okun talk at new school by prakash loungani
DESCRIPTION
Okun's LawTRANSCRIPT
1
Does One Law Fit All?Cross-Country Evidence on Okun’s Law
Laurence Ball (Johns Hopkins University)Davide Furceri (IMF)Daniel Leigh (IMF)
Prakash Loungani (IMF)
September 10, 2013
Outline of Talk
1. Evidence for 20 OECD countries– Review of “Okun’s Law: Fit at 50?” (Ball, Leigh and Loungani, 2013)– New Evidence on “A Law for the Ages? Okun’s Law by Age Group”
2. Okun Outside OECD– Comparison of Estimates for OECD and non-OECD countries– Determinants of the Okun Coefficient
3. Addressing Issues of Data Quality and Relevance of Unemployment-Based Okun’s Law
– Data Quality: Closer Look at Evidence for Latin America– Relevance: Preliminary Work using Employment
3
Okun’s Law 101
• Okun (1962)
• “Levels” version:
Ut – Ut* = β (Yt – Yt
*) + εt, β < 0,
• “Changes” version:
ΔUt = α + β ΔYt + ωt
• Textbooks say U.S. coefficient β = –0.5.
4
Accusations Against Okun’s Law
• It’s unstable• “An Unstable Okun’s Law, Not the Best Rule of Thumb” (Meyer and
Tasci, St. Louis Fed, 2012)
• It’s dead• “The Demise of Okun’s Law” (Robert Gordon, 2011)
• Recoveries have become “jobless”
• It broke down during Great Recession• April 2010 WEO (“Okun’s Law and Beyond”)
Reasons to care about Okun’s Law
• Along with other indicators, it helps us assess whether an increase in unemployment is cyclical or structural
• If short-run decreases in output are due to declines in aggregate demand, it tells us monetary and fiscal policies may be important in solving ‘the unemployment problem’
• It suggests need for a two-handed approach to solving unemployment (both aggregate demand and aggregate supply matter)
• Statistical check on the quality of GDP and labor market data
• Initial increase cyclical rather than structural
• Greater uncertainty about relative proportions now, but signs that cyclical component remains important in most countries
• Beveridge curve quite stable; moreover shifts may not be sign of increase in natural rate (Diamond 2013)
• Other measures of mismatch back to normal• Lack of deflation not a sign of small
unemployment gap
• Stability of Okun’s Law suggests jobs will return if the growth returns.
Unemployment during the Great Recession
6
Debate in the Blogosphere:Krugman vs. Kocherlakota
“Why is unemployment remaining high? Because growth is weak — period, full stop, end of story. Historically, low or negative growth has meant rising unemployment, fast growth falling unemployment (Okun’s Law) … what we’ve been seeing lately is well within the normal range of noise.”Paul Krugman, July 9, 2011
• Minneapolis Fed president sees mismatch driving unemployment rateMinneapolis PostSep. 10, 2010
• A mismatch of worker skills, the location of available jobs and a less mobile workforce has added 2.5 points to the nation’s unemployment rate.
Present Debate:Kocherlakota vs. Kocherlakota
• A Federal Reserve Governor Announced A Surprising Change-Of-Mind About How He Sees The Economy, Business Insider, Sep. 20, 2012
• Kocherlakota … used to be something of a hawk, arguing that the Fed couldn't do much about unemployment, because the unemployment crisis was "structural" not "cyclical."
• Minneapolis Fed president sees mismatch driving unemployment rateMinneapolis PostSep. 10, 2010
• A mismatch of worker skills, the location of available jobs and a less mobile workforce has added 2.5 points to the nation’s unemployment rate.
9
Debate in the Blogosphere: Krugman vs. McKinsey
“There’s no hint in these data that we’ve entered new territory in which decent growth fails to create jobs; the problem is that we haven’t had decent growth.” Paul Krugman, July 9, 2011
“The U.S. jobs challenge today stems from a pattern of jobless recovery that does not conform to the classic cyclical view of recession and recovery. So while healthy GDP growth will be essential [for a return to full employment], it will probably not be sufficient.... it will require major efforts in education, regulation, and even diplomacy.” McKinsey Global Institute, 2011
A Debate since the 1970s …
• “There is sometimes the naïve belief that unemployment must be due to a defect in the labor market, as if the hole in a flat tire must always be at the bottom, because that is where the tire is flat” (Solow, 2000).
• "It takes a heap of Harberger triangles to fill an Okun's gap.” (Tobin, 1977)
“We impute the higher [European] unemployment to welfare states' diminished ability to cope with more turbulent economic times, such as the ongoing restructuring from manufacturing to the service industry, adoption of new information technologies, and a rapidly changing international economy. “(Ljungqvist and Sargent, 1998)
11
What We Do … and What We Find• We examine fit of Okun’s Law:
• In the U.S. since 1948 • In 20 advanced economies since 1980
• What we conclude:• It is a law (at least by the standards of macroeconomics)• Strong and stable in most countries• Exceptions exaggerated and/or quantitatively small
BUT:• substantial variation in coefficient across countries
» for reasons only partly understood
12
Deriving Okun’s Law
(1) Et – Et* = γ (Yt – Yt
*) + ηt γ > 0
(2) Ut – Ut
* = δ (Et – Et *) + μt δ < 0
• We expect γ < 1.5 (labor as quasi-fixed factor)
• We expect |δ| <1 (procyclical labor force participation)
(3) Ut – Ut* = β (Yt – Yt
*) + εt β < 0
• β = γδ, |β|<1.5, and εt = μt + δ ηt.
