the potential gains from international migration...from migration.however, such increases in...

International migration can generate substan-tial welfare gains for migrants, their countriesof origin, and the countries to which they mi-grate. The main focus of this report is on gainsfrom the remittances that migrants send home(discussed in chapters 4–6); chapters 2 and 3address the economic costs and benefits ofmigration and the impact of migration onpoverty. In this chapter, we use an economicmodel to estimate the size of the welfare gainsresulting from migration from developing tohigh-income countries.1 It must be recognizedat the outset that the model fails to capturesome known costs and benefits of migration;that the results are dependent on the specifica-tion of the model and its key parameters;and that the model cannot incorporate socialor political considerations.2 The results of thissimulation do not provide a precise forecast ofthe likely impact of migration; instead, theyprovide a consistent framework that offers in-sights into (a) the economic gains that can beexpected from changes in policy or circum-stances, and (b) the channels through whichmigration affects welfare—and both are diffi-cult to measure in reality. The conclusionsdrawn from the model are supported by sev-eral empirical studies, and they hold up wellunder various alternative assumptions formodel specification and parameters.

In chapter 3, we complement this model-based approach to measuring the gains frommigration with a review of the economic liter-ature, which covers the implications for

migrants and for their origin countries. Herewe can refer to a broader range of economicissues than are captured by the model, al-though without the ability to quantify that themodel-based simulation provides.

Starting from the base-case forecast of eco-nomic activity described in chapter 1, we in-troduce an additional increase in migrationfrom developing to high-income countries suf-ficient to raise the labor force of high-incomecountries by 3 percent over the period2001–25. The assumed increase, roughly one-eighth of a percentage point a year, is close tothat observed over the 1970–2000 period. Weimply no judgment concerning whether suchan increase is likely or politically feasible, butrather view the rise in migration as an exoge-nous shock. As discussed in chapter 3, pres-sures to migrate are likely to rise over the nextfew decades, but the actual size of the migrantflows will depend heavily on political deci-sions in destination countries. This exercisepresents us with the following key findings.

The expected decline in the labor force inhigh-income countries will increase depen-dency ratios, which could add to the benefitsfrom migration. However, such increases inmigration are unlikely to be large enough tohave a significant impact on dependency ra-tios in high-income countries.

Under the assumptions adopted in this model-ing exercise, the rise in migration—small

25

The Potential Gains fromInternational Migration

2

relative to the labor force of high-income coun-tries, but large relative to the existing stock ofmigrants—would generate large increases inglobal welfare. Migrants, natives in destinationcountries, and households in origin countrieswould experience gains in income, although mi-grants already living in high-income countrieswould see a decline in wages relative to the basecase. Estimates of these gains and losses are par-ticularly sensitive to assumptions about the de-gree of differentiation among workers (betweennatives and migrants and between old and newmigrants), the impact of migrants on fiscal bal-ances, and the extent of remittances.

Empirical studies of the impact of migrationon natives’ wages have had mixed results. Inthis simulation exercise, the rise in migrationleads to a small decline in average wages inhigh-income countries relative to the baseline,which one would anticipate from a labor sup-ply shock. But the decline has a barely per-ceptible impact on the long-term growth ratefor wages.

Native households in high-income countriesenjoy a rise in income, on average, as returnsto capital increase, offsetting the mild declinein wages. The impact on developing countriesis nearly the reverse, with wage income risingas labor-market conditions for workers im-prove, while returns on capital decline withthe smaller supply of workers. In developingcountries the gain from increased remittancesgreatly exceeds that from changes in factorreturns.

The economic benefits for high-incomeeconomies could be even larger than those pre-dicted by the model, due to several factors: themodel excludes the increased productivity ofmigrants (and the benefits to their offspring)over time; investment levels could increasesubstantially in response to higher returns tocapital; labor-force participation could riseamong natives with the greater availability ofmigrant labor (for household help, for exam-ple); the labor market would become moreflexibile, and diversity would increase.

The costs of adjusting to increased migrationand the gains from migration depend, in part,on the investment climate. Adjustment costsas a result of migration will be lower if moreflexible labor markets and more efficient cap-ital markets in high-income economies reducetransitional unemployment and the cost of re-placing capital as economies adjust to the risein immigration. Similarly, developing coun-tries with strong investment climates will beable to use increased remittances more effi-ciently, and enable workers who do not mi-grate to respond to improved labor marketconditions. The cost of adjustment may alsobe lower if migration is spread over timerather than concentrated in spurts.

A principal conclusion from this exercise isthat migration can generate significant eco-nomic gains for migrants, origin countries,and destination countries—but migration alsocan have important political and social conse-quences. For example, natives in destinationcountries may become concerned about main-taining cultural identity in the middle of agrowing diversity, which also has implicationsrelative to minority languages and other issuessurrounding the integration of migrants. Tosome extent, opposition to migration is drivenby these concerns, and not by an economiccalculation of the gains and losses.

We begin with a discussion of recent trendsand discuss how migration to high-incomecountries has grown over the past 30 years.We then turn to the prospects for migration,including the intense pressures generated bydemographic changes. We describe the base-case scenario for migration and the model-based analysis of the welfare gains from in-creased migration. We conclude with issuesthat the model does not consider.

International migration trends

Migration to high-income countries has acceleratedThe United Nations (UN) estimates that mi-grants account for some 3 percent of the

G L O B A L E C O N O M I C P R O S P E C T S 2 0 0 6

26

world’s population, or about 175 millionpersons.3 The stock of immigrants to high-income countries increased at about 3 percentper year from 1980 to 2000, up from the2.4 percent pace in the 1970s (table 2.1). Atthat rate of growth, the share of migrants inhigh-income countries’ population almostdoubled over the 30-year period, and popula-tion growth (excluding migration) fell from0.7 percent per year in the 1970s to 0.5 per-cent in the 1990s. Immigration has had a par-ticular impact on population growth in severalhigh-income countries. For example, withoutimmigration Germany, Italy, and Swedenwould have experienced a decline in popula-tion in the past few decades (OECD 2005;IOM 2005). By contrast, migration to devel-oping countries rose by only 1.3 percent peryear from 1970 to 2000. With rapid popula-

tion growth, the share of migrants in develop-ing countries’ population (excluding the for-mer Soviet Union) fell (figure 2.1).4

Most high-income countries saw immigra-tion rise by at least 2 percent per year from1980 to 2000.5 This increase reflected, inpart, increased demand for services accom-panying rising incomes, global competitionfor highly educated workers as technologicaladvances boosted the premium for skills, thegrowth of networks of immigrants in high-income countries that facilitated new immi-gration, and increased refugee movements.Almost 70 percent of the increase in immi-gration is accounted for by the United Statesand Germany, which together make up lessthan 40 percent of the population of thehigh-income countries. In the United States,the Immigration Reform and Control Act(IRCA) of 1986, which provided permanentstatus to 2.7 million migrants, facilitated fur-ther immigration through rules governingfamily reunification and may have encour-aged further irregular immigration (Passel2005) by encouraging expectations of futureamnesties.6 Germany saw a large inflow ofethnic Germans following the breakup of theSoviet Union (Dustmann and Glitz 2005), aswell as an increase in temporary migrationunder bilateral agreements.

Though the stock of migrants has acceler-ated sharply relative to the population in theindustrial countries, in some respects the com-position and patterns of international migra-tion have exhibited continuity over the pastfew decades. The share of female migrantshas remained almost unchanged (47 percent

T H E P O T E N T I A L G A I N S F R O M I N T E R N A T I O N A L M I G R A T I O N

27

Percentage

Source: United Nations.

Note: a. Excluding countries of the former USSR.

World Developedcountries

Developingcountries

Developingcountriesa

Figure 2.1 International migrants as ashare of destination countries’ population

0

2

1

3

7

5

9

6

4

8

20001970

Table 2.1 Growth in international migration by destination, 1970–2000Percent change per year in stock of migrants

1970–80 1982–90 1990–2000

World 2.0 4.4 1.3

High-income countries 2.4 2.9 3.1Developing countries 1.8 5.5 –0.1

Excluding former USSR 1.9 2.1 0.0Former USSR 0.5 25.0 –0.3


of global migrant populations in 1970, com-pared with 49 percent in 2000—figure 2.2),although women are the great majority ofmigrants from some countries. More womentoday are migrating as independent wageearners, rather than to accompany their hus-bands (IOM 2005). Migration continues to beheavily determined by geographic proximity(from Mexico to the United States, fromNorth Africa to Southern Europe, and fromEastern to Western Europe), as well as bycolonial ties (from Latin America to Spain andfrom a number of Sub-Saharan African coun-tries to Belgium, France, Portugal, and theUnited Kingdom—OECD 2005). The majorcountries of destination continue to admit thelargest share of permanent immigrants forfamily reunification (or, in the case of the EUcountries, for humanitarian or refugee resettle-ment), although some countries are refocusingtheir migration policy toward economic(largely skilled) immigration (figure 2.3).7 Butinternational migration is also changing, par-ticularly in the direction of flows. For exam-ple, more Asians are today seeking work inother Asian countries rather than in the Mid-dle East (Wickramasekera 2002; OECD 2005;IOM 2005), while more Latin Americans areturning to Europe for work opportunities, inaddition to North America.

It should be emphasized that the migrationdata on which these judgments are basedtend to be unreliable and incomplete. Manycountries and international agencies do notdistinguish between regular and irregularmigration or among types of temporary migra-tion. Some record migrants’ country of birth;others their nationality (OECD 2005).National estimates of the number of migrantscan be vastly different depending on whether“migrant” is defined as foreign born or of for-eign nationality.

Migration is set to increaseIt is likely that the number of people who wishto migrate from developing to high-incomecountries will rise over the next two decades.About 31 percent of developing countries’population is below the age of 14, comparedwith 18 percent in high-income countries. Wecan thus anticipate a large influx in the agecategories most suitable for emigration, aslifetime earnings from migration tend to belargest for those emigrating early in theirworking life. The surge in immigration sincethe 1980s has established large diasporas inhigh-income countries, which help to reducethe costs and risks of migration (see chap-ter 3). The demand for immigrant services inhigh-income countries will also rise as the


28

Percentage


World Developedcountries

Developingcountries

Figure 2.2 Share of females ininternational migration

43

44

45

49

47

51

48

46

50

Note: Stock data, reported by countries of destination.

2000

1970

Percent


Workers Reunitingfamilies

Refugee/asylum

Figure 2.3 Immigration to selectedcountries, reasons for admittance, 2001

0

10

20

60

40

80

50

30

70

Note: Stock data, reported by destination countries.

France

Australia

UnitedStates

UnitedKingdom

aging of the population shrinks the workforceand increases demand for services thatimmigrants can supply (such as nursing care).As income standards rise, the demand forother services that employ migrants (such ashousehold and restaurant help) should growrapidly. The intensifying competition forskilled workers may also draw migrants, espe-cially from countries with strong systems ofhigher education.

Policies in destination countries can affect migrationForecasts of migration flows remain problem-atic. But with the underlying demand for andsupply of migrants likely to increase in com-ing decades, the number of migrants will de-pend on policy decisions governing admit-tance and the effectiveness of efforts to policeborders and enforce workplace rules. Opposi-tion to immigration may grow as the numberof migrants increases, as it did in majorcountries of destination before World War I.But it is likely that the main policy issue willbe how best to manage and live with in-creased migration. In the simulations that fol-low, we explore the impact of an increase inmigration to 2025 in line with recent histori-cal experience.

The demographic challenge

The labor force in the high-incomecountries is set to declineA key driver in the demand for internationalmigrants over the next 20 years will be slow-ing growth, and then decline, of the laborforce in high-income countries. The age groupthat supplies the bulk of the labor force(15–65 years old) is expected to peak near500 million in 2010, and then fall to around475 million by 2025 (figure 2.4). In Japan thisage group has already begun to shrink, whilein Europe the peak will be reached in2007–08. In the other high-income countries,the peak will occur later—around 2020 forthe United States and 2015 for the rest. As-

suming no change in labor-force participationrates, the high-income countries may loseabout 20 million workers by 2025, relative topeak employment.8

The expected decline in the labor force isaccompanied by a rise in the overall depen-dency ratio, defined as the ratio of nonworkersto workers. For the high-income countries as agroup, this ratio is forecast to remain at justunder one through 2009. However, by 2025,100 workers will be supporting 111 depen-dents, largely reflecting the increased numberof the elderly (also, in most countries the num-ber of children under 15 will fall). The largestrise in the dependency ratio will be in Europe.If we focus more narrowly on the number ofelderly per worker, every 100 European work-ers now support 36 elderly people; by 2025they will have to support 52. In Japan 100workers will support 60 elderly in 2025.

