labour dynamics in macro models

Labour dynamics in macro modelsShock transmission through labour markets

Ekkehard Ernst1

1Employment Trends UnitInternational Labour Organization

[email protected]

http://ekkehard.ernst.free.fr

IMFApr 18th, 2012

E. Ernst (ILO) ILO Modelling Washington, 2012 1 / 27

Understanding growth and employment The traditional approach

The ILO Global Employment Trends approach

GET is based on a tight employment-growth linkI Okun’s law as methodological backgroundI Allows regional and country-specific employment projectionsI Useful for imputation of missing data

Employment by different labour market segmentsI Metholodolgy can be applied at the level of individual labour

market segmentsI Allows to differentiate for employment by:

I AgeI GenderI SectorI Status (self-employment, informal employment)


Understanding growth and employment A new method

Considering labour dynamics more closely

Unemployment is the result of labour flow dynamicsI In the current crisis both unemployment in- and outflows explain

unemployment dynamicsI Different factors help explain their movements ⇒analysis

necessary for precise policy recommendationsI Characteristic patterns of both flow types over the business cycle

0.01

60.

018

0.02

00.

022

0.02

4Jo

b de

stru

ctio

n ra

te

0.0

0.1

0.2

0.3

0.4

Job

crea

tion

rate

1970 1980 1990 2000 2010

Job creation rate Job destruction rate

Note: No data available for Argentina, China, India, Indonesia and Saudi Arabia. Shaded areas correspond to global recessions Source: GET Labour flows model


Understanding growth and employment A new method

Labour flows and the business cycle

Okun’s law approach insufficientI In- and outflows are mutually dependent via employment stocksI Strong cyclical variations of coefficients make the traditional

approach impracticalI Calls for integrating flows in a (small) macro model

Pre−crisis peak (2008)

Great recession (2009)

0.15

0.20

0.25

0.30

0.35

Coe

ffici

ent e

stim

ate

1970 1980 1990 2000 2010

Confidence interval Time−varying estimate

Elasticitiy of job creation rates (wrt. to GDP growth, time−varying)

−0.

03−

0.02

−0.

010.

000.

01C

oeffi

cien

t est

imat

e

1970 1980 1990 2000 2010

Confidence interval Time−varying estimate

Elasticitiy of job destruction rates (wrt. to GDP growth, time−varying)


Understanding growth and employment Overview of today’s talk

Overview

1 Modelling unemployment flowsResearch strategyWhat drives unemployment dynamics?

2 A macro “wrapper”Modelling and estimation strategyA finance-augmented double Phillips curve

3 Estimation and model dynamicsData and methodologyEstimating the labour flow macro-modelModel dynamics under shocks

4 Shock transmission under different policy settings

5 Concluding remarks


Modelling unemployment flows Research strategy

Aiming at a fully estimated model

In the tradition of macro-econometric modelsI All parameters are estimated; no calibrationI All variables are based on observablesI Instrumentation of expectation variables

Including policy reaction functionsI Taylor rule for monetary policyI Government spending rule for fiscal policyI Monetary-fiscal interactions via public debt



Model ideas I

Flow model of the labour marketI Empirical formulation of standard matching modelI Full and separate account of unemployment in- and

outflowsI See, e.g., Carlsson et al. (2006)

Financial frictions modelI Real-share prices affect investment and long-term

interest rates (e.g., Phelps, 1994)I Yield curve with sticky long-term interest ratesI Productivity shocks as medium-term drivers



Model ideas II

Wage-price dynamicsI Double Phillips curve (e.g. Flaschel et al. 1997; Erceg

et al. 2000):I Reduced-form wage bargaining curveI Hybrid Phillips curve

Labour market policies interact with structural shocksI Use reduced-form strategies to model structural policies

I Employment protectionI Wage policies/bargaining institutions

I Policy identification through parameter changes


Modelling unemployment flows What drives unemployment dynamics?

An overview of the model flows I

I Decomposing unemployment dynamics into...

4Ut =4Lt −4ETt = INt −OUTt

I ...Labour force growth and...

4Lt = α3 + β314Lt−1 + β324ut−1 + β33Taxt

I ...Employment growth (i.e. the net effect of job creation and destruction)

4ETt = JobCreationt −JobDestructiont



An overview of the model flows II

I Job creation

JobCreationt = β11ETt−1 + β12wt + β13ADt + β14rt + β15Invt +

β16JobCreationt−1

I Job destruction

JobDestructiont = β21TFPt + β22rt + β23REERt + β24ADt + β25wt +

β26JobDestructiont−1

I Wage determination:

wt = α4 + β41Kt + β42CBt + β434ut−1 + β44Taxt



Putting the pieces together

Substituting the flow equations:

OUTt = JobCreationt

INt = JobDestructiont + ∆Lt−1

Hence:

OUTt = β̃11OUTt−1 + β̃12X JobCreationt + β̃144ETt−1

INt = β̃21INt−1 + β̃22X JobDestructiont + β̃244Lt−1


A macro “wrapper” Modelling and estimation strategy

Modelling methodology IStep-by-step estimation

Step 1: Identify base-line equationsI Macro variables to affect unemployment flowsI Reduced-form panel estimatesI System-GMM used to control for endogeneityI Results published in Ernst (2011)