13
Estimating Okun’s Law(3) Ut – Ut
* = β (Yt – Yt *) + εt β < 0
• We usually measure Ut* and Yt
* with HP filter.
• Several tests of robustness• With HP
» Alternate values of HP smoothing parameter» Addressing end-point problem
• Without HP» Use of forecast errors » Use of CBO measure of Ut
* and Yt *
» Use of “changes” specification
(4) ΔUt = α + β ΔYt + ωt , holds if U * and ΔY * constant.
14
U.S. EVIDENCE ON OKUN’S LAW
15
Results: U.S. Annual Data 1948-2011Levels equation: Ut – Ut
* = β (Yt – Yt *) + εt ,
Changes equation: ΔUt = α + β ΔYt + ωt
Note: OLS standard errors. ***, **, and *: sig. at the 1, 5, and 10 percent level.
λ = 100 λ = 1,000 Changes
β -0.411*** -0.383*** -0.405***(0.024) (0.023) (0.029)
α 1.349***(0.116)
Obs 64 64 63
Adjusted R 2 0.817 0.813 0.752
Levels
16
Results for U.S., 1948-2011(SUR, joint estimation of equations 1-3, annual data, λ = 100)
Note: Standard errors in parentheses. ***, **, and *: sig. at the 1, 5, and 10 percent level.
`
Okun’s Law for Employment Estimate Adjusted R 2
γ 0.543*** 0.610
Unemployment-Employment Relationδ -0.728*** 0.798
Okun’s Law for Unemploymentβ -0.405*** 0.820
Obs
p -value for H0: β = γδ
64
0.378
17
Okun’s Law: U.S. Fit(Levels specification, natural rates based on HP filter, annual data)
1982194919581961
198319752009
2010
196019951962199319631991
20111996
199219761971
197019941959
1997
1954
1964
1974
1950
1981
20032002
194819801957
19841998197719722001
20041985
200819871986
19901965195619992005
19671969198820061989
20001968
2007
19781952
1955
1951197919731966
1953
-2-1
01
2U
ne
mp
loym
en
t ga
p
-6 -4 -2 0 2 4Output gap
= 100
1982
195820111961
200920101983
1949
1960
1975
1962199119931963
1992
1959
199519941996
198119761984
19641957
1980
1948
19541950
1997
19851971
19702008
1977
1974
1986198719901956
19981988
2003
1965
2002
1989
1972 2004200119781979
1955
2007196719992005
195220061951
196919531968197319662000
-2-1
01
23
Un
em
plo
yme
nt
gap
-6 -4 -2 0 2 4Output gap
= 1,000
18
Okun’s Law: U.S. Fit(“Changes” specification, annual data)
2009
19821958
1954
1974
1949
200819801991
1975
1970
2001
2011
2002
199020071956
1957
1961
20101960
1995
196719812003
20061993200519691979
1987
19711992
19862004198919961952
199419881985200019981963
19971983
1977
195319991968
197219761978
19641973
196219651966
1959
1984
1955
1951
1950
-2-1
01
23
4C
han
ge in
une
mp
loym
ent
ra
te
-4 -2 0 2 4 6 8Change in log of real GDP
19
Okun Stability Test, 1948-2011(Test for stability of Okun coefficient, β, at unknown date, annual data)
Note: F-statistic, inner 70 percent of sample. Critical value from Andrews (2003).
10 percent critical value
02
46
8
1950 1960 1970 1980 1990 2000 2010
= 100 = 1,000 Changes
20
Results for U.S.: Quarterly Data(OLS, levels specification: Ut – Ut
* = β(L) (Yt – Yt *) + εt , 1948Q2-2011Q4)
Note: Standard errors in parentheses. ***, **, and *: sig. at the 1, 5, and 10 percent level.
1,600 1,600 16,000 16,000
β0 -0.428*** -0.245*** -0.411*** -0.213***(0.015) (0.0230) (0.013) (0.0286)
β1 -0.133*** -0.153***(0.0345) (0.0447)
β2 -0.116*** -0.0794***(0.0230) (0.0286)
β0 + β1 + β2 -0.494*** -0.445***(0.0126) (0.0119)
α
Obs 256 256 256 256
Adjusted R 2 0.767 0.865 0.795 0.852
Hodrick-Prescott filter λ
21
Replication of Okun (1962) and More(OLS, changes specification: ΔUt = α + β(L) ΔYt + εt)
Note: Standard errors in parentheses. ***, **, and *: sig. at the 1, 5, and 10 percent level.
