Collective Bargaining
Bruno Van der Linden
Université catholique de Louvain
November 29, 2016.
ESL (UCL)
1 / 110
Outline
1
Perfectly competitive labor market if no unions
Right-to-manage
Efficient contracts
The holdup problem
Equilibrium effects
2
Non competitive labor market if no unions
3
Empirical evidence
4
Extensions
Incomplete information
5
References
ESL (UCL)
2 / 110
Introduction
Outline of the introduction
Historical perspective
Nowadays
1
2
“What do unions do” in Continental Europe?
What are the effects of unions?
Some facts
Union goals
Bargaining theory
ESL (UCL)
3 / 110
Introduction
Historical perspective
“Labor problems and unionization”
Richard T. Ely (1886) The Labor Movement in America
Beatrice and Sidney Webb (1897) Industrial Democracy (U.K.)
are the first analyses of “labor movements” and unions. These books
and the followers were characterized by
An interdisciplinary approach (→ Industrial relations after WWII);
An inductive approach (a lot of case studies);
An historical and comparative approach;
Preoccupation with social reforms.
ESL (UCL)
4 / 110
Introduction
Historical perspective
Already at the end of the 19th/ beginning of the 20th century, a division
between
→ The “labor specialists” (on the whole strong advocates of unionism)
→ and the “economic theorists” (stressing the monopoly aspects of
unions). Even among the latter:
“Marshall, Pigou, Taussig and other leading theorists were
troubled by the ‘peculiarities’ of the labor market – the fact
that the worker sells himself with his services, that his
immediate financial need may place him at a disadvantage in
negotiating with employers, that he is influenced by
non-pecuniary motives, that he has limited knowledge of
alternative opportunities, and that there are objective barriers
to free movement of labor” (Reynolds, 1951)
ESL (UCL)
5 / 110
Introduction
Nowadays
Unions are widespread in Continental Europe and are an extremely
complex “institution”.
Questions arise such as:
1
“What do unions do” in Continental Europe?
2
What are the effects of unions?
ESL (UCL)
6 / 110
Introduction
“What do unions do” in Continental Europe?
Explicit bargaining over wages.
Explicit bargaining over employment is rather unusual.
Often explicit bargaining over working conditions (e.g. working
hours). See Section 1.1.3 of Cahuc and Zylberberg (2004).
Unions also bargain over
Grievance and arbitration procedures to settle a dispute (see e.g. p.
424-426 of Cahuc, Carcillo and Zylberberg, 2014),
The rules governing promotions and discipline (Not covered by the
book),
The rules governing discharges (firing rules like seniority rules or
last-in first-out - LIFO - rule). Not covered by the book.
Severance payments (the book emphasizes this)
ESL (UCL)
7 / 110
Introduction
“What do unions do” in Continental Europe?
In addition, in some countries :
→ Unions and employers jointly manage the social security system,1
→ They take part to management of many institutions that control the
functioning of (part of) the economy :
⇒ “Corporatism”, a rather vague notion that refers to strong
coordination between employers, unions and the government.
Reference: Teulings and Hartog (1998)
Many differences within Europe. See e.g.
http://www.worker-participation.eu/
National-Industrial-Relations/Countries
1
See Boeri, Brugiavini and Calmfors (2001). An example is the so-called “Ghent
system” in which unemployment insurance schemes are run by trade unions and
partially subsidized by the State (e.g. Finland, Sweden, Denmark and Belgium).
ESL (UCL)
8 / 110
Introduction
What are the effects of unions?
According to a well-known reference, Nickell and Layard (1999),
“...unions raise unemployment and reduce labor input (i.e.
hours/population). These effects are, however, offset if
unions and employers can coordinate their wage bargaining
activities” (p. 3055).
This chapter presents some basic tools to deal with unions and shows
that underlying assumptions are needed to show that unions cause
unemployment.
Comprehensive analyses about unions and their effect are available in
Booth (1995), Booth, Burda, Calmfors, Checchi, Naylor and Visser
(2000) and Boeri, Brugiavini and Calmfors (2001).
ESL (UCL)
9 / 110
Introduction
Some facts
Indicators of “union power”
More information on the Moodle webpage (not compulsory). See the
file du_caju_et_al_Institutional_features_of_wage_bargaining_in_22_EU_countries.pdf
Union density = the proportion of wage-earners who are unionized
Collective bargaining coverage = the proportion of wage-earners who
are covered by collective agreements.
France (FRA): Union density ≈ 10%; Coverage ≈ 90%
(qualitatively similar in Germany (DEU), The Netherlands
(NLD),...);
Nordic countries (and Belgium): density higher than in the
previous group and coverage > density;
U.S.: Union density ≈ Coverage ≈ 15%
ESL (UCL)
10 / 110
Introduction
Some facts
Indicators of “union power”: Density and Coverage in 2006
100,0
Austria
France
Belgium
90,0
TheNetherlands
80,0
Danemark
Italy
Spain
Norway
70,0
Unioncoverage(%)
Sweden
Finland
Germany
Portugal
60,0
Australia
50,0
40,0
Poland
30,0
UK
Canada
20,0
USA
Japan
10,0
0,0
0,0
10,0
20,0
30,0
40,0
50,0
60,0
Uniondensity(%)
70,0
80,0
90,0
100,0
Figure: Coverage vs density: Cross-country comparison
ESL (UCL)
11 / 110
Union density in selected countries, 1970−2009/10
90.0
Union density (%)
eclining
ted by
mpirical
onomic
helming
eclining
a loss
d firms
ses the
nism is
he nonuse” is
Introduction
1970
80.0
1980
70.0
1990
2000
60.0
2009/10
50.0
40.0
30.0
20.0
10.0
0.0
AU
DE
FR
JP
NL
NO
SE
UK
US
Source: ICTWSS Database. Online at: http://uva-aias.net /208.
ESL (UCL)
Figure: Long-run trends
12 / 110
Introduction
Some facts
Indicators of “union power”
In Continental Europe, the amount of heterogeneity is substantial !
(Recognized) unions are often by law the “institution” that has the
right to negotiate wages (See Cahuc, Carcillo and Zylberberg,
2014, p. 411-2, 461);
The same pay between unionized and non-unionized workers is
often the (enforceable?) rule;
Under certain conditions, a mandatory extension of collective
agreements to all firms of the sector can be frequent. In some
countries, under certain conditions, firms have the opportunity of
opting out of a collective agreement.
In the U.S., “if a majority of workers vote in favor of the union, the law
required the management to bargain in ‘good faith’ with the recognized
union” (DiNardo and Lee, 2004, p. 1385).
ESL (UCL)
13 / 110
Introduction
The bargaining level
In principle, collective bargaining can take place at different levels:
The firm or the establishment;
The sector;
The region;
On a national scale;
For some issues an international scale (EU, ILO).
In practice, there is often an overlap between negotiations occurring at
several levels, with or without (much) explicit or implicit coordination.
Furthermore, in some countries the government intervenes in the
bargaining process setting legal minimum wages, fixing wage growth
norms, freezing wage levels during a period of time, and the like.
ESL (UCL)
14 / 110
Introduction
Union goals
Do unions maximize an objective function?
This is an old debate.
♦ Dunlop (1944) answered yes.
♦ Ross (1948) answered no: “Of all participants in economic life, the
trade union is probably least suited to purely economic analysis”.
Currently, there are 3 views:
1
The dominant one assumes that unions maximize an objective
function. Developed by Cahuc, Carcillo and Zylberberg (2014).
2
The most important alternative assumes: (i) union members have
heterogeneous preference over, say, the wage and (ii) perfectly
democratic union with heterogeneous members (choice through a
vote). Briefly mentioned later on.
3
Union leadership has discretionary power and possible conflicts of
objectives with the members. Ignored later on.
ESL (UCL)
15 / 110
Introduction
Union goals
Main assumptions in the book of CCZ = Cahuc, Carcillo and Zylberberg (2014)
(Mostly) a static setting
A union cares for an exogenous number N of (typically)
homogeneous “members” (N = “the size of the union”.
Interpretations: Formal members, employed members, the whole
labor force...).
Each of the N “members” supply one unit of labor. The net real
wage paid is denoted w ≥ an exogenous reservation wage w
called the “outside option”. Interpretations: See below.
An employed worker attains a level of utility v (w). The latter is
increasing and concave (risk aversion).
Perfect information about preferences, profits,...
ESL (UCL)
16 / 110
Introduction
Union goals
Main assumptions in the book of CCZ.
The union is here assumed
→ to care only for its “members” and
→ to have a utilitarian objective function Vs that trades off employment
and net wages. 2
Assuming an equal treatment of all “members” if employment falls
short of the size of the union (i.e. if L < N): (CCZ p. 427)
Vs = ` · v (w) + (1 − `) · v (w), where ` = min(1, L/N)
(1)
Note: If w is an unemployment benefit. Empirical evidence that the
utility loss in case of unemployment is larger than the loss of money.
Captured by the concavity of v (·) or a different utility function ṽ (·) when
unemployed.
2
The literature has also used other functional forms for Vs . See e.g. p. 429 of CCZ.
ESL (UCL)
17 / 110
Introduction
Union goals
Indifference curves of the union
ESL (UCL)
18 / 110
Introduction
Bargaining Theory
Rubinstein (1982); Chapter 16 of Osborne (2004); CCZ p. 415-422.
Standard assumptions (see however CCZ p. 422):
Infinite horizon.
Two impatient players who have to share a time-invariant “pie”.
