New Evidence on the Earnings Growth of Foreignborn Workers in the United States,

Transcription

1 New Evidence on the Earnings Growth of Foreignborn Workers in the United States, PLEASE DO NOT CITE Any opinions expressed herein are those of the author and not those of the Social Security Administration SSA Jae Song

2 Total Workers and Percentage of Foreign-born, aged 25-60, , 1% sample 18% 1,200,000 16% 14% 1,000,000 12% 800,000 10% 8% 600,000 6% 400,000 4% 2% 200,000 0% % foreign-born (left) Total workers (right) 2

3 Background United States has experienced: 1) an increase in the number of foreign-born workers; and 2) a shift in their regions of origin. Earlier studies showed that foreign-native wage gaps increased through the early 1990s. What has happened in wage gaps in more recent years? Changes in the gap are due to differentials in 1) wage growth ( catch-up ); 2) wages of new entrants ( entry cohort quality ); 3) wages of exit-ers ( emigration of low wage workers ) The lack of a good data source (complete panel data) has been a major problem. Limitations of survey based data include short panel data, small sample size, response error, non-random attrition, top coding, etc. 3

4 This study: 1. Examines 1) wage progression of foreign-born workers; 2) entry cohort quality ; 3) mechanics behind wage growth 2. Uses new, large, and accurate administrative panel data that represents the general US working age population in Uses native workers who are in the labor market (or enter) at the same time as control group 4. lny = Xβ + (1 d)(γysi) + e: Previous studies (wrt a typical native worker) lny = Xβ + (1 d)γ f YSL f + dγ n YSL n + e : current study, where YSI = year since immigration; YSL = year since US labor market entry; d = 1, if native 4

5 A few findings: The gaps (mean and median) have been decreasing since the mid-1990s: While mean gaps have nearly disappeared by 2012, median gaps have decreased to 10% level in the same year. Longitudinal estimates show much slower wage catch-up. Relatively slower convergence (similar growth rate) for women s wages; most of past studies focus on men only. Improvement in entry cohorts quality since mid-1990s decreased entry wage gap & improved subsequent relative wage growth 1986 immigration reform has a significant impact on entry wage differential as well as on subsequent wage growth Foreign-born workers have relatively higher prob of exit right after entry; and relatively higher prob of staying (lower prob of switch) with the same employer right after entry, conditional on working Foreign-born workers tend to work for firms with a high % of foreign-born upon entry and then gradually move to firms with a lower % of foreign-born Foreign-born workers tend to work for relatively small-sized firms upon entry and gradually move to larger firms. 5

6 Previous Studies First generation of studies: used single cross-sectional data (e.g., Chiswick (JPE,1977) 1970 Census) 1. Foreign-born workers initially earn less than native-born, but their earnings grow faster -- earnings equal to that of native-born after 10 to 15 years. 2. Regress log of earnings on years of schooling, labor market exp (= age-schooling-5), years since arrival. Second generation: used repeated cross-sectional data (e.g., Borjas (JEL, 1994)) 1. Cohort effects due to deceasing cohort quality (changing in countries of origin), the cross-sectional age earnings profile of foreign-born becomes steeper 2. Requires longitudinal data to track over time, uses decennial census from 1970 to Earlier arrivals earn nearly 20% more than natives, but late arrivals ( ) earn more than 30 % less than natives Third generation: used longitudinal data (90, 91 SIPP matched capped earnings) & repeated cross sections (Lubotsky (JPE, 2007)) 1. Slower convergence of earnings growth with Longitudinal data % growth in first 20 years 2. Emigration of low-wage foreign-born workers causes the overestimation of assimilation in repeated cross-sectional data. Fourth generation: 6

7 Road map Data sources SSA Numident file Description of X-sectional wage gap Wage assimilation longitudinal Entry wage relative to natives entered at the same time Wage growth mechanics Summary 7

8 Data sources SSA Numident file 1. Master file of assigned SSNs 2. Main source: Form SS-5, Application for Social Security Card 3. Dates of birth and death, place of birth, citizenship code, enumeration cycle date, entry code, race, and sex 4. Potential Issues missing POB, missing cycle date Note on SSN (see Puckett [2009]) 1. SSN structure: Area number first three digits; Group number second two digits; Serial number last four digits 2. SSNs with the same area and group numbers issued at approximately same time. 3. SSNs are NOT recycled. 4. Any missing SSN issue date can be easily imputed 8

