1

Subject to Change

Centre for Economic Performance Working Paper

Do Relative Gains to Migration Vary? A Comparison of the Labour Market

Performance of Indians in the UK, the USA and India

Augustin de Coulon1 and Jonathan Wadsworth

2

1. Institute of Education, University of London and Centre for Economic

Performance, the London School of Economics

2. Royal Holloway College, University of London and Centre for Economic

Performance, the London School of Economics and IZA Bonn

Corresponding author:

Augustin de Coulon

Centre for Economic Peformance

London School Of Economics

London WC2A 2AE

Tel: (00 44) 207 955 7063

E-Mail: a.de-coulon@lse.ac.uk

2

Abstract

We compare the labour market performance of immigrants from a single origin country,

in two host countries with different migration policies, relative to their (counterfactual)

position had they remained in their source country. While most studies of immigration

focus on the absolute income differences between countries, we argue that relative gains

to migration may also influence the migration decision and that these gains will be

influenced by differential relative costs across the skill distribution. Using data on Indian

immigrants in the United States and Great Britain matched to comparable data on

individuals who remained in India, we show that the average Indian immigrant will

experience a fall in their relative ranking in the wage distribution compared to the

position they would have achieved had they remained in the origin country. The fall in

relative rankings is larger for immigrants to the US than to the UK, and largest of all for

those with intermediate skills.

Key words. Immigration, Wages, Relative Gains

JEL Classification Number J6

1

1. Introduction

Do relative rewards influence the migration decision? Three recent studies have

broadened the analysis of the labour market returns to immigration by comparing the

outcomes of similar countrymen (but not women) in both the host and the origin

countries. Bauer et al. (2002) compare the economic performance of the Portuguese in

Germany with that of similar individuals who remained in the home country (hereafter:

stayers). They observe that these Portuguese immigrants earned more than similar stayers

in Portugal. In a similar study, Hartog and Winkelmann (2003) compare the labour

market position of the Dutch in the Netherlands and in New-Zealand. Their findings

suggest that this sub-set of immigrants were also substantially better off, on average, in

terms of lifetime earnings, than if they had remained in the Netherlands. Chiquiar and

Hanson’s (2005) study compares the economic performance of Mexicans in Mexico and

in the US, finding that Mexican immigrants to the US tend to be drawn from the middle

to upper regions of the Mexican wage distribution. Such immigration patterns will tend to

raise wage inequality in the origin country.

These papers all use one origin and one host country. Our paper introduces a

second host country (the US and the UK), and compares the performance of immigrants

from one origin country (India) with similar individuals who remained in India.1 Every

year around 500,000 individuals leave India to go and live elsewhere, (Srivastava &

Saikumar 2003). Individuals born in India comprise the third largest country group of

immigrants in the United States and the largest group of immigrants in the UK. The U.K.

1 Antecol, Cobb-Clark and Trejo (2003a, 2003b, 2003c) investigate the effect of different migration policies on the employment and wage assimilation of all immigrants in three different host countries (Canada, Australia and the US). The results rely on the assumption that immigrants originating from different countries behave similarly in the host countries.

2

was the largest industrialised host country for Indian immigrants until 1970. Since then,

the USA and Canada have become the dominant industrialized host countries, (Srivastava

& Saikumar (2003) )2. The two host countries are therefore important destination sources

for Indian immigrants but have different migration policies that reflect, in part, each

countries pattern of demand and areas of comparative advantage. The two countries also

have different levels of wage inequality.

Extending the number of host countries allows us to investigate two related

predictions linked to the extended Roy model of self-selection (Borjas, 1987, 1999). First,

given two potential host countries, with different wage inequality and migration policies

and therefore different costs of migration, what is the comparative benefit of the decision

to move to either country? Second, we argue that the usual focus on absolute gains

(where individuals make their location choice based on an income maximising decision)

misses one important aspect of the migration decision. An individual’s relative position

in the wage distribution may also be an argument in the utility function, (Clark and

Oswald 1998), and so location decisions could also depend on the purchasing power or

status that moving to a job in the host country could command. We therefore compare

the position of similar Indian individuals in the wage distributions of India, the US and

the UK with the aim of assessing the relative wage gains from migration given

differential migration and assimilation costs across host countries.

