Resource Reallocation and Aggregate Productivity: Firm dynamics in Korean Manufacturing

Transcription

1 Resource Reallocation and Aggregate Productivity: Firm dynamics in Korean Manufacturing Minho Kim Jan KDI-Brookings Workshop

2 1. Motivation 2. Data 3. Methodology: Decomposing aggregate productivity growth 4. Empirical results 5. Additional materials on misallocation 2

3 Motivation Sources of aggregate productivity growth 1. Allocative efficiency gains associated with shifting labor and capital out of small, less-productive firms into large, moreproductive firms 2. Technical efficiency gains associated with innovation, better management, etc Estimate the aggregate productivity growth in the Korean manufacturing industry in recent two decades Investigate the role of each source in the aggregate productivity growth 3

4 Data Plant-level data from 1995 to 2013 in the Korean Mining and Manufacturing Survey censored census establishments with at least 10 employees in manufacturing industries unbalanced panel of 52,496 plants (1995) => 65,272 (2013) In 2012, compared to the manufacturing industry, the data represent Number of plants Employees Gross output 17% 72% 87% Source: 1) Gross output: Input-Output table from the Bank of Korea 2) Number of plants and employees: Bank of Korea 4

5 Data (Thousands) 1991~2014 Manufacturing Employment and value added growth (Trillion Won) Employment (left) Value added (right) Source: Statistics Korea Korean Mining and Manufacturing Survey, 5

6 Methodology Estimate manufacturing plant s contribution to the change in aggregate final demand Use Petrin and Levinsohn (2012) s definition of APG (aggregate productivity growth) and quantify the contribution of resource reallocation and technical efficiency to the APG. Decompose at any level of aggregation (age, size, industry) Petrin, White, Reiter (2011) find that APG in U.S. manufacturing between 1976 and 1996 was 2.2% on average and the contribution of resource reallocation was larger than the contribution of technical efficiency growth. Kwon, Narita, Narita (2015) find that the contribution of resource reallocation declined during 1990s and was negative when a financial crisis occurred in the late 1990s. 6

7 Methodology: Decomposing aggregate productivity growth Aggregate productivity growth (APG) : the change in final demand minus the change in the aggregate expenditures on labor and capital. APG i P i dy i i f W if dx if where Y i : final demand, X if : inputs, W if : input costs Using the national accounting identity, i P i Y i = i VA i, APG = i dva i i f W if dx if 7

8 Methodology: Decomposing aggregate productivity growth APG = (1) productivity gains from technical efficiency (TE) + (2) resource reallocation across plants (RE) + (3) net entry of plants (NE) APG t = TE t + RE t + NE t TE t = i Ct D it ln A it where A it : TFP D it =gross output/aggregate value added: Domar weight x it: = x i,t 1 + x it 2 TE is the sum of weighted plant-level changes in TFP using the ratio of plant-level revenue to aggregate final demand as the weight. 8

9 Methodology: Decomposing aggregate productivity growth APG t = TE t + RE t + NE t RE t = where RE ft f = f i Ct D it (ε if sift) ln X ift D it =gross output/aggregate value added: Domar weight ε if = Y i Y i : elasticity of output to input, X X if X if if f labor L, capital K, materials M s ift W ift X ift /Y it : ratio of costs to gross output RE is the weighted sum of the change in input. The weight ε if sift is the gap between the marginal product and the unit cost of input. RE increases when inputs are reallocated to plants with larger gap from plants with smaller gap. 9

10 Methodology: Decomposing aggregate productivity growth APG t = TE t + RE t + NE t NE t = iεεt D it 1 f s ift - i χt 1 D i,t 1 1 f s if,t 1 where D it =gross output/aggregate value added: Domar weight s ift W ift X ift /Y it : ratio of costs to gross output NE is output (net of input use) added by the net entry of plants 10

11 Productivity Estimation Estimation of Productivity ln Y it = ln A it + α K ln K it + α L ln L it + α M ln M it Estimate α K, α L and α M, the elasticity parameters of capital, labor and materials, for each 3 digit industries (82 industries) - apply Wooldridge, Levinsohn and Petrin (2009) method to control simultaneity issue and selection bias - Estimated parameters are the elasticity of output to input,. ε if = α f 11

