Royal Economic Society 1 KEYNESIAN CONTROVERSIES ON WAGES Two - - PowerPoint PPT Presentation

Royal Economic Society

1

KEYNESIAN CONTROVERSIES ON WAGES Two Hypotheses from The General Theory 1 Changes in Money Wages and in Real Wages Keynes: “.........in the case of changes in the general level of wages, ................When money-wages are rising,......it will be found that real wages are falling; and when money-wages are falling, real wages are rising.” pp. 9-10.

2

2 Changes in Real Wages and in Employment “... with a given organisation, equipment, and technique, real wages and the volume of output (and hence of employment) are uniquely correlated, so that, in general, an increase in employment can only

• ccur to the accompaniment of a decline in the rate
• f real wages....The real wage...................has a

unique (inverse) correlation with the volume of employment.” This correlation traces out firms’ labour demand functions.

3

Examining the Facts John Dunlop, Economic Journal, Sept.1938 Lorie Tarshis, Economic Journal, March 1939 Henry Richardson, Economic Journal, Sept.1939 John Dunlop, ReviewofEconomicStudies,June 1939

• J. M. Keynes, Economic Journal, March 1939

John Dunlop, QuarterlyJournalofEconomics,1941 John Dunlop, Wage Determination under Trade Unions, 1944 **************************************** Product wages v. Consumption wages; Wages v. Earnings

4

Tarshis: “Mr. Keynes appears to be mistaken” Dunlop: Keynes’ hypothesis of a negative correlation between changes in real wages and changes in output was “untenable”. Keynes’ Response Keynes: “.....one of the most surprising, yet best- established, facts in the whole range of economic statistics,.........is the stability of the proportion of the national dividend accruing to labour, irrespective.....of the phase of the trade cycle.....a bit

• f a miracle.” Why was Keynes surprised?

5

If S = (w.L) ÷ (p X) , log S = log (w/p) - log (X/L) d S / d t = d [(w/p)] / d t - d [(X/L)] / d t if d [(w/p)] / d t = d [(X/L)] / d t , d S / d t = 0 The characterisation of labour markets in The General Theory is not one of its strong elements. Keynes: “Every trade union will put up a resistance to a cut in money-wages, however small. But....no trade union would dream of striking on every

• ccasion of a rise in the cost-of-living....”

6

Nevertheless, the correlation between changes in

• utput and in real wages became an important

identifying feature of macroeconomic models. The Post-War Research 1 The Cyclical Movement of Real Wages Macro-economics: A large literature yielded a wide variety of results. There is no empirical regularity. Micro-economics: using longitudinal observations

• n individual workers finds a mild pro-cyclical

movement

7

2 The Rigidity of Money Wages use observations on individual workers to plot the frequency distribution of wage changes between two years, for those who remain with the same employer and for those who switch employers. Y a noticeable spike at no change in nominal wages; this spike is higher when price inflation low. If money wages are rigid, what accounts for this? Adjustment costs?

8

Ask managers: Roger Kaufman (1984) Truman Bewley (1999) ...damages “employees’ morale”. Dunlop (1938) had referred to the “loss in morale” and the consequent “adverse effect on output” if an employer imposed wage cuts on disgruntled employees.

9

Conclusion The movement in real wages over the business cycle: a lot of effort has been directed to an issue of questionable importance. Nominal wage rigidity seems present. If wages are more rigid than prices, why did Keynes’ critics find weak or no empirical support for his hypothesis that real wages rise when money wages are falling and real wages fall when money wages are rising?

Royal Economic Society

Understanding the Great Recession

Lawrence Christiano Martin Eichenbaum Mathias Trabandt

Royal Economic Society Conference, University of Manchester 2015

.

Background

Background

• GDP seems to have su§ered a permanent (10%?) fall since

2008.

Background

• GDP seems to have su§ered a permanent (10%?) fall since

2008.

• Trend decline in labor force participation accelerated after the

‘end’ of the recession in 2009.

Background

• GDP seems to have su§ered a permanent (10%?) fall since

2008.

• Trend decline in labor force participation accelerated after the

‘end’ of the recession in 2009.

• Unemployment rate rose sharply, and has now been falling

Background

• GDP seems to have su§ered a permanent (10%?) fall since

2008.

• Trend decline in labor force participation accelerated after the

‘end’ of the recession in 2009.

• Unemployment rate rose sharply, and has now been falling, but

— fall appears not to reflect increased health of labor market.

Background

• GDP seems to have su§ered a permanent (10%?) fall since

2008.

• Trend decline in labor force participation accelerated after the

‘end’ of the recession in 2009.

• Unemployment rate rose sharply, and has now been falling, but

— fall appears not to reflect increased health of labor market. — primarily, reflects the fall in labor force participation.

Background

• GDP seems to have su§ered a permanent (10%?) fall since

2008.

• Trend decline in labor force participation accelerated after the

‘end’ of the recession in 2009.

• Unemployment rate rose sharply, and has now been falling, but

— fall appears not to reflect increased health of labor market. — primarily, reflects the fall in labor force participation.

• Vacancies rose sharply after the o¢cial end of 2007 recession,

but unemployment fell relatively little (‘shift in Beveridge curve’, ‘mismatch’).

Background

• GDP seems to have su§ered a permanent (10%?) fall since

2008.

• Trend decline in labor force participation accelerated after the

‘end’ of the recession in 2009.

• Unemployment rate rose sharply, and has now been falling, but

— fall appears not to reflect increased health of labor market. — primarily, reflects the fall in labor force participation.

• Vacancies rose sharply after the o¢cial end of 2007 recession,

but unemployment fell relatively little (‘shift in Beveridge curve’, ‘mismatch’).

• Investment and consumption weak.

Background

• GDP seems to have su§ered a permanent (10%?) fall since

2008.

• Trend decline in labor force participation accelerated after the

‘end’ of the recession in 2009.

• Unemployment rate rose sharply, and has now been falling, but

— fall appears not to reflect increased health of labor market. — primarily, reflects the fall in labor force participation.

• Vacancies rose sharply after the o¢cial end of 2007 recession,

but unemployment fell relatively little (‘shift in Beveridge curve’, ‘mismatch’).

• Investment and consumption weak.
• Inflation fell, but by only a small amount given the evidence of

economic weakness.

Questions

• What were the key forces driving U.S. economy during the

Great Recession?