SampleData
β0 -0.307*** -0.233*** -0.286*** -0.218***(0.036) (0.0303) (0.018) (0.0160)
β1 -0.168*** -0.137***(0.0327) (0.0168)
β2 -0.0394 -0.0767***(0.0307) (0.0160)
β0 + β1 + β2 -0.441*** -0.432***(0.0380) (0.0200)
α 0.305*** 0.424*** 0.244*** 0.359***(0.061) (0.0524) (0.023) (0.0215)
Obs 51 51 255 255
Adjusted R 2 0.584 0.758 0.494 0.663
1948Q2-1960Q4 1948Q2-2011Q4Vintage data Current data
22
Okun’s Law: U.S. Fit, Quarterly Data(Actual and fitted values of unemployment rate, 1948Q2-2011Q4)
Note: Fitted value of Ut based estimate of Ut – Ut* = β (Yt – Yt
*) + εt with λ = 1,600.
02
46
810
12
1950 1960 1970 1980 1990 2000 2010
actual fitted
23
Okun’s Law: U.S. Fit, Quarterly Data(Actual and fitted values of unemployment rate gap, 1948Q2-2011Q4)
Note: Fitted value of Ut – Ut* based estimate of Ut – Ut
* = β (Yt – Yt*) + εt with λ = 1,600.
-2-1
01
23
1950 1960 1970 1980 1990 2000 2010
actual fitted
24
JOBLESS RECOVERIES?
25
Okun’s Law vs. “Jobless Recoveries”
• Popular view:• “Output Came Back, Employment Didn’t” (NPR, 2011)
• Our view:• Okun’s Law holds (as shown in previous slides)• Confusion because recent output recoveries have been
slow.• Point is recognized by some observers
» Krugman (2011)» Gali et al. (2012)
26
A Recovery that Looks Jobless(U.S. during the Great Recession)
Note: HP filter trends through 2007. Assumption: Ut* and ΔYt
* constant thereafter.
pre-recession trend
.85
.9.9
51
2007 2008 2009 2010 2011
Log of GDP
pre-recession normal
46
810
2007 2008 2009 2010 2011
Unemployment Rate
27
A Recovery that Looks Job-full(U.S. During the 1981 Recession)
Note: HP filter trends through 1980. Assumption: Ut* and ΔYt
* constant thereafter.
pre-recession trend
.8.8
5.9
.95
1
1980 1981 1982 1983 1984 1985 1986
Log of GDP
pre-recession normal
78
910
1980 1981 1982 1983 1984 1985 1986
Unemployment Rate
28
EVIDENCE ON OKUN’S LAW FOR 20 OECD COUNTRIES
29
Cross-country Estimates, 1980-2011(OLS, levels specification: Ut – Ut
* = β (Yt – Yt *) + εt , λ = 100, annual data)
Note: Standard errors in parentheses. ***, **, and *: sig. at the 1, 5, and 10 percent level.
β Obs Adj. R 2 β Obs Adj. R 2
Australia -0.536*** 32 0.80 Japan -0.152*** 32 0.65Austria -0.136*** 32 0.21 Netherlands -0.511*** 32 0.62Belgium -0.511*** 32 0.54 New Zealand -0.341*** 32 0.59Canada -0.432*** 32 0.81 Norway -0.294*** 32 0.62Denmark -0.434*** 32 0.72 Portugal -0.268*** 32 0.62Finland -0.504*** 32 0.77 Spain -0.852*** 32 0.90France -0.367*** 32 0.68 Sweden -0.524*** 32 0.62Germany -0.367*** 32 0.51 Switzerland -0.234*** 32 0.44Ireland -0.406*** 32 0.77 UK -0.343*** 32 0.60Italy -0.254*** 32 0.29 USA -0.454*** 32 0.82
30
Cross-country Sub-sample Stability(OLS, levels specification, λ = 100, annual data, 1980-2011)
Note: Standard errors in parentheses. ***, **, and *: sig. at the 1, 5, and 10 percent level.
βpre-95 βpost-95 p -value βpre-95 βpost-95 p -value
Australia -0.552*** -0.433*** 0.405 Japan -0.109*** -0.209*** 0.008Austria -0.134* -0.137** 0.974 Netherlands -0.713*** -0.336*** 0.006Belgium -0.634*** -0.310** 0.053 New Zealand -0.317*** -0.426*** 0.363Canada -0.500*** -0.287*** 0.006 Norway -0.319*** -0.247*** 0.410Denmark -0.490*** -0.369*** 0.205 Portugal -0.221*** -0.463*** 0.007Finland -0.610*** -0.297*** 0.001 Spain -0.793*** -0.923*** 0.205France -0.400*** -0.335*** 0.470 Sweden -0.648*** -0.362*** 0.046Germany -0.427*** -0.270** 0.232 Switzerland -0.211*** -0.274*** 0.516Ireland -0.462*** -0.382*** 0.359 UK -0.419*** -0.215*** 0.045Italy -0.142 -0.358*** 0.110 USA -0.447*** -0.464*** 0.829
31
Summary of Cross-country Estimates
• Strong relationship in most countries. • Coefficient β falls significantly at 5 percent level in 5
countries, rises significantly in 2. • Average β is –0.43 in first sample, –0.35 in second. • Correlation of countries’ βs across periods = 0.50.
