WP 17.2 Age and Productivity
Older workers' productivity and training:
evidence from Europe
ANNA RUZIK-SIERDZINSKA (CASE),
MACIEJ LIS, MONIKA POTOCZNA (IBS),
MICHELE BELLONI, CLAUDIA VILLOSIO (CERP- COLLEGIO CARLO ALBERTO)
NEUJOBS WORKING PAPER NO. 17.2
2nd NEUJOBS Validation Event: Employment 2025 - How Will Multiple
Transitions Affect the European Labor Market?
9-11 April 2014 IZA
WP 17.2 Age and Productivity
Motivation
The literature on changes in ability to perform working tasks with age
concludes that it rises in the first 10 years of working life due to
general education and learning-by-doing and peaks at about 30-35
years old. Then it becomes stable until around 50, when it starts
declining.
The process of declining productivity is rather slow and varies greatly
among workers strongly depending on both personal and job
characteristics (Göbel and Zwick 2009), as well as on the type of tasks
performed and the level of human capital of workers.
Increasing life expectancy, together with low economic growth, has put
pressure on financial equilibrium of many pension systems in Europe
and in other industrialized countries  Increase in the average
retirement age out of the labour force and limited access to early
retirement.
WP 17.2 Age and Productivity
In a context of an extended working life, skills attained at school may
be obsolete for older workers  lifelong learning (LLL) is widely
recognized among the international research and policy community as
a key element to increase or, at least, limit the decline in productivity
of older workers (see, e.g., OECD 2006).
LLL “all learning activity undertaken throughout life, with the aim of
improving knowledge, skills/competences and/or qualifications for
personal, social and/or professional reasons” (European Commission
2001)
UNESCO “Lifelong learning principles, if systematically implemented,
will be able to contribute to more just and equitable societies”.
Scarce empirical evidence on the effect of LLL (training) on older
workers’ productivity.
In this paper, we focus on older workers and analyse the short-term
impact of training undertaken later in life (“older age training”).
WP 17.2 Age and Productivity
Theoretical foundations
According to Human capital theory (Becker 1964,1993), training
activities are investments in human capital that are expected to raise
future productivity. The costs of this investment are lower wages
when trained and direct training costs.
The human capital theory is based on the assumption of a perfectly
competitive labour market, and usually decompose training into
general or specific according to whether acquired skills are portable
or not across firms. The theory also predict who (firms or workers)
would finance and receive the returns of general vs. specific training
However, more recently, new theoretical developments have
challenged this framework and helped explain some aspects at odds
with human capital theory (i.e. firms financing general training, no
evidence of wage cuts during training).
WP 17.2 Age and Productivity
Literature on the effect of training on productivity
1. Direct test through the estimation of a production function.
Requires firms’ data on value added and/or turnover.
• Ichniowski, Shaw and Prennushi (1997), Black and Lynch (2001),
Dearden et al., (2006), Göbel and Zwick (2010), Heywood et al. (2010)
find a positive effect of training
2. Indirect test through the estimation of training effect on workers’
wages
• Positive effect for the US: Lynch (1992) Loewenstein and Spletzer
(1998), Parent (1999), Veum (1995) and Frazis and Loewenstein
(2005)
• Positive effect for the UK: Booth (1991), Booth (1993), Blundell et al.
(1996, 1999), Arulampalam and Booth (2001)
• Positive effect for Norway (Shone 2004), Switzerland (Gerfin 2004),
and Portugal (Budria and Pereira, 2007)
• Unclear effects for Germany (Piske 2001, Mühler et al. 2007,
Kuckulenz and Zwick 2004) and France (Maurin 2000, Fougère et al.
2001)
• Prevalence of positive effects in cross-country analyses (OECD 1999,
Bassanini et al. 2007, Ok and Tergeist 2003)
WP 17.2 Age and Productivity
Our focus on older workers
From theory = i) lower training incidence; ii) lower returns to training
Empirical evidence specifically on older workers is scarce as it is
mostly singled out from general population analyses and often
restricted to age<=55.
