The costs of climate policies in a second-best world with labour market imperfections

Abstract

This article explores the critical role of labour market imperfections in climate stabilisation costs formation. To do so, we use a dynamic recursive energy-economy model that represents a second best world with market imperfections and short-run adjustments constraints along a long-term growth path. We show that the degree of rigidity of the labour markets is a central parameter and we conduct a systematic sensitivity analysis of the model results to this parameter. When labour markets are represented as highly flexible, the model results are in the usual range of existing literature, i.e. less than 2% GDP losses in 2030 for a stabilisation target at 550ppm CO 2 equivalent. But when labour markets rigidities are accounted for, mitigation costs increase dramatically. In a second time, the article identifies accompanying measures, namely labour subsidies, which guarantees against the risk of large stabilisation costs in the case of high rigidities of the labour markets. That vision complements the usual view that mitigation is a long-term matter that depends on technology, innovation, investment and behavioural change. Here we add the warning that mitigation is also a shorter-term issue and a matter of transition on the labour market.

Full text

Page 1

1
The costs of climate policies in a second best world with labour market imperfections

Céline Guivarcha,*, Renaud Crassousa, Olivier Sassia and Stéphane Hallegattea,b
a Centre International de Recherche sur l’Environnement et le Développement, Nogent-sur-Marne, France
b Ecole Nationale de la Météorologie, Météo-France, Toulouse, France
* Corresponding author at: CIRED, 45bis, Av. de la Belle Gabrielle, F-94736 Nogent-sur-Marne, France.
Tel.: +33 1 43 94 73 86; fax: +33 1 43 94 73 70.
E-mail address: guivarch@centre-cired.fr (C. Guivarch).


Abstract

This article explores the critical role of labour market imperfections in climate stabilisation costs
formation. To do so, we use a dynamic recursive energy-economy model that represents a second best
world with market imperfections and short-run adjustments constraints along a long-term growth path.
We show that the degree of rigidity of the labour markets is a central parameter and we conduct a
systematic sensitivity analysis of the model results to this parameter. When labour markets are
represented as highly flexible, the model results are in the usual range of existing literature, i.e. less
than 2% GDP losses in 2030 for a stabilisation target at 550ppm CO2 equivalent. But when labour
markets rigidities are accounted for, mitigation costs increase dramatically. In a second time, the
article identifies accompanying measures, namely labour subsidies, which guarantees against the risk
of large stabilisation costs in the case of high rigidities of the labour markets. That vision complements
the usual view that mitigation is a long-term matter that depends on technology, innovation,
investment and behavioural change. Here we add the warning that mitigation is also a shorter-term
issue and a matter of transition on the labour market.

Keywords: labour market rigidities, climate policies costs, second best world, sensitivity analysis.

Page 2

2
Introduction

There is now a global scientific consensus that halving current world GHG emissions before 2050 is
necessary to limit the high risks associated with anthropogenic climate change, and this consensus is
flagged as a long run collective target by an increasing number of policymakers. Such a mitigation
target will necessarily require deep structural and technological change in the economic production
system and in the use of energy, materials and lands around the world. According to the last IPCC
report, one can be confident about the feasibility of this challenge thanks to existing or future
technologies: the global macroeconomic mitigation cost is estimated below five percent of GDP in
2050, even for the most stringent emission constraints (IPCC 2007). However these IPCC results are
submitted to a critical caveat that is precised in Box SPM-3: ‘Most models use a global least cost
approach to mitigation portfolios and with universal emissions trading, assuming transparent markets,
no transaction cost, and thus perfect implementation of mitigation measures throughout the 21st
century’. Face to these first-best assumptions, one may wonder whether the imperfections of the real
world are likely to weaken the robustness of the range of GDP variations due to ambitious mitigation
policies.
A significant body of literature is devoted to the analysis of existing market imperfections that are
known to raise barriers to the adoption of optimal behaviours and the diffusion of efficient
technologies. Notably, the topic has been at the heart of the debate about ‘no-regret potentials’ that
opposed top-down and bottom-up modellers in the nineteen’s. While it remains uneasy to identify and
assess market imperfections, comparison of the results of optimal planning models on one side and
market simulation models1 on the other side can convey rough ideas on the magnitude of the existing
barriers to changes in technology and behaviour.
Beyond that, there may be other imperfections, out of the technological sphere, but also critical, for
example on labour, capital and other non energy markets. Disappointingly, those imperfections are
insufficiently analysed in the economic literature about climate policies, even though they are likely to
condition the efficiency and the net cost of climate policies. The ‘visible part’ of the iceberg includes
carbon leakages (that hangs on both trade and capital flows) and investment crowding-out. Labour
market imperfections were also considered in the past but have been almost underwater for a decade.
Indeed, one has to rewind to the double dividend controversy that occurred during the nineties to find
several articles that explored the links between mitigation policies and labour market dynamics
(Bovenberg and van der Ploeg 1994); (Welsch 1996); (Carraro et al 1996); (Bovenberg and van der
Ploeg 1996) (Bovenberg 1999). The general issue was then to determine to what extent the
replacement of a fraction of payroll taxes by a CO2 tax could generate an extra economic benefit that

