Structural reforms in DSGE models
- 1. Structural reforms in DSGE models a plead for sensitivity analysis B. Campagne, A. Poissonnier Insee/Crest/Ecole Polytechnique Society for Economic Measurement Conference 22nd July 2015 This work reflect the opinions of the authors and not necessarily the views of Insee. Campagne, Poissonnier (Insee) DSGE models’ sensitivity SEM 2015 1 / 29
- 2. 1 Motivation 2 A standard new-Keynesian DSGE model 3 Structural reforms : labour and goods markets deregulation 4 Structural reforms : a sensitivity analysis 5 Conclusion Campagne, Poissonnier (Insee) DSGE models’ sensitivity SEM 2015 2 / 29
- 3. Plan 1 Motivation 2 A standard new-Keynesian DSGE model 3 Structural reforms : labour and goods markets deregulation 4 Structural reforms : a sensitivity analysis 5 Conclusion Campagne, Poissonnier (Insee) DSGE models’ sensitivity SEM 2015 3 / 29
- 4. Increasing use of DSGEs for policy purposes DSGE models supplanted macroeconometric models, now widely used in governments, central banks and international institutions : Quest III (EC), NAWM or EAGLE (ECB), GIMF (IMF), SIGMA (FED)... Vast literature using DSGE models for the evaluation of structural reforms (ie. deregulation of goods and labour markets) : Quest III : Roeger et al. (2008), D’Auria et al. (2009), Annicchiarico et al. (2013), Varga et al. (2014) GEM : Bayoumi et al. (2004) or Everaert and Schule (2006, 2008) for European countries OECD : Cacciatore et al. (2012) Central Banks : Forni et al. (2010) Campagne, Poissonnier (Insee) DSGE models’ sensitivity SEM 2015 4 / 29
- 5. Motivation In order to conduct policy evaluations in DSGE models : Do we understand how structural reforms affect the economy ? Are DSGE models qualitatively stable enough ? Are DSGE models quantitatively stable enough ? Campagne, Poissonnier (Insee) DSGE models’ sensitivity SEM 2015 5 / 29
- 6. Our main results Structural reforms simulations in a Smets and Wouters’ type of model are in line with Blanchard and Giavazzi (2003) The consumption-leisure arbitrage plays a key role in the transmission of structural reforms to the economy Therefore, the Frisch elasticity (as well as the risk aversion, and habits) are crucial for the quantitative assessment of reforms Welfare analysis shows households might loose upon reform Campagne, Poissonnier (Insee) DSGE models’ sensitivity SEM 2015 6 / 29
- 7. Plan 1 Motivation 2 A standard new-Keynesian DSGE model 3 Structural reforms : labour and goods markets deregulation 4 Structural reforms : a sensitivity analysis 5 Conclusion Campagne, Poissonnier (Insee) DSGE models’ sensitivity SEM 2015 7 / 29
- 8. A three-slide overview of the model Core structure New Keynesian model à la Smets and Wouters (2003) : nominal rigidities on prices and wages, real rigidities on capital ... and all the "traditional" ingredients Households’ utility : non-separable (King et al., 2002), with consumption multiplicative habits (Abel, 1990 or Gali, 1994) Ut(τ) = 1 1−σc Ct(τ) Ct−1 N −hc 1−σc 1 − κ(1 − σc)lt(τ)1+σl σc Alternative models : with no consumption habits with labour habits with non Ricardian agents à la Campbell and Mankiw Campagne, Poissonnier (Insee) DSGE models’ sensitivity SEM 2015 8 / 29
- 9. A three-slide overview of the model Calibration Detailed steady state calibration for the Eurozone taken as a closed economy (in line with Smets and Wouters, 2005) DATA MODEL EA (12) EA Output in 2000 (GDP)∗ 6943 6943 Output per capita growth rate∗∗ 1.2% 1.2% Working age population in 2000 ∗∗∗ 135.9 135.9 Hours worked per week and working age capita (since 2000) 34.3 34.3 Gross Op. Surplus to VA 46% 42% Gross wages to VA 53% 58% Nominal 3 month Euribor∗∗ 3.8% 4.0% Inflation∗∗ 2.0% 2.0% Private consumption to GDP ratio 56.3% 56.9% Public consumption to GDP ratio 19.7% 19.7% Investment to GDP ratio 21.9% 23.5% Sources : Eurostat with data averaged from 1995 to 2007. ∗ in billion e in current prices ∗∗ annualised ∗∗∗ aged from 15 to 64 in millions Campagne, Poissonnier (Insee) DSGE models’ sensitivity SEM 2015 9 / 29