Two rational players informed about each other’s preferences.
On even dates, player 1 proposes a partition which player 2
accepts or refuses. On odd dates, player 2 has the initiative.
Subgame perfect equilibria. Both players with discount factor < 1
agree on a partition of the “pie” at the outset of the game.
Let rU and rΠ be the discount rate resp. of the union and the firm
owner ⇒ the “bargaining power of the union” is defined as
rΠ
rU + rΠ
If rU → 0, then γ → 1. As rU /rΠ increases, γ → 0. Impatience reduces
the bargaining power and conversely.
γ=
ESL (UCL)
19 / 110
Introduction
Bargaining Theory
Π = Π(w) is the profit function
Π is a “reservation level of profit” (e.g. the firm fires all the workers
and recruit other ones or the firm shuts down and relocate in
another region). Π = firm’s “outside option” (w is the worker’s
outside option).
V0 et Π0 are the levels reached during the negotiation in case of a
strike (or other action like work-to-rule, go-slow). Called “inside
options”.
Under certain conditions, the solution of the Rubinstein bargaining
game converges to the following generalized Nash solution:
γ 1−γ
max Vs − V0
Π − Π0
w
s.to w ≥ w and Π ≥ Π
ESL (UCL)
(2)
20 / 110
Perfectly competitive labor market if no unions
Sec. 1. Perfectly competitive labor market if no unions
After this introduction, we start with the standard implicit assumption:
The labor market would be perfectly competitive in there were no
unions.
The supply of labor to a single firm is flat provided that
w ≥w
where w can be
the value out of the labor force;
the competitive wage if there is a unique unionized firm located in
an otherwise competitive labor market;
a mix of wages in other firms and of unemployment benefits if
unemployment prevails for whatever reason (e.g. a binding legal
minimum wage) and an unemployment insurance system exists.
ESL (UCL)
21 / 110
Perfectly competitive labor market if no unions
Assumptions about Vs − V0
Let w0 be the net income of a worker during a strike (without resorting
to outside opportunities). Remembering the union’s objective (1),
Vs − V0 =
(L/N)[v (w) − v (w)] + [v (w) − v (w0 )]
v (w) − v (w0 )
if L < N
if L ≥ N
(3)
To simplify expressions, CZ, as many authors, assume that w0 = w.
Is it a sensible assumption? Few data on strike payments!
⇒ Difficult question.
Example: In the Belgian metal industry in 2008, the strike payment
= 25e/day the first week, 31e the second one. If w is proxied by the
level of unemployment benefits, the level of unemployment benefits in
Belgium (1st year) ranged between 24 and 44e/day in 2008 in case of
a sufficient record of employment.
ESL (UCL)
22 / 110
Perfectly competitive labor market if no unions
Assumptions about Π − Π0
Consider a single-input firm3 with revenue function
R(L) with R 0 > 0, R 00 < 0.
The (real) profit function is then simply a function of the endogenous
pair (w, L):
Π(w, L) = R(L) − w · L
CCZ assume Π0 = 0 (no production, no fixed cost).4
3
Union models where firms have more than one input are considered e.g. by
Manning (1994), CCZ p. 445, Van der Linden (2002).
4
Krueger and Mas (2004) argue that Firestone’s decision to hire replacement
workers during a strike led to a serious loss in product quality. Mas (2008) provides
evidence that Π0 could be < 0 if there is a labor dispute.
ESL (UCL)
23 / 110
Perfectly competitive labor market if no unions
Right-to-manage
A static right-to-manage model in PARTIAL
equilibrium: one firm-one union setting
The right-to-manage = “The union and the firm bargain over w knowing
that, conditional on w, the firm chooses the level of employment L that
maximizes profits” (The timing matters: L is chosen after w).
That is, ∀w, L is given by the demand curve
R 0 (L) − w = 0 ⇒ Ld (w) = R 0−1 (w).
Along an iso-profit curve
R 0 (L) − w · dL − L · dw = 0
and hence the slope of an iso-profit curve is given by:
R 0 (L) − w
dw
=
dL
L
in a two-dimensional space (L, w).
ESL (UCL)
24 / 110
Perfectly competitive labor market if no unions
Right-to-manage
Iso-profit curves
w
Isoprofit curves
Π(w,L) = constant
Ld(w)
. . . .
Profit
increases
L
ESL (UCL)
25 / 110
Perfectly competitive labor market if no unions
Right-to-manage
Assumptions
By assumption, the firm bargains with the (recognized) union:
This can be imposed by law (as in many countries of Western
Continental Europe).
In the anglo-saxon countries, this setting is called “closed shops”
or “union shops”, whereby a new employee has to join the local
union within a certain period of time after hiring (OECD,
Employment outlook, 2004).
We start by assuming L ≤ N.
By assumption, the negotiated wage is the same for all workers (no
discrimination).
By assumption, at w = w, the profit level is positive (an unexplained
rent to be shared).
ESL (UCL)
26 / 110
Perfectly competitive labor market if no unions
Right-to-manage
The bargaining problem
The maximization
γ 1−γ
max Vs − V0
Π − Π0
w
(4)
s.to w ≥ w and Ld ≤ N
becomes under (1) and the assumption w0 = w:
h
iγ
d
max L (w)/N [v (w) − v (w)]γ [Π(w)]1−γ
w
s. to Ld ≤ N, w ≥ w
(5)
(6)
where Π(w) the profit function if L = Ld (w), i.e.
Π(w) ≡ Π(w, Ld (w)) = R(Ld (w)) − w · Ld (w).
CCZ search for an interior solution to (4). It is easy to take account of
the “outside option” constraint Π ≥ Π.
ESL (UCL)
27 / 110
Perfectly competitive labor market if no unions
Right-to-manage
The monopoly union (γ = 1)
Graphical exposition
The union acts as a Stackelberg leader and the firm is the follower.
ESL (UCL)
28 / 110
Perfectly competitive labor market if no unions
Right-to-manage
The monopoly union (γ = 1)
Analytical solution
Ignore for a while the constraints Ld ≤ N and w ≥ w.
wM = arg max Ld (w) [v (w) − v (w)]
(7)
Notations:
The absolute value of the elasticity of labor demand is
ηwL (w) = −(w/Ld (w))(dLd (w)/dw) > 0
The first-order condition (f.o.c.) is then
1
v (wM ) − v (w)
= L
0
wM · v (wM )
ηw (wM )
(8)
As the right-hand side is positive, one has wM > w.
ESL (UCL)
29 / 110
Perfectly competitive labor market if no unions
Right-to-manage
The (static) monopoly union (γ = 1)
Analytical solution
In sum,
If Ld (wM ) ≤ N
then w = wM , L = Ld (wM )
d
otherwise L = N, w | L (w) = N
(9)
(10)
i.e. the second case is a corner solution where all members are
employed.
Exercise 3 in the file Exercices_Unions.pdf invites you to check that
static models overrates the impact of (a monopoly) union on
employment compared to dynamic models.
ESL (UCL)
30 / 110
Perfectly competitive labor market if no unions
Right-to-manage
The general case 0 ≤ γ ≤ 1
Consider an interior solution to Problem (5). The f.o.c is an implicit
equation in w (see CCZ for the proof) :
γ
v (w) − v (w)
= µs ≡ L
0
wv (w)
γηw (w) + (1 − γ)ηwΠ (w)
(11)
where ηwΠ (w) = −(w/Π(w))(dΠ(w)/dw) ≥ 0
• Shocks to R(·) affect w if ηwL or ηwΠ vary. Real rigidity 6= general.
• If γ > 0, then w > w, ⇒ µs is a mark-up.
∂µs
∂γ
>0
• The higher ηwL or ηwΠ , the lower µs . Intuition?
L
Π
w (w)
w (w)
• Sufficient conditions: µs < 1, ∂η∂w
≥ 0, ∂η∂w
≥ 0.
ESL (UCL)
31 / 110
Perfectly competitive labor market if no unions
Right-to-manage
The general case 0 ≤ γ ≤ 1
Approximation and wage rigidity
Taylor expansion of order 1 of v (w) implies
w≈
v (w)−v (w)
wv 0 (w)
≈1−
w
w.
So,
w
>w
1 − µs
Particular case leading to totally rigid real wages:
σ
Let v (w) = wσ , σ ≤ 1, σ 6= 0.
00 (w)
= 1 − σ ≥ 0 (constant)
relative risk aversion ≡ −w·v
0 (w)
v
h
σ i
v (w)−v (w)
⇒ wv 0 (w) = σ1 1 − w
w
ESL (UCL)
32 / 110
Perfectly competitive labor market if no unions
Right-to-manage
The general case 0 ≤ γ ≤ 1
Particular case with real wage rigidity
Assuming also an iso-elastic revenue function
R(L) = A · Lα , A > 0, α ∈]0, 1[
⇒ ηwL = 1/(1 − α) et ηwΠ = α/(1 − α), independent of A and w!
⇒ µs =
γ(1−α)
γ+α(1−γ) ,
hence µs ∈]0, 1[ if γ > 0
Then, Equation (11) becomes (check!):
w=
w
1/σ
[1 − σ · µs ]
, with
∂w
∂w
> 0,
>0
∂w
∂µs
(12)
and real wages are fully rigid ( = not affected by multiplicative shocks
i.e. on A).
