Skills and the labour market Research ideas Contributors: CEPII, CBS, University of Groningen, WIIW, WIFO, IVEI.

Skills and the labour marketResearch ideas

Contributors: CEPII, CBS, University of Groningen, WIIW,

WIFO, IVEI

Potential research ideas • capital-skill complementarity (CBS, University Groningen,

WIIW)• skill-biased technology change (e.g. through information

technology capital and R&D capital) (NIESR, WIIW, University of Groningen, IVIE)– Skill premium and technological change– Relative skill/unskilled employment and technological change– Impact of ICT capital on highly skilled workers, low skilled workers

and intermediate skill workers (job polarisation)• changes in supply of skills and technological change (IVEI,

University of Groningen)• role of industry structure (WIIW)• impact of trade and outsourcing (WIFO)• demand of labour for specific types of workers (WIIW)

– older and younger workers (“age biased technological change”)– specific occupations such as engineers and technicians

Potential research ideas • relationship between output growth and employment evolution by

skill level and age groups

• Impact of overeducation on wages (NIESR)

• Product market structure and labour market institutions (NIESR)– joint effects on productivity, employment and wages

– Effects on returns to skills and investment in human capital.

• Other determinants of heterogeneous labour demand

• Determinants of labour demand at the industry level (ln H, = f(ln W/P, Y, K, R&D))

Potential research ideas • Novelty: consistent and uniform data set covering up N=72

industries, T=30, J countries

• Econometric methods; issues to address

– SUR

– adjustment costs, dynamic panel data methods

– T>N = > single equation: dynamic heterogeneous panel data method, Pesaran, Shin, Error correction model applied to panel data

– N>>T =>dynamic panel data model (first difference GMM, system GMM)

– Heterogeneity across industries

– Endogeneity of wages

Role of industry structure in skill upgrading • size of low skill industries is decreasing (=> between shifts effect)• within industry shifts• shift-share analysis; decomposition of the change in the

employment share of highly skilled/unskilled workers:– contribution to the total change which results from employment shifts

between sectors of different skill intensities. – contribution to the total change which results from the shift towards skilled

workers within a sector• Extension: occupational structure• Aim:

– within and between contributions change over time– differences between manufacturing and non-manufacturing– Between effect differ across EU countries

capital-skill complementarity

• Krusell et al. (2000): – capital-embodied technological change alone can account for most of

the variations in the skill premium

– use of quality-adjusted prices for a number of durable equipment categories such as office and computing equipment including peripheral equipment and accounting machinery (OCAM), communication equipment, general industrial equipment and

transportation equipment. • Relative employment equation, LS/LU=f(WS/WU, C, IT-C,

R&D)

• System of labour demand equations

• Measurement: (i) CSC may differ across industries, (ii) structural breaks over time, (iii) robust to measurement of capital (fixed/variable factor, quality adjustment) and skills

Skill-biased technological changeBerman, Bound and Griliches (1994):

– change in the cost share of non-production workers is positively related to the industry's initial ratio of investment in computers to total investment.

– one third of the change in the non-production wage bill share can be explained by the computer variable

•

• DSN: change in non-production wage bill share• DlnK/Y: change in capital-output ratio• R/Y: R&D intensity• CI/I: ratio of investment in computers to total investment • Pn/Pj relative wages between non-production and production workers

0 1 2 3 4ln( / ) ln( / ) ( / ) ( / )N n jS P P K Y CI I R Y e

Impact of outsourcing and trade

• Measure of outsourcing: intermediate goods imports from the same industry)

• Impact of exports (exports generated by imported materials) on the demand for heterogeneous labour

• Combing trade statistics and EUKLEMS data

Product market structure and labour market institutions: effects on returns to skills and

investment in human capital

• Link between PMC and wage inequality

• data

Previous literature • Morrison-Paul Siegel (2001): US manufacturing, 73-89, four types of labour; results: high-tech CSC is

significant (explains 78% of the increase of LH)

