Ageing Workforce,
Productivity and Labour costs
of Belgian Firms
Vandenberghe, Vincent (IRES- UCL)
Waltenberg, Fabio (CEDE, Universidade Federal
Fluminense)
Université
Catholique de
Louvain
ZEW seminar, Mannheim
June 15, 2010
Presentation outline
1. Motivation
2. Existing literature
3. Methodology
4. Data
5. Results and conclusions
2
1. Context, motivation
• Policy and scientific context
- Ageing population, political initiatives to increase older
empl. rates but (very) low employment in some EU
countries (e.g. Belgium, France, Luxembourg)
- Existing literature looks mainly at…
• the consequences of an ageing population, in terms of
welfare cost or growth (Gruber and Wise, 2004)
• the retirement behaviour of older individuals (replacement
rates, pension, early-retirement schemes, role of health, jointdecision within households…) (Mitchell & Fields, 1983)
<=> supply side
- Not so much the determinants of the labour demand by
firms (e.g. labour costs, productivity...)
<=> demand side
- Despite country-level evidence suggesting that it could
matter
3
1. Context, motivation (cont)
Belgium
4
1. Context, motivation (cont)
Belgium
5
1. Context, motivation (cont)
proc glm data=silc.corr;
model emplg= rwage rp /solution;
run;
Standard
Parameter
Intercept
rwage
rp
Estimate
0.20
-.58
0.17
Error
0.16800741
0.17990096
0.22314542
t Value
Pr > |t|
1.23
-3.26
0.76
0.2312
0.0038
0.4560
6
1. Context, motivation (cont.)
• Our main motivation here is to answer two
questions
- Do ageing workforces negatively affect productivity
performance of firms? [growth/ GDP]
- Are employers willing to (re)employ older workers?
[Employment rate]
=> Key assumption: a sizeable negative productivityvs. labour costs gap is likely to adversely affect the
labour demand for older workers
7
2. Existing literature on age,
productivity (and labour costs)
-
Individual-level data
“Individual job performance is found to decrease from around
50 years of age, which contrasts with life-long increases
in wages. Productivity reductions at older ages are
particularly strong for work tasks where problem
solving, learning and speed are needed, while in
jobs where experience and verbal abilities are
important, older individuals’ maintain a relatively high
productivity level.” (Skirbekk, 2004: SURVEY)
8
2. Existing literature (cont.)
-
Country-level data
“(…) large macro-data panel (…) explores the impact of the
age composition of the labour force on levels and growth
rates of output per worker as well as on total factor
productivity (TFP). The results point to an inversely Ushaped relationship between the share of workers in
different age groups (...) the impact of projected
ageing of the labour force on productivity and percapita growth could be really substantial in some cases”
(Werding, 2007)
9
- Firm-level data***
• Hellerstein et al. (1999) [USA]: wages and productivity tend to
grow with age, but no significant gap.
• Malmberg, Lindh, & Halvarsson (2006) [Sweden]: an
accumulation of high shares of older adults in Swedish
manufacturing plants does not seem to have a negative effect on
plant-level productivity
• Gründ & Westergård-Nielsen (2008) [DK]: find that mean
age (and age dispersion) in Danish firms are inversely u-shaped
related to firm productivity
• Skirbekk, (2008) [International survey]: The most common finding
from these studies is a hump-shaped relation between job
performance and age. Of the 14 studies considered, 11 find a
productivity decline in the 50s relative to the 30s and 40s, two have
inconsistent results, while one finds that productivity peaks among
10
the oldest workers.
Aubert & Crépont (2003), Economie & Statistiques
[France], productivity rises with age until around the age of 40,
before stabilizing, a path which is very similar to those of wages.
A wage-productivity gap is observed only for workers aged more
than 55
Dostie (2006), IZA [Canada]
obtains concave (inversely U-shaped) age-productivity profiles.