12 Results 1995~2013 Manufacturing Aggregate productivity growth decomposition TE RE NE APG 12

13 Results 1995~2013 Manufacturing Aggregate productivity growth decomposition APG TE RE NE

14 Methodology: Decomposing aggregate productivity growth Baily, Hulten, Campbell (1992) is commonly used to compute aggregate productivity growth. BHC t = i D it ln A it - i D it 1 ln A it 1 BHC t = TE t + BHC_RE t + BHC_NE t TE t = i Ct D it ln A it BHC_RE t = BHC_NE t = i Ct ln A it D it iεεt D it ln A it - i χt 1 D it 1 ln A it 1 BHC reallocation is the weighted change in shares with the weight of plant-level TFP. 14

15 Results: APG VS BHC Aggregate productivity growth APG VS BHC APG Final Demand Growth BHC_APG 15

16 Results: APG VS BHC Reallocation effects APG VS BHC RE BHC_RE 16

17 Results: APG VS BHC Net entry effects APG VS BHC NE BHC_NE 17

18 Results: Reallocation during crisis APG decomposition during Asian Financial Crisis (1997~1998) PL_NE PL_RE TE_WLP PL 18

19 Results: Reallocation during crisis APG decomposition during Global Financial Crisis (2008~2009) PL_NE PL_RE TE_WLP PL 19

20 Results: Role of young plants Aggregate productivity growth Young (up to 5 years old) VS Old APG APG_Y5 APG_O5 20

21 Results: Role of small plants Aggregate productivity growth Small (up to 300 employees) VS Big APG APG_S300 APG_B300 21

22 Results: Productivity by size Productivity by plant s size (number of employee) Labor productivity Total factor productivity 22

23 Results: Productivity by size Total factor productivity Productivity by plant s size Domar weighted sum 23

24 Results: Productivity by size Proportion of number of plants by plant s size 24

25 Results: Productivity by size Proportion of number of employees by plant s size 25

26 Results: Productivity by size Proportion of value added by plant s size 26

27 Results: Productivity by age Productivity by plant s age Labor productivity Total factor productivity 27

28 Results: Productivity by age Total factor productivity Productivity by plant s age Domar weighted sum 28

29 Results: Productivity by age Proportion of number of plants by plant s age 29

30 Results: Productivity by age Proportion of number of employees by plant s age 30

31 Results: Productivity by age Proportion of value added by plant s age 31

32 Additional materials Misallocation and Manufacturing TFP in Korea by Kim, Oh, Shin (2017) Apply Hsieh and Klenow (2009) to assess degree of resource misallocation in the Korean manufacturing between 1982 and 2007 We find an improvement in allocative efficiency during the first decade and a reversal after

33 Additional materials Misallocation and Manufacturing TFP in Korea Firm faces firm specific wedges that affect marginal product of inputs max Lsi,K si = (1 τ Ysi )P si Y si ωl si (1 + τ Ksi )RK si Revenue productivity (TFPR) is proportional to the geometric average of marginal products of capital and labor. TFPR si P si A si σ 1 σ R α s α ω 1 α s 1 α s 1 + τksi α s 1 τ Ysi TFP s = Y s K s α s L s 1 α s M s i=1 A si TFPR s TFPR si σ 1 1 σ 1 33

34 Additional materials Misallocation and Manufacturing TFP in Korea Industry-level TFP is TFP s = Y s K s α s L s 1 α s M s i=1 A si TFPR s TFPR si σ 1 1 σ 1 Without distortions, M s TFP s = A s A si i=1 TFPR s = TFPR si i σ 1 1 σ 1 Ratio between the final goods produced with and without distortions Y Y eff = S s=1 TFP s A s θ s = S s=1 M s i=1 A si A s TFPR s TFRP si σ 1 θ s σ 1 34

35 Additional materials 35

36 Additional materials 36

37 Additional materials 37