Step 2: Identify relevant fiscal interactionsI Labour flow model generically refers to aggregate demandI Possibility to set up specific fiscal interactions such as:

I Wage- vs. Non-wage public consumptionI Direct vs. indirect taxationI Labour market programmes (ALMP, UB)

I Results published in Ernst and Rani (2011)


A macro “wrapper” Modelling and estimation strategy

Modelling methodology II

Step 3: Estimate macro modelI Introduce macro-economic closure: Modified Euler equationI Introduce endogenous policy rulesI Estimate using GMM

Step 4: Simulate model and reform scenariosI Model simulation using DynareI Reform scenarios through parametric changeI Analysis of shock transmission:

I Financial shocks (share prices)I Productivity shocks

I Analyse impact on unemployment dynamics


A macro “wrapper” A finance-augmented double Phillips curve

Financial frictions and labour flows

Share price dynamicsI Financial accelerator effect due to variations in real share

prices, Ft

I Gross-fixed capital formation also depends on publicinvestment and real long-term interest rates:

Kt = Kt−1 +Ft−1 +G It−1 +LPt−1 + rL

t−1

Yield curveI Wedge between long- and short-term interest ratesI Short-term rates determined by household expectations and

policy interventionsI Long-term rates with persistence determined by:

I Share pricesI Net government lending


A macro “wrapper” A finance-augmented double Phillips curve

A double Phillips curve

Price inflation

πt = πt−1 +Eπt+1 +REERt−1 +Gapt−1

Wage inflation

wt = wt−1 +ETt−1 + πt−1 +Eπt+1 +TFPt−1

Output gap dynamics

Gapt = wt +OUTt−1 + INt−1 +GovConst−1


Estimation and model dynamics Data and methodology

A word on the data and methodology

I Unemployment flows come from Elsby et al. (2008)I Constructed on the basis of information regarding unemployment duration atdifferent duration lengths

I Complemented by similar information for more years and other countries toimprove coverage

I Extended coverage possible using imputation methods with broadly similarresults

I Information on share price dynamics is based on OECD share price index(OECD Main Economic Indicators) deflated by CPI

I Macro indicators come from the OECD Economic Outlook databaseI Fixed effects have been accounted for through de-meaning:

dXit = Xit −Xi ·+X··


Estimation and model dynamics Estimating the labour flow macro-model

Estimation results: Labour block

INt−1 ∆LFPRt−1 ∆Prodt−1 RIRSt−1 TaxIndt−1 Gapt ∆Wagest−1

(1) INt 0.620*** -7.084*** -5.702*** 0.008*** 1.284*** -0.018*** 0.254***

(0.021) (0.901) (0.434) (0.001) (0.313) (0.001) (0.090)

OUTt−1 ETRt UCCt ∆Wagest ∆INVt Gapt

(2) OUTt 0.595*** 1.812*** -0.009*** -0.940*** 3.303*** 0.024***

(0.017) (0.140) (0.001) (0.122) (0.240) (0.001)

OUTt INt

(3) ∆ETt 0.015*** -0.019***

(0.001) (0.002)

∆Wagest−1 πt−1 E {πt+1} ∆Prodt ∆ETt−1

(4) ∆Wagest 0.640*** 0.178** 0.045* 0.487*** 0.255***

(0.040) (0.031) (0.027) (0.082) (0.042)

∆Prodt−1 ∆TFPt−1

(5) ∆Prodt 0.751*** 0.033**

(0.065) (0.004)



Estimation results: Macro block

E {RIRSt+1} GAPt NLGQt πt−1

(6) RIRSt 1.157*** 0.094*** -0.116*** -9.426***

(0.018) (0.021) (0.040) (1.384)

RSharet RIRLt−1 RIRSt

(7) RIRLt 0.700*** 0.419** 0.493***

(0.186) (0.012) (0.012)

RSharet−1 ∆GovInvt−1 ∆Prodt−1 RIRLt−1 ∆ETt−1

(8) INVt 0.004* 2.687*** 0.678*** -0.001*** 0.547***

(0.002) (0.921) (0.116) (0.000) (0.034)

OUTt−1 INt−1 ∆Wagest GovConst−1

(9) GAPt 1.090*** -4.390*** 4.971*** 8.062***

(0.131) (0.278) (1.357) (2.698)

πt−1 E {πt+1} ∆ToTt−1 ∆Wagest−1

(10) πt 0.449*** 0.533*** -0.049*** 0.041

(0.018) (0.028) (0.014) (0.029)



Estimation results: Fiscal block

GovConst−1 ∆ETt−1

(11) GovConst 0.973*** 0.025*

(0.025) (0.015)

GovInvt−1 ∆ETt−1

(12) GovInvt 0.959*** 0.028***

(0.042) (0.007)

OUTt−1 INt−1

(13) Taxt 0.010*** 0.003

(0.001) (0.003)

GovConst GovInvt Taxt

(14) NLGQt -88.626*** -122.280*** 106.385***

(4.530) (9.230) (3.129)


Estimation and model dynamics Model dynamics under shocks

Unemployment and productivity shocks

Reduction in total factor productivity leads to...