Significant wage-productivity gap occurs with educated males
aged 55 +
Ilmakunnas & Mliranta(2007) [Finland]. Older workers
separations are correlated with higher productivity, lower cost=>
higher profits
Göbel, Ch. and Zwick, Th. (2009) [Germany] find that
productivity increases with the share of employees until the age
of 50-55 and only decreases slightly afterwards
van Ours, J.C & Stoeldraijer, L. (2010), [Netherlands]
find little evidence of an age related pay-productivity gap
11
3. Methodology
Equ.1: productivity
log Yit = α log LitA +ß logKit +γ Fit + it
where: Yit is the firm’ value added
and LitA a “labour quality index” à-la-Hellerstein
LitA = ∑k λk Likt = µref Lit + ∑k (µk - µref) Likt
µk being the productivity of type (e.g. age) k workers
12
Assuming
k=0  18-29
k=1  30-49
k=2  50-65
[ref]
log Yit ≡ yit = A + α li,t + η0 Pi0t + η2 Pi2t+ß kit +γFit +  it
with
Pi2t= Li2t/Li1t
η2 = α (λ2 – 1) and λ2= µ2/µ1;
13
Equ.2: labour costs
ln LCit = ln π1 + ln Lit + 0Pi0t + 2Pi2t + it
with 2 ≡ Φ2-1= π2/ π1 -1
π being the relative labour cost of the
considered type of workers
Key question
λ2= ??? Φ2
λ2= relative productivity of 50-65
Φ2= relative labour cost of 50-65
14
Identification challenge
yit = A + α li,t + η0 Pi0t + η2 Pi2t+ß kit +γFit +  it

it
= δi + ωit + εit
δi unobservable (time-invariant) heterogeneity between firms
ωit short-term (asymmetrically) observed productivity
shocks, ωit
εit random error

E(εit) = 0
15
Production/productivity (cont.)
We report the results of several estimations methods: OLS, firstdifference, within (fixed-effect), System GMM à-la BlundellBond
Our preferred approach = proxying the short-term productivity
shocks ωit using with demand for intermediate inputs
(Levinshon & Petrin, 2003)
intit =I(ωit , kit)
[5]
Assuming this function can be inverted the residual  it becomes
δii + ωit(intit) + εit
[6]
with ωit(intit) that can be approximated by a polynomial expansion
in int.
16
4. Our Data
•
Employers-employees matched data
–
~10.000 firms with 20+ workers (BELFIRST- BNB)
–
using firm identifiers, we are able to inject information
from banque Carrefour de la sécurité sociale on the age
of (all) workers employed by these firms: ~1.200.000
workers
–
…..we do not need to assign workers to firms using
matching methods like in Hellerstein et al. (1999)
•
Data aggregated at firm level
•
Long Panel 1998-2006 (9 years)
17
• Information on firms from the (now dominant) service
sector, where administrative and intellectual work is
predominant
• Like Aubert & Crépon (2003) and Dostie (2006), we have
a measure of firms’ productivity (the net valued added),
which is measured independently from firms’ wage cost
• Contrary to Dostie (2006), we do have a measure of
firms’ capital stock, such that no imputation method is
required.
18
Mean age and value added per employee
Belgium
0
1000
VA_L
2000
3000
4000
Denmark
20
30
40
50
60
70
magey
June 09
19
5. Results
20
Estimating age differencials. Calculating the produtivity/labour cost gap
lnY; Y being value added (productivity) or labour cost
Method:
1-OLS
2-Within (firm fixed
effects)
3-First Differences
4-Intermediate inputs 5-Lagged IV (system
(Levinsohn-Petrin) GMM Blundell-Bond)
6-Within ( firm fixed
effects+ intermediate
inputs LP)
Share of 18-29 workers
-0,324
0,009
0,078
-0,334
0,233
0,022
p-value
0,0000
0,5134
0,000
0,0000
0,0129
0,2043
Share of 50-65 workers
-0,253
-0,293
-0,164
-0,295
-0,287
-0,321
p-value
0,0000
0,0000
0,000
0,0000
0,0066
0,0000
Controls
capital, number of
employees + fixed
effects: year, nace1,
region
76.512
capital, number of
employees + fixed
effects: firm, year
capital, number of
employees + fixed
effects: year, nace1,
region
66.615
capital, number of
employees + fixed
effects: year, nace1,
region
61.975
capital, number of
employees + fixed
effects: year, nace1,
region
59.971
capital, number of
employees + fixed
effects: firm, year
Productivity equation
η2
Nobs.