In the developing countries the laborforce will expandDeveloping countries show considerable di-versity in demographic trends, but overall thebulge of youths born over the last two decadesis now entering the labor force, the number ofelderly is as yet still rising slowly, and thenumber of births is falling rapidly. Thusdeveloping countries are forecast to addnearly one billion workers to the world’s laborforce by 2025, again assuming no change inthe labor-force participation rate, and depen-dency ratios are expected to fall.

The expected expansion of the labor forcein developing countries, coupled with largewage premiums in high-income countries,means that migration could help reduce de-pendency ratios in high-income countries.However, increases in immigration sufficientto have a noticeable impact on dependency ra-tios would have to be very large. The scenariodiscussed below envisions an increase in thelabor force in high-income countries of 3 per-cent through migration, or a hike of nearly50 percent in working migrants in high-incomecountries. Even if migrants come with noelderly, the dependency ratio in the host coun-


29


30

Europe Japan

Figure 2.4 Labor force and dependency rates

Labor force, millions

Dependency rate Dependency rate

170

175

180

185

190

200

195

1.3

0.8

0.9

1.0

1.1

1.2

2001

2004

2007

2010

2013

2016

2019

2022

2025

1.3

0.8

0.9

1.0

1.1

1.2

2001

2004

2007

2010

2013

2016

2019

2022

2025

1.3

0.8

0.9

1.0

1.1

1.2

2001

2004

2007

2010

2013

2016

2019

2022

2025

1.3

0.8

0.9

1.0

1.1

1.2

2001

2004

2007

2010

2013

2016

2019

2022

2025

1.3

0.8

0.9

1.0

1.1

1.2

2001

2004

2007

2010

2013

2016

2019

2022

2025

1.3

0.8

0.9

1.0

1.1

1.2

2001

2004

2007

2010

2013

2016

2019

2022

2025


40

45

50

55

60

70

65

United States Other high-income countries


Dependency rateLabor force, millions Dependency rate

140

145

150

155

160

170

165


50

55

60

65

70

80

75

Total, Developing countriesTotal, High-income countries

Dependency rate Dependency rate

470

475

480

485

490

500

495


2,400

2,600

2,800

3,000

3,200

3,600

3,400

Source: World Bank databanks (DDP).

Note: The dependency rate is measured as the ratio of the nonworking population to working population.

Labor force Dependency rate

tries would fall by only about 3 percent undersuch a scenario. In the case of Japan, it wouldlower the number of elderly dependents in2025 from 60 per 100 workers to 59 per 100workers—barely a dent. Nevertheless, asdiscussed in more detail below, selectivemigration—for example, of experienced andskilled workers—can help mitigate the transi-tional costs of financing pension benefits forrapidly aging populations in high-incomecountries.

Migration and its developmentimpact

To illustrate the potential gains fromincreased migration, we compare the base-

case forecast for output and consumptionin chapter 1 with an alternative scenario, inwhich the stock of migrant workers is allowedto increase in the high-income countries soas to raise the overall stock of workers by 3 per-cent (a movement of 14.2 million workersfrom developing countries to high-incomecountries by the year 2025). A firstapproximation of the global gains from such ascenario is simply to calculate the income gainsaccruing to the new migrant workers—thiswill reflect the gains to the global economy,because it approximates the increase in globalproductivity derived from equipping the mi-grants with more and improved capital andtechnology. This back-of-the-envelope calcula-tion yields an increase in gross wage income of$772 billion in 2025.9 As we will see later,when corrected for differences in prices thatmigrants face in high-income versus develop-ing countries, and taking into account otherimpacts of migration (on prices, for example)as calculated by the model, global gains fall to$356 billion—an 0.6 percent increase in globalincome. The scenario is particularly beneficialto developing countries relative to high-incomecountries. The aggregate percentage gain todeveloping countries (including the new mi-grants) is 1.8 percent, whereas the gains tonatives in high-income countries amount to0.4 percent relative to baseline income. These

numbers hold up well as an approximation ofthe gains to global output, regardless of vari-ous assumptions made about taxes, non-wageincome distribution, key model parameters,and other factors.

Our modeling exercise uses a global gen-eral equilibrium model to measure the impactof migration (box 2.1).10 One of the purposesof the global model is to verify the basic intu-ition described above—that migration pro-duces a sizeable global gain. But it also is apowerful tool to evaluate distributionalimpacts—between skilled and unskilled work-ers, between native- and foreign-born workers,between capital and labor, and acrossregions—and to show how these distribu-tional impacts vary with policy choices andparameters (for example, the role of fiscalpolicies or the propensity to remit).

The assumption is that migrants as ashare of population remain constant in the baseline scenarioWe begin with a base case for global eco-nomic activity (outlined in chapter 1), demo-graphic trends (described at the outset of thischapter), and for migration. For the base case,the proportion of migrants in each region re-mains the same over time—somewhat con-trary to the trends of the last two decades.This does not imply that gross migration isstagnant, or even declining. The stock of mi-grants in any year will equal the previousstock of migrants, plus new migrants, less theattrition through death and return migration.We chose a relatively neutral assumption be-cause of the difficulty in forecasting thesecomplex processes. For some countries—forexample Japan and those in Europe—the as-sumption results in an absolute decline in thestock of migrant workers. This decline paral-lels the overall decline in the European andJapanese labor forces.11 For the high-incomecountries as a group, the stock of migrantworkers would increase by some 760,000between 2001 and 2025, just a small incrementfrom the estimated 27.8 million in 2001. Themain issue, however, is not the base case,


31

but rather the impact of deviations fromit—although significantly different base as-sumptions could affect the deviations aswell.

According to the base-case scenario, mi-grant workers would make up about 6 percentof the labor force of high-income countries in2025, though with sharp differences acrossregions and skills (table 2.2). The vast major-ity of migrant workers are unskilled—some25.3 million migrant workers out of a pro-jected total of 28.5 million, or 7.8 percent ofhigh-income countries’ labor force. Skilled mi-grants, on the other hand, represent just 2.2 per-cent of the total skilled workforce on average.

There are welfare implications ifmigration rises significantlyThe alternative scenario involves a rise inmigration sufficient to increase the labor force

of high-income countries by 3 percent, phasedin from 2010 through 2020.12 As migrantsmake up about 6 percent of high-incomecountries’ labor force, a 3 percent rise in thelabor force (through migration) implies a50 percent increase in the number of migrantworkers. This may seem like a large change,but the resulting stock of migrants in Europe,Japan, and the United States would remain afar smaller share of population than currentlevels in some high-migration countries. (InAustralia, for example, about a quarter of thepopulation are migrants, in Canada 19 per-cent, in Kuwait 50 percent). The percentageincrease in migrants is large in Japan (as thebaseline share of migrants is relatively low),and lower in the United States. The increasecorresponds to an annual growth rate ofabout 1.9 percent, somewhat slower than the


32

The underlying analytical framework used in thischapter is the World Bank’s standard global

general equilibrium model—LINKAGE—which hasbeen used in previous reports for trade policy analy-sis. It has been modified to differentiate between mi-grant and native workers and to incorporate remit-tances. The model is based on release 6.0 of theGTAP database (base year 2001), developed by theGlobal Trade Analysis Project (www.gtap.org), aglobal network of researchers and policymakers en-gaged in the quantitative analysis of internationalpolicy issues. It is supplemented for use in ourmodel with a new database developed jointly byGTAP and the University of Sussex (Parsons andothers 2005). That database contains a comprehen-sive estimate of bilateral stocks of migrants for 226countries and territories.

While the new migration database is undergoingconstant improvements as new data become avail-able and obvious errors are corrected, its developershave done a remarkable amount of detective work,largely in national data sources. The GTAP centerhas used this underlying migration database to builda bilateral migration database for the 87-region levelof aggregation of the main GTAP database—

Box 2.1 The model used in this studyincluding estimates of population and the stock ofworkers, both skilled and unskilled (Walmsley,Ahmed, and Parsons 2005).a World Bank data (de-scribed in more detail in chapter 4 of this report)was used to provide the total level of remittances,and the bilateral stock of migrants was used to esti-mate the bilateral remittance flows subject to theoverall total flows.

The standard horizon for the LINKAGE model hasbeen 2015. For the work described here, the modelhorizon has been extended to 2025, in part becausedemographic dynamics play a more important roleover the longer-term horizon, and in part to allowfor more time to phase in the increase in migration.

aThe 87 regions of GTAP have been aggregated into 21 re-gions for the purposes of this study. Six of these are high-in-come regions using World Bank definitions—the EuropeanUnion and the European Free Trade Area, Canada, the UnitedStates, Japan, Australia/New Zealand, and the newly-industrial-izing economies. The fifteen developing countries/regions in-clude China, the Philippines, India, Russia, Turkey, SouthAfrica, and Mexico as individual countries, plus 6 regions thatrepresent the remaining countries in each geographical area.

average increase over the period 1980–2000.Moreover, the growth rate is unbalanced, withan annual increase of only 1.5 percent in un-skilled workers, but 3.8 percent in skilledworkers. A number of additional assumptionsare critical to the results.

First, the high-income countries’ laborforce of both skilled and unskilled workersincreases by 3 percent.13 As the share of skilledworkers among migrants is much smaller thanthe share of skilled workers among high-income country natives, the shock results in amuch larger percentage increase for skilledmigrants. The number of unskilled migrantworkers increases by 39 percent, while thenumber of skilled migrant workers rises by138 percent.14

Second, the share of migrants by region oforigin remains constant; in other words, thenew migrants reflect the same allocation byregion of origin as existing ones. Thus ifMexicans constitute 30 percent of foreign mi-grants in the United States in the base case,they maintain the same share after the increasein migration. This assumption is made to sim-plify the analysis, although it does fail to

reflect the likely migration pressures impliedby large differences in demographic trends insending regions (for example, Sub-SaharanAfrica versus Latin America).

Third, foreign workers are assumed tobring family members in proportion to thedependency ratio in their home country. As aresult, the total number of migrants in high-income countries increases from 65 million(6.5 percent of high-income countries’ popula-tion) in the baseline for 2025, to 93 million (9 percent of population) after the shock. Thisassumption can change the average depen-dency ratio of the host country. It canalso have other implications not modeledexplicitly—including fiscal impacts, becausethe families of new migrants may require ad-ditional public services (such as schooling),not fully compensated by the taxes paid by thenew migrants.

Fourth, remittances are assumed to be afixed proportion of migrants’ labor income,equal to the level in the base year. The averagefor developing countries is 17 percent, althoughthe level varies with the migrant’s origin anddestination countries. New migrants are


33

Table 2.2 Labor force structure in the base case and after increases in migrantsIn millions except where noted

Baseline Migration shock

2001 2025 Change in millions 2001–25 Change in percent 2001–25

High-income countriesTotal labor force 480.8 474.0 14.2 3.0Developing-country migrant workers 27.8 28.5 14.2 49.9

Unskilled 24.6 25.3 9.8 38.6Skilled 3.1 3.2 4.5 137.9

Developing-country migrant workersas share of total labor force, percenta 5.8 6.0 8.8Unskilled, percent 7.4 7.8 10.5Skilled, percent 2.1 2.2 5.0

Developing countriesTotal labor force 2,596.2 3,561.0 –14.2 –0.4

Unskilled 2,395.9 3,294.3 –9.8 –0.3Skilled 200.4 266.7 –4.5 –1.7

Source: Initial 2001 data from migration database under development by GTAP/University of Sussex (Parsons and others 2005and Walmsley, Ahmed, and Parsons 2005). Scenarios based on World Bank assumptions.Note: a. The percentage of migrant workers as a share of the total labor force is assumed to be the same for each individualregion of the model throughout 2001–25, but the share averaged across all developed regions will change through aggregationeffects.

assumed to send remittances to their homecountry at the same rate (relative to income)as existing migrants.15

Returns to households

The gains from increased migration are largeWith the labor force moving, it is best toassess the effects on real income in terms ofhouseholds as opposed to the national level(as is typically done in analyses of trade re-form). Households are broken down into fourgroups. First are the native households inhigh-income countries.16 Second are previousmigrants from developing countries now liv-ing in high-income countries, that is, thosewho were in place in the baseline scenario.Third are native households in developingcountries—households that do not migrate.17

And finally, we have the households of thenew migrants. Each household’s welfare isbroken down between the change in privateconsumption and the change in the consump-tion of public services.