Questions

• What were the key forces driving U.S. economy during the

Great Recession?

• Mismatch in the labor market?

Questions

• What were the key forces driving U.S. economy during the

Great Recession?

• Mismatch in the labor market?
• Why was the drop in inflation so moderate?

To answer our questions we need a model

• Model must provide empirically plausible account of key

macroeconomic aggregates

— employment, vacancies, labor force participation, job finding rate, unemployment rate, real wages — output, consumption, investment, ..

To answer our questions we need a model

• Model must provide empirically plausible account of key

macroeconomic aggregates

— employment, vacancies, labor force participation, job finding rate, unemployment rate, real wages — output, consumption, investment, ..

• Novel features of labor market

— Endogenize labor force participation. — Derive wage inertia as an equilibrium outcome.

To answer our questions we need a model

• Model must provide empirically plausible account of key

macroeconomic aggregates

— employment, vacancies, labor force participation, job finding rate, unemployment rate, real wages — output, consumption, investment, ..

• Novel features of labor market

— Endogenize labor force participation. — Derive wage inertia as an equilibrium outcome.

• Estimate model using pre-2008 data.

To answer our questions we need a model

• Model must provide empirically plausible account of key

macroeconomic aggregates

— employment, vacancies, labor force participation, job finding rate, unemployment rate, real wages — output, consumption, investment, ..

• Novel features of labor market

— Endogenize labor force participation. — Derive wage inertia as an equilibrium outcome.

• Estimate model using pre-2008 data.
• Use estimated model to analyze post-2008 data.

Questions and Answers

Questions and Answers

• What forces drove real quantities in the Great Recession?

— Shocks to financial markets key drivers, even for variables like labor force participation.

Questions and Answers

• What forces drove real quantities in the Great Recession?

— Shocks to financial markets key drivers, even for variables like labor force participation. — Government shocks not important: because of size and timing (consistent with zero lower bound (ZLB) literature).

Questions and Answers

• What forces drove real quantities in the Great Recession?

— Shocks to financial markets key drivers, even for variables like labor force participation. — Government shocks not important: because of size and timing (consistent with zero lower bound (ZLB) literature).

• Mismatch in the labor market?

— Not a first order feature of the Great Recession.

Questions and Answers

• What forces drove real quantities in the Great Recession?

— Shocks to financial markets key drivers, even for variables like labor force participation. — Government shocks not important: because of size and timing (consistent with zero lower bound (ZLB) literature).

• Mismatch in the labor market?

— Not a first order feature of the Great Recession.

• Low labor force participation reflects response to bad labor

market prospects.

Questions and Answers

• What forces drove real quantities in the Great Recession?

— Shocks to financial markets key drivers, even for variables like labor force participation. — Government shocks not important: because of size and timing (consistent with zero lower bound (ZLB) literature).

• Mismatch in the labor market?

— Not a first order feature of the Great Recession.

• Low labor force participation reflects response to bad labor

market prospects.

— We account for ‘shift’ in the Beveridge curve without resorting to structural shifts in the labor market.

Questions and Answers

• Why was the drop in inflation so moderate?

Questions and Answers

• Why was the drop in inflation so moderate?

— Prolonged slowdown in TFP growth during the Great Recession.

Questions and Answers

• Why was the drop in inflation so moderate?

— Prolonged slowdown in TFP growth during the Great Recession. — Rise in cost of firms’ working capital as measured by spread between corporate-borrowing rate and risk-free interest rate.

Questions and Answers

• Why was the drop in inflation so moderate?

— Prolonged slowdown in TFP growth during the Great Recession. — Rise in cost of firms’ working capital as measured by spread between corporate-borrowing rate and risk-free interest rate. — Both forces exert countervailing pressure on inflation.

Outline

• Provide a very crude outline of the model.

— focus on the novel labor market features.

Outline

• Provide a very crude outline of the model.

— focus on the novel labor market features.

• Very rough description of the results.

Outline

• Provide a very crude outline of the model.

— focus on the novel labor market features.

• Very rough description of the results.
• For details, see the paper.

Labor Market

Labor Market

Employment* E* Non,par/cipa/on* N* Unemployment* U*

Labor Market

Employment* E* Non,par/cipa/on* N* Unemployment* U* ,Household*labor*force*decision* ,Split*between*U*and*E*determined*by*job,finding*rate.*

E0

1

X

t=0

tU( ~ Ct);

~ Ct = h (1 ! !) (Ct)" + ! " CH

t

#"i 1

Labor Market

Employment* E* Non,par/cipa/on* N* Unemployment* U* ,Household*labor*force*decision* ,Split*between*U*and*E*determined*by*job,finding*rate.*

E0

1

X

t=0

tU( ~ Ct);

~ Ct = h (1 ! !) (Ct)" + ! " CH

t

#"i 1

h ! " # i CH

t

=1!Lt

Labor Market

Employment* E* Non,par/cipa/on* N* Unemployment* U* ,Household*labor*force*decision* ,Split*between*U*and*E*determined*by*job,finding*rate.* E0

X

tU( ~ Ct);

max fCt;Lt;CH

PtCt + PI;tIt + Bt+1 " RK;tKt + (Lt ! lt) PtDt + ltWt + Rt!1Bt ! Tt ! Kt+1 = (1 ! $K) Kt + It

Labor Market

Employment* E* Non,par/cipa/on* N* Unemployment* U* Bargaining* Three*types*of*worker,firm*mee/ngs:* *i)*E*to*E*,*ii)*U*to*E,*iii)*N*to*E**

Modified version of Hall-Milgrom

Modified version of Hall-Milgrom

• Nature of bargaining between workers and firms has the

consequence that sensitivity of wages to general business conditions is reduced somewhat.

Modified version of Hall-Milgrom

• Nature of bargaining between workers and firms has the

consequence that sensitivity of wages to general business conditions is reduced somewhat.

• After expansionary shock, rise in wages is relatively small.

— Important for capturing the dynamic e§ects of monetary policy and technology shocks.

Estimated Medium-Sized DSGE Model

Estimated Medium-Sized DSGE Model

• Standard empirical NK model (e.g., CEE, ACEL, SW):

— Calvo price setting frictions, but no indexation. — Habit persistence. — Variable capital utilization. — Working capital. — Adjustment costs: investment, labor force. — Taylor rule.

