The Unbalanced Growth Hypothesis and the Role of the State: the Case of China’s State-owned Enterprises

Carsten A. Holz Hong Kong University of Science & Technology, and USC

The Stanford Center for International Development 26 April 2011
1

Motivation: Unbalanced Growth Hypothesis
Albert Hirschman (1958):
questions if a big push can overcome a low-level equilibrium trap of development questions the ability of governments to finance simultaneous industrialization across all sectors -> create “inducement mechanisms”

Unbalanced growth hypothesis: (1) Identify high-linkage sector(s) = sectors that create high profit opportunities in other sectors
Supply bottlenecks create profit opportunities in upstream industries -> induce further investment. Producers propagate further uses of new products: profit opportunities in downstream industries -> induce further investment.

(2) Policy recommendation for government: invest in high-linkage sectors
-> new profit opportunities in other sectors -> private entrepreneurs enter those other sectors

=> high economy-wide growth

2

China in the reform period
Question: Has the Chinese state retained state-owned enterprises predominantly in high-linkage sectors and thereby promoted rapid economic growth?

Answers: A high degree of linkage does promote economy-wide growth. State ownership is not strategically concentrated in high-linkage sectors.

3

1. Literature, question
a. Literature b. China and the unbalanced growth hypothesis

2. The linkage effect on growth 3. Linkage indicators 4. Evidence 5. Conclusions

4

1976. Literature. Rostow (1956). Lewis (1954). 1982) Unbalanced growth hypothesis implicit Industrial policy discussion (next two slides) 5 . Nurske (1953) Application of the unbalanced growth hypothesis Yotopoulos and Nugent (1973.1984). Literature Unbalanced growth hypothesis Hirschman (1958) vs/and Rosenstein-Rodan (1943.1. question a.

and Sabel (2008) industrial policy “in the small” (putting mechanisms into place to identify and remove roadblocks facing existing economic activities) vs.UGH: at early stages of economic development: all products associated with industrialization(?) 6 . Rodrik.UGH: criterion for making strategic bets Rodrik (2010) “good industrial policy attempts to enhance the relative profitability of nontraditional products that face large information externalities or coordination failures…” --.Unbalanced growth hypothesis implicit: industrial policy discussion Cohen (2007) vertical (sector-specific) vs. horizontal (framework) policy --.Hirschman’s unbalanced growth hypothesis (UGH) = vertical Hausmann. “in the large” (strategic bets on industries one would want to see develop) --.

incl. Hirschman: role for the government includes directly productive activities 7 .cont. 2006): “the government is only focused on providing complementary inputs to the market” vs. Wade (2009): industrial policy as “any sectorally or activity-targeted intervention” with government leadership or followership of the market (leading decentralized private producers to do something they would otherwise not do.UGH: government leadership. in directly productive activities Pack and Saggi (2006): industrial policy: “any type of selective government intervention or policy that attempts to alter the structure of production in favor of sectors that are expected to offer better prospects for economic growth in a way that would not occur in the absence of such intervention in the market equilibrium” --.UGH: in favor of sectors that are expected to offer better prospects + Role of government: Industrial policy discussion (Rodrik. betting on some of what the private sector is already doing) --.

adoption of the modern enterprise system. China and the Unbalanced Growth Hypothesis Pre-reform period: balanced growth Planning textbook (Li. 1983): planned economy is superior to the market economy because it leads to “balanced and continuously rapid economic development of social production” and the “rational use” of resources -> Basic industrial structure is in place at the beginning of the reform period -> How to develop faster? Reform period: opportunity/need(?) for unbalanced growth strategy State-owned enterprise (SOE) reform: Ongoing. but in particular during the period 1998-2000: 抓大放小. withdrawal of the state from specific sectors -> Is the state making strategic choices from which sectors to withdraw? Is the strategy designed to optimize economy-wide growth? + Fiscal decentralization -> incentives for localities to develop + Creation of national markets -> scope for specialization 8 .b.

local specialization incentives for local governments to promote economic growth Chinese provinces the size of a European country Why focus on SOE policy? other government policies (explicit industrial policies) hard to quantify WTO prohibition for many types of industrial policy SOEs as key mechanism through which the state engages in the economy 9 .The nice thing about China: avoid difficulties of cross-country studies 8 regional. 31 provincial observations all subject to similar institutional and cultural environment cohesion within a nation state increasingly national markets.

