The Place Premium

The Place Premium

Lant Pritchett(paper with Michael Clemens, CGD and Claudio

Montenegro, World Bank)LEP Lunch/Development Seminar

Sept 29, 2008

Outline of the presentation

• Empirical estimates of wages differences for observationally equivalent workers on opposite sides of the US Border

• Addressing the issue of migrant self-selection– Simulation with residuals– Data from Latin America– One true experiment– Comparison with macro growth accounting– Experiences with spatially distinct but open borders

• Comparisons of (adjusted) wage gaps with other “similar” numbers (wage discrimination, etc.)

Drilling down through wage surfaces

ln(w)

X (e.g. education, age)

Bolivian (born, educated) workers in Bolivia

USS (born, educated) workersIn USA

Bolivian (born, educated) workersIn USA

)(w

)(w

Bol

USA0

X

XR

New collection of data sets

• 2,015,411 formal-sector wage-earners in 43 countries

• 42 different countries wage surveys—of wage earners– Wages (converted to monthly, PPP)– Country of birth– Amount and country of schooling– Age/experience– Gender– Rural/urban

Comparison of our wage survey results with labor value added per worker

SLE

ETH

TCD

NGA

UGA

YEM

BGD

NPL

KHMVNM

HTIGHA

CMRPAK

BOL

INDIDN

HND

ECU

LKA

JAMNIC

PHL

GUYMARPRY

PERTHA

EGY

GTM

COL

JOR

VEN

PANBLZ

BRA

DOMTURURYCRIZAF

MEXCHL

ARG

USA

62.

51

252

505

001

000

200

04

000

Avg

. P

PP

wag

e, lo

g sc

ale

62.5 125 250 500 1000 2000 4000PPP Labor income, 0.65 share, log scale

45 deg. line Cubic fit w/o HND

The “formal sector” is a bigIssue for the poor African Countries In the sample

We just drop Honduras

Combine with PUMS US Census

• Wages of individuals, with country of birth and age at arrival plus– Schooling– Age– Sex– Urban/rural residence

Results of the wage surface drilling: foreign born, foreign educated (late arrivers), high school or less educated, 35

year old, males, in urban areas in USA vs home

Ratio of US to country wages

Comparing foreign born, foreign educated in US to in home

R0

Predicted annualized wages (2000 PPP)

In US In Home Absolute gap

Mean 7.3 5.1 $20,764 $5,352 $15,411

Median 6.2 4.1 $19,972 $4,675 $15,438

Selected Countries of Interest

Nigeria (2nd highest) 13.5 14.9 $18,394 $1,238 $17,155

Haiti 23.5 10.3 $17,428 $1,690 $15,738

India 10.9 6.3 $23,024 $3,684 $19,340

Philippines 6.2 3.8 $18,436 $4,820 $13,615

Brazil 5.0 3.8 $23,725 $6,302 $17,423

Mexico 3.8 2.5 $17,650 $6,971 $10,679

Dom Rep. (lowest) 3.3 2.0 $17,897 $8,984 $8,912

0

5

10

15

20

25

30Ye

me

n

Nig

eri

a

Egyp

t

Hai

ti

Cam

bo

dia

Sie

rra

Leo

ne

Gh

ana

Ind

on

esi

a

Pak

ista

n

Ve

ne

zuel

a

Cam

ero

on

Vie

tnam

Ind

ia

Jord

an

Ecu

ado

r

Bo

livia

Sri L

anka

Ne

pal

Ban

glad

esh

Uga

nd

a

Eth

iop

ia

Gu

yan

a

Ph

ilip

pin

es

Pe

ru

Bra

zil

Jam

aica

Ch

ile

Nic

arag

ua

Pan

ama

Uru

guay

Gu

ate

mal

a

Co

lom

bia

Par

agu

ay

Sou

th A

fric

a

Turk

ey

Arg

en

tin

a

Me

xico

Be

lize

Thai

lan

d

Co

sta

Ric

a

Mo

rocc

o

Do

min

ican

Re

p.

Ro

Estimates of R0 (predicted wages of observationally workers across the US border) for 42 countries with

95% confidence intervals

38/42 can reject bigger than 1.5

32/42 cannot reject bigger than 4

All kinds of comparability issues: but the biggest is PPP

• Gross versus net• Inclusion of benefits (in-

kind, entitlements) or not• Valuation of workplace

amenities (e.g. safety regulation)

• But we suspect the biggest is imputation of the location of consumption (in US versus home)

Estimates of R0 at various fractions at PPP versus official exchange rates

100%(base)

80% 40% 0%

Mean 5.1 5.8 8.3 16.4

Median 4.1 4.9 7.2 13.9

• Remittances about 20 percentFor Mexicans• Remittances/savings about 60For Philippines overseas workers• Think “optimal” savings of temporaryworker

How much of the observed wage differentials of observationally equivalent workers represent

border restrictions vs. selection or home preference?

