
ers
770
Jani
tors
and
sext
ons
Pack
ers
and
ship
pers
772
Mid
wiv
esPo
rter
san
dhe
lper
s(i
nst
ores
)77
3Po
licem
enan
dde
tect
ives
Oth
erpe
rson
sin
trad
ean
dtr
ansp
orta
tion
(nec
)78
3U
sher
s,re
crea
tion
and
amus
emen
t78
5W
atch
men
(cro
ssin
g)an
dbr
idge
tend
ers
790
Serv
ice
wor
kers
,exc
ept
priv
ate
hous
ehol
d(n
.e.c
.)
TO
TAL
NO
OC
CU
PAT
ION
No
occu
pati
on98
0K
eeps
hous
e/ho
usek
eepi
ngat
hom
e/ho
usew
ife
981
Impu
ted
keep
ing
hous
e(1
850-
1900
)98
2H
elpi
ngat
hom
e/he
lps
pare
nts/
hous
ewor
k98
3A
tsc
hool
/stu
dent
984
Ret
ired
985
Une
mpl
oyed
/wit
hout
occu
pati
on98
6In
valid
/dis
able
dw
/no
occu
pati
onre
port
ed98
7In
mat
e99
1G
entl
eman
/lad
y/at
leis
ure
995
Oth
erno
n-oc
cupa
tion
alre
spon
se
47
Tabl
eA
-2.M
atch
ing
ethn
icgr
oups
and
coun
trie
s.
Ethn
icgr
oup
Ethn
icit
ies
(fro
mad
min
istr
ativ
eda
ta)
Cou
ntry
ofbi
rth
(fro
mC
ensu
s)H
igh
retu
rn?
Larg
e?
Brit
ish
ansc
estr
yEn
glis
hEn
glan
d,C
anad
a(E
nglis
h),A
ustr
alia
No
Yes
Iris
hIr
elan
dN
oYe
sSc
otch
Scot
land
No
Yes
Wel
shW
ales
No
No
Fren
chD
utch
and
Flem
mis
hBe
lgiu
m,N
ethe
rlan
dsN
oN
oFr
ench
Fran
ce,C
anad
a(F
renc
h)N
oYe
sSo
uth
Euro
pean
sIt
alia
n(N
orth
),It
alia
n(S
outh
)It
aly
Yes
Yes
Port
ugue
sePo
rtug
alYe
sN
oSp
anis
hSp
ain
Yes
No
His
pani
csM
exic
anM
exic
oN
oYe
sSp
anis
h-A
mer
ican
Cen
tral
and
Sout
hA
mer
ica
No
No
Cub
an,W
est
Indi
an,A
fric
an(B
lack
)A
llC
arri
bean
sIs
land
sYe
sN
oG
erm
ans
Ger
man
s,G
erm
anH
ebre
ws
Aus
tria
,Ger
man
y,Sw
itze
rlan
dN
oYe
sSc
andi
navi
ans
Finn
ish
Finl
and
No
No
Scan
dina
vian
Den
mar
k,N
orw
ay,S
wed
enN
oYe
sR
ussi
ans
and
othe
rsLi
thua
nian
,Rus
sian
,Rus
sian
Heb
rew
sR
ussi
aN
oYe
sPo
lish
Pola
ndYe
sYe
sR
oman
ian,
Rom
ania
nH
ebre
ws
Rom
ania
Yes
No
Oth
erEu
rope
Bohe
mia
nan
dM
orav
ian,
Rut
heni
an,S
lova
kBo
hem
ia(C
zech
oslo
vaki
a)Ye
sYe
sSe
rbia
nan
dM
onte
negr
in,C
roat
ian
and
Slov
enia
n,Bu
lgar
ia,Y
ugos
lavi
aYe
sYe
sBu
lgar
ian,
Dal
mat
ian,
Bosn
ian,
Her
zego
vini
anG
reek
Gre
ece
Yes
Yes
Mag
yar
Hun
gary
No
Yes
Oth
erC
hine
seC
hina
Yes
No
East
Indi
anIn
dia
Yes
No
Japa
nese
Japa
nN
oN
oPa
cific
Isla
nder
Paci
ficIs
land
sN
oN
oSy
rian
,Tur
kish
,Arm
enia
nTu
rkey
Yes
No
Oth
ers,
othe
rH
ebre
ws,
Kor
ean
Kor
ea,o
ther
coun
trie
sYe
sN
o
48