Aging, technology and productivity

1
Aging, technology and productivity

• Francesco Daveri
• Università di Parma,
• and IGIER

Mika Maliranta ETLA
BDL/CAED Conference Sept 2005, Cardiff
2
Motivation

• Aging of the workforce at times of fast technical
  change -- a crucial policy issue
• Particularly so at times when fast spreading of
  new technologies makes skills rapidly obsolete
• Yet aging also brings about experience and
  seniority for workers
• Message of growth accounting exercises
• Dont worry, productivity growth will take care
  of the aging problem
• Aging (and education) has a negligible impact on
  productivity growth
• Innovative and absorptive capacity
• Importance of efficiency in innovation, and
  technology implementation in particular
  (Jovanovic 1997)
• Wheres the balance? This is the object of our
  paper

3
Why Finland

• Finland ideal laboratory for the study of such
  issues
• From a substantive point of view
• aging workforce
• IT revolution around the Nokia cluster
• Rapidly changing industry structures old
  economy industries forest and basic metals -gt
  new economy industries, electronics industry in
  particular. We pick four of them (forest, basic
  metals, machequipment, electronics)
• It has good data
• Register data of companies, plants and individuals

4
Basics on aging and productivity

• Three sources of skills accumulation
• Seniority (company/plant specific skills)
• (Potential) experience (general skills)
• Formal schooling (general skills)
• gt Aging increase 1) and 2) but makes 3)
  increasingly obsolete/ineffective (the role of
  employer-provided training?!)
• Twofold effect of aging on productivity
• Positive
• ageing ?experience (learning-by-using
  technology)
• Negative
• ageing skill deterioration
• ageing ability or willingness to adopt new
  technologies (?)
• But picture is more complex

5
Previous studies

• Ideally estimate determinants of individual
  ability, looking for discontinuities before and
  after major tech change -- Yet individual ability
  usually unobserved
• Most studies employ earnings data, at individual,
  plant or company level
• Typical result
• Aging and productivity (earnings) relation is
  concave or follows inverted-U shape
• Shape possibly different across industries -
  evidence of age-related bias of new technology

6
Our empirical strategy (1/2)

• Effects of aging on productivity only imperfectly
  captured by shape of age-earnings relationship
• Older workers may appropriate higher share of
  value added, independently of their sheer
  productivity at work
• Responses of productivity and wages to aging are
  thus possibly different. How?
• Theory of human capital
• Lazears incentive theory

7
Our empirical strategy (2/2)

• Make use of a matched employer-employee data set
• to study differences in productivity by worker
  age by plant data (i.e. grouped individuals)
• to study differences between industries in
  differences by age (differences-in-differences)
• Old technology industries
• Forest industry
• Basic metals
• New technology industries
• Machinery etc.
• Electronics
• Different industries are analyzed separately
• Focus on the years 1996-2002

8
Data sources

• We use Finnish Longitudinal Employer-Employee
  Data (FLEED)
• Consist of
• Employment Statistics Data
• Business Register on Establishments and Companies
• Manufacture Census on plants
• RD survey
• ICT survey
• Representative data compiled from different
  register sources
• Links are not perfect gt non-random samples in
  the analysis?
• Plants employing at least 20 persons are included
  in the analysis
• We focus on the years 1996-200

9
TFP measurement

• Labor characteristics calculated byStatistics
  Finland for all of the workers
• We calculate TFP at the plant level
• TFPexpln(Y/L)-(1-a(it))ln(K/L)
• TFP, Y/L, K/L vary across plants, industries
  years
• Income shares a(it) are industry-year specific
  (but time series are smoothed with a nonlinear
  filter)
• Corrected for self-employment
• Need for using cost shares? (high profitability
  in Electronics industry)
• gt some sensitivity analysis has been done
• Capital measured by the Perpetual Inventory
  Method
• Labor input by hours worked

10
Empirical framework

• Estimate TFP regressions at the plant level
• ln(TFP)pit ?E EXPpit ?T TENpit ?S SCHOOLpit
  ? CONTROLSpit
• ?Y YEARit epit
• p1,2,..,Ni
• iforest, basic metals, mach equipment,
  electronics
• t 1994, .., 2002

11
Empirical framework

• 1st step growth accounting framework, with TFP
  computed as a residual under CRS and perfect
  competition
• 2nd step TFP related to plant characteristics
• Experience of personnel
• Seniority of personnel
• Schooling of personnel
• Yearly dummies
• Plant size (three groups)
• Plant age (plants may go back to the 1970s)
• To reduce simultaneity risk old workers hired in
  old, less productive, plants
• Hiring and separation rates (delayed one period)
• To reduce selectivity risk older workers may be
  the least productive workers stuck in their plant
  of origin

12
Econometric issues

• Descriptive analysis with OLS and Monte Carlo
  simulations
• Flexible specifications (several powers of the
  variables of interest, 3th power for age and
  seniority and 2nd for education )
• Confidence intervals for expected values
  (profiles)
• Useful in assessing the results
• GMM system estimations
• Useful in tackling simultaneity problems
• Problems with more flexible specifications (use
  of linear model)
• Outliers
• Some extreme outliers removed from regressions
  using Hadi (1992, 1994) method
• Variables employed in the removal labor
  productivity and wage in logs, capital intensity
• Importance of controlling other determinants
• Plant age!! (effect on seniority profiles)
• Plant size
• To be done later? Treatment of sample attrition
• Heckman correction

13
OLS analysis schooling, wage and TFP (with
control for plant age and worker turnover)
TFP
Wage
14
OLS analysis potential experience, wage and TFP
(with control for plant age and worker turnover,
without seniority)
TFP
Wage
15
OLS analysis seniority, wage and TFP (with
control for plant age and worker turnover, and
for potential experience) (FIG 3)
TFP
Wage
16
OLS analysis seniority, wage and TFP (WITHOUT
control for plant age and with control for worker
turnover, and for potential experience)
TFP
Wage
17
OLS analysis The estimated effects of plant
cohorts on TFP and wages (seniority, potential
experience and worker turnover included)
18
OLS analysis with Monte Carlo simulations
numerical implications of Figure 3

19
(No Transcript)
20
(No Transcript)
21
(No Transcript)
22
(No Transcript)
23
Very, very preliminary conclusions

• Seniority-based skills are more important in old
  economy industries than in new economy
  industries
• Schooling is more important in new economy
  industries
• The relative wages do not always correspond to
  the relative productivity levels
• Exerts pressure on plant structures (creative
  destruction) within industries
• Productivity-tenure profile is sensitive to the
  inclusion of control for plant age
• Decline starts earlier and is steeper when plant
  age is not controlled
• gt Mobility of older workers from old to
  new/young plants is important