Misallocation, Establishment Size, and Productivity

Transcription

1 Misallocation, Establishment Size, and Productivity Pedro Bento West Virginia University Diego Restuccia University of Toronto November 15, / 23

2 Motivation Large Income Differences Across Countries consensus: mostly explained by productivity differences evidence of resource misallocation across establishments in poor countries (Hsieh & Klenow 2009) explains significant portion (but not most) of variation in aggregate productivity establishment-level productivity also varies across countries (Hsieh & Klenow 2009, Pagés-Serra 2010, Gal 2013) not explored much in misallocation literature 2 / 23

3 What We Do in our model, establishments take misallocation into account when investing in productivity Bhattacharya, Guner, & Ventura (2013) and Hsieh & Klenow (2014) include similar mechanism, combined with (several) other extensions simplify, isolate effect of misallocation on investment and entry separated from static effect through distorted output decisions 3 / 23

4 What We Find if distortions are random; no effect of misallocation on investment or establishment size model collapses to Hsieh and Klenow (2009) if distortions are correlated with productivity; less investment in productivity larger impact of misallocation on aggregate productivity smaller establishments 4 / 23

5 Evidence: Establishment Size Are establishments smaller in poor countries? previous evidence inconclusive no standardized data for large number of countries 5 / 23

6 Evidence: Establishment Size Are establishments smaller in poor countries? previous evidence inconclusive no standardized data for large number of countries we construct new dataset hundreds of sources: census, business registries,... standardized data for 134 economies persons engaged per establishment representative of all manufacturing establishments 5 / 23

7 Evidence: Establishment Size (134 economies) ASM Establishment Size (log scale) MYS SGP ARE LUX LIE TTO DEU PRI UKR MAC MCO QAT AUT BGR MNP GBR CANUSA DNK JPN KWT BRAVEN LVA IRL PHL GEO MKD ROULTU RUS TWN NLD FRO CHE VIR BEL ABW GUM FRA ARG KAZ KGZMDA POL MEX HUN PYF ISR HRV EST SVN NZL SWE TON HKG ESP MDG GHA PERZAF THAMUS BHRKOR NPL LKA BTN SLVCOL URY CPV JOR TUR AUS BRN SMR STP MLT SAU PRT CZE ITA FIN BGDSDN KHM BOL BIH SRB PAN ETH CMR VNM HND PRYMNG DZA TUN UGA RWA ALB LBY ECU SVK NOR CYP ALA MAR AND BMU GRL LAO IRN PSE IDN MNE NIC GRC SYR UVK REUNCL IND YEM GUF PLW SLE MTQ MDV MWI GLP BEN GDP per Capita (log scale) elasticity: 0.27 (0.04) 6 / 23

8 Evidence: Establishment Size (107 large economies) Establishment Size (log scale) ETH ARE TTO DEU PRI UKR AUT CANUSA BGR GBR DNK JPNLD KWT BRAVEN LVA TWN RUS CHE IRL PHL GEO MKD ROULTU FRA BEL ARG KAZ KGZMDA POL ISR MEX HUN HRV EST SVN NZL SWE HKG ESP GHA PERZAF THAMUS BHRKOR MDGNPL LKA BTN SLVCOL URY SAU PRT JOR TUR CZE ITAAUS FIN BGDSDN KHM BOL BIH SRB PAN CMR VNM DZA TUN RWA HND PRYMNG ALB LBY UGA ECU SVK NOR CYP MAR LAO IRN PSE IDN MNE NIC GRC SYR UVK REU IND YEM SLE MWI BEN MYS SGP GDP per Capita (log scale) QAT elasticity: 0.33 (0.04), population > 0.5 million 7 / 23

9 Model: Environment standard model of monopolistic competition, but; endogenous entry entrants invest to determine productivity abstract from heterogeneity, establishments identical ex ante 8 / 23

10 Model: Environment ( N Final-Good Firm: Y = 0 y σ 1 σ i di ) σ σ 1 N: number of intermediate-good firms y i : demand for input i σ: the constant elasticity of substitution between varieties. Intermediate Firm: y i = s i l i s i : productivity, l i : labor demanded upon entry, choose s i by investing c S Ys θ i owner forgoes market wage w while running firm exogenous probability of firm death λ 9 / 23

11 Model: Environment each firm i faces tax τ i on output assume τ i depends on productivity s i ( sī ) γ (1 τ i ) = s s: average productivity γ: elasticity of distortion w.r.t. productivity 10 / 23

12 Model: Equilibrium Steady-State Decentralized EQ: prices and allocations constant given prices P i, final-good firm maximizes profits P i = Y 1 1 σ y σ i given w, R, Y, l i maximizes per-period profits l i = (1 τ i) σ s σ 1 i w σ ( σ 1 given w, R, Y, s i maximizes life-time profits σ ) σ Y, π i = wl i σ 1 free-entry: investment = life-time profits forgone wages labor-market clears: 1 = N (E [l i ] + 1) 11 / 23