1 However it should be noticed that the status of most simulation models is unclear, since they represent perfect
markets with no imperfection, but calibrated on real imperfect markets. Their output is then unclear, somehow
hybrid between the economic potential –what is profitable- and the market potential -what is achievable with the
current conditions.

Page 3

3
could eventually offset the costs of mitigation. The answer has not been really cleared up, as the
assessment of the magnitude of the double dividend critically depends on the intricate representation
of the labour market in models.
Since then, few contributions tried and study quantitatively the importance of labour market
imperfections on the costs of climate policies. To our knowledge, the study of Babiker and Eckaus
(2007) is the most developed, in the sense that the authors (i) recognized the importance of the issue,
(ii) modified the EPPA model to assess the impact of sector-specific labour and wage rigidities, (iii)
demonstrate that such imperfections may increase mitigation costs, (iv) eventually show that
additional appropriate labour policies, namely outplacement assistance and wage subsidies, could
offset the cost increase. It may appear surprising that, after this innovative demonstration that labour
market dynamics are critical in the assessment of climate policies2, the EPPA team never used again
the modified version of the model in the dozens of that followed.
By shelving the issue, one runs the risk of missing the appropriate parallel policies that could reconcile
employment and climate, and finally ease a safe stabilization of our climate. This article tries and ‘re-
open the box’ to explore the critical role of labour market imperfections in climate policy costs. Our
approach differs from the previous ones since labour and capital market imperfections are intrinsically
represented in our model, Imaclim-R. This model is a dynamic recursive energy-economy model, built
to handle three well-known methodological challenges: (i) consistently hybridizing disaggregated
technical potentials with general equilibrium constraints (Hourcade and al. 2006), (ii) representing
short-run adjustments constraints along a long term growth path (Solow 2000), and (iii) allowing for
market imperfections, suboptimalities and adaptive expectations3. This new modelling paradigm has
been recognized earlier as producing an important transitional slowdown during two decades after the
start of stringent climate policies (Edenhofer and al. 2006); (Hourcade and Crassous 2008). As these
results are clearly out of the range of other cost assessments in the literature, we have been challenged
to identify the very sources of these costs and to understand what policies could smooth the transition
while achieving the same environmental results. This exploration led us to understand the critical role
of labour market. Technically, in each region of the model, the labour market is modelled through an
aggregate regional wage curve that links real wages to the unemployment rate (Blanchflower and
Oswald 1995).
Assuming that this representation can encapsulate most of labour markets imperfections, we found that
the calibration of the wage curve is a critical parameter that could shift the model from our initial high-

2 In another modelling report, the same team recognized that ‘a third limitation is that, like capital, labour is
treated as being in inelastic supply. This, combined with the full employment assumption that is standard in
many CGE models, implies that the reduction in labour demand associated with the decline in fossil fuel and
energy-using sectors cannot generate unemployment. Instead, the wage falls, allowing the labour market to clear
and surplus labour to move to the rest of the economy, where it is re-absorbed.’ (Wing 2004)

3 IPCC, 2007

Page 4

4
cost simulations to scenarios in which policy costs are within the usual range. It would be a limited
outcome if we had to interpret the sensitivity of the model results to such a hidden and uncertain
parameter as a harmful weakness; on the contrary we intend to stress that this sensitivity reveals the
importance of labour market dynamics on climate policy costs, which is ignored in all models based
on a full utilization of the labour force. While we do not pretend to represent labour market accurately
enough to derive precise policy recommendations, our conclusion is that achieving ambitious
reductions at a reasonable macro cost, namely within the range of costs assessments gathered by the
IPCC, will certainly require specific parallel policies on the labour market. As an example, in the last
part of the article, we test a recycling policy that re-allocate the tax revenue to labour subsidies and we
find that such a policy succeeds in lowering mitigation costs, whatever the calibration of the wage
curve.