Natives in high-income countries gain$139 billion in real income, or 0.4 percent ofthe baseline, as a result of the rise in migration(table 2.3). Nonmigrating households in de-veloping countries see a rise in real income of

nearly 0.9 percent from baseline levels.18 Asignificant portion of the increase is due to theremittances from the new migrants, with someimprovement in labor-market conditions forremaining workers. Those who are likely tolose—in the absence of any compensatorymechanism—are the existing migrants inhigh-income countries, who are relativelyclose substitutes for the new migrants. Theirprivate consumption would decline by over9 percent and overall consumption (includingpublic services) by 6 percent compared tobaseline levels.

New migrants and their countriesof origin reap benefits (through remittances) The main gains come from the higher incomesthe new migrant workers can earn in the des-tination country relative to what they wouldhave earned in their country of origin. Newmigrants earn $481 billion in real (after-tax)income in 2025 over the base case. However,the dollar increase in income overestimates thewelfare gains for migrants. Essentially, anadditional $1 spent in the high-income coun-tries does not provide the same amount ofwelfare as an additional $1 spent in the homecountry, because prices are higher in high-income countries. Whereas the prices of


34

Table 2.3 Change in real income across households in 2025 relative to baseline

Real income adjusted forReal income cost of living

Private Public Total Private Public Total

Natives in high-income countries 139 –1 139 139 –1 139Old migrants in high-income countries –88 0 –88 –88 0 –88Natives in developing countries 131 12 143 131 12 143New migrants 372 109 481 126 36 162World total 554 120 674 308 48 356

Natives in high-income countries 0.44 –0.01 0.36 0.44 –0.01 0.36Old migrants in high-income countries –9.41 –0.02 –6.02 –9.41 –0.02 –6.02Natives in developing countries 0.94 0.44 0.86 0.94 0.44 0.86New migrants 584 607 589 198 203 199World total 1.20 1.15 1.19 0.67 0.45 0.63

Source: World Bank model simulations.

Change, $ billions Change, $ billions

Change, % Change, %

traded goods (for example, cars and electron-ics) are the same worldwide, at least in princi-ple, the prices of nontraded goods and services(for example, housing and haircuts) are muchhigher in high-income countries.

A simple example may clarify the idea.Take a household of two persons living intheir home country. One works and earns$200. The other does not work. Each spends$100, half on tradable goods (each priced at$1) and half on nontradable goods (likewisepriced at $1). Now the worker moves to ahigh-income country and earns $700. Assumethat spending patterns do not change. Theworker remits $200 back to the home country,so the income (and welfare, in money terms)of the other doubles. The new migrant buysthe same goods—50 units of tradable goodsand 50 units of nontradable,19 but the price ofthe latter is now $9 and not $1. The migrantthus spends $500, but welfare is unchanged,because the basket of purchased goods isidentical.

Welfare evaluations are of course morecomplex than this simple example illustrates.For one thing, new migrants will have to adjusttheir spending patterns to deal with their newenvironment. Heating oil and warm clothesare necessities that will not boost a migrant’swelfare above what it was in the home coun-try. For another, the decision to migrate is nottaken for simply static reasons; there are sig-nificant dynamic reasons for migrating—forexample, better opportunities for one’s chil-dren that are not captured in this simple frame-work. Nonetheless, the difference in purchas-ing power illustrated in the example is a strongmotivation for migrating, even on a temporarybasis. The more wage income earned in high-income countries that can be spent in lower-income countries, the greater will be the wel-fare benefits. Box 2.2 provides additionaldetail on the computation and interpretationof global welfare gains from migration.

To account for the change in prices facedby the new migrants, their “new” consump-tion in the destination country is adjusted toaccount for differences in the cost of living,

using purchasing power parity (PPP) exchangerates from the World Bank’s database.20 Thusinstead of an increase of $481 billion, the risein welfare for new migrants is $162 billion.21

Table 2.3 shows the change in these com-ponents for the four household groups and theworld. Measured in national accountingterms, that is, with no adjustments for the dif-ference in the cost of living for the new mi-grants, global real income rises under themodel by 1.2 percent relative to the baseline,or 0.6 percent with the cost-of-living adjust-ment. Global private consumption increases inreal terms by $308 billion in 2025 (with thecost-of-living adjustment), with real govern-ment expenditures increasing by an additional$48 billion. The total real gain—with equalweight for high-income—and developing-country gains—is $356 billion, with justunder half accruing to the new migrants,though natives in both high-income and devel-oping countries also are better off. In percent-age terms—where relative weights betweenhigh-income and developing countries areirrelevant—the scenario clearly indicates thatthe relative gains are much higher fordeveloping-country households than high-income country households, rivaling gainsfrom global reform of merchandise trade.

Obviously, global income and global gainswould also be larger if expressed in PPP terms.As the percentage increase in welfare for mi-grants living (originally) in developing coun-tries is larger than the percentage increase forthose living in high-income countries, a switchto PPP measures would also increase theglobal gains as a percentage of global income.If in the migration scenario presented here thegains are PPP-adjusted, the global gains wouldamount to 0.9 percent of global income in thebaseline, instead of 0.6 percent using the EVaggregation. This scenario illustrates that mi-grants living (originally) in developing coun-tries gain the most from migration in percent-age terms.

The impact of higher migration on prices ismild in aggregate in high-income countries,with a small decline in the average price of


35


36

Box 2.2 Calculating and interpreting global welfaregains from migration

national accounting standards. However, the stan-dard real income measure is still a good approxima-tion of the welfare gains for the other households inthe model.

To the extent that new migrants remit part oftheir income to their country of origin and thatincome is spent in that country of origin, the increasein the cost of living that new migrants face is notrelevant. Therefore, the EV measure of remittances islarger than the same nominal income spent by thenew migrant in the host country. This differenceillustrates the incentive for new migrants to remitincome home.

Aggregation. The second issue relates to the inter-pretation of the “global” gains. Typically, to deriveaggregate or global gains, EV (expressed in a com-mon currency, typically the U.S. dollar) is summedacross all households. For individual persons or ho-mogeneous groups this EV aggregation, expressed asa percentage of original income, is a good approxi-mation of the change in welfare (or more precisely, itis a good indication of the change in welfare).e

However, no clear link exists between global wel-fare and the aggregation of EV across heterogeneousgroups, because we do not know how to weigh indi-vidual welfare across heterogeneous groups (a partic-ularly difficult issue in aggregating across countriesat very different stages of development, as is donehere). For example, while most groups gain from mi-gration in the scenario discussed in the text, somelose. The fact that the change in global welfare (ex-pressed as the aggregation of EV across groups) ispositive does not mean that the welfare gains of thewinners are considered more important than the wel-fare losses of the losers. Thus, global gains as ex-pressed in aggregate EV should not be interpreted asa value judgment on how to weigh individual orlocal welfare gains.

The aggregation of EV across groups does,however, have a useful interpretation, which islinked to the notion of compensation and Paretooptimality. As long as the global gains are positive—using the standard practice of adding up EVs acrosshouseholds—then it is possible through redistribu-tion to compensate households that lose (so that noone is worse off relative to the baseline scenario),

Two sets of issues arise with respect to the so-called global gains from a policy shock. First,

how should the gains of specific groups be evaluatedand how do the gains compare with traditional mea-sures, such as GDP or national accounting stan-dards? Second, how should the gains be aggregatedover groups and countries, and how should theaggregated gains be interpreted?

Evaluation of the welfare gains of specificgroups. In standard applications of general equilib-rium (GE) models, the welfare impacts of specificgroups are evaluated using a concept from welfaretheory called equivalent variation (EV). The con-cept is relatively straightforward. Welfare changesas a result of changes in nominal income andchanges in prices. EV calculations summarize thiswelfare change in terms of an equivalent changein income alone, showing by how much incomeat original prices would have to change toachieve the same change in welfare as observed ina simulation.a

For most households, the standard notion of thechange in real income, that is, the difference in nomi-nal income adjusted by the change in the CPI, is agood approximation of EV.b

This is not the case for new migrants, however.There is no standard price index that can be used asa deflator for the change in the nominal gains for thenew migrants, since the prices they face in their newhost country have no linkages to the prices they paidin their home countries. GE and macro models typi-cally calibrate base-year prices in each region to one(or unit value) by choosing corresponding volumeunits.c

This approach does not allow one to take into ac-count the price increases that new migrants face asa result of their migration. In the simulations, themacro PPP exchange rate (as an approximation ofthe rise in prices faced by migrants from developingto high-income countries) has been used to adjust thegains to the migrants—although this is just an ap-proximation of the true welfare gains.d

Because of the cost-of-living adjustment to thewelfare gain of new migrants, the real gain reportedis no longer equal to real income gains ofcountries—and real output gains—measured using

absorption (private consumption, private in-vestment, and government spending) of 0.1percent. However, prices of some key nontrad-ables decline by larger amounts—0.8 percenton average for public services (includinghealth-related services) and 0.2 percent for con-struction and recreational services. These pricedeclines will be even sharper for specific sub-sectors where migrant workers are concen-trated (for example, household help), for whichwe currently have no comprehensive data.

The allocation of the gains across develop-ing countries depends on various factors, in-cluding the skill loss and the resulting impacton production, the locations to which mi-grants move and the relative wage differential,and the propensity to remit. By developing re-gion, the gains to households under the modelvary from 0.6 percent for Europe and CentralAsia to 1.1 percent for South Asia and LatinAmerica and the Caribbean (table 2.4).

For the new migrants, the real incomegains—cost-of-living adjusted—increase bynearly 200 percent. There are large differencesacross regions, with the highest gains (in per-centage terms) accruing to migrants from Sub-Saharan Africa (619 percent) and the lowestto migrants from the Middle East and NorthAfrica and Europe and Central Asia. The mainreason for the disparity is the relative differen-tial between wages in origin and destinationcountries. Variations in wages paid to mi-grants from different regions in destinationcountries are minor, whereas there are verywide variations in wages in countries of ori-gin. For example, the average wage for a mi-grant in Europe in the base year is about$16,500—with only minor variation acrossmigrants. However, the average wage in Sub-Saharan Africa is only $470, whereas in theMiddle East and North Africa it is $2,700.Thus, the migrant from Sub-Saharan Africa


37

Box 2.2 (continued)developing countries in global aggregates would in-crease in the measurement of both global incomelevels and global welfare gains. However, the per-centage increase in income for developing countrieswould not be affected.

aOne of the advantages of the EV measure is that it trans-forms the ordinal concept of welfare into a cardinal concept ofincome. While it is impossible to measure how much one welfarelevel differs from another (one can only conclude that one level ispreferred to another), the corresponding increase in income canbe measured, and the size of the increase has a clear meaning.

bFor example, in trade-reform scenarios, the change in theprice index is a relatively good approximation of the welfareimpact, since the new price is approximately the old price lessthe tariff.

cThere are exceptions. For example, in the case of climate-change models, it is necessary to know the relative prices of thedifferent fuels to accurately determine the carbon tax.

dSee Timmer and van der Mensbrugghe (2005) for moredetails.

eThe size of the change in individual welfare is undeter-mined, since welfare is an ordinal concept.

when some households are better off. In that sensethe global gain can be compared with an equal risein global output plus redistribution.

In this report we maintain this standard practiceof reporting EV aggregates, making the gains compa-rable to global gains in many other studies.