ESL (UCL)
33 / 110
Perfectly competitive labor market if no unions
Right-to-manage
Bargained wages and unemployment
As Blanchard and Giavazzi (2003), in a fully unionised economy, let’s
assume quite naturally that the “outside option”, w, is negatively
affected by the unemployment rate u:
w = f (u), with f 0 (u) < 0
Starting from an equation like (12), it is easily seen that:
ln[w] = ln[f (u)] − (1/σ) · ln [1 − σ · µs ]
(13)
Such a relation is called a “wage curve”. It relates the wage level (not
its growth rate as in a Phillips curve) to the level of the unemployment
rate (See the section devoted to empirical work).
ESL (UCL)
34 / 110
Perfectly competitive labor market if no unions
Right-to-manage
Unions & occupational health and safety (OHS)
An extension to the right-to-manage model by Donado and Wälde (2012)
Main assumptions (OHS is only discussed here):
Workers value the wage and OHS, denoted by a scalar s: their
utility is v (w, s), vs > 0, vss < 0.
Firm’s revenue is negatively affected by s: A(s)R(L), A0 < 0.
Labor supply is positively affected by s (e.g. the share of days a
worker is healthy and can work)
Under full information about job characteristics, to attract workers
in low s jobs, firms should provide a higher wage. This is the
compensating wage differentials approach (Rosen, 1986).
Donado and Wälde (2012) assume instead
→ total individual ignorance of the impact of job characteristics on s.
So, they choose to supply labor on the basis of w only while
→ unions have a superior (collective) learning technology in
comparison to individuals.
ESL (UCL)
35 / 110
Perfectly competitive labor market if no unions
Right-to-manage
Unions and OHS
To simplify the analysis as much as possible, Donado and Wälde
(2012) assume
Wages are not set through bargaining but are as in a competitve
framework while
Unions have a monopoly power on s.
Unions’ objective function is L v (w, s), where labor demand L is
known (by unions) to be negatively related to s via A(s). Note that
unions ignore the impact of s on those who do not work (those
who are sick).
The authors conclude:
“A laissez faire approach in which firms set safety standards
is suboptimal as workers are not fully informed of health risks
associated with jobs. Safety standards set by better informed
trade unions are output and welfare increasing.”
ESL (UCL)
36 / 110
Perfectly competitive labor market if no unions
Right-to-manage
A quick look at the democratic union
A monopoly union story (e.g. Booth, 1984)
If the vote has the following (restrictive) characteristics:
majority rules;
agents vote sincerely without pre-vote communication;
they are voting on a single variable (the wage);
the union members’ utility functions are heterogeneous but admits
only one maximum (wage);
Then, the decision expresses the preferences of the median voter.
If the median voter
faces a low risk of being laid-off (say, thanks to a first-in last-out
rule),
does not value the utility of other members,
Then, the chosen wage can be “very high” (and employment, low).
Should be less true in a dynamic setting with forward-looking agents
and endogenous membership.
ESL (UCL)
37 / 110
Perfectly competitive labor market if no unions
Efficient contracts
Weakly efficient negotiation (partial equilibrium)
CZ p. 397
Can the union-firm pair do better than adopting the right-to-manage
solution?
Even though there is not much evidence that firms and unions bargain
over employment, let us see what would be the outcome of a bargain
over w and L:
max [L/N]γ [v (w) − v (w)]γ [Π(w, L)]1−γ s. to Ld ≤ N, w ≥ w
w,L
(14)
F.O.Cs:
L:
w:
ESL (UCL)
R 0 (L) − w
γ
γ Π
+ = 0 ⇔ w = R 0 (L) +
Π(w, L)
L
1−γ L
L
γv 0 (w)
−(1 − γ)
+
=0
Π(w, L) v (w) − v (w)
(1 − γ)
(15)
(16)
38 / 110
Perfectly competitive labor market if no unions
Efficient contracts
From the two F.O.Cs, an implicit relation between the bargained levels
of L and w that is independent of γ.
It is called the “contract curve” and defined by:
w − R 0 (L) =
v (w) − v (w)
v 0 (w)
Totally differentiating this equality w.r. to w and L yields the slope of
the contract curve: See next slide.
Note: From (16) and a Taylor expansion of order 1, one immediately
derives that:
γ Π(w, L)
w ≈w+
1−γ
L
The real wage is equal to the outside option plus a share of profit per
worker.
ESL (UCL)
39 / 110
Perfectly competitive labor market if no unions
ESL (UCL)
Efficient contracts
40 / 110
Perfectly competitive labor market if no unions
Efficient contracts
Is the weakly efficient negotiation plausible?
→ With risk averse workers, employment is higher than at the
“alternative wage” w (the latter can be interpreted as the competitive
wage). With this conclusion about the “(weakly) efficient contract”, it is
hard to argue that unions cause “too low a level of employment”.
→ However, bargained contracts typically do not stipulate an
employment level. It is easily seen that the firm will increase profits by
reneging the contract, namely by moving towards the labor demand
curve at the contractual wage rate. So, it is argued, efficient contracts
are not plausible.
→ Espinosa and Rhee (1989) challenge this conclusion by considering
a repeated game with credible punishments if the firm deviates from an
efficient agreement.
In a subgame perfect equilibrium, the solution can lie anywhere
between the right-to-manage solution and the efficient solution
according to the time preference of the agents.
ESL (UCL)
41 / 110
Perfectly competitive labor market if no unions
Efficient contracts
Strongly efficient contracts (partial equilibrium)
CCZ p. 437; See Booth (1997) for an early treatment.
Nothing prevents unions and firms to bargain over variables other
than employment and wages.
There is some evidence that unions and firms bargain over the
level of unemployment benefits or “severance payments”, i.e. a
payment given to union members who are redundant.
Assume a static setting where
A union representing risk-averse workers bargain simultaneously
over w and a benefit b that is added to w (i.e. income union
members get under all circumstances if they are redundant in this
firm).
The firm owner keeps the “right-to-manage”.
Still, perfect information (no incentive problems: those who are
instructed to work offer a unit of labor)
ESL (UCL)
42 / 110
Perfectly competitive labor market if no unions
Efficient contracts
Whatever the probability (` = L/N < 1) of being employed, compared
to a labor contract C = (w, b) that
pays w to those who get a job (assuming w > w) and
b to those who are redundant, with w + b < w,
A risk-averse union prefers (ex ante) a contract Cˆ = (ŵ, b̂) such that
earnings are lower in case of employment: the wage is
ŵ = `w + (1 − `)(w + b) < w
a benefit b̂ is transferred to redundant workers, with
b̂ = ŵ − w ⇒ w + b̂ = ŵ
⇔ Perfect insurance against the unemployment risk
ESL (UCL)
43 / 110
Perfectly competitive labor market if no unions
Efficient contracts
CCZ show that the employer is indifferent between these two contracts
ˆ in the sense that the wage bill is the same. Hence, “this
C and C,
conclusion does not depend on the employer’s attitude to risk”
(CCZ2014, p. 438).
In addition, they show that the profit with Cˆ is:
Π(L) = R(L) − w · L − b̂ · (L + N − L)
Hence, the firm chooses the employment level according to R 0 (L) = w.
Given this level of employment,
b̂ = arg max (v (w + b) − v (w))γ (Π(L))1−γ .
CCZ call this the “strongly efficient contract”.
With bargaining over wages and severance pay, it is hard to argue that
“unions cause too low a level of employment”. (Nickell and Layard,
1999)!
ESL (UCL)
44 / 110
Perfectly competitive labor market if no unions
Efficient contracts
Critical assessment:
The strongly efficient contract assumes perfect information:
Information is actually imperfect.
Imperfect information about effort on the job ⇒ a trade off
between insurance and incentives.
Furthermore, in more general models where firms are allowed to
recruit above N (CCZ p.443-5), the property
R 0 (L) = w
remains true only if “two-tier” contracts are allowed (i.e. identical
employees are paid differently whether they are incumbent workers
(“insiders”) or newcomers (“entrants”)). Is this realistic? Not
exceptional but still relatively rare in reality since discriminatory
(according to e.g. Bewley, 1998, and Fehr and Falk, 1999).
⇒ the “strongly efficient contract” should be seen as a limit case.
ESL (UCL)
45 / 110
Perfectly competitive labor market if no unions
Efficient contracts
Relaxing the assumption L ≤ N
CCZ p. 443-5
It is possible to build a right-to-manage model where the firm can hire
additional workers (“outsiders”) if all incumbent workers (“insiders”),
who bargain over working conditions, keep their job.
As already alluded to before, this model has quite distinct properties
whether
There is a single (real) wage, w, paid to all employees (could
capture a requirement of fairness):
Message: bargaining power of insiders can be detrimental
to employment.
Or identical workers can get lower wages if they are newcomers.
Message: bargaining power of insiders is a source
of discrimination not a cause of unemployment.
Note: This wage differential is not due to an on-the-job training cost. It
is pure discrimination.
ESL (UCL)
46 / 110
Perfectly competitive labor market if no unions
The holdup problem
Consider now a revenue function R(K , L) increasing and concave in
both capital (K ) and labor. For simplicity, assume that L is fixed.
Let r designate the user cost of capital and Π = R(K , L) − w L − r K .
The FOC wrt to K writes:
RK (K ∗ , L) = r .
Message:
“The irreversible character of investment gives the firm an
incentive to under-invest when the union cannot make a
credible commitment not to renegotiate wages once the
equipment has been installed. In this situation, the union
knows that every strike costs the firm rK , whereas the strike
has a cost of zero if it is impossible to renegotiate wages. The
union can thus demand a larger share of the profits in the first
case, which provokes a reduction in investment.” (CCZ, p.
