Potential Growth and Structural Unemployment in Spain, EMU and the US

  • 404 views
Uploaded on

Many estimates of structural unemployment are very procyclical …

Many estimates of structural unemployment are very procyclical
In most cases, these procyclicality of structural unemployment is the main
cause of the procyclicality of potential growth
On this respect, the evidence for the Spanish structural unemployment rate
estimated by European Commission is a clear example
This procyclicality affects the estimation of important gaps in policy making
as, for example, the cyclical component of budget balance

  • Full Name Full Name Comment goes here.
    Are you sure you want to
    Your message goes here
    Be the first to comment
    Be the first to like this
No Downloads

Views

Total Views
404
On Slideshare
0
From Embeds
0
Number of Embeds
0

Actions

Shares
Downloads
7
Comments
0
Likes
0

Embeds 0

No embeds

Report content

Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
    No notes for slide

Transcript

  • 1. Potential Growth and Structural Unemploymentin Spain, EMU and the USRafael DoménechChief Economist for Developed EconomiesBrussels, May 20131/34
  • 2. . MotivationWhenever unemployment stays high for an extended period, it is common tosee analyses, statements, and rebuttals about the extent to which the highunemployment is structural, not cyclical.Peter Diamond, 20132/34
  • 3. . MotivationMany estimates of structural unemployment are very procyclicalIn most cases, these procyclicality of structural unemployment is the maincause of the procyclicality of potential growthOn this respect, the evidence for the Spanish structural unemployment rateestimated by European Commission is a clear exampleThis procyclicality affects the estimation of important gaps in policy makingas, for example, the cyclical component of budget balance3/34
  • 4. . MotivationUnemployment rate and its structuralcomponent, Spain 1980-2012Source: European Commission7%9%11%13%15%17%19%21%23%25%27%1980 1984 1988 1992 1996 2000 2004 2008 2012Growth of GDP and potential growth, Spain1981-2012Source: European Commission-4%-3%-2%-1%0%1%2%3%4%5%1981 1985 1989 1993 1997 2001 2005 20094/34
  • 5. . MotivationUnemployment rate and its structuralcomponent, Spain 1980-2012Source: OECD7%9%11%13%15%17%19%21%23%25%27%1980 1984 1988 1992 1996 2000 2004 2008 2012OECD estimate also procyclical, but lessthan in the case of the ECNAIRU increase: 4.7pp (OECD) vs 11.2(EC)Debate about the interactions betweenshocks and institutions5/34
  • 6. . A useful decomposition of GDPGDP can be decomposed in terms of the working-age population, L15−64t :GDPt =GDPtL15−64tL15−64tor, in growth of rates∆ ln GDPt = ∆ ln(GDPtL15−64t)+ ∆ ln L15−64tAdditionally, GDP per working-age population can be decomposed asGDPtL15−64t=GDPtHtHtLdt(1 − ut)LstL15−64twhere H is the total numbers of hours worked, Ld is total employment and Lsis labour supply.6/34
  • 7. . A useful decomposition of GDPThe decomposition of GDP per hour using the production function approachimplies usually the specification of a Cobb-Douglas production function suchaslnGDPtHt= ln At + α ln(KtHt)where capital is, in some cases, corrected by capacity utilization.7/34
