Lingnan (University) College Nov. 8, 2013 Growth of TFP in Chinese Domestic Firms:FDI Spillovers or Institutional Effects? Dianchun Jiang & Yu Zhang Institute of Int l Economics, Nankai University
I. Motivation II. Theoretic re-examination III. Model IV. Conclusion
I. Motivation
Spillover of FDI : Standard theoretic predictions Possible channels of technology diffusion Competition (Wang and Blomström, 1992). Local firms are under pressure to use existing technology more efficiently, or to search for more advanced technology, yielding productivity gains; Acquisition of human capital.(haaker,1999). Movement of labor from MNEs to existing firms, or to start new firms can generate productivity improvement Imitation (Das, 1987) is the classic transmission mechanism for new products and processes. Linkage to local industries (Aitken & Harrison, 1991) The effects depend on the absorptive ability of local industries, and threshold phenomenon may exist. (Glass and Saggi,1998)
Worldwide Empirical results on Spillovers of FDI Summary In general, empirical evidences on spillover effects of the presence of FDI are mixed. Among studies in various host countries, some authors found statistically significant positive spillover effects, but others reported no or even negative effects. There is no sign that FDI knowledge diffusion may be systematically different between developed and developing country groups. Transition economies, however, seem to have a much lower chance to benefit from FDI spillovers ------- Among the 42 studies listed in Gorg & Greenaway (2004), 10 out of 17 studies on developed countries find evidences of positive spillover effects of FDI, while the other 7 studies find no effects; 10 out of 17 studies on developing economies support the positive effect hypothesis, while the other 7 studies find no effects or significant negative effects. Only one out of 8 studies using data from transition economies obtains positive and significant FDI spillover effects.
Studies on Developed Countries(Gorg and Greenaway, 2004) 1 Girma and others(2001) UK 1991-96 Panel Micro 0 2 Girma and Wakelin(2001) UK 1980-92 Panel Micro 0 3 Harris and Robinson(2004) UK 1974-95 Panel Micro 0 4 Girma and Wakelin(2002) UK 1988-96 Panel Micro 0 5 Girma(2002) UK 1989-99 Panel Micro 0 6 Girmaand Gorg(2002) UK 1980-92 Panel Micro 0 7 Barrios and Strobl (2002) Spain 1990-94 Panel Micro 0 8 Caves(1974) Australia 1966 Cross-sectional Industry + 9 Globerman(1979) Canada 1972 Cross-sectional Industry + 10 Liu and others(2000) UK 1991-95 Panel Industry + 11 Driffield (2001) UK 1989-92 Cross-sectional Industry + 12 Haskel and others (2002) UK 1973-92 Panel Micro + 13 Ruane and Ugur (2002) Ireland 1991-98 Panel Micro + 14 Dimelis and Louri (2002) Greece 1997 Cross-sectional Micro + 15 Keller and Yeaple (2003) United States 1987-96 Panel Micro + 16 Gorg and Slrobl (2003) Ireland 1973-96 Panel Micro + 17 Castellani and Zanfei (2002b) France, Italy, Spain 1992-97 Panel Micro + for Italy; -for Spain; 0 for France
Studies on Developing Economies (Gorg and Greenaway, 2004) 1 Aitken and Harrison(1999) Venezuela 1976-89 Panel Micro - 2 Lopez-Cordova(2002) Mexico 1993-99 Panel Micro -,0 3 Haddad and Harrison(1993) Morocco 1985-89 Panel Micro and Industry 0 4 Kokko and others(1996) Uruguay 1990 Cross-sectional Micro 0 5 Kathuria(2000) India 1976-89 Panel Micro 0 6 Kokko and others(2001) Uruguay 1988 Cross-sectional Micro 0 7 Kugler(2001) Colombia 1974-98 Panel Industry 0 8 Blomstrom and Persson(1983) Mexico 1970 Cross-sectional Industry + 9 Blomstrom(1986) Mexico 1970/1975 Cross-sectional Industry + 10 Blomstrom and Wolff(1994) Mexico 1970/1975 Cross-sectional Industry + 11 Kokko(1994) Mexico 1970 Cross-sectional Industry + 12 Kokko(1996) Mexico 1970 Cross-sectional Industry + 13 Blomstrom and Sjoholm(1999) Indonesia 1991 Cross-sectional Micro + 14 Sjoholm(1999a) Indonesia 1980-91 Cross-sectional Micro + 15 Sjoholm(1999b) Indonesia 1980-91 Cross-sectional Micro + 16 Chuang and Lin(1999) Taiwan 1991 Cross-sectional Micro + 17 Gorg and Strobl(2002) Ghana 1991-97 Panel Micro +