Different migration policies may have differential effects on the quality of a job

match. An optimal migration policy would allocate unfilled vacancies in the host country

to immigrants with the requisite qualifications. Because of frictions caused, for example,

2 The annual flow of Indian immigrants to the U.K. reached a peak of 15,000 in 1970 and has subsequently fallen back to around 5,000 a year. The flow of Indian immigration to the US. and Canada is currently around 25,000 and 18,000, respectively, a year.

3

by migration quotas, inefficient transfer of skills or problems of assimilation, a perfect

job match may not be easily attainable and this will affect an individual’s position in the

wage distribution. We therefore compare immigrant performance in the UK and the US

relative to the position they would (theoretically) have attained, had they stayed in the

sender country. If the cross-country transfer of skills is inefficient, then immigrants may

be matched to jobs that are not suited for their qualifications. This would mean that they

would appear lower down the wage distribution of the host country, other things equal,

than in the sender country. In addition comparisons of immigrants with similar

characteristics in countries with different immigration-related institutions could facilitate

analysis of the relative efficiency of both migration policies.

We find that the absolute wage gains (defined as mean earnings differences

adjusted for Purchasing Power Parity, hereafter PPP, and controlling for observed

characteristics) are, around 500% for immigrants to the US and the UK, This average

comprises larger percentage gains for less skilled workers and smaller gains for college

educated workers. However, Indian immigrants will typically have a lower ranking in the

wage distribution of both host countries compared to the position had they remained in

the source country. These relative losses are highest for workers with intermediate

qualifications and lowest for those with lower levels of education. The relative fall in the

wage distribution is higher in the US than in the UK.

The layout of the paper is as follows. In Section 2, we discuss the model used and

the institutional features of immigration policy in the U.S. and in the U.K. In Section 3,

the data sets are presented together with the different samples used in our investigations.

4

In Section 4, the main results are introduced, while Section 5 presents our matching

estimates. We give concluding comments in Section 6.

2. Immigration Institutions in the U.S.A and in the U.K.

US migration policy as regards Indian nationals is influenced both by the country-specific

absolute limits on numbers of immigrants in place since the 1960s and the system of

preference categories embedded in visa applications skewed toward skilled workers or

workers in areas in which labor is in short supply, (Jasso, Rosenzweig and Smith (2000) ).

The 1990 Immigration Act introduced a system of visas allowing entry of high-tech

workers with a sponsoring employer/residence for up to 6 years (and with the subsequent

chance of applying for US citizenship) and raised the share of employment based visas in

the total immigrant flow. In practice this has meant that immigration from India has been

oriented towards highly qualified immigrants since the beginning of the 1990’s.

As regards the UK, policy has at times, sought to encourage Indian immigrants

from both ends of the skill distribution. In part this is because over the previous four

decades, UK migration policy has been greatly influenced by its links with former

colonies, of which India was the most populous. Before 1962, any commonwealth or

Irish citizen had the right of entry into the UK. A system of work permits was introduced

after that point and the principle of right of entry to commonwealth citizens was

abolished in 1973, replaced by a system of work permits, (again skewed toward skilled

workers in short supply) and entry rights for dependents that has continued, subject to

periodic modifications, to this day. A recent report from the Home Office, (Home Office,

2002), suggests that these legislative changes did not change significantly the

5

composition of (the reduced numbers of) Indian immigrants subsequently allowed to

settle.3

Theoretical Framework

We frame our empirical investigations with a simple extension of the Roy model of self-

selection (Borjas, 1987, Chiquiar and Hanson, 2005), whereby individuals in the origin

country face a host country wage distribution and potential immigrants face either an

origin or a host country wage distribution. Wages are assumed to depend on observed

human capital characteristics of the individual, and the rewards to those characteristics.

Characteristics unobserved to the researchers such as motivation and dynamism are the

captured by a random error term. In practice the discussion here is restricted to the effects

of observed characteristics but could be extended to unobserved one without significantly

affecting the main implications. If only one country is considered as a potential

destination, then the decision to move is made comparing income in the host and the

origin country net of migration costs. Absolute incomes gains are expected to be large at

every level education for migration from a developing to a developed country. Let the

wage depends on observed human capital characteristics of the individual, Xi and the

rewards to those characteristics, but also on characteristics unobserved to the researchers

such as motivation, or dynamism and captured by the random error term ε. Assuming,

only one host country (H) for potential migrants who are originally located in the origin

country (O):

Ln(wiO) = βo xi + εo (1)

Ln(wiH) = βH xi + εH (2)

3 The percentage of immigrants coming from India is significantly higher in the UK than in the US, (respectively 13.5% and 5% of all new immigrants with work permits, see OECD, 2001).