9 Data sources (cont.) SSA Master Earnings file (MEF) 1. Uncapped total wages (W-2 level), 1978 to Capped Social Security covered wages, 1951 to 2012 SSA Employee-Employer (EE-ER) file (1% sample) 1. Geo code (state and county code for firm and/or residence) 2. Standard Industrial Classification (SIC) code 3. And other firm level information, such as total wage, 4. 1% file available from 1957-; (0.1% file from 1936) 9

10 Data construction First, select the 1% sample from the SSA Numident file. 1. Selected by a stratified cluster design based on certain serial digits of the SSN. 2. a random sample, represent the general population of those who have ever obtained a valid SSN. Then add individuals annual earnings, Social Security coverage (topcoded) from 1951 to 2012 and total wage (not top-coded) from 1978 to 2012 from the MEF. Also add 1) the first year of work (since 1936); 2) number of years worked, 1936 to 1950; 3) EINs associated with individuals annual wage, 1978 to Further, using SSA s 1% EE-ER file, we add geographic code (state and county) and 4-digit 1987 SIC code associated with employers. 10

11 Add following firm-level information: number of workers; percentage of foreign-born (and by regions of origin); average wage; average age of workers, from separate tabulation from the Master Earnings File. This 1% sample covers the same samples in the Continuous Work History Sample (CWHS). Census data includes illegals but not institutionalized. SSA data excludes illegals but includes institutionalized Examples of using the 1% data: Song and Manchester (2007); Sabelhaus and Song (2009); Gorodnichenko, Song and Stolyarov (2013); Kopczuk, Saez and Song (2010)--commerce and industry workers. 11

12 Sample restrictions & grouping Restrict samples to those age Exclude those born in Puerto Rico Grouping foreign-born workers by age upon arrival - arrived at age <=16 or 16<arrival age. Grouping foreign-born workers, by region of origin 1) Far East (Asia_FA); 2) Asia-South (Asia_SO); 3) Central America (CenAC); 4) Europe- East, including former Soviet (EuropeE); 5) Europe- North, including South (Europe_N); 6) Mexico (Mexico); 7) Middle East (MiddleE); 8) Latin America, including Caribbean (SouthA); 9) Sub-Sahara (subsa); and 10) Canada(Canada). 12

13 Would missing POB (place of birth) cause problems? Figure 1 Number of SSNs issued each year and the percentage of unknown place of birth, (10% Numident file). Figure 2 Number of SSNs issued by birth year and the percentage of missing place of birth variable, 1875 to Figure 3 Number of SSNs issued by selected regions of origin and the year of Issuance,

14 Figure 1: Number of SSN Issued and Percentage with Missing POB, by Issue Year, , , , , SSN issued (in,000 left) % Unknown POB (right) 14

15 600 Figure 2: Number of SSNs Issued and Percentage with Missing POB, by Birth Year, SSN issued (in,000 left) % Unknown POB (right) 15

16 Figure 3: Number of SSNs Issued, by Issue Year and Region of Origin, , ,000 80,000 60,000 40,000 20, Asia Europe Mexico America_SC 16

17 Figure 4: Number of Workers Aged and Percentage of Missing POB, , , , , % 4, , N workers (in,000) % Unknown POB (%, right) 17

18 Highlights A total of 39 million SSNs were issued in 1936 and 1937 Percentage of missing POB dropped below 5% starting from 1966 Percentage of missing POB also dropped below 5% starting from those who were born in 1931 or later. Most of those cases with missing country/state code (and foreign birth indicator) were SSNs issued to those born in 1925 or earlier. Those who were 60 or younger as of 1980 (those born in 1920 or later) appear to have fairly complete information on POB (country of origin). In 1978, the POB was unknown for about 0.6% (less than 1%) of workers aged 25 to 64. The percentage has further decreased to 0.039% by

19 Table 1: Description of data, every 10 th year & 2012 Foreign -born (age at Native entry>16) year Worker characteristics N 654, , , , ,618 26,438 42,685 62,521 70,328 67,814 Log (wage)* Men White Black Other Age Years since FWK Years since SSN Age FWK Log(wage FWK)* Employer characteristics Region Northeast Midwest South West Other/unknown Industry (SIC division) A Agriculture, Forestry B: Mining C: Construction D: Manufacturing E Transportation,Comm F: Wholesale Trade G: Retail Trade H: Finance, Insur I: Services J: Public Admin K: Nonclassifiable/unknown