448)
ESL (UCL)
47 / 110
Perfectly competitive labor market if no unions
The holdup problem
Timing of decisions:
1
The firm chooses a level of capital (irreversible choice);
2
Wage bargaining over w (L fixed).
Backward induction:
a) The wage bargain
For given K the bargaining problem is then written as follows:
γ
max [v (w) − v (w)] [R(K , L) − wL − rK − (−rK )]1−γ
w
Let w(K ) designate the solution to this problem.
b) The forward-looking firm chooses K̂ knowing this K 7→ w(K )
relationship:
RK (K̂ , L) = w 0 (K̂ ) L + r
CCZ show that w 0 (K̂ ) > 0 ⇒ K̂ < K ∗ . This is the holdup problem.
ESL (UCL)
48 / 110
Perfectly competitive labor market if no unions
Equilibrium effects
Monopolistic competition and unions
Blanchard and Giavazzi (2003)
This is not covered by CCZ. If needed the paper is on the Moodle
webpage.
Basic assumptions
Monopolistic competition in the goods market ⇒ endogenous
rents.
(Weakly) efficient Nash union-firm bargaining.
Two periods:
Short run: A fixed number m of homogeneous firms and goods
Long run: The number of firms is endogenous (entry condition)
An exogenous number L of homogeneous workers/consumers.
ESL (UCL)
49 / 110
Perfectly competitive labor market if no unions
Equilibrium effects
Monopolistic competition and unions
Specific assumptions
The utility of worker j in each period is given by the following CES
utility function:
"
Vj = m−1/σ
m
X
#σ/(σ−1)
(σ−1)/σ
Cij
(17)
i=1
Cij is worker j’s consumption of good i (price Pi )
elasticity of substitution σ ≡ σ · g(m) > 1, g 0 (m) > 0.
Labour supply LSj ∈ {0, 1} every period.
The entire income is spent on consumption. There are no savings
nor capital in the model.
ESL (UCL)
50 / 110
Perfectly competitive labor market if no unions
Equilibrium effects
Monopolistic competition and unions
Specific assumptions
When not employed, nominal income = Pf (u), with f 0 (u) < 0. P is
the price index
!1/(1−σ)
m
1 X 1−σ
Pi
P≡
m
i=1
In each period, the budget constraint of individual j is:
m
X
Pi Cij = Wj · Lj + Pf (u) · (1 − Lj )
i=1
where Wj = the nominal wage and Lj ∈ {0, 1}.
All firms have the same simple technology: Yi = Li
Firms’ (real) entry costs, c, are proportional to output
ESL (UCL)
51 / 110
Perfectly competitive labor market if no unions
Equilibrium effects
Efficient union-firm i bargaining
Assumptions and notations.
union goal = the wage bill Wi · Li (risk-neutrality; no role here for a
size of the union)
Nash product
Union component = the surplus to workers from working in firm i
For each individual,
P
It she is employed
firm i, m
i=1 Pi Cij = Wj Lj = Wi
Pin
m
If not employed, i=1 Pi Cij = Pf (u)
Firm’s component = (Pi Yi − Wi Li ) − 0 = (Pi − Wi )Li
Exogenous union bargaining power γ
So, in any firm i, the bargaining problem can be written as:
Max Ω = γlog((Wi − Pf (u))Li ) + (1 − γ)log((Pi − Wi )Li )
Wi ,Li
ESL (UCL)
52 / 110
Perfectly competitive labor market if no unions
Equilibrium effects
Short-run partial equilibrium
Taking total demand Y , P and u as given, firm i and its union negotiate
the level of employment Li and the wage Wi . They solve the static
problem:
Max
Wi ,Li
Ω ≡ γlog(Wi − Pf (u)) + γlogLi
(18)
+(1 − γ)log(Pi − Wi ) + (1 − γ)logLi
s.to
Yi = Li
−σ
Pi
Y
. This defines a
However, the demand for good i is Yi = m
P
one-to-one relationship between Li and Pi .
Taking it into account, the problem can be written as Max Ω(Wi , Pi ).
Wi ,Pi
So, ...
ESL (UCL)
53 / 110
Perfectly competitive labor market if no unions
Equilibrium effects
Short-run partial equilibrium
F.O.C.s
∂Ω
=0 ⇒
∂Pi
⇒
−σ(Y /m)(Pi /P)−σ−1 (1/P)
1−γ
+
=0
(Y /m)(Pi /P)−σ
Pi − Wi
1−γ
σ
=
Pi
Pi − Wi
∂Ω
γ
1−γ
=0 ⇒
=
∂Wi
Wi − Pf (u)
Pi − Wi
⇒ γPi = Wi − (1 − γ)Pf (u)
(19)
(20)
Hence,
(1 − γ)(Pi − Pf (u)) = Pi − Wi
Pi
σ
⇒
=
f (u)
P
σ−1
ESL (UCL)
(21)
54 / 110
Perfectly competitive labor market if no unions
Equilibrium effects
Short-run partial equilibrium
1
1
= σ·g(m)−1
, with ∂µ/∂m < 0, ∂µ/∂σ < 0. Then, the
Let µ(m) ≡ σ−1
real wage in firm i is given by (use the F.O.Cs):
Wi
P
Pi
P
σ
= (1 − γ)f (u) + γ
f (u)
σ
−
1
σ
= 1+
− 1 γ f (u)
σ−1
= (1 + γµ(m))f (u)
(22)
= (1 + µ(m))f (u)
(23)
Higher mark-ups µ → higher rents → higher real wages, especially if γ
is high. All this is conditional on Y , P and u.
ESL (UCL)
55 / 110
Perfectly competitive labor market if no unions
Equilibrium effects
Short-run symmetric general equilibrium
Because the model is symmetric, all firms choose the same price level
in equilibrium:
Pi = P ∀i ∈ {1, ..., m}
So,
1 = (1 + µ(m))f (u)
µ(m) but not γ fixes u (higher µ ⇒ higher u!). Hence,
1 + γµ(m)
Wi
=
P
1 + µ(m)
higher µ(m) ⇒ lower real wages
higher γ ⇒ higher real wages ⇒ deregulation understood as lowering
γ reduces real wages without affecting unemployment: In the short-run
(i.e. number of firms m being fixed), a clear loss for unions.
ESL (UCL)
56 / 110
Perfectly competitive labor market if no unions
Equilibrium effects
Long-run symmetric general equilibrium
In the long run, rents must cover entry costs (that - for convenience are assumed to be proportional to output). Per unit of output:
Wi
Wi
Pi
−
= 1−
P
P
P
1 + γµ(m)
= 1−
1 + µ(m)
(1 − γ)µ(m)
=
1 + µ(m)
= c
1
c
⇒ µ(m) =
=
(24)
σ · g(m) − 1
1−γ−c
| {z }
⊕
ass.>0
c
1
⇒ f (u ) =
=1−
(25)
1 + µ(m)
1−γ
⇒ Wi /P = 1 − c
(26)
ESL (UCL)
57 / 110
Perfectly competitive labor market if no unions
Equilibrium effects
Long-run “general” equilibrium
In sum in the long run,
1
the real wage is not affected by the workers’ bargaining power γ
because the supply of firms is fully elastic in the long run: Firms
simply reach the break-even point (profit per worker is c).5
2
a decrease in γ raises profits and leads to more active firms, more
competition and lower mark-ups. Moreover, unemployment
shrinks (compare with the short-run!)
3
a decrease in the entry cost c has a positive effect on the the real
wage.
4
a decrease in the entry cost favours entry and reduces the
unemployment rate.
Extensions: See e.g. Felbermayr and Prat (2011).
5
On p. 884, Blanchard and Giavazzi provide two interpretations for c, one of them
implying positive pure profits under free entry.
ESL (UCL)
58 / 110
Perfectly competitive labor market if no unions
Equilibrium effects
Empirical evidence
on product market regulation using a panel of countries
Empirically we have shown that the significant product
market de-regulation experienced in the 1990s by some
OECD countries was associated with an increase in
competition as measured by average firm profitability. Such
exogenous increases in competition are further associated
with increases in aggregate employment and the real wage.
(Griffith, Harrison and Maccartney, 2007, p. C162)
We find that, for the average OECD country, high and
long-lasting unemployment benefits, high tax wedges and
stringent anti-competitive product market regulation (PMR)
increase aggregate unemployment. (Bassanini and Duval,
2009)
ESL (UCL)
59 / 110
Non competitive labor market if no unions
Sec. 2. Non competitive labor market if no unions
What’s the impact of unions if the labor market is not perfectly
competitive in the absence of unions?
Three different settings will be considered.
In the absence of unions, the labor market could be characterized as
1
A monopsony (i.e. a single firm or buyer faces many sellers of
labor)
2
A monopsonistic competitive market (based on the “Equilibrium
search model”; CCZ p. 306-314)
3
A labor market with matching frictions (based on Chap. 9 of CCZ)
ESL (UCL)
60 / 110
Non competitive labor market if no unions
1. Monopsony and Unions
Falch and Strom (2007) assume
A (non-discriminating) monopsony (in partial equilibrium) with one
input
A labor supply curve which is less than perfectly elastic.
Conclusions:
1
The model predicts an inverted-U shaped relationship between
the bargaining power of the union and employment.
2
“Bargaining power in the hands of trade unions may give an
efficient outcome because ‘medium’ powerful unions generate an
outcome equal to the ‘competitive’ solution” (The authors’
conclusion, p. 206)
Monopsony is however a limit case.