6

As is usual in these models we assume that εi ~ N(0, σ2i ) and εo and εH have correlation

coefficient ρ. If ρ>0 then characteristics are rewarded similarly in both countries.

We now introduce, π, the costs of migration.

Often the migration literature gives relatively little attention to the costs of

migration, be it direct transportation costs, barriers to entry imposed by the host country,

indirect psychic costs such as loss of social networks, loss of visits, contacts with the

extended family. Yet Borjas (1991) acknowledges that the predictions of the Roy model

could be changed if the costs of migration vary with earnings potential, rather than being

constant or randomly distributed across individuals. The greater the variation in migration

costs by skill, the more costs can offset the simple predictions of the Roy model,

(Chiquiar and Hansen, 2005) .

Clearly the work permit/visa systems in place in both the U.S. and the U.K. make

the costs of immigration higher for less skilled workers. However this alone, would not

be expected to affect an individual’s position in the wage distribution. Yet anecdotal

evidence shows that some immigrants end up in the host countries lower down the wage

distribution than expected, (see for example Guardian, 02/04 and migration news). First,

it may be that the transfer of skills is not perfect. This may be caused by language and

familiarization issues often discussed in the assimilation literature, (Chiswick, 1978).

However, there may also be other institutional impediments. For example, nurses who

qualified in India may be impeded from getting their qualifications validated by

unfamiliarity with the system or the actions of third parties.4 Secondly if there are non-

4 Doctors and nurses in the UK with qualifications obtained abroad are required to find a supervised job placement in order to do conversion training and have their qualifications “adapted” . Guardian, 02/04 suggests that the system is open to abuse, in that adaptations are either only available in more menial jobs or, in some cases, validation may be delayed. Migration news reports that some colleges in India

7

linearities in the returns to education across occupations and in particular if those non-

linearities vary across the home and host countries, this too could change the relative

position in the wage distribution. For example, the relative return to an occupation may

differ if there are differential labour shortages for particular occupations in the host

relative to the sender countries.

In short, potential moves down the income distribution may offset the potential

absolute income gains for some individuals. If that is the case, then immigration flows are

lower than would be expected from a simple extended Roy model. We therefore allow

the costs of migration to vary according to an individual’s position in the wage

distribution: πi= µπ - piπ where µπ are costs of migration that are constant across

individuals and the variable cost pi>0 (pi<0), depends on whether individuals move up

(down) the income distribution of the host country compared to the position in the wage

distribution of the origin country.

3. Data

We use three data sets, each containing individual data on pay and a set of

individual control variables. For the U.S. we utilise the year 2000 1% sub-sample5. We

select individuals aged 16 to 65 employed in a full-time job the week previous to the

interview earning an hourly wage of between 1 and 200 US$ . Using the country of birth

and detailed categorization of race data, we define Indians immigrants as those born in

India and who fall into the “Asian Indians” category. We split the sample into four

education groups that can be compared across the data sets in each country. These

"guarantee" students an H-1B visa, usually by having a US-based contractor attest that the individual is needed to fill a US job. However, once in the US, there may not be a job. 5 This can be downloaded from the IPUMS web site http://www.ipums.org/usa/index.html

8

categories correspond to those who left school before 16, those with some post-

compulsory state education, those with some vocational education and those with a

college degree. To focus on the decision made after education has been completed, we

further select on those who migrated after the age at which education was completed.

This gives us a US sample of around 900,000 individuals, of whom around 3,700 came

from India after completing their education.

For the second host country, namely the UK, we use various waves of the

quarterly Labour Force Survey6 (LFS). The LFS being a rotating panel, where each

individual is interviewed for five quarters in a row, we take wage responses from the

8,000 working individuals interviewed for the first time in each quarter. To boost the

sample size we pool all individuals interviewed over the period 1997 to 2002. We again

select only individuals of working age, between 16 and 60/65 (respectively for women

and men) and again define “Indian immigrants” as someone born in India of Indian

ethnicity and who migrated from India after completing their education. For the host

wage distribution we select only those with a full-time weekly wage between 15 and

3000 £UK. This gives us a UK sample of around 75,000 individuals, of whom around

400 came from India after completing their education.