20 Foreign -born (age at Native entry>16) year Firm size (number of workers) Log (N workers) to 4 workers to to to to to 5, ,001 or more Other characteristics of firm Average age of workers % men Average wage (all workers)* Average wage (native* Average wage (foreign)* % foreign-born Regions of origin (%) Asia_FA Asia_SO Canada CenAC EuropeE EuropeN Mexico MiddleE SouthA SubSA Other * : wage values are all in 2005 constant $ 20

21 Overview of X-Sectional Wage Gap, The one-percent sample of workers aged 25 to 60; Men and women separately; Older entrants (arrival age>16); Changes in mean, median, and distribution; 21

22 Mean & Median Wage Gaps MeanDiff_M MedianDiff_M MeanDiff_W MedianDiff_W 22

23 Wage Gap: All Percentiles Arrival age > 16, Men Percentile diff_1980 diff_2000 diff_1990 diff_

24 Wage Gap: All Percentiles Arrival age <= 16, Men Percentile diff_1980 diff_2000 diff_1990 diff_

25 Wage Gap: All Percentiles Arrival age > 16, Women Percentile diff_1980 diff_2000 diff_1990 diff_

26 X-sectional Wage Inequality 0.55 FigureA7: Wage Inequality (Gini), Native Foreign 26

27 Longitudinal wage catch-up 1) Conventional specification : lny = Xβ + αage + (1 d)γysi + e, d = 1, if Native 2) Since AGE = AgeFWK + YSL(or YRsinceFWK), we estimate lny = Xβ + αagefwk + (1 d)γ f YSL f + dγ n YSL n + e 3) For samples with nonzero wage at year = t, follow their earnings through 2012; Consider t = 1978, 1982, 1988, 1992, 1998, & 2002; Alternatively, lny ijt = a + βδ jt + γδ t + δδ j + θ X ijt + e ijt, where t = t,,2012; and j=0 (native), 1 (foreign), Δ are year, group, and interaction dummies; Control individual and firm characteristics; Fixed effects individual, firm (establishment) 27

28 Full Panel EST STD.E EST STD.E Men (N=17,127,485) foreign agefwk agefwk yrsincefwk*foreign (yrsincefwk*foreign)^ yrsincefwk*native (yrsincefwk*native)^ Women (N=15,501,043) foreign agefwk agefwk yrsincefwk*foreign (yrsincefwk*foreign)^ yrsincefwk*native (yrsincefwk*native)^

29 Estimation II: Cohort samples: 1978 cohorts (work in 1978 or entered by 1978), 1982, ; Follow native and foreign-born workers until 2012; Interested parameters: interaction between year since base year and foreign-born; Full controls, control individual fixed effects, control employer(firm) fixed effects, control both individual and firm fixed effects. 29

30 Longitudinal wage catch-up estimates 1978 working cohorts 1978 cohorts Men Women fixed, full control indivi fixed, fixed, firm fixed, both full control indivi relative to a typical native worker Yrsince*Foreign est std.e (Yrsince*Foreign)^2 est std.e Foreign est std.e fixed, firm fixed, both Yrsince*Foreign est std.e (Yrsince*Foreign)^2 est std.e Yrsince*Native est std.e (Yrsince*Native)^2 est std.e Foreign est std.e

31 1982 working cohorts 1982 cohorts Men Women relative to a typical native worker full control fixed, indivi fixed, firm fixed, both full control fixed, indivi Yrsince*Foreign est std.e (Yrsince*Foreign)^2 est std.e Foreign est std.e fixed, firm fixed, both Yrsince*Foreign est std.e (Yrsince*Foreign)^2 est std.e Yrsince*Native est std.e (Yrsince*Native)^2 est std.e Foreign est std.e

32 32

33 33

34 34

35 35

36 Estimation III: Cohort samples: 1978 cohorts (work in 1978 or entered by 1978), 1982, ; Include year dummies and interaction (year*foreign) dummies; Mean and percentile regressions; Control both individual and firm characteristics. Estimates of interaction dummies are reported in next slides - progression of wage gap 36