Yusuf et al. 2003): SOE profitability not inferior if circulation taxes and capital intensity are controlled for (+ debunking a number of other myths about intrinsic systematic inferiority of SOEs) Brandt and Zhu (2000.SOE literature Lardy (1998).) Here: positive externalities via linkages? 10 . 2001. Jefferson and Su (2006). 2010): negative externalities of SOEs (lead to cycles of growth and inflation etc. (2006). and many more: point to inefficiency of SOEs. 2003): policy-determined burdens place SOEs at disadvantage (distorted output prices. need for privatization Lin et al. high capital intensity. social burdens) Holz (2002. (1998.

Literature.1. Evidence 5. b. The linkage effect on growth a. Earlier research Critique of the use of quantitative measures in testing the UGH 3. question 2. Linkage indicators 4. Conclusions 11 .

not changes in gross output (c) Take into account government 12 .2. The linkage effect on growth a. level (1) Testing for unbalanced growth (i) calculate Hirschman-compliance index for each of 36/39 countries: corr. 1973) Calculate linkage coefficients (6. Earlier research: Yotoupolos and Nugent (QJE. 18 sectors) for each of 2 groups of countries: developed countries (6 I-O Tables). coeff.5 prob. Take square root. underdeveloped countries (5) use identical backward linkage coefficients if not different at 0.) economy-wide growth rate. Differently here: (a) High linkage coefficient of a sector need not imply high growth in that sector (b) Sectoral linkage coefficients measure profit opportunities. of (i) and economy-wide growth -> insignificant (2) But: “balanced-growth” version: conformance across sectors to a particular imbalance pattern defined by sectoral linkage coefficients leads to high growth Squared deviations of sectoral growth rates from the weighted (by sectoral linkage coeff. Weigh squared deviations by sector’s share in GDP. Standardize by economy-wide growth rate. coeff. sectoral linkage indices and sectoral growth rates (ii) across countries: corr.

little further developments re linkage effect on growth: Discussion of Yotoupolos and Nugent (QJE. and Badillo (2009) wp on threshold effects of technological integration for growth 13 . 1973) in the QJE in 1976 (with a rejoinder by Yotoupolos and Nugent) Nugent/Yotoupolos sequel-extension (1982) Ortiz. Castro.Since then.

(1995). rubber and oil palm export industries on development of domestic light engineering industry (Thoburn. AndreossoO’Callaghan and Yue (2002). Cai and Leung (2004) 2. Ichimura and Wang (2003). Li and Xue (1998). 1973) Effect of changes in intra. Coady and Jie (1993).Other linkage-related literature 1. Linkage effects from demand for capital goods in Malaysia’s tin. AndreossoO’Callaghan and Yue (2004) Input-output literature for China w/o reference to linkage coefficients: Polenske and Chen (1991). Dietzenbacher and Linden (1997). Sonis et al.and interregional backward linkage coefficients on gross outputs of European Union countries (Sonis et al. Miller and Lahr (2001).. Empirical exploration of linkage effects --. Xu and Liu (2004) 14 . 1996) Literature on social accounting matrices 3.not on growth. Heimler (1991). Clements (1990). Extensive theoretical discussion of variations of linkage coefficients For ex.. Calculation of linkage coefficients for China Bhalla and Ma (1990). Cella (1984). For ex. Li (1992). Diamond (1976) Schultz (1977).: Strassert (1968).

a victim of the tendency to subordinate economic hypotheses to the restrictive requirements of elementary regression and correlation analysis. of another sector? differences in linkage coefficients across sectors? (too narrow) (2) Measurement of linkages: how to operationalize? 15 .: largest value for backward linkage in grain mill products… therefore grow grain? Low forward linkages in machinery and transport equipment: selling to final demand But no fundamental difference between a stimulus from agriculture towards establishing a tractor assembly plant vs. Critique of the use of quantitative measures in testing the UGH Hirschman: “excessive reliance should obviously not be placed on these rankings.” (1) (a) Reduction of book-length theory of unbalanced growth to an argument about linkages (b) Meaning of “unbalanced growth?” different sectors growth at different speed? level of dev. an insecticide mixing plant McGilvray (1977): “It is regrettable that an original and valuable contribution to an understanding of economic development processes has been emasculated and oversimplified. of one sector imposes a constraint on the dev.b. based as they are on a mental experiment subject to numerous qualifications” Assumption that a country’s development started with the industry in question Ex.

Conclude: non-falsifiable UGH? Abandon attempts to test the UGH Krugman (1994): fall of the “high development theorists” Here: specific focus. using Chinese data: (1) To what extent do linkage effects impact on economic growth? (2) To what extent does the Chinese government consider linkage effects in decisions on the distribution of state ownership across sectors? 16 .