Estimated Medium-Sized DSGE Model

• Standard empirical NK model (e.g., CEE, ACEL, SW):

— Calvo price setting frictions, but no indexation. — Habit persistence. — Variable capital utilization. — Working capital. — Adjustment costs: investment, labor force. — Taylor rule.

• Our labor market structure.

Estimated Medium-Sized DSGE Model

• Standard empirical NK model (e.g., CEE, ACEL, SW):

— Calvo price setting frictions, but no indexation. — Habit persistence. — Variable capital utilization. — Working capital. — Adjustment costs: investment, labor force. — Taylor rule.

• Our labor market structure.
• Estimation strategy: Bayesian impulse response matching.

— Shocks to monetary policy, neutral and investment-specific technology. — Our model performs well relative to this metric.

The U.S. Great Recession

The U.S. Great Recession

U.S. Great Recession: Target Gap Ranges

U.S. Great Recession: Target Gap Ranges

The Great Recession in the U.S.

• Gov. Cons. & Invest. (%)

Shocks Driving the Recession

• Two financial shocks.

Shocks Driving the Recession

• Two financial shocks.
• A TFP shock.

Shocks Driving the Recession

• Two financial shocks.
• A TFP shock.
• A government spending shock.

Shocks Driving the Recession

• Two financial shocks.
• A TFP shock.
• A government spending shock.
• All shocks are computed from the data.

Shocks Driving the Recession

• Two financial shocks.
• A TFP shock.
• A government spending shock.
• All shocks are computed from the data.

— But, we must assume time series representations for the shocks, so that agents agents can forecast them in real time.

Monetary Policy in the Great Recession

• From 2008Q3 to 2011Q2:

— Taylor-type feedback rule subject to the ZLB.

Monetary Policy in the Great Recession

• From 2008Q3 to 2011Q2:

— Taylor-type feedback rule subject to the ZLB.

• After 2011Q2: ‘forward guidance’

— following 1 year transition, ‘Evans rule’

Monetary Policy in the Great Recession

• From 2008Q3 to 2011Q2:

— Taylor-type feedback rule subject to the ZLB.

• After 2011Q2: ‘forward guidance’

— following 1 year transition, ‘Evans rule’ — keep funds rate at zero until either unemployment falls below 6.5% or inflation rises above 2.5%.

Solving the Model

• Very substantial nonlinearity due to monetary policy

— Zero lower bound and forward guidance.

Solving the Model

• Very substantial nonlinearity due to monetary policy

— Zero lower bound and forward guidance.

• We do stochastic simulation on the actual nonlinear equations,

starting in 2008Q3.

— Agents forecast future values of shock given history of past shocks.

Solving the Model

• Very substantial nonlinearity due to monetary policy

— Zero lower bound and forward guidance.

• We do stochastic simulation on the actual nonlinear equations,

starting in 2008Q3.

— Agents forecast future values of shock given history of past shocks. — Use a version of extended path to do the calculations (see paper).

The U.S. Great Recession: Data vs. Model

The U.S. Great Recession: Data vs. Model

Figure 8: The U.S. Great Recession: Data vs. Model

The U.S. Great Recession: Data vs. Model

Figure 8: The U.S. Great Recession: Data vs. Model

Decomposing What Happened into Shocks

Decomposing What Happened into Shocks

• Our shocks roughly reproduce the actual data.

Decomposing What Happened into Shocks

• Our shocks roughly reproduce the actual data.
• We investigate the e§ect of a shock by shutting it o§.

— Resulting decomposition is not additive because of nonlinearity.

Decomposing What Happened into Shocks

• Our shocks roughly reproduce the actual data.
• We investigate the e§ect of a shock by shutting it o§.

— Resulting decomposition is not additive because of nonlinearity.

• Results:

— Financial shocks - account for the biggest e§ects on real quantitites.

• especially the financial shock measured by interest rate spreads.

Decomposing What Happened into Shocks

• Our shocks roughly reproduce the actual data.
• We investigate the e§ect of a shock by shutting it o§.

— Resulting decomposition is not additive because of nonlinearity.

• Results:

— Financial shocks - account for the biggest e§ects on real quantitites.

• especially the financial shock measured by interest rate spreads.

— Government spending - relatively small role.

Decomposing What Happened into Shocks

• Our shocks roughly reproduce the actual data.
• We investigate the e§ect of a shock by shutting it o§.

— Resulting decomposition is not additive because of nonlinearity.

• Results:

— Financial shocks - account for the biggest e§ects on real quantitites.

• especially the financial shock measured by interest rate spreads.

— Government spending - relatively small role. — TFP - plays an important role in preventing drop in inflation.

Decomposing What Happened into Shocks

• Our shocks roughly reproduce the actual data.
• We investigate the e§ect of a shock by shutting it o§.

— Resulting decomposition is not additive because of nonlinearity.

• Results:

— Financial shocks - account for the biggest e§ects on real quantitites.

• especially the financial shock measured by interest rate spreads.

— Government spending - relatively small role. — TFP - plays an important role in preventing drop in inflation.

Conclusion

• Bulk of movements in economic activity during the Great

Recession due to financial frictions interacting with the ZLB.

— ZLB has caused negative spending shocks to push the economy into a prolonged recession.

Conclusion

• Bulk of movements in economic activity during the Great

Recession due to financial frictions interacting with the ZLB.

— ZLB has caused negative spending shocks to push the economy into a prolonged recession.

• Findings based on looking through lens of a NK model:

— firms face moderate degrees of price rigidities, — no sticky wages. — endogenous labor force participation, standard labor market variables.

Conclusion

• Bulk of movements in economic activity during the Great

Recession due to financial frictions interacting with the ZLB.

— ZLB has caused negative spending shocks to push the economy into a prolonged recession.

• Findings based on looking through lens of a NK model:

— firms face moderate degrees of price rigidities, — no sticky wages. — endogenous labor force participation, standard labor market variables.

• No (or little) evidence for ‘mismatch’ in labor market.

Conclusion

• Bulk of movements in economic activity during the Great

Recession due to financial frictions interacting with the ZLB.

— ZLB has caused negative spending shocks to push the economy into a prolonged recession.

• Findings based on looking through lens of a NK model:

— firms face moderate degrees of price rigidities, — no sticky wages. — endogenous labor force participation, standard labor market variables.

• No (or little) evidence for ‘mismatch’ in labor market.
• Modest fall in inflation is not a puzzle once fall in TFP and

risky working capital channel are taken into account.