• O’Mahony, Robinson, Vecci (2004): US, UK, F, D, 1970-2000, three/four types of labour: IT is the

major factor

• Chun (2000): US all industries, 1960-1995; IT CSC is significant, explains 25% of the increase of LH

• Fitzenberger (1999) Germany, non-manufacturing, measure: inolut coefficients of the

computer/electrical industry; not significant

• Other evidence: Krusell et al. (2000), Machin and Van Reenen (1998), Green, Felstead and Gallie

(2000), Riley and Young (1999), Green, Felstead and Gallie (2000), Hansson (2000), Mellander (2000)

for Sweden, Lindquist and Skjerpen (2000) for Norway, Strauss-Kahn (2003) , Goux and Maurin (2000)

for France

• Theoretical lit: Caselli (1999)

• Summary. Studies agree on IT CSC but differ with respect to the magnitude of the impact

Empirical model I

• Labour demand model for each skill group

– Lnit: total annual hours of highly, medium and unskilled workers

– Yit: value added in constant prices

– WPnit: hourly wage deflated by the value added deflator

– lnpitt: price index of information equipment and software

– YRS: Average years of schooling in the working age population

• Estimation method: Fixed effects model

.lnlnlnln 54321 nitinntntnnitnitnnnit tßYRSßPITßWPßYßßL

Empirical model II• System of factor demand equations derived from a flexible

cost function

• Generalised box-Cox cost function

• variable inputs: xnt = (xhnt, xsnt, xunt, xmnt)

• input prices as pnt = (phnt, psnt, punt, pmnt)

21

' , ; 1, , , 2' , ;exp

P ZCp x nt nt np z a nt nC nt nt nP Zp x C nt nt nnt n

0

0

2

2

for

for

' ' '0

1 1, ;

2 2Cn pn nt z nt nt pp nt nt pz nt nt zz ntnt nt nC A P A Z P A P P A Z Z A ZP Z

)

Empirical model II• inputs

– xhnt annual working hours of highly skilled workers (workers with a university degree/univ. entrance degree

– xsnt annual working hours of workers with apprenticeship training– xunt annual working hours of workers with compulsory

school – xmnt total materials in constant prices

• Input prices (normalized to 1 in 1980)– phnt hourly wages of highly skilled workers– psnt hourly wages of medium skilled workers– pumt hourly wages of low skilled workers– phnt price index total material inputs

• fixed factors and total variable costs– ynt gross output in constant prices– knt net capital stock in constant prices– t time trend, alternatively:

– (i) average years of schooling of the working age population – (ii) price index of information equipment and software

• c sum of labour costs and total materials)

Empirical model II

The components of Z, P:

0

0

ln 1

11

for

for

z

zZ

jnt

jntjnt j=k, y.

0

0

ln

)'/(

1

11

1

for

for

P

PP

Pjnt

njntjnt

jnt j = h, s, u, m.

)

This system of four input demands is derived by the application of Shepard's lemma.

ntntntntnt ezpxx /),,,(*

Empirical model II

)

• Estimation method: non-linear SUR with fixed effects.

• number of parameters: 24 due, two Box-Cox parameters, 4 x 21 industry dummies

• estimation problems – non-stationarity of the data

– A second problem is the potential endogeneity of wages

• Elasticities of factor demand:– Two inputs are substitutes (complements) if the cross-

price elasticity is significantly positive (negative).

.,,,,,*

*

ushmjii

p

p

i j

jip j

Empirical model II

)

• impact of the net capital stock:

• Complementarity/substitutability :

• output elasticities:

• The time "elasticities'':

• elasticities of the different labour inputs with respect to the price of information processing equipment and software:

• The impact of the supply of skilled workers:

0/0 ikik

.,,,,*

*

ushmii

k

k

iik

.,,,,*

*

ushmii

y

y

iiy

.,,,,1*

*

ushmiit

iit

.,,,,*

*

ushmii

pit

pit

iipit

.,,,1*

*

ushiiyrs

iipit

Hypotheses

)

• Hypothesis 1: Capital-skill complementarity is found if capital and skilled labour are complements while capital and unskilled labour are substitutes.

