Presentation is loading. Please wait.

Presentation is loading. Please wait.

How do innovation and imitation change the short run impact of GDP on unemployment ? Boussemart J.P. Briec W. Tavéra C.

Similar presentations

Presentation on theme: "How do innovation and imitation change the short run impact of GDP on unemployment ? Boussemart J.P. Briec W. Tavéra C."— Presentation transcript:

1 How do innovation and imitation change the short run impact of GDP on unemployment ? Boussemart J.P. Briec W. Tavéra C.

2 Introduction The Okuns law relationship as an empirical regularity (Okun, 1962) : Developments – Theoretical background (Prachowny 1993,…) – Empirical analysis of the dynamic effects of GDP on unemployment (Crespo-Cuaresma 2003, Silvapulle et al. 2004) : OLC(Expansion) < OLC (Recession) Okuns law as a demand driven macroeconomic mechanism real GDP = +1% unemployment rate = -0.3 pt of %

3 A simple graphic version of the OL (Bureau of Economic Analysis for US datas)

4 Introduction This paper aims at : – Reexamining the supply side aspect of the OL mechanims (distinction potential-observed real output) – Re-visiting the OL by introducing the influence of technical change and technological distance – Evaluating the short-run impact of technology- driven output movements on unemployment

5 Technical progress and catching up Innovation and imitation : the simple diagram Technological leader Innovation Shifts of the Technological frontier Follower country imitation

6 Technical progress and catching up Innovation and imitation as complementary proceses (Benhabib-Spiegel 994, Acemoglu- Aghion-Zilibotti, 2002)

7 Specification 1: an interaction-augmented- version of the OL relationship First order effects : linear effects Non linear effects : Squared variables Interaction terms (cross-terms)

8 Specification 2 : OL relationship with threshold Threshold variable Z : Technical progress or Technological distance with the leader Estimation method suggested by Hansen (1999) : Min square estimate of the threshold Test for significativeness of the threshold

9 The measure of productivity gaps The technological gap is measured in terms of TFP levels between any country i and the leader (Malmquist index : Färe et al. 1994). At time t, the production set is defined as T = {, X can produce Y} : T satisfies strong disposability and convexity assumptions and we assume constant returns to scale The distance between country i and the world frontier can be decomposed into two components : – The time change in the technical efficiency – The geometric mean if the shift of the frontier

10 Technical change and productivity gaps Productivity variation in country i Technological gap variation (catching up) Movement of the technological frontier (technical change) Pays i (initial) Pays i (final) Leader (initial) Leader(final) Inputs X Output Y

11 Data Annual data, oecd countries : Austria, Belgium, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, Netherlands, Portugal, Spain, Sweden, UK, and USA. 400 observations for each variable Equilibrium levels of output and unemployment : HP filter

12 Empirical results 0 : a preliminary analysis of the basic OL model


14 Empirical results 1 : the interaction- augmented-version of the OL relationship

15 First order linear approximation of the impact of GDP on unemployment rate : – D TC with k = – D TC with k = 1 The total effect of a 1% rise in output on unemployment variation is twice the first order effect for a technological distance close to 16% The impact of a 1% rise in output on unemployment variation is zero when technical change is close to 9.2% % Very rapid increases in the rhythm of technical change can thus lead to a reversal of the traditional effect on unemployment movements in the short run

16 Empirical results 2 : the threshold version of the OL relationship



19 The short run impact (in absolute value) of GDP movements on unemployment rate is : larger when the technological distance is large (imitation) close to zero for countries close to the technological frontier smaller when the size of technical progress is large (innovation)

20 Some concluding remarks The OL relationship does not contain only demand induced macroeconomic mechanims The origins of variations in TFP matter for determining the total impact of GDP movemements on unemployment, even in the short run Imitation and innovation generate second order non linear mechanisms that can boost or mitigate the traditional first order OLC Our results lend suppport to recent empirical papers which show that the ouput-unemployment relationship might be dominated by permanent shocks rather than by temporary shocks only (Sinclair 2009)