Studies on Transition Economies (Gorg and Greenaway, 2004) 1 Djankov and Hoekman(2000) Czech Republic 1993-96 Panel Micro - 2 Kinoshita(2001) Czech Republic 1995-98 Panel Micro 0 3 Bosco(2001) Hungary 1993-97 Panel Micro 0 4 Konings(2001) Bulgaria Poland Romania 1993-97 1994-97 1993-97 Panel Micro - 0-5 Damijan and others (2001) Bulgaria. Czech Republic. Estonia. Hungary. Poland. Romania. Slovakia. Slovenia 1994-1998 Panel Micro 0 or + only for RO 6 Li and others (2001) China 1995 Crosssectional Industry + 7 Smarzynska-Javorcik (2004) Lithuania 1996-2000 Panel Micro 0 8 Zukowska-Gagelmarm (2000) Poland 1993-97 Panel Micro -
Researches Focusing on China Existing empirical results are quite optimistic, though some adverse evidences exist. Among the 10 studies testing spillovers of FDI in China summarized in Hale and Long (2006), 8 studies report unambiguously positive and significant results. Another literature summary in Yan (2004), which includes 14 studies on spillover effects of FDI in Chinese industries published in Chinese journals, shows that all authors announced positive results. Recent papers: Ran et al. (2007), using panel data from 19 industries and 30 provinces in China, finds that while the regional disparity has been growing, the net effect of FDI is positive. Buckley et al. (2007) also confirms that positive spillovers from FDI exist universally in Chinese industries, but the magnitudes may differ across source countries of the investment or industries.
10 studies testing spillovers of FDI in China summarized in Hale and Long (2006)
Puzzles 1. Why transition economies as a group cannot perform as well as the other two groups in benefiting from FDI spillover? 2. Whether the existing empirical work has exaggerated the role of FDI in promoting the productivity of Chinese industries? If so, what s wrong?
Explanations Based on Data Used or model designing Aggregate data bias Most researches discussed use either provincial level data or industry level data, without excluding FIEs from the sample. Given that FIEs are more productive on average than local firms, including these firms apparently exaggerate the positive effects of FDI (Hale and Long, 2006); Failure to control for endogeneity of FDI FDI is more likely to happen to places or industries where the local firms have higher productivity, therefore a positive correlation between FDI magnitude and productivity of domestic firms may simply reflect the location decision by foreign investors rather than the positive spillover effects of their investment. No studies exist explicitly realize the dynamic institution in transition countries is a key factor in evaluating the FDI effects in these countries. Though some author began to notice the institutional development as a necessary condition for spillovers of inward FDI (ex. Hatani, 2009) Some aspects of institutional changes, such as financial freedom, are controlled by a few authors
II Theoretic re-examination
Two basic characteristics of economic institutions in China Underdeveloped The reason why China is unlikely to outperform other economies in benefiting from FDI spillovers Poor protection of property right Unstable regulations Regulated prices in specific private business fields Important roles of government in SOEs governance and operations Biased rules for firms with different ownership, e.g. resource acquiring and entry permission (Huang, 2003). Market fragmentation in provinces or cities Keeping on improving The fact overlooked completely by existing studies and causes bias estimation As the results of the gradualist s economic reform, China has successfully restructured its economy from a planned system to a basically market-oriented system during last three decades.
Value of Knowledge with Institution constraints Under-pricing of new technology and business wisdom lowers firms incentives to learn or adopt new technologies, therefore hinder the possible channels of spillovers from FIEs. Profit can be made from various non-economic means other than management and technology Popular corruptions Common fake and shoddy goods Fragmentation of domestic markets limits possible profits of any new technology Inefficient protection of knowledge property steals profit from the innovator, though it does encourage imitation.