Royal Economic Society

Unemployment Fluctuations, Match Quality, and the Wage Cyclicality of New Hires

Mark Gertler1, Christopher Huckfeldt2, Antonella Trigari3

March 31, 2015 Royal Economic Society Conference 2015

What we do

• 1. Present new panel data evidence on the cyclical behavior of

wages for new hires versus existing workers

• 2. Develop model of unemployment fluctuations consistent with this

evidence

• 3. Model is variant of Mortensen/Pissarides that features:

bargaining)

Why we do it

◮ Long (and controversial!) tradition of incorporating wage

stickiness in macro models to improve empirical performance

◮ Pissarides critique: Existing evidence suggests greater cyclicality

• f wages for new hires than for existing workers (e.g., Bils, 1985)

margin

misplaced

Addressing the Pissarides critique

◮ Our take: evidence reflects compositional effects associated with

procyclical movements in match quality for job changers

◮ Typical regression recover estimate of new hire effect by pooling

new hires from unemployment and new hires from other jobs

Ward, 1992)

during expansion (Barlevy, 2002)

◮ Implication: New hire contract effect not separately identified

from composition effect

Our approach and main findings

◮ Construct new panel data set that permits distinguishing new

hires that are job changers vs. those coming from unemployment

◮ Show no new hire effect for workers hired from unemployment

◮ Develop a search and matching model with staggered wage

contracting and on-the-job search to explain

• vs. from unemployment

◮ Current work: results robust with complementarity btwn leisure

and consumption (Chodorow-Reich and Karabarbounis, 2014)

Data

◮ Survey of Income and Program Participation, 1990-2012

◮ Large, representative sample ◮ Interviews every four months ◮ High-frequency structure allows for construction of precise

measurements of job tenure and wages

◮ Can separate new hires between job changers and those coming

from unemployment

Existing econometric framework, e.g. Bils (1985)

log wijt = x′

ijtπx +πu ·Ut +πn ·I(newijt)+πnu ·I(newijt)·Ut +αi +eijt ◮ xijt: observables for individual i in job j at time t ◮ I(newijt): indicator for new hire ◮ αi: person fixed effect

Key finding: πnu < 0 Two observations:

• 1. New hire interaction does not vary by type of job transition
• 2. Unobserved match quality ⇒ possible estimation bias for job

changers

“Bils regressions” and new hire effect

1990-2012 sample (1) (2) Unemployment rate

• 0.162***
• 0.448***

(0.0582) (0.0920)

• Unemp. rate · I(new)
• 1.247***
• 0.997**

(0.2477) (0.4465) Estimator FE FD

• No. observations

379,104 321,397

Robust standard errors in parenthesis * p<0.10, ** p< 0.05, *** p<0.01

◮ Pissarides (2009) interpretation: flexible wages for new hires

Our econometric framework

log wijt = x′

ijtπx + πu · Ut

+ πEE

n

· I(newijt & EE) + πENE

n

· I(newijt & ENE) + πEE

nu · I(newijt & EE) · Ut + πENE nu

· I(newijt & ENE) · Ut + αi + eijt

◮ Allow separate coefficients for new hires from employment (EE)

and new hires from non-employment (ENE)

◮ Estimate in fixed effects and first differences ◮ Several measures of EE and ENE

Job changers (EE) vs. new hires from unempl. (ENE)

1990-2012 sample (1) (2) (3) (4) UR

• 0.160***
• 0.160***
• 0.159***
• 0.159***

(0.0582) (0.0582) (0.0582) (0.0582) UR· I( new & EE)

• 1.921***
• 1.927***
• 1.920***
• 1.926***

(0.4696) (0.4403) (0.4696) (0.4403) UR· I( new & ENE)

• 0.326

0.120

• 0.487

0.005 (0.5086) (0.5636) (0.5616) (0.6353) UR· I( new & LTU) – – 0.963 0.964 – – (1.1325) (1.1325) P(πEE

) 0.019 0.004 0.046 0.011 Unemp spell for ENE 0+ 1+ (0,9] (1,9]

• No. observations

375,649 375,649 375,649 375,649

• No. of fixed effects

56,878 56,878 56,978 56,878

Robust standard errors in parenthesis * p<0.10, ** p<0.05, *** p<0.01

Composition bias and new hire effect for job changers

log wijt = x′

ijtπx + πu · Ut + πn · I(newijt) + πnu · I(newijt) · Ut + αi + eijt

eijt = qij + εijt

◮ qij: unobserved match quality ◮ Procyclical match quality ⇒ Cov(∆qij, I(newijt) · ∆Ut) < 0 ◮ Cov(∆qij, I(newijt) · ∆Ut) < 0 ⇒ ˆ

πnu biased downward

Model

◮ Starting point: RBC with search and matching, perfect

consumption insurance (Merz, 1995; Andolfatto, 1996)

◮ Variations:

Moscarini and Postel-Vinay, 2013)

◮ No wage flexibility for new hires! ◮ Evaluate model’s ability to match micro and macro data

Vacancies, searchers and matching

◮ Each firm employs nt workers in good matches and bt workers in

bad matches

◮ Labor quality lt

lt = nt + φbt

with

0 < φ < 1 = (1 + φγt)nt

with

γt = bt nt

◮ Posts υt vacancies

◮ Exogenous separation probability, 1 − ν

Vacancy, searchers and matching, cont.

◮ Searchers:

¯ st = ¯ ut

• unemployed

+ νςbt¯ bt + νςn¯ nt

• OTJ search

+ (1 − ν) ςu

• ¯

nt + ¯ bt

• separated within

period ◮ Matching function:

¯ mt = σm¯ sσ

t ¯

υ1−σ

t ◮ Job finding rates for good and bad:

pn

t

= ξ ( ¯ mt/¯ st) pb

t

= (1 − ξ) ( ¯ mt/¯ st)

◮ Job filling rates for good and bad:

qn

t

= ξ ( ¯ mt/¯ υt) qb

t

= (1 − ξ) ( ¯ mt/¯ υt)

• 1 − νςbt¯

bt + νςn¯ nt ¯ st

• 13 / 25

Firms

◮ Firms choose hiring rate xt and capital kt to max firm value Ft

Ft = max

xt,kt

• ztkα

t l1−α t

− wtlt − rtkt − κ 2 x2

tlt + Et

• Λt,t+1Ft+1
• s.t.