- 10. A three-slide overview of the model Structural parameters Technology Technology parameter α 0.3 Smets and Wouters (2005) Depreciation rate δ 0.025 idem Capital rigidity S 6.17 idem TFP growth rate g 0.003 Eurostat Population size N 135 922 100 Eurostat Monetary policy Inflation Π∗ 1.005 Consensus, ECB Smoothing parameter ρ 0.85 Smets and Wouters (2005) Weight on inflation rπ 1.4 idem Weight on output gap ry 0.11 idem Prices and Wages Substitutability between goods θ 6 QUEST III, GEM, NAWM, EAGLE Substitutability between workers θw 4 idem Price rigidity ξ 0.90 Smets and Wouters (2005) Wage rigidity ξw 0.92 idem Price indexation γp 0.29 idem Wage indexation γw 0.90 idem Preferences Households discount factor β 0.9983 SS constraint Risk aversion σc 1.13 Smets and Wouters (2005) Inverse Frisch elasticity σl 2 idem Consumption habits hc 0.61 idem Labour habits hl 0 idem Weight on labour disutility κ 5812.38 SS constraint on hours worked Campagne, Poissonnier (Insee) DSGE models’ sensitivity SEM 2015 10 / 29
- 11. Plan 1 Motivation 2 A standard new-Keynesian DSGE model 3 Structural reforms : labour and goods markets deregulation 4 Structural reforms : a sensitivity analysis 5 Conclusion Campagne, Poissonnier (Insee) DSGE models’ sensitivity SEM 2015 11 / 29
- 12. Channels of transmission of structural reforms What should we expect ? Blanchard and Giavazzi (2003) monopolistic competition, entry costs and wage Nash bargaining production = employment Reform short-run transition long-run ↓ bargaining power of workers profits ↑, and real wages ↓ more firms enter, labour and real wages ↑ labour ↑ and real wages stagnate ↑ substitutability between goods real wages and labour ↑ profits ↓ so firms exit neutral ↓ entry costs neutral as fixed number of firms firms start entering real wages and labour ↑ DSGE with labour search and matching (Cacciatore et al., 2012) for labour market reforms, labour ↓ in the short-run worsening trade-off between short-run pains and long-term gains no endogenous growth, so decreasing mark-ups to 0 is optimal Campagne, Poissonnier (Insee) DSGE models’ sensitivity SEM 2015 12 / 29
- 13. Channels of transmission of structural reforms ...in Smets and Wouters’ model Capital demand curve θ−1 θ αζ1−α ¯K ¯L α−1 = ¯rk Capital supply curve ¯rk = 1+¯r ¯Π − 1 + δ Labour demand curve θ−1 θ (1 − α)ζ1−α ¯K ¯L α = RW Labour supply curve RW = θw θw−1 (1 + σl)¯cf(¯l) profits ¯d = 1 θ labour share RW¯L ¯Y = (1 − α)θ−1 θ capital share ¯rk ¯K ¯Y = αθ−1 θ θ ↑ Figure: Decomposition of revenues Campagne, Poissonnier (Insee) DSGE models’ sensitivity SEM 2015 13 / 29
- 14. Channels of transmission of structural reforms ...in Smets and Wouters’ model (graphically) ¯K ¯RW¯rk ¯L Capital market Labour market D S S D Campagne, Poissonnier (Insee) DSGE models’ sensitivity SEM 2015 14 / 29
- 15. Channels of transmission of structural reforms Product market deregulation : ↑ substitutability between goods dividends ↓ and factors’ remuneration ↑ (D1 to D’) increased demand for factors translates into ↑ marginal productivity and further ↑ demand (D’ to D2) positive wealth effect, consumption ↑ and labour supply ↓ (S1 to S2) ¯K ¯RW¯rk ¯L Capital market Labour market D1 D’ D2 S2 S1 S1 D1 D’ D2 At steady state capital ↑, labour ↑, production ↑ consumption ↑, real wages ↑ back Campagne, Poissonnier (Insee) DSGE models’ sensitivity SEM 2015 15 / 29
- 16. Channels of transmission of structural reforms Labour market deregulation : ↑ substitutability between workers lower marker power of workers implies an increased labour supply for a same wage (S1 to S’) increased labour supply translates into ↑ marginal productivity of capital and then labour (D1 to D2) consumption-leisure arbitrage adjusts (S’ to S2 - wealth effect) but unchanged real wages in the long-run At steady state capital ↑, labour ↑, production ↑ in the same proportions consumption ↑, real wages stagnates back ¯K ¯RW¯rk ¯L Capital market Labour market D1 D2 S1 S2 S’ S1 D1 D2 Campagne, Poissonnier (Insee) DSGE models’ sensitivity SEM 2015 16 / 29
- 17. Channels of transmission of structural reforms Quantitative simulations Figure: Production Figure: Real wage Figure: Total utility Steady state variations upon reforms in p.p. Best practices = standard exercise of convergence towards the three best EU performers (UK, SW, DK) Utility decreases (partly due to consumption habits) Campagne, Poissonnier (Insee) DSGE models’ sensitivity SEM 2015 17 / 29
- 18. Channels of transmission of structural reforms Transition dynamics Product market deregulation ↑ factors’ remuneration translates into ↑ in the real marginal cost rk ↑ and then capital, real wages ↑ slowly so labour overshoots investment ↑ as rk ↑, financed by labour and savings Labour market deregulation immediate shift in the labour supply curve and labour overshoots to finance consumption and investment as marginal productivity ↑, capital overshoots Campagne, Poissonnier (Insee) DSGE models’ sensitivity SEM 2015 18 / 29