An alternative approach to calculating globalgains would be to add up changes in income mea-sured in PPP terms. The rationale for that alterna-tive is that because prices of nontraded goods arelower in developing countries, the addition of a dol-lar to a developing country would enable the pur-chase of a larger amount of goods and services thanin an high-income country. In that case, both baseincome and gains for new migrants and for thosewho remain in developing countries would beroughly three times as large as reported here. Thisis true for all gains, whether they come from migra-tion itself, from remittances, or from changes inwages and prices in developing countries. As a re-sult, the share of those who live (originally) in

will gain much more in both absolute and per-centage terms than one from the Middle Eastand North Africa.

The impact of migration on tradewould be mildWhether migration and trade are substitutesfor each other is an old debate. For exam-ple, in the discussions leading up to the sign-ing of NAFTA—the free trade agreementamong the United States, Canada, andMexico—one of the key arguments was thattrade would replace migration and reducethe pressure for Mexicans to migrate to theNorth. Likewise, allowing for increasedmigration—for example of unskilledworkers—could reduce trade, because itwould enable the high-income countries tocontinue producing low-skill-intensive prod-ucts at competitive cost.

Evidence of the link between trade andchanging the comparative advantage emergesin the migration scenario described here. Forexample, the largest gains in export revenuefor high-income countries come in agriculture,clothing, other manufacturing, recreationalservices, and public services—all labor-intensive sectors, the first four being relativelyintensive in unskilled workers and the last inskilled workers.

Change in comparative advantage has onlya mild impact on trade flows in this scenario,however, as migration affects trade through

several channels, some of which increase, andothers that decrease, trade flows:

• First, the rise in incomes due to migra-tion produces a small rise in global tradeflows, with regional differentiation (be-cause income gains differ considerablyamong regions). In addition to higher in-comes, the rise in migration changes thesize of regional economies, with implica-tions for their demand for imports andability to export.

• Second, the nature of the shock assumedin our model differs from the standarddebate over trade and migration. Theshare of skilled workers in total migrantsis larger in the shock than in actual mi-gration over the recent past. A largeproportion of skilled workers will findemployment in nontraded sectors—forexample, as doctors and nurses—ratherthan in producing traded goods. Thiswill have general equilibrium effects tothe extent that the price of nontradedgoods will decline by more than the priceof traded goods. Thus there will be a rel-ative shift to nontraded goods and apotential reduction in demand for im-ports of traded goods. Overall, the largershare of skilled versus unskilled workersdoes tend to reduce trade flows.

• Third, the increase in remittances pro-vides an opportunity for developing


38

Table 2.4 Real income impacts across developing regionsChange in 2025 relative to the baseline, adjusted for differences in cost of living

Natives in region New migrants from region

$ billions Percent $ billions Percent

Total developing 143 0.9 162 199East Asia and Pacific 37 0.7 32 215South Asia 21 1.1 2 175Europe and Central Asia 14 0.6 25 138Middle East and North Africa 18 0.9 11 134Sub-Saharan Africa 7 0.9 7 619Latin America and the Caribbean 47 1.1 85 224


countries to import more and export less,as their current-account balance will in-crease by the size of the remittances ($98billion in net terms). The model resultsshow that total imports into developingcountries would increase by $58 billionin 2025 (1.1 percent relative to the base-line), as aggregate exports decline by$40 billion (0.7 percent).22 The changein remittances leads to an appreciation ofthe real exchange rate and therefore aloss in relative export competitiveness.23

For instance, the output price index indeveloping countries rises by 0.6 percenton average, whereas it declines by 0.1 per-cent for high-income countries.

In summary, the scenario provides evidencethat changes in comparative advantage due tomigration do influence trade flows. However,overall migration and trade are not substitutesfor each other, because migration has manyother economic effects that have more powerto stimulate or reduce trade. One implicationof this finding is that migration policies shouldnot be pursued because of their specific impacton trade flows. Likewise, in trade policies theimpact on migration should not be a mainfocus.24 Trade and migration policies shouldbe evaluated on their own merits.

Migrants’ impact on government fiscalaccounts is broadly neutralThe assumption concerning the level ofconsumption of public goods and services bynew migrants has important implications forindividual gains, and global gains, under themodeled scenario. We assume that the newmigrants’ level of consumption of publicgoods and services equals the amount they payin taxes, that is, their impact on the publicbudget is revenue-neutral. This is broadly con-sistent with the available evidence (box 2.3).To provide some sense of how different ap-proaches would affect the scenario results, wepresent two alternative assumptions regardingthe distribution of public goods and servicesto the new migrants (table 2.5). The defaultassumption had a largely neutral impact forexisting residents in the host country. Underanother assumption—new migrants pay taxesbut receive no benefits from public goods andservices—existing residents, native and mi-grant, enjoy a rise in real incomes of $126 bil-lion ($117 billion for natives and $9 billionfor existing migrant households). Note thatthe global welfare gains increase as well, sincethe income accruing to natives (and existingmigrants) is not adjusted for the differences inthe cost of living between developing andhigh-income countries.25 A second extreme


39

Table 2.5 Impact of different assumptions on the consumption of public goods and services by selected groups in 2025Change in cost-of-living-adjusted real income in 2025; billions of dollars

Defaultassumption—

“New” migrants “New” migrants receive benefits receive no public “New” migrants equal to their benefits but pay receive per capita

taxes taxes average benefit

Private Public Total Public Total Public Total

Natives in high-income countries 139 �1 139 117 256 �85 54Old migrants in high-income countries �88 0 �88 9 �79 �6 �94Natives in developing countries 131 12 143 12 143 12 143New migrants 126 36 162 �18 108 75 201World total 308 48 356 120 428 �4 304


assumption is that new migrants receive thesame amount in public benefits as the averagehousehold in the destination country. Thiswould imply a net positive transfer to the newmigrant households, since they would receivemore in public benefits than they paid in

taxes.26 In this case natives in high-incomecountries would lose $85 billion in aggregatepublic goods and services, although thisamount would not translate one-for-one into abenefit for new migrants due to the cost-of-living adjustment. These simulations underline


40

Box 2.3 The impact of immigrants on fiscal balances60s generally imposed a long-term burden. Studiesthat follow immigrants over time generally concludethat in net-present-value terms, immigrants and theirdescendants tend to contribute more in terms of taxrevenues than they absorb in expenditures, but the or-ders of magnitude are typically small (OECD 1997).Intergenerational models are sensitive to the discountrate used and assumptions concerning the allocationof the fiscal burden over future generations.

Third, the computation will depend on the levelof skills, experience, education, and fertility of immi-grants. Rowthorn (2004) calculates that skilled mi-grants to the United States typically make a largepositive contribution to the fiscal balance, whereasunskilled immigrants cost more on average than thetaxes they pay. Storesletten (2000) calculates that thenet-present-value contribution of the average high-skilled immigrant to the U.S. budget is $96,000; themedium-skilled immigrant’s contribution is �$2,000;and low-skilled immigrant’s contribution is�$36,000.

The results may change over time, as migrantcharacteristics and government policies change. Theprobability that an immigrant to the United Stateswill receive public benefits has risen since the 1970s,probably due to an increasing share of immigrantsfrom poorer countries (Gustmann and Steinmeier1998).

An issue of particular concern has been the im-pact of migration on government-financed pensions.Likely increases in immigration can make only asmall net contribution to strengthening the financingof pensions in the United States (Fehr, Jokisch, andKotlikoff 2004), although selecting immigrants forworking age and high skill levels could improve thepicture (Storesletten 2000). By contrast, increases inimmigration could make a significant contributionto financing pensions in Germany (Bonin, Raffel-huschen, and Walliser 2000) and Spain (Collado,Iturbe-Ormaetxe, and Valera 2004).

Immigrants’ net contribution to fiscal revenues isusually considered to be small. The net fiscal im-

pact of immigration on the United States has beenminimal (Coppel, Dumont, and Visco 2001;Auerbach and Oreopoulos 1999). The U.S.Binational Study on Migration (1997) found that ir-regular migrants did impose a significant fiscal bur-den on state and local government. However, schoolexpenses accounted for the bulk of these costs, and(as the authors note) education is an investmentthat may readily be recovered in greater future pro-ductivity. Moreover, Lee and Miller (2000) foundthat the overall fiscal consequences of altering thevolume of immigration to the United States wouldbe quite small. Gott and Johnston (2002) andSriskandarajah, Cooley, and Reed (2005) estimatedthat immigrants made a positive contribution topublic finances in the United Kingdom. Gustafssonand Osterberg (2001) found that new immigrants toSweden generated a net fiscal cost, but this turnedinto a positive contribution after a few years. Nanaand Williams (1999) found that immigrants to NewZealand had a positive fiscal impact. Bonin,Raffelhuschen, and Walliser (2000) found that thenet fiscal contribution of immigrants to Germanycould be significant if the government selects forskills.

Calculations of the net fiscal cost of immigrationare fraught with difficulties, for several reasons.

First, the computation at any point in timedepends heavily on the methodology used, whatexpenditures and revenues are included, which publicservices should be regarded as pure public goods (andthe extent of economies of scale in expenditures), andwhether households or individuals are considered.

Second, static calculations of the current net fiscalimpact fail to take into account the age structure ofthe immigrant population. Smith and Edmonston(1997) found that immigrants arriving between theages of 10 and 25 years produced fiscal benefits undermost scenarios, while immigrants arriving in their late

the effect of public policy on the distributionof gains from migration.

Additional gains from migration can be substantialThe gains for migrants from this scenarioessentially provide the same message as ear-lier estimates. In their seminal paper, Walmsleyand Winters (2003) estimate that a relax-ation on the movement of temporary work-ers on the same order as that modeledhere—that is, 3 percent of the labor force ofthe high-income countries—would yieldglobal income gains of $150 billion (using a1997-based comparative static model). Theresult from our scenario that is roughlycomparable to their figures (that is, globalgains before adjustment for cost of livingand measured relative to 2001, rather than2025) are more than double their results.27

However, our figures are comparable withthe more recent work done by Walmsleyand her colleagues.28 One of the key reasonsfor the increase in the global welfare impactis a reevaluation of the assumed wage dif-ferential between the home and host coun-try. In their initial work, Walmsley andWinters had assumed that new migrantsmade up 50 percent of the difference be-tween the home and host country’s wages.Their new assumption (used in our model aswell) is 75 percent, based in part on the factthat the migrants are permanent rather thantemporary. Hamilton and Whalley (1984)and Moses and Letnes (2004) have shownthat removing all restrictions on labormovement, admittedly not a realistic sce-nario, would yield a huge increase in worldoutput. Overall, these papers suggest thatlabor-market restrictions are imposing amuch larger burden on the global economythan are trade restrictions. The World Bank’strade model suggests that removing all re-maining merchandise trade barriers wouldyield $287 billion in global real income gainsin 2015. For the purpose of comparison, whenthe gains from the two different scenarios—

those from an increase in migration, and thosefrom global trade reform—are scaled to thesame reference year, 2001, the gains fromtrade reforms are $155 billion versus $175 bil-lion from the migration scenario.29 Thisleaves little doubt that easing restrictions onthe movement of labor could provide a sig-nificant boost to the global economy. More-over, in comparison with the most recentwork on global merchandise trade reform,the gains from an increase in migration aremore balanced toward income increases fordeveloping countries relative to developedcountries. In a study by Anderson, Martin,and van der Mensbrugghe (2005), the gainsto high-income and developing countries are0.6 and 0.8 percent, respectively, relative tobaseline income. In the scenario modeledhere, the income increases are 0.4 percentfor native households in high-income coun-tries and 1.8 percent for developing coun-tries (including the new migrants).

Returns to factors of production

Four critical factors determine the distribu-tion of gains from migration among skilled

workers, unskilled workers, and owners ofcapital: (a) the size of the increase in migra-tion; (b) the distribution of nonwage income(profits); (c) the degree of substitution betweenworkers by region of origin; and (d) the degreeof substitution or complementarity betweenworkers and capital. We have already positedthat the increase in migration is large, with anaverage increase in the migrant labor force ofaround 50 percent over a 20-year period, andcomparable (if somewhat less) to the rise in theshare of migrants in high-income country pop-ulation over 1970–2000. In the absence of anyspecific data on the source of migrant income,we assume that migrants—both existing andnew—receive no nonwage income. In essence,their real income will be driven by changes inwages. The effects of this simple assumptionon the distribution of gains are significant, andthe implications of relaxing it are discussedbelow.


