How do R&D expenditures influence national total factor productivity and technical efficiency?

Transcription

1 CENTER FOR MACROECONOMIC ANALYSIS AND SHORT-TERM FORECASTING Phone: , How do R&D expenditures influence national total factor productivity and technical efficiency? Alexander Apokin, Irina Ipatova (CMASF)

2 1. Motivation Cross-country productivity studies widely use the growth accounting framework (Barro, 1999) assumes full utilization of production factors total factor productivity (TFP) (the Solow residual ) is assumed to represent technical progress (Acemoglu, 2008) We decompose the TFP dynamics with cross-country comparison on a macrolevel Our TFP decomposition separately includes: the dynamics of the global technological frontier the dynamics of the national technological frontiers the dynamics of technical efficiency of each economy We examine whether R&D expenditures significantly influence TFP growth in the framework based on Kumar and Russell (2002) 2

3 1. Motivation What we did in this paper: estimated the impact of R&D expenditures on technical efficiency (TE) and total factor productivity (TFP) on the span of 5 to 10 years How we did this: estimated the structure and dynamics of TFP in a three-factor production function model framework using SFA and DEA methods used both 22-country OECD-1990+Russia sample and a larger 65-country sample for modeled TE and TFP as dependence on R&D expenditures and control variables We found that: there was a large and significant influence of 5- and 10-year lagged R&D expenditures both on TFP and technical efficiency 3

4 The Outline 1. Motivation 2. Literature 3. Data and Methods 4. TFP and Technical Efficiency Estimates 5. R&D Expenditures Impact 6. Conclusions 4

5 2. Literature Method development: SFA: Farrell (1957); Meeusen, Van den Broeck (1997); Aigner et al. (1977); DEA: Charnes et al. (1978) TFP decomposition: Fare and Primont (1994); Kumar and Russell (2002); O Donnell (2008, 2012) TFP of a growth arithmetic approach: how to extract: Solow (1957); Mankiw et al. (1992); Klenow and Rodríguez-Clare (1997); Timmer et al. (2010); Fernald (2014) the drivers: King and Levine (1993); Barro (1996); Coe and Helpman (1995); Djankov et al. (2003) within-sector productivity studies: Harberger (1998), Jorgenson et al. (2000), Timmer et al. (2007), De Vries et al. (2012) Estimation of technical inefficiency on a macro-level: Kumar and Russell (2002) Hsieh and Klenow (2009) marginal products Bartelsman et al. (2013), Jones (2013) an input-output multiplier R&D spillover externalities: Coe and Helpman (1995); Guellec and De La Potterie (2002) 5

6 The Outline 1. Motivation 2. Literature 3. Data and Methods 4. TFP and Technical Efficiency Estimates 5. R&D Expenditures Impact 6. Conclusions 6

7 3. Data and Methods Two samples: OECD-1990+Russia countries that were in OECD in 1990 and probably form global technological frontier + Russia (22 countries) 65 countries including a lot of emerging economies (each of them had more than 0.1% World GDP in 2011) Endogenous variable is GDP (PWT) Exogenous variables are: physical capital stock, K (AMECO, WIOD and PWT) labour, L (PWT) human capital stock, H (PWT) GDP and K are in constant PPP 7

8 3. Data and Methods Total factor productivity change: N M qit, xit output and input vectors of firm i in period t it it, it it aggregate output and input (nonnegative, nondecreasing and linearly homogeneous) functions Qit Q QI hs hs, it TFPIhs, it X X XI Q where Q( q ) X X( x ) TFPI hs, it hs, it XI hs it TFP index QI, it hs hs, it, aggregate output and input quantity indices 8

9 3. Data and Methods TFP change decomposition (O Donnell, 2008): TFP TFP TFPE TFP OTE OSME where TFP it * TFP t TFPE * * it t it t it it TFP of firm i in period t maximum TFP using technology of period t TFP efficiency OTE it output technical efficiency it output scale and mix efficiency OSME it TFPI TFP TFPE TFP OTE OSME * * t it t it it hs, it * * TFPs TFPEhs TFPs OTEhs OSMEhs 9

10 3. Data and Methods DEA is optimization problem of mathematical programming which estimates deterministic production frontier SFA is econometric method which estimates stochastic production frontier lngdp ln K ln L ln H v u it 0 1 it 2 it 3 it it it 2 2 it v it u v N(0, ), u N (0, ), i 1,...,65, t 1990,...,2011 Battese, Coelli (1988): Pooled regression, sample TE i u i e i ˆ E i i 10

