Motivation & factsModelResults
Conclusion
The Changing Structure of Immigration to theOECD: What Welfare E¤ects on Member States
M. Burzynski, F. Docquier, H. Rapoport
ZiF Conference on "In Search of the Global Labour Market"
Bielefeld, Oct 16-18, 2017
Motivation & factsModelResults
Conclusion
What do we do?
I ∆ structure of immigration ( ∆ origin mixI E¤ect on natives�welfare (real income) in the OECD memberstates over last 25 years (1990-2015):
I Non-estimation paper: quantitative theoryI Combination of channels of transmissionI Comparative study on 20 selected OECD countriesI Comparison of three recent immigration waves !!!
I Shed light on...I Dynamics of welfare e¤ects of immigrationI Comparison between the post crisis and pre-crisis wavesI Di¤erences across countries vs across immigration waves
Motivation & factsModelResults
Conclusion
Map of the talk
1. Motivation & facts2. Model
3. Results
4. Concluding remarks
Motivation & factsModelResults
Conclusion
Origin mix!
" The most recent waves have signi�cantly lower earnings and labor force
participation rates, work fewer weeks, and have higher unemployment
propensities than earlier waves. [...] One single factor, the changing national
origin mix of the immigrant �ow, is mostly responsible for these historical
trends" (Borjas, 1993)
"A few source country characteristics explain over two-thirds of the variance of
welfare recipiency rates across national origin groups, and changes in the
average source country characteristics of the foreign-born population between
1970 and 1980 can account for most of the rise in immigrant welfare use that
occurred over the decade" (Borjas-Trejo, 1993)
Context: drastic policy reform!
Motivation & factsModelResults
Conclusion
Origin mix!
Immigration to the United States19th Century + 1920-1965 After 1965
National Origins Formula 1965 Immig & Nat Act
Immigration of Europeans Immig from Mexico, Asia, Africa
Motivation & factsModelResults
Conclusion
Origin mix!
Similar context: immigration to Australia1901-1945 After 1970
White Australia Policy Multiculturalism
Immigration of Europeans Immigration from Asia
Motivation & factsModelResults
Conclusion
Origin mix?
Origin mix in last 25 years: other factors?
I Schengen (1990) + EU enlargment (2004)I Points-based system (e.g., Canada 1980�s, UK 2008-10)I Civil wars and con�icts (e.g., European migrant crisis, 2015)I Economic imbalances (rather stable)I Demographic imbalances: Dev/HI population ratio
I 3.0 in 1960I 5.3 in 2016 !!!!
Motivation & factsModelResults
Conclusion
Facts
Changing size and origin mixRising in�ows from LDCs
Data for EU15 (black), USA (red), Canada, Australia, Japan & Switz.
Motivation & factsModelResults
Conclusion
Facts
Immigration to OECD member states:
I Sharp rise in the share of immigrants btw 1960 and 2010I From 4.5% to 11% (+6.5 pp), on average
I And changing origin-mixI Immig from HIC: stable 3%I Immig from LDC: from 1.5% to 8%I Share from LDC: from 1/3 to 3/4
=) More and more invididuals originating from countries that areeconomically, geographically, institutionally, culturally distant=) Does it matter? How is it perceived?
Motivation & factsModelResults
Conclusion
Perceptions
Changing origin mix =) xorit /xdesttLower education and productivity at origin?
Motivation & factsModelResults
Conclusion
Perceptions
Changing origin mix =) xorit /xdesttLower participation rates at origin?
Motivation & factsModelResults
Conclusion
Perceptions
I Common portrayal of recent immigrants (opinion polls):I Less and less educated, less and less productiveI Remain poor and inactive: access to welfare systemsI Induce wage/job losses, increasing inequality, �scal de�citsI Induce contamination or epidemiological e¤ects...
I But migrants self-select on many attributes:I DIOC database on population structure in OECD member states
I By dyad (220 origin and 20 destination countries)I By age, education, labor market status
I Let�s focus on the stock of immigrants in 2010...
Motivation & factsModelResults
Conclusion
Back to facts
Selection by education and participationOrigin (X) versus Dyad (Y)
Motivation & factsModelResults
Conclusion
Back to facts
Ori/Dyad correlation and corridor sizeCorr (Ori, Dyad) = 0.22(edu) vs 0.05(lab) ( 6= for top-50 corridors)
Motivation & factsModelResults
Conclusion
Back to facts
I Correlations "origin/dyadic" characteristics are smallI Sign of self-selection and/or partial assimilationI Still, they are >0 and larger for main dyads (0.30(edu), 0.55(lab))I Large enough for changing origin mix to a¤ect welfare?