This article is organized as follows. The first section comments the literature on the links between
climate policy assessments and the representations of the labour market. The second section describes
the Imaclim-R model we use in this article. The third section presents and comments simulation
results. The last section discusses the results and concludes.

1. Modelling interactions between labour market imperfections and climate policies

1.1 Quick review of the existing models

Almost all numerical models used in climate policy studies assume a perfect labour market and
neglect unemployment issues, even complex computable general equilibrium models whose
comparative advantage is supposed to represent subtle macro feedbacks. In general, labour supply is
equal to active population multiplied by an exogenously increasing productivity, and is fully flexible
across all sectors, so that it always remains fully utilized, with its price equalizing its marginal
productivity in all (CES) production functions.

This representation contrasts with the imperfections of real world labour markets. Indeed, frictions
arise in the labour markets due to geographic immobility, time consuming job search process, specific
skills required for a specific job sector. Moreover wages are not fully flexible: wage rigidities are
linked to work contracts, unions’ power, laws on minimum wage etc. Since Keynes, the study of
labour market rigidities and their macroeconomic implications is one of the central preoccupations of

Page 5

5
macroeconomists4. The attention devoted to this issue in the macroeconomic field and the refinement
of theories and representations of labour market rigidities proposed in the literature stand out against
the very poor representation of labour markets in most energy-economy-environment (E3) models.
Notable exceptions are E3MG (Köhler et al. 2006), SGM (Fawcett and Sands, 2005) and the version
of EPPA mentioned in the introduction (Babiker and Eckaus, 2007). Technically, in E3MG,
employment is output-driven and real wages are linked to the unemployment rate through an
econometric equation; in SGM, labour supply stems from a leisure-labour trade-off, while labour
demand remains elastic, so that there is no involuntary unemployment. The most refined
representation of labour markets is probably in the version of EPPA by Babiker and Eckaus (2007),
which includes the inertia of inter-sector reallocation (a share of the labour force is sector specific in
the short-run) and wage rigidities (the model takes into account a constraint of minimum nominal
wage equal to the calibration year wage). However this representation of labour markets rigidities is
not included in the standard version of EPPA that relies on perfect labour markets.
Our approach differs from the previous one since labour markets imperfections are represented in the
basic version of our model and are, as we will see, critical to explain the magnitude of the climate
mitigation costs it calculates.
The next section details Imaclim-R architecture, in which the labour market is output-driven while real
wages are linked to unemployment through regional wage curves.



1.2 The Imaclim-R model: partial factor utilisation and short-run adjustments

a. Model architecture and major features

IMACLIM-R is a hybrid recursive general equilibrium model of the world economy that is split into 12
regions and 12 sectors (Crassous and al. 2006), (Sassi and al. 2009). It is hybrid in two senses: (1) It is
a hybrid model in the classical sense: its structure is designed to combine Bottom-Up information in a
Top-Down consistent macroeconomic framework. Energy is explicitly represented in both money
metric values and physical quantities so as to capture the specific role of energy sectors and their
interaction with the rest of the economy. The existence of explicit physical variables allows indeed a
rigorous incorporation of sector based information about how final demand and technical systems are

4 The seminal contributions in this field are too numerous to single out a few references in this text. The
interested reader may refer to the recent article by Blanchard and Galí (2007), who analyse the implications of
the introduction of real wage rigidities in the New Keynesian Model and give an overview of recent
developments in the theories and representations of labour markets rigidities.

Page 6

6
transformed by economic incentives. (2) It is hybrid in the sense of Solow (2000)5, i.e. it tries and
bridges the gap between long-run and short-run macroeconomics, as efforts were devoted not only to
model long-term mechanisms but also focus on transition and suboptimal pathways through possible
underutilization of production factors. We seek, indeed, to capture the transition costs with a modeling
architecture that allows for endogenous disequilibrium generated by the inertia in adapting to new
economic conditions due to both imperfect foresight and non flexible characteristics of equipment
vintages available at each period (putty-clay technologies). In the short run, the main available
flexibility lies in the rate of utilization of capacities, which may induce excess or shortage of
production factors, unemployment and unequal profitability of capital across sectors.