SOEs with Institution constraints SOEs: sheltered by governments, crippled students in markets Exclusive accession to productive resources makes some SOEs immune from competition; Soft budget constraint gives a easier way to make money; Lacking of self-decision rights; Rigid personnel system leads brain drain suffering Established R&D programs collapse in M&A
Private Firms with Institution constraints Local Private Firms: disabled competitors due to adverse legislation and regulations Political risk curtails firms life circle and encourage short-term behaviors rather than tech innovations and learning; Hostile financing position makes it difficult to raise enough capital for employee training and R&D programs. Entry prohibition in many areas denies any chance of competition, and restrains possible benefit from economy of scope; Active innovators tend to be merged by foreign competitors, and most of them actually expect that happen in order to change their ownership nature.
FIEs with Institution constraints Average quality of technology transferred to the affiliates may be much lower than expected Super-national Treatment engenders institutional rent; Lack of enough identified local competitors; Risks from changeable policies; Weak protection of intellectual property rights Lack of linkage with local firms, esp. in backward linkage Underdeveloped institutions in terms of not only changeable regulations and the weak protection of intellectual property rights, but also within an institutional context for production in emerging economies, including fragmented supply chains and disorganised inter-industry linkages discourage MNEs from close interactions with local firms (Hatani, 2009).
Why Changing institutions are important in Valuing FIEs Contribution? 4 3.5 Labor Share of Non-SOEs TFP of Domestic Firms 0.9 0.8 3 0.7 2.5 0.6 0.5 2 0.4 1.5 0.3 1 0.2 0.5 0.1 0 2011 2010 2009 2008 2007 2006 2005 2004 2003 2002 2001 2000 1999 1998 1997 1996 1995 1994 1993 1992 1991 1990 1989 1988 1987 1986 1985 1984 1983 1982 1981 1980 0 Data Sources: TFP: authors calculated; Labor: Yearbook of Chinese Industrial Economy. * From 1999 on, only enterprises with total products of 5 million or more included
Institutional Change in Chinese Economy Property right protection improved Private capital is more encouraged and expands rapidly SOEs are restructured or privatized More liberalized Financial System (incl. development of stock market, banking industry, and more tolerant of informal financing activities) The governmental intervention in market is considerably limited 8 7.5 7 6.5 6 5.5 5 4.5 4 3.5 3 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 Index of marketization of Chinese economy Data Source: G. Fan and X.L. Wang (2011
Dynamic institution as a driving power for domestic firms productivity As institutions do matter (North, 1976), gradual institutional evolution must have been continuously affecting the Chinese economic development and the productivity. Some Examples The efficiency of private owned firms is 42.46% higher than that of SOEs (Yao & Zhang, 2003) Due to the institutional change, transaction costs saved in China are estimated about 1. 64% of GDP annually during 1991~ 2002 (Y. Jin, 2005). A 1%improvement in institutions generates on average a 5% increase in output per worker (Hall and Jones,1999; Tiago etal., 2005) Two reform measures carried out by Chinese central government in the late 1990s alone, namely an enormous wave of ownership restructuring due to the formal endorsement of private property rights, and the launch of a large scale labour retrenchment program, can explain about 30% of the TFP growth of Chinese manufacturing industries during 1998-2003 (Pandey and Dong, 2009 ).
Implication on empirical studies Solow residual (TFP growth) dy da dk dl Y A K K L L Closed Economy:? human capital; R&D Open Economy: + FDI spillover Transition Economy: + institutional factors Previous studies may have overstate the FDI spillovers in China. The essential problem is that they have neglected Chinese dynamic nature, and therefore fail to separate FDI spillovers from productivity improvement associated with institutional changes in the economy. Any effort to evaluate the impact of FDI on the development of China (and other transition economies) should draw into institutional change into the researching scenery.