ESL (UCL)
61 / 110
Non competitive labor market if no unions
2. Monopsonistic competition
In the chapter about job-search, equilibrium search models are
discussed under the assumptions:
Search frictions on the labor market,
The unemployed and the employed job-seekers seek for a (better)
job,
A large number of firms compete to attract workers by making
take-it-or-leave-it wage offers.
In this setting, despite competition, firms keep some monopsony
power in equilibrium.
ESL (UCL)
62 / 110
Non competitive labor market if no unions
We will not explicitly introduce unions in this setting.
We will assume that
Centralized or sectoral collective bargaining fixes a wage floor
(call it z);
This wage floor is enforced;
Individual firms compete to attract workers, by posting wages ≥ z
(taking z as exogenous);
The arrival rate of job offers is endogenous
Workers and jobs are homogeneous6 .
The presentation is based on Mortensen (2000). This paper merges
the setting of Burdett-Mortensen and the one of Pissarides.
6
See Cai, Gautier, Teulings and Watanabe (2014) that, in an extension with
heterogeneous workers and firms, studies the pros and cons of bargaining at the
sectoral versus the firm level.
ESL (UCL)
63 / 110
Non competitive labor market if no unions
Assumptions
In a continuous time setting with infinitely-lived agents, assume the
following stationary environment:
A1 Rational forward-looking and homogeneous risk-neutral agents
who only care about their income. For simplicity, the instantaneous
value in unemployment is nil (no UB, no cost or search).
A2 Job-seekers choose to reject or accept a job offer, if any. An
accepted wage remains constant all along the employment spell.
Rejected offers (i) cannot be recalled, (ii) lead to no sanction.
A3 On-the-job search. No threshold wage above which searching on
the job no longer pays. λu = λe = λ designate the equal
exogenous arrival rates of job offers respectively for the
unemployed and the employed. So, with [A1], the reservation
wage x = 0.
ESL (UCL)
64 / 110
Non competitive labor market if no unions
Assumptions continued
A4 Knowing the exogenous wage floor z, firms choose their wage
offer w ≥ z and commit to pay that wage. To currently employed
workers, firms send wage offers ignoring their current wage. The
worker’s current employer does not respond to the outside offer.
A5 Constant exogenous job destruction rate, 0 < q < +∞.
A6 Constant exogenous discount rate, r .
A7 A continuum of workers and firms, each of unitary mass.
ESL (UCL)
65 / 110
Non competitive labor market if no unions
Assumptions continued
A8 Workers and vacancies are matched randomly. The matching
function measures the rate of contacts between v vacancies, u
unemployed and 1 − u employed:7
M(v , u, 1 − u).
By [A3] and [A4],
M(v , u, 1 − u) = M(v , 1).
Furthermore, the flow of contacts verifies:
λ · (u + 1 − u) = M(v , 1).
So, one can write λ = λ(v ). With standard assumptions about
function M, λ(v ) is increasing and concave. The Inada conditions
are assumed (λ(0) = 0; λ0 (0) = +∞).
7
v , u and 1 − u are rates, the denominator being the normalised size of the labor
force.
ESL (UCL)
66 / 110
Non competitive labor market if no unions
Given the similarities with the equilibrium search model, the
presentation will be brief and focused on differences.
u=
q
q + λ(v )
(27)
If H(w) is the CDF of the wage offers,
the fraction of those employed at a wage w or less, i.e. the wage CDF
G(w) verifies:
G(w) =
ESL (UCL)
q · H(w)
with H(w) ≡ 1 − H(w)
q + λ(v ) · H(w)
(28)
67 / 110
Non competitive labor market if no unions
Expected discounted profit
The cdf H(·) being given
The expected discounted profit of a filled job paid w is denoted Πe (w).
)
Let m(v ) = λ(v
v be the contact rate per vacancy and h be the flow cost
of posting a vacancy. Since firms set wages, the expected discounted
profit of a vacant job Πv
r Πv = max{−h + m(v ) [u · 1 + (1 − u)G(w)] (Πe (w) − Πv )},
w≥z
(29)
where
r Πe (w) = y − w + q + λ(v ) · H(w) (Πv − Πe (w))
(30)
where y is the marginal product (y > z).
The higher the wage paid, the bigger the retention rate λ · H(w) and
the lower the instantaneous profit.
ESL (UCL)
68 / 110
Non competitive labor market if no unions
Free entry
In equilibrium with free entry of vacancies, Πv = 0.
This property can then be introduced in (29) and (30) to yield two
expressions for Πe (w).
Equating them and using (27) and (28) lead to
y −w
h·v
q
h
·
(31)
=
= max
w≥z q + λ(v ) · H(w) r + q + λ(v ) · H(w)
m(v )
λ(v )
A steady-state equilibrium is a vacancy rate v and a wage offer
distribution H such that the value of hiring workers is optimal and equal
for every wage of the support of H.
In what follows, the equilibrium vacancy rate is characterised and then
the distribution and the support of H are defined.
ESL (UCL)
69 / 110
Non competitive labor market if no unions
Equilibrium vacancy rate
At the wage floor z, unemployed workers accept job offers since x = 0.
The optimal lower bound of the support of the wage distribution is
therefore z with H(z) = 0.
For w = z, (31) can be written as an implicit equation in v (conditional
on z), namely:
(y − z)λ(v )
q
(32)
h·v =
q + λ(v ) r + q + λ(v )
There are two equilibria:
v1 = 0 This solution is unstable since a small increase in v raises
the return to vacancy creation more than the cost
and v2 > 0, which is stable
ESL (UCL)
70 / 110
Non competitive labor market if no unions
Stable solution
h.v
R.H.S. v1
ESL (UCL)
v2
v
71 / 110
Non competitive labor market if no unions
Equilibrium wage offer distribution
given v = v2
There is an equilibrium non degenerate wage distribution on a support
[z, w].8
Any equilibrium wage offer w ≥ z in the support of H must yield the
same profit.
∀w ∈ [z, w], H(w) solves:
q
y −w
q
y −z
·
=
·
,
q + λ(v2 ) r + q + λ(v2 )
q + λ(v2 ) · H(w) r + q + λ(v2 ) · H(w)
where w is explicitly given by this equality since H(w) = 1.
Note: The difference between contractual wages (w) and the wage
floor (z) has been documented: see e.g. Cardoso and Portugal (2005).
8
Here, w has the meaning that should not be confused with "wages in other firms"
used earlier in these slides.
ESL (UCL)
72 / 110
Non competitive labor market if no unions
Implications of the wage floor z
1) On employment
Looking at the graphical representation of (32),
for any v , a rise in z lowers the curve on the R.H.S. of (32).
Hence, necessarily a higher wage floor lowers v2 .
So, λ(v2 ) shrinks as well and hence the unemployment rate rises.
6= conclusion from the partial equilibrium analysis of Falch and Strom
(2007).
But, here, contrary to the latter paper, total labor supply is exogenously
fixed.
2) On efficiency
Assume the simplification r → 0.
With risk-neutral agents and in the absence of any value to
unemployment, a benevolent utilitarian planner would choose v so as
to maximize output net of recruiting costs.
ESL (UCL)
73 / 110
Non competitive labor market if no unions
Wage floor and efficiency
This planner then solves the following problem:
max y (1 − u) − h · v = y
v
λ(v )
−h·v
q + λ(v )
(33)
Let v ∗ denote a solution to this problem. v ∗ verifies:
y
qλ0 (v ∗ )
=h
[q + λ(v ∗ )]2
(34)
The assumption λ00 (v ) < 0 is a sufficient condition.
ESL (UCL)
74 / 110
Non competitive labor market if no unions
Is the decentralized economy efficient?
Let us compare
v ∗ to the decentralised equilibrium v2 without wage floor z = 0.
Condition (32) evaluated at r = z = 0 becomes:
h · v2 =
q · y · λ(v2 )
(q + λ(v2 ))2
(35)
Combining (34) and (35) yields:
y
qλ0 (v ∗ )
qλ(v2 )/v2
qλ0 (v2 )
=
h
=
y
>
y
[q + λ(v ∗ )]2
[q + λ(v2 )]2
[q + λ(v2 )]2
(36)
given the assumptions about the function λ(·).
qλ0 (v )
Now, since [q+λ(v
is a decreasing function of v , one can conclude
)]2
that
v2 > v ∗
ESL (UCL)
75 / 110
Non competitive labor market if no unions
The decentralized economy is inefficient
Put another way, in the decentralised equilibrium without a wage floor,
too many vacancies are created.
The intuition for this result is the following. Vacancies that are created
to hire unemployed people have a social value (they generate a social
gain of y ).
However, vacancies are also created to attract employed people in
better paid jobs. The social gain of this is zero.
That is the reason why the laissez-faire economy (i.e. monopsonistic
competition when z = 0) creates too many vacancies.
ESL (UCL)
76 / 110
Non competitive labor market if no unions
A well-chosen wage floor is efficient
A positive wage floor z ∗ equal to
v ∗ λ0 (v ∗ )
y 1−
λ(v ∗ )
would be needed to induce the optimal number of vacancies (and the
optimal unemployment rate). Mortensen (2000) concludes that
∗ 0
∗
λ (v )
Available empirical estimates of the elasticity [ v λ(v
∗) ]
suggests a value for the ratio z/y somewhere in the range
between 40% and 60%. (Mortensen (2000), p. 288)
Implementation of z:
A union could fix an appropriate wage floor;
An equivalent legal minimum wage could be chosen by the State.
Notice that the union could be better informed about y and the function
λ than the State.