76.512
61.975
Labour cost equation
2
Share of 18-29 workers
-0,450
-0,122
-0,108
-0,491
0,088
-0,118
p-value
0,0000
0,0000
0,0000
0,0000
0,1830
0,0000
Share of 50-65 workers
-0,191
-0,012
-0,020
-0,202
-0,169
-0,0085
p-value
0,0000
0,3559
0,2039
0,0000
0,0247
0,5999
Controls
fixed effects: year,
nace1, region
77.696
fixed effects: firm, year
fixed effects: year,
nace1, region
66.615
fixed effects: year,
nace1, region
61.973
fixed effects: year,
nace1, region
60.713
fixed effects: firm, year
Nobs.
77.696
61.973
Productivity vs labour cost differentials
productivity (λ ) 18-29
0,63
1,01
1,16
0,62
1,26
1,03
Labour cost (Φ ) 18-29
0,55
0,88
0,89
0,51
1,09
Gap (λ-Φ ) 18-29
0,08
0,14
0,27
0,11
0,18
0,88
0,15
productivity (λ ) 50-65
0,71
0,57
0,66
0,66
0,67
0,55
Labour cost (Φ ) 50-65
0,81
0,99
0,98
0,80
0,83
Gap (λ-Φ ) 50-65
-0,10
-0,41
-0,32
-0,14
-0,16
1,01
-0,47
21
η2 = α (λ2 – 1); λ2= µ2/µ1; 2 ≡ Φ2-1= π2/ π1 -1
Testing the significance of the gap (pooled data)
Wald Hyp. Test
System FGLS
18-29
50-65
System OLS
18-<30
50-<65
Production
diff. (λ):
ref=30-49
Labour-cost
diff (Φ):
ref=30-49
Gap (λ-Φ)
χ2
Prob>χ2
0.93
0.78
0.86
1.01
0.08
-0.23
19.71
81.73
0.0000
0.0000
0.98
0.86
0.12
9.76
0.0018
0.59
1.01
-0.42
46.43
0.0000
(λ=Φ)
22
Testing the significance of the gap (by sector)
Wald Hyp. Test
Production diff.
(λ): ref=30-<50
Labour-cost
diff (Φ):
ref=30->50
(λ=Φ)
Gap (λ-Φ)
χ2
Prob>χ2
35.20
17.91
0.0000
0.0000
Industry
System FGLS
18-29
50-65
System OLS
18-29
50-65
1.05
1.04
0.88
1.03
0.01
-0.15
1.15
0.90
0.24
17.19
0.0000
0.68
1.03
-0.35
14.01
0.0002
Commerce
System FGLS
18-29
50-65
System OLS
18-29
50-65
0.96
0.70
0.87
0.99
0.09
-0.29
5.22
24.38
0.0224
0.0000
1.00
0.87
0.12
2.23
0.1354
0.53
0.99
-0.46
12.15
0.0005
Service
System FGLS
18-29
50-65
System OLS
18-29
50-65
0.77
0.74
0.78
1.02
-0.01
-0.28
0.08
30.67
0.7798
0.0000
0.78
0.78
0.00
0.00
0.9476
0.58
1.02
-0.44
16.61
0.0000
23
Testing the significance of the gap (firm-size)
Wald Hyp. Test
Production
diff. (λ):
ref=30-49
Labour-cost
diff (Φ):
ref=30-49
(λ=Φ)
Gap (λ-Φ)
χ2
Prob>χ2
0.76
57.22
0.3841
0.0000
Small firms (<50)
System FGLS
18-29
50-65
System OLS
18-29
50-65
0.90
0.91
0.76
1.01
-0.01
-0.24
0.91
0.88
0.03
0.57
0.4490
0.61
1.01
-0.40
34.03
0.0000
Medium-size firms (50-99)
System FGLS
18-29
50-65
System OLS
18-29
50-65
1.07
0.82
0.84
1.08
0.23
-0.26
28.79
16.75
0.0000
0.0000
1.22
0.87
0.34
16.74
0.0000
0.53
1.08
-0.55
12.38
0.0004
Big firms (100 +)
System FGLS
18-29
50-65
System OLS
18-29
50-65
0.99
0.78
0.76
0.94
0.23
-0.16
25.06
5.41
0.0000
0.0201
1.08
0.78
0.30
10.88
0.0000
0.62
0.94
-0.32
3.71
0.0541
24
Other robustness checks
- Sub-sample of (big) firms properly
reporting on part-time work
- Sub-samble of (big) firms reporting on
human capital attainment of recruits and
separating workers + share of blue-collar
workers
=> no major qualitative impact on estimates
25
Conclusion
•
•
•
An increase of 10 percentage points in
the share of older workers (>50) in a firm
depresses its added value by 3.2%
(preferred model & cross-model average)
Large productivity differential for olders
workers, only partially compensated by
lower relative labour costs
…which could (negatively) affect the
labour demand for older workers.
