The Microeconomic

Determinants of Older

Employment

Evidence from Belgium and Abroad

Vandenberghe, Vincent (IRES-ESL-UCL)

NBB-BNB

21 Feb 2014

Brussels

1

2

Outline

1. Context

2. Tentative framework: a (labour) supply &

demand story

3. Empirical evidence (with a focus on Belgium)

- Country-level

- Firm-level

4. Some success stories

5. Policy recommendations

1. Context

Ageing (rising health & pension costs)…

… and the understantable desire to expand

an older employment rate that remains

relatively low

3

4

Weakness of older Employment rate in

Belgium

Women

Men

2. Framework

• The existing literature, and most policy

initiatives, have focused on supply-side

determinants (i.e. the role of pension,

early- retirement schemes in fostering

withdrawal for the labour market…)

• But one should not ignore the role of

demand-side barriers (reluctance of firms

to (re)employ older workers) 5

6

Wage

Older Employment

Supply

Demand

L2L1

Supply

7

Supply-Side Demand-side

- Generosity of replacement

earnings

- (Early) pensions benefits

- Unemployment benefits

- Disability benefits

- Health or family

considerations

- Seniority wages

- Employment protection

- Severance costs

_______

- Productivity decline

3. The Empirical Evidence

3.1 Country-level evidence

– What OECD/EU-SILC country-level data

reveal

3.2 Firm-level evidence

(Bel-first + social security data)

- The Employment consequences of steeper

wage-age gradients in Belgium

- The productivity and profitability

consequences of larger shares older workers

8

3.1. Country-level : Youth vs Old: no evidence

of crowding out … on the contrary

9

AUS

AUT

BEL

CAN

CHL

CZE

DNK

ESTFIN

FRA

DEU

GRC

HUN

ISL

IRLISR

ITA

JPN

KOR

LUX

MEX

NLD

NZL

NOR

POL

PRT

SVK

SVN

ESP

SWECHE

TUR

GBR

USAOECD

30

40

50

60

70

80

90

100

30 35 40 45 50 55 60 65 70 75 80

Employment rate, 20-24 year-olds (%)

Percentage of 55-59 year-olds and 20-24 year-olds in employment, 2009)

Employment rate, 55-59 year olds, per cent

Country-level – OECD 2006

OERj= f (BCj, SCHOOLj, PENSIONj, EMPL_PROTj, WAGE SETTINGj,

TAXESj, REL_WAGEj, GENDER*AGE j)

10

Countries: AUS,AUT,BEL,CZE,DEU,DNK,ESP,FIN,FRA,GBR,ITA,NLD,NOR,POL,PRT,SVK,SWE,USA

11

Belgium

12

3.2 Firm-level The Employment consequences

of steeper wage-age gradients in Belgium

13

LnWAGE ij = α + βj AGEij + ρ lnYj

Old Employment Outcomej = f( 𝜷𝒋, Controlsj)

14

Firm-level –The productivity and profitability

consequences of larger shares older workers

Productivity

lnYit=f( Kit ,Lit , AGE SHARESit)

Gross profits (employability)

ln(Yit/Wit)= g( Kit ,Lit , AGE SHARESit)

Key results

4. Success stories

4.1. Successful firms

15

=>BMW “Oldsters Beat Youngsters in BMW

Assembly-Line Test”• In 2007, BMW set up an experimental assembly line with older

employees to see whether they could keep pace. The production

line features stools to spare aging backs, adjustable-height work

benches, and wooden floors instead of rubber to help hips swivel

during repetitive tasks

• The verdict: not only could they keep up, the older workers did a

better job than younger staffers on another line at the same factory

• Today, many of the changes are being implemented at plants across

the company.

16

=> Horndal steel-plant in central

Sweden • Between 1930 and 1950 annual growth rate in

productivity of 2.5 percent

• Horndal had a very aged workforce

– In 1930, 34% of the workers at Horndal were

50+

– In 1950, 50% …

• Horndal suggests that workforce ageing is not a

problem for productivity

17

4.2 A successful country: Sweden

*Horizon: extension of the right to stay in employmentfrom 65 to 67 (2001), discussions to extend to 69

* Supply-side: The pension reform of the 1990’s (fromdefined benefit to defined contribution) introduces strongfinancial incentives to stay active

Simultaneoulsly, the alternative exit routes have been closed

*Demand-side:

• Moderate seniority wage (compare to Belgium, France), declining wage >55

• Reduced social contribution >65

• More attention to workplace ergonomy

• Higher incidence of training18

19

20

21

Source: Eurostat, 2007

5. Policy Recommendations

1. Boost supply

1.1. Expand Horizon (legal age of retirement)

1.2. Make it less financially attractive to retire (replacement rate)

1.3. Better control alternative exit routes (disability…)

2. Entice demand

2.1. Combat age-related productivity decline

- Invest in ergonomy/work environment (BMW…) [1]

- Develop training among 40-50 year-old workers (Sweden)

2.2. Labour costs [2]

- Moderate seniority-based wage rises [3]

- Conditional on social partners committing to [1],[2], [3],

selectively reduce tax wedge on older labour

3. Improve intermediation (activate older unemployed)

– “Seek and Ye Shall Find” J. van Ours (2012)