  • 8. . A useful decomposition of GDPHere, we use an alternative approach decomposing the log of GDP (gdp) interms of the trend and cyclical components of the log of GDP perworking-age population (y ≡ ln(GDP/L15−64)) and of the latter variable(l = ln L15−64). Sincegdpt ≡ yt + ltthengdpt = gdpt + gdpct = yt + lt + yct + lctwhere, as usual, the bar over the variables represents the trend componentsand the superscript c denotes the cyclical component.The variable that we use to identify the cycle is the unemployment rate.Two reasons for the choice of this variable:▶ The economic relevance of the unemployment rate▶ Its correlation with other components in the decomposition of GDP(capacity utilization, activity rate, growth of working-age population, etc.)8/34
  • 9. . The Okuns lawOur identification scheme is based on the Okuns law (e.g., Ball et al., 2013)and its usefullness to identify the trend component of GDP and theunemployment rate (e.g., Doménech and Gómez, 2006):ut − ut = β(gdpt − gdpt)+ εtAn alternative to the preceding equation is to use GDP per working agepopulation (y) instead of GDP:ut − ut = β (yt − yt) + εt9/34
  • 10. . The Okuns lawIn both specifications, the problem is that the trend components of u, gdp,and y are not observed.The usual approach is to estimate these trend components using theHodrick-Prescott filter.With annual data, most researchers have used a smoothing parameterbetween 100 (e.g., Backus and Kehoe, 1992) and 400 (e.g., Correia, Nevesand Rebelo, 1992, or Cooley and Ohanian, 1991).These values are above 6.25, which corresponds to the standard value of1600 used with quarterly data (Ravn and Uhligh, 2002).10/34
  • 11. . Preliminary evidenceut − ut = β (yt − yt) + εtEstimates of the Okuns Law, 1980-2012GDP y(1) (2) (3) (4) (5) (6)USA β −0.52(12.8)−0.51(13.6)−0.48(11.4)−0.53(12.4)−0.51(12.6)−0.50(11.9)R2 0.84 0.85 0.80 0.83 0.83 0.81DW 1.61 0.91 0.53 1.54 0.81 0.58EMU β −0.41(10.9)−0.48(14.5)−0.48(16.6)−0.49(11.1)−0.45(13.5)−0.41(13.4)R2 0.79 0.87 0.90 0.79 0.85 0.85DW 1.24 1.05 0.98 1.22 0.83 0.58Spain β −0.98(11.9)−0.96(15.1)−0.91(13.6)−0.97(11.1)−1.07(17.3)−1.07(20.2)R2 0.82 0.88 0.85 0.79 0.90 0.93DW 1.59 0.63 0.35 1.63 1.05 0.89λ 6.25 100 400 6.25 100 400λ is the smoothing parameter of the HP filter11/34
  • 12. . Preliminary evidenceSimilar values of β for the USA and EMU, clearly higher (in absolute terms)for SpainThe estimated values of β are very statistically significant and robust tochanges in λHigh R2High autocorrelation of residuals (low DW), particularly for high values of λSimilar results for GDP and GDP per working-age population12/34
  • 13. . An unobserved component modelWe asumme that GDP per working-age population can be decomposed as:yt ≡ yt + yctyt = γyt + yt−1γyt = γyt−1 + ωγt,where ωγt is i.i.d.Therefore, we assume stochastic growth for the trend component, that is,that trend GDP is I(2)13/34
  • 14. . An unobserved component modelIn the same vein, the unemployment rate can be decomposed as:ut ≡ ut + uctut = ut−1 + ωutwhere ωut is i.i.d.Therefore, we assume that the unobserved component of the unemploymentrate is I(1)Additionally:ut − ut − ρut−1 = β (yt − yt) − ρβ (yt−1 − yt−1) + εtPrevious equation is a more general case (allowing for autocorrelation),which collapses to the standard specification when ρ = 014/34
  • 15. . An unobserved component modelWe can write all previous equations in state-space form:ytyt−1utut−1 =2 −1 0 01 0 0 00 0 1 00 0 1 0yt−1yt−2ut−1ut−2 +ωyt0ωut0ytutut − βyt − ρ(ut−1 − βyt−1) =1 0 0 00 0 1 0−β ρβ 1 −ρytyt−1utut−1 +yctuctvutwhereσ2yc = λσ2ωy, σ2uc = λσ2ωu= µλσ2ωy, σ2vu= γλσ2ωy15/34
  • 16. . ResultsParameters estimates, 1980-2012γ µ σωy λ β ρUSA 2.49(3.17)0.56(4.05)0.002(13.9)61.6(4.91)−0.50(11.3)0.82(8.18)EMU 1.19(2.57)0.36(4.05)0.002(13.9)67.6(4.93)−0.41(11.0)0.81(10.3)Spain 4.09(2.21)1.94(4.07)0.003(13.9)65.3(4.49)−1.10(11.0)0.81(6.61)16/34