Evidence 5.1. The linkage effect on growth 3. Linkage indicators 4. Literature. question 2. Conclusions 17 .

Input-Output Table X11 + X12 + … + X1n + Y1 = X1 Ax+y=x => x = (I-A)-1 y Define input coefficient a12 = X12 / X2 etc. Col. D: depreciation. value Intermed. 1 X11 X12 … X1n U1 C1 I1 E1 M1 Y1 X1 sales of 2 X21 X22 … X2n U2 C2 I2 E2 M2 Y2 X2 sector i … … … … … … … … … … … … n Xn1 Xn2 … Xnn Un Cn In En Mn Yn Xn Sum int. i sum of (I-A)-1 is the BL of sector i 18 . T: net taxes on production. Con. X1 X2 Xn L: labor remuneration. W is value added. Linkage indicators Intermediate purchases Sum Final demand Sum Gross of sector j interm. purch. Inv.) dem.3. Con. outp. The sum of sectoral value added equals the sum of final demand (income approach GDP equals expenditure approach GDP). Exp.: consumption. W1 W2 … Wn Gr.: exports. V1 V2 … Vn L L1 L2 … Ln Primary D D1 D2 … Dn inputs T T1 T2 … Tn S S1 S2 … Sn Sum prim. S: operating surplus. val. inp. Exp.: investment (gross fixed capital formation). Imports final output 1 2 … n sales (neg. Inv.

excluding sector i. decrease if sector i were absent? -> Linkage measures in the literature are in terms of output Output linkage of a particular sector measures how a unit increase in final demand or gross output in one particular sector affects output across the sum of all sectors (possibly excluding the original change. Intermediate sales / GOV (2) BL and FL based on Leontief inverse. 2001) of a sector: By how much does an economy’s total output in all sectors. Gosh inverse (3) LSD and GSD (Cai and Leung. 2004) (4) Index of vertical integration INT (Heimler.Five sets of sectoral linkage measures (1) Intermediate purchases / GOV. 1991) (5) Total linkage index (Miller and Lahr. or the sector in which the original change occurred) Here: profit opportunities created Profit linkage of a particular sector = how a unit increase in final demand or gross output in one particular sector affects profit across the sum of all sectors (…) Multiply output effect that a particular sector experiences by its ratio of operating surplus to output 19 .

need a comprehensive measure of linkage for each locality: “coefficient of interdependence” = weighted sum of all sectoral linkage coefficients. with as weights the sectoral output values (2) Measure of variation for linkage coefficients given an equal depth of sectoral interdependence. additional manipulations (where needed below) (1) For comparisons across geographic entities.Two standard. linkage coefficients of approximately equal value across sectors may have a different impact on growth than linkage coefficients that differ across sectors 20 .

where v is a vector of sectoral ratios of value added to gross output value. and H = (I Aii . per unit of value added in sector i absolute reduction in output outside sector i (in sector[s] j) if sector i is eliminated total output change outside sector i (in sector[s] j) due to one-unit output change in sector i 21 . and VAi is value added of sector i e’(xj – xjR) = e’(αjj Aji H yi) + e’(αjj Aji H Aij αjj yj) where αjj = (I-Ajj)-1 . vertical integration (Heimler. outside sector i (in sector[s] j).Aij αjj Aji)-1 TLi = e’(xj –xjR) / Xi share of intermediate purchases in inputs of sector j share of intermediate sales in output of sector i total (economy-wide) output change due to one-unit change in final demand for sector j total output (input) change due to one-unit change in value of primary inputs into sector i total output change due to one-unit change in output of sector i total output change due to one-unit change in input of sector i value added generated by sector i. 2 Leontief .Linkage Indicators of Individual Sectors 1 Hirschman . (2004) 4 Index of . effect: Leontief – HEM BL(1)j = ∑iXij / Xj = ∑iaij FL(1)i = ∑jXij / Xi = ∑jbij BL(2)j = ∑iαij = jth element of e’(I-A)-1 FL(2)i = ∑jβij = ith element of (I-B)-1 e LSD = BL(3)i = 1 + e’ (I-Ajj)-1 Aji GSD = FL(3)i = 1 + Bij (I-Bjj)-1 e INTi = [v’ (I-Ajj)-1 Aji Xi] / VAi . inverse Gosh inverse 3 Cai/Leung . 1991) 5 Total linkage .

state ownership. Linkage indicators 4. Evidence a.1. State ownership (i) Ownership data (ii) Linkage and state ownership c. The linkage effect on growth 3. Conclusions 22 . and real GDP growth 5. question 2. Literature. Linkage coefficients (i) Choice of input-output table (ii) Linkage values b. Linkage.