41

Substituting between migrant and nativeworkers determines who gainsThe key issue of who reaps the benefits in-volves the degree of substitution among dif-ferent workers. The allocation of demand forworkers assumes differentiation among work-ers from different regions. This is done in twosteps. First, “similar” workers are bundled to-gether into “native” workers and “foreign-born” workers.30 In the second step, thesetwo bundles are decomposed into labor de-mand by region of origin. We assume thatthere is more differentiation between a nativeand a foreign-born worker (that is, a lowersubstitution elasticity) than between twoworkers from different countries of originwithin each of the two aggregate bundles. Forexample, in the case of the United States, em-ployers see a greater difference between a U.S.worker and a generic immigrant from a devel-oping country than between a Mexican and aSalvadoran worker. The implication is that arise in the supply of migrants has a greater im-pact on old migrants than on native workers,which plays a key role in the distributionaloutcomes of the increase in migration. Theassumption of labor demand differentiationoperates for both skilled and unskilled laborin the model.

In the default case, we assume that wagesare flexible, with a substitution elasticity be-tween unskilled migrants and natives that isroughly comparable to that implied in theconclusions of the meta analysis in Longhi,Nijkamp, and Poot (2004); they concludethat a “one percentage point increase in theproportion of immigrants in the labor forcelowers wages across the investigated studiesby only 0.119 percent.” (See box 2.4 for a re-view of empirical studies of the impact of mi-gration on wages.) In the scenario describedhere, the 50 percent increase in the stock ofmigrants raises their proportion of the laborforce by about 3 percentage points, produc-ing a 0.5 percent decline in the wages of na-tives.31 We also assume perfect substitutionbetween new and old migrants (the large ma-jority of both categories being unskilled). The

empirical evidence of the extent to which mi-grants are substitutes for natives or for exist-ing migrants is sparse. Thus in addition toexploring the implications of the assumptionsmade, we also devote attention to alternativeassumptions.

Finally, in a departure from previouswork but in line with a developing consen-sus, we assume that skilled workers are nearcomplements with capital (meaning that theyare more productive, and thus earn higherreturns, when used together with capital),whereas unskilled workers are substitutesfor capital and skilled labor.32 This specifi-cation has important consequences for thedistributional impacts of increased migra-tion. Whereas investment rises with in-creased income, the overall increase in thestock of capital is modest, so that the rise inthe supply of skilled workers is not matchedby an equivalent increase in capital. Thus themarginal productivity of additional skilledworkers declines, provoking a decline in thewage of skilled workers (by more than thefall in the wages of unskilled workers).

Increased migration can generatesubstantial changes in income distributionamong workers and owners of capitalThe change in factor returns is depicted infigure 2.5. In the high-income countries only


42

% change in factor returns, 2025, relative to baseline

Source: World Bank simulations.

Capital

Developing countries

High-incomecountries

Figure 2.5 Factor returns and migration

Unskillednative

workers

Unskilledforeign-

bornworkers

Skillednative

workers

Skilledforeign-

bornworkers

�25

�20

0

�10

10

�5

�15

5


43

Box 2.4 Empirical studies of the impact of immigration on wages

increased the unemployment rate by 0.2 percentagepoints. This shows more adjustment through em-ployment than through earnings compared to U.S.studies, which may be due to French rules governingwages (Dustmann and Glitz 2005). The comparisonunderlines the importance of the investment climate,and in particular, labor-market flexibility, in the effi-cient absorption of migrants. Some studies show thatimmigration reduces native unemployment in thelong term (Poot and Cochrane 2004), presumablybecause increased consumption demand from immi-grants raises the demand for labor.d

Thus some articles support the view that unskilledimmigrants are relatively close substitutes for nativeworkers (without attempting to distinguish betweenthe effects on native workers and old migrants). How-ever, the share of low-skilled native workers in desti-nation countries is falling. The share of U.S. adultswith less than a high-school education declined from47 percent in 1970 to 22 percent in 1998 (Massey2000). About 90 percent of new native entrants to theU.S. labor force in 2004 had completed high school(U.S. Labor Survey 2005). By contrast, the averagemigrant from rural Mexico has six years of educationand does not speak English (Mora and Taylor 2005).Many low-skill immigrants may have such limited ed-ucation and language skills that they do not competewith native low-skilled workers at all, but insteadtake jobs that natives are unwilling to do. In this view,the rise in immigration in high-income economiessince 1980 has been accompanied by increases in na-tive educational levels; essentially natives moved outof certain kinds of jobs, creating a demand for immi-grant labor.

aThe researchers do attempt to correct for endogeneity byusing instrumental variables.

bExamples include the Mariel boatlift from Cuba (Card1989), the repatriation of Algerians of European origin to France(Hunt 1992), the inflow of workers to Austria after the break-down of the communist regimes (Winter-Ebmer and Zweimuller1999), and the return of Portuguese from Africa in the 1970s(Carrington and de Lima 1996).

cStill, Card (2001) finds no evidence that immigration intoan area leads to offsetting net outflows of workers.

dSee Gross (1999) for this result for France.

Most cross-sectional studies find that immigrantshave no impact, or a very limited impact, on

the wages or employment of natives (LaLonde andTopel 1997 and Borjas, Freeman, and Katz 1997 forthe United States; Pischke and Velling 1994 forGermany). However, cross-sectional approaches re-late wage differences across local labor markets tothe share of immigrants in each market. If immi-grants are attracted to high-wage areas, which islikely, it is difficult to identify the exogenous impactof immigrants on wages.a Studies of sudden, politi-cally driven inflows of immigrants likewise fail to de-tect a significant impact on natives’ wages or employ-ment in affected areas.b However, native workersmay adjust to large, sudden inflows of immigrants bymoving to other areas (or through reduced inflowsfrom other areas), again obscuring the relationshipbetween immigration and labor-market outcomes.c

This problem has encouraged the use of paneltechniques that can discern the combined effects oftime and cross-sectional effects. Some panel studieshave found a significant impact on the wages of un-skilled natives, who in addition have suffered de-clines in wages due to skill-biased technical changeand increased trade. Borjas (2003a), analyzed the im-pact of immigration in the United States across dif-ferent levels of skills and experience and estimatesthat immigration reduced the wages of native high-school dropouts in the United States by 8.9 percentfrom 1980 to 2000. Jaeger (1996) finds that immi-gration lowered the real wage of U.S. high-schooldropouts by as much as 3.6 percent in the 1980s.DeNew and Zimmermann (1994) find that a onepercentage point rise in the share of migrants in thelabor force reduces the wages of blue-collar workersby almost 6 percent. By contrast, Dustmann and oth-ers (2003) find that immigration has little impact onnative wages (or employment) in the UnitedKingdom, including for the low-skilled.

Where wages are relatively inflexible, an inflow ofmigrants may affect employment levels rather thanwages. Angrist and Kugler (2002) find that increasedimmigration in Europe is associated with a signifi-cant decline in native employment, particularly forthe low-skilled. Hunt (1992) finds that a one per-centage point rise in the share of immigrants in theFrench labor force (following Algerian independence)

capital enjoys an increase in returns under themodel—with wages declining for all laborcategories, skilled and unskilled, native andforeign-born. With essentially only a laborshock, the scarcity value of capital increases.The negative impact on unskilled native wagesis small, at around 0.3 percent, depending onthe assumed elasticity of substitution betweenmigrant and native workers.33 The greaterimpact is felt by existing, unskilled migrants,whose wages decline by more than 10 per-cent.34 At least two factors mitigate thatdecline. First, labor markets are not com-pletely segmented, so that part of the adjust-ment falls on native workers. Second, othergeneral equilibrium effects are at work, suchas a relative shift in the demand for unskilledworkers as the price of capital (combined withskilled labor) rises and a relative shift in de-mand toward goods that use unskilled laborintensively, raising the relative demand for un-skilled workers.

The impact of the shock on the wages ofskilled workers is greater than for unskilled.Wages decline by 1.1 percent on average forskilled natives, significantly more than forunskilled natives. Old skilled migrants suffer awage decline of 20 percent, which is doublethat of old unskilled migrants. The impact onskilled workers is larger than for unskilledbecause skilled workers are assumed to be nearcomplements with capital; with capital in-creasing only slightly, this would tend to drivedown skilled wages. And the impact is largeston old skilled migrants because the rise inmigration of skilled workers is large relative tothe stock of old skilled migrants, and the newmigrants are assumed to be closer substitutesfor skilled migrants than are native workers.

The greater impact of migration on skilledthan unskilled wages is not at first sight con-sistent with the limited evidence available. Inthose studies that find any significant impactof migration on native wages, the largest im-pact tends to be on unskilled wages (seeabove). Our seemingly contrary result arisesfor three reasons. First, unskilled immigrationto high-income countries has been much

larger than skilled migration, so that the wageimpact of unskilled migration is easier to de-tect in empirical work. Second, the shockmodeled here represents a one-time increase inskilled migration that is larger than the exist-ing stock, which has built up over time. Andthird, the model assumes little change in thecapital stock, while increased investment in re-sponse to migration would dampen the fall inskilled wages, a point to which we return inthe conclusion to this chapter.

The impact in developing countries isnearly the reverse. Capital returns suffer andlabor returns improve, with larger improve-ments for skilled workers than for unskilledworkers. The magnitudes differ because therelative size of the shock differs. For example,the decline in unskilled workers in developingcountries is only 0.3 percent, versus 1.7 per-cent for skilled workers.

Assuming that all capital income accruesto native households, native households inhigh-income countries are on aggregate betteroff after the shock, with real incomes increas-ing by 0.4 percent. That is, the increase incapital income more than offsets the loss inwage income. Part of the old migrants’ 6 per-cent decline in real income is due to the as-sumption that they own no capital, so enjoyno nonwage income.35 An alternative, ex-treme assumption is that on a per capita basis,old migrants receive the same amount of non-wage income as natives. This alternativewould reduce old migrants’ loss to 3.4 percentof base real income.

To summarize, the new migrants areclearly the large winners, particularly in per-centage terms. Under the assumptions of themodel, existing migrants are likely to belosers—though the extent of their loss willdepend on their degree of substitutabil-ity with native workers and their share ofnonwage income. Native households in bothhigh-income and developing countries arebetter off. The sources of their gains, though,are very different (figure 2.6). In the high-income countries the gains are generated byhigher returns to capital—somewhat offset by


44

lower wages. The gains for natives in high-income countries would be lower if we as-sumed a more even distribution of capital (to-ward migrants) and a greater degree of laborsubstitutability. In developing countries, thegains to natives essentially are generated byhigher wage income and higher remittances—somewhat offset by lower returns to capital.These gains would be lower should thepropensity of the new migrants to remit belower than average.

If migrants are viewed the same asnatives, then increased migrationreduces natives’ wagesThe degree of labor-market differentiationplays a critical role in determining the effect ofthe increase in migration on native and for-eign households. An alternative scenario—maintaining the same increase in migration—assumes that employers are perfectlyindifferent to hiring native workers versus for-eign-born workers.36 This empirical issue is

linked to real-world dynamics since, over thelong run, differences in labor characteristicscould fade as migrants adjust to their newenvironment and as employers cease to seethem as different.37

The impacts of the alternative scenario onfactor returns are shown in figure 2.7. Themost notable impact is that native wages (forskilled and unskilled workers) decline by morewhen natives and migrants are viewed as per-fect substitutes for each other, while the wagesof the foreign-born decline by significantlyless.38 For skilled workers, the average declinebecomes negligible; the burden of adjustmentis spread out more evenly between native andforeign-born workers. For the given shock,and depending on the assumed elasticity ofsubstitution between foreign and native work-ers, the impact on native wages ranges froma slight increase to a decline of 1 percent(box 2.5).