Technically, the model can be labelled as ‘recursive dynamic’, since it generates an energy-economy
trajectory by solving successive yearly static equilibria of the economy, interlinked by dynamic
modules. Within the static equilibrium, domestic and international markets for all goods – except
factors such as capital and labour – are fully cleared by a unique set of relative prices that depend on
the behaviours of representative agents on the demand and supply sides. The calculation of this
equilibrium determines the following variables: relative prices, wages, labour, quantities of goods and
services, value flows.
Within each yearly static equilibrium, the behaviour of producers is not represented by a flexible
production function allowing for substitution between factors. These substitutions are treated between
two equilibria in sector-specific dynamic modules. Producers are therefore assumed to operate under
short-run constraints of (i) a fixed maximal production capacity Capk,i, defined as the maximum level
of physical output achievable with the equipment built and accumulated previously, and (ii) fixed
input-output coefficients representing that, with the current set of embodied techniques, producing one
unit of a good i in region k requires fixed physical amounts ICj,i,k of intermediate goods j and lk,i of
labour. In this context, the only margin of freedom of producers is to adjust the utilisation rate
Qk,i/Capk,i according to the relative market prices of inputs and output, taking into account increasing
costs when the production capacities utilization rate approaches one6. This represents a different
paradigm from usual production specifications, since the ‘capital’ factor is not always fully operated.
Between two static equilibria, technical choices are flexible but they modify only at the margin the
input-output coefficients and labour productivity embodied in existing equipment vintages that result
from past technical choices. This general putty-clay assumption is critical to represent the inertia in
technical systems.

5 Solow (2000) : ‘I can easily imagine that there is a « true » macrodynamics, valid at every time scale. But it is
fearfully complicated [...] At the five-to-ten-year time scale, we have to piece things together as best we can, and
look for a hybrid model that will do the job.’
6 Following (Corrado and Mattey, 1997), we assume that this is generally caused by higher labour costs due to
extra hours with lower productivity, costly night work and more maintenance works.

Page 7

7
Our model is calibrated on 2001 data from GTAP 6 database (Dimaranan and McDougall 2002) that
provides, for the year 2001, a set of balanced input-output tables of the world economy, detailed in 87
regions and 57 sectors. We modified the original GTAP-6 dataset (i) to aggregate regions and sectors
according to the IMACLIM-R mapping (ii) to make it fully compatible with the 2001 IEA energy
balances7.

Our model growth engine is composed of exogenous demographic trends (UN World Population
Prospects, medium scenario, United Nations, 2005) and exogenous trends of labor productivity, as in
Solow’s neoclassical model of economic growth (Solow 1956). To build these trends we draw on
stylized facts from the literature, in particular the convergence assumption (Barro and Sala-i-Martin
1992) and two empirical analyses on economic convergence, one investigating the past trends by
Maddison (1995), and the other one looking at future trends, by Martins and al. (2005). We retained a
“leader”, the US, whose labor productivity growth trend lies between 2% today and 1.65% in the long
run. The other regions labor productivity trends catch up with the leader’s, i.e. their labor productivity
growth is higher all the more as their absolute labor productivity is far from the leader’s level.

The two sets of assumptions on demography and technical change, although exogenous, only prescribe
potential growth. Effective growth results endogenously from the interaction of these driving forces
with short-term constraints: (i) available capital flows for investments and (ii) rigidities, such as fixed
technologies, immobility of the installed capital across sectors or rigidities in real wages, which may
lead to partial utilization of production factors (labor and capital). The next section details this last
point.

b. Labour market representation

As already mentioned in the previous section, producers operate in static equilibria with a fixed input
of labour per unit of output. This labour input, corresponding to labour productivity, evolves between
two yearly equilibria following exogenous trends of labour productivity.
Three of the model features explain the possibility of under-utilisation of labour as a production factor,
and thus of unemployment. First, rigidity of real wages, represented by a wage curve as we will see in
the following, can prevent the wages to fall at the level allowing the labour market to clear. In other
terms, the wages are adjusted instantaneously to the economic context in the static equilibrium, but not
in an optimal manner. Second, in the static equilibrium, the fixed technologies (Leontief coefficients
even for labour input) prevent substitution among factors on the short run. And third, the installed

7 This process of building hybrid input-output matrices is very precisely discussed in (Sands et al., 2005).

Page 8

8
productive capital is not mobile across sectors, which creates rigidities in the reallocations of
productions between sectors when relative prices change.

In each region k, each sector employs the labour force lk,i
.Qk,i, where lk,i is the unitary labour input (in
hours worked) and Qk,i the production. The underutilization of the labour force, equivalently referred
to as unemployment rate8 in the following, zk is therefore equal to one minus the ratio of employed
labour force in all sectors over Lk, the total labour if all active population was fully employed:
= −∑
.
,,
1
k i
l
k i
i
k
k
Q
z
L
⋅
No endogenous mobility of workers between regions is allowed in the model, thus twelve separate
labour markets are represented.

