Transcript Document

CMTEA 2008
The future of Europe in a world of uncertainties
Romania, Iaşi, September 25-27, 2008
Modeling migration flows:
explanations and policy implications
(the case of Luxembourg)
[email protected]
Luxembourg in Europe
Paris
The context (1)
• Migration
– migration = change of place of residence and workplace
( residential move)
– functional labour market areas (flma)  administrative
areas
– crossing borders: internal vs international
• Cross-border commuting
– travel daily or weekly from residence to workplace, not
necessarily, but generally within flma
• Luxembourg: commuters  cross-border workers (CBW)
The context (2)
• Importance of migrations and commuting for
Luxembourg
– average population growth: 0.9%
• net migration flows explain 60% of demographic growth
• today, 40% of the 0.5 million inhabitants are foreigners
– average employment growth: 2.6%
• commuters take 2/3 of net newly created jobs and make up
40% of total employment
• As a result, 60% of employed workers are
foreigners
• Another illustration: pop. aged 15-64: 322 000
total employment: 319 000
The context (3)
• This research takes place in the context of the
overall modelling of the Luxembourg economy
– estimated standard macro-model
– endogenize labour supply through modelling of
migrations and commuters
• Stylized facts
– net earnings and unemployment differentials with
neighbouring countries
• net earnings are higher, about 40%
• unemployment is lower, some 5 percentage points
– housing prices are higher in (and around) Lux. (>100%)
• other living costs (food, cothing) are less different
The context (4)
• Why this research might be interesting (for others)?
– not so many time series studies in the migration context (factors
influencing migrations)
– few time series studies that apply cointegration testing and error
correction techniques
– not many studies that compare factors affecting simultaneously
migrations and commuting
– this work could easily be extended to other regions/countries,
experiencing high in/out-flows of labour
• NB it is ongoing work, paper not finalised…
Literature review (1)
• Causes of migration
–
–
–
–
–
gravity models
human capital
income / leisure
job search + matching
equilibrium / disequilibrium
• Consequences of migration
– wages
– productivity
– demographic trends
Literature review (2)
• Gravity models
– based on the Newtonian law of gravitation
– not derived from theoretical modelling of economic
behaviour
– but widely used, with good results, can be estimated
• Mij = G * Pi0 * Pj 1 * Dij 2
–
–
–
–
–
Mij = migration from i to j
G = constant term
Pi = population of origin («weight»)
Pj = population in destination area
Dij = distance between both destinations
Literature review (3)
• Modified gravity models
– include variables linked to economic behaviour
– no formal derivation but taken from other theories
• Mij = G * Pi0 * Pj 1 * Dij 2 * Xi 3 * Xj 4
– Xi, Xj = economic and other variables related to regions
i and j
– Job opportunities, earnings, unemployment, housing
prices, risk, geographic characteristics (amenities),
political situation, etc...
The model (1)
• Data 1980-2006, yearly
• Test / impose restrictions on model coefficients
– taking ratios of independent variables: (Xi/ Xj)
–  reduction of the number of parameters to be
estimated
• Other simplifications
– drop Pi (foreign population) and Dij (distance)
• foreign population varies much less
• two “countries”: Luxembourg and “the rest of the world”
(ROW, to be defined)  aggregate flows
The model (2)
• ln(Mk/P) = 0k + 1k * ln(L/P) + 2k * ln(Yj/Yi) + 3k
* ln(Uj/Ui) + 4k * ln(HPj/HPi) + k
– j = Lux; i = ROW
– k = in, out, com:
• in: in-migration (flow)
• out: out-migration (flow)
• com: commuters (stock)
•
•
•
•
L = tot. labour demand in Lux.: 1in, com > 0; 1out = 0
Y = relative earnings: 2in, com >0; 2out<0
U = rel. unempl. rates: 3in, com < 0; 3out > 0
HP = rel. house prices: 4in < 0; 4out, com > 0
The model (3)
• Some precisions on the variables
– Migrations (Min, out, com):
• in, out = total (gross) flow
• com(muters) = stock of foreign workers travelling daily or
weekly from B, F, D to L
– Labour demand (L) = total domestic employment in Lux.
– Per capita earnings (Y):
• B, F, D (country wise); source = OECD (“Taxing wages”) 
after taxes and social transfers
– Unemployment rate (U):
• neighbouring regions (from B, F, D), Nuts3; source =Eurostat
– House prices (HP):
• neighbouring regions (from B, F, D), different sources
Estimation results (1)
• Order of integration
– all variables (ratios) entering the equations are I(1)
• Estimation of level equations (1st step of EngleGranger two step procedure)
– OLS, stationarity of residuals  cointegration?
