The Absorption of Immigrants and its Effects on the Thai Wage Structure

Dilaka LathapipatThailand Development Research Institute

ObjectivesTo study the diffusion of low-skilled labour in different provinces, and how local industries respond to these relative supply shiftsThis will help us understand how industries across geographical areas absorb inflows of low-skilled immigrants

Mostly from the Greater Mekong Sub-region and represent around 5 percent of the 2007 labour force

To examine the (long-run) effects on wages – Thai and foreign wages in various education-experience groups

Diffusion of SkillsUse 2007 Q3 Labour Force Survey

Most recent coverage of (registered) immigrant workersAround three quarters of immigrants are unregistered, and the census undercount could be as high as 80 percent

Divide workers (men and women ages between 16 and 65 years) into 4 schooling groups (LP, UP, HS, CO)

Lower Primary: Schooling<=3 yearsUpper Primary: 4 years=<Schooling<=6 yearsHigh School: 7 years=<Schooling<=12 yearsCollege: Schooling>12 years

Diffusion of SkillsOur (weekly) labour supply measure:

“total weekly hours” plus “desired additional hours”Captures the extent of involuntary underemploymentDiscard those workers not in the labour force (who reported zero weekly hours and were not seeking work)For the unemployed workers, we impute their missing laboursupplies using information from the rest of the sampleUnemployment rate using our broader measure is 2.87%Conventional headline measure of unemployment is 1.81%Foreign supply accounts for 0.91 percent of total supplyEach worker therefore has a fraction of hours employed and hours unemployed

Diffusion of SkillsCard and Lewis (2005) decompose variation in the overall fraction of a skill group employed in a given city from the national average into 3 components:

Between industry (B)Within industry (W)Interaction term (I)

To make statements about the relationships between skill supplies, industry structure, and unemployment we modify the CL modelInstead, we decompose variation in the overall fraction of a skill group supplied in a given province from the national average

Diffusion of SkillsThis necessarily entails a 4th component:

Unemployment (U)Begin with the identity:

( ) 1 ( )( ) ( )( ) ( ) ( )

( ) ( ) ( )

( ) ( )

( ) ( ) ( )

i ii i

jj

ij j i

ujj

i ij j uj

L p u ps p N pL p L p L p

N p N ps p

L p N p

p p s pλ ω

= = +

= +

= +

∑

∑

∑where L(p) and N(p) are hours supplied and employed in

province p, schooling group and industry are indexed by i and j, and ui(p) is the hours unemployed in province p

Diffusion of SkillsThe employment share of industry j indicates the size of the industry in the locality:

( ) ( ) / ( )j jp N p L pλ ≡

While the share of skill group i employed in industry jindicates the skill intensity of the industry:

( ) ( ) / ( )i ij j jp N p N pω ≡

We decompose the gap between si(p) and the national average into the 4 components:

( ) ( ) ( ) ( ) ( )i is p s B p W p I p U p− = + + +

Diffusion of Skillswhere ( ) ( )

( ) ( )

( ) ( ) ( )

( ) ( )

ij j jj

i ij j jj

i ij j j jj

i iu u

B p p

W p p

I p p p

U p s p s

ω λ λ

λ ω ω

λ λ ω ω

⎡ ⎤= −⎣ ⎦

⎡ ⎤= −⎣ ⎦

⎡ ⎤⎡ ⎤= − −⎣ ⎦ ⎣ ⎦

= −

∑∑∑

( ) ( )

( ) ( )

( ) ( )

( ) ( )

i iB B B

i iW W W

i iI I I

i iU U U

B p s p s

W p s p s

I p s p s

U p s p s

α β ε

α β ε

α β ε

α β ε

⎡ ⎤= + − +⎣ ⎦⎡ ⎤= + − +⎣ ⎦

⎡ ⎤= + − +⎣ ⎦⎡ ⎤= + − +⎣ ⎦

And estimate the following regressions:

where the 's necessarily sum to 1β

Diffusion of SkillsUnder the strict Hecksher-Olin condition, variation in the shares of a skill supply across provinces is absorbed by expansion or contraction of industries that employ the skill type more intensively, that is, via the between industry component B(p), and with no change in relative wagesWe formally test the hypothesis to see if βB=1Focus on the two schooling groups with less than a high school education