10 20 30 40 50−0.8

−0.6

−0.4

−0.2

0Productivity growth

10 20 30 40 50−1

−0.5

0

0.5Gross fixed capital formation

10 20 30 40 500

5

10

15

20Unemployment

10 20 30 40 50−60

−40

−20

0Output gap


Estimation and model dynamics Model dynamics under shocks

Transmission of financial shocks

Adverse shock to real share prices and its effect on the real economy

10 20 30 40 500

2

4

6

8Unemployment

10 20 30 40 50−10

−8

−6

−4

−2

0Output gap

10 20 30 40 50−4

−3

−2

−1

0Gross fixed capital formation

10 20 30 40 50−0.03

−0.02

−0.01

0Real wage growth


Shock transmission under different policy settings

Policy change scenarios

I How to model shifts in structural policies?I Structural model with reduced-form elementsI Allows changes in certain parameters to be identified with policy changes:

I Shock transmission on unemployment flowsI Reactivity of wages to productivity shocks

I Employment protection legislation...I ...affect the level of unemployment flows, but this has no effect in ourlinearized model

I ...also affect the elasticity of flows wrt shocksI Dynamic effects depend on the transmission margin that is affected mostI Question: Why do some countries with high EPL seem to have suffered lessfrom financial crisis?

I Wage policies/collective bargaining institutionsI Minimum wages affect wage level⇒Level effects onlyI Collective bargaining affects elasticity of wages wrt productivity,unemployment and inflation



Dynamic effects of changes in EPL I

Higher EPL =⇒ weaker reaction of unemployment to adverse financial shocks

-0.100.10.20.30.40.50.6

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49Unemployment

(deviation from steady state)

Years after shock to real share pricesBaseline Lower reactivity of flows to shocks after increasing EPL



Dynamic effects of changes in EPL II

Decomposing shifts in unemployment dynamics: Biggest effect stems from theinterest rate channel

-30

-25

-20

-15

-10

-5

0

5

10

1 6 11 16 21 26 31 36 41 46

Co

mp

osi

tio

n o

f ch

an

ge

s in

un

em

plo

ym

en

t d

yn

am

ics

Years after shock to real share prices

Interest rate transmission Output gap transmission Wage growth transmission

Investment transmission Total effect of reforms



Changes in the wage pass-through

Changes in the reactivity of wages have substantial effects on the dynamics ofunemployment

-0.015

-0.01

-0.005

0

0.005

0.01

0.015

1 6 11 16 21 26 31 36 41 46

Un

em

plo

ym

en

t re

act

ion

to

pro

du

ctiv

ity

sh

ock

(De

via

tio

n f

rom

ste

ad

y s

tate

)

Years after adverse productivity shock

Baseline Lower reactivity of wages to employment conditions

Higher real wage rigidity Lower sensitivity of wages to inflation


Concluding remarks

Lessons learned and outlook

What have we learned so far?I Linear structural model allows full estimation of all parametersI Pass-through play important role for labour flows and unemploymentI Allows detailed analysis of transmission mechanisms

Next stepsI Labour flow data available for 60+ countriesI Develop open economy and global model, based on estimation of regional

blocksI Country-specific pass-through effects rather than fixed effectI Allow for more direct policy interactions (include policy variables in the model

set-up)I Need more precise (empirical) information on flow elasticities wrt structural

policies


Bibliography

Bibliography

I Carlsson, M.; Eriksson, S.; Gottfries, N. 2006. Testing Theories of JobCreation: Does Supply Create Its Own Demand? Discussion Paper, No. 2024(Bonn: Institute for the Study of Labor (IZA)).

I Erceg, C. J.; Henderson, D. W.; Levin, A. T. 2000. "Optimal monetarypolicy with staggered wage and price contracts", Journal of MonetaryEconomics, Vol. 46, No. 2, pp. 281-313.

I Ernst, E. 2011. Determinants of unemployment flows. Labour marketinstitutions and macroeconomic policies. Discussion paper, No. 209 (Geneva:International Institute for Labour Studies), available at:http://www.ilo.org/public/english/bureau/inst/download/dp209_2011.pdf

I Ernst, E.; Rani, U. 2011. “Understanding unemployment flows”, OxfordReview of Economic Policy, Volume 27, No. 2, pp. 268–294.

I Flaschel, P.; Franke, R.; Semmler, W. 1995. Dynamic macroeconomics.Instability, fluctuations, and growth in monetary economies. (Cambridge,MA.: MIT Press).

I Phelps, E. 1994. Structural Slumps. (Cambridge, MA.: Harvard UniversityPress).


http://www.ilo.org/public/english/bureau/inst/download/dp209_2011.pdf

labour dynamics in macro models

Economy & Finance