– Results fail to confirm cointegrating relationship
(McKinnon critical values) but residuals “optically”
stationary…
Estimation results (2)
• Error correction models
– Dynamic ECM only works for CBW: cointegration
clearly confirmed by t-test on error-correction
parameter (Banerjee 1998)
– Others: retain static LR parameters ↔ Engle-Granger
two-step (or Zivot 2000)
• Endogeneity bias:
– to what extent the immigration rate does it cause (some
of) the independent variables (for example house
prices)?
•  to be studied
– other non-tackled issues: small sample bias, outliers…
Estimation results (3)
Table: Migration equations (elasticities*)
Independent variables 5
Test statistics
Error
correction
term
F-stat. (joint significance)
LM test stat.
0.23 0.06
0.45 0.06
0.93 0.01
0.45
0.51
0.32
0.34
0.35
0.16
Ajusted R-squared
Student
-0.45 -2.95
-0.25 -1.60
-0.12 -6.39
S.E. regression
0.66
-0.25
n.s. 0
n.s.
0.05
0.12 2
1.75 -1.33*** 1.67*** 0
Value
Dummies (number)
Rel. house prices
1.00
n.a.
1.00
Unemp. ratio
n.s.
0.07
0.06
Rel. revenues
Employment
0.27
-0.09
n.s.
0.06
0.19* -0.09***
Rel. revenues3
Employment
n.a.
0.64
0.34
n.a.
n.a. 1.17***
Rel. house prices5
0.35
n.s.
n.s.
Long run
Unemp. ratio4
In-migration
Out-migration
Cross-border workers 6
Lagged immigration
Dependent var. 1
Lagged dep.
Short run (variables in first difference)
Serial
correlation of
residuals (p
values)
* All variables in log-form
All migration variables are expressed as migration rates , i.e. migration flow divided by total population; the stock of cross.border workers is divided by total
employment
2
Coefs on indep. variables are elasticities; *=10% significance level; **=5%; ***=1%; No * ==> not significant (n.s.) at the 10% level (short run) except for the first
two equations where the long-run part is calibrated.
3
Revenues Lux. / Rev. abroad
4
Unemployment Lux. / Ue abroad
5
House prices Lux. / H. p. abroad
6
Contrary to the other equations, the cross-border equation is estimated in one step (dynamic ECM), hence significance levels on variables in the long run part (no
* ==> not significant)
7
n.a. = not applicable
1
Estimation results (4)
Final long-run specifications:
log(Min) = log(L) + 0.66*log(Yj/Yi) – 0.25*log(Uj/Ui)
log(Mout) = log(P) + 0.05*log(Uj/Ui) + 0.12*log(HPj/HPi)
log(Mcom) = log(L) + 1.75*log(Yj/Yi) – 1.33*log(Uj/Ui)
+1.67*log(HPj/HPi)
Simulations (1)
• Set up a model linking the labour market with
population dynamics:
– 3 migration equations
– population dynamics (linked to migrations)
– unemployment
• 2 simultanous feedback variables: population + unemployment
• But: partial model
– no feedback from unemployment to prices/wages
– total domestic employment (L) = exogenous
Simulations (2)
Rel. Disp.
income
Net Migrations
Total
population
Natural
movement
Commuters
Labour force
(act. pop.)
Exogenous
var.
Rel.
House
prices
Rel. unempl.
rate
Labour
demand /
Total empl.
Resident
unempl.
Endogenous var.,
(behavioural)
Resident
employment
Definition var.
Simulations (3)
• Integrate the “new” migration equations into a
complete macro-model:
–
–
–
–
–
–
wage equation (WS-PS), depending i.a. on UE
wage-price spiral
price-competitiveness
employment is endogenous
capacity constraints
etc…
• Simulate the same shocks in both set-ups (partial
and complete)
Simulations (4)
• Simulations: generate shocks to main RHS
variables:
– domestic labour demand and unemployment
– foreign unemployment, house prices and labour
earnings
• Rationality of the shocks:
– test impact of national policies acting on the labour
market: higher employment, lower unemployment
– reproduce stylized facts: higher unemployment in
bordering regions, lower net wages and house prices
Simulations (5)
• 10% increase in labour demand (in Lux.)