I Develop a model accounting for...I Dyadic disparities in immigrant characteristicsI Main channels of transmission of migration shocksI Impact of successive immigrant cohorts (∆ origin mix)
Motivation & factsModelResults
Conclusion
MechanismsParameterization
Map of the talk
1. Motivation & facts
2. ModelI MechanismsI Parameterization
3. Results
4. Concluding remarks
Motivation & factsModelResults
Conclusion
MechanismsParameterization
General structure
I Static modelI No trade, no physical capitalI Country = independent entity with
I Heterogeneous individualsI Heterogeneous �rmsI �Passive�government
I Exogenous demographic size: Noa,sI With o = n vs.o = (f 1, f 2, ..., f F ), s = (h, l), a = (y , o)
I Endogenize the e¤ect of ∆Noa,s on the state of the economy
Motivation & factsModelResults
Conclusion
MechanismsParameterization
General structure
I Channels of transmission in the benchmarkI Labor market: endogenous wages (Ottaviano-Peri) + endogenousparticipation rates + exogenous unemployment rates
I Public �nances: endogenous income tax rate (Storesletten)I Market size: endogenous prices (Krugman; Aubry et al.)
I Robustness with less consensual mechanismsI Schooling externalities + Cultural diversity + Diaspora
I Microfoundations...
Motivation & factsModelResults
Conclusion
MechanismsParameterization
Individuals
I Max utility
Uoa,s = Coa,s �
φoa,s (1� γoa,s )1+η
1+ η
I Subject to
C oa,s =
�Z B
0coa,s (i)
ε�1ε di
� εε�1
Z B
0coa,s (i)p(i)(1+ υ)di = (1� γoa,s )w
os + T
oa,s
I where wos = (1� uos )wos (1� τ) + uos δwosI This determines (C oa,s ,γ
oa,s ,U
oa,s ) and aggregate labor supply
Motivation & factsModelResults
Conclusion
MechanismsParameterization
Firms
I Continuum of monopolistic producers with a measure B:I Nested CES production framework
y(i) = Aq(i) = A�
θ1h(i)σ1�1
σ1 + (1� θ1)l(i)σ1�1
σ1
� σ1σ1�1
h(i) =
�θ2hn(i)
σ2�1σ2 + (1� θ2)hf (i)
σ2�1σ2
� σ2σ2�1
l(i) =
�θ2ln(i)
σ2�1σ2 + (1� θ1)l f (i)
σ2�1σ2
� σ2σ2�1
I Max pro�t s.t. techno + demand: price + labor demandI Entry costs = ψ units of q(i): B and aggregate labor demand
Motivation & factsModelResults
Conclusion
MechanismsParameterization
Government
I Government budget constraint:
(v + τ)Y = δ ∑o ,a,s Noa,s (1� γoa,s )u
os w
os +∑o ,a,s N
oa,sT
oa,s
I Consumption tax = v (exo); Income tax = τI Unemployment replacement rate = δ (exogenous)I Public cons + Group-speci�c transfers: T oa,s (exogenous)
I This determines τ
Motivation & factsModelResults
Conclusion
MechanismsParameterization
Monopolistic equilibrium
De�nition
For a set of common param. fε, η, σ1, σ2g, a set of dest-speci�cvar.