ESL (UCL)
77 / 110
Non competitive labor market if no unions
3. Monopoy union and matching frictions
Cahuc, Carcillo and Zylberberg (2014) p. 605-6
Consider a centralized monopoly union that cares for the well-being of
the unemployed by maximizing rVu in steady state. This union
chooses the wage for all workers (matches). Using notations of the
matching model,
rVe = w + q (Vu − Ve ) ⇒ Ve =
w + qVu
r +q
The latter can be substituted in
rVu = z + θm (θ) (Ve − Vu ) ⇔ (r + θm (θ)) Vu = z + θm (θ) Ve
Then,
rVu =
z(r + q) + w θm(θ)
.
r + q + θm(θ)
(37)
Interpret!
ESL (UCL)
78 / 110
Non competitive labor market if no unions
The union is supposed to maximize (37) wrt w subject to the vacancy
supply curve (a right-to-manage assumption) :
w =y−
h(r + q)
m (θ)
(38)
where m(θ) is the rate at which vacant jobs are filled.
Substituting this relationship into (37), the problem becomes a
maximization wrt to θ of
z(r + q) + y θm(θ) − h(r + q)θ
r + q + θm(θ)
Check that the FOC of this maximization leads to the following implicit
equation for tightness :
(y − z)(1 − η)
h
=
,
r + q + θm(θ)η
m(θ)
(39)
where η is the absolute value of the elasticity of m(θ) with respect to
θ, 0 < η < 1.
ESL (UCL)
79 / 110
Non competitive labor market if no unions
Message: When r → 0, Eq. (39) is the condition characterizing θ when
the social planner maximizes net output in steady state.
Why does the monopoly union endowed with objective (37) maximize
net output?
Because, when r → 0, rVu is equivalent to net output under the
Beveridge curve and along the vacancy supply curve (38):
rVu = z
q
θm(θ)
q
+y
− hθ
= y + [z − y − hθ]u.
q + θm(θ)
q + θm(θ)
q + θm(θ)
Pissarides (1986) develops this point for any value of r . Then, he
shows that if the union (also) takes care of the utility of the employed
by maximizing
αVu + (1 − α)Ve , 0 ≤ α ≤ 1,
Then the monopoly union wage (and hence the unemployment rate)
will be inefficiently high if r > 0 (because when r → 0, Ve → Vu ).
ESL (UCL)
80 / 110
Non competitive labor market if no unions
Generalization with homogenous workers
Insider-outsider model with matching frictions
Consider a union that only cares about “insiders” whose intertemporal
value is Ve .
Assume that the inside option during if the negotiation is interrupted is
equal to Vu
♦ If the marginal product of labor is constant, so that Πe (resp., Πv )
designates the intertemporal value of any filled (resp., vacant) position,
Then the collective wage bargain is arguably solving
max (Ve − Vu )γ (Πe − Πv )1−γ
w
i.e. the same problem as under individual bargaining in Chap. 9 of
CCZ2014. So, the efficient bargaining power should be given by the
Hosios condition. Extended out of steady state by Krusell and Rudanko (2016).
♦ If the marginal product of labor is decreasing, this is no more true.
See e.g. Bauer and Lingens (2014).
ESL (UCL)
81 / 110
Empirical evidence
Sec. 3. Empirical evidence
Main message of empirical research:
Let “bargaining status” mean
- Unionized vs non-unionized worker (US, UK,...),
- Firms or workers covered by collective agreements or not (Western
Europe),
- Centralized vs decentralized bargaining, or ...
Main message of the book:
Hard to find clear-cut conclusions about the effects of “bargaining
status”.
Only some papers exploit (quasi-natural) experiments and
produce convincing evidence about the causal effect of
unionization. Examples: DiNardo and Lee (2004), Lee and Mas
(2012) in specific contexts.
ESL (UCL)
82 / 110
Empirical evidence
Why?
1
Selectivity: Confound “bargaining status” with relevant
- unobserved worker characteristics (worker’s or firm’s choice);
- unobserved firm-level characteristics;
Panel data allow to deal with time-invariant unobserved
individual/firm effects.
2
The “bargaining status” is an endogenous variable (e.g. being or
not a union member is related to the wage hikes a union may
obtain).
Panel data: Identification of the effect of “bargaining status”
possible thanks to individuals who change status. Exogeneity of
this change is dubious.
3
To test hypotheses, one often has to make assumptions about
unions’ preferences and revenue functions. Some tests are
sensitive to the specific assumption made by the researcher.
ESL (UCL)
83 / 110
Empirical evidence
Impact on wages
Common wisdom (Anglo-Saxon countries): Unionized workers earn
higher wages than similar non-unionized ones.
This view has been challenged by DiNardo and Lee (2004), who
develop a convincing identification strategy.
Specificity of US rules concerning union recognition (more details on p.
1389):
At the initiative of a group of employees (helped by a labor union),
the National Labor Relations Board holds (under certain
conditions) a secret-ballot election at the work site.
A simple majority (50% plus one vote) is required to recognize the
union in a given firm.
Then the union becomes “the exclusive bargaining agent for the
unit, and the employer is obligated to negotiate ‘in good faith’ with
that union.” (p.1389)
ESL (UCL)
84 / 110
Empirical evidence
Pro-union vote share (actual or would be)
and collective bargaining power.
1394
QUARTERLY JOURNAL OF ECONOMICS
FIGURE I
Theoretical Relation between Employer Outcome and Vote Share
ESL (UCL)implicit expectation that wages and benefits will be higher fol-
85 / 110
Empirical evidence
ESL (UCL)
86 / 110
Empirical evidence
Regression-discontinuity designs (RDD)
Formal treatment
Let
y = the outcome of interest in each firm (e.g. the average wage);
D = 1 if the union is recognized (otherwise D = 0);
V = the pro-union vote share in the election;
X = observable predetermined characteristics determining V and
y;
ε and u are unobservable determinants;
β = the parameter of interest.
= Xγ + Dβ + ε
1 if V > 1/2
D =
0 otherwise
V = Xδ + u
y
ESL (UCL)
(40)
(41)
(42)
87 / 110
Empirical evidence
Identifying assumptions
OLS essentially computes
E [y | X = x, D = 1] − E [y | X = x, D = 0]
which is biased if ε and u are correlated so that
E [ε | V > 0.5] − E [ε | V ≤ 0.5] 6= 0
Identifying assumptions for RDD:
1
There is some ex ante uncertainty in the vote share (u).
2
The density of u (and hence of V) conditional on X and ε is
continuous (at the threshold V = 0.5)
Then (by Bayes rule), the density of (X, ε) conditional on V is
continuous at V = 0.5.
ESL (UCL)
88 / 110
Empirical evidence
Consequently, it can be shown that
lim [E (y |V = 0.5 + ∆) − E (y |V = 0.5 − ∆)] = β
∆→0+
(43)
Internal validity: If the relation between X and the vote share is
discontinuous (around V = 0.5), the assumptions are not valid.
Two ways of presenting RDD evidence:
1
Graphical plots of E (y |V) and E (X|V) as a function of vote share
categories (“bins”);
2
Approximating E (y |V) and E (X|V) by flexible polynomials with an
intercept shift at V = 0.5 and estimating the parameters.
Here, only the first approach is considered and I only look at E (y |V).
ESL (UCL)
89 / 110
Empirical evidence
Evidence in the manufacturing industry
Unit of observation = a manufacturing establishment.
Period of observation: 1984-2001.
First, DiNardo and Lee (2004) report evidence that barely winning an
election has a lasting impact on legal recognition of the union:
Almost no recognition (neither immediately, nor in the coming
years) when V ≤ 0.5;
Almost always a (rapid) recognition when V > 0.5. ( MORE ).
Second, they provide evidence on the impact of recognition on wages:
Solid circles = the means of the hourly wage by union vote share
category for establishment-year observations in the years that
follow the election;
Open circles = the same but before the year of the election
(“placebo test”);
Solid triangles = the means of post-election wage deviated from
the establishment-specific mean during years before the election.
( MORE )
ESL (UCL)
90 / 110
Empirical evidence
BACK
ESL (UCL)
91 / 110
Preelection Mean 28,785. For definition of preelection and postelection periods,
see note to Figure VIII.Empirical evidence
BACK
FIGURE IXb
Log(Production Hourly Wage), Pre- and Postelection,
by Union Vote Share, LRD
Note: Observations: Preelection 38,870, Postelection 28,929, Postelection minus
Preelection Mean 28,790. For definition of preelection and postelection periods,
see note to Figure VIII.
ESL (UCL)
92 / 110
Empirical evidence
In the paper
There are many robustness checks.
Some other outcomes are also considered (example below).
Interpretation:
Small effects that cannot be detected by their research design;
The effect of union recognition is truly non significantly different
from zero. The authors argue that this is the right interpretation.
Open issues
Here, the measured effect is identified at the 50% threshold.
What if the union is recognized with a much higher share of votes
(“stronger unions”)?
ESL (UCL)
93 / 110
Empirical evidence
FIGURE VIIIa
Observation of a Contract Expiration Notice, Pre- and Postelection,
by Union Vote Share, LRD
FIGURE VIIIb
Log(Production Hours/1000), Pre- and Postelection, by Union Vote Share, LRD
Note: Observations: Preelection 38,870, Postelection 28,929, Postelection minus
Preelection Mean 28,790. Preelection period include the years of observation in
the LRD that are strictly before the year of the election. Postelection period
include the years that are in the same year or later than the year of the election.