• Six different methods/data for examining wage selection, all of which suggest our predicted mean wage ratios of observationally equivalent workers over-state wage ratios of equal intrinsic productivity workers by between 1 and 1.4.

The question of selection on unobservable

• Our estimates of compare what those who moved to US make versus what those who are observationally equivalent make in home.

• But those who did move might have made more than the o.e. counter-parts so R0 overstates the gain

• We are not talking about the upper end but the low skill end—people making 10$/hour

• Not obvious that there is positive self-selection on unobservable productivity in the home market—theory is that people would maximize the gain from moving if either:

– productivity is a market match phenomena (e.g. having an uncle with a good business), or

– Individually differential obstacles (e.g. family unification visas)

then one might expect zero or negative selection.

0.0

0.2

0.4

0.6

0.8

Kern

el

densi

ty

0 5 10 15Component plus residual from ln(wage) regression

USA born, USA res, USA educ IND born, IND res, IND educ

IND born, USA res, USA educ IND born, USA res, IND educ

IndiaR0 compares means

Could compare to otherpercentile of the home distribution of unobservables,e.g. 70th

1st approach: Wage ratios under various assumptions about where in the home distribution

of unobservables migrants came

50th 70th 90th 95th

Median across countries 4.5 3.4 2.1 1.6

Ratio to assumption of 50th 1.34 2.20 2.85

Selected Countries of Interest

Nigeria 10.34 6.92 4.24 3.49

Haiti 8.76 4.08 1.34 0.86

India 7.05 5.16 3.28 2.6

Philippines 3.77 2.73 1.76 1.44

Brazil 4.23 3.2 2.03 1.6

Mexico 3.32 2.44 1.57 1.24

Dominican Rep. 2.95 2.26 1.71 1.07

Wage ratios of equally productive workers at various assumptions of source of migrants in

distribution of unobservables

0

1

2

3

4

5

6

7

8

9

10

Egyp

tYe

men

Nig

eria

Sier

ra L

eone

Jord

anVe

nezu

ela

Indo

nesi

aPa

kist

anVi

etna

mIn

dia

Nep

alCa

mer

oon

Cam

bodi

aEc

uado

rBa

ngla

desh

Sri L

anka

Gha

naG

uyan

aBo

livia

Jam

aica

Braz

ilCh

ileTu

rkey

Ethi

opia

Uga

nda

Phili

ppin

esPa

nam

aPe

ruN

icar

agua

Colo

mbi

aPa

ragu

ayU

rugu

ayBe

lize

Arge

ntina

Gua

tem

ala

Mex

ico

Cost

a Ri

caSo

uth

Afri

caM

oroc

coH

aiti

Thai

land

Dom

inic

an R

ep.

Re

30th

50th

70th

90th

95th

2nd Approach: Data from the Latin American Migration Project (LAMP)

• Tracks migrants from seven Latin American countries and does surveys in their origin localities of non-migrants

• Wage histories of migrants including last wage before migrating

• Compare wages of migrants before moving and non-migrants, with distribution of residuals

Distribution of the unobserved component on wages (residuals) in home for migrants and non-

migrants: Mexico0.

00.

20.

40.

60.

8K

ern

el d

ensi

ty

0 2 4 6 8 10ln(wage)

Migrant in home Non-migrant in home

Mean migrant at 53rd

Percentile of non-migrants

0.0

0.2

0.4

0.6

0.8

Kern

el

densi

ty

0 5 10 15Component plus residual from ln(wage) regression

USA born, USA res, USA educ MEX born, MEX res, MEX educ

MEX born, USA res, USA educ MEX born, USA res, MEX educ

Mexico

Actual distributionOf residuals for MexicoSo we can compute 50th

Of movers to 53rd of home

Distribution of the unobserved component on wages (residuals) in home for migrants and non-

migrants: Haiti0.

00.

10.

20.

30.

4K

ern

el d

ensi

ty

2 4 6 8 10 12ln(wage)

Migrant in home Non-migrant in home

Mean migrant at 61st

Percentile of non-migrants

0.0

0.2

0.4

0.6

0.8

Kern

el

densi

ty

-5 0 5 10 15Component plus residual from ln(wage) regression

USA born, USA res, USA educ HTI born, HTI res, HTI educ

HTI born, USA res, USA educ HTI born, USA res, HTI educ

Haiti

Typical migrant percentile in distribution of non-migrants' unobserved component of wages

Mean migrant: 53 50 54 58 51 61 69

Median migrant: 49 50 50 50 50 64 62

Ratio of migrant home wage to non-migrant home wage, conditional on observables = exp(βmigrant)

1.07 1.00 1.10 1.19 1.06 1.46 1.42

US wage (our data)

1471 1553 1561 1606 1491 1452 1714

Non-migrant wage (our data)

581 529 443 775 749 141 452

Ro 2.53 2.94 3.52 2.07 1.99 10.31 3.79

Re 2.37 2.93 3.19 1.74 1.89 7.07 2.67

Ro/Re 1.07 1.00 1.10 1.19 1.06 1.46 1.42

México Guatemala Nicaragua Costa Rica Dominican Rep.