32
Okun’s Law and the Great Recession(Peak-to-trough output and unemployment changes)
Notes: Similar to Figure 3.1 in April 2010 WEO.ΣΔU and ΣΔY = cumulative peak-trough changes. Adjusted R2 = –0.03.
USA
GBR
AUT
BEL
DNK
FRA
DEU
ITANLDNOR
SWE
CHE
CAN
JPN
FIN
IRL
PRT
ESP
NZL
02
46
8
U
0 2 4 6 8 10- Y
33
Okun’s Law and the Great Recession(Peak-to-trough changes, adjustment for recession duration, T)
Note: Adjusted R2 = 0.54.
USA
GBR
AUT
BEL
DNK
FRA
DEU
ITANLD NOR
SWE
CHE
CAN
JPN
FIN
IRL
PRT
ESP
NZL
02
46
8
U
1 2 3 4 T + Y
ΔUt = α + β ΔYt
ΣΔU = αT + β Σ ΔY
34
Okun’s Law and the Great Recession(Adjustment for recession duration and country-specific Okun coefficients)
Note: αi and βi = country-specific Okun coefficients, T = duration. Adjusted R2 = 0.76.
USA
GBR
AUT
BEL
DNK
FRA
DEU
ITA NLDNOR
SWE
CHE
CAN
JPN
FIN
IRL
PRT
ESP
NZL
02
46
8
U
0 2 4 6 8i T + i Y
35
SOURCES OF VARIATION IN OKUN’S LAW COEFFICIENTS
36
Cross-Country Variables(Okun coefficient vs. candidate variable)
Note: Average unemployment rate denotes 1980-2011 mean. OECD overall employment protection index: 1985-2011 mean based on available data.
USA
GBR
AUT
BEL
DNK
FRADEU
ITA
NLD
NOR
SW E
CHE
CAN
JPN
FIN
IRL
PRT
ESP
AUS
NZL
-.8
-.6
-.4
-.2
0O
kun
coef
ficie
nt
0 5 10 15Average unemployment rate
USA
GBR
AUT
BEL
DNK
FRADEU
ITA
NLD
NOR
SWE
CHE
CAN
JPN
FIN
IRL
PRT
ESP
AUS
NZL
-.8
-.6
-.4
-.2
0O
kun
co
eff
icie
nt
0 1 2 3 4OECD overall employment protection index
37
Individual Stories
• Large coefficient in Spain: temporary labor contracts
• Three smallest coefficients:
– Japan: lifetime employment tradition
– Switzerland: migrant labor
– Austria: a puzzle
38
Conclusions for OECD Countries
• Strong, stable relationship in most countries.
– Little evidence of jobless recoveries or breakdown in the Great Recession.
• Substantial cross-country variation only partly understood.
39
A LAW FOR THE AGES?OKUN’S COEFFICIENT BY AGE GROUP
Okun’s Law Coefficient: Age Groups (based on “levels” specification)
Spain
Swed
enAu
strali
aFin
land
Portu
gal
Cana
daUn
ited K
ingdo
mNe
ther
lands
Switz
erlan
dJap
an
-2.1
-1.8
-1.5
-1.2
-0.900000000000001
-0.600000000000001
-0.3
0 15-24
Spain
Swed
enFin
land
Switz
erlan
dCa
nada
Belgi
umNo
rway
Portu
gal
Japa
nIta
ly
-2.1
-1.8
-1.5
-1.2
-0.900000000000001
-0.600000000000001
-0.3
0 25-34
Okun’s Law Coefficient: Age Groups(based on “levels” specification)
Spain
Austr
alia
Finlan
dSw
eden
Belgi
umUn
ited K
ingdo
m Franc
eSw
itzer
land Ita
lyAu
stria
-2.1
-1.8
-1.5
-1.2
-0.900000000000001
-0.600000000000001
-0.3
0 35-44
Portu
gal
Cana
daUn
ited K
ingdo
mBe
lgium
Franc
eNo
rway
Swed
enAu
stria Ita
lyAu
strali
a
-2.1
-1.8
-1.5
-1.2
-0.900000000000001
-0.600000000000001
-0.3
0 45-54
Okun’s Law Coefficients: Age Groups(based on “levels” specification)
15-24 25-34 35-44 45-54 55-64-1.2
-0.9
-0.599999999999999
-0.299999999999999
1.33226762955019E-15
0.300000000000002AustraliaAustriaBelgiumCanadaDenmarkFinlandFranceGermanyItalyJapanNetherlandsNew ZealandNorwaySwedenSwitzerlandUnited KingdomUnited States
OKUN OUTSIDE OECD
Growth Strategies vs. Jobs Strategies
Quotes from World Bank’s report on “Jobs” (World Development Report 2013)
“From a statistical point of view, the relationship between growth and employment (or unemployment) shows substantial variation over time, across countries, and across sectors. In light of this diversity, a given rate of growth does not guarantee a given level of job creation or a given composition of employment.”
