The Aggregate Implications of Innovative Investment. by A. Atkeson and A. Burstein. Discussion by erm, July 2017

Transcription

1 The Aggregate Implications of Innovative Investment by A. Atkeson and A. Burstein Discussion by erm, July 2017

2 If policy changed to induce more innovation, What happens to Output Productivity Welfare?

3 Develop model nesting many others Use discipline of national accounts

4 Welfare gains are potentially huge But, results very sensitive to Model choice Parameter estimates

5 AB s main result doesn t rely on cross-section data Aggregate implications of innovative investment: Can estimate medium-run growth Can t yet estimate long-run growth or welfare

6 Firm output: y t = K ρ It (kα Tt kγ It l1 α γ t ) Capital accumulation k Tt+1 =(1 δ T )k Tt + x Tt k It+1 =(1 δ I )k It + A rt K φ 1 It x It }{{} y rt

7 Given α, γ, define: log Z t =logy t α log K Tt (1 α γ)logl t Then growth rates are: g Zt =(ρ + γ)g KIt g KIt =log(1 δ I + Y rt /K It ) And AB s elasticity is: g Zt ḡ Z ϵ zr = log Y rt log Ȳr ( ) exp ḡki 1+δ I (ρ + γ) exp ḡ KI

8 ( exp ḡki 1+δ I ϵ zr (ρ + γ) exp ḡ KI ) ρ + γ =1/3(basedonvarietymodels) ḡ Z ḡ KI =.0123(basedonBEA/BLSdata) =.0369(fromdefinition) δ I? 0.0, ϵ zr = , ϵ zr = , ϵ zr =1/3

9 ( exp ḡki 1+δ I ϵ zr (ρ + γ) exp ḡ KI ) ρ + γ =1/3(basedonvarietymodels) ḡ Z ḡ KI =.0123(basedonBEA/BLSdata) =.0369(fromdefinition) δ I? Not known use investment/value ratio (X I /V I )instead

10 ( exp ḡki 1+δ I ϵ zr (ρ + γ) exp ḡ KI ) ρ + γ =1/3(basedonvarietymodels) ḡ Z ḡ KI =.0123(basedonBEA/BLSdata) =.0369(fromdefinition) δ I? Not known use investment/value ratio (X I /V I =.08) X I :NIPA+imputationforentrants V I : dividends/(interest rate growth rate)

11 ( exp ḡki 1+δ I ϵ zr (ρ + γ) exp ḡ KI ) (ρ + γ) X I V I ρ + γ =1/3(basedonvarietymodels) ḡ Z ḡ KI =.0123(basedonBEA/BLSdata) =.0369(fromdefinition) δ I? Not known use investment/value ratio (X I /V I =.08)

12 ( exp ḡki 1+δ I ϵ zr (ρ + γ) exp ḡ KI ) (ρ + γ) X I V I =.027 ρ + γ =1/3(basedonvarietymodels) ḡ Z ḡ KI =.0123(basedonBEA/BLSdata) =.0369(fromdefinition) δ I? Not known use investment/value ratio (X I /V I =.08)

13 ϵ zr =2.7% Used no information from LBD or GHK Relied on only one dubious parameter (ρ)

14 ϵ zr =2.7% Used no information from LBD or GHK Relied on only one dubious parameter (ρ) What does this imply for future growth and welfare?

15 10% Increase in Innovation Investment

16 φ φ 10% Increase in Innovation Investment

17 φ φ 10% Increase in Innovation Investment

18 To rationalize both Observed growth in Y, L and Large negative spillovers Requires high growth in A r

19 g Ar Recall y t = K ρ It (kα Tt kγ It l1 α γ t ) y rt = A rt K φ 1 It x It Growth in output (with φ =1): ḡ Y }{{}.025 = ρ + γ g Ar (1 α)(2 φ) (ρ + γ) }{{}}{{} (1 φ)(1 α γ) g (1 α)(2 φ) (ρ + γ) }{{} L }{{}.008 0

20 g Ar Recall y t = K ρ It (kα Tt kγ It l1 α γ t ) y rt = A rt K φ 1 It x It Growth in output (with φ = 1.6): ḡ Y }{{}.025 = ρ + γ g Ar (1 α)(2 φ) (ρ + γ) }{{}}{{} (1 φ)(1 α γ) g (1 α)(2 φ) (ρ + γ) }{{} L }{{}

21 g Ar Recall y t = K ρ It (kα Tt kγ It l1 α γ t ) y rt = A rt K φ 1 It x It Growth in output (with φ = 1.6): ḡ Y }{{}.025 = ρ + γ g Ar (1 α)(2 φ) (ρ + γ) }{{}}{{} (1 φ)(1 α γ) g (1 α)(2 φ) (ρ + γ) }{{} L }{{} Choice of φ also matters for welfare...

22 Depend importantly on Spillovers Appropriation... which are both hard to estimate

23 Proportion κ stolen k It+1 =(1 δ I )k It +(1+κ)q rt x It κq rt k It X It K It New impact elasticity: ϵ zr = ϵ zr/(1 + κ) How to interpret κ? Weak IP protection? Quid pro quo policy? How to measure κ?

24 No appropriation: 1% rise in X I /Y 2.3 to 15% rise in welfare depending on φ With appropriation: 1% rise in X I /Y possibly negative welfare! Optimal policy may involve barriers to new firms, products

25 Need estimates for some key parameters AB on right track Match theory to NIPA Should look at cross-country evidence Don t see that they need LBD or GHK