Structural policies and economic resilience to shocksR. Duval, J. Elmeskov and L. Vogel (OECD)

1 March 2007

Motivation of the paper

• Business cycle fluctuations have become smaller• But it is unclear whether business cycles have become

more synchronised, except perhaps within the euro area

• Sources of cyclical divergence:- Idiosyncratic shocks- Divergent responses to common shocks: have English-

speaking and Nordic OECD countries been more “resilient” than Continental European ones lately? Is there a “double dividend” from structural reforms?

Purpose of this paper: study extent to which structural and macro policy settings affect resilience

Stylised features of business cycles

• Three simple OG measures: OECD, HP filter, Baxter-King filter

• Cross-country divergence of OGs has declined• But unclear whether this just reflects smaller amplitude

of cycles within countries or also greater synchronisation of cyclical positions:

- “Normalised” stdev of OGs has not come down- Marginally increased coincidence of turning points- No decline in idiosyncratic vs common comp. of OGs…

- …except perhaps within the euro area

])[(][ itititititGAP εγγγλεγλ +−++=++=

“common” component “idiosyncratic” component

One explanation: resilience differs• Resilience to shocks comprises at least 2 dimensions:- Extent to which shocks are damped (A)- Speed with which output reverts to potential (B)

• Labour/product market regulations (LMPMR) are expected to have ambiguous effects on resilience (e.g. cost-push or technology shocks in basic NK framework):

- Wage/price stickiness increases A but reduces B- Real wage rigidities reduce A but increase B• Strong monetary policy transmission mechanisms good for

resilience in general: they increase both A and B

This is largely an empirical issue Set up empirical framework which disentangles A and B Cross-country time-series analysis of OG patterns

Econometric methodology• Start with NLS estimation of simple panel equation:

- Coef λt is a common unobserved shock (Blanchard-Wolfers, EJ 2000) resilience to common shocks

- Coefs ϕi and are γi capture country-specific persistence and amplification mechanisms

- AR(2) specification suggested by residual auto-correlation test + sinusoidal pattern of OGs

- Sample: 20 OECD countries, 1982-2003, annual data• Alternative concepts of OG:- OECD OGs- Sensitivity analysis (needed due to bias/inconsistency

risks): HP OG, BK OG and U gap estimates; unobserved and observed shocks

( ) itiitititiit GAPGAPGAP εαγληϕ ++++−= −− 1)( 21

Persistence of shocks:1

coefficientImplied half-life of output gaps

(in years)

Amplification of shocks:1

coefficient

Estimate for the US:

USA 0.44 1.67 0.41

Estimates for other OECD countriesand test for statistical differences in coefficients with respect to the US:

AUS 0.34 1.3 0.24* *

AUT 0.47 1.8 -0.85***

BEL 0.45 1.7 -0.25*

CAN 0.27 1.1 0.34** **

CHE 0.41 1.6 -0.16*

DEU1 0.42 1.6 -0.15

DNK 0.31 1.2 -0.49* * **

ESP 0.54 2.3 -0.41** ** **

FIN1 0.49 2.0 0.24

FRA 0.50 2.0 -0.53***

GBR 0.41 1.6 -0.30*

IRL 0.49 1.9 0.37

ITA 0.47 1.9 -0.37**

JPN 0.50 2.0 -0.57***

NLD 0.50 2.0 -0.55**

NOR 0.57 2.5 -0.97*** *** ***

NZL 0.38 1.4 -0.61***

PRT 0.56 2.4 -0.38*** *** **

SWE1 0.40 1.5 0.18Time dummies yesObservations 434

R-squared 0.85Non-linear least squares.

Output gap equations with country dummies (20 OECD countries, 1982-2003)

iηϕ iγ

Preliminary estimation results• OG persistence: low in English-speaking and Nordic

countries, high in continental European countries and Japan

• Initial impact of common shocks: opposite patterns, i.e. high in English-speaking and Nordic countries, low in continental European countries and Japan

• OG persistence (initial impact of shocks) positively (negatively) correlated across countries with labour and product market regulatory indicators

• This is descriptive analysis, not econometric evidence:- In practice, coefficients are estimated over 22 observations- Assumes unchanged business cycle patterns over time

Expand econometric framework to allow for role of (time-varying) policies and institutions

Exploring the structural policy determinants of resilience

• Write persistence and amplification coefs as a function of indicators of labour/product market regulation:

where the Xj s and Xk s include:- Unemployment benefit replacement rate (UBRR)- Stringency of employment protection legislation (EPL)- Union coverage- Degree of “corporatism” in industrial relations - Stringency of product market regulation (PMR)

• Positive ϕj means Xj increases OG persistence• Positive γk means Xk increases initial impact of shock

itik

kkit

ktitit

j

jjit

jit XXGAPGAPXXGAP εαγληϕϕ ++

−++−

−+= ∑∑ −− )(1)()( ..21..