Growth Strategies vs. Jobs Strategies
Quotes from World Bank’s report on “Jobs” (World Development Report 2013)
• “These elasticities show great variability over time and space, too, making it difficult to forecast net job creation …”
– “in Tanzania growth elasticities of employment declined from 1.04 in the period 1992–96 to 0.27 in the period 2004–08. Similar trends have been reported for Ethiopia, Ghana, and Mozambique.”
– “In Latin America, recent estimates show that growth elasticities of employment were much lower during the global financial crisis than in previous crises. In other words, the Great Recession produced comparatively less net employment destruction in that region.”
Okun’s Law Coefficients: All Countries(based on “change” specification)
Spain
Poland
Tunisi
a
Slovak
Rep
ublic
Finlan
d
United
King
dom
France
Netherl
ands
Bosnia
and H
erzeg
ovina
South
Africa
Belgium
Mold
ova
Switzerl
and
Georg
ia
Panam
a
Nicarag
ua
Albania
Armen
ia
Turke
y
Austri
aIta
ly
Roman
ia
Belaru
s
Domini
can R
epub
lic
Azerb
aijan
China
Uzbek
istan
-1.2
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
Average=-0.25839
Okun’s Law Coefficients: All Countries(based on “level (HP100)” specification)
Spain
Slovak
Rep
ublic
Finlan
d
Netherl
ands
Belgium
Denmark
France
United
King
dom
Portug
al
Bosnia
and H
erzeg
ovina
Switzerl
and
Korea
Pakist
an
Costa
Rica
Mold
ova
Hong K
ong S
AR
Mex
ico
Philipp
ines
Peru
Argen
tina
Hondu
ras
Seych
elles
Roman
ia
Azerb
aijan
Kyrgy
z Rep
ublic
Singap
ore
El Salv
ador
-1-0.8-0.6-0.4-0.2
00.20.4
Average=-0.23952
Country Groups
Total: 81 Countries• 24 Advanced Economies
• 20 Emerging Market Economies• 11 Frontier Market Economies
• 26 Other Developing Economies
Spa
inA
ustr
alia
Uni
ted
Sta
tes
Can
ada
Irel
and
Fin
land
Sw
eden
Uni
ted
Kin
gdom
Den
mar
kF
ranc
eG
reec
eN
ethe
rlan
dsP
ortu
gal
New
Zea
land
Isra
elG
erm
any
Bel
gium
Sw
itze
rlan
dN
orw
ayH
ong
Kon
g S
AR
Aus
tria
Ital
yJa
pan
Sin
gapo
re
-1
-0.8
-0.6
-0.4
-0.2
0
Average=-0.3032
Okun’s Law Coefficients: Advanced Economies
(based on “change” specification)
Notes: Ball and others (2013) do not cover Greece, Israel, Hong Kong SAR, and Singapore. These countries are marked with green.
Okun’s Law Coefficients: Advanced Economies(based on “level (HP100)” specification)
Notes: Ball and others (2013) do not cover Greece, Israel, Hong Kong SAR, and Singapore. These countries are marked with green.
Spain
United
Stat
es
Nether
lands
Sweden
Greec
e
Denm
ark
Franc
e
United
Kin
gdom
Israe
l
Norway
Italy
Austri
a-1
-0.8-0.6-0.4-0.2
0
Average=-0.37687
Okun’s Law Coefficients: Emerging Market Economies
(based on “change” specification)
Colom
biaChi
le
Polan
dEgy
pt
Hunga
ry
Czech
Rep
ublic
South
Afri
ca
Brazil
Philip
pine
s
Mex
icoKor
ea
Russia
Taiwan
Pro
vinc
e of C
hina
Turke
yPer
u
Thaila
nd
Mala
ysia
Indo
nesia
Mor
occo
China
-0.8
-0.6
-0.4
-0.2
0
Average=-0.23075
Okun’s Law Coefficients: Emerging Market Economies
(based on “level (HP100)” specification)
Colom
biaChi
le
Polan
dEgy
pt
Hunga
ry
South
Afri
ca
Czech
Rep
ublic
Korea
Taiwan
Pro
vinc
e of C
hina
Brazil
Mex
ico
Philip
pine
s
Russia Per
u
Thaila
nd
Turke
y
Mala
ysia
China
Mor
occo
Indo
nesia
-0.8
-0.6
-0.4
-0.2
0
Average=-0.23154
Okun’s Law Coefficients: Frontier Market Economies
(based on “change” specification)
Tunisi
a
Lithua
nia
Vietna
m
Bulga
ria
Argen
tina
Croati
a
Jord
an
Kazak
