A Global View of Creative Destruction Chang-Tai Hsieh, University of - - PowerPoint PPT Presentation

A Global View of Creative Destruction

Chang-Tai Hsieh, University of Chicago and NBER Pete Klenow, Stanford and NBER Ishan Nath, University of Chicago

October 2020 University of Toronto Trade Workshop

1 / 42

Our questions

1

Do ideas flow across countries?

2

Does trade boost the pace of global creative destruction?

3

Are there big dynamic gains from trade?

2 / 42

What we do and find

Document 2 key facts about export turnover in U.S. manufacturing Analyze a 2-country Klette-Kortum model Calibrate the model and carry out counterfactuals Relative to autarky, current trade flows result in:

◮ 0.46 percentage points faster long run growth ◮ ∼ 37% higher consumption-equivalent welfare

3 / 42

Related recent papers

Evidence on dynamic benefits of trade Bloom, Draca and Van Reenen (2016) Aghion, Bergeaud, Lequien, and Melitz (2018) Models of trade and growth Alvarez, Buera and Lucas (2013) Sampson (2016) Akcigit, Ates and Impullitti (2018) Buera and Oberfield (2020) Perla, Tonetti and Waugh (2020)

4 / 42

Outline

1

Facts about export reallocation

2

Baseline model with learning from sellers

3

Alternative models with learning from producers

4

Dynamic gains from trade (or openness more generally)

5 / 42

Datasets

U.S. Census of Manufacturing All establishments with employees 1987, 1977, ... 2012 firm-level exports and domestic sales 300–375k establishments per Census year We calculate moments for firms World Trade Database Bilateral country exports in 4-digit SITCs 1982–2003 Focus on U.S. manufacturing exports

6 / 42

U.S. export reallocation across categories

1982–2003 five-year changes across 540 4-digit manufacturing SITCs Export reallocation rate 13.0% Category exit rate 7.8% Source: World Trade Database

7 / 42

U.S. export reallocation across firms

1987–2012 five-year arc growth rates across firms S.D. of export growth 1.72 S.D. of domestic sales growth for exporting firms 1.20 Variance ratio 2.05 Source: U.S. Census of Manufacturing

8 / 42

Overall job reallocation

Five-year reallocation rates Job creation rate 31.4% Job destruction rate 36.6% Job destruction from larger firms 30.7% Larger = above mean employment in the 1st year of a 5-year period Source: U.S. Census of Manufacturing

9 / 42

Productivity and size of exporters vs. non-exporters

Labor productivity and employment is higher at exporters than at non-exporters But there is much overlap in the exporter and non-exporter distributions of labor productivity and employment Labor productivity = revenue per worker See also Bernard, Eaton, Jensen and Kortum (2003)

10 / 42

U.S. labor productivity distribution in 2012

11 / 42

U.S. employment distribution in 2012

12 / 42

Outline

1

Facts about export reallocation

2

Baseline model with learning from sellers

3

Alternative models with learning from producers

4

Dynamic gains from trade (or openness more generally)

13 / 42

Preferences

Representative consumer in each country U = 1 ln Cj dj U ∗ = 1 ln C∗

j dj

Fixed set of varieties Each country consumes all varieties Home = U.S., Foreign = rest of OECD = *

14 / 42

Static production technology

Yj = AjLj Y ∗

j = A∗ jL∗ j

Aj and A∗

j are the best home and foreign blueprints

A′

j and A∗′ j are the second-best home and foreign blueprints

15 / 42

Markups under Bertrand competition

Home market            Exported or non-traded Imported Aj max

• A′

j, ωA∗

j

τ

• A∗

j/τ

max A∗′

j

τ , Aj ω

• Foreign market

           Imported Non-traded or exported Aj/τ max A′

j

τ , ωA∗ j

• A∗

j

max

• A∗′

j , Aj ωτ

• τ > 1 is the symmetric tariff on all traded goods

ω is the relative wage (home relative to foreign)