11 The Outline 1. Motivation 2. Literature 3. Data and Methods 4. TFP and Technical Efficiency Estimates 5. R&D Expenditures Impact 6. Conclusions 11

12 4. TFP and Technical Efficiency Estimates DEA (Färe-Primont aggregate function) SFA: Cobb-Douglas vs. translog specification inefficiency error term distribution: half-normal exponential truncated normal pooled regressions: homoscedasticity heteroscedasticity panel regressions: Time-Invariant, TI Time-Varying Decay, TVD Two-Step Models, FE-2S/RE-2S (Heshmati et al., 1995) «True» Random Effect Model, TRE, «True» Fixed Effect Model, TFE (Greene, 2005) 12

13 4. TFP and Technical Efficiency Estimates Estimation results for panel models (log GDP) Explanatory variables TI model, TI model, OECD-1990+Russia FE-2S model, TFE model, 0.54*** 0.50*** 0.52*** 0.56*** ln(capital) (0.02) (0.04) (0.02) (0.02) ln(number of 0.36*** 0.51*** 0.43*** 0.46*** employed) (0.02) (0.05) (0.03) (0.03) 0.83*** 0.50*** 0.80*** 0.42*** ln(human capital) (0.08) (0.11) (0.09) (0.08) 4.24*** 4.80*** 3.70*** - Constant (0.14) (0.37) (0.16) *** 0.21*** 0.10*** 0.17*** (0.05) (0.04) (0.01) (0.01) 0.11*** 0.07*** 0.08*** 0.02*** (0.002) (0.002) (0.003) (0.01) Number of observations Number of countries logl

14 4. TFP and Technical Efficiency Estimates Estimation results for pooled models (log GDP) ln(capital) Explanatory variable ln(number of employed) ln(human capital) Constant 2 ln u Total R&D expenditures per researcher, 5 years lag Business R&D expenditures per researcher, 5 years lag Merchandise trade, % GDP Constant 2 ln Model 1, Model 2, Model 3, OECD-1990+Russia 0.73*** 0.29*** 0.33*** (0.03) (0.05) (0.05) 0.20*** 0.64*** 0.66*** (0.02) (0.06) (0.05) 0.16* 1.22*** 0.96*** (0.09) (0.15) (0.11) 2.16*** 6.34*** 6.16*** (0.28) (0.64) (0.63) *** - - (3.01) *** *** - (6.11) (7.52) 0.01*** 0.04*** -0.16*** (0.01) (0.01) (0.04) -1.43*** -2.71*** - (0.43) (0.54) - Index of economic freedom Telephone lines ** *** (0.02) - (0.04) * - Adjusted roads density - (0.0004) *** 14.97*** 12.36*** (2.60) (3.47) (2.76) v -3.37*** -4.06*** -4.66*** Constant (0.09) (0.12) (0.13) Number of observations logl

15 4. TFP and Technical Efficiency Estimates We calculate TFP and TE estimates and test the economic adequacy of our model TFP estimates by comparing them with: each other estimates from other issuers: OECD Conference Board PWT All TFP estimates are highly correlated even though PF coefficients vary The results, with some exceptions, indicate very strong correlation between our estimates and those made by OECD, Conference Board and PWT 15

16 4. TFP and Technical Efficiency Estimates TFP estimates, Model 2 Poland Hungary Czech Republic Slovak Republic 16

17 The Outline 1. Motivation 2. Literature 3. Data and Methods 4. TFP and Technical Efficiency Estimates 5. R&D Expenditures Impact 6. Conclusions 17

18 5. R&D Expenditures Impact We estimate over 500 regressions for TFP and TE through a cyclic selection procedure using: control variables from three groups (WB\WDI data): structural infrastructural institutional two relative measures of R&D expenditure: intensity (% GDP) density (per researcher) total expenditures business expenditures 18