hstan
Pakist
an
Roman
ia
Ukrain
e-0.5
-0.4
-0.3
-0.2
-0.1
0
Average=-0.19679
Okun’s Law Coefficients: Frontier Market Economies
(based on “level (HP100)” specification)
Lithua
nia
Tunisi
a
Croati
a
Vietna
m
Pakist
an
Bulga
ria
Jord
an
Argen
tina
Kazak
hstan
Roman
ia
Ukrain
e-0.6-0.5-0.4-0.3-0.2-0.1
0
Average=-0.21871
Okun’s Law Coefficients: Other Developing Economies
(based on “change” specification)
Slova
k Rep
ublic Ira
n
Urugu
ay
Costa
Rica
Georg
ia
Nicara
gua
Kyrgy
z Rep
ublic
Alger
ia
Belaru
s
Domin
ican
Repub
lic
Azerb
aijan
El Salv
ador
Sudan
-0.5-0.4-0.3-0.2-0.1
00.1
Average=-0.13956
Okun’s Law Coefficients: Other Developing Economies
(based on “level (HP100)” specification)
Slova
k Rep
ublic
Urugu
ay
Venez
uela
Alban
ia
Mol
dova
Domin
ican
Repub
lic
Hondu
ras
Seych
elles
Georg
ia
Azerb
aijan
Sudan
Uzbek
istan
Serbi
a-0.6
-0.4
-0.2
0
0.2
Average=-0.12769
Okun’s Law Coefficients: Summary(group average)
All Economies(81)
Advanced Economies
(24)
Emerging Markets(20)
Other Developing Economies
(26)
Frontier Market Economies
(11)
-0.4
-0.3
-0.2
-0.1
0
Change Specification Level Specification ( HP 100 ) Level Specification ( HP 12 )
Distribution of Okun’s Law Coefficients:(in percent, based on “levels” specification)
<-0.8 (-0.8, -0.6) (-0.6, -0.4) (-0.4, -0.2) (-0.2, 0) (0, 0.2) (0.2, 0.4)0
102030405060
OECD Countries (30)
<-0.8 (-0.8, -0.6) (-0.6, -0.4) (-0.4, -0.2) (-0.2, 0) (0, 0.2) (0.2, 0.4)0
102030405060
Non-OECD Countries (51)
Distribution of Okun’s Law R-sq:(in percent, based on “levels” specification)
<0.2 (0.2, 0.4) (0.4, 0.6) (0.6, 0.8) >0.80
20
40
60OECD Countries (30)
<0.2 (0.2, 0.4) (0.4, 0.6) (0.6, 0.8) >0.80
20
40
60Non-OECD Countries (51)
Okun's Law Coefficients: Regional Country Groups
Based on "level" Based on "change"
mean median mean median
South Asia, East Asia, and the Pacific -0.166 -0.161 -0.138 -0.108
Americas -0.239 -0.215 -0.224 -0.214
Europe -0.362 -0.367 -0.289 -0.282
Middle East and North Africa -0.154 -0.184 -0.173 -0.145
Emerging Europe and CIS -0.187 -0.136 -0.188 -0.152
Sub-Saharan Africa -0.178 -0.163 -0.170 -0.145
Other Robustness Checks
Panel A. HP 12 Panel B. SURE
Panel C. Innovation Panel D. IV
0.5
11.5
22.5
De
nsity
-.6 -.4 -.2 0 .2HP12
0.5
11.5
22.5
De
nsity
-.8 -.6 -.4 -.2 0 .2sure
0.5
11.5
22.5
De
nsity
-.8 -.6 -.4 -.2 0 .2bq
0.5
11.5
2D
ensity
-1 -.5 0iv
Determinants of Okun’s Law coefficient(OECD and Non-OECD)
0 10 20 30 40-1
-0.8-0.6-0.4-0.2
00.20.4
Average Unemployment Rate
Average Unemployment Rate
Oku
n’s
Law
Coe
ffici
ent
0 25 50 75-1
-0.8-0.6-0.4-0.2
00.20.4
Informality
Informality
Oku
n’s
Law
Coe
ffici
ent
0 3 6 9 12-1
-0.8-0.6-0.4-0.2
00.2
Labor Market Flexibility
Labor Market Flexibility
Oku
n’s
Law
Coe
ffici
ent
0 3 6 9 12-1
-0.8-0.6-0.4-0.2
00.2
Product Market Flexibility
Product Market Flexibility
Oku
n’s
Law
Coe
ffici
ent
Determinants of Okun's Law Coefficients (OECD and Non-OECD) (1) (2) (3) (4) (5)Average Unemployment -0.009*** -0.017*** -0.014*** [-3.466] [-4.723] [-3.785] Informality 0.003*** 0.003*** 0.004*** [4.162] [5.254] [5.705] Services (% of GDP) -0.003*** 0.000 [-6.953] [-0.015] Constant 0.034*** -0.235*** 0.117*** -0.157*** -0.195*** [2.897] [-7.991] [6.589] [-5.066] [-3.058] N 102 92 96 92 87R2 0.107 0.161 0.34 0.329 0.355 Note: t-statistics are in parenthesis. ***, **, and * indicates significance at 1, 5, and 10 percent levels, respectively.