16 / 42

Traded and non-traded goods

Ordering products so that Aj/A∗

j is decreasing in j

j ∈ [0, x] are traded and produced at home j ∈ (x, x∗) are non-traded j ∈ [x∗, 1] are traded and produced abroad The cutoff products x and x∗ are determined by Ax τ = ωA∗

x,

Ax∗ = ωA∗

x∗

τ When τ = 1, x = x∗ and all products are traded

17 / 42

Labor market clearing

L = 1 Lj dj L∗ = 1 L∗

j dj

Lj = 0 for an imported variety, L∗

j = 0 for an exported variety

Exogenous innovation (does not use labor)

18 / 42

Balanced trade

The relative wage ω is pinned down by balanced trade: I∗ · x = I · (1 − x∗) I and I∗ denote nominal GDP at home and abroad LHS = home country exports RHS = home country imports

19 / 42

GDP and markups

I = µwL 1 − 1−τ

τ

· (1 − x∗) and I∗ = µ∗w∗L∗ 1 − 1−τ

τ

· x 1 µ ≡ x∗

1 µj dj + 1 τ ·

x

1 µf

j dj

x∗ + x/τ and 1 µ∗ ≡ 1

x 1 µ∗

j dj + 1

τ ·

1

x∗ 1 µ∗f

j dj

1 − x + (1 − x∗)/τ

20 / 42

Equilibrium consumption wages

ln W = x∗ ln Aj µj

• dj +

1

x∗ ln

• A∗

j

µ∗

j

· ω τ

• dj

ln W ∗ = x ln Aj µj · 1 ω τ

• dj +

1

x

ln

• A∗

j

µ∗

j

• dj

21 / 42

Arrival rates of quality improvements

Home Foreign Innovation by incumbents

λ λ∗

Innovation by entrants

η η∗

Pareto draws build on A of the current seller into the domestic market The average improvement in quality (over the seller) is

1 θ−1

22 / 42

Expected growth rate for symmetric countries

Autarky (λ + η) · 1 θ − 1 Free trade

• λ +

η + λ∗ + η∗ · 1 θ − 1

23 / 42

Reflecting barrier

The bottom ψ percent of qualities redraw from the top 1–ψ percent each year Maintains a stationary quality distribution Allows us to match the empirical trade elasticity In the spirit of Perla, Tonetti and Waugh (2020)’s endogenous diffusion

24 / 42

Data moments used for calibration

Export share of revenues (home) U.S. mfg 2012 10.2% Trade elasticity from halving τ Head and Mayer (2014) –5 Revenue per worker exp./non-exp. U.S. mfg 2012 1.066 Employment share of entrants U.S. mfg 2012 14.4% Employment home/foreign U.S., OECD mfg 1995–2008 0.389 Value added per worker home/foreign U.S., OECD mfg 1995–2008 1.29 TFP growth rate U.S. mfg 1995–2008 3.01% Exports in 75th/25th SITC U.S. mfg 1983–2002 20

Sources: U.S. Census of Manufacturing U.S. BLS Multifactor Productivity Database KLEMS for OECD countries World Trade Database

25 / 42

Parameter estimates

θ Shape parameter of innovation draws 10.8 λ Innovate rate, home incumbents 13.5%

• η

Innovation rate, home entrants 2.6%

• µ∗

Innovation rate, foreign incumbents + entrants 12.3% τ Gross tariff rate 1.50 ψ Reflecting barrier for product quality 1.0%

26 / 42

Discrete number of varieties

We simulate an economy with a finite number of varieties The number of varieties matters for category export reallocation We start the smallest category with 1 variety Then the number of varieties rises at the exponential rate ǫ with the ranking of the category ǫ = 0.915% makes U.S. 75th/25th exports = 20, as in the World Trade Database Result: 41,911 varieties in total (average of ∼ 78 varieties across the 540 categories)

27 / 42

Baseline model growth vs. tariffs

28 / 42

Baseline model relative wage vs. tariffs

29 / 42

Baseline model trade elasticity vs. tariffs

30 / 42

Baseline model quality dispersion

31 / 42

What if knowledge diffusion is independent of trade?