19 5. R&D Expenditures Impact TFP growth models and R&D density, % GDP Exogenous variables Model 1, Model 2, Model 3, OECD Russia Model TI, Model TI, OECD Russia Model FE-2S, Model TFE, DEA, DEA, OECD Russia Total R&D expenditures per 1 researcher, 5 years lag, $M Merchandise trade, % of GDP Index of economic freedom Telephone lines TFP growth rates, 1 year lag TFP level, 1 year lag Constant 14.10*** 13.73** 25.97** 14.16** 25.04** 13.86** 12.58** 10.55* 24.01* (5.38) (5.93) (13.09) (5.86) (12.23) (5.81) (5.49) (6.08) (12.69) 0.07*** 0.06*** 0.14*** 0.06*** 0.14*** 0.06*** 0.07*** 0.08*** 0.16*** (0.02) (0.02) (0.03) (0.02) (0.03) (0.02) (0.02) (0.02) (0.03) *** *** *** - - (0.73) - (0.72) (0.77) 0.10*** 0.08** ** ** 0.07** 0.07* - (0.03) (0.04) - (0.04) - (0.04) (0.03) (0.04) *** 0.14*** *** *** 0.15*** 0.18*** - (0.05) (0.05) - (0.05) - (0.05) (0.05) (0.05) *** *** *** *** -0.02*** *** -0.02*** *** *** (0.12) ( ) (0.0002) (0.001) (0.002) (0.001) (0.003) (6.15) (6.38) 9.23*** ** ** ** 5.40* *** (2.75) (2.28) (6.25) (2.39) (5.80) (2.42) (2.67) (2.98) (5.95) Number of observations R-squared Number of countries

20 5. R&D Expenditures Impact TFP growth models and R&D intensity, per researcher Exogenous variables Model 1, Model 2, Model 3, OECD Russia Model TI, Model TI, OECD Russia Model FE- 2S, Model TFE, DEA, DEA, OECD Russia Total R&D expenditures, 10 years lag, % GDP Manufacturing value added, % of GDP Index of economic freedom TFP level, 1 year lag Constant 3.06* 6.20*** 7.69*** 6.04*** 6.98*** 5.94*** 4.70*** 4.30** 8.00*** (1.82) (2.00) (1.96) (1.96) (1.78) (1.94) (1.80) (2.00) (2.01) 1.06*** 1.16*** 1.52*** 1.16*** 1.59*** 1.14*** 1.03*** 1.10*** 1.61*** (0.16) (0.18) (0.22) (0.18) (0.20) (0.17) (0.16) (0.18) (0.23) 3.89*** 3.34** 4.67*** 3.53** 5.39*** 3.36** 3.05** 4.94*** 6.23*** (1.39) (1.50) (1.55) (1.48) (1.43) (1.46) (1.37) (1.52) (1.62) -1.61*** *** *** *** -0.03*** -0.01*** -0.04*** *** *** (0.25) (0.0001) (0.0004) (0.001) (0.004) (0.001) (0.01) (13.81) (13.92) *** *** *** * ** (11.26) (11.76) (14.39) (11.71) (12.72) (11.63) (11.33) (12.69) (14.09) Number of observations R-squared Number of countries

21 Marginal effects 5. R&D Expenditures Impact 14 Marginal effects, Model 2,, R&D expenditures, % GDP, 10 years lag 21

22 Marginal effects 5. R&D Expenditures Impact 14 Marginal effects, Model 2,, CHN THA ARG UKR CZE BRA IND ITA IRL IDN MAR ECU MEX TUN TUR PAK ZAF COL BGR PRT KAZ GRC ESP PER RUS SWE KOR JPN DEU FIN AUT BEL USA FRA NLD DNK GBR NOR R&D expenditures, % GDP, 10 years lag 22

23 5. R&D Expenditures Impact Marginal effects, Model 2,,

24 The Outline 1. Motivation 2. Literature 3. Data and Methods 4. TFP and Technical Efficiency Estimates 5. R&D Expenditures Impact 6. Conclusions 24

25 6. Conclusions After the controls, 5 to 10 year lags of R&D expenditures significant impact on TFP and technical efficiency growth The elasticity of TFP growth rates w.r.t. to 10-year lag of R&D intensity (per 1.0% increase of R&D/GDP) is: for the sample: 3.0 to 6.0 for the OECD-1990+Russia sample: 7.0 to 8.0 The elasticity of TFP growth rates w.r.t. to 5-year lag of R&D density (per $1000 increase of R&D expenditure per researcher) is: for the sample: p.p. for the OECD-1990+Russia sample: p.p. 25

26 Thank you! 26