• Hypothesis 2: Technological change measured as the price index of information equipment and software favour higher skills and reduces the demand for low-skilled workers.

• Hypothesis 3: Own-wage elasticities in absolute values decrease with the skill level

• Hypothesis 4: Substitution possibilities between different labour inputs are higher than between labour and non-labour inputs.

• Hypothesis 5: Unskilled workers can be substituted more easily for materials than both medium-skilled workers and highly skilled workers.

Data and summary statistics

)

• annual two-digit industry data for Austrian industries for the period 1980-2003

• data sources: National Accounts, calculations based on micro census, wage and salary statistics

• Annual hours worked: number of employees X actual hours per employee x 52 working weeks

• Capital stock: PIM • NIPA: price index of Information Processing

Equipment

Data and summary statistics

)

mean st.dev max min working hours of highly skilled workers, h 4.1 3.1 12.2 -0.4 working hours of medium skilled workers, s 1.4 4.7 14.9 -3.6 working hours of unskilled workers, u -3.0 4.0 6.0 -8.8 total materials, m 3.3 3.1 14.7 -0.5 net capital stock 1.5 1.9 4.3 -1.7 hourly wage index of highly skilled workers, ph 3.9 0.6 4.8 1.8 hourly wage index of medium skilled workers, ps 4.1 0.3 5.0 3.8 hourly wage index of unskilled workers, pu 3.9 0.3 4.8 3.2 price index of total materials 1.9 0.7 3.1 0.6 price index of gross output 2.1 1.1 4.4 0.1 gross output constant prices 2.6 1.9 7.3 -1.4 average years of schooling of working age population 0.4 0.4 0.4 price index information equipment and software -3.8 -3.8 -3.8

Annual percentage changes in inputs, output, wages and prices, 1980-2003

summary statistics – evolution of quantities

)

manufacturing

0.00

0.50

1.00

1.50

2.00

1980

1981

1982

1983

1984

1985

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

highly skilled labourmedium-skilled labourunskilled labourtotal materialsnet capital stockgross output c.p

summary statistics – evolution of quantities

)

non-manufacturing

0.00

0.50

1.00

1.50

2.00

2.50

3.00

3.50

1980

1982

1984

1986

1988

1990

1992

1994

1996

1998

2000

2002

highly skilled labourmedium-skilled labourunskilled labourtotal materialsnet capital stock

summary statistics – evolution of prices

)

manufacturing

0.0

0.5

1.0

1.5

2.0

2.5

3.0

1980

1982

1984

1986

1988

1990

1992

1994

1996

1998

2000

2002

price index of highly skilled labour

price index of medium-skilled labour

price index of unskilled labour

price index of total materials

price index of gross output

price index ofinformation processing equipment & software

summary statistics – evolution of prices

)

non-manufacturing

0.00

0.50

1.00

1.50

2.00

2.50

3.00

1980

1982

1984

1986

1988

1990

1992

1994

1996

1998

2000

2002

price index of highly skilled labour

price index of medium-skilled labour

price index of unskilled labour

price index of total materials

price index of gross output

price index information processingequipment & software

Empirical results II

)

Results based on the factor demand system

• IPES investment reduce the demand for both unskilled workers and medium-skilled workers; with a large magnitude

• IPES have a positive but small impact on highly-skilled workers

• positive impact of net capital stock on all types of labour; highest impact on highly skilled workers

• demand for unskilled workers is quite elastic to changes in wages.