37 Mean Gap: Men who Arrived Age > e s t m e a n g a p t=1978 t=1982 t=1986 t=1990 t=1994 t=

38 Median Gap: Men who Arrived Age > e s t m e d i a n g a p t=1978 t=1982 t=1986 t=1990 t=1994 t=

39 Mean Gap: Women who Arrived Age > e s t m e a n g a p t=1978 t=1982 t=1986 t=1990 t=1994 t=

40 Median Gap: Women who Arrived Age > e s t m e d i a n g a p t=1978 t=1982 t=1986 t=1990 t=1994 t=

41 Entry cohort quality: Use native worker who entered at the same time as control group; Base cohort: those who entered in Samples who entered the labor market in ; Two sets of estimates 1) relative entry wage differentials and 2) relative wage evolutions using diff-in-diff; First estimates: relative entry wage (up to 5 years since entry) differentials Comparing across entry cohorts DD: d1 foreign and native; d entrants & entrants EntryW t Native EntryW t Foreign - EntryW t0 Native EntryW t0 Foreign lny ijt = a + βδ jt + γδ t + δδ j + θ X ijt + e ijt, Where t = t,,2007 (entry year); and j=0 (native), =1 (foreign arrived>16),; Δ are cohort year, group, and interaction dummies 41

42 Second estimates: 1) Differentials in relative wage evolutions since entry 2) Wage evolutions of 1981, 1982, entry cohorts relative to the wage evolution of entrants 3) Diff-in-diff-in-diff (3D) -D1: foreign & native; D2:entry Year & subsequent Years; D3: entry & other entry cohorts EntryW Native Foreign t EntryW t EntryW Native Foreign t0 EntryW t0 - EntryW Native Foreign t EntryW t EntryW t0 Native EntryW t0 Foreign cohort (or later entrants) cohort lny ijtk = a + β t Δ t + β j Δ j +β k Δ k + γ jt Δ jt +γ tk Δ tk + γ jk Δ jk + δ jtk Δ jkt +θ X ijtk + e ijtk, Where t = t0+1,,2012;k= , , ,.(entry year); and j=0 (native), =1 (foreign arrived>16),; Δs are year, entry year, group, and interaction dummies 42

43 Estimates I: Entry Wage Differential, Men Relative to Entrants 0.5 Estimated Entry Wage Differentials, Men Entry Age >

44 Entry Wage Differential, Women Relative to Entrants 0.8 Estimated Entry Wage Differentials, Women Entry Age >

45 Estimates II: Wage Growth by Entry Cohorts, Men Relative to Entrants

46 Wage Growth by Entry Cohorts, Women Relative to Entrants

47 Highlights Increase in cohort quality since mid-90s could be a cause of the decline in the wage gap since mid-90s 1) Relative entry wages have improved 2) Subsequent wage growth has improved Wage assimilation (longitudinal) has been occurring slowly Slowest convergence: entry and 1990 working cohorts Relative earnings growth rate varies by cohorts 47

48 Wage evolution mechanism: Job switching & sorting Regardless whether the catch-up occurs or not, foreignborn workers wages have been growing faster. How foreign-born workers are different from natives in terms of: work participation; exit; job switch/stay; choice of firms i) percentage of foreign-born workers; ii) size of firm Simple regression of interested variables on individual characteristics, year dummies, and quadratics of year since US labor market entry 48

49 Transition Prob.: Exit- work(t) to not work(t+1) Native_M Foreign_M Native_F Foreign_F 49

50 Transition Prob.: Stay with the same employer, t to t Native_M Foreign_M Native_F Foreign_F 50

51 Transition Prob.: Switch employer, t to t Native_M Foreign_M Native_F Foreign_F 51

52 Exit-YR Since FWK Native_M Foreign_M Native_W Foreign_W 52

53 Stay-YR since FWK Native_M Foreign_M Native_W Foreign_W 53

54 Switch-YR since FWK Native_M Foreign_M Native_W Foreign_W 54

55 % Foreign-born YR since FWK Native_M Foreign_M Native_W Foreign_W 55

56 Firm size (workers) - YR since FWK Native_M Foreign_M Native_W Foreign_W 56

57 Are low-wage foreign-born workers more likely to exit from the labor market? Regress exit on interaction of lagged wage and foreign-born, year dummies, and other characteristics MEN WOMEN EST STD.E EST STD.E yrsincefwk foreign lcwage (-1) foreign*lcwage(-1)

58 Highlights Foreign-born workers have relatively higher prob of exit right after entry Foreign-born workers have relatively higher prob of staying (lower prob of switch) with the same employer right after entry, conditional on working Foreign-born workers tend to work for firms with a high % of foreign-born at the entry and gradually move to firms with a lower % of foreign-born Foreign-born workers tend to work for relatively smallsized firms at the entry and gradually move to larger firms 58

59 Wrap up Wage gap for men has been deceasing since late 1990s Foreign-born workers wages grow faster, but catch-up? Improved new entry cohorts quality since late 1990s Wage evolution mechanism -- Job switching & sorting 59