  • 17. . ResultsUnemployment rate and its structuralcomponent, Spain 1980-2012Source: BBVA Research7%9%11%13%15%17%19%21%23%25%27%1980 1984 1988 1992 1996 2000 2004 2008 2012Structural unemployment rate relativelystable from 1980Consistent with the absence ofstructural reforms in the labour marketStructural unemployment has increased4 pp during the latest crisis17/34
  • 18. . ResultsUnemployment rate and its structuralcomponent, USA 1980-2012Source: BBVA Research3%4%5%6%7%8%9%10%1980 1984 1988 1992 1996 2000 2004 2008 2012Unemployment rate and its structuralcomponent, EMU 1980-2012Source: BBVA Research3%4%5%6%7%8%9%10%11%12%1980 1984 1988 1992 1996 2000 2004 2008 201218/34
  • 19. . ResultsGrowth of GDP per working-age population,Spain 1981-2012Source: BBVA Research-4.5%-3.5%-2.5%-1.5%-0.5%0.5%1.5%2.5%3.5%4.5%1981 1985 1989 1993 1997 2001 2005 2009Significant reduction of GDP per wapgrowth in the years of the housing andfinancial bubblesPotential growth similar to the crisis inthe second half of the 70s andfirst 80s ...... but not negative: close to 1% andvery similar to the potential growthobserved in the USA and EMU19/34
  • 20. . ResultsGrowth of GDP per working-age population,USA 1981-2012Source: BBVA Research-5.0%-3.0%-1.0%1.0%3.0%5.0%1981 1985 1989 1993 1997 2001 2005 2009Growth of GDP per working-age population,EMU 1980-2012Source: BBVA Research-5.0%-3.0%-1.0%1.0%3.0%5.0%1981 1985 1989 1993 1997 2001 2005 200920/34
  • 21. . ResultsGrowth of working-age population, Spain, EMUand USA 1981-2012Source: BBVA Research-1.5%-1.0%-0.5%0.0%0.5%1.0%1.5%2.0%2.5%1981 1985 1989 1993 1997 2001 2005 2009USA  EMU  Spain  USA: stable growth of WAP around 1%EMU: slightly negative trend, growtharound 0.3%Spain: (1) very volatile growth, (2)immigration, and (3) negative growthsince 201021/34
  • 22. . Extension I: investment and unemployment ratesInvestment and unemployment rates, Spain, EMUand USA 1981-2012Source: BBVA Research80  81   82  83  84  85  86  87  88  89  90  91  92  93   94  95  96  97  98  99  00  01  01  03  04  05  06  07  08  09  10  11  12  80  81  82  83  84  85  86  87  88  89  90  91  92  93  94  95  96  97  98  99  00  01  01  03  04  05  06  07  08  09  10  11  12  80  81  82  83  84  85  86  87  88  89  90  91  92  93  94  95  96  97  98  99  00  01  01  03  04  05  06  07  08  09  10  11  12  0.14  0.16  0.18  0.20  0.22  0.24  0.26  0.28  0.30  0.32  0.02   0.06   0.10   0.14   0.18   0.22   0.26  Investment  rate  Unemployment  rate  Spain  80  84  EMU  USA  Significant and relatively stable negativecorrelation (-0.79)%USA: stable correlation since mid 80sSpain: stable correlation since 1985,with a shift of 4 pp in theunemployment rate since 2009.22/34
  • 23. . Extension I: investment and unemployment ratesWe extend the UCM with a new state for the investment rate:ytyt−1utut−1irtirt−1=2 −1 0 0 0 01 0 0 0 0 00 0 1 0 0 00 0 1 0 0 00 0 0 0 1 00 0 0 0 1 0yt−1yt−2ut−1ut−2irt−1irt−2+ωyt0ωut0ωirt023/34
  • 24. . Extension I: investment and unemployment ratesWe also change the UCM with a new measurement equation:ytutirtut − βuyt − ρu(ut−1 − βuyt−1)ut − βirirt − ρir(ut−1 − βirirt−1)= A¯yt¯yt−1¯ut¯ut−1¯irt¯irt−1+yctuctirctvutvirtwhereA =1 0 0 0 0 00 0 1 0 0 00 0 0 0 1 0−βy ρyβy 1 −ρy 0 00 0 1 −ρir −βir ρirβir24/34
  • 25. . Extension I: resultsParameters estimates, 1980-2012γy µy σωy λ βy ρy γir µir βir ρirUSA 1.87(3.40)0.67(3.92)0.002(17.8)54.6(5.91)−0.49(12.1)0.79(7.26)3.35(3.57)0.63(4.08)−0.72(9.13)0.86(9.22)EMU 0.75(2.54)0.36(4.05)0.002(13.9)66.6(6.04)−0.40(12.8)0.80(11.8)1.64(3.41)0.36(4.01)−0.64(8.84)0.74(9.00)Spain 3.10(2.45)2.37(4.46)0.003(17.8)55.7(5.59)−1.00(11.5)0.81(7.12)0.94(1.27)1.02(4.32)−1.16(14.3)0.71(9.39)25/34