Because increasing migration constitutes aclear labor-supply shock, one would expect it


45

Figure 2.6 Source of gains for native workers

Change in real income in 2025, $ billions Change in real income in 2025, $ billions

Native workers in high-income regions

100

50

0

�200

�150

�100

�50

150

200

250

300

350

150

0

�100

�50

50

100

200


Capita

l

Labo

r

Public

Remitta

nces

Tota

l

Native workers in developing regions

Capita

l

Labo

r

Public

Remitta

nces

Tota

l

to affect wages (or employment, where move-ment in wages is constrained). But it is impor-tant to view these changes in dynamic terms.First, assuming differentiation between nativeand foreign-born workers, the impact on na-tive workers’ wages in high-income countriesis slight even in absolute terms (–0.04 percentfor unskilled workers and –0.4 percent forskilled workers). More important, in dynamicterms, these changes alter the rate of growthof wages over the next two decades onlyslightly. In the base case, nominal wages willincrease by 3.6 and 4.7 percent, respectively(average annual growth between 2001 and2025), for unskilled and skilled workers inhigh-income countries. With an increase in mi-gration, the growth rate is unchanged for un-skilled workers and drops to 4.6 percent forskilled workers. Even in the worst-case sce-nario for native workers, where foreign-bornworkers are assumed to be perfect substitutesfor natives, the dynamic trends are almost ex-actly the same as in the baseline scenario.

The effects of the modeled shock are obvi-ously larger for existing migrants. With labordifferentiation, their long-term wage growthtrend drops to 3.1 percent for unskilled work-ers and 3.9 percent for skilled workers fromtheir baseline trend of 3.6 and 4.9 percent—

still positive but significantly lower. In themore optimistic scenario for existing mi-grants, where native workers bear a largerpart of the burden, the trend growth in wagesis virtually identical to the baseline, decliningto 3.4 and 4.8 percent, respectively, for un-skilled and skilled workers.

The assumption of perfect substitutabilityof native and migrant labor has a small impacton the global income gain ($379 billion insteadof $356 billion), but significant distributionaleffects. First, in high-income countries perfectsubstitution implies a more pronounced pro-capital bias, as native labor suffers a largerloss. However, the negative impact on existingmigrant households is much smaller: less than1 percent, compared with the 6 percent suf-fered with labor differentiation. This has apositive impact for developing countries, be-cause their loss in remittances from existingmigrants drops dramatically. Second, with per-fect substitution the new migrants benefit froma larger wage differential and thus a higher in-come gain. And again, for developing coun-tries, this translates into higher remittances.Overall, the change in real net remittances is$129 billion under perfect substitutability, asopposed to $88 billion. The bottom-line is thatthe real income of new migrants increases by250 percent, as opposed to 200 percent, andthe gains for natives in developing countriesrise to 1.2 percent, instead of 0.9 percent.

Several other parameters affect the relativeimpact of increased migration on wages andreturns to capital. For example, as shown inbox 2.5, the relative impact will depend on thesubstitution between capital and labor. Thelesser the degree of substitution (or the lessflexible the economy), the greater will be thenegative impact on wages. Most econometricevidence suggests a capital–labor substitutionelasticity of around 1, somewhat higher thanthat used in the model.39 Another crucial as-sumption in the standard model is that skilledworkers and capital are near complements. Analternative would be to assume that unskilledand skilled workers were both substituteswith capital. This would moderate the decline


46

% change relative to baseline


Capital

Perfectsubstitution

Standard

Figure 2.7 Factor returns and migrationin high-income countries, 2005

Unskillednative

workers

Unskilledforeign-

bornworkers

Skillednative

workers

Skilledforeign-

bornworkers

�25

�20

0

�10

5

�5

�15


47

Box 2.5 Increased migration and its impact on wages

and native workers are perfect substitutes, thenwages of native workers would decline by 1 percent,that is, by the same amount as the aggregate wage.

The figure shows the impact on wages—of bothforeign and native workers—using different as-sumptions about their substitutability. The shock isa 50 percent increase in the stock of foreign work-ers with the same assumptions used for the numer-ical example in the table. The impacts on wagesconverge only at very high levels of substitutability.In actual econometric work, it might be hard to es-

The impact of increased migration on wages canbe summarized by a few simple formulas. The

key parameters are the share of foreign workers inthe economy, the capital-labor elasticity of substitu-tion, and the substitution between native and foreignworkers. These relations are summarized in the tablebelow.a The values refer to an economy where for-eign labor is 5 percent of the labor force (sf

l). Thecapital-labor substitution elasticity (�v) is 0.9; andthe substitution (�l) between native and foreignworkers is 4.b The point elasticity is given in thethird column of the table. The estimated impact of a50 percent increase in the stock of foreign workers issimply the point elasticity multiplied by 50. Theactual impact comes from a calibrated numericalmodel. The two are relatively close, despite the sizeof the shock. This is because the results are largelydriven by the low share of foreign workers in theeconomy; therefore most of the economy stays nearits initial equilibrium. The aggregate wage falls byonly 1 percent. This is allocated across the differentworkers depending on their substitutability. With anelasticity of 4 (used in the default scenario), migrantworkers see a 10 percent decline in wages; domesticworkers see only a 0.5 percent decline. If migrants

Summary of elasticities with respect to an increase in foreign workers

Point Estimated ActualExpression Description elasticity impact impact

�W, F � � Aggregate wage �0.0222 �1.1 �1.0

�FW, F � � � Wage of foreign workers �0.2597 �13.0 �10.1

�NW, F � � Wage of native workers �0.0097 �0.5 �0.5slf sk

�v

slf

� l

slf sk

�v

sln

� l

slf sk

�v

timate the substitution elasticity with great pre-cision, particularly since the shocks are unlikelyto be as considerable as those modeled here.

aThese relations are for a small single-sector closed econ-omy but line up relatively well with the impacts from theglobal model. Because it is closed and single-sector, the rela-tions may not hold exactly because of other general equilib-rium effects. See van der Mensbrugghe 2005b.

bThe other parameters are the capital share (sk), the shareof foreign labor in output (sf � sl.sf

l), and the share of nativeworkers in the labor force (sn

l).

% change in wages

Source: World Bank staff calculation.

Substitution elasticity between native and foreign workers

Wage impact from a 50 percent increase inforeign workers, under various substitutionassumptions

�35

�25

�30

�20

0

�10

5

�5

�15

1.5 2.5 3 3.5 4 8 16 32 64

Foreignworkers’wages

Nativeworkers’wages


48

of skilled wages and the rise in the returns tocapital, essentially because there would bemore (assumed) flexibility in the economy.When capital and skilled workers are comple-ments, a sharp increase in skilled workers,without a concomitant increase in capital,raises the scarcity value of capital. If we makecapital and skilled labor more substitutable,the decline in the wages of skilled workers willcreate more demand for them and dampen thenegative impact on their wages.

Caveats—what the model leaves out

While the scenarios discussed here providea wealth of insights, they do not address

many important aspects of the impact ofmigration. Using side calculations, it is possibleto get a sense of two such aspects.

The model does not account for changesin migrant characteristics over timeThe model assumes that migrant characteris-tics do not change over time. This is useful inhighlighting the immediate impact of migra-tion, but less realistic over the medium term.As migrants remain in the destination coun-try, they tend to take on the characteristics ofnative workers. For example, they learn orimprove their fluency in the language, theybetter understand (and may tend to adopt)the social mores of the destination country,and they may become more educated. Em-ployers are likely to view a migrant that haslived in the country for 20 years as beingmore similar to a native worker than a mi-grant who arrived yesterday. This issue hasseveral implications for the calculation of thegains from migration. Migrants who havespent a longer time in the destination countrywill be less perfect substitutes for new mi-grants, mitigating the drop in their wagespredicted by the model. Similarly, as thedegree of labor differentiation declines, theimpact on native wages will rise, thus reduc-

ing native households’ gains. On the otherhand, increased productivity as migrants im-prove their education may generate largergains to owners of capital and could benefitnative workers through spillover effects suchas training. Remittances may decline as mi-grants become more removed from the origincountry.

The process of catch-up in productivity isnot captured in our current model, but wehave done some side calculations to see howthe results could be affected. The catch-up rateand workers’ length of stay are two factorsthat affect catch-up. How long does it take theaverage migrant to achieve the level of pro-ductivity of native workers? Borjas (2003b)provides mixed evidence on this point formigrants in the United States. First, he showsthat the catch-up rate depends very much onwhen the migrants arrived, with earlier mi-grants doing better than later migrants. Sec-ond, he finds no absolute convergence, withmigrants’ wages remaining below those of na-tive workers.

The second factor relates to workers’length of stay. The longer workers stay, thebetter placed they are to improve their skillsand adapt to local work practices, includinglanguage skills (if necessary). At one extreme,all migrants may be assumed to be temporaryworkers staying for a short period to returnhome permanently. Or they may be assumedto be permanent workers arriving young andwith high educational attainment or acquiredskills.

Under the most optimistic scenario, wherecatch-up occurs within a year, our simple,calibrated model predicts gains in the outputof high-income countries that are about25 percent higher than in the case of no catch-up.40 Under a more plausible scenario, wherethe process of catch-up takes 10 years andannual attrition41 is around 10 percent, theoutput gain in high-income countries is about12 percent higher than with no catch-up. Itappears, therefore, that the catch-up phenom-enon could boost the gains from migrationsubstantially.


49

The model does not account for thepotential of migration to spur higherinvestmentThe model generates only a modest rise in thecapital stock as a result of the increase in mi-gration. Increased returns to capital, and thusinterest rates, do increase savings. However,this effect is marginal, in keeping with empiri-cal estimates of the responsiveness of savingsto changes in interest rates. In reality, higherinvestment in response to the higher returns tocapital may be financed by capital inflows,which do not change in the model. There issome evidence, however, that large immigra-tion can attract capital flows. For example,Davis and Weinstein (2002) argue that skilledlabor, unskilled labor, and capital are all at-tracted to the United States, owing to U.S.technological superiority. The mass migrationfrom Europe to the new world before WorldWar I encouraged large inflows of capital.

Higher investment would lessen the declinein wages suffered by skilled workers, dampenthe rise in the return to capital, and increasethe demand for unskilled workers in the high-income countries—but it could have the oppo-site effects in developing countries. To verifythis intuition, we simulated the same migra-tion shock and added to the shock a 0.4 per-cent decline in the level of investment indeveloping countries, with a concomitanttransfer of these resources to high-incomecountries.42 Those assumptions indeed have apositive impact for the high-income countriesand dampen the capital-income gains andlabor-income losses. Overall, the gain fornative households in high-income countriesimproves by 4.5 percent (from $139 billion to$145 billion). But it comes at the expense ofnatives in developing countries, whose incomegain drops from $143 billion to $125 billion,a drop of 12.5 percent. The global gains fall to$345 billion, a 3 percent fall. This suggests—in the absence of an increase in savings—thatthe potential reallocation of global savingstoward high-income countries is negative atthe global level and that capital is moreproductive in developing countries.

Other factors not covered by the modelare more difficult to quantifyThere are various costs associated with migra-tion that the model does not take intoaccount. One issue concerns adjustment costs,as changes in the technology of productionand in the mix of goods imply transitional un-employment and changes in the pattern ofinvestment. The magnitude of these costs de-pends in part on the structure of labor-marketinstitutions (such as constraints on hiring andfiring and minimum-wage legislation) and onthe efficiency of capital markets. Countrieswith more flexible labor markets and soundbanking, stock, and bond markets are likely toexperience lower adjustment costs, underlin-ing the importance of the investment climatefor realizing the potential gains from migra-tion. The size of adjustment costs will alsodepend on whether migration is concentratedor spread over time. This point also has policyimplications. If a country anticipates needingmigrants in the future or recognizes that mi-gration pressures are bound to rise due to de-mographic changes, it would be better toloosen constraints on migration earlier andmore gradually than to be confronted with asharp rise later on. Migration also involvesdirect costs, including transportation andtransitional expenses, as well as the noneco-nomic costs suffered by migrants separatedfrom their families (for example, the impacton children raised without one or both par-ents—see chapter 3). In general these are ei-ther short-term costs that should not greatlychange the calculation of benefits from per-manent migration or problems that declineover time as migrants and families adjust topermanent changes or take steps to reunite.

Several other issues that may affect thegains from migration are impossible to quan-tify. First, our model does not distinguish be-tween irregular and regular migrants. If mi-grants are irregular, they may be paid lowerwages (see chapter 3), which would reducetheir welfare gains (and remittances) relativeto the model results, while it increases thegains of natives. However, irregular migrants

also may impose costs on destinationcountries—among them the costs of enforce-ment (as governments seek to limit what somemay view as undesirable changes in the coun-try’s culture and demographic characteristics);a possible burden on public spending, whichmay be higher for irregular migrants (see box2.3); and the potential for other forms ofillegality generated by the presence of a large,undocumented group of foreign workers.