   Laborforce: lt+1 = (ρt + xt) lt (ρt is retention rate)

Composition: γt+1 = γ(γt, xt, ρn

t , ρb t, qn t , qb t)

(γt ≡ bt/nt)

◮ Optimal hiring and capital:

κxt = Et {Λt,t+1Jt+1} rt = αztˇ kα−1

t

with Jt(γt, wt, st) ≡ F(lt, γt, wt, st) lt and ˇ kt ≡ kt lt

Workers

◮ Value of unemployment

Ut = ub + Et

• Λt,t+1
• pn

t ¯

V n

t+1 + pb t ¯

V b

t+1 + (1 − pt) Ut+1

• ◮ Value of worker in a match of type i = n, b

V i

= wit − [νc(ςit) + (1 − ν)c(ςu)] +Et

• Λt,t+1
• ν
• (1 − ςitpn

V n

• + (1 − ν)
• ςupn

V n

V b

• 15 / 25

Search intensity

◮ Problem:

¯ V b

= max

wbt − ν [c(ςbt) + (1 − ν)c(ςu)] +Et

• Λt,t+1
• ν
• (1 − ςbtpn

V b

V n

• + (1 − ν)ςu
• pn

V n

V b

• with convex search costs:

c(ςbt) = ς0 1 + ης ς1+ης

bt ◮ Solution:

ς0ςης

• Λt,t+1pn

• ¯

V n

V b

• 16 / 25

Staggered Nash bargaining

◮ Firms bargain infrequently with good matches to solve

max

w∗

(V n

t − Ut)η (Jt)1−η

s.t. wt+1 =

• wt with probability λ

w∗

t+1 with probability 1 − λ ◮ Bad workers receive fraction φ of contract wage w∗ t ◮ Average wage evolves as

¯ wt ≃ (1 − λ) ¯ w∗

t + λ ¯

wt−1

◮ New hires entering in between contract periods receive current

contract wage ⇒ no extra wage flexibility for new hires!

Wage growth of job changers

• ¯

g

JC t

= ω ¯ g

w t

• fundamental

component

+

• 1 − ω
• cw

t

• compositional

component

where

• ¯

g

w t =

¯ wt − ¯ wt−1 and

• cw

t = πBG

δBG,t−1 − πGB δGB,t−1

• δGB,t−1 = πγ

¯ γt + πς ¯ ςbt

◮ δXY,t−1 is X-to-Y share of job flows initiated at t − 1 ◮

cw

t is procyclical due to procyclicality of

δBG,t−1

¯ ςbt

Calibration

Parameter values Discount factor β 0.997 = 0.991/3 Capital depreciation rate δ 0.008 = 0.025/3 Production function parameter α 0.33 Technology autoregressive parameter ρz 0.983 = 0.951/3 Technology standard deviation σz 0.0075 Elasticity of matches to searchers σ 0.4 Bargaining power parameter η 0.5 Matching function constant σm 1.0 Search cost elasticity ης 1.0 Renegotiation frequency λ 0.917 (4 quarters)

Jointly calibrated parameters

Parameter Description Value Target φ Inverse productivity 0.70 Average E-E wage premium increase (4.80%) ξ

• Prob. of good

0.02 Average E-N-E wage match decrease (6.20%) κ Hiring cost 165.73 U-E probability parameter (0.45) ς0 Scale parameter of 0.06 E-E probability search cost (0.029) ub Flow value of 1.91 Relative value, unemployment non-work (0.501) 1 − υ Separation 0.05 E-U probability probability (0.026)

◮ Steady state/parameter restriction: ςu = ςn = ˜

ςb

Aggregate statistics

y w n + b u v U.S. Economy, 1964:1-2013:02 Relative St. Dev. 1.00 0.48 0.64 5.74 6.38 Autocorrelation 0.88 0.88 0.94 0.92 0.92 Correlation with y 1.00 0.57 0.79

• 0.87

0.91 Model Economy, λ = 11/12 (4 quarters) Relative St. Dev. 1.00 0.45 0.31 5.41 8.15 Autocorrelation 0.85 0.96 0.89 0.89 0.86 Correlation with y 1.00 0.64 0.92

• 0.92

0.98 Model Economy, λ = ∞ (Flex wages) Relative St. Dev. 1.00 0.74 0.16 2.76 4.73 Autocorrelation 0.82 0.80 0.91 0.91 0.88 Correlation with y 1.00 1.00 0.83

• 0.83

0.98

Wage semi-elasticities

Semi-elasticities of wages w/r.t. unemployment SIPP Model, 4Q Model, flex UR

• 0.37
• 0.53
• 2.41

UR· I( new)

• 1.14
• 1.08
• 2.64

TFP, productivity and output

◮ Output:

yt = ztkα

t

• nt + φbt

1−α

◮ Measured output:

yt = zt 1 + φγt 1 + γt 1−α kα

t

• nt + bt

1−α

◮ Loglinearized measured TFP:

• zt − (1 − α)

˜ γ 1 + ˜ γ 1 − φ 1 + φ˜ γ γt

◮ Extra term is procyclical

→ γ falls during an expansion

TFP, productivity and output, con’t

(TFP = zt + et)

Conclusion

◮ No evidence that new hire wages are more flexible than those for

existing workers

models

◮ Developed model of unemployment fluctuations with

• 1. Wage rigidity
• 2. Variation in match quality
• 3. On-the-job search

◮ Model consistent with both macro and micro evidence

Supplementary Slides

Recent literature

◮ Barattieri, Basu, Gottschalk (2012)

◮ Martins, Solon, and Thomas (2012)

workers

◮ Hagedorn and Manovskii (2013)

contemporaneous unemployment

◮ Haefke, Sonntag and van Rens (2013)

but not statistically significant

Martins, Solon, and Thomas

Sample selection

◮ Men, ages 20-60; drop individuals attending school full-time, self

employed, armed forces, permanent disabilities; hours ∈ [10, 100]

◮ We only use wage observation from last month of wave (SIPP

seam effect)

◮ Drop observations where individual holds multiple jobs ◮ Drop observations with top-coded or imputed wages ◮ Drop observations with wage below minimum wage minus two ◮ Wage measure unreliable at beginning and end of job spell

Measurement

◮ We use direct measure of hourly wage when available ◮ Otherwise construct hourly wage from job-specific earnings divided by

hrs/wk × wks/mth

◮ Wages deflated with PCE ◮ Unemployment: Males, 20 yrs+ ◮ Longterm unemployed: duration > 9 months ◮ Job tenure: beginning of period retrospective information, then update

for each month observed working for pay

◮ EE transitions: change in job ID across two months, both months

worked for pay

◮ ENE transitions: change in job ID, intervening month(s) w/o work for

pay

◮ New hire: tenure < 4 months

Recalls

◮ SIPP maintains unique job identifiers for each job, except if a

worker spends an entire wave in non-employment (1996+)

◮ If worker returns to previous job after full wave in

non-employment, the job is given a new ID

◮ Potential for mis-labeling recalls as ENE transitions ◮ Our solution:

associated with that job

prior to wave of non-employment

Recalls, cont.