The incidence (and return) of training is often found to decrease with
age (e.g. Dostie and Leger 2011, Zwick 2011, Bassanini et al. 2007
Warr and Fay, 2001).
Provided explanations:
•
•
•
Training is less effective for older because it’s not tailored (Gobel and
Zwick 2010, Zwick 2011).
Managers believe that older employees are less able or willing to learn
(Warr and Birdi 1998).
Institutional barriers (e.g. early retirement schemes Fouarge and
Schils 2009)
Our research looks explicitly at the impact of training on older
workers’ wages in an internationally comparable setting.
WP 17.2 Age and Productivity
Data 1/2
We use data from the “Survey of Health, Ageing and Retirement in
Europe” (SHARE). www.share-project.org
SHARE started in 2004 and is a multidisciplinary, cross-national biannual household panel survey. The target population consists of
individuals 50+, plus their spouses or partners irrespective of age.
The questionnaire includes individual and household characteristics,
ranging from physical, mental and psychological health to socio–
economic status, from housing to social support and expectations for
the future.
Four waves of SHARE are currently available (the 3rd is SHARELIFE),
covering 12 (first wave) to 17 (fourth wave) European countries.
The common questionnaire and interview mode, and the
standardization of procedures ensure cross-country comparability
(Börsch-Supan, and Jürges 2005).
WP 17.2 Age and Productivity
Data 2/2
Some data limitations force us to use only wave1 and 2 of SHARE
Training variable in SHARE
Activities in last MONTH: Attended an educational or training course
(wave 1 and wave 2)
Activities in last YEAR: Attended an educational or training course
(wave 4)
Wage variable in SHARE
Hourly net wage (taken home payment, net of tax, national insurance, pension and
health contributions) computed from last taken home payment, frequency
of payments and hours worked in a week
Available ONLY for wave 1 and wave 2
For the robustness checks we used information from SHARELIFE
 We measure the impact of training undertaken in wave 1 on the hourly
wage reported in wave 2 (one to three years later)
WP 17.2 Age and Productivity
Older-age training participation rate by country
Source: SHARE wave 1 and 2
Northern European countries have the highest levels of participation in training at
older ages, followed by the Continental countries. Very low incidence of training at
older ages in Southern European countries.
WP 17.2 Age and Productivity
The empirical strategy
log wit   xit'    itn   it
(1)
where:
log(wit) is the log of hourly wages of individual i at time t
x is a vector of exogenous demographic and job-related individual’s
characteristics (age, gender, education, tenure, sector, occupation, country,
year of interview)
 is a dummy variable equal to one if individual i participated to any training
activity at time t-n
ε is a random term
Endogeneity of

Given the data limitations described in previous slides a fixed effects panel data
model to get rid of individual unobserved heterogeneity (e.g. «ability») cannot be
estimated. Our strategy: control of unobserved heterogeneity by inclusion
of cognitive ability measures
WP 17.2 Age and Productivity
Descriptive 1/2
Equation (1) is estimated on a panel composed by SHARE wave 1 and wave 2
of employees receiving positive wage and working 15-70 hours per week in
11 countries (AT, SE, NL, ES, IT, FR, DE, EL, CH, BE)
0
Log weekly wage
distribution by training
status
.5
1
1.5
Sample of 2312 individuals, 17.8% receiving training
1
2
3
lwage
Untrained workers
4
Trained workers
5
WP 17.2 Age and Productivity
Descriptive 2/2
Training incidence by individual characteristics
Workers' characteristics
Male
Female
Share of trainees
14.5%
21.8%
Private sector
Public sector
16.7%
21.5%
Up to lower Secondary education (ISCED 0-1-2)
Upper and post-Secondary education (ISCED 3-4)
Tertiary education (ISCED 5-6)
6.7%
17.3%
27.4%
Source: SHARE wave 1
WP 17.2 Age and Productivity
Results