22

Appendices

23

24

0,1

0,15

0,2

0,25

0,3

0,35

0,4

Source: ISSP2005

A high share of involuntary retirement (55-64)

25

Implicit contrats

What needs to change if the (implicit) age of retirement is raised

Age

wage 1

Productivity

6555

wage 2

100

Productivity

Wage

A

B

25

26

Implicit contrats and wage dynamics in case of job change

Seniority 1

wage 1

Productivity

Employer 1 Employer 2

wage 2

Seniority 2

100

Productivity

Wage

55

27

Implicit contrats and negative shocks- Impact on departure age

Age/seniority

Productivity

Wage

Wage

Productivity

End of implicit

contract

100

A*

B*

5552

28

Average earnings by age: full-time male workers, 2008

Source: D'Addio et al. (2010) based on

OECD Earnings Database.

29

Model and Econometrics

Equ.1:average productivity

ln (Yi,t /Li,t )= lnA + α ln QLi,t +ß lnKi,t – lnLi,t

where: Yi,t is the firm’s value added

and QLit a labour quality index à-la-Hellerstein-Neumark

QLit = ∑k µi,j Li,j,t = µi,ref Li,t + ∑j ≠ref (µi,j - µi,ref) Li,j,t

µj reflecting the productivity of type j workers (e.g.

university)

30

• Assuming same marginal product across firms,

(no subscript i) and taking logs QL becomes:

ln QLi,t = ln µref + lnLi,t+ ln (1+ ∑j ≠ref (λj - 1) Pi,j,t)

Where

λj≡µj/µref is the relative productivity of type j

workers

Pi,j,t ≡ Li,j,t/Li,t the proportion of type j

workers (eg.young, prime age (ref.), old…)

31

Since ln(1+x)≈ x, we can linearize

Ln QLit = ln µref + ln Li,t + ∑j ≠ref (λj - 1) Pi,j,t

And the production function becomes:

ln(Yi,t /Li,t)= lnA+ α [ln µref + ln Li,t + ∑j ≠ref (λj -1) Pijt]

+ ß lnKit – lnLik

or, equivalently

ln (Yit /Lit)= B + (α-1)lit + ∑j ≠ref ηj Pijt + ß kit

where:

– B=lnA+α ln µref

– ηj = α (λj – 1); λj=µj/µref j ≠ref.

– lit=lnLit; kit=lnKit

32

Equ.2: average labour costs

Wit /Lit= ∑j πj Lijt / Lit =πref + ∑j ≠ref.(πj - πref) Lijt/ Lit

Taking the logarithm and using log(1+x)≈ x, we

can approximate this by:

ln(Wt /Lit)= Bw + ∑k ≠ref ηWj Pi,j,t

where

– Bw =ln π0

– ηWj = (Φj – 1)

– Φj≡ πj/ πref j ≠ref

33

Formulating the key hypothesis test is now

straightforward

Assuming spot labour markets and cost-

minimizing firms the null hypothesis of no

impact on the productivity-labour cost ratio

for type j worker implies ηj = ηwj.

Any negative (or positive) difference between

these two coefficients can be interpreted as

a quantitative measure of the disincentive

(incentive) to employ the category of workers

considered.

34

Econometric specificationThe hyp. test = easily implemented if one adopts strictly

equivalent econometric specificationsln (Yit /Lit)= B + (α-1)lit + ∑j ≠ref ηj Pitj + ß kit + γFit + εit

ln (Wit /Lit)= Bw+ (αw-1)lit + ∑j ≠ref ηw

j Pitj + ßw kit + γwFit + εwit

Taking the difference ln (Yit )- ln (Wit )=BG+ αGlit + ∑j ≠ref η

Gj Pitj + ßG kit + γGFit

+ εGit

where:

BG= B -Bw; αG=α-αw ;ηGj= ηj-η

wj;

ßG= ß- ßw ;γG= γ-γw and εGit =εit -εw

it

ηGj = direct estimate of null hypothesis of no impact

on the productivity-labour cost ratio

35

ln (Yit /Lit)= B +

+ ß kit + γFit + εit

with εit = ϴi + τit + σit

ϴi unobservable (time-invariant) heterogeneity between firms

τit short-term (asymmetrically) observed productivity shocks

σit random error E(σit) = 0

The identification problem

36

One can deal with θi by resorting to first differences

or firm fixed effects

The biggest challenge= coping with τit

=> two methods:

IV: lagged values Pi,j,t-1 ; Pi,j,t-2 as instruments (Aubert

and Crépon, 2003, 2007; van Ours & Stoeldraijer,

2011)

Intermediates-as-proxy/more structural approach

Olley & Pakes (1998), Levinsohn & Petrin (2003),

Ackerman, Caves and Fraezer (2006).

37

ACF uses intermediates/materials to proxy the short-term

productivity term

intit =f(τit , kit, qlit)

Assuming this function is monotonic, it can be inverted

τit =f-1(intit, kit, qlit)

Leading to

ln (Yit /Lit)= B+ φ qlit + ß kit + γFit + θi + f-1(intit, kit, qlit) + σit

=> Our specificity it to combine this strategy with fixed

effects to eliminate θi => FE-ACF

38

39

40

41

42

Robustness checks

43

Training