Panel

• No. of jobs post

% Potential recalls 4 month empl. gap 1996 13,360 33.20 2001 8,605 38.81 2004 16,675 49.22 2008 23,776 52.49

Recalls, cont.

◮ Separate start date questions for jobs that start before or during

current wave

◮ 1996 panel: “Did [FIRST AND LAST NAME] begin [HIS \HER]

employment with [NAME OF EMPLOYER] at some time between [MONTH1] 1st and today?” (variable STRTJB).

(STRTREFP)

year, month, and date that the job began (variables STRTMONJB, STRTJYR, STRTJMTH).

Composition bias and new hire wage cyclicality

Bad match Job Changer EXPANSION RECESSION

t w

RECESSION Good match Continuing worker

Hiring and retention rates

Laborforce dynamics: lt+1 = (ρt + xt) lt

◮ Retention rate: ρt ≡ ρn t + φγtρb t

1 + φγt , ρi

t = ν(1 − ςitptξ) for i = n, b ◮ Hiring rate: xt ≡

• mn

t + φmb t

• vt

lt , mi

t = qi tvt for i = n, b

Royal Economic Society

Does the New Keynesian Model Have a Uniqueness Problem?

Lawrence J. Christiano Martin Eichenbaum Benjamin K. Johannsen

Disclaimer: The views expressed here are those of the authors and do not necessarily reflect the views of the Federal Reserve Board, the FOMC, or anyone else associated with the Federal Reserve System.

Introduction

• The more we look at macroeconomic models, the more

equilibria we find in those models.

Introduction

• The more we look at macroeconomic models, the more

equilibria we find in those models.

• So, analysis inevitably involves equilibrium selection.

Zero Lower Bound

• Krugman (1998) and Eggertsson and Woodford (2003)

launched a literature on the analysis of the New Keynesian model (NK) when the nominal rate of interest is at its zero lower bound (ZLB).

Zero Lower Bound

• Krugman (1998) and Eggertsson and Woodford (2003)

launched a literature on the analysis of the New Keynesian model (NK) when the nominal rate of interest is at its zero lower bound (ZLB).

– This model lies at the core of the way many think about the events of recent years and about the appropriate policy response.

Zero Lower Bound

• Krugman (1998) and Eggertsson and Woodford (2003)

launched a literature on the analysis of the New Keynesian model (NK) when the nominal rate of interest is at its zero lower bound (ZLB).

– This model lies at the core of the way many think about the events of recent years and about the appropriate policy response.

• Eggertsson-Woodford consider equilibria in which the economic

variables are constant in the ZLB and jump to a particular steady state when the ZLB is over.

Striking Results in the Literature

• Although in most models the government spending multiplier is

quite small,

Striking Results in the Literature

• Although in most models the government spending multiplier is

quite small,

– In the ZLB the multiplier can be very large.

Striking Results in the Literature

• Although in most models the government spending multiplier is

quite small,

– In the ZLB the multiplier can be very large. – Multiplier bigger the more flexible are prices.

Striking Results in the Literature

• Although in most models the government spending multiplier is

quite small,

– In the ZLB the multiplier can be very large. – Multiplier bigger the more flexible are prices.

• Similarly,

Striking Results in the Literature

• Although in most models the government spending multiplier is

quite small,

– In the ZLB the multiplier can be very large. – Multiplier bigger the more flexible are prices.

• Similarly,

– Output collapse in ZLB can be substantial, and worse the more flexible are prices.

Recent Findings (Mertens-Ravn, Braun-K¨

• rber-Waki)
• There are at least two ZLB equilibria:

Recent Findings (Mertens-Ravn, Braun-K¨

• rber-Waki)
• There are at least two ZLB equilibria:

– one equilibrium has the striking properties reported in the literature,

Recent Findings (Mertens-Ravn, Braun-K¨

• rber-Waki)
• There are at least two ZLB equilibria:

– one equilibrium has the striking properties reported in the literature, – the other one has very different properties.

Recent Findings (Mertens-Ravn, Braun-K¨

• rber-Waki)
• There are at least two ZLB equilibria:

– one equilibrium has the striking properties reported in the literature, – the other one has very different properties.

• The existing literature didn’t notice the other equilibrium!

Recent Findings (Mertens-Ravn, Braun-K¨

• rber-Waki)
• There are at least two ZLB equilibria:

– one equilibrium has the striking properties reported in the literature, – the other one has very different properties.

• The existing literature didn’t notice the other equilibrium!

– reflects linearization solution strategy used in that literature.

Recent Findings (Mertens-Ravn, Braun-K¨

• rber-Waki)
• There are at least two ZLB equilibria:

– one equilibrium has the striking properties reported in the literature, – the other one has very different properties.

• The existing literature didn’t notice the other equilibrium!

– reflects linearization solution strategy used in that literature.

• But, multiple equilibria is a huge problem.

Recent Findings (Mertens-Ravn, Braun-K¨

• rber-Waki)
• There are at least two ZLB equilibria:

– one equilibrium has the striking properties reported in the literature, – the other one has very different properties.

• The existing literature didn’t notice the other equilibrium!

– reflects linearization solution strategy used in that literature.

• But, multiple equilibria is a huge problem.

– Government spending multiplier hard to define in these circumstances.

Our Results

• We argue that stability under learning can select unique

equilibrium in ZLB

Our Results

• We argue that stability under learning can select unique

equilibrium in ZLB

– That equilibrium has the striking properties stressed in the literature.

Our Results

• We argue that stability under learning can select unique

equilibrium in ZLB

– That equilibrium has the striking properties stressed in the literature.