Addressing the multicollinearity issue• There is a multicollinearity issue• Here it is addressed in two alternative ways:1. “Statistical tournament”:- Estimate all possible models with 1 institution- Eliminate institution i if not significant- Based on new (reduced) set of institutions, re-estimate all

possible models with 2 institutions- Eliminate institution j if not always significant- …etc…and stop process when final model selected2. Principal component analysis construct a “synthetic”

indicator of LMPMR:Labour and product market regulation = 0.42*(replacement rate) + 0.45*(EPL) + 0.48*(union coverage) - 0.51*(low corporatism) + 0.37*(PMR)

1 2

Final model selected from statistical tournament

With synthetic indicators of labour and product market regulation

Persistence coefficients:1.085 1.065

[24.96]*** [23.82]***0.425 0.397

[11.39]*** [10.04]***

Effect of institutions on persistence:

EPL 0.128[4.77]***

PMR

Labour and product market regulation1 0.090[4.20]***

Effect of institutions on amplification of shocks:

EPL

PMR -0.537[5.93]***

Labour and product market regulation1 -0.147[2.93]***

Observations 434 434R-squared 0.83 0.83Non-linear least squares. Absolute value of t statistics in brackets.* significant at 10%; ** significant at 5%; *** significant at 1%.

Final output gap equations with labour and product market regulation indicators (20 OECD countries, 1982-2003)

kγ

φ

jϕ

η

Monetary factors and resilience• OG dynamics in aftermath of shocks depends also on

monetary policy response and strength of transmission mechanisms

• This is tentatively captured here by adding to the baseline equation with LMPMR indicator:

- Household mortgage debt: captures strength of one major transmission channel (negatively correlated with mortgage market regulation, and proxy for financial market flexibility)

- Flexible exchange rate regime dummy variable: captures whether monetary policy is able to react

- Financial intermediation (credit/stock market value traded)

Significant and robust negative effect of household mortgage debt on OG persistence

1 2 3 4 5

With synthetic indicators of labour and product

market regulation alone

'= 1 + household

mortgage debt

'= 1 + flexible exchange

rate regime

'= 1 + financial

intermediation

Final model with synthetic indicators of labour and

product market regulation and monetary factors

Persistence coefficients:1.065 1.060 1.057 0.918 1.060

[23.82]*** [22.81]*** [23.02]*** [11.76]*** [22.98]***0.397 0.392 0.395 0.382 0.392

[10.04]*** [9.54]*** [9.85]*** [7.51]*** [9.56]***

Effect of institutions on persistence:

Labour and product market regulation1 0.090 0.083 0.103 0.079 0.073[4.20]*** [3.80]*** [3.64]*** [3.71]*** [3.81]***

Household mortgage debt2 -0.546 -0.478[2.01]** [2.03]**

Flexible exchange rate regime 0.073[0.87]

Financial intermediation 0.0002[1.90]*

Effect of institutions on amplification of shocks:

Labour and product market regulation1 -0.147 -0.136 -0.195 -0.281 -0.136[2.93]*** [2.74]*** [2.91]*** [1.50] [2.77]***

Household mortgage debt2 0.076[0.13]

Flexible exchange rate regime -0.286[1.38]

Financial intermediation -0.003[1.16]

Country fixed effects yes yes yes yes yesTime dummies yes yes yes yes yesObservations 434 412 434 410 412R-squared 0.83 0.83 0.83 0.85 0.83Non-linear least squares. Absolute value of t statistics in brackets.* significant at 10%; ** significant at 5%; *** significant at 1%.