VARIABLES
training
log tenure
female
public
age-50
(age-50)^2
No education (ISCED 0)
Primary and lower Secondary
education (ISCED 1-2)
Upper and post-Secondary education
(ISCED 3-4)
(i)
OLS
Dep. Variable= log
weekly wage
0.0648***
(0.0213)
0.0672***
(0.00873)
-0.167***
(0.0163)
-0.0635***
(0.0198)
0.0174***
(0.00567)
-0.00170***
(0.000427)
-0.274***
(0.0554)
-0.223***
(0.0261)
-0.161***
(0.0202)
participation to activities
(ii)
(iii)
Instrumental Variables
Dep. Variable= participation
Dep. Variable= log
into training
weekly wage
0.0919
(0.137)
0.0670***
(0.00870)
-0.168***
(0.0166)
-0.0627***
(0.0200)
0.0176***
(0.00566)
-0.00170***
(0.000423)
-0.273***
(0.0554)
Taking part in training
activities increases
wages of older workers
by up to 6.5 %. This
return is sizable: it is
comparable to having a
upper and postsecondary instead of a
-0.0738***
-0.221***
and lower
(0.025) primary (0.0278)
education
-0.0284 secondary
-0.161***
(0.020)
degree. (0.0203)
0.0333***
0.0074
(0.008)
0.028*
(0.016)
-0.0349*
(0.019)
-0.0042
(0.006)
-0.0001
(0.000)
-0.0586
(0.054)
(0.004)
Observations
R-squared
2,312
0.431
Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1
Estimation includes also sector, occupation, country and year of interview dummies
2,312
0.430
WP 17.2 Age and Productivity
Control of the potential endogeneity of training by
including cognitive ability measures
(a) Inclusion of cognitive abilities and health AT TIME OF
TRAINING
Indicator of health = Self reported health status [1 (poor) to 5 (excellent) scale]
Indicator of cognitive ability = Average of numeracy skills, verbal fluency and ability
to recall a list of words. The indicator ranges 0-10
VARIABLES
Training
Dep. variable= Log Dep. variable= Log
hourly wage
hourly wage
0.065***
(0.021)
2.74***
(0.11)
0.064***
(0.021)
-0.003
(0.008)
0.015*
(0.008)
2.71***
(0.13)
2,312
0.431
2,295
0.431
Health at the time of training
Cognitive abilities at the time of training
Constant
Observations
R-squared
Source: SHARE wave 1 and 2
Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1
Estimation includes also sector, occupation, country and year of interview dummies
Results turn out
to be barely
affected by the
inclusion of these
additional
variables in the
model
WP 17.2 Age and Productivity
Control of endogeneity by cognitive ability measures
(b) Inclusion of EARLY LIFE circumstances (by exploiting
SHARELIFE)
•
Recent evidence of positive association between favorable early life
circumstances and later life cognitive ability (Guven and Lee 2011, dal
Bianco et al. 2013)
•
Havari and Mazzonna 2011 provide evidence about the quality of
retrospective assessments of older individuals in SHARE data
Indicator of early life cognitive abilities= Average of relative position to
others in math and language when ten. The indicator ranges 1-5
CS010 RELATIVE POSITION TO OTHERS MATHEMATICALLY WHEN TEN
CS010a RELATIVE POSITION TO OTHERS LANGUAGE WHEN TEN
Now I would like you to think back to your time in school when you were 10 years old. How
did you perform in Math (Language) compared to other children in your class? Did you perform
much better, better, about the same, worse or much worse than the average?
1. Much better
2. Better
3. About the same
4. Worse
5. Much worse
WP 17.2 Age and Productivity
Control of endogeneity by cognitive ability measures
(b) Inclusion of EARLY LIFE circumstances (by exploiting
SHARELIFE)
•
Recent evidence of positive association between favorable early life
circumstances and later life cognitive ability (Guven and Lee 2011, dal
Bianco et al. 2013)
•
Havari and Mazzonna 2011 provide evidence about the quality of
retrospective assessments of older individuals in SHARE data
Indicator of parents’ cultural background = Enough books in the
parental home to fill one bookcase (>26 books) [Flores and Kalwij 2013]
CS008 NUMBER OF BOOKS WHEN TEN
Approximately how many books were there in the place you lived in when you were 10? Do not
count magazines, newspapers, or your school books.
