Growth in Euro Area Labour Quality

1
Growth in Euro Area Labour Quality
Guido Schwerdt (European University
Institute) and Jarkko Turunen (ECB)
OECD Workshop on Productivity Analysis and
Measurement Bern, 17 October 2006
2
Motivation

• Composition of euro area work force changes over
  time
• Share of workers with higher education tends to
  increase
• Workers with different education levels, work
  experience and skills move in an out of
  employment
• Raw measures of labour input such as total hours
  worked or employment provide biased measures of
  actual labour input
• Adjusting for labour quality is important for
  understanding sources of labour productivity
  growth and fluctuations in labour input over the
  business cycle
• Evidence suggests that positive labour quality
  growth contributes significantly to growth in
  labour productivity (Jorgensen, 2004)

3
Significant increase in the share of workers with
university level education in the euro area
Figure 1. Total hours worked by education
(percentages)
Sources Labour Force Survey and authors
calculations. Note Low refers to those with
lower secondary education or less, medium to
those with upper secondary education and high to
those with tertiary education.
4
Related questions

• Has there been a shift in the composition towards
  workers with lower skills in the 1990s?
• What drives sustained decline in euro area labour
  productivity growth?
• How should economic policies be designed to
  further improve knowledge and innovation
  (mid-term review of the Lisbon agenda)

5
Outline

• Literature
• How do we measure labour quality?
• Main results
• Index of labour quality for the euro area and
  some euro area countries
• Robustness to alternative assumptions
• Changes in composition over the business cycle?
• Decomposition of labour productivity growth the
  case of disappearing TFP growth!

6
Literature

• Studies on the US provide methodological
  background
• Bureau of Labor Statistics (BLS) (1993), Ho and
  Jorgenson (1999), Aaronson and Sullivan (2001)
• Few studies on euro area countries
• Jorgenson (2004) for France, Germany and Italy
• Brandolini and Cipollone (2001) for Italy, Card
  and Freeman (2004) for Germany, Melka and Nayman
  (2004) for France
• OMahony and Van Ark (2004) provide sectoral
  evidence for France, Germany and the Netherlands
• No evidence for the euro area as a whole, country
  evidence too scattered to draw firm euro area
  conclusions

7
Measuring labour quality an overview

• Task construct an estimate of labour quality
  adjusted labour input in the euro area
• Combining information from microdata of
  individuals with official aggregate data
• Step 1 Take data of individual wages and
  personal characteristics from the European
  Community Household Panel (ECHP) and construct
  weights for worker groups (by age, sex, education
  and country) using wage regressions
• Step 2 Combine weights with data of hours worked
  for each worker group from the European Labour
  Force Survey (LFS)
• Alternative estimates using entirely microdata
  based regression method in Aaronson and Sullivan
  (2001)

8
Method, step 1 Weights

• Dependent variable is individual real hourly wage
• Regressors are dummy variables for age and
  education, equation is estimated using weighted
  OLS separately for males and females and for each
  country (30 times 12 worker-country groups)
• Predicted wages for each worker-country group are
  used to construct weights (the share of worker
  group i of total labour compensation)

9
Method, step 2 Quality index

• Growth in total labour input is constructed as
  the weighted growth in total hours for worker
  group i

• Growth in labour quality is defined as the
  difference between growth in total labour input
  and unweighted growth in hours worked

10
Data

• Micro data from ECHP
• Longitudinal information on wages and other
  individual characteristics (e.g. education, age,
  gender)
• All euro area countries for the 1994-2001 time
  period
• Aggregate data from European LFS
• Hours worked and employment for worker groups
  cross-classified by education, age, gender and
  country (also sectors and full/part-time status)
• All euro area countries for (currently) the
  1983-2004 time period
• Note Breakdown by education only available from
  1992 onwards additional information from the
  Luxembourg Income Study (LIS) and the German
  Socio-Economic Panel (GSOEP) is used

11
Caveats

• Common to all studies of labour quality
• Individual wages are assumed to accurately
  reflect productivity differences union
  bargaining, search frictions, discrimination etc.
  suggest that this assumptions is likely to be
  violated
• Proxies for measuring composition are imperfect
  e.g. work experience is inaccurately proxied by
  age, no measure of quality of education exists
• Specific to our study
• Data for detailed classification pre-1992 is
  partly intrapolated
• Assume fixed weights, i.e. that returns to
  individual characteristics do not change over
  time evidence for Europe suggest that relative
  wages are rigid (e.g. Brunello and Lauer, 2004)
• Measurement error in survey data?