Haití Perú

Results from 7 countries

• The medians of the migrant and non-migrants are exactly the same for 5 of the 7—the selection is mostly an upper tail thing

• Using the means to adjust out Ro estimates lowers them by a ratio of between 1 (no adjustment for Guatemala) to 1.46 (Haiti)

• In no country is the typical migrant from as high as the 70th percentile of non-migrants (which, from table above, implies an adjustment of 1.34 using the actual residuals data).

3rd Approach: Comparison with experimental estimates of wage effects

• Movers from Tonga to New Zealand chosen from applicants based on a lottery

• OLS wage ratio: 6.14 (chosen versus all stayers)

• Experimental wage ratio: 4.91 (foreign wages of randomly selected chosen versus home wages of applicants).

• Bias from not correcting for selection: 6.12/4.91=1.25

4th Approach: Comparison to macro growth decomposition (Hall and Jones)

Hall and Jones estimates

Ratio wage based

estimates to macro

accountingR

estimates,

Ratio of USA A and K

to country A and K

Ratio of USA A

to country

A

Median 3.82 3.07 2.44 1.25

Average 5.11 3.69 2.71 1.39

Average without fouroutliers 4.53 3.92 2.90 1.16

5th approach: Use comparisons of average wages of observationally equivalent in home

and foreign (allowing for country specific schooling)

• Doesn’t involve movers at all—so should understate the marginal mover if there is positive selection.

• In fact, these are larger than bilateral estimates• But one has to correct for the quality of

schooling as S in Bolivia is not S in USA• Under various plausible adjustments of S

“evaporation” suggest selection at most increases R0 by factor of 1.2

6th approach: wage ratios in spatially distinct but legally integrated labor

markets: Puerto Rico

1.3 1.4 1.4 1.5 1.6 1.8 1.8

0

1

2

3

4

5

Guam=1.36

When borders were open wage ratios above 2 caused massive mobility, leading to wage

convergence0

12

34

5

1830 1840 1850 1860 1870 1880 1890 1900 1910 1920 1930

Germany Great BritainI reland I talyNorway Sweden

Shall I compare thee to a summer’s day…thou are much bigger

• Wage discrimination—comparing wage discrimination against disfavored social groups within borders to consequence of local of birth/citizenship/market access based wage differentials

• Border differentials in prices of goods or capital

• Impacts of poverty programs

Our average cross-border wage differential (5.1) is larger by a factor of 3 than racial

discrimination in the US in 1939

1.1 1.1 1.3 1.4 1.4 1.4 1.61.9

0

1

2

3

4

5Using our wage data we can estimate the largest discrimination against females in the world, Pakistan, 3.1

Using historical data one can estimate the gap between marginal product (rental price) and subsistence wage for 19th century North American slaves: around 3.8

Estimates of the remaining price gaps across countries

0

10

20

30

40

50

60

Me

an

pe

rce

nta

ge

ab

so

lute

va

lue

d

iffe

ren

ce

in p

rod

uc

er

pri

ce

s a

cro

ss

g

oo

ds Canada-USA

Germany-USAUK-USAJapan-USA

Source: Bradford and Lawrence, 2004

Combination of small price gaps and large wage gaps implies the estimated gains from even minor relaxations in labor mobility are big relative to the

largest gains in remaining trade liberalization

79.5 86

305

Bil

lio

ns

Doubling net ODA

Net gains to developingcountries fromliberalization in Doharound

Value of welfare gains tocurrent developing countryresidents (including gainsto movers)from 3% ofOECD labor force increase

`

Source: Winters et al 2004

Comparing estimated gains from anti-poverty interventions in poor countries to wage differences

Intervention Country

Present-value lifetime income

increment due to

intervention (US$ at PPP)

Annual wage difference of

observationally equivalent

male low skill worker

Weeks of US work equivalent to lifetime NPV of intervention

Microcredit Bangladesh 700 $14,891 2.4Anti-

sweatshop Indonesia 2,700 $17,478 8.0Additional year

of schooling Bolivia 2,250 $15,455 8.0

Deworming Kenya 71 $16,265 0.2

Conclusion

• Massive gaps in wages between observationally equivalent workers in 42 poor countries—average 5.1, median 4.1--$15,000 per year (PPP)

• The bulk of the evidence suggests that the self-selection might cause these to overstate gains from movement of unskilled workers by a modest amount (scale back by between 1 and 1.4)

• These make the wage differentials across borders:– Bigger than any wage discrimination– Bigger than any price distortion due to borders– Bigger than any poverty impact

by factor multiples (if not orders of magnitude)