For India, we build our sample of control groups using the dataset collected by the

National Sample Survey Organisation (NSSO), who have conducted representative

household surveys in India since 1950. The 55th round was conducted between July 1999

and June 2000, with 820,000 individuals interviewed in around 100,000 households.

After elimination of non valid data, we end up with 108,622 individuals aged 15 to 65 in

6 The UK Census does not contain information on wages.

9

full-time work.7 This group comprises the set of stayers with which we compare relative

positional rankings of immigrants, though it will, of course, include any return

immigrants. We convert all weekly wage data into US dollars using the IMF Purchasing

Power Parity Index, prevailing in the year 2000, (IMF 2000).

4. Descriptive statistics

As a first indication of our focus on distributional issues, Figure 1 gives the weekly wage

distribution, PPP adjusted, of full-time employees observed in the three countries. We

observe, not surprisingly, that the Indian wage distribution is to the right of the wage

distributions in the UK and even more so in the US. We also observe that the three

distributions overlap. Of more interest, is the confirmation that inequality appears much

higher in India.8 Table 1 shows that the Gini coefficients are the highest in India, and also

higher in the US compare to the UK. This holds both for men and women. Wage

inequality amongst Indian women is particularly high. In absolute terms, an individual at

the 90th percentile of the Indian distribution earns more than someone at the 10th

percentile in the UK, but less than those at the 10th percentile in the US.

Indians living in Great Britain arrived earlier, on average, than other immigrants

to Great Britain, (Table 2). Indian immigrants comprise a similar fraction of the U.K.

and US (full-time) workforces, but the immigrant shares are small relative to both the

host and the sender country, around 0.3% in the host countries. It also appears that the US

received rather more Indian immigrants than the U.K. in the years immediately before

7 We select those individuals working at least 3.5 days per week (working time is given by total numbers of ½ days at work). 8 The bimodality of the Indian data is not caused by our definition of full-time work. The same pattern occurs for those working 7 days a week.

10

2000. With regards to education profiles (Table 3), Indian immigrants are similar to those

of other immigrants in the U.K. There are both more highly qualified and more lower

qualified immigrants than in the British-born population. Note also the similar education

profiles of Indian immigrants to Britain and stayers in full-time jobs in India. In contrast,

the education profiles of Indian immigrants to the US are rather different to those of other

immigrants and of the US born population. There are many more highly qualified

individuals among Indian immigrants than among other immigrants or among the US

born.

Table 4, shows that in the UK, the lower qualified arrived earlier than those with

higher qualifications and this is also the case for other immigrants. In US, the average

length of stay in the country is similar across skill groups and also similar compared to

other immigrants.9 The participation of Indian immigrant women is slightly higher in

the UK than in the US, (Table 5). In both host countries, this proportion is much higher

that in India. The proportion of young people is lower amongst immigrants than amongst

the native populations, as opposed to the general rule that immigrants are on average

younger than natives. The selection in our samples of immigrants of only those who

completed their studies in India together with the large proportion of graduates most

probably explains this pattern.

Section 5. Empirical Results

We begin, in Table 6, with an examination of the absolute gains to migration. To make

this comparison, we pool the Indian stayers data with those of Indian immigrants found in

the U.S. or U.K. data. We then run simple regressions of the log weekly wage on a set of

9 Our sample may be affected by no-random return migration, and therefore comments should be taken as conditional on those flows.

11

controls together with a dummy variable to indicate if the individual is an immigrant.

Table 6 shows that Indian immigrants in the US earn more, on average, than Indian

immigrants in the UK when compared to Indians working in India. The absolute gains to

immigration, net of observable differences, are in the order of 1.9 log points.10 If anything,

the absolute gains net of controls are higher in Britain than in the US, particularly

absenting occupation and industry controls. When we disaggregate by year of entry, there

does not seem to be much difference across the cohorts (Table 7).

When we split the sample by level of education, (Table 8), it appears that the

largest absolute gains are made by those with the lowest levels of educational attainment

– with little difference between immigrants to the U.S. and the U.K. and the lowest those

with intermediate levels of education. College-level educated workers gain more in

absolute terms by moving to the United States, Secondary level workers gain more by

moving to the U.K. 11 Both these patterns are consistent with the patterns of wage

inequality in the two host countries.12

In Table 9 we examine the relative gains to migration, comparing an individual

immigrant’s percentile ranking in the wage distribution with that of similar individuals in

the Indian stayers’ wage distribution. In both host countries, there appears to be a

significant fall in the immigrant’s wage ranking compared to those in the origin country.