21 How many true values are there for the Okuns Law coefficient? One or Two ? A meta-analysis of empirical results Roger Perman (a) - Gaetan Stephan (b) - Christophe Tavéra (b) University of Strathclyde CREM, CNRS – Université de Rennes 1

22 Loi dOkun Exemple : Etats-Unis, , données trimestrielles, Coefficient moyen = -0.41

23 Objectif / Methode Objectif : estimer le coefficient dOkun Méthode : – Ne pas utiliser un nouvelle base de données – Utiliser les estimations obtenues dans la littérature et les caractéristiques des analyses économétriques correspondantes

24 Les catégories de modélisations

25 Méthode déchantillonnage : Etape 1 Recherche darticles dans Econlit avec critères : – mots clés : Okuns Law – Output-unemployment relationship – Presence dun abstract (verification estimation présente) – Publication après 1980 – Presence dans Econlit en décembre 2010 Papiers identifiés : 97

26 Méthode déchantillonnage : Etape 2 Exclusion des articles – Ne contenant pas une estimation originale de la loi dOkun – Ne précisant pas suffisamment les caractéristiques de lestimation (période, etc.) – Contenant des estimations de modèles non linéaires de la loi dOkun Papiers retenus : 30

27 Cycle de vie de la publication

28 Homogénéisation des estimations

29 Caractéristiques statistiques de léchantillon

30 Meta régression : Biais et tests BiaisTests Type 1Test FAT de Stanley Type 2Galbraith Plot

31 Meta régression : tests de biais

32 Meta régression multuvariée DiversSampleFrequencyCountryModelEndogen ous Filter Firstyear Lastyear Pubyear SampTS SampPA FreqY FreqSQ CountDED CountDING Count Reg ModSTA ModDYN Othexo Noothexo Neq1 NeqN EndY EndU Level Delta FiltLT FiltHP FiltBK FiltBN FiltUC FiltMOD Principales dummies retenues pour la régression multivariée

33 Quelques résultats sur les biais Test dabsence de biais de type 1

34 Quelques résultats de la méta régression multivariée Whole sample Unemployment sub- sample Output sub-sample OLS STEPWISE procedure IRLS procedure STEPWISE then IRLS STEPWISE then IRLS Constant -240,409 (-2,01)-189,478 (-1,73)-194,449 (-3.00) (-4.23) Precision -0,400 (-3,08)-0,454 (-10,23)-0,528 (-9,44) (-9.09) (-13.79) SAMPPA -0,261 (-1,74)-0,292 (-2,06)-0,174 (-1,80) FREQSQ 0,152 (1,37)0,149 (4,95)0,186 (4,38) (5.40) (-11.56) COUNTDING 0,188 (3,83)0,193 (4,40)0,225 (4,83) (7.44) REG 0,334 (2,67)0,333 (2,76)0,293 (3,71) (2.86) MODDYN 0,117 (2,36)0,151 (3,55)0,145 (2,96) (10.07) OTHEXO 0,138 (2,16)0,186 (3,48)0,218 (5,54) (-5.74) NEQN -0,057 (-1,65)-0,058 (-1,88) ENDY -0,437 (-3,35)-0,455 (-5.00)-0,390 (-6,22) LEVEL -0,124 (-1,71) (-7.48)1.274 (8.69) FILTLT -0,153 (-1,09) FILTHP -0,031 (-0,54) FILTBK -0,160 (-1.00) (2.58) FILTBN -0,300 (-1,20) FILTUC -0,019 (-0,16) FILTMOD 0,545 (0,88) AVGYEAR 0,120 (1,99)0,095 (1,71)0,097 (2,96) (4.20) LOGNOBS R2 0,652 0,6430, F-test (P. value) Reset test (P. value)0.061 (0.80)0.691 (0.41)0.024 (0.87) (0.33)0.557 (0.46)

Download ppt "How do innovation and imitation change the short run impact of GDP on unemployment ? Boussemart J.P. Briec W. Tavéra C."

Similar presentations

Ads by Google