Output gap equations with synthetic indicators of labour and product market regulation(20 OECD countries, 1982-2003)

jϕ

kγ

φ

η

Back to resilience

• Estimated impact of LMPMR is ambiguous: persistence vsamplification (impact of mortgage regulation is not)

• Simulate final equation for different values of LMPMR indicator to see how the latter affects resilience according to 3 alternative criteria:

- Time needed for OG to close in aftermath of shock (A)- Cumulative output loss (B)- Output gap volatility (C)Warning: These are not welfare criteria

LMPMR increases A and B but reduces C• One can also use final equation to simulate how each

country (based on its current set of policy settings) is expected to score on A, B and C

Simulated degree of resilience (based on final equation, using 2003 values of policy and institutional indicators)

0.0

1.0

2.0

3.0

4.0

5.0

6.0

USA

CH

E

GBR NZL

CAN AU

S

DN

K

NLD JPN

DEU

NO

R

SWE

IRL

ESP

PRT

FIN

BEL

AUT

FRA

ITA

Time T needed for output to get back to potential(in years, following a 1 percentage point negative common shock to output gaps)

Simulated degree of resilience (continued)(based on final equation, using 2003 values of policy and institutional indicators)

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

CH

E

NLD

DN

K

GBR USA NZL

DEU AU

S

PRT

NO

R

CAN

SWE

JPN

IRL

ESP

FIN

BEL

AUT

FRA

ITA

Cumulative output loss between 0 and T(as a percentage of output, following a 1 percentage point negative common shock to output gaps)

Simulated degree of resilience (continued)(based on final equation, using 2003 values of policy and institutional indicators)

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

NLD

CH

E

DN

K

PRT

NO

R

DEU

SWE

IRL

FIN

ESP

BEL

FRA

AUT

NZL

AUS

GBR JP

N

CAN USA IT

A

Output gap volatility(number of squared standard deviations of the common shock, assuming there are no idiosyncratic shocks)

Summing up• LMPMR dampens initial impact but increases persistence• EPL (persistence) and PMR (initial impact) robust• This is theoretically consistent with view that labour/product

market rigidities primarily increase wage/price stickiness• Overall impact of LMPMR on resilience depends on criterion

used. Normative implications not explored here• Mortgage market regulation detrimental (tentative though)• Labour, product and financial market settings seem able to

account for some recent business cycle stylised facts: English-speaking/Nordics vs Continental Europeans

• Agenda for future work: - Further explore monetary transmission channels- Should one explore non-systematic effect of macro policies?

If so, is this feasible in a univariate framework?

Thank you

Tables and charts

Data before 1991 refer to Western Germany. BP gap data for Korea start only in 73:1. BP data stop in 05:4 due to the filtering method. LUX, MEX, TUR and Eastern European countries excluded.Source: OECD Economic Outlook 80 database and authors' calculations.

Cyclical convergence accross 23 OECD economies 1970-2006Standard deviation of unweighted output gaps

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

4.5

5.0

1970 75 80 85 90 95 2000 05

HP GAP BP GAP OECD GAP

Data before 1991 refer to Western Germany. BP gap data for Korea start only in 73:1. BP data stop in 05:4 due to the filtering method. LUX, MEX, TUR and Eastern European countries excluded.

Source: OECD Economic Outlook 80 database and authors' calculations.

Cyclical synchronisation accross 23 OECD economies 1970-2006Standard deviation of unweighted output gaps divided by the average absolute output gap

0.0

0.5

1.0

1.5

2.0

2.5

1970 75 80 85 90 95 2000 05

HP GAP BP GAP OECD GAP

Data before 1991 refer to Western Germany. BK gap data for Korea start only in 73:1.LUX, MEX, TUR and Eastern European countries excluded.

Source: OECD Economic Outlook 80 database and authors' calculations.

Cycles are defined as lasting five quarters at least. Turning points require that the upturn or downturn lasts at least over two subsequent quarters.

Business cycle turning points accross 23 OECD economiesAverage number of turning points over current and previous three quarters

HP output gap measure

6

5

4

3

2

1

0

1

2

3

4

5

6

1970 75 80 85 90 95 2000 05

Peaks Troughs

Data before 1991 refer to Western Germany. LUX eand SI excluded.Source: OECD Economic Outlook 80 database and authors' calculations.

Business cycle turning points accross euro membersAverage number of turning points over current and previous three quarters

HP output gap measure

Cycles are defined as lasting five quarters at least. Turning points require that the upturn or downturn lasts at least over two subsequent quarters.