The unconditional percentile rankings for the whole sample, suggest a relative gain to fall

for immigrants to the US and a small fall for immigrants to the UK. However much of

10 For British migrants in the US, the immigration premiums are much lower 11 From results not shown here, it appears that secondary workers in India receive a larger wage premium compared to the low qualified than in the other countries, perhaps because these basic skills don’t attract same premium in developed countries as in developing world. 12 The nature of our data mean that we are unable to control directly for potential selectivity in the migrant pool. Table A1 in the appendix compares returns to education of immigrants with their U.S. born counterparts. The returns are similar which suggests little (observable) difference between the two stocks.

12

this effect is driven by the differential composition of immigrants across the two host

countries. When we condition on personal characteristics and job characteristics the

estimated relative rankings become significantly negative. When the sample is split by

educational attainment, the percentile ranking decline is much larger in the US than in the

U.K. So Indian immigrants to the US appear to do relatively worse than similar

immigrants to the UK who in turn do relatively worse than similar Indians in India. The

largest ranking falls are amongst Indian immigrants with secondary level of qualification.

These conclusions do not change much with the addition of controls. Table A3 in the

appendix shows that matching by propensity score produces similar results. When we

disaggregate by entry cohort, Table 10, the largest percentile declines in the U.S. appear

to be found amongst more recent immigrants. In contrast more recent immigrants from

India to the UK appear to do relatively better than immigrants from the sixties and

seventies.

6. Conclusions

While most studies of immigration focus on the absolute income differences between

countries, we argue that relative gains to migration may also influence the migration

decision and that these gains will be influenced by differential relative costs across the

skill distribution. Using data on Indian immigrants in the United States and Great Britain

matched to comparable data on individuals who remained in India, we show that the

average Indian immigrant will experience an absolute gain in the order of 600%.

However the typical immigrant will also experience a significant fall of up to 30 points in

their relative ranking in the wage distribution compared to the position they would have

achieved had they remained in the origin country.

13

References

Antecol, Cobb-Clark and Trejo (2003a), “Human Capital and Earnings of Female

Immigrants to Australia, Canada, and the United States.” In: Host Societies and

the Reception of Immigrants, La Jolla, CA: Center for Comparative Immigration

Studies, University of California, San Diego.

Antecol, H., Kuhn, P. and Trejo, S. (2003b), Immigration Policy and the Skills of

Immigrants to Australia, Canada and United States, Journal of Human Resources,

38(1), pp. 192-218.

Antecol, H., Kuhn, P. and Trejo, S. (2003c), Assimilation via Prices or Quantities? Labor

Market Institutions and Immigrant Earnings Growth in Australia, Canada, and the

United States, IZA Discussion Paper No 802, Bonn, Germany

Bauer, T., Pereira P., Michael Vogler, M. and Zimmerman, K., (2002) ‘Portuguese

Migrants in the German Labor Market: Performance and Self-Selection’, International Migration Review, Vol 36, No. 2, Summer 2002, pp. 467-491.

Borjas, G., (1987), ‘Self-Selection and the Earnings of Immigrants’, American Economic

Review, Vol. 77, September, pp. 531-553.

Borjas, G., (1991), ‘Immigration and Self-Selection’, in Immigration, Trade and the Labor Market (eds.) J. Abowd and R. Freeman, NBER , University of Chicago Press.

Borjas, G., (1987), ‘Self-Selection and the Earnings of Immigrants’ American Economic

Review, pp. 531-553.

Borjas, G., (1999), The Economic Analysis of Immigration, in Handbook of Labor

Economics, Volume 3A, edited by Orley Ashenfelter and David Card, North-

Holland, pp. 1697-1760

Chiquiar D., and Hanson, G., (2005), ‘International Migration, Self-Selection and the Distribution of Wages: Evidence from the United States and Mexico’, Journal of

Political Economy, April 113, pp. 239-281. Chiswick, B. (1978), ‘The Effect of Americanization on the Earnings of Foreign-Born

Men,’ Journal of Political Economy, October, pp. 897-922.