Determinants of Okun's Law Coefficients (OECD and Non-OECD) (1) (2) (3)Average Unemployment -0.017*** -0.019*** -0.020*** [-4.072] [-4.569] [-4.763] Informality 0.004*** 0.002*** 0.002*** [5.468] [2.786] [3.062] Services (% of GDP) 0.001 0.001 0.002 [0.946] [1.116] [1.660] Labor Market Flexibility -0.015** -0.012* [-2.298] [-1.735] Product Market Flexibility -0.036*** -0.029** [-3.110] [-2.369] Constant -0.151** -0.014 -0.019 [-2.089] [-0.155] [-0.204] N 74 71 69R2 0.412 0.467 0.488 Note: t-statistics are in parenthesis. ***, **, and * indicates significance at 1, 5, and 10 percent levels, respectively.
Data QualityRelevance
What we’ve done thus far:
1. Looked more intensively at Latin American countries, for which we had a better data set
2. Re-done analysis using employment (rather than unemployment) – this work is preliminary
New Data Set on Unemployment for Latin America and the Caribbean
Laurence Ball, Nicolás De Roux and Marc Hofstetter, Unemployment in Latin America and the Caribbean, IMF WP 11/252, 2011
• “A major reason that Latin American unemployment is understudied is lack of data. Unemployment statistics are fragmentary, and there are big differences in how unemployment is measured across countries and over time. Therefore, a major part of our project involves data gathering.”
• Ball et al compile a data set for 19 countries that covers far more years than the commonly-used IADB data.
– “To derive our data, we have gone country by country to figure out how unemployment was measured in different periods. We have made judgments about which changes in methodology are small enough to ignore, and how to adjust for larger changes. In some cases we can splice different unemployment series together using periods in which they overlap. When in doubt, we have sought advice from people at the agencies that produce unemployment data. “
1979
1980
1981
198219831984
1985
1986
198719881989
1990
1991
1992
1993
1994
1995
1996
19971998
19992000
20012002
2003
2004200520062007
20082009
20102011
-50
510
u_ch
g
-10 -5 0 5 10y_chg
1979-2011 (33 obs) slope: -0.22 (se = 0.05)Adjusted R-squared = 0.36 RMSE = 1.72
1978
19791980
19811982
1983
1984
1985
198619871988
1989
1990
1991 1992
1993
1994
1995
1996
1997
1998
19992000
2001
2002
2003
2004
20052006
20072008
2009 2010 2011
-4-2
02
46
ur_s
tar1
00
-20 -10 0 10y_star100
1978-2011 (34 obs) slope: -0.11 (se = 0.05)Adjusted R-squared = 0.08 RMSE = 1.95
1978
19791980
19811982
1983
1984
1985
1986
19871988
1989
1990
19911992
19931994
1995
1996
1997
1998
19992000
2001
2002
2003
20042005
2006
20072008
2009 2010 2011
-20
24
ur_s
tar1
2
-15 -10 -5 0 5y_star12
1978-2011 (34 obs) slope: -0.24 (se = 0.05)Adjusted R-squared = 0.39 RMSE = 1.24
Argentina
19831984
1985
1986
19871988
1989
1990
1991
1992
199319941995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
20072008
2009
20102011
-2-1
01
2u_
chg
-5 0 5 10y_chg
1983-2011 (29 obs) slope: -0.21 (se = 0.04)Adjusted R-squared = 0.45 RMSE = 0.68
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996 1997
19981999
2000
2001
2002
2003
2004
2005
2006
2007
20082009
20102011
-2-1
01
2ur
_sta
r100
-10 -5 0 5 10y_star100
1982-2011 (30 obs) slope: -0.22 (se = 0.03)Adjusted R-squared = 0.59 RMSE = 0.55
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
19961997
1998
1999
2000
2001
2002
2003
2004
2005
2006 2007
2008
2009
20102011
-1-.