We chose to model labour markets imperfections through an aggregate regional wage curve that links
real wage levels to the unemployment rate. This representation is based on theories developed in the
80s and early 90s, when a generation of macroeconomic models arose in which an aggregate wage
curve, or wage setting curve, is the primary distinguishing feature (an overview can be found in
Layard et al., 2005; Lindbeck, 1993; or Phelps, 1992). The novel approach of these models, when
introduced, was to replace the conventional labour supply curve with a negatively-sloped curve linking
the level of wages to the level of unemployment. The interpretation of this wage curve is given either
by the bargaining approach (Layard and Nickell, 1986) or the wage-efficiency approach (Shapiro and
Stiglitz, 1984). Both interpretations lie on the fact that unemployment represents an outside threat that
leads workers to accept lower wages when the threat is important. The bargaining approach emphasize
the role of workers (or unions) power in the wage setting negotiations, power which is weakened when
unemployment is high. The wage-efficiency approach stands from the firms’ point of view and argues
the firms set wage levels so as to discourage shirking; this level is lower when the threat of not finding
a job back after being caught shirking gets higher. The wage curve specification allows the theories to
be consistent with both involuntary unemployment and the fact that real wages fluctuate less than what
the paradigm of the conventional flexible labour supply curve gives. Microeconometric evidence for
such formulation was given in a seminal contribution by (Blanchflower and Oswald 1995).

In practice, the wage curve for each region k in our model is implemented through the relation:
kkk
kkk
wwref
pindref
z
awf
pind zref






=⋅⋅
,

8 Contrary to the definition by the US Bureau of Labour Statistics, the unemployment is here expressed in terms
of worked hours and not in terms of persons.

Page 9

9
where w is the hourly nominal wage level, pind the consumption price index, z the unemployment rate,
ref indexes represent the value at the calibration date. pindref is derived from the final consumption
prices and volumes at the calibration date, wref is calibrated from the total salaries per sector in GTAP
6 database and the shares of labour force per sectors in the International Labour Organisation statistics.
By default, aw is calibrated to 1 and evolves in parallel to the labour productivity so that unitary real
wages are indexed on labour productivity. zref represents the underutilization of the labour force at the
calibration date. f is a function equal to one when the unemployment rate is equal to its calibration
level, and negatively sloped, representing a negative elasticity of wages level to unemployment9.

There remain important uncertainties on the values to give to the elasticity of wages level to
unemployment and to the underutilization of the labour force at the calibration date. By default, we
assume all regions labour markets are identical and set the underutilization of the labour force at 10%10
and the wage curve elasticity at -0.1 for all regions, which is a value emerging from many econometric
studies: (Blanchflower and Oswald 1995), (Blanchflower and Oswald 2005). But, it is uncertain
enough to justify a systematic sensitivity analysis of our model’s results, which is the topic of the
following section.

2. Numerical experiments and results

2.1 Experimental protocol

We use the model described in previous section to perform two sets of scenarios: a set of
‘reference’ scenarios, i.e. without climate policies, and a set of ‘550ppm’ scenarios with climate
policies starting in 2010 represented by carbon pricing so as to fit a given global emissions profile
corresponding to stabilisation target at a concentration of 550ppm all gases.11 This places us in a “cost-
efficiency” context.
For each set of scenarios we run the model with alternative values for the elasticity of the wage curve
exploring a wide interval from zero to high values. For each scenario, the wage setting function is thus
recalibrated so that it has the chosen elasticity at calibration point. Low values represent very rigid
labour markets whereas high values get closer to perfect labour markets. An infinite elasticity would

9 Choosing a functional form and calibrating the function is particularly tricky notably due to the lack of reliable
data to fully inform the functioning of the labour markets worldwide. We chose a function of the form
()()
1tanh
a c z
⋅−⋅
, and calibrate the parameters a and c so as to have the desired value and elasticity at the
calibration point.
10 Obviously, this is a limitation of the current calibration of the model and future developments will look into
the possibility to differentiate labour markets per regions. However, one important difficulty lies in the lack of
reliable data on the underutilization of the labour forces in all regions, in particular due to informal economy,
very diverse accounting rules for unemployment rates and variations in hours worked per person across
countries.
11 By default, the carbon tax revenues are rebated to households in a lump-sum manner.