26
Conclusion (cont.)
•
•
The dominant service sector does not
seem to offer working conditions that
mitigate the negative relationship
between age and productivity
Older workers in smaller firms display a
larger productivity gap, and their
productivity is less aligned on labour
costs. Small firms might be less inclined
to employ them
27
Other stylised facts
28
Other stylised facts (cont.)
Profitablity of firms located in Belgium and workforce age
(intervalles de confiance au seuil de 2,5% confidence interval).
Year1998-2006
Source : Belfirst & Carrefour
Note : Profits : value added/labour cost, centered using year and NACE4 fixed effects. Age data
correspond to the 75th percentile of the firm’s age distribution. Resutls on display are obtained using non
29
prarametric estimation methods
Productivity/labour cost gaps and employment contract à-la-Lazear
Mandatory
departure from
firm 1
Relative levels of
productivity and age
(100=age average)
Wage 1
Wage 2
B
100
Productivity
A
Firm 1
Firm 2
Age/seniority 1
Age/seniority 2
30
References
• Aubert. P. and B. Crépon (2003). “La productivité des salariés âgés : une
tentative d’estimation”. Economie et Statistique. 368. 95-119.
• Dostie. B. (2006). Wages. Productivity and Aging. IZA. Discussion Paper
No. 2496. Bonn. Germany.
• Göbel, Ch. and Zwick, Th. (2009), "Age and productivity: evidence from
linked employer-employee data," ZEW Discussion Papers 09-020, ZEW Zentrum für Europäische Wirtschaftsforschung / Center for European
Economic Research.
• Grund and Westergård-Nielsen (2008). International Journal of
Manpower. Vol. 29(5). pp. 410-422
• Hellerstein. J.K. and Neumark. D. (1995). ‘Are Earning Profiles Steeper
than Productivity Profiles: Evidence from Israeli Firm-Level Data’. The
Journal of Human Resources. vol. 30. 1. pp. 89-112.
• Ilmakunnas, P. and M. Maliranta, (2007), Ageing, Labour Turnovers and
Firm Performance, ETLA DP, No 102, The Research Institute of the
Finnish Economy, Helsinki
31
References (cont.)
• Levinsohn. J. and A. Petrin (2003). Estimating production functions using
inputs to control for unobservables. Review of Economic Studies. 70 (2).
317-341
• Malmberg. B. Lindh. T & Halvarsson. M., (2005). Productivity
consequences of workforce ageing -Stagnation or a Horndal effect?.
Arbetsrapport No 2005:17. Institute for Futures Studies. Stockholm.
• Skirbekk, V. (2004), Age and individual productivity: a literature survey,
In: Feichtinger, G. (Editor): Vienna yearbook of population research 2004.
Vienna: Austrian Academy of Sciences Press, pp. 133-153.
• Skirbekk, V. (2008), Age and productivity capacity: Descriptions, causes
and policy options, Ageing Horizons, 8, pp. 4-12.
• van Ours, J.C & Stoeldraijer, L. (2010), Age, Wage and Productivity, IZA
Discussion Papers 4765, Institute for the Study of Labor (IZA), Bonn.
• Werding, M. (2007). "Ageing, Productivity and Economic Growth: A
Macro-level Analysis," PIE/CIS Discussion Paper 338, Center for
Intergenerational Studies, Institute of Economic Research, Hitotsubashi
32
University