14

Clark, A. and Oswald, A., (1998), ‘Comparison-concave utility and following behaviour in social and economic settings’, Journal of Public Economics Vol. 70, pp. 133–155

Hartog J., Winkelmann R., (2003) Comparing Migrants to Non-Migrants: The Case of

Dutch Migration to New Zealand, Journal of Population Economics 16(4): 683-

705.

Home Office, (2002), Migration: An Economic and Social Analysis, Research,

Development and Statistic Occasional Paper No 67, Government Statistical Office,

Home Office, London, UK.

Jasso, G., Rosenzweig, M. and Smith, J. (2000). "The Changing Skill of New

Immigrants to the United States: Recent Trends and Their Determinants." pp.185-225 in George J. Borjas (ed.), Issues in the Economics of Immigration. Chicago, IL: University of Chicago Press.

OECD, (2001), SOPEMI, Trends in International Migration, Continuous Reporting

System on Migration, Annual Report, Paris.

Sjaastad, L, (1962), The Costs and Returns of Human Migration, Journal of Political Economy, 70(5), 80-93.

Srinavasta, R. and Saikumar, S., (2003), ‘An Overview of Migration in India, its Impacts

and Key Issues’, in Migration Development Pro-Poor Policy Choices, DFID

Conference Papers, www. livelihoods.org

15

Figure 1. Kernel Density Estimates of Log Weekly Wages (Full-Time Workers):

India, U.S. and G.B (2000)

density

Log Weekly Wage

India U.S. G.B.

0 .5 1 2 3 4 5 6 7 8 9 9.5

.1

.2

.3

.4

.5

.6

.7

16

Table 1. Wage inequality in India, UK and US in 2000 (full-time workers)

India Britain U.S.

Total 10 2.78 4.84 5.54 50 3.69 5.46 6.39 90 5.21 6.15 7.22 Gini .509 .305 .393 Men 10 3.00 4.93 5.66 50 3.88 5.55 6.51 90 5.23 6.25 7.34 Gini .484 .303 .398 Women 10 2.44 4.73 5.44 50 3.13 5.28 6.22 90 5.03 5.92 7.00 Gini .567 .277 .358 Note: Log of weekly wages for full-time employees, corrected for PPP’s (IMF, 2004).

Table 2. Average years since migration of immigrants to the UK and US

Great Britain United States

Origin Country India Other immigrants

India Other immigrants

Mean 15.3 (11.2)

12.3 (11.3)

11.2 (8.8)

13.2 (9.4)

25th percentile 5 3 3 5 Median 14 9 9 11 75th percentile 24 20 17 20 % of all full-time employees 0.2 6.8 0.3 7.5 Note: Sample is individuals in full-time work who completed education in the host country and subsequently moved. Standard-errors in brackets.

17

Table 3. Educational Attainment (Full-Time Employees)

India Great Britain United States

% Share GB born

Indian immigr

ants

Other immigrants

US born

Indian immigrant

s

Other immigran

ts Primary 67.6 28.4 64.7 57.8 6.7 6.2 39.8 Secondary 16.8 38.3 11.2 12.5 34.3 10.7 24.9 Some college 3.0 17.2 7.3 10.3 31.6 9.7 15.1 College 12.7 16.1 16.9 19.4 27.4 73.4 20.3

Table 4. Average years since migration in UK and US by qualification

Great Britain U.S.A.

Indian Other Indian Other Low Qualifications 18.9 11.8 11.2 13.3 Secondary 18.1 15.7 12.6 13.1 Some college 13.5 10.8 15.4 14.6 College 13.2 10.8 12.6 12.1

Table 5. Other Sample Characteristics of Full-Time Employees

Indian Stayers

Indian Immigrants to G.B.

Indian Immigrants to U.S.A.