Page 10

10
be equivalent to the full employment assumption, but it is not compatible with Leontief specification
for the production function. Therefore, we limit the interval and we stop our exploration at the upper
value of seven.

2.2 Results

There are several different concepts of climate mitigation costs: technical cost, macroeconomic cost
and welfare cost (see for instance Hourcade and Ghersi, 2009). This article focuses on the
macroeconomic cost of mitigation, which will be measured through the global GDP variations, as it is
common practice in the IPCC. We do so to be able to compare our model results to the range of
existing costs evaluations reported in the last IPCC report. We, thus, first show the results of the
numerical experiments we conducted in terms of discounted real GDP losses12 over the period 2010-
2050 between each 550ppm scenario and the corresponding reference scenario13.
Figure 1 presents the global discounted GDP variations over the period 2010-2050 between
each of the ‘550ppm’ scenarios, and the corresponding ‘reference’ scenario with the same wage curve
elasticity, plotted against the wage curve elasticity. As it is beyond the scope of this article to discuss
what suitable discount rate should be taken, Figure 1 gives the curves for two contrasted discount rates
that correspond to the values used by the US Office of Management and Budget to analyze policy
decisions (OMB, 2003; see Appendix D, OMB Circular A-4). The discount rate chosen shifts a little
the curve up for higher discount rates, but does not change its shape. It clearly appears that the lower
the wage curve elasticity, the higher the global GDP losses. This result is not surprising considering
that the elasticity of the wage curve determines the balance between the adjustment of the economy in
prices (high elasticity) or in quantities (low elasticity). The wage curve could be interpreted as playing
the role of a spring anchoring the aggregated production quantities through labour intensity to the
value Lk.(1-zrefk), with a constant spring inversely related to the elasticity of the wage curve (the more
the wage curve is elastic, the less the spring is stiff). For a fully flexible wage curve, the aggregate of
production quantities is anchored to the Lk.(1-zrefk) value. The rigidity of the wage curve gives the
possibility to move away from the anchor, thus adding to the impact of the reallocation of production
across sectors on real GDP, an impact on the general level of activity.



12 We measure the real GDP with the Laspeyres index of quantities.

13 Note that we do not have a single reference scenario but one per value of the elasticity of the wage curve
tested. Indeed the elasticity influences the results, GDP growth in particular, in the scenario without climate
policy as well. Therefore each 550ppm scenario has to be compared with the scenario without climate policy but
with the same representation of the labour market, provided that the climate policies do not modify the
functioning of the labour market.

Page 11

11
Global discounted GDP variations
(% Reference scenarios)
-7%
-6%
-5%
-4%
-3%
-2%
-1%
0%
-8-6-4-20
Wage Curve elasticity
Discounted GDP variations
(ppp value)
3%
7%
discount rates

Figure 1 : Global GDP variations discounted over the period 2010-2050 between stabilisation
scenarios and corresponding reference scenarios, depending on the wage curve elasticity, and for two
discount rates (3% and 7%).

When we get close to fully flexible labour markets (high elasticity, in absolute value), costs
are limited, which somehow connects our model’s results to existing literature. Using the IPCC AR4
measures, our model results are ranging from 0.5% to 1.5% global GDP losses in 2030 for the wage
curve elasticity going from -7 to -1, and form 0.9% to 1.6% in 2050 for the same wage curve elasticity
range. It makes our results compatible with the intervals given in the IPCC AR4: 2030 GDP losses
ranging from 0.2% to 2.5% in 2030, and from slightly negative (actual gains) to 4% loss in 2050, with
median values of 0.6% in 2030 and 1.3% in 2050.14 It remains that the high values of the wage curve
elasticity necessary to replicate IPCC results are not compatible with econometrically estimated
values. It is especially true because what matters here is downwards flexibility of wages, which was
evaluated to even lower values than upwards flexibility of wages (Akerlof et al. 1996), (Kahn 1997),
(Altonji and Devereux 1999) and (Dickens et al. 2007).

When moving to lower values, in absolute value, of the wage curve elasticity (below 1), the slope of
the curve is steep and mitigation costs increase dramatically, this interval of wage curve elasticities
explored being more compatible with econometrically estimated elasticities. We acknowledge that

14 Note that the carbon price emerging in our simulations in around 50$/tCO2 in 2030 and around
100$/tCO2 in 2050, with slight variations depending on the value of the wage curve elasticity. This
order of magnitude is in the range given by the IPCC AR4: 30-150$/tCO2 in 2050 for stabilisation
levels between 535 and 590ppm CO2-eq.