  • 26. . Extension I: resultsSpain: structural unemployment rate, 1980-2012Source: BBVA Research7%9%11%13%15%17%19%21%23%25%27%1980 1984 1988 1992 1996 2000 2004 2008 2012When the investment rate is taken intoaccount, the structural unemploymentrate is similar to the previous estimateIn most of the years, the previousestimate is inside the new confidenceintervalNevertheless, the structuralunemployment rate increases slightlymore in the latest years26/34
  • 27. . Extension I: resultsUnemployment rate and its structuralcomponent, USA 1980-2012Source: BBVA Research3%4%5%6%7%8%9%10%1980 1984 1988 1992 1996 2000 2004 2008 2012Unemployment rate and its structuralcomponent, EMU 1980-2012Source: BBVA Research3%4%5%6%7%8%9%10%11%12%1980 1984 1988 1992 1996 2000 2004 2008 201227/34
  • 28. . Extension II: assuming that u is I(2)Now we assume that in the UCM the structural unemployment rate is I(2):ytyt−1utut−1irtirt−1=2 −1 0 0 0 01 0 0 0 0 00 0 2 −1 0 00 0 1 0 0 00 0 0 0 1 00 0 0 0 1 0yt−1yt−2ut−1ut−2irt−1irt−2+ωyt0ωut0ωirt028/34
  • 29. . Extension II: assuming that u is I(2)Spain: structural unemployment rate, 1980-2012Source: BBVA Research7%9%11%13%15%17%19%21%23%25%27%1980 1984 1988 1992 1996 2000 2004 2008 2012Similar u to the one estimated by the ECBetter results for Okuns law andinvestment equation when u is assumedto be I(1)I(2) assumption not supported by unitroot tests29/34
  • 30. . Extension II: assuming that u is I(2)First difference of the unemployment rate:Spain, USA and EMU, 1980-2012Source: BBVA Research-­‐0.03  -­‐0.02  -­‐0.01  0.00  0.01  0.02  0.03  0.04  0.05  0.06  0.07  1980   1984   1988   1992   1996   2000   2004   2008   2012  (1-­‐L)  Unemployment  rate  Spain  EMU  USA  The first difference of u seems to bestationaryHigher volatility in SpainHigh correlation of ∆u30/34
  • 31. . Extension II: assuming that u is I(2)Dickey-Fuller Unit Root Testt-statistic 5% critical valueUSA u -2.97 -2.96 Reject I(1)∆u -4.21 -1.95 Reject I(2)EMU u -3.20 -2.95 Reject I(1)∆u -3.61 -1.95 Reject I(2)Spain u -2.09 -2.95 Accept I(1)∆u -2.66 -1.95 Reject I(2)31/34
  • 32. . Implications for the structural budget deficitUnemployment and budget balance, Spain1980-2012Source: BBVA Research808182838485868788899091 92939495969798990001 0203040506070809101112(f)12e6e-12-10-8-6-4-20247 9 11 13 15 17 19 21 23 25 27Budgetbalance(%GDP)Unemployment rate (%)1.7%  -­‐5.9%  21.67%  (2012)  10.96%  (2006)  18%  (2012)  -­‐3%  The structural budget deficit estimatedby the EC is also prcyclicalWith our estimates of structuralemployment the structural deficit is lessprocyclicalLarger structural deficits in the boomand smaller in the crisis32/34
  • 33. . Extensions (work in progress)Inclusion of financial variables (as in Borio et al., 2013)Labour market variables (vacancies and the Beveridge curve)Unemployment gap and wage inflation. Is inflation helpful for estimating theunemployment gap? Globalization, composition effects, etc.33/34
  • 34. . ConclusionsEstimates of structural unemployment based only in the information contentof wages or price inflation are often very procyclicalOther economic variables (such as the GDP, investment rates, etc.) containuseful information about the cyclical and structural components ofunemployment ratesBased on this information, our estimates show a more stable behaviour ofthe structural unemployment rateThe unemployment rate and its structural component also contains veryuseful information to asses the fiscal stance of budget balances34/34