3

2

1

0

1

2

3

1970 75 80 85 90 95 2000 05

Peaks Troughs

Idiosyncratic relative to common fluctuations over time

Standard deviation of idiosyncratic component relative to standard deviation of common component

HP output gap BP output gap

1973-1989 1990-2006 1973-1989 1990-2006

France 0.7 0.8 0.7 0.7 Belgium 0.9 0.7 0.9 0.6 Germany 0.7 1.1 0.6 1.0 Austria 1.0 0.9 0.9 0.9 Spain 1.0 1.0 1.0 0.9 Italy 1.1 0.8 1.0 0.8 Netherlands 1.2 0.9 0.9 0.8 Canada 1.2 1.3 1.1 1.3 United Kingdom 1.4 1.0 1.3 1.0 Japan 1.1 1.4 1.0 1.5 Denmark 1.5 1.1 1.5 0.9 Sweden 1.5 1.0 1.4 1.0 United States 1.4 1.2 1.4 1.2 Switzerland 1.6 1.0 1.6 1.0 Australia 1.4 1.4 1.4 1.4 Norway 1.6 1.4 1.6 1.3 Ireland 1.5 1.8 1.6 1.5 Portugal 2.1 1.7 2.1 1.6 Finland 1.8 2.1 1.7 2.1 Iceland 1.9 2.8 2.0 2.2 Korea 2.3 3.3 2.3 3.4 Greece 3.2 2.1 2.4 1.2 New Zealand 3.4 2.1 2.7 2.0

Average 1.5 1.4 1.4 1.3

Source: OECD Economic Outlook 80 database and calculations.

Idiosyncratic relative to common fluctuations in the euro area

Standard deviation of idiosyncratic component relative to standard deviation of common component

HP output gap BP output gap

1973-1989 1990-2006 1973-1989 1990-2006

France 0.5 0.5 0.3 0.3 Belgium 0.7 0.5 0.4 0.3 Germany 0.7 0.7 0.4 0.4 Austria 0.9 0.6 0.4 0.3 Spain 0.9 0.5 0.5 0.3 Italy 0.9 0.6 0.5 0.4 Netherlands 1.2 0.5 0.6 0.2 Ireland 1.3 1.2 0.7 0.6 Portugal 1.7 0.9 1.0 0.6 Finland 1.6 1.6 1.1 1.3 Greece 2.7 1.5 1.5 0.6

Average 1.2 0.8 0.7 0.5

Source: OECD Economic Outlook 80 database and calculations.

Persistence of shocks: coefficient Amplification of shocks: coefficient

Benefit replacement rate 0.12 -0.39*

EPL for regular contracts 0.62*** -0.43**

PMR 0.58*** -0.46**

Collective bargaining coverage 0.29 -0.23

Low corporatism -0.52*** 0.54***

Labour and product market regulation(synthetic indicator) 0.5** -0.51**

Source : Authors' estimates on the basis of country-specific persistence and amplification coefficients estimated in Table 2.1

coefficients and labour/product market policy indicatorsCross-country correlation coefficients between persistence/amplification

(based on simpleregressions of country-specific coefficients on the average value of each policy indicator over the period 1982-2003, using bootstrapped critical values to assess statistical significance)

iγiϕ

Correlation coefficients between labour and product market regulation indicators

Benefit replacement

rate

EPL PMR Collective bargaining coverage1

Low corporatism

Benefit replacement rate 1

EPL 0.29 1

PMR 0.15 0.37 1

Collective bargaining coverage1 0.52 0.44 0.40 1

Low corporatism -0.53 -0.57 -0.38 -0.48 1

1. time-invariant indicator (country average over the period 1980-2000)

Correlation coefficients, 1982-2003

1 2 3 4 5

Benefitreplacement rate

EPL Low corporatism Bargaining coverage

PMR

Persistence coefficients:1.077 1.081 1.069 1.123 1.112

[22.96]*** [24.71]*** [23.54]*** [23.98]*** [25.46]***0.395 0.425 0.390 0.393 0.399

[9.62]*** [11.19]*** [9.55]*** [9.68]*** [10.27]***

Effect of institutions on persistence:

Benefit replacement rate 0.002[0.75]

EPL 0.141[5.11]***

Low corporatism -0.279[3.40]***

Collective bargaining coverage 0.004[2.63]***

PMR 0.099[3.15]***

Effect of institutions on amplification of shocks:

Benefit replacement rate -0.007[1.09]

EPL -0.242[2.93]***

Low corporatism 0.493[2.47]**

Collective bargaining coverage -0.002[0.51]

PMR -0.508[6.15]***

Observations 434 434 434 434 434R-squared 0.82 0.83 0.82 0.82 0.83Non-linear least squares. Absolute value of t statistics in brackets.* significant at 10%; ** significant at 5%; *** significant at 1%.

Output gap equations with individual labour and product market regulation indicators (20 OECD countries, 1982-2003)

kγ

φ

jϕ

η

END