Page 12

12
wage curves are only one representation of labour market rigidities among a variety of alternative
representations. Moreover, the value of the wage curve elasticity valid for each countries is still
debated: the value of -0.1 found by Blanchflower and Oswald (1995) for twelve countries has been
confirmed by many studies (Blanchflower and Oswald, 2005 give an impressive list of such studies)
but has been challenged by others, for instance Nijkamp et al. (2005), who show that reported
elasticities do vary, even excluding outlier, between -0.5 and +0.1. However, without taking a stand on
what the “right” representation of labour market rigidities is or on what the “exact” elasticity of the
wage curve is, we show that taking into account labour market imperfections leads to higher
macroeconomic costs of climate mitigation than in the case of perfect labour markets.

Our results are not fully comparable to those of Babiker and Eckaus (2006) because we consider
different climate policies (Kyoto Protocol emissions caps throughout the 2010-2100 horizon in
Babiker and Eckaus, 2006 vs. a 550ppm stabilisation target here) and different ways of representing
labour market rigidities (shares of sector specific labour and minimum nominal wages vs. wage
curves). Nevertheless, we may note that both studies show that labour market imperfections induce
larger macroeconomic costs of climate mitigation than perfect labour markets. The magnitude of the
impact is, however, different. Babiker and Eckaus (2006) study finds relatively small impacts: US
GDP variations between the climate policy scenario and the reference scenario in 2100 equal -3.5% of
the reference GDP without labour market imperfections and -5% with labour market imperfections.
For Japan, the same indicator changes from -10% to -12% when labour market imperfections are taken
into account, whereas it has almost no impact in Europe. Our results conclude to more significant
impacts of labour markets rigidities on mitigation costs: the discounted world GDP loss varies from
0.35% with the wage curve elasticity equal to -7 (close to perfect labour market) to almost 3.5% with
an elasticity equal to -0.1.

Page 13

13
Global GDP variations
(% Reference scenario)
-7%
-6%
-5%
-4%
-3%
-2%
-1%
0%
2010 2015 20202025
2030 20352040
2045 2050
real GDP variations (ppp aggregation)
-0.1 -2
wage curve elasticity calibrated to:

Figure 2 : Global GDP relative variation in ppp values between a stabilisation scenario corresponding
to a 550ppm concentration target and the corresponding reference scenario, for two alternative
elasticities of the wage curve calibrated to -0.1 (black diamonds) and -2 (grey crosses) respectively.

Figure 2 shows the temporal profile of global GDP relative variation between a stabilisation scenario
and the corresponding reference scenario, for two alternative elasticities of the wage curve calibrated
to -0.1 and -2 respectively. It illustrates another particularity in our model’s results: it exhibits
transition costs whereas a majority of models give costs increasing over time. In the last decade of
the time horizon, global GDP partially catches up its baseline level. This partial catch-up is
explained by (i) induced technological change which lowers costs of low carbon techniques
(Crassous et al., 2006) and (ii) lower vulnerability to peak oil in the stabilisation scenario than
in the baseline15. The economic vulnerability to peak oil is linked to the imperfect expectation
of oil price steep increase16 and investment decisions that reveal inadequate when oil
producers are constrained by resources depletion. In stabilisation scenarios, this imperfect
expectation is partially corrected by carbon pricing: technical change and consumption
structure change induced by climate policies reduce the economic dependence on oil and
lower tensions on oil markets. Exploring further the issue of the time profile of GDP losses is
beyond the scope of this paper, but Figure 2 calls into question the common attitude of most

15 Investigating the respective influence of both mechanisms is beyond the scope of this paper and will be
analysed in detail in a follow-up paper.
16 For instance, in the baseline scenario with a wage curve elasticity calibrated to -0.1, the endogenous
international oil prices rise from 85$/bbl in 2040 to 184$/bbl in 2050.

Page 14

14
economists who claim to be ‘neoclassical on the long run’17. When transposed to the field of climate
change this attitude leads to use preferably models based on a full utilization of production factors
along stabilized growth pathways because climate stabilization is interpreted mostly as a long run
challenge while rigidities in the labour market and unemployment are considered as a short to medium
run issue. We argue that this position is restrictive because it does not allow studying the transitional
pathways that lead to a low carbon economy. The goal of climate policies – avoiding a dangerous
climate change – is obviously relevant on the long run only, but climate policies themselves are likely
to impact the short and medium run before delivering their final outcome. We demonstrate here that a
mechanism considered as purely short-run has impacts over the long-term: stronger rigidities in the
labour markets lead to a larger reduction in growth rates over three decades in climate stabilisation
scenarios compared to baseline scenarios.