50
.51
1.5
ur_s
tar1
2
-10 -5 0 5 10y_star12
1982-2011 (30 obs) slope: -0.20 (se = 0.04)Adjusted R-squared = 0.50 RMSE = 0.47
Brazil
19791980
1981
1982
1983
1984
1985
198619871988 1989
1990
19911992
19931994 1995
1996
19971998
1999
2000
2001
20022003
2004
20052006
2007
2008
2009
20102011
-50
510
15u_
chg
-15 -10 -5 0 5 10y_chg
1979-2011 (33 obs) slope: -0.57 (se = 0.08)Adjusted R-squared = 0.59 RMSE = 2.16
19781979
1980
1981
19821983
1984
198519861987
19881989
19901991
199219931994 1995
199619971998
1999
20002001
20022003
2004 2005 2006
20072008
2009
20102011
-10
-50
510
ur_s
tar1
00
-10 -5 0 5 10 15y_star100
1978-2011 (34 obs) slope: -0.47 (se = 0.07)Adjusted R-squared = 0.54 RMSE = 2.00
1978
1979
1980
1981
1982
1983
1984
198519861987
1988 1989
19901991
199219931994
19951996
1997
1998
1999
20002001
20022003
20042005 2006
20072008
2009
20102011
-10
-50
5ur
_sta
r12
-5 0 5 10 15y_star12
1978-2011 (34 obs) slope: -0.56 (se = 0.07)Adjusted R-squared = 0.65 RMSE = 1.33
Chile
19791980
19811982
1983
19841985
1986
19871988
19891990
19911992
1993
1994
1995
1996
1997
19981999
2000
20012002
2003
2004
2005
2006
20072008
2009
20102011
-2-1
01
23
u_ch
g
-10 -5 0 5 10y_chg
1979-2011 (33 obs) slope: -0.20 (se = 0.03)Adjusted R-squared = 0.55 RMSE = 0.68
1978
1979
19801981
1982
1983
1984
19851986
19871988
1989
1990 19911992
1993 1994
1995
1996
1997
1998
19992000
20012002
2003
2004
2005
200620072008
2009 20102011
-2-1
01
23
ur_s
tar1
00
-10 -5 0 5 10y_star100
1978-2011 (34 obs) slope: -0.17 (se = 0.04)Adjusted R-squared = 0.39 RMSE = 0.77
1978
1979
198019811982
1983
1984
19851986
19871988
1989
1990 199119921993
1994
1995
1996
1997 1998
1999 2000
20012002
2003
2004
2005
2006 20072008
20092010
2011
-2-1
01
2ur
_sta
r12
-5 0 5 10y_star12
1978-2011 (34 obs) slope: -0.22 (se = 0.03)Adjusted R-squared = 0.63 RMSE = 0.47
Mexico
Based on y_chg
β Obs Adjusted
R² RMSE
Argentina -0.221*** (0.0524) 32 0.352 1.751
Brazil -0.215*** (0.0438) 29 0.452 0.685
Chile -0.580*** (0.0840) 32 0.601 2.177
Colombia -0.622*** (0.0885) 32 0.609 1.048
Costa Rica -0.221*** (0.0496) 32 0.378 0.960
Ecuador -0.355** (0.152) 21 0.181 1.982
El Salvador 0.000950 (0.116) 23 -0.048 1.254
Honduras 0.0421 (0.115) 21 -0.045 1.313
Jamaica -0.154** (0.0716) 32 0.105 1.109
Mexico -0.207*** (0.0339) 32 0.540 0.690
Nicaragua -0.170* (0.0928) 32 0.070 2.039
Panama -0.204*** (0.0357) 32 0.506 0.975
Peru -0.118*** (0.0382) 32 0.216 1.298
Uruguay -0.279*** (0.0524) 32 0.468 1.332
Venezuela -0.212*** (0.0318) 32 0.584 1.199
Average -0.234 0.331
Okun’s Coefficients for Latin America
Distribution of Employment Responsiveness
Okun’s Law Coefficients:Unemployment Elasticities vs. Employment Elasticities
Determinants of Okun's Law Coefficients (Employment Elasticities) (1) (2) (3) (4) (5)
Average Employment Growth -0.017 -0.036* -0.023 [-0.885] [-1.793] [-1.040] Informality -0.006*** -0.007*** -0.004** [-3.889] [-4.257] [-2.264] Services (% of GDP) 0.006*** 0.005*** [5.164] [2.990] Constant 0.148*** 0.383*** -0.153*** 0.465*** 0.132 [4.172] [5.610] [-2.815] [5.715] [0.966] N 82 72 76 72 67R2 0.01 0.178 0.265 0.214 0.314 Note: t-statistics are in parenthesis. ***, **, and * indicates significance at 1, 5, and 10 percent levels, respectively.
Determinants of Employment Elasticities (Non-OECD) (1) (2) (3)Average Employment Growth -0.026 -0.039 -0.031 [-1.109] [-1.543] [-1.251] Informality -0.004* -0.003 -0.003* [-1.884] [-1.522] [-1.716] Services (% of GDP) 0.008*** 0.004 0.006** [4.173] [1.577] [2.433] Labor Market 0.086** 0.084** [2.644] [2.549] Product Market 0.007 0.009 [0.566] [0.693] Constant -0.625** 0.095 -0.589** [-2.272] [0.653] [-2.094] N 54 54 52R2 0.455 0.349 0.45 Note: t-statistics are in paranthesis. ***, **, and * indicates significance at 1, 5, and 10 percent levels, respectively.
Okun Outside OECDTentative Conclusions
1. Law holds better-than-(we)-expected.
2. Coefficient for non-OECD countries is smaller (around -0.2 on average) than for OECD countries, but there is considerable heterogeneity
3. Variation in Okun’s coefficient can be accounted for partially by country characterics such as degree of informality and product market flexibility