• zero substitutability relationship between skilled and unskilled workers

• substitutability relationship between unskilled workers and material inputs

• output elasticity of unskilled workers: 1.12

• time elasticities: significantly positive for university graduates and significantly negative for unskilled labour

• years schooling: positive impact on highly-skilled labour, negative on medium and unskilled workers

Empirical results II

)

*( , )i jx p *hx *

sx *ux *

mx

price elasticities ph 0.85 0.02 -0.81 -0.01 (2.24) (0.11) (-3.06) (-0.69) ps 0.02 0.23 0.16 -0.10 (0.16) (1.45) (0.80) (-7.80) pu -0.67 0.06 -0.78 0.12 (-2.68) (0.23) (-3.39) (9.07) pm -0.12 -0.44 1.62 -0.02 (-0.80) (-7.39) (7.65) (-0.91)

other elasticities output, zy -0.12 -0.19 1.12 1.24 (-0.64) (-1.44) (3.62) (50.34) net capital stock, zk 2.12 0.37 0.85 0.06 (6.66) (2.46) (4.59) (1.84) time trend z 0.016 0.005 -0.073 -0.001 (2.46) (0.63) (-3.94) (-0.42) price index IT equipment and -0.05 0.49 1.38 softeware (-0.50) (5.59) (9.59) av. years of t 0.40 -0.14 -2.25 schooling (2.09) (-1.99) (-13.17)

Empirical results II

)

Results based on the standard labour equations (estimated separately)

• Impact of IPES on the demand for unskilled workers is negative and highly significant; magnitude is large

• Robust when time trend and years of schooling are included

• Very small impact of IPES on medium and highly skilled workers

• Again: own-wage elasticities (absolute values) highest for unskilled workers, output elasticities increase the lower the skill level is

• Years of schooling also contribute the growing skill shares the industry level

Empirical results I

)

ln (wu/pva) -1.40 *** -1.41 *** *** -1.40 ***-9.72 -9.82 -9.33

ln VA 0.82 *** 0.84 *** *** 0.85 ***7.07 7.41 7.50

trend 0.059 **2.04

yrschool -0.55 *** -1.25 ***-4.93 -2.57

ln pit 0.50 *** 0.80 ***4.82 2.66

constant 10.99 *** 5.26 *** *** 18.10 ***9.41 323.40 3.640.91 0.91 0.92

Single equation estimates: fixed effects results,

dep. Var: log (annual hours unskilled worker)

Nobs: 504; N=21

Empirical results I

)

Single equation estimates: fixed effects results,

dep. Var: log (annual hours medium-skilled worker)

Nobs: 504; N=21

ln (ws/pva) -0.79 *** -0.68 *** -0.76 ***-6.87 -5.53 -6.19

ln VA 0.36 *** 0.43 *** 0.39 ***3.90 4.75 4.29

trend -0.02-0.85

yrschool 0.38 ** 0.95 *3.39 1.90

ln pit -0.20 ** -0.03-2.01 -0.11

constant 2.26 * 6.20 *** -3.641.94 348.29 -0.71

Adj-R2 0.88 0.88 0.88

Empirical results I

)

Single equation estimates: fixed effects results,

dep. Var: log (annual hours highly skilled worker)

Nobs: 504; N=21

ln (wh/pva) -0.65 *** -0.57 *** -0.69 ***-5.50 -4.67 -5.83

ln VA 0.35 *** 0.39 *** 0.30 ***2.94 3.68 2.60

trend 0.010.52

yrschool 1.15 *** 0.6311.71 1.29

ln pit -0.93 *** -0.18-11.33 -0.66

constant -6.65 *** 5.41 *** -1.26-6.43 292.44 -0.25

Adj-R2 0.93 0.93 0.93

conclusions

)

• price decrease of IT equipment tends to reduce the demand for low-skilled workers; but little impact on the two upper skill level

• traditional CSC is the major factor explaining the demand for highly skilled workers

• All types of skills benefit from an increase in capital• results are robust when supply side effects are taken into account• demand for unskilled labour is more wage-elastic than the

demand for medium-skilled labour.• zero substitutability between different types of labour• material inputs are a substitute for unskilled labour