The previous results indicate that stabilisation costs may race out of control in case of high rigidities
on the labour markets. For us, these results do not mean that ambitious mitigation actions should not
be undertaken because of high mitigation costs, but rather that we should concentrate on how to
reduce these costs if they eventually reveal higher than what classical literature gives. More precisely,
as there is uncertainty on the functioning of the labour markets today and a fortiori on how they will
change over time, accompanying measures are needed to reduce the risk of steep increase of
mitigation costs. To do so, we focus here on measures that tend to offset the rise in production costs
due to carbon pricing and are directed towards labour. In the following, we assess how using carbon
tax revenues to lower payroll taxes or subsidise labour18 changes mitigation costs. Figure 3 shows that
this policy largely offsets GDP losses. The curve of discounted GDP losses as a function of the wage
curve elasticity is now almost flat; the steep increase of stabilisation costs when wages are rigid does
not occur with the tax-recycling policy tested here. Our results are therefore consistent with the
literature on the double dividend and labour market imperfections, which tended to show that the more
flexible the labour market, the lower the magnitude of the double dividend (Welsch 1996). Figure 3
shows that for very low absolute values of the wage curve elasticity, global discounted GDP variations
between the stabilisation scenario and the corresponding reference scenario may even become
positive, in the case of a 3% discount rate. This result is due to the facts that in the long run, the
stabilisation scenario exhibits actual GDP gains compared to the reference and that the lower discount
rate gives more importance to these gains. These GDP gains confirm the fact that our model represents

17 Solow (2000): ‘ I can easily imagine that there is a "true" macrodynamics, valid at every time scale. But it is
fearfully complicated, and nobody has a very good grip on it. At short time scales, I think, something sort of
"Keynesian" is a good approximation, and surely better than anything straight "neoclassical." At very long time
scales, the interesting questions are best studied in a neoclassical framework, and attention to the Keynesian side
of things would be a minor distraction.’

18 By default, the revenues of the carbon tax are rebates to households in a lump-sum manner.

Page 15

15
a second best world, or more precisely that our baselines are not optimal in terms of GDP growth.
Here our results corroborate those of Babiker and Eckaus (2006), who conclude that “if the one type of
interference with the markets is imposed, in this case the imposition of emissions restrictions, and
there are labor market imperfections, an offsetting policy, e.g. wage subsidies, as an example, can
ameliorate, and possibly eliminate the negative effects”.


Global discounted GDP variations
(% Reference scenarios)
-6.0%
-5.0%
-4.0%
-3.0%
-2.0%
-1.0%
0.0%
-8-6-4-20
Wage Curve elasticity
Discounted GDP variations
(ppp value)
with labour subsidies
without labour subsidies
Discount rate 7%


Global discounted GDP variations
(% Reference scenarios)
-7.0%
-6.0%
-5.0%
-4.0%
-3.0%
-2.0%
-1.0%
0.0%
1.0%
-8-6-4-20
Wage Curve elasticity
Discounted GDP variations
(ppp value)
with labour subsidies
without labour subsidies
Discount rate 3%
Figure 3 : Discounted global GDP variations between stabilisation scenarios and corresponding
reference scenarios, depending on the wage curve elasticity, with (grey rounds) or without (black
diamonds) labour subsidies in the stabilisation scenarios, for two discount rates 7% (left panel) and 3%
(right panel).



3. Conclusion

The aim of the study was to understand why our model’s results were at the higher end of existing
mitigation costs estimations. This led us to show that the representation of rigidities in labour markets
is critical in the assessment of mitigation costs. When labour markets are highly flexible, the
mitigation costs are very limited in the model (less than 2% real GDP losses in 2050), which brings
our results back in the usual range of costs assessments. But these results are based on unrealistic
values of the elasticity of the wage curve (especially for downward flexibility). Using more realistic
values of this elasticity leads to a dramatic increase of mitigation costs. We admit the use of wage
curves is only a very stylised representation of labour markets imperfections; there are many
alternative representations of rigidities in wage adjustment (nominal wage stickiness, minimum wage
etc) or in labour mobility (costs of labour reallocation across sectors). Nonetheless, in the absence of
certainty on the functioning of labour markets, our results constitute a warning that there is a risk that