G.B. natives U.S. natives

Female 22.2 37.8 30.0 38.0 43.3 % < age 35 36.0 21.4 24.0 39.2 31.6 %>= age 50 20.1 30.7 31.1 20.7 24.3

18

Table 6. Absolute Gains to Migration: Log(weekly wage)

Indians in United States wrt Indian stayers Indians in Great Britain wrt Indian stayers

Immigrant 2.838 1.875 1.846 2.176 2.116 1.84 (0.016)** (0.015)** (0.016)** (0.029)** (0.030)** (0.025)** Demographic No Yes Yes No Yes Yes Occupation No No Yes No No Yes Industry No No Yes No No Yes N 106776 106776 106776 104800 104799 104799

Adjusted R2 0.177 0.578 0.69 0.026 0.505 0.633

Robust standard errors in brackets; ** indicates significant at 5%;

Table 7. Absolute Gains to Migration by Entry Cohort: Log(weekly wage)

Indian in United States wrt stayers Indian in Great Britain wrt Indian stayers

Sixties 3.148 2.026 2.08 2.053 2.016 1.741 (0.076)** (0.075)** (0.057)** (0.061)** (0.072)** (0.062)** Seventies 3.004 1.955 1.953 2.096 2.047 1.799 (0.041)** (0.037)** (0.028)** (0.053)** (0.058)** (0.046)** Eighties 2.751 1.798 1.834 2.051 1.976 1.762 (0.028)** (0.026)** (0.022)** (0.064)** (0.062)** (0.049)** Nineties 2.807 1.882 1.774 2.38 2.32 1.967 (0.022)** (0.021)** (0.021)** (0.047)** (0.049)** (0.042)** Demographic No Yes Yes No Yes Yes Occupation No No Yes No No Yes Industry No No Yes No No Yes Adjusted R2 0.177 0.578 0.69 0.026 0.505 0.633

19

20

Table 8. Absolute Gains to Migration by Education: Log(weekly wage)

Indian in United States wrt stayers Indian in Great Britain wrt stayers

College Immigrant 1.902 1.85 1.611 1.635 1.558 1.347 (0.018)** (0.018)** (0.022)** (0.053)** (0.058)** (0.048)** Demographic No Yes Yes No Yes Yes Occupation No No Yes No No Yes Industry No No Yes No No Yes Secondary Immigrant 1.799 1.495 1.672 1.82 1.51 1.606 (0.029)** (0.030)** (0.029)** (0.060)** (0.067)** (0.064)** Demographic No Yes Yes No Yes Yes Occupation No No Yes No No Yes Industry No No Yes No No Yes Low Immigrant 2.459 2.461 2.187 2.355 2.394 2.081 (0.046)** (0.050)** (0.048)** (0.032)** (0.033)** (0.032)** Demographic No Yes Yes No Yes Yes Occupation No No Yes No No Yes Industry No No Yes No No Yes Notes; see Table 6.

21

Table 9. Relative Percentile Gains to Migration by Education: Log(weekly wage)

Indian in United States wrt stayers Indian in Great Britain wrt stayers

Total Immigrant 14.593 -13.672 -16.054 -4.815 -5.984 -14.806 (0.612)** (0.563)** (0.604)** (1.413)** (1.213)** (1.090)** Demographic No Yes Yes No Yes Yes Occupation No No Yes No No Yes Industry No No Yes No No Yes College Immigrant -9.629 -10.915 -18.013 -7.513 -9.409 -12.904 (0.641)** (0.645)** (0.750)** (2.560)** (2.648)** (2.188)** Demographic No Yes Yes No Yes Yes Occupation No No Yes No No Yes Industry No No Yes No No Yes Secondary Immigrant -24.079 -32.965 -28.42 -12.658 -20.341 -17.934 (1.240)** (1.230)** (1.159)** (3.484)** (3.554)** (3.414)** Demographic No Yes Yes No Yes Yes Occupation No No Yes No No Yes Industry No No Yes No No Yes Low Immigrant -6.913 -6.639 -16.824 -3.866 -3.598 -2.053 (1.909)** (1.989)** (1.925)** (1.498)** (1.498)* (1.447) Demographic No Yes Yes No Yes Yes Occupation No No Yes No No Yes Industry No No Yes No No Yes