�u0s , θ1, θ2,A,ψ, δ,T
na,s ,T
fa,s/T na,s , v
, a set of dyadic var.�
φoa,s ,Noa,s
, the monopolistically competitive equil. is a set�
wos , coa,s ,γ
oa,s , `
n, hn, `f , hf , y , p,P,B, τof endogenous var. that
satis�es the following conditions:
(i) individuals maximize their utility(ii) optimal employment and zero-pro�t conditions hold(iii) labor markets clear(iv) the government budget is balanced
Motivation & factsModelResults
Conclusion
MechanismsParameterization
Map of the talk
1. Motivation & facts
2. ModelI MechanismsI Parameterization
3. Results
4. Concluding remarks
Motivation & factsModelResults
Conclusion
MechanismsParameterization
Parameterization
Common parameters (consensus levels from 9 literature):I ES between varieties of goods: ε = 7.0 (Feenstra, 1994)I ES between HS and LS: σ1= 2.0 (Ottaviano-Peri, 2012)I ES between Nat and Immig: σ2= 20.0 (Ottaviano-Peri, 2012)I Elasticity of labor supply: 1/η = 0.1 (Evers et al, 2008)
Motivation & factsModelResults
Conclusion
MechanismsParameterization
Parameterization
Other parameters: calibration on the year 2010 to exactly �t
I Population and labor force data (Noa,s ): DIOCI Labor force data (1� γoa,s , u
os ): DIOC
I Income data (Y ,wos ): OECD.Stat + Educ at a glance + Buchel et
al. (2008)
I Fiscal data (v , τ, δ,T oa,s ): Annual national accounts + SOCX +
household surveys + OECD (2007)
Motivation & factsModelResults
Conclusion
MechanismsParameterization
Parameterization
Parameters with some country variations:
I Disutility of labor (φoa,s ) matches participation rates by dyad/skillI Technological parameters (θ1, θ2) match wage distributionI TFP (A) matches GDPI Entry cost (ψ) = nb of days to create a �rm (US=1)
I Fiscal parameters (v , τ, δ,T na,s ) match OECD/SOCX dataI Transfers to immigrants (T fa,s/T
na,s ) match �scal contribution of
immigration (OECD, 2007)
Motivation & factsModelResults
Conclusion
MechanismsParameterization
Parameterization
Table 1. Country-speci�c parametersParam Description Mean s.d. Source/Moment
φoa,s Lab disutility (�US) 1.675 1.487 Matches γoa,suos Unemp rates 0.095 0.072 Matches DIOC
θ1 Firms�pref HS 0.557 0.050 Matches wh/w lθ2 Firms�pref native 0.527 0.040 Matches wns /w fsA TFP (�US) 0.894 0.294 Matches GDP
ψ Cost of entry (�US) 1.435 0.952 Nb. days
δ Replacement rate 0.600 0.300 Un. Exp/GDP
T na,s Pub transf (�GDPpc) 0.321 0.089 Gov. Exp/GDP
T fa,s Pub transf (�T na,s ) 1.066 0.467 Fiscal cont. im.
v Cons tax rate 0.173 0.042 OECD data
Motivation & factsModelResults
Conclusion
Welfare e¤ectsChannelsOrigin mixCaveatsRobustness
Map of the talk
1. Motivation & facts
2. Model
3. ResultsI Welfare e¤ectsI ChannelsI Origin mixI CaveatsI Robustness
4. Concluding remarks
Motivation & factsModelResults
Conclusion
Welfare e¤ectsChannelsOrigin mixCaveatsRobustness
Welfare e¤ects
I Assessing the "causal" impact of immigration usingcounterfactuals:
I Eliminating immigrants arrived in 2001-2010 (pre-crisis 1)I Eliminating immigrants arrived in 1991-2000 (pre-crisis 2)I Adding immigrants arrived in 2011-2015 (post crisis)
I Welfare e¤ect for the native population in 2010I Shocks have di¤erent sizes ) semi elasticity
∆Una,s/Una,s∆m
=(Una,s )WithMig � (Una,s )NoMig
∆m.(Una,s )NoMig
Motivation & factsModelResults
Conclusion
Welfare e¤ectsChannelsOrigin mixCaveatsRobustness
Welfare e¤ects
Welfare and inequality e¤ects:
I Average welfare e¤ect on natives (∆C nyCny
/∆m)
I Inequality e¤ect ((∆C ny ,hC ny ,h
� ∆C ny ,lC ny ,l
)/∆m)
Motivation & factsModelResults
Conclusion
Welfare e¤ectsChannelsOrigin mixCaveatsRobustness
Welfare e¤ects
Semi-elasticiy of average real income to m
(i) Last 3 waves have increased the average real income of natives
(ii) Strong and persistent variations across countries
(iii) Post-crisis is usually less bene�cial
Motivation & factsModelResults
Conclusion
Welfare e¤ectsChannelsOrigin mixCaveatsRobustness
Welfare e¤ects
Semi elasticity of HS-to-LS income ratio
(i) Heterogeneous inequality e¤ects of immigration
(ii) Increasing inequality in continental EU (but no loss for LS)
(iii) Post-crisis wave is more inegalitarian in 2/3 of countries
Motivation & factsModelResults
Conclusion
Welfare e¤ectsChannelsOrigin mixCaveatsRobustness
Channels
What are the channels of transmission?
I Inequality e¤ect is due to wage e¤ect (skill-speci�c)I As for average welfare gains...