Second, immigrants may improve the effi-ciency of employment from the perspective offirms by providing a source of labor that caneasily be employed in new geographic loca-tions, and hired or fired in response to changesin cyclical conditions. Piore (1986) describeshow many migrants (at least initially) tend toview their stay as temporary, filling jobs withlower salaries and less stability than those ofnatives.43 Large numbers of immigrants workin construction, which facilitates new develop-ments in areas that require a mobile laborforce. However, over time, migrants will be-come more permanent and demand jobs simi-lar to those held by natives.

Third, our model does not reflect the socialor economic implications of increasing diver-sity in the destination country. The social im-pact lies outside our present scope, but diver-sity has potential economic costs and benefitsthat should be considered. Some writers arguethat increased diversity has an economicvalue. Glasser, Kolko, and Saiz (2001) empha-size the role of a rich variety of services andconsumer goods in enhancing the attractive-ness of cities. Florida (2002) relates an indexof diversity to a concentration on high-tech-nology industries. Ottaviano and Peri (2004)find that cultural diversity has a net positiveeffect on the productivity of U.S.-born citi-zens. By contrast, Schiff (1998) uses a theoret-ical model to underline how a society’s sharedvalues can reduce the cost of transacting busi-ness, owing to higher trust and easier enforce-ability of sanctions. Thus immigration, whichincreases diversity, may lower productivity byraising transaction costs. Finally, Alesina andFerrara (2004) find that increases in ethnic

diversity are associated with lower growthrates, holding all else equal. However, diversitymay be more beneficial to growth at higher in-come levels.44 Clearly much will depend onthe kinds of diversity involved: immigrantswho rely on national affinities to cement loy-alty to violent gangs presumably have a verydifferent impact on growth and welfare thanimmigrants who open ethnic restaurants.

Fourth, the model may not fully capturethe beneficial effect of immigration on in-creasing the supply of labor in the service sec-tor. Although reductions in the prices of ser-vices are captured, the resulting expansion inthe supply of native labor (as more parentscan afford child care and workers have moretime to devote to their jobs) is not.

Fifth, the model does not reflect the possi-bility that skilled migration may lowergrowth in origin countries, for example, be-cause of positive externalities from the pres-ence of skilled workers or increases in theprice of services that require technical skills(see chapter 3).

Finally, the model assumes constant re-turns to scale, while immigration may bemore beneficial if significant sectors enjoyincreasing returns to scale. Increasing returnsmay be derived, for example, from fixed pro-duction costs, network effects (the unit priceof providing telephone service falls as the cus-tomer base grows), reduced transport andcommunications costs (as the local market ex-pands), or increased productivity due to inter-actions among highly skilled workers. In theirrole as consumers and workers, immigrantsmay facilitate an expansion of the market,thereby raising productivity by increasing re-turns. On the other hand, large inflows of im-migrants may induce congestion, strainingpublic transportation systems, for example, orbidding up the price of land. Such effects areparticular to the sector and geographic areainvolved, so it is difficult to draw broad con-clusions. However, skilled immigrants havemade significant contributions to high-tech-nology sectors that are subject to increasingreturns to scale.


50

These qualifications to the scenario resultsillustrate how model exercises must abstractfrom reality to provide quantitative measuresof the impact of migration. Some of the is-sues that the model does not consider wouldlikely be small in the medium term (adjust-ment costs, transportation costs) or wouldtend to increase the economic benefits of mi-gration (improved productivity of migrantsover time, greater labor-market flexibility andsupply of labor). Other issues would increasebenefits to destination countries, while poten-tially harming origin countries (higher invest-ment, economies of scale). Still others mayhave both economic and social effects, with-out lending themselves to determinations oftheir direction and size (diversity, irregularmigration).

Notes1. In keeping with the overall thrust of this report,

we focus here on South-North migration, although it isimportant to recognize that a large portion of migrantsfrom developing countries move to other developingcountries.

2. Some readers may also find the chapter too tech-nical, as it necessarily deals with detailed specificationissues—for example, the degree of differentiation be-tween native and migrant workers, the fiscal impact ofmigrants, and how to take into account the change inprices between developed and developing countrieswhen evaluating gains to migrants.

3. Data on the stocks of migrants are generallytaken from census reports in countries of destination andthus include both regular and irregular migrants. How-ever, irregular migrants tend to be less likely to reporttheir immigrant status, so the estimate of total mi-grants is probably low.

4. The breakup of the Soviet Union and emergenceof 15 new independent countries in 1991 created newpopulations of “international” migrants without mi-gration having taken place.

5. The exceptions were Belgium, France, Ireland,Portugal, and the United Kingdom.

6. In their regression equation explaining immigra-tion to the United States, Hatton and Williamson (2002)calculate that IRCA doubled the Mexican immigrationrate from 1989 to 1991.

7. Germany, Ireland, and the Czech Republic are inthe process of establishing new immigration regimes,with a major focus on economic migration. The EU is

also discussing the Green Paper on an EU Approach toManaging Economic Migration (EU 2005).

8. These numbers will be moderated to the extentthat labor-force participation rates in the 65� cohortare positive, if small. Moreover, labor-force participa-tion rates for the elderly are likely to increase as pen-sions and benefits stagnate or decline with fiscal pres-sures and as life expectancy rates continue to increase.We may also witness an increase in labor-force partici-pation rates among people of working age.

9. These global gains are comparable to the recentfindings by Walmsley and Winters (2003), when ad-justed for the size of the economy in 2001 relative tothe projected size of 2025.

10. The model’s specification is described in vander Mensbrugghe (2005a).

11. “Migrants” refers to migrants from developingcountries unless otherwise stated.

12. The phase-in period is somewhat arbitrary.Because of its 10-year implementation, it minimizesadjustment costs to some extent. The five-year periodbetween 2020 and 2025 enables an assessment of long-run steady-state impacts.

13. This is by design. An alternative would be toincrease the stock of migrants in proportion to theircurrent structure—by host region and skill level. In thiscase, the largest proportional increase would be for un-skilled workers in the United States.

14. A switch of 14 million workers from develop-ing to high-income countries has only a small impact(a decline of 0.4 percent) on aggregate employment indeveloping countries, albeit with potentially greaterconsequences among the relatively more scarce skilledworkers.

15. Many factors determine the level of remittances.For example, new migrants may leave many dependentsin their home countries, which would tend to raise re-mittances. On the other hand, at least in the short-term,moving and start-up costs could lower remittances.

16. For simplicity, migrants from other high-income countries are added to the true natives.

17. Again, for simplicity, all migrants in developingcountries—both from rich and developing countries—are lumped together for the purposes of the aggregateanalysis.

18. The impact on households other than the“new” migrants is not affected by cost-of-living adjust-ments. Since these households do not move, they face thesame system of prices, and thus their change in realincome simply depends on the standard real incomemeasure.

19. This assumes a perhaps implausible Leontiefutility function but the purpose is simply to illustratethe point that corrections need to be made for differ-ences in the cost of living.


51

20. Were the true prices available, one could do astandard equivalent variation calculation that wouldtake into consideration the change in prices.

21. In the high-income countries, new migrants’total real consumption is $562 billion, compared with$80 billion in the baseline, hence the real increase of$481 billion. When the $562 billion is adjusted for thedifference in the cost of living, real consumption, asperceived from the point of view of the new migrant, isonly $244 billion, taking the change in real incomedown to $162 billion. The cost-of-living adjustmentaverages 2.3, lower than the 3.1 GDP-weighted aver-age PPP of developing countries. This occurs becausemiddle-income countries (with a relatively low PPP ad-justment) have a higher weight in migration than in de-veloping countries’ GDP.

22. High-income countries, on the other hand, seea substantial rise in exports, $211 billion (2.2 percent),and a more modest $113 billion rise in net imports(1.2 percent), with imports from developing countriesdeclining by $23 billion. This implies that although alarge part of the increase in high-income exports can beattributed to the increase in remittances, a significantportion is also coming through intraregional tradeamong high-income countries driven in part by chang-ing comparative advantage.

23. A standard “Dutch disease” effect of foreigninflows.

24. Studies of the impact of trade reform indeveloping and industrial countries tend to show thatwages in developing countries rise relatively more—particularly for unskilled workers—than in industrialcountries, but those changes are relatively minor com-pared to the initial gap in wages. For example, unre-ported results from Anderson, Martin, and van derMensbrugghe (2005) show that full merchandise tradereform would increase unskilled real wages in develop-ing countries by 3.7 percent (unweighted average), butby only 0.7 percent in industrial countries. This couldinduce a small reduction in the incentive to migrate,but it would not substantially alter the significant wagemultiple of 4 to 5 (taking into account cost-of-livingdifferentials).

25. The cost-of-living adjustment for the new mi-grants treats their consumption of public goods andservices the same as their private consumption—that is,it is adjusted by the same PPP factor.

26. The assumption is that the new migrant house-holds come with the dependency ratio of their homecountry.

27. There will be compositional impacts in trans-lating gains from 2025 to 2001, since developing coun-tries are growing on average more rapidly than high-income countries.

28. Results presented at the eighth annual confer-ence on Global Economic Analysis held in Lübeck,Germany. See http://www.gtap.agecon.purdue.edu/events/conferences/2005/program_day3.htm.

29. The full merchandise trade reform scenario iswith a standard model and ignores any beneficial im-pact through higher trade-induced productivity orscale economy effects. Note that the gains from reformof services trade could be multiples of merchandisetrade reform. See Anderson, Martin, and van derMensbrugghe (2005).

30. In the case of high-income countries, ‘similar’workers would be migrant workers from other high-income countries. Other migrant workers are bundledtogether in a so-called ‘foreign-born’ aggregate.

31. One would expect the elasticity to increase asthe proportion of migrants in the population increases(box 2.5).

32. See, for example, Bchir and others (2002).33. Box 2.5 shows how wages—native and

foreign—are related to an increase in the stock of mi-grants. Two parameters are crucial—the substitutionbetween native and foreign workers and the share offoreign workers in the labor force.

34. The general equilibrium elasticity is only 0.27for unskilled workers; for roughly a 40 percent increasein supply, wages decline by around 10 percent.

35. Migrants from other high-income countries(also assumed to have no nonwage income) see only asmall change in their real incomes, as their wages areclosely linked to the wages of native workers.

36. Observed wage differentials can arise from acombination of two effects—differences in productiv-ity and differentiated labor demand. If labor isperfectly substitutable, then the equilibrating condi-tion is the equality of efficiency wages, that is,productivity-adjusted nominal wages. If labor is differ-entiated, efficiency wages are no longer necessarilyequalized, and the equilibrium wage will be determinedby supply and demand conditions for the differentiatedlabor.

37. The empirical evidence on “catch-up” is lim-ited. In the case of migrants to the United States, Bor-jas (2003b) shows that migrants who arrived in the1960s almost caught up with natives within a 10–15year period. Those who arrived in the 1970s made lessprogress in closing the gap with natives. However, thewages of migrants arriving in the 1980s actually fellfurther behind those of natives after a 10-year period.The scenarios described in this chapter assume nochange in the relative productivity of migrants. Such anassumption would require a more elaborate specifica-tion of migrants to capture their changing compositionover time, similar to modeling capital vintages. By


52

ignoring the catch-up process, our results may under-estimate the longer-term gains from migration.

38. The changes in wages by region of origin areidentical for all workers in each high-income region,but due to aggregation effects, this will not necessarilybe true when averaging across regions.

39. The model has a vintage structure with a lowersubstitution elasticity for “old” or installed capital anda higher substitution elasticity for “new” capital. Theactual substitution will be a weighted average of theold and the new vintages, with a higher average forcountries with relatively high rates of investment.

40. This follows directly from the assumption thatthe productivity level of migrants is initially 75 percentthat of natives.

41. The attrition rate will be a combination offactors—return migration, retirement, and death. Thefirst factor is probably most important the first year,whereas the other two factors will depend on the ageof the migrant.

42. The value of 0.4 percent was chosen because itcorresponds to the change in the number of workers indeveloping countries—though it should be noted thatthe change in workers represents a change in the stocklevel, whereas the change in investment is a change inflows.