III Econometric Model
Model Let the production function for domestic department be of Hicksian Neutral type: Y( t) A( t) F( K, L) and assume the TFP be FDI Spillovers domestic firms decision Log form: Institutional effects
Specifications Equation without institution consideration: (1) Equation with control for institutional changes: Equation taking into account the interaction of FDI and institution (2) (3)
Variables and Data Data Panel data consists of 31 provinces (cities) during 1999-2009. Collected from Yearbook of Chinese Industrial Economy and Yearbook of Chinese Science and Techlodgy. Explained Variable TFP: Total factor products of domestic industrial division (ie. industries excluding FIEs), where TFP is calculated in Data Envelop Analysis (DEA) model (Fare et al., 1994). Key Explaining variables FDI: A Composite index of FIEs presentation: First calculate fdi it, the average of FIEs shares in Fixed Capital, Total Products and Total labor Employed Standardized by FDI fdi / fdi it it min Z: Two different measures: Z m : modified marketnization index based on Fan et al.(2000-2008), in which factors related to FIEs are removed (standardized as the way in FDI). Z P, the average of Non-SOEs shares in Fixed Capital, Total Products and Total labor Employed (standardized as the way in FDI).
Variables and Data Control variables (of domestic firms) R it, R i,t-1 : Ratio of R&D expenditure in total products, of period t and t-1 PGDP: Real GDP per capita, k: Capital per capita, calculated by (Fixed capital)/(labor), controlling for the development level of areas. TB: Technological Basis, measured by the Pure Efficiency (Farrell efficiency) of domestic firms obtained in DEA. Additional explaining variables TFP i,t-1 : lagged value of TFP, in regard to that TFP reflects firms cumulative knowledge capacity, and also to control for other unobservable factors. -- Dynamic panel data (DPD) models. HMT: FDI source variable, defined by the share of HMT firms in all FIEs (average value of the measures in Fixed Capital, Total Products and Total labor Employed).
Regression Methods Each equation is estimated independently in two methods: Fixed effect approach, with cross section weighting adjustment (but only report results of eq (1) here) Generalized Method of Moments (GMM, Hansen, 1982 ), to cope with the problem of potential endogeneity, as well as autocorrelation and heteroscedasty. Arellano-Bover/Blundell-Bond estimation Predetermined: R, PGDP, k and Z Endogenous: FDI, HMT; maximum lags of predetermined and endogenous variables for use as instruments: 4 Regressions are run in Stata 12 Software package.
Explaining Variables Estimated Results of Eq. (1) FE(AR) Sys-GMM Coefficient t -statistics Probability Coefficient t -statistics Probability ln(tfp i,t-1 ) 0.8972 10.30 0.000 0.2287 4.85 0.000 ln(r i,t-1 ) 0.0978 4.42 0.000 0.0908 4.37 0.000 ln(r it ) -0.0374-2.79 0.000 0.0143 0.84 0.401 ln(pgdp it ) 0.1086 3.50 0.000-0.0181-0.68 0.495 ln(k it ) 0.0632 2.16 0.032 0.0470 1.98 0.048 ln(tb it ) 0.0616 4.78 0.000 0.1091 6.21 0.000 ln(fdi it ) 0.0381 2.72 0.007 0.0278 2.09 0.036 ln(hmt it ) -0.0027-0.30 0.768-0.0037-0.35 0.724 C -0.7758-2.52 0.012-0.3200-2.04 0.041 R 2 0.9413 Sigma_e 0.0484 F Statistic 100.5753[0.0000] Hausman Test 31.4229 [0.0001] Sargan Test 23.8012 [1.0000] Arellano-Bond AR(1) -3.6309 [0.0003] Arellano-Bond AR(2) 0.5025[0.6153]
Explanation Housman test confirms the fixed effects. Values of F- statistic and R-square seem to be quite satisfactory. But the estimators are not consistent due to endogenity. Sagan test identify the validity of the IVs, and Arellano- Bond test does not reject the hypothesis of no autocorrelation in levels, which is required for GMM. In both estimations FDI are reported to have statistically positive effects on domestic firms productivity, but the effects in GMM are much weaker, both in size and significance. The endogenity of FDI does cause upward bias. HMT s share has no significant effects, which means that HMT firms are as good as MNEs from other countries in regard of knowledge sharing. Coefficients for other explaining variables have expected signs.