12
Continuous increase in euro area labour quality
Figure 2. Labour quality growth (index 1983 100)
Sources Authors calculations.
13
Changes in quality growth over time and
significant impact on quality adjusted labour
input
Figure 3. Labour quality adjusted labour
input (average annual growth rate)
Sources Authors calculations.
14
Validation comparing country results with
existing estimates
Figure 4. Comparing country estimates (average
annual growth rate)
Sources Authors calculations and Jorgensen
(2004).
15
Robustness

• Validation using country estimates supports
  robustness of our calculation for the euro area
• Alternative data and methods available for a
  shorter time period
• Robust to including additional determinants of
  labour quality (sector and part/full-time
  status) 1993-2004 average annual growth goes
  from 0.61 to 0.65
• Robust to accounting for changing weights using
  the regression approach 1995-2001 average annual
  growth goes from 0.47 to 0.44

16
Changes in composition over the business cycle
are expected to be countercyclical

• Previous evidence suggests that labour quality
  growth is likely to be counter-cyclical (Aaronson
  and Sullivan, 2001 and Solon et al. 1994)
• Down-skilling in upturns as the share of
  workers with lower skills increases firms lower
  skill requirements, and increased likelihood of
  finding a job and possibly higher wages encourage
  lower skilled workers to enter the labour market
• Up-skilling in downturns (in reverse)

17
Confounding factors and data weaknesses

• Business cycle effects may be confounded by the
  impact of changing trends
• Labour market reforms in the late 1990s may have
  resulted in increased participation of lower
  skilled workers
• Demographics ageing of the baby boom generation
• Measure of skills may not be accurate enough to
  fully capture cyclical effects
• Unobserved characteristics (e.g. motivation)
  matter, but are likely to be correlated with
  observables

18
Some evidence of lagged countercyclicality in
euro area labour quality
Figure 5. Trend/cycle decomposition (log levels
and deviations from trend)
Sources Authors calculations. Note The trend
and cycle have been extracted using a band-pass
filter (with cycle length between 2 and 8 years).
19
Education and work experience the main
determinants of growth in human capital
Figure 6. Main determinants of labour quality
growth (annual growth rates)
Sources Authors calculations.
20
Implications for measuring total factor
productivity

• Best practice in productivity measurement
  suggests taking into account labour quality
  adjustment (OECD, 2001)
• Decomposing labour productivity (measured per
  hours worked) growth into
• Capital deepening (growth in capital services per
  hours worked)
• Labour quality growth
• TFP growth (residual)
• Previous decompositions of euro area labour
  productivity have not considered quality
  adjustment, thus overestimating TFP growth (e.g.
  ECB, 2004, Vijselaar and Albers, 2004)

21
The case of the disappearing TFP growth!
Figure 7. Decomposition of labour productivity
growth (averages of annual growth)
Sources ECB calculations. Except for the
estimate of labour quality growth, data are from
the Groningen Growth and Development Centre.
22
Other (possible) applications

• Done
• Quality-adjusted measure of wage growth lower
  growth in quality adjusted real wages, weak
  cyclicality
• Quality of the available labour force quality
  growth of the unemployed higher than for the
  employed in the late 1990s
• Further research
• Forecasts of labour quality growth looking
  forward, population ageing may lower the
  contribution of human capital to growth
• Reconsider previous work using TFP estimates for
  the euro area

23
Summary and conclusions

• A robust estimate of labour quality growth in the
  euro area
• Positive labour quality growth in the past 20
  years
• Some (weak) evidence of changes in composition
  over the business cycle, especially in the 1990s
  down-skilling in upturns and up-skilling in
  downturns
• Implications for productivity growth
• Approximately 1/3 of labour productivity growth
  due to improvements in labour quality
• Accounting for labour quality lowers estimates of
  euro area TFP growth