I Average price index (∆PP /∆m)
I Aggregate employment rates (∆e/∆m)I Income tax rate (∆τ/∆m)
Motivation & factsModelResults
Conclusion
Welfare e¤ectsChannelsOrigin mixCaveatsRobustness
Channels
E¤ect on the price index (∆PP /∆m < 0)
Virtually una¤ected: very strong persistence, bene�cial everywhere
Motivation & factsModelResults
Conclusion
Welfare e¤ectsChannelsOrigin mixCaveatsRobustness
Channels
E¤ect on employment rates (∆e/∆m ? 0)
Slightly smaller in 2011-15: very strong persistence
Motivation & factsModelResults
Conclusion
Welfare e¤ectsChannelsOrigin mixCaveatsRobustness
Channels
E¤ect on income tax rates (∆τ/∆m < 0)
Bene�cial everywhere... but usually less bene�cial in 2011-15
Motivation & factsModelResults
Conclusion
Welfare e¤ectsChannelsOrigin mixCaveatsRobustness
Immigration waves
Why have bene�cial e¤ects decresed over time...
I Mechanical e¤ect of immigration onI Support ratio (young-to-retiree) (∆s/∆m)I Proportion of college-educated workers (∆h/∆m)
Motivation & factsModelResults
Conclusion
Welfare e¤ectsChannelsOrigin mixCaveatsRobustness
Immigration waves
E¤ect on support ratio (∆s/∆m� 100)
Virtually una¤ected: always >0, especially in countries with low Ret/Pop
Motivation & factsModelResults
Conclusion
Welfare e¤ectsChannelsOrigin mixCaveatsRobustness
Immigration waves
E¤ect on human capital (∆h/∆m� 100)
Strong persistence across waves... but post-crisis wave usually less educated
Motivation & factsModelResults
Conclusion
Welfare e¤ectsChannelsOrigin mixCaveatsRobustness
Caveats
I Comparability of waves?I 1991-2000 vs 2001-2010: older migrants (less educated) + negativeselection in return migration (those who stayed are more educated)
I 2001-2010 vs 2011-2015: calibrated using of ∆stock instead ofstock by year of entry + using the origin mix to predict theeducation structure + smaller size
I More...
Motivation & factsModelResults
Conclusion
Welfare e¤ectsChannelsOrigin mixCaveatsRobustness
Caveats
I Comparability of waves?I 1991-2000 vs 2001-2010I 2001-2010 vs 2011-2015
I Role of origin mix? Does the changing origin mix between
1991-2000 and 2001-10 predict the ∆ in skill structure?
I Size: non linearities? If we multiply the 2011-15 immigration wave
by two, does it change the semi-elasticity of average real income to m?
I Endogeneity of migration? Mt can a¤ect Mt-1 or Mt+1
Motivation & factsModelResults
Conclusion
Welfare e¤ectsChannelsOrigin mixCaveatsRobustness
Caveats
Role of origin mix: hmig91�00 Non linearities: ∆U/(U∆m)
CF: bhmig91�00 = ∑i ωi91�00h
i01�10 Assuming 2016-20 = 2011-15
Corr(hmig91�00,bhmig91�00) =. 97 Corr(2011-15,2011-20) =.99
Motivation & factsModelResults
Conclusion
Welfare e¤ectsChannelsOrigin mixCaveatsRobustness
Robustness
Do the welfare e¤ects depend on...
I Common elasticities (ε, σ2, η)I Potential technological externalities (schooling, birthplacediversity, diaspora)
I Immigrant characteristics (φ, u, h) identical to natives
Focus on the 2001-2010 wave!
Motivation & factsModelResults
Conclusion
Welfare e¤ectsChannelsOrigin mixCaveatsRobustness
Robustness
Sensitivity to elasticities
Larger gains if ε large, η smaller - Negligible e¤ects on inequality
Motivation & factsModelResults
Conclusion
Welfare e¤ectsChannelsOrigin mixCaveatsRobustness
Robustness
Sensitivity to TFP externalities
Schooling externalities matter in some countries - No e¤ects on inequality
Motivation & factsModelResults
Conclusion
Welfare e¤ectsChannelsOrigin mixCaveatsRobustness
Robustness
Sensitivity to immigrants�characteristics
What matters for inequality: skill composition of immig. population
Motivation & factsModelResults
Conclusion
Map of the talk
1. Motivation & facts
2. Model
3. Results
4. Concluding remarks
Motivation & factsModelResults
Conclusion
Conclusion
Welfare impact of immig. over last 25 years:
I Large variations across destinations and skill groups
I Variations across countries > Variations across waves (limitedimpact of origin mix over the last 25 years)
I In relative terms, the post crisis is usually less bene�cialI Due to changing origin-mix ) lower h, smaller �scal gainsI Still, no negative e¤ect on native citizens in virtually all countries
I Gradual squeeze of welfare gain from immigration?
Motivation & factsModelResults
Conclusion
Thanks for your attention
Motivation & factsModelResults
Conclusion
Appendix
Size of immigration waves (∆m)