Estimated Results of Eq. (2)(Sys-GMM) Explaining Variables Z=Z P Z=Z M ln(tfp i,t-1 ) 0.1187 3.14 0.002 0.1334 2.81 0.005 ln(r i,t-1 ) 0.0409 1.69 0.091 0.0513 2.63 0.009 Coefficient t -statistics Probability Coefficient t -statistics Probability ln(r it ) 0.0210 0.77 0.444-0.0054-0.35 0.728 ln(pgdp it ) -0.0980-1.12 0.265-0.0681-2.91 0.004 ln(k it ) 0.1158 1.73 0.083 0.0948 4.16 0.000 ln(tb it ) 0.0726 2.61 0.009 0.0717 4.27 0.000 Z it 0.2214 6.42 0.000 0.0660 3.15 0.002 ln(fdi it ) 0.0231 1.28 0.201 0.0086 0.69 0.491 ln(hmt it ) -0.0069-0.21 0.830-0.0137-1.43 0.152 C 0.0393 0.007 0.941-0.3389-2.45 0.014 Sargan Test 20.8584 [1.0000] 22.9175[1.0000] Arellano-Bond AR(1) -3.5974 [0.0003] -3.5181[0.0004] Arellano-Bond AR(2) 0.6477[0.5172] 1.1254[0.2604]
Explanation New added institutional variable has a strong significant positive coefficient, in both cases with different institution measures. This confirms that the institution changes drive domestic firms productivity up. The coefficient for FDI becomes insignificant, in sharp contrast to the results in eq. (1). Implications: Besides R&D activities, technology basis and other domestic firms own factors, it is the underlying institutional improvement, rather than the presence of FDI, to be the major factor promoting local firms productivity.
Estimated Results of Eq. (3)(Sys-GMM) Z=Z P Z=Z M Coefficient t -statistics Probability Coefficient t -statistics Probability ln(tfp i,t-1 ) 0.1143 2.34 0.019 0.1231 2.54 0.011 ln(r i,t-1 ) 0.0424 2.20 0.028 0.0272 1.78 0.076 ln(r it ) 0.0029 0.18 0.853 0.0108 0.70 0.485 ln(pgdp it ) -0.0687-3.41 0.001-0.0811-3.73 0.000 ln(k it ) 0.1006 4.61 0.000 0.1093 5.00 0.000 ln(tb it ) 0.0789 4.90 0.000 0.0790 5.15 0.000 Z it 0.1620 4.53 0.000 0.0467 2.99 0.003 ln(fdi it ) -0.0016-0.14 0.888-0.0622-3.21 0.001 ln(hmt it ) -0.0158-1.68 0.092-0.0169-1.77 0.077 Explaining Variables Z it ln(fdi it ) 0.0091 3.02 0.003 0.0257 5.40 0.000 C -0.2680-2.13 0.033 0.0948 0.61 0.539 Sargan Test 15.3284 [1.0000] 18.1032 [1.0000] Arellano-Bond AR(1) -4.0483 [0.0001] -3.4741 [0.0005] Arellano-Bond AR(2) 0.8204 [0.4120] -0.6207[0.5348]
Explanation The coefficient for the interaction term, Z*ln(FDI), is significantly positive. But the interaction consists of two-way effects. Compared to unconditional effects in Eq.(2), the conditional marginal effects in Eq. (3) become as the following: Since ˆ 0, ˆ 0, FDI has effects on domestic firms TFP if and only if the institution develop to some critical level. In both cases of Z, ˆ 0, ˆ 0, implying that marketization drives domestic firms TFP up independent of the presence of FDI. FDI from HMT areas are more disappointing, while MNEs from other countries may play a positive role in knowledge diffusion.
IV Conclusions and Future Research
Conclusions 1. FDI spillover may be a completely different story in a transition economy, both in mechanism and in size. 2. There are no evidence for Intra-provincial spillovers of FDI as a whole in China. Earlier studies have largely exaggerated FDI spillovers in China due to failure to control for institution effects. 3. The growth of Chinese domestic firms' productivity is mainly a result of the underlying institutional changes, rather than the presence of FDI. 4. An improved institutional environment tends to be helpful for the local firms to benefit from the knowledge diffusion of FDI. 5. An extraordinary route of spillovers may exist in China and other transition economies -institutional channel, either in positive or negative way.
Future Research Firm level data Inter-industrial effects; FDI effects on economic growth, etc.
Thank you!