– increases resident employment and commuters (CBW)
• impact on CBW stronger (except for the two first years in
partial model) for a transition period, but, in the LR,
convergence towards increase of 10%
– part in newly created jobs: 2/3 commuters; 1/3 resident
– resident unemployment only decreases initially
• decrease in resident unemployment attracts new foreign
workers  unsustainable
– Full model: multiplier effects  impact on total
employment > 10%  decrease in resident UE a little
stronger
Partial model
Complete model
25
25
20
20
15
15
In-migration
10
In-migration
10
CBW
CBW
5
4.5
4.5
3.5
3.5
2.5
2029
2027
2025
2023
2021
2019
2017
2015
2013
2007
2029
2027
2025
2023
2021
2019
2017
-10
2015
-10
2013
-5
2011
-5
2009
0
2007
0
Out-migration
2011
Out-migration
2009
5
2.5
Unemployment rate (%
points)
1.5
Unemployment rate (%
points)
1.5
2028
2025
2022
-3.5
2019
-3.5
2016
-2.5
2013
-2.5
2010
-1.5
2007
-1.5
2028
Activity rate
2025
-0.5
2022
Activity rate
2019
-0.5
2016
Migration rate (% of
tot. pop.)
2013
0.5
2010
Migration rate (% of
tot. pop.)
2007
0.5
Partial model
Complete model
25
25
20
20
Cross-border
employment (CBW)
15
Cross-border
employment (CBW)
15
Total employment
10
80
80
70
70
2029
2027
2025
2023
2021
2019
2017
2015
60
50
2029
2027
2025
2023
2021
2019
2017
2015
2029
2027
2025
2023
2021
2019
2017
20
2015
20
2013
30
2011
30
2009
40
2007
40
% of new jobs taken
by CBW
2013
% of new jobs taken
by CBW
2011
50
% of new jobs taken
by res. employment
2009
% of new jobs taken
by res. employment
2007
60
2013
2007
2029
2027
2025
2023
2021
2019
2017
2015
2013
0
2011
0
2009
5
2007
5
Resident employment
2011
Resident employment
2009
10
Total employment
Complete model
3.5
18
16
3.0
14
2.5
GDP (vol.)
2.0
National demand (vol.)
1.5
Exports (vol.)
1.0
GDP deflator
Other exports
12
Exports of goods (vol.)
10
Exports of services
(vol.)
8
6
Consumption of nonresidents (vol.)
4
0.5
2030
2020
2015
2010
2030
2020
-2
2015
-0.5
2010
0
2007
0.0
2007
2
Simulations (6)
• 1 ppt decrease in domestic unemployment (UE)
– the initial decrease in domestic UE increases foreign
labour supply…
– …which pushes up UE in L
• there is a 1:1 substitution between resident workers and CBW
– as a result, the decrease in UE is almost completely
reversed
• only in the complete model is there a sligthly bigger decrease
in resident UE, because migrations increase less…
• …due to lower net wages (overall negative demand shock)
Partial model
Complete model
0.50
0.50
0.25
0.25
Activity rate (% of tot.
pop.)
2007
2028
2025
2022
2019
2016
-0.75
2013
-0.75
2010
-0.50
2007
-0.50
2.5
2.5
2.0
2.0
1.5
1.5
In-migration
1.0
2028
-0.25
2025
Activity rate (% of tot.
pop.)
2022
-0.25
2019
Unemployment rate (%
points)
2016
0.00
2013
Unemployment rate (%
points)
2010
0.00
In-migration
1.0
CBW
2029
2027
2025
2023
2021
2019
2017
2015
2013
2011
2029
2027
2025
2023
2021
2019
2017
-1.0
2015
-1.0
2013
-0.5
2011
-0.5
2009
0.0
2007
0.0
Out-migration
2009
Out-migration
CBW
0.5
2007
0.5
Partial model
Complete model
1.0
0.5
GDP (vol.)
National demand (vol.)