ESL (UCL)
94 / 110
Empirical evidence
Effects on employment
Right-to-Manage vs Efficient bargaining.
(Weakly) efficient bargaining with a general union objective Vs (w, L)
leads to the following contract curve:
∂Vs /∂L
R 0 (L) − w
=−
,
L
∂Vs /∂w
(44)
(w)
the right-hand side being − v (w)−v
if Vs is assumed to be (1).
Lv 0 (w)
Empirical strategy:
With many specifications of Vs (but not all!), (44) defines an
implicit relationship between L, w and w. Then, one specifies an
econometric equation of the type (i (t) designating the plant (time))
log Lit = α0 + α1 zit + α2 log wit + α3 log w it + εit
(45)
where zit captures factors affecting in particular the marginal
revenue function R 0 (·).
ESL (UCL)
95 / 110
Empirical evidence
Next, one needs to make assumptions on how to measure w it .
Notice that under the right-to-manage assumption, conditional on the
wage, the employment level is not affected by the outside option w it .
→ Fragile conclusions according to CCZ.
♦ Some papers reject the (weakly) efficient model, others do not.
♦ Recently, Dobbelaere and Mairesse (2013) conclude that imperfect
competition in the product market and (weakly) efficient bargaining in
the labor market is the predominant regime in the 38 French
manufacturing industries (1978-2001).
ESL (UCL)
96 / 110
Empirical evidence
Effects on (un)employment
Direct estimations.
♦ Direct estimations of a link
“Legislation influencing union power” → employment”
(e.g. Thatcher’s reforms in the 1980s in the UK) lead to mixed results.
♦ Using a Regression Discontinuity Design (in the US),
DiNardo and Lee (2004) conclude that union recognition has
insignificant effects on employment.
Sojouner, Grabowski, Frandsen, Chen and Town (2014) find
significant negative effects on staffing levels9 in nursing homes.
9
The mean number of nursing hours per resident day.
ESL (UCL)
97 / 110
Empirical evidence
Effects on (un)employment
Country panel analyses
♦ Bassanini and Duval (2009) (hereafter BD2009):
consider a panel of 20 OECD countries over the period 1982 2003
study the relationship between the unemployment rate and a
range of OECD harmonized indicators capturing in particular
“labor market institutions”.
Difficulty: How to address the potential endogeneity of those
indicators? Is this approach able to identify causal effects?
Here I focus only on the basic specification and results.
In the paper, they deal with interactions between “institutions” and
so on.
ESL (UCL)
98 / 110
Empirical evidence
BD2009
Basic static specification:
X
Uit =
βj · Xitj + χ · OGit + αi + λt + εit , i = 1, ..., 20, t = 1982, ..., 2003
j
(46)
where
Uit = standardized unemployment rate in % among the population
aged 15-64;
Xitj = labor market indicator j in country i and year t ;
OGit = The “output gap”, i.e. the relative gap between observed
GDP and potential GDP;10
αi country fixed-effect, λt time fixed-effect and εit the error term.
10
Potential gross domestic product (GDP) is defined in the OECD’s Economic
Outlook publication as the level of output that an economy can produce at a constant
inflation rate.” (Source:
http://stats.oecd.org/glossary/detail.asp?ID=2094).
ESL (UCL)
99 / 110
Empirical evidence
Institutions Xitj
Source: Bassanini and Duval (2006) and BD2009
On the labor market:
“Replacement rate” = “average unemployment benefit
replacement rate across two income situations (100% and 67% of
APW (Average Production Worker) earnings), three family
situations (single, with dependent spouse, with spouse in work)
and three different unemployment durations (1st year, 2nd and 3rd
years, and 4th and 5th years of unemployment).”
The “tax wedge” expresses the sum of personal income tax and
all social security contributions as a percentage of total labor cost
for a single-earner couple with two children earning 100% of APW
earnings.
“Union density”: Trade union density corresponds to the ratio of
wage and salary earners that are trade union members, divided
by the total number of wage and salary earners (OECD)
ESL (UCL)
100 / 110
Empirical evidence
Institutions Xitj
Sources: Bassanini and Duval (2006) and BD2009
On the labor market (C’ted):
“EPL”: OECD summary indicator of the stringency of Employment
Protection Legislation.
The “High corporatism” dummy variable takes value 1 when
bargaining is centralised or co-ordinated and zero otherwise
(OECD)
On the goods markets:
“PMR”: “OECD summary indicator of regulatory impediments to
product market competition in seven non-manufacturing industries.””
ESL (UCL)
101 / 110
Empirical evidence
ESL (UCL)
102 / 110
Empirical evidence
Messages (significant effects):
1
Positive correlation between the gross (net) replacement ratio and
the unemployment rate;11
2
Idem for the tax wedge;
3
More competition on the goods market (i.e. lower PMR) is
associated with lower unemployment;
4
“Corporatism” is negatively correlated with the unemployment
rate;12
The effect of union density (as a proxy for union power) is often not
significant! Either hard to find a detrimental effect of unions on the
unemployment rate or density is a poor proxy of union power...
11
“in Bassanini and Duval (2006) ... the positive impact of unemployment benefits
on unemployment diminishes and can even collapse in countries that offset their
detrimental effects through extensive active labor market policies.” (BD2009)
12
“Yet, this effect is identified by only four within sample shifts in the type of
bargaining system, and therefore it should be seen as somewhat more
tentative.”(Bassanini and Duval, 2006, p.12)
ESL (UCL)
103 / 110
Empirical evidence
Robust conclusions? The 1998-2008 period.
Unemployment rate at t
ln (ret100)t−1
UBRRt−1
UnionDensityt−1
wcoordt−1
EPLt−1
outputgapt
inflchanget
iratet
Opennesst
R2
Hansen J test
Anderson Rubin F test
OLS
–6.617**
0.055**
0.089
–0.502**
1.084***
–0.440***
–0.113
–0.359***
0.049***
0.90
IV
–22.069***
0.125***
0.005
–0.536***
1.384***
–0.510***
–0.195*
–0.582**
0.053***
0.88
0.3579
0.0269
Table: 21 OECD countries over the period 1998-2008. Significance levels: ∗:
10%, ∗∗: 5%, ∗ ∗ ∗: 1%. p-values of Hansen J over-identification tests and of
Anderson and Rubin F test of significance of endogenous regressors are
provided. Instruments are Taxconsol and Leftism (lagged twice). Source:
Lehmann, Lucifora, Moriconi and Van der Linden (2015).
ESL (UCL)
104 / 110
Empirical evidence
Other effects of unionization
Effects of unions on productivity:
Inspired by Hirschman (1970), Freeman and Medoff (1984) claim that
unions can help enhance firm productivity by reducing turnover rates
and by promoting changes in working methods or production
techniques.
Small positive effect on productivity, at least in the US.
Often fragile conclusions.
Positive effect confirmed by RDD Evidence in US nursing homes
by Sojouner, Grabowski, Frandsen, Chen and Town (2014).
Effects on profits:
According to the identification strategy (Lee and Mas, 2012),
one can get substantial negative effects on firm’s equity value
(Difference in differences method) or negligible effects
(regression discontinuity design like in DiNardo and Lee).
ESL (UCL)
105 / 110
Empirical evidence
Effects on investment (the hold-up problem):
Theory: there should be under-investment in unionized firms if
→ investment is irreversible and
→ the union cannot make a credible commitment not to renegotiate
wages once the equipment has been installed.
Identifying this effect is difficult for the same reasons as above.
• A number of papers conclude that unionization has a negative effect
on firms’ investment (see CCZ, p. 465).
• This conclusion has however been challenged by e.g. Card,
Devicienti and Maida (2013): “Whether there is holdup or not, however,
depends on whether the wage bargaining process allows the firm to
recoup its investment costs before splitting the rents with employees,
and not on rent-sharing per se."
Using matched employer-employee data for an Italian region and
developing an IV strategy, they conclude that “workers receive a share
of the rents that remain after the costs of capital are fully deducted”
(i.e. no hold-up problem).
ESL (UCL)
106 / 110
Empirical evidence
Wages and unemployment
A clear distinction between:
The Phillips Curve (Phillips, 1958), which is a relationship
between wage growth and unemployment; studied with aggregate
time-series methods.
The Wage Curve (Blanchflower and Oswald, 1994), which is a
(logarithmic) relationship between the wage level and
unemployment in the local area; studied with individual data.
An unemployment elasticity of approximately –0.1 is found in many
countries all over the world.
Example: For the UK, Bell, Nickell and Quintini (2002) conclude that
“The long-run elasticity of average regional wages with respect to
regional unemployment is in the range 0.11–0.13 (...). The long-run
elasticity of individual wages with respect to regional unemployment is
around 0.053.”
Why is there such a regularity? This is still not well understood.
ESL (UCL)
107 / 110
Extensions
Incomplete information
Sec. 4. Extensions
Incomplete information
Up to know, information is perfect or “complete”.
The threat of a disagreement is important but an agreement is found
instantaneously. Why would the player wait in the presence of
discounting, loss of production and earnings in case of a strike?
An implication of this is that strikes are nonsense. However, strikes are
part of reality.
Explanations:
Irrational behaviour (“emotions”,...) or bounded rationality;
Asymmetric information (see Kennan, 1986, and Kennan and
Wilson, 1993, for surveys):
Firm’s profitability unobserved by the union ⇒ equilibria with
strikes are then possible (for example, “screening” or “signalling”
equilibria according to specificity of the game; Kennan and
Wilson, 1993, as of p. 55; not compulsory)
ESL (UCL)
108 / 110
Extensions
Incomplete information
Explanations based on asymmetric information have been criticized
because if the length of time between offers and counter-offers goes to
zero, so does the actual duration of a delay (or strike).