Diffusion of Skills

-0.05

0.00

0.05

0.10

0.15

-0.05 0.00 0.05 0.10 0.15

Excess Fraction of Lower Primary Labour

Betw

een-

Indu

stry

Com

pone

nt

-0.05

0.00

0.05

0.10

0.15

-0.05 0.00 0.05 0.10 0.15

Excess Fraction of Lower Primary Labour

With

in-I

ndus

try

Com

pone

nt

-0.2

-0.12

-0.04

0.04

0.12

0.2

-0.2 -0.12 -0.04 0.04 0.12 0.2

Excess Fraction of Upper Primary Labour

Betw

een-

Indu

stry

Com

pone

nt

-0.2

-0.12

-0.04

0.04

0.12

0.2

-0.2 -0.12 -0.04 0.04 0.12 0.2

Excess Fraction of Upper Primary Labour

With

in-I

ndus

try

Com

pone

nt

Diffusion of SkillsMuch flatter slopes for the between industry absorption of lower and upper primary labourLimited evidence of HO-style absorption of excess supplies of these two skill groupsIndustries in low-skilled labour-abundant provinces are more intensive in their use of this type of labourSubstantial portions of excess supplies of lower and upper primary labour are associated with higher unemployment rates

7.2 % of excess fraction of lower primary labour absorbed by unemployment, and 6.3% for upper primary labour

Diffusion of SkillsTable 2.1: Absorption of Excess Lower Primary and Upper Primary Workers

Diffusion of SkillsTable 2.3: Absorption of Excess Lower Primary Thai and Immigrant

Workers

Diffusion of SkillsTable 2.4: Absorption of Excess Upper Primary Thai and Immigrant

Workers

Diffusion of SkillsTables 2.3 and 2.4 further disaggregate lower and upper primary workers into Thai and Foreign portionsMost striking result is the much lower between industry absorption of immigrant relative to Thai excess supplies

For lower primary groups – 1.8% of excess foreign supply is associated with shifting industry structure, while the corresponding number for Thais is 21.6%For upper primary groups – for immigrants, and for Thais

A closer look into the contribution of selected industries reveals that sector-specific estimates of βB’s are all zeroes for immigrant labour

ˆ 0.4%Bβ =ˆ 30.2%Bβ =

Diffusion of SkillsComparing the absorption contribution of agribusiness sector reveals that

While the entire immigrant excess supplies are absorbed through the within industry component, the exact opposite is observed for similarly-educated Thai labourEvidence of HO-style absorption of Thai excess supplies in this industryEvidence that employers regard foreign workers as temporary

Excess supplies of immigrants are essentially uncorrelated with their rates of unemployment

Since the sample contains only registered workers, this observation is not at all surprising

Effects on WagesThe small role of shifting industry structure in the absorption of immigrant excess supplies indicates that unemployment rates and/or the wage structure must adjustIn addition to language barriers and other factors, we have established evidence that immigrants and natives are imperfect substitutes in production due to their temporary worker statusFramework to study wage changes borrowed from Ottavianoand Peri (2008) – OP from here onDue to data availability, we evaluate variations in wages and employment across geographical areas rather than through time as in OP

Effects on WagesWe also make the implicit assumption that industries across provinces adjust their physical capital to accommodate differences in relative skill suppliesOur model can thus be considered a long-run modelWe begin by defining a general production function:

( , )p p pY F K N=Following OP, the labour CES aggregate, N, is defined as:

1 1 1HL

HL HL HLHL HL

p Lp Lp Hp HpN N N

σσ σ σσ σθ θ

− − −⎡ ⎤= +⎢ ⎥⎢ ⎥⎣ ⎦

where NL

and NH

are aggregate measures of low-

and high-educationlabour, σHL is the elasticity of substitution parameter, and the θ’s are the

relative efficiency parameters

Effects on Wages

The detailed education group – denoted by k – further nests workers in 3 potential experience groups:

The broad education groups are CES aggregates of the 4 detailed education groups defined earlier:

1 1 1LL

LL LL LLLL LL

Lp LPp LPp UPp UPpN N N

σσ σ σσ σθ θ

− − −⎡ ⎤= +⎢ ⎥⎢ ⎥⎣ ⎦

1 1 1HH

HH HH HHHH HH

Hp HSp HSp COp COpN N N

σσ σ σσ σθ θ

− − −⎡ ⎤= +⎢ ⎥⎢ ⎥⎣ ⎦

1 1EXP

EXP EXPEXP

kp kj kjpjN N

σσ σσθ

− −⎡ ⎤= ⎢ ⎥⎢ ⎥⎣ ⎦∑

Group j=1: Experience<=14 yearsGroup j=2: 15 years=<Experience<=29 yearsGroup j=3 : Schooling>=30 years