22

Table 10. Relative Returns by Entry Cohort

Indian in United States wrt stayers

Indian in Great Britain wrt stayers

Total College Second Low Total College Second Low Sixties 3.244 -12.009 -22.522 -53.835 -22.155 -24.377 -30.779 -15.212 (2.652) (2.656)** (6.412)** (1.983)** (2.781)** (5.291)** (9.772)** (3.004)** Seventies -4.073 -14.787 -25.382 -10.729 -19.16 -17.092 -19.282 -15.945 (1.229)** (1.370)** (2.482)** -5.925 (2.162)** (5.343)** (4.944)** (2.678)** Eighties -10.424 -19.363 -25.824 -15.067 -21.863 -16.987 -28.286 -19.725 (0.919)** (1.154)** (1.739)** (2.441)** (1.983)** (4.259)** (7.092)** (2.294)** Nineties -13.162 -19.111 -34.228 -20.74 -9.762 -6.601 -5.823 -5.772 (0.912)** (1.121)** (1.804)** (3.127)** (1.773)** (2.878)* -5.883 (2.545)* Demographic Yes Yes Yes Yes Yes Yes Yes Yes Occupation Yes Yes Yes Yes Yes Yes Yes Yes Industry Yes Yes Yes Yes Yes Yes Yes Yes Observations 106719 15373 21470 69876 104742 13630 21025 70087 R-squared 0.588 0.314 0.462 0.445 0.594 0.325 0.451 0.44 Robust standard errors in parentheses. Final column of British movers to U.S. not available because of small immigrant sample size with low qualifications.

23

Table A1. Sample Means

Indian Immigrants in US

Indian Immigrants in UK Indians stayers in India

Age 41.5 41.6 35.3 (10.2) (10.3) (11.4) Education: dropouts 0.06 0.68 0.67 (0.004) (0.02) (0.001) Secondary educ 0.11 0.07 0.17 (0.006) (0.01) (0.001) Some college 0.10 0.05 0.03 (0.005) (0.01) (0.001) College 0.73 0.19 0.13 (0.009) (0.02) (0.001) Female 0.30 0.38 0.22 (0.009) (0.02) (0.001) Single 0.10 0.06 0.20 (0.005) (0.01) (0.001) Main Occupations 3(24%) Production (25%) 18(33%) 10(12%) Health Care (12%) 17(10%) 16(9%) Maintenance (9%) 19(8%) 17(8%) Construction (9%) 15(8%) 21(8%) Management (8%) 8(7%) Main industries Manufacturing (32%) Agriculture (32%) Health (14%) Manufacturing

(14%) Finance/Real Estate

(11%) Public Services (11%)

Retail (10%) Construction (8%) Transport (9%) Education (7%) N 2515 515 104261

24

Table A2. Returns to Education Movers and Stayers (Log weekly wages full-time

workers)

US stayers

Indian movers to US

GB movers to US

GB stayers

Indian movers to GB

US movers to GB

Female -0.327 -0.300 -0.558 -0.283 -0.266 -0.455 (0.001)** (0.031)** (0.041)** (0.004)** (0.049)** (0.081)** Single -0.111 -0.093 0.039 -0.080 0.118 0.135 (0.002)** (0.050) (0.067) (0.005)** (0.107) (0.121) Secondary 0.227 0.112 0.389 0.184 0.278 -0.059 (0.002)** (0.058) (0.129)** (0.005)** (0.073)** (0.178) Some college

0.420 0.456 0.593 0.400 0.541 0.336

(0.002)** (0.061)** (0.126)** (0.005)** (0.111)** (0.168)** College 0.841 0.942 1.109 0.679 0.773 0.528 (0.002)** (0.049)** (0.126)** (0.006)** (0.063)** (0.159)** Constant 5.665 5.724 5.849 5.086 5.777 5.709 (0.019)** (0.219)** (0.172)** (0.023)** (0.175)** (0.632)** Observations 899807 2515 1342 73867 515 263 R-squared 0.28 0.24 0.32 0.34 0.29 0.24

Robust standard errors in parentheses, * significant at 5%; ** significant at 1%

25

Table A3. Propsensity Score Matching Estimates of Relative Position in the wage

distribution in India and in the UK

Percentile Ranking

Indian Immigrants (Treatment)

Indians in India (Control)

Difference

USA 64.3 Unmatched 49.7 +14.6 Radius Matching 79.1 -14.8 Local linear regression 80.4 -16.1 Kernel matching: Epanechnikov (bandwidth=0.01) 78.9 -14.5 Epanechnikov (bandwidth=0.05) 80.2 -16.2 Tricube (bandwidth=0.001) 80.3 -16.3 UK Unmatched 45.0 49.8 -4.7 Radius Matching 45.1 60.7 -15.6 Local linear regression 45.0 63.1 -18.1 Kernel matching: Epanechnikov (0.01) 45.0 68.3 -23.3 Epanechnikov (0.05) 45.0 68.4 -23.4 Tricube (0.001) 45.3 65.8 -20.7