• Reconcile our findings with Mertens-Ravn.

Our Results

• We argue that stability under learning can select unique

equilibrium in ZLB

– That equilibrium has the striking properties stressed in the literature.

• Reconcile our findings with Mertens-Ravn.

– They conclude that learning is not an effective equilibrium selection device.

Our Results

• We argue that stability under learning can select unique

equilibrium in ZLB

– That equilibrium has the striking properties stressed in the literature.

• Reconcile our findings with Mertens-Ravn.

– They conclude that learning is not an effective equilibrium selection device. – They follow the standard approach to learning in the literature (Evans and Honkapohja 2001).

Our Results

• We argue that stability under learning can select unique

equilibrium in ZLB

– That equilibrium has the striking properties stressed in the literature.

• Reconcile our findings with Mertens-Ravn.

– They conclude that learning is not an effective equilibrium selection device. – They follow the standard approach to learning in the literature (Evans and Honkapohja 2001). – But, we argue that that approach is incomplete.

Why Care About Stability Under Learning?

• RE models typically assume that agents have an extraordinary

amount of information about the structure of the economy.

Why Care About Stability Under Learning?

• RE models typically assume that agents have an extraordinary

amount of information about the structure of the economy.

• We hope that these models are useful abstractions for thinking

about a world where these assumptions aren’t literally satisfied.

Why Care About Stability Under Learning?

• RE models typically assume that agents have an extraordinary

amount of information about the structure of the economy.

• We hope that these models are useful abstractions for thinking

about a world where these assumptions aren’t literally satisfied.

– Lucas (1978), Evans (1983, 1985), Marcet and Sargent (1989), Evans and Honkapohja (1995, 2001), McCallum (2002).

Why Care About Stability Under Learning?

• RE models typically assume that agents have an extraordinary

amount of information about the structure of the economy.

• We hope that these models are useful abstractions for thinking

about a world where these assumptions aren’t literally satisfied.

– Lucas (1978), Evans (1983, 1985), Marcet and Sargent (1989), Evans and Honkapohja (1995, 2001), McCallum (2002).

• Suppose a RE equilibrium wasn’t stable if agents made small

errors in forming expectations.

Why Care About Stability Under Learning?

• RE models typically assume that agents have an extraordinary

amount of information about the structure of the economy.

• We hope that these models are useful abstractions for thinking

about a world where these assumptions aren’t literally satisfied.

– Lucas (1978), Evans (1983, 1985), Marcet and Sargent (1989), Evans and Honkapohja (1995, 2001), McCallum (2002).

• Suppose a RE equilibrium wasn’t stable if agents made small

errors in forming expectations.

• Then that equilibrium wouldn’t be empirically interesting.

Why Care About Stability Under Learning?

• RE models typically assume that agents have an extraordinary

amount of information about the structure of the economy.

• We hope that these models are useful abstractions for thinking

about a world where these assumptions aren’t literally satisfied.

– Lucas (1978), Evans (1983, 1985), Marcet and Sargent (1989), Evans and Honkapohja (1995, 2001), McCallum (2002).

• Suppose a RE equilibrium wasn’t stable if agents made small

errors in forming expectations.

• Then that equilibrium wouldn’t be empirically interesting.
• This perspective guides our analysis of equilibria in NK model.

The Model

• Standard NK model analyzed non-linearly

The Model

• Standard NK model analyzed non-linearly

– Representative household,

The Model

• Standard NK model analyzed non-linearly

– Representative household, – Monopolistically competitive firms face price-adjustment costs (Rotemberg, 1982),

The Model

• Standard NK model analyzed non-linearly

– Representative household, – Monopolistically competitive firms face price-adjustment costs (Rotemberg, 1982), – Government

The Model

• Standard NK model analyzed non-linearly

– Representative household, – Monopolistically competitive firms face price-adjustment costs (Rotemberg, 1982), – Government

• In appendix we perform non-linear analysis of model with Calvo

pricing.

The Model

• Standard NK model analyzed non-linearly

– Representative household, – Monopolistically competitive firms face price-adjustment costs (Rotemberg, 1982), – Government

• In appendix we perform non-linear analysis of model with Calvo

pricing.

– Similar conclusions.

The ZLB

• As in Eggertsson-Woodford (2003) we assume there is an

increased desire to save (decrease in the discount rate, rt).

The ZLB

• As in Eggertsson-Woodford (2003) we assume there is an

increased desire to save (decrease in the discount rate, rt).

– rt = rℓ ≤ rh at time zero

The ZLB

• As in Eggertsson-Woodford (2003) we assume there is an

increased desire to save (decrease in the discount rate, rt).

– rt = rℓ ≤ rh at time zero – rt jumps to rh > 0 with probability 1 − p

The ZLB

• As in Eggertsson-Woodford (2003) we assume there is an

increased desire to save (decrease in the discount rate, rt).

– rt = rℓ ≤ rh at time zero – rt jumps to rh > 0 with probability 1 − p – rh is an absorbing state

The ZLB

• As in Eggertsson-Woodford (2003) we assume there is an

increased desire to save (decrease in the discount rate, rt).

– rt = rℓ ≤ rh at time zero – rt jumps to rh > 0 with probability 1 − p – rh is an absorbing state – Ct and πt return to steady state when rt = rh.

The ZLB

• As in Eggertsson-Woodford (2003) we assume there is an

increased desire to save (decrease in the discount rate, rt).

– rt = rℓ ≤ rh at time zero – rt jumps to rh > 0 with probability 1 − p – rh is an absorbing state – Ct and πt return to steady state when rt = rh.

• Let Cℓ

t and πℓ t be the equilibrium values of consumption and

inflation assuming rt = rℓ.

ZLB Equilibria

• The NK equilibrium conditions give us a mapping
• Cℓ

t

πℓ

t

• = FZLB
• (Cℓ

t+1)e, (πℓ t+1)e

ZLB Equilibria

• The NK equilibrium conditions give us a mapping
• Cℓ

t

πℓ

t

• = FZLB
• (Cℓ

t+1)e, (πℓ t+1)e

• In any RE equilibrium
• Cℓ

πℓ

• = FZLB
• Cℓ, πℓ

ZLB Equilibria

• The NK equilibrium conditions give us a mapping
• Cℓ

t

πℓ

t

• = FZLB
• (Cℓ

t+1)e, (πℓ t+1)e

• In any RE equilibrium
• Cℓ

πℓ

• = FZLB
• Cℓ, πℓ
• With RE, the equilibrium conditions collapse to f(πℓ) = 0.