Effects on Wages

Equating the value of marginal productivity of a generic domestic worker to the wage rate, it can be shown that:

The Nkj’s are CES aggregates of foreign and domestic workers within the same kj skill groups

1 1 1IMMI

IMMI IMMI IMMIIMMI IMMI

kjp Dkj kjp Fkj kjpN D F

σσ σ σσ σθ θ

− − −⎡ ⎤= +⎢ ⎥⎢ ⎥⎣ ⎦

1 1 1ln( ) ln ln( ) ln ln( ) ln

1 1 1 1 ln( ) ln ln( )

1 ln ln( )

Dbkjp p p bp bp kpHL HL bb

kp kj kjpbb EXP EXP IMMI

Dkj kjpIMMI

w N N

N N

D

ζ θ θσ σ σ

θσ σ σ σ

θσ

⎛ ⎞= + + − − + −⎜ ⎟

⎝ ⎠⎛ ⎞ ⎛ ⎞

− + − − +⎜ ⎟ ⎜ ⎟⎝ ⎠⎝ ⎠

−

Effects on WagesThe same wage equation for foreign workers (changing the notation from D to F)

Estimation of Elasticity of Substitution ParametersHere we only show the estimation of σIMMI :Taking the difference between the previous equation for Fand D:

1ln ln lnFbkjp Fkj kjp

Dbkjp Dkj IMMI kjp

w Fw D

θθ σ

⎛ ⎞ ⎛ ⎞ ⎛ ⎞= −⎜ ⎟ ⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟ ⎜ ⎟

⎝ ⎠ ⎝ ⎠ ⎝ ⎠

The above expression can be estimated by running the following regression with detailed education-by-experience fixed effects :

Effects on Wages

ˆexp( ) 1ˆ ˆ and ˆ ˆ1 exp( ) 1 exp( )kj

Fkj Dkjkj kj

II I

θ θ= =+ +

1ln lnFbkjp kjpkj bkj

Dbkjp IMMI kjp

w FI u

w Dσ⎛ ⎞ ⎛ ⎞

= − +⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟⎝ ⎠ ⎝ ⎠

The efficiency parameters, θ’s, (normalised to sum to 1) can be estimated as follows:

The estimated σIMMI , θFkj and θDkj are then used to construct the aggregate labour input Nkj using the expression:

ˆˆ ˆ1 1 ˆ 1ˆ ˆˆ ˆˆ

IMMIIMMI IMMI IMMI

IMMI IMMIkjp Dkj kjp Fkj kjpN D F

σσ σ σσ σθ θ

− − −⎡ ⎤= +⎢ ⎥⎢ ⎥⎣ ⎦

Note that the Dkjp

and Fkjp

are observed in the sample

Effects on WagesTable 3.1: Regression Estimates of the Elasticity of Substitution Parameters

Effects on WagesThe relative magnitudes of the estimates do make senseUsing these estimates together with data on wages and supplies of each skill group, we can simulate the effects on Thai wages of any immigrant supply shocks using the following expression:

3 3

1 1

3

1

1 1 1ˆ ˆ ˆ

1 1 1 1 ˆ ˆ ˆ ˆ

Dbkjp Fcqip cqip cqip Fbqip bqip bqip

c B q E i q b iDbkjp HL p p cqip HL bb bp bp bqip

Fbkip bkip bkip F

ibb EXP kp kp bkip EXP IMMI

w w F F w F Fw w N F w N F

w F F ww N F

σ σ σ

σ σ σ σ

∈ ∈ = ∈ =

=

Δ Δ Δ⎛ ⎞= − −⎜ ⎟

⎝ ⎠Δ⎛ ⎞ ⎛ ⎞

− − − −⎜ ⎟ ⎜ ⎟⎝ ⎠⎝ ⎠

∑∑∑ ∑∑

∑ bkjp bkjp bkjp

kjp kjp bkjp

F Fw N F

Δ

The effects on foreign wages is simulated by1

ˆFbkjp Dbkjp bkjp

Fbkjp Dbkjp IMMI bkjp

w w Fw w FσΔ Δ Δ

= −

Effects on WagesSimulated doubling of foreign labour force across all skill groups yields the following wage responses

Effects on WagesEven a small degree of imperfect substitutability between native and foreign workers can cause substantial differences in wage responsesImmigration adversely affect wages of foreign workers much more than those of low-skilled ThaisInflows of low-skilled immigrants found to raise the productivity of high-skilled nativesYounger workers are found to suffer greater wage losses than older workers since most immigrants are from the younger groups