- 19. Plan 1 Motivation 2 A standard new-Keynesian DSGE model 3 Structural reforms : labour and goods markets deregulation 4 Structural reforms : a sensitivity analysis 5 Conclusion Campagne, Poissonnier (Insee) DSGE models’ sensitivity SEM 2015 19 / 29
- 20. Long term elasticities of output and utility Elasticities when performing : a product market deregulation d ¯Y ¯Y = α 1−α + 1 (1+σl)(1−hl)(1+B) (1 + iy cy ) dθ θ θ−1 > 0 d ¯U ¯U = (1 − σc)(1 − hc) α 1−α − iy cy + 1− θw−1 θw θ−1 θ (1−α)(1−hl) (1−hc)cy (1+σl)(1−hl)(1+B) (1 + iy cy ) dθ θ θ−1 a labour market deregulation d ¯Y ¯Y = 1 (1+σl)(1−hl)(1+B) dθw θw θw−1 > 0 d ¯U ¯U = (1 − σc)(1 − hc) 1− θw−1 θw θ−1 θ (1−α)(1−hl) (1−hc)cy (1+σl)(1−hl)(1+B) dθw θw θw−1 → Importance of the households’ utility and habits calibration → Output enhancing but ambiguous sign for utility Campagne, Poissonnier (Insee) DSGE models’ sensitivity SEM 2015 20 / 29
- 21. Calibration of the utility function A strong dependency on the Frisch elasticity Figure: Goods market reforms Figure: Labour market reforms A higher σl implies a flatter labour supply curve, that is a smaller ↑ in labour when the demand curve shifts up for PMR PMR a smaller shift in the labour supply curve for LMR LMR Campagne, Poissonnier (Insee) DSGE models’ sensitivity SEM 2015 21 / 29
- 22. Calibration of the utility function Habit formation Figure: PMR - Standard Figure: PMR - hc = 0 Figure: PMR - hl = 0.5 Consumption habits Negative externality No impact on real variables (steady state and transition) Strong impact on welfare Labour habits Positive externality Scale effect on real variables (steady state and transition) Ambiguous impact on welfare Identical conclusions for LMR Campagne, Poissonnier (Insee) DSGE models’ sensitivity SEM 2015 22 / 29
- 23. Introducing non-Ricardian households Sensitivity Impact of reforms in p.p. depending on the share of non-Ricardian agents Low sensitivity for labour market reforms, agents react similarly Strong impact for product market reforms, up to 1 p.p. of production Campagne, Poissonnier (Insee) DSGE models’ sensitivity SEM 2015 23 / 29
- 24. Introducing non-Ricardian households Inequality implications of reforms For PMR, as dividends ↓ Ricardian households experience a negative wealth effect For LMR, profits are untouched. Undifferentiated behaviours. Campagne, Poissonnier (Insee) DSGE models’ sensitivity SEM 2015 24 / 29
- 25. Introducing non-Ricardian households Utility decomposition Figure: PMR - Representative Figure: PMR - Ricardian Figure: PMR - non-Ricardian For PMR, aggregate utility increases due to non-Ricardian agents Same decomposition for LMR, with a decrease in utility Campagne, Poissonnier (Insee) DSGE models’ sensitivity SEM 2015 25 / 29
- 26. Plan 1 Motivation 2 A standard new-Keynesian DSGE model 3 Structural reforms : labour and goods markets deregulation 4 Structural reforms : a sensitivity analysis 5 Conclusion Campagne, Poissonnier (Insee) DSGE models’ sensitivity SEM 2015 26 / 29
- 27. In short... Our main results The effect of structural reforms in DSGE models strongly relies on the consumption/leisure arbitrage Changes in the definition and the calibration of the utility function have strong impact on the size of gains Economic-enhancement is not always in line with welfare improvement → need for sensitivity analyses when evaluating structural reforms On the agenda Expand this analyses to a fully-fledged institutional-like model Investigate the relevance of reforms’ transmission channels in DSGE models Campagne, Poissonnier (Insee) DSGE models’ sensitivity SEM 2015 27 / 29
- 28. Plan 6 Appendix Campagne, Poissonnier (Insee) DSGE models’ sensitivity SEM 2015 28 / 29
- 29. Real rigidities and markups Monopolistic competition between goods (resp. workers) : existence of market powers Goods market differentiated goods market power : θ θ−1 demand curve : yt = pt Pt −θ Yt P * Labour market (Erceg and al., 2000) differentiated workers market power : θw θw−1 demand curve : lt = wt Wt −θw Lt N Competition level degree of substitutability between goods/workers Campagne, Poissonnier (Insee) DSGE models’ sensitivity SEM 2015 29 / 29