13 Model: Equilibrium labor-market clearing + optimal l i ; Y = N 1 E [ s σ 1 (1 τ) σ 1] σ σ 1 σ 1 (1 N) E [s σ 1 (1 τ) σ ] or [ ( ) σ 1 ] 1 Y = N 1 σ 1 σ 1 MRPL (1 N) E s σ 1 MRPL i ( ) MRPL i = P i y i l i = σ w σ 1 (1 τ i ) if no investment, same as Hsieh & Klenow (2009) if distortions random, same as Hsieh & Klenow (2009) 12 / 23

14 Model: Equilibrium now use (1 τ i ) = ( ) s ī γ s Y (σ 1)σ 1 s σ(1 γ) 1 (1 ρ)w σ 1 σ σ i s σγ ρ(λ, R): discount factor value of entry = w 1 ρ c SYs θ i optimal investment: c S Ys θ = E[π] [σ(1 γ) 1] (1 ρ) θ free-entry: E[π][θ+1 σ(1 γ)] θ = w 13 / 23

15 Model: Results N = [θ + 1 σ(1 γ)] θσ + 1 σ(1 γ) ( ) 1 ( σ(1 γ) 1 θ 1 s = N θ(1 ρ)σc S ) 1 θ aggregate investment share : λ[σ(1 γ) 1] σθ(1 ρ) 14 / 23

16 Model: Results correlated distortions (γ) discourage productivity investment entrants invest lower fraction of profits on productivity this increases value of entry, so free entry implies number of establishments must increase to lower value of entry to zero 15 / 23

17 Model: Results When distortions more correlated with productivity (higher γ): higher number of establishments lower employment per establishment lower establishment-level productivity aggregate productivity increasing in both establishment-level productivity and number of establishments could be higher or lower 16 / 23

18 Calibration quantify impact of correlated distortions on average employment and productivity calibrate model economy to U.S. manufacturing benchmark: γ US = 0.13 from Hsieh & Klenow (2014) 17 / 23

19 Quantitative Exercise Table: Model Results across Hypothetical Correlated Distortions γ γ Size Productivity Investment Output 0.13 (γ US ) % % % % % (γ India ) % 0.27 (0.17) 18 / 23

20 Evidence: Correlated Distortions Do poor countries have higher γ s? Hsieh & Klenow (2014): India, Mexico higher γ s than U.S. World Bank s Enterprise Surveys: establishment-level data for low- and middle-income countries Hsieh & Klenow (2009) method to back out within-industry distributions of distortions and productivity use regressions to estimate γ s for 62 countries result: γ higher in poorer countries 19 / 23

21 Evidence: Correlated Distortions and GDP per Capita GDP per Capita (log scale) USA ESP IRL SVN SVK CZE EST HUN HRVRUS POL LTU TTO KAZ PAN LVA MEX ROU TUR URYBGR ARG SRB MUS ZAF THA BRAUKR MKD COL ALB BIH PER ECU SLV DZA UVK GEO MNG PRY IDN MDAMAR NIC LKA BOL JOR PHL KGZ VNMHND IND PSE YEM GHA LAO BGD BEN UGANPL MDG MWI ETH Productivity Elasticity of Distortions elasticity: (0.86) 20 / 23

22 Evidence: Correlated Distortions and Average Employment Establishment Size (log scale) USA ESP TTO UKR BGR BRA LVA LTU IRL ROU RUS GEO MKD PHL POL HUN KAZ MEX KGZMDA ARG SVN HRV EST ZAF THA COLPER GHA MUS URY SLV MDGLKA NPL JOR TUR CZE BGD SRB PAN VNM BIH SVKHNDETHALB PRY BOL DZA UGA ECU MNG MAR LAO NIC PSE IDN UVK IND YEM MWI BEN Productivity Elasticity of Distortions elasticity: (0.50) 21 / 23

23 Evidence: Correlated Distortions and R&D Intensity R&D Intensity (%, log scale) USA ESP EST IRL SVN CZE BRA RUS HUN ZAF LTU UKR HRV IND POL SVK MAR TUR ARG BGR LVA MDA ROU MEX URYUGAJOR MUS SRB ECU MNG THA GHA MDG COL ETHLKA KGZKAZ GEO MKD VNM PAN PER BOL ALBPHL TTO IDN DZA PRY NPL HND LAO NIC BIH Productivity Elasticity of Distortions elasticity: (1.04) 22 / 23

24 Conclusion systematic evidence that poor countries have: smaller establishments less investment in productivity more strongly correlated idiosyncratic distortions if establishments take misallocation into account when investing in productivity; model can account for above facts large impact on aggregate productivity combined with Hsieh & Klenow (2009), misallocation can explain 6-fold difference in size, establishment-level productivity, and aggregate TFP between U.S. and India 23 / 23