0.0
GDP deflator
-0.5
2.5
2.5
2.0
2.0
1.5
1.5
2029
2027
2025
2023
2021
2019
2017
2015
2013
2011
2029
2027
2025
2023
2021
2019
2017
Resident employment
2015
2029
2027
2025
2023
-1.5
2021
-1.5
2019
-1.0
2017
-1.0
2015
-0.5
2013
-0.5
2011
0.0
2009
0.0
2007
Total employment
0.5
2013
Resident employment
2011
0.5
Cross-border
employment (CBW)
1.0
2009
Cross-border
employment (CBW)
2007
1.0
2009
2007
-1.0
Simulations (7)
• Modifiy (foreign, exogenous) variables that act on
foreign labour supply:
– unemployment
– earnings
– house prices
• Modifiy these variables in a way to emphasize
stylized facts:
– higher UE, lower earnings and lower house prices in the
neighbouring regions
Simulations (8)
• Results:
– in all cases, increased foreign labour supply depresses
resident employment and increases res. UE
– the initial negative impact on GDP reverses after some
periods, due to the favorable evolution of price
competitiveness (fall in domestic prices)
– in case of a fall in foreign house prices, the negative
demand shock lasts longer (although the amplitude of
the results of the shocks on the national variables can
generally not be compared)
Impact on in-migration
Impact on cross-border employment
3.0
7.0
2.0
Increase in foreign UE,
partial
6.0
Increase in foreign UE,
partial
1.0
Increase in foreign UE,
complete
5.0
Increase in foreign UE,
complete
0.0
Decrease in foreign net
w ages, partial
4.0
Decrease in foreign net
w ages, partial
-1.0
Decrease in foreign net
w ages, complete
3.0
Decrease in foreign net
w ages, complete
-2.0
Decrease in foreign house
prices, partial
2.0
Decrease in foreign house
prices, partial
-3.0
Decrease in foreign house
prices, complete
1.0
Decrease in foreign house
prices, complete
Impact on resident employment
2029
2027
2025
2023
2021
2019
2017
2015
2013
2011
2009
2007
2029
2027
2025
2023
2021
2019
2017
2015
2013
2011
2009
0.0
2007
-4.0
Impact on resident UE
1.6
0.0
Increase in foreign UE,
partial
-1.0
-2.0
Increase in foreign UE,
complete
-3.0
1.4
Increase in foreign UE,
partial
1.2
Increase in foreign UE,
complete
1.0
-4.0
Decrease in foreign net
w ages, partial
0.8
Decrease in foreign net
w ages, partial
-5.0
Decrease in foreign net
w ages, complete
0.6
Decrease in foreign net
w ages, complete
Decrease in foreign house
prices, partial
0.4
Decrease in foreign house
prices, partial
Decrease in foreign house
prices, complete
0.2
Decrease in foreign house
prices, complete
-6.0
-7.0
-8.0
2029
2027
2025
2023
2021
2019
2017
2015
2013
2011
2009
2029
2027
2025
2023
2021
2019
2017
2015
2013
2011
2009
2007
2007
0.0
-9.0
Impact on GDP (vol.), complete model
2.0
Impact on national demand, complete model
1.0
Increase in foreign UE (+1
% point)
1.5
1.0
Increase in foreign UE (+1
% point)
0.5
0.0
Decrease in foreign net
w ages (-5%)
0.5
0.0
Decrease in foreign house
prices (-10%)
-0.5
Decrease in foreign net
w ages (-5%)
-0.5
-1.0
Decrease in foreign house
prices (-10%)
Impact on total employment, complete model
2.5
2029
2027
2025
2023
2021
2019
2017
2015
2013
2011
2009
2007
2029
2027
2025
2023
2021
2019
2017
2015
2013
2011
2009
2007
-1.5
Impact on GPD deflator, complete model
0.0
2.0
Increase in foreign UE (+1
% point)
Increase in foreign UE (+1
% point)
-0.5
1.5
-1.0
1.0
Decrease in foreign net
w ages (-5%)
Decrease in foreign net
w ages (-5%)
-1.5
0.5
Decrease in foreign house
prices (-10%)
0.0
-0.5
-2.0
Decrease in foreign house
prices (-10%)
2029
2027
2025
2023
2021
2019
2017
2015
2013
2011
2009
2007
2029
2027
2025
2023
2021
2019
2017
2015
2013
2011
2009
2007
-2.5
Conclusions (1)
• Econometric evidence confirms the importance of
earnings, unemployment and house prices for
explaining cross-border worker’s (commuters)
movements
• Estimations of migration equations are less robust
(econometrically), but the obtained coefficients are
sensible
Conclusions (2)
• A positive demand shock on the national economy,
having an impact on employment and/or
unemployment, increases foreign labour supply,
possibly as much as to reverse, partially or totally,
the positive initial impact of the favourable shock
• Increased foreign labour supply, due to
unfavourable exogenous causes (negative shocks
on foreign economies), is generally positive for the
domestic economy, after some lags, with the
exception of unemployment, that increases
Thank you very much for your
attention
Questions?