43. He notes that this trend may be changing, astechnology and globalization encourages smaller-scaleproduction and more permanent immigration.

44. This may occur because “the productivity ben-efits of skill complementarities are realized only whenthe production process is sufficiently diversified,” orbecause high-income economies are able to develop in-stitutions that help them cope better with the potentialfor conflict inherent in ethnic diversity (Alesina andFerrara 2004).

ReferencesAlesina, Alberto, and Eliana La Ferrara. 2004. “Ethnic

Diversity and Economic Performance.” CentroStudi Luca D’Agliano Development StudiesWorking Papers 193. Milan and Turin. December.

Anderson, Kym, Will Martin, and Dominique van derMensbrugghe. 2005. “Market and Welfare Impli-cations of Doha Reform Scenarios.” In Agricul-tural Trade Reform and the Doha DevelopmentAgenda, ed. Kym Anderson and Will Martin.New York: Palgrave Macmillan.

Angrist, Joshua, and Adriana Kugler. 2002. “Protectiveor Counter-Productive? European Labour MarketInstitutions and the Effect of Immigrants on EUNatives.” Centre for Economic Policy ResearchDiscussion Paper 3196. London.

Auerbach, Alan J., and Philip Oreopoulos. 1999. “An-alyzing the Fiscal Impact of U.S. Immigration.”American Economic Review Papers and Proceed-ings 89(2): 176–80.

Bchir, Mohamed Hedi, Yvan Decreux, Jean-LouisGuérin, and Sébastien Jean. 2002. “MIRAGE, aComputable General Equilibrium Model forTrade Policy Analysis.” CEPII Working Paper2002-17. Paris. December.

Bonin, Holger, Bernd Raffelhuschen, and Jan Walliser.2000. “Can Immigration Alleviate the Demo-graphic Burden?” Finanz Archiv 57(1): 1–21.

Borjas, George J. 1994. “The Economics of Immi-gration”, Journal of Economic Literature 32December (1994): 1667–1717.

———. 2003a. “The Labor Demand Curve IS Down-ward Sloping: Reexamining the Impact of Immi-gration on the Labor Market.” Quarterly Journalof Economics. November.

———. 2003b. “The Economic Integration ofImmigrants in the United States: Lessons for Pol-icy.” WIDER Discussion Paper DP2003/78.World Institute for Development EconomicsResearch, United Nations University, Helskinki.December.

Borjas, George, Richard B. Freeman, and LawrenceKatz. 1997. “How Much Do Immigration andTrade Affect Labor Market Outcomes?” Brook-ings Papers on Economic Activity 1: 1–90.

Butcher, Kristin F. and David Card. 1991. “Immigra-tion and Wages—Evidence from the 1980s.”American Economic Review 81(2): 292–6.

Card, David. 1989. “The Impact of the Mariel BoatLift on the Miami Labor Market.” NationalBureau of Economic Research Working Paper3069. Cambridge, MA.

———. 2001. “Immigrant Inflows, Native Outflows,and the Local Labor Market Impacts of HigherImmigration.” Journal of Labor Economics 19:22–64.

Carrington, W., and P. de Lima. 1996. “The Impact of1970s Repatriates from Africa on the PortugueseLabor Market.” Industrial and Labor RelationsReview 49(2): 330–47.

Collado, M. Dolores, Inigo Iturbe-Ormaetxe, andGuadalupe Valera. 2004. “Quantifying the Im-pact of Immigration on the Spanish WelfareState.” International Tax and Public Finance 11:335–53.

Coppel, Jonathan, Jean-Christophe Dumont, andIgnazio Visco. 2001. “Trends in Immigration andEconomic Consequences.” Economics Depart-ment Working Paper 284. Organisation for Eco-nomic Co-operation and Development, Paris.


53

Davis, Donald R., and David E. Weinstein. 2002.“Technological Superiority and the Losses fromMigration.” National Bureau of EconomicResearch Working Paper 8971. Cambridge, MA.

DeNew, John P., and Klaus F. Zimmermann. 1994.Native Wage Impacts of Foreign Labor. Journal ofPopulation Economics 7: 177–92.

Dustmann, Christian, and Albrecht Glitz. 2005. “Im-migration, Jobs and Wages: Theory, Evidence andOpinion.” Centre for Research and Analysis ofMigration, Department of Economics, UniversityCollege London.

Dustmann, Christian, Francesca Fabbri, Ian Preston,and Jonathan Wadsworth. 2003. “The LocalLabour Market Effects of Immigration in theUK.” Home Office Online Report 06/03.London.

EU (European Union). 2005. “Green Paper on an EUApproach to Managing Economic Migration.”COM(2004)811 final, Commission of the Euro-pean Communities, Brussels. November.

Fehr, Hans, Sabine Jokisch, and Laurence Kotlikoff.2004. “The Role of Immigration in Dealing withthe Developed World’s Demographic Transition.”National Bureau of Economic Research WorkingPaper 10512. Cambridge, MA.

Florida, Richard. 2002. “Bohemia and Economic Ge-ography.” Journal of Economic Geography 2:55–71.

Glasser, Edward L., Jed Kolko, and Albert Saiz. 2001.“Consumer City.” Journal of Economic Geogra-phy 1: 27–50.

Gott, Ceri, and Karl Johnston. 2002. “The MigrantPopulation in the UK: Fiscal Effects.” RDS Occa-sional Paper 77. Home Office, London.

Gross, Dominique. 1999. “Three Million Foreigners,Three Million Unemployed: Immigration andthe French Labor Market.” IMF Staff WorkingPaper 99/124. International Monetary Fund,Washington, DC.

Gustafsson, B., and T. Osterberg. 2001. “Immigrantsand the Public Sector Budget: Accounting Exer-cises for Sweden.” Journal of Population Eco-nomics 14(4): 689–708.

Gustmann, Alan L., and Thomas L. Steinmeier. 1998.“Social Security Benefits of Immigrants and U.S.Born.” National Bureau of Economic ResearchWorking Paper 6478. Cambridge, MA.

Hamilton, Bob, and John Whalley. 1984. “Efficiencyand Distributional Implications of Global Restric-tions on Labour Mobility.” Journal of Develop-ment Economics 14: 61–75.

Hunt, Jennifer. 1992. “The Impact of the 1962 Repa-triates from Algeria on the French Labor

Market.” Industrial and Labor Relations Review45: 556–72. April.

IOM (International Organization for Migration).2005. World Migration 2005. Geneva.

Jaeger, David A. 1996. “Local Labor Markets, Admis-sion Categories, and Immigrant LocationChoice.” Unpublished paper. Hunter College,New York. June.

LaLonde, Robert J., and Robert H. Topel. 1997. “Eco-nomic Impact of International Migration and theEconomic Performance of Migrants.” In Hand-book of Population and Family Economics,ed. Mark Rosenzweig and Oded Stark. Amster-dam: North-Holland.

Lee, Ronald, and Timothy Miller. 2000. “Immigration,Social Security, and Broader Fiscal Impacts.”American Economic Review Papers and Proceed-ings 90(2): 350–54.

Longhi, Simonetta, Peter Nijkamp, and Jacques Poot.2004. “A Meta-Analytic Assessment of the Effectof Immigration on Wages.” Population StudiesCentre Discussion Papers 47. University ofWaikato, Hamilton, New Zealand. December.

Massey, Douglas. 2000. “Higher Education and SocialMobility in the United States, 1940–1998.” Paperdelivered at Association of American UniversitiesCentennial Meeting, April 17, Washington, DC.

Mora, Jorge, and J. Edward Taylor. 2005. “Determi-nants of Migration, Destination and SectorChoice: Disentangling Individual, Household andCommunity Effects.” In International Migration,Remittances, and Development, ed. CaglarOzden and Maurice Schiff, Washington, DC:World Bank.

Moses, Jonathon W., and Bjørn Letnes. 2004. “TheEconomic Costs to International Labor Restric-tions: Revisiting the Empirical Discussion.”World Development 32(10): 1609–26.

Nana, Ganesh, and Julian Williams. 1999. “Fiscal Im-pacts of Migrants to New Zealand.” Report tothe New Zealand Immigration Service, Auckland.

Ottaviano, Gianmarco, and Giovanni Peri. 2004. “TheEconomic Value of Cultural Diversity.” NationalBureau of Economic Research Working Paper10904. Cambridge, MA.

OECD (Organisation for Economic Co-operation andDevelopment). 1997. Trends in InternationalMigration. Paris.

———. 2005. Trends in International Migration. Paris.Parsons, Christopher R., Ronald Skeldon, Terrie L.

Walmsley, and L. Alan Winters. 2005. “Quantify-ing the International Bilateral Movements ofMigrants.” Paper presented at the Eighth AnnualConference on Global Economic Analysis, June9–11, Lübeck, Germany.


54

Passel, Jeffrey S. 2005. “Estimates of the Size andCharacteristics of the Undocumented Popula-tion.” Pew Hispanic Center, Washington, DC.

Piore, Michael J. 1986. “Perspectives on Labor MarketFlexibility.” Industrial Relations 25: 146–66.Spring.

Pischke, Jorn-Steffen, and Johannes Velling. 1994.“Wage and Employment Effects of Immigrationto Germany: An Analysis Based on Local LaborMarkets.” Centre for Economic Policy ResearchDiscussion Paper 935. London. March.

Poot, Jacques, and Bill Cochrane. 2004. “Measuringthe Economic Impact of Immigration: A ScopingPaper.” Immigration Research Programme,New Zealand Immigration Service, Auckland.December.

Rowthorn, Robert. 2004. “The Economic Impact ofImmigration.” Civitas Online Report. Civitas:The Institute for the Study of Civil Society,London.

Schiff, Maurice. 1998. “Trade, Migration and Welfare:The Impact of Social Capital.” In H. Singer, N.Hatti, and R. Tandon (eds.) Globalization, Tech-nology, and Trade in the 21st Century. Vol. 19,New World Order, Delhi: B. R. Publishing.

Smith, James P., and Barry Edmonston. 1997. TheNew Americans: Economic, Demographic andFiscal Effects of Immigration. Washington, DC:National Academy Press.

Sriskandarajah, Dhananjayan, Laurence Cooley, andHoward Reed. 2005. “Paying Their Way: TheFiscal Contribution of Immigrants in the UK.”Institute for Public Policy Research, London.

Storesletten, Kjetil. 2000. “Sustaining Fiscal Policythrough Immigration.” Journal of Political Econ-omy 108(21).

Timmer, Hans, and Dominique van der Mensbrugghe.2005. “Migration, PPP and the Money Metric ofWelfare Gains.” Development Economics Depart-ment, World Bank, Washington, DC.

U.S. Binational Study on Migration. 1997. BinationalStudy: Migration Between Mexico and the UnitedStates. www.utexas.edu/lbj/uscir/binational.html.

U.S. Labor Survey. 2005. Bureau of Labor Statistics.http://bls.gov/cps/home.htm.

van der Mensbrugghe, Dominique. 2005a. “LINKAGETechnical Reference Document: Version 6.0.” Un-published paper. World Bank, Washington, DC.

———. 2005b. “Derivation of Output and Wage Elas-ticities Relative to an Increase in Migrants.”Development Economics Department, WorldBank, Washington, DC.

Walmsley, Terrie L., S. Amer Ahmed, and ChristopherR. Parsons. 2005. “The GMig2 Data Base: AData Base of Bilateral Labor Migration, Wagesand Remittances.” GTAP Research Memoran-dum No. 6. Center for Global Trade Analysis,Purdue University. September.

Walmsley, Terrie Louise, and L. Alan Winters. 2003.“Relaxing the Restrictions on the TemporaryMovements of Natural Persons: A SimulationAnalysis.” Centre for Economic Policy ResearchDiscussion Paper Series 3719. London. January.

Wickramasekera, Piyasiri. 2002. “Asian Labor Migra-tion: Issues and Challenges in an Era of Global-ization.” International Labor Office, Interna-tional Migration Program. Geneva.

Winter-Ebmer, Rudolf, and Josef Zweimuller. 1999.“Do Immigrants Displace Young Native Workers:The Austrian Experience.” Journal of PopulationEconomics 12: 327–40.


55

the potential gains from international migration...from migration.however, such increases in...

Documents