1. None or very few (0-10 books)
2. Enough to fill one shelf (11-25 books)
3. Enough to fill one bookcase (26-100 books)
4. Enough to fill two bookcases (101-200 books)
5. Enough to fill two or more bookcases (more than 200 books)
WP 17.2 Age and Productivity
Results of the inclusion of EARLY LIFE circumstances on
returns to training
Dependent variable= Log hourly wage
VARIABLES
Training
0.065***
(0.021)
0.063***
(0.024)
0.023*
(0.012)
2.74***
(0.11)
2.73***
(0.13)
0.058**
(0.024)
0.017
(0.012)
0.084***
(0.019)
2.65***
(0.13)
2,312
0.431
1,755
0.466
1,750
0.472
Early life cognitive ability
Enough books in parental home (bookcase)
Constant
Observations
R-squared
Source: SHARE wave 1, 2 and 3 (SHARELIFE)
Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1
Estimation includes also sector, occupation, country and year of interview dummies
WP 17.2 Age and Productivity
Further robustness checks
Trained and untrained workers do not appear to be
different in the wage earned before training
Results hold also when restricted the analysis to those workers
undertaking training for work-related purposes
WP 17.2 Age and Productivity
Countries heterogeneity
Is impact of training on older workers’ wage different
across countries?
Returns to training are found to be higher in Continental and
Southern than in Northern European countries
(Higher training incidence for Northern countries, followed by
Continental and by Southern countries)
Our results confirm the existence of a negative association
between incidence and returns to training previously found in other
studies which do not explicitly focus on older workers (see, e.g.,
Bassanini et.al., 2007)
WP 17.2 Age and Productivity
Conclusion
Growing consensus on the importance of training investments for keeping workers’ skills
updated and offsetting the productivity decline which is combined with the ageing process.
From our study: taking training increases wages in the short-term by about 5.8% - 6.5%
for employees 50+ residing in one the eleven analyzed European countries (Robust).
This return is comparable to moving from primary and lower secondary to upper and
post-secondary education and comparable to what already found in literature.
Training incidence is higher in Northern EU countries, returns to training are higher in
Central and Southern EU countries
Longer working lives also means more profitable investments in adult training and
learning for firms and workers.
In addition:
 (a) wage returns may underestimate positive effect of training
 (b) training may be beneficial also on other measures of performance of older workers:
employability, early retirement, job security ….
Avoiding older workers’ skills obsolescence through training may lead to higher
productivity and lower unemployment, and increase economic growth.
WP 17.2 Age and Productivity
WP 17.2 Age and Productivity
WP 17.2 Age and Productivity
Further robustness checks 1/2
Estimation of the wage equation on wage pre-training
(i)
Log wage after training
(ii)
Log wage before training
Training
0.065***
(0.021)
0.043
-0.041
Constant
2.74***
(0.11)
2.49***
-0.24
2,312
0.431
2,021
0.174
Dependent variable:
Observations
R-squared
Source: SHARE wave 1 and 2
Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1
Estimation includes also sector, occupation, country and year of interview dummies
WP 17.2 Age and Productivity
Further robustness checks 2/2
MOTIVATION to training:
1.
2.
3.
4.
5.
6.
7.
8.
To meet other people
To contribute something useful
For personal achievement
Because I am needed
To earn money
Because I enjoy it
To use my skills or to keep fit
Because I feel obligated to do it
(80% of training obs.)
(i)
Log wage after training
(ii)
Log wage on the subsample
Training
0.065***
(0.021)
0.061***
-0.023
Constant
2.74***
(0.11)
2.73***
-0.11
2,312
0.431
2,244
0.435
Dependent variable:
Observations
R-squared
Source: SHARE wave 1 and 2
Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1
Estimation includes also sector, occupation, country and year of interview dummies