Suppose U.S. draws with probability z∗ on its own best producers, 1 − z∗ on the best OECD products And the OECD draws with probability 1 − z on its own best producers, z on the best U.S. products Such “disembodied” spillovers are isomorphic to baseline model at z = x and z∗ = x∗ But dynamic gains from lower tariffs will differ if {z, z∗} are fixed while {x, x∗} move ...

32 / 42

Disembodied spillovers and growth vs. tariffs

33 / 42

Outline

1

Facts about export reallocation

2

Baseline model with learning from sellers

3

Alternative models with learning from producers

4

Dynamic gains from trade (or openness more generally)

34 / 42

Alternative model assumptions

Learning from domestic producers

◮ When innovating on an imported variety:

fraction κ of draws on sellers fraction 1 − κ on (dormant) domestic producers

Research specialization

◮ Fraction ν of draws on all products ◮ Fraction 1 − ν of draws on products a country produces

35 / 42

Alternative models and targeted moments

κ = 1 κ = 0.1 κ = 0.1 Data ν = 1 ν = 1 ν = 0.1 TFP Growth 3.0% 3.0% 3.0% 3.0% U.S.-OECD Relative Wage Gap 29.0% 29.0% 29.0% 29.0% U.S. Export Share of Revenues 10.2% 10.2% 10.2% 10.2% Employment Share of Entrants 14.4% 14.4% 14.4% 14.4% Trade Elasticity 5.0 5.0 1.4 0.0 log(Revenue/Worker) Exporters vs. Non-Exporters 6.6% 6.6% 1.0%

• 0.6%

36 / 42

Alternative model implications for export reallocation

κ = 1 κ = 0.1 κ = 0.1 Data ν = 1 ν = 1 ν = 0.1 Country-level export reallocation Export reallocation rate 13.0% 12.8% 5.4% 0.6% Category exit rate 7.8% 2.9% 0.6% 0.0% Firm-level export volatility SD of export growth 1.72 1.78 1.54 1.66 S.D. of domestic sales growth 1.20 1.08 1.23 1.66 Variance Ratio 2.05 2.72 1.57 1.00

37 / 42

Outline

1

Facts about export reallocation

2

Baseline model with learning from sellers

3

Alternative models with learning from producers

4

Dynamic gains from trade (or openness more generally)

38 / 42

Welfare gains from cutting tariffs in half (in our baseline model)

U.S. OECD Static gains according to the ACR formula 3.0% 1.2% Static gains in our model 4.2% 5.0% Change in growth in our model 0.17% 0.17% Dynamic* gains in our model 9.6% 15.5% * PDV of consumption with τ equal to 1.25 relative to 1.50 (and discount rate of 5%)

39 / 42

Welfare gains vis a vis autarky (in our baseline model)

U.S. OECD Static gains according to the ACR formula 1.1% 1.0% Static gains in our model 13.3% 15.5% Change in growth in our model 0.46% 0.46% Dynamic* gains in our baseine model 37.0% 104% * PDV of consumption with τ equal to 1.5 relative to 4 (discount rate of 5%)

40 / 42

Conclusion

Documented 2 key facts about export reallocation in U.S. manufacturing

◮ ∼ 13% reallocation rate across 4-digit categories ◮ firm-level export growth is twice as dispersed as domestic sales growth

Analyzed a 2-country model of creative destruction and growth In our baseline model, current trade (relative to autarky):

◮ raises growth in TFP by ∼ 0.46% per year ◮ raises higher consumption-equivalent welfare by ∼ 37%

41 / 42

Potential follow-up research

Optimal R&D subsidies

◮ Need a Global Technical Change Accord?

China’s Coupling and Decoupling from the World Innovation Engine Frictions to job reallocation in response to creative destruction

◮ Implications for consumption inequality and the distribution of welfare gains

42 / 42