Other explanations for strikes under complete information:
If the union has to choose between striking and continuing to work
under the “old wage contract”, there exist multiple
subgame-perfect equilibria, some of which with a strike (according
to Fernandez and Glazer, 1991)
A preemptive action, to the extent that it is credible, to confer a
bargaining advantage (Appelbaum, 2008).
ESL (UCL)
109 / 110
Extensions
Incomplete information
Enlarging the view about human beings can also be a way of
interpreting strikes:
“Revolt takes many forms. Sometimes it stems from
desperation; there is nothing to lose. Sometimes it stems
from a dying moment, when the tide of history is drowning the
losers, when just standing up is an act of defiance. The
miners’ strike in Britain in 1984 was like that. It was
resistance against loss of a way of labouring that had turned
adversity into a community of shared identity.” (Preface of
Standing (2016))
ESL (UCL)
110 / 110
References
Appelbaum, E. (2008) “Strategis militancy and the probability of
strikes in union-firm bargaining”.
Labour Economics, 15:315–333.
Bassanini, A. and R. Duval (2006) “The determinants of
unemployment across OECD countries: Reassessing the role of
policies and institutions”.
OECD Economic Studies, 42:7–86.
Bassanini, A. and R. Duval (2009) “Unemployment, institutions,
and reform complementarities: re-assessing the aggregate
evidence for OECD countries”.
Oxford Review of Economic Policy, 25(40–59).
Bauer, C. and J. Lingens (2014) “Does collective wage bargaining
restore efficiency in a search model with large firms?”
Economic Journal, 124:1066–2014.
Bell, B., S. Nickell and G. Quintini (2002) “Wage equations, wage
curves and all that”.
ESL (UCL)
110 / 110
References
Labour Economics, 9:341—360.
Bewley, T. F. (1998) “Why not cut pay?”
European Economic Review, 42:459–490.
Blanchard, O. and F. Giavazzi (2003) “Macroeconomic effects of
regulation and deregulation in goods and labor markets”.
Quarterly Journal of Economics, 118:879–907.
Blanchflower, D. and A. Oswald (1994) The wage curve.
MIT Press.
Boeri, T., A. Brugiavini and L. Calmfors (2001) (Eds.), The role of
unions in the 21st century.
Oxford University Press, Oxford, UK.
Booth, A. (1984) “A public choice model of trade union behaviour
and membership”.
Economic Journal, 94:883–898.
Booth, A. (1995) The economics of the trade union.
ESL (UCL)
110 / 110
References
Cambridge University Press, Cambridge.
Booth, A. (1997) An analysis of firing costs and their implications
for unemployment policy.
In D. Snower and G. de la Dehesa, editors, Unemployment policy :
Government options for the labour market. Cambridge University
Press, Cambridge.
Booth, A., M. Burda, L. Calmfors, D. Checchi, R. Naylor and
J. Visser (2000) “What do unions do in europe? prospects and
challenges for union presence and union influence”.
Working Paper, Fondazione Rodolpho Debenedetti, Italy.
Cahuc, P., S. Carcillo and A. Zylberberg (2014) Labor economics.
MIT Press, Cambridge.
Cahuc, P. and A. Zylberberg (2004) Labor economics.
MIT Press, Cambridge.
ESL (UCL)
110 / 110
References
Cai, X., P. Gautier, C. Teulings and M. Watanabe (2014) “Collective
versus decentralized wage bargaining and the efficient allocation
of resources”.
Labour Economics, 26:34–42.
Card, D., F. Devicienti and A. Maida (2013) “Rent-sharing, holdup,
and wages: Evidence from matched panel data”.
Review of Economic Studies, doi:10.1093/restud/rdt030.
Cardoso, A. and P. Portugal (2005) “Contractual wages and the
wage cushion under different bargaining settings”.
Journal of Labor economics, 23(4):875–902.
DiNardo, J. and D. S. Lee (2004) “Economic impacts of new
unionization on private sector employers : 1984–2001”.
Quarterly Journal of Economics, 119:1383–1441.
Dobbelaere, S. and J. Mairesse (2013) “Panel data estimates of
the production function and product and labor market
imperfections”.
ESL (UCL)
110 / 110
References
Journal of Applied Econometrics, 28(1):1–46.
Donado, A. and K. Wälde (2012) “How trade unions increase
welfare”.
Economic Journal, 122(1010):990–1009.
Dunlop, J. (1944) Wage determination under trade unions.
Macmillan, New-York.
Espinosa, M. and C. Rhee (1989) “Efficient wage bargaining as a
repeated game”.
Quarterly Journal of Economics, 49:565–588.
Falch, T. and B. Strom (2007) “Wage bargaining and monopsony”.
Economics Letters, 94:202–207.
Fehr, E. and A. Falk (1999) “Wage rigidities in a competitive
incomplete contract market. an experimental investigation”.
Journal of Political Economy, 107:106–134.
ESL (UCL)
110 / 110
References
Felbermayr, G. and J. Prat (2011) “Product market regulation, firm
selection and unemployment”.
Journal of the European Economic Association, 9(2):278–317.
Fernandez, R. and J. Glazer (1991) “Striking for a bargain between
two completely informed agents”.
American Economic Review, 81:240–252.
Freeman, R. and J. Medoff (1984) What do unions do?
Basic Books, New-York.
Griffith, R., R. Harrison and G. Maccartney (2007) “Product market
reforms, labour market institutions and unemployment”.
Economic Journal, 117(C142–C166).
Hirschman, A. (1970) Exit, voice and loyalty.
Harvard University Press, Cambridge (MA).
Kennan, J. (1986) The economics of strikes.
ESL (UCL)
110 / 110
References
In O. Ashenfelter and R. Layard, editors, Handbook of Labor
Economics, volume 2, pp. 1091–1137. North-Holland.
Kennan, J. and R. Wilson (1993) “Bargaining with private
information”.
Journal of Economic Literature, 31(45–104).
Krueger, A. and A. Mas (2004) “Strikes, scabs, and tread
separations: Labor strife and the production of defective
bridgestone/firestone tires”.
Journal of Political Economy, 112(2):253–289.
Krusell, P. and L. Rudanko (2016) “Unions in a frictional labor
market”.
Working Paper 16-07/R, Federal Reserve Bank of Philadelphia.
Lee, D. and A. Mas (2012) “Long-run impacts of unions on firms:
New evidence from financial markets, 1961-1999”.
Quarterly Journal of Economics, 127:333–378.
ESL (UCL)
110 / 110
References
Lehmann, E., C. Lucifora, S. Moriconi and B. Van der Linden
(2015) “Beyond the labour income tax wedge: The
unemployment-reducing effect of tax progressivity”.
International Tax and Public Finance, Forthcoming:1–36.
Manning, A. (1994) “How robust is the microeconomic theory of
the trade union ?”
Journal of Labor Economics, 12(3):430–459.
Mas, A. (2008) “Labour unrest and the quality of production:
Evidence from the construction equipement resale market”.
Review of Economic Studiesv, 75(229–258).
Mortensen, D. (2000) Equilibrium unemployment with wage
posting: Burdett-Mortensen meet Pissarides.
In H. Bunzel, B. Christensen, P. Jensen, N. Kiefer and
D. Mortensen, editors, Panel data and structural labour market
models. Elsevier Science Publisher (North-Holland), Amsterdam.
ESL (UCL)
110 / 110
References
Nickell, S. and R. Layard (1999) Labor market institutions and
economic performance.
In O. Ashenfelter and D. Card, editors, Handbook of Labor
Economics, Volume 3C. Elsevier Science Publisher
(North-Holland), Amsterdam.
Osborne, M. (2004) An introduction to game theory.
Oxford University Press.
Phillips, A. (1958) “The relation between unemployment and the
rate of change of money wage in the United-Kingdom, 1861-1957”.
Economica, 25:283–299.
Pissarides, C. (1986) “Trade unions and the efficiency of the
natural rate of unemployment”.
Journal of Labor Economics, 4(4):582–595.
Reynolds, L. (1951) The structure of labor markets.
Harper and brothers, New-York.
ESL (UCL)
110 / 110
References
Rosen, S. (1986) The theory of equalizing differentials.
In O. Ashenfelter and R. Layard, editors, Handbook of Labor
Economics, Volume 1. North-Holland, Amsterdam.
Ross, A. (1948) Trade union wage policy.
University of California Press, Berkeley and Los Angeles.
Rubinstein, A. (1982) “Perfect equilibrium in a bargaining model”.
Econometrica, 50:97–109.
Sojouner, A., D. Grabowski, B. Frandsen, M. Chen and R. Town
(2014) “Impacts of unionization on quality and productivity:
Regression discontinuity evidence from nursing homes”.
Working Paper 8240, IZA Institute for Labor.
Standing, G. (2016) The corruption of capitalism: Why rentiers
thrive and work does not pay.
Biteback Publishing.
ESL (UCL)
110 / 110
References
Teulings, C. and J. Hartog (1998) Corporatism or competition?
labour contracts, institutions, and wage structures in international
comparison.
Cambridge University Press, Cambridge.
Van der Linden, B. (2002) “Is basic income a cure for
unemployment in unionied economies? a general equilibrium
analysis”.
Annales d’Economie et de Statistique, 66:81–105.
ESL (UCL)
110 / 110