• 0.002
• 0.001

Are both equilibria stable?

• Assume that agents form beliefs so that

(Cℓ

t+1)e = Cℓ t−1 and (πℓ t+1)e = πℓ t−1

Are both equilibria stable?

• Assume that agents form beliefs so that

(Cℓ

t+1)e = Cℓ t−1 and (πℓ t+1)e = πℓ t−1

• How do Cℓ

t and πℓ t evolve if (Cℓ t+1)e = Cℓ or (πℓ t+1)e = πℓ

and rt remains at rℓ for a long time.

Properties of the ZLB Equilibria

• The high-consumption equilibrium is stable.

Properties of the ZLB Equilibria

• The high-consumption equilibrium is stable.
• The low-consumption equilibrium is repelling.

Properties of the ZLB Equilibria

• The high-consumption equilibrium is stable.
• The low-consumption equilibrium is repelling.
• We conclude that the low consumption equilibrium is not

empirically interesting as a RE equilibrium.

Properties of the ZLB Equilibria

• The high-consumption equilibrium is stable.
• The low-consumption equilibrium is repelling.
• We conclude that the low consumption equilibrium is not

empirically interesting as a RE equilibrium.

• At the stable equilibrium

Properties of the ZLB Equilibria

• The high-consumption equilibrium is stable.
• The low-consumption equilibrium is repelling.
• We conclude that the low consumption equilibrium is not

empirically interesting as a RE equilibrium.

• At the stable equilibrium

– The government spending multiplier is large.

Properties of the ZLB Equilibria

• The high-consumption equilibrium is stable.
• The low-consumption equilibrium is repelling.
• We conclude that the low consumption equilibrium is not

empirically interesting as a RE equilibrium.

• At the stable equilibrium

– The government spending multiplier is large. – Multiplier bigger the more flexible are prices.

Properties of the ZLB Equilibria

• The high-consumption equilibrium is stable.
• The low-consumption equilibrium is repelling.
• We conclude that the low consumption equilibrium is not

empirically interesting as a RE equilibrium.

• At the stable equilibrium

– The government spending multiplier is large. – Multiplier bigger the more flexible are prices. – Output collapse is worse when prices are more flexible.

Speed of Divergence

• Unstable equilibrium

Speed of Divergence

• Unstable equilibrium

– Divergence is extremely rapid

Speed of Divergence

• Unstable equilibrium

– Divergence is extremely rapid – If divergence from the unstable equilibrium were slow, the unstable equilibrium might be empirically interesting.

Relationship to Mertens and Ravn (2014)

• Find that divergence from unstable equilibrium is slow.

Relationship to Mertens and Ravn (2014)

• Find that divergence from unstable equilibrium is slow.

– Sharp contrast to our result.

Relationship to Mertens and Ravn (2014)

• Find that divergence from unstable equilibrium is slow.

– Sharp contrast to our result.

• Mertens-Ravn assumption: each firm chooses its price knowing

the aggregate of all other firms’ prices.

Relationship to Mertens and Ravn (2014)

• Find that divergence from unstable equilibrium is slow.

– Sharp contrast to our result.

• Mertens-Ravn assumption: each firm chooses its price knowing

the aggregate of all other firms’ prices.

– This is the standard approach (Evans-Honkapohja 2001)

Relationship to Mertens and Ravn (2014)

• Find that divergence from unstable equilibrium is slow.

– Sharp contrast to our result.

• Mertens-Ravn assumption: each firm chooses its price knowing

the aggregate of all other firms’ prices.

– This is the standard approach (Evans-Honkapohja 2001)

• Obviously, firms cannot literally know the aggregate of all other

firm’s prices at the time they set their own price.

Relationship to Mertens and Ravn (2014)

• Find that divergence from unstable equilibrium is slow.

– Sharp contrast to our result.

• Mertens-Ravn assumption: each firm chooses its price knowing

the aggregate of all other firms’ prices.

– This is the standard approach (Evans-Honkapohja 2001)

• Obviously, firms cannot literally know the aggregate of all other

firm’s prices at the time they set their own price.

• Our assumption:

Relationship to Mertens and Ravn (2014)

• Find that divergence from unstable equilibrium is slow.

– Sharp contrast to our result.

• Mertens-Ravn assumption: each firm chooses its price knowing

the aggregate of all other firms’ prices.

– This is the standard approach (Evans-Honkapohja 2001)

• Obviously, firms cannot literally know the aggregate of all other

firm’s prices at the time they set their own price.

• Our assumption:

– Each firm chooses its price based on a belief about the aggregate of all other firms’ prices.

Relationship to Mertens and Ravn (2014)

• Find that divergence from unstable equilibrium is slow.

– Sharp contrast to our result.

• Mertens-Ravn assumption: each firm chooses its price knowing

the aggregate of all other firms’ prices.

– This is the standard approach (Evans-Honkapohja 2001)

• Obviously, firms cannot literally know the aggregate of all other

firm’s prices at the time they set their own price.

• Our assumption:

– Each firm chooses its price based on a belief about the aggregate of all other firms’ prices. – Firms update their beliefs when beliefs are incorrect.

Conclusion

• Many macro models have multiple RE equilibra.

Conclusion

• Many macro models have multiple RE equilibra.

– We consider this phenomenon in the ZLB.

Conclusion

• Many macro models have multiple RE equilibra.

– We consider this phenomenon in the ZLB.

• If we require that RE equilibria are robust to small deviations

from RE, then we find a unique stable equilibrium in the ZLB.

Conclusion

• Many macro models have multiple RE equilibra.

– We consider this phenomenon in the ZLB.

• If we require that RE equilibria are robust to small deviations

from RE, then we find a unique stable equilibrium in the ZLB.

– We find divergence from the unstable equilibrium is fast.

Conclusion

• Many macro models have multiple RE equilibra.

– We consider this phenomenon in the ZLB.

• If we require that RE equilibria are robust to small deviations

from RE, then we find a unique stable equilibrium in the ZLB.

– We find divergence from the unstable equilibrium is fast.

• This analysis is a reminder of the importance of understanding

how actual agents actually form beliefs.

Royal Economic Society