2000. 1985. apply to all provinces within a region Unit of analysis: province Three periods: 1994-1997 [1997 I/O Table] 1997-2000 Remember: SOE reform program: 1998-2000 2000-2003 23 .M) Regional: 1987: seven regions. Evidence a. 1987. 1990. 2005 Provincial: mostly “secret” – would be ideal (but would be w/o detailed domestic X. 1995. nine sectors 1997: eight regions. Linkage coefficients (i) Choice of input-output table National: 1981. 1983.4. 1997. 1992. 17 sectors -> USE. 2002.

. Potential for local substitution of domestically imported goods is ignored. the regional input coefficients capture the impact of a change in ‘final demand for the products of this region’ on ‘production in this region’ (only) They also ignore the impact of a change in ‘final demand for the products of another region’ on ‘production in this region. 24 . but to a corresponding change in the production in this region (or another region). not in other localities.’ Why this choice? Local government through its SOE decisions wishes to influence economic growth in its locality. (b) International imports: no choice: competitive A change in final demand for the products of this region will not lead to imports from abroad.e.How to handle trade flows? (a) Interregional trade flows and international exports: domestically non-competitive I.

do the linkage patterns across sectors differ between linkage indicators? Question 2: For any one specific linkage indicator. do the values differ across sectors and across regions? 25 . for each of the 17 sectors in each of the 8 regions.(ii) Linkage values Calculate the values of the 5 sets of linkage indicators (8 distinct indicators). both output and profit linkage indicators. Ask two questions Question 1: In any one region.

Two (of the four) BL indicators and two (of the three) FL indicators are positively correlated with the total linkage indicator (TL). Linkage patterns of different linkage indicators across sectors are similar for different regions.If not. -> TL captures sth of everything? Profit linkage indicators: No clear pattern! -> difficult to narrow down the choice of indicators In the following. then one representative linkage indicator suffices for the remainder of the analysis. do the linkage patterns across sectors differ between linkage indicators? --. GSD 26 . Focus on one region: Beijing-Tianjin -> Table Output linkage indicators: Backward linkage measures tend to be correlated with BL measures. FL) + Hirschman trusts more in BL than in FL… alternatives: INT. focus on TL (of more interest than breakdown BL vs. BL and FL measures tend to be negatively correlated.Question 1: In any one region. Forward linkage measures tend to be correlated with FL measures. LSD.

785 0.724 ***0.267 FL(1) 1.047 ***0.649 ***0.169 **0.248 FL(2) 1.196 0.607 -0.641 TL 1.868 **-0.Correlation Coefficients Across 17 Sectors (Region Beijing-Tianjin.557 ***0.000 ***0.141 BL(2) 1.125 LSD 1.300 -0.690 BL(2) 1.000 **0.605 -0.000 *0.411 ***0. *** 1%.416 0.827 ***-0.000 **-0.199 ***-0.vs.974 ***0.154 ***-0.212 FL(2) 1.518 ***-0.000 Same type of linkage indicator. the (output) linkage coefficient is the share of intermediate purchases in inputs of sector j.000 -0.347 -0.346 LSD 1.306 **-0.472 INT 1.806 ***0.000 **0. profit-linkage ***0.892 ***-0. output.240 -0.925 ***0.823 For the meaning of the individual column and row labels see earlier table.689 0.502 0.000 -0.566 TL 1.343 **0. 1997) BL(1) FL(1) BL(2) FL(2) LSD GSD INT TL Output-linkage indicators BL(1) 1.568 -0.335 -0.678 -0.582 *-0.000 *-0.000 Profit-linkage indicators BL(1) FL(1) 1.146 GSD 1.000 ***-0.000 0. Significance levels: * 10%. ** 5%.000 -0.373 -0.448 ***0.629 -0. Source: manipulation of data from SIC (2005).304 -0.392 0.905 GSD 1.130 -0.334 INT 1.582 ***0.432 0.931 **-0. 27 . BL(1) values in the case of profit-linkage coefficients are not meaningful. with no possibility to transform this coefficient into a profit measure.000 ***0.730 ***0.970 *0.000 0.

-. INT. 28 .If they differ across sectors. GSD also differ across sectors.Each region’s set of linkage indicators will be applied to all provinces within the region. CV ≠ 0. regions similar. Findings (Table) Total Output Linkage (TOL) values differ substantially across sectors. (Same for TOL. (Data problems?) Patterns of TPL coefficients across any two regions tend to be different. then different regions should have different distributions of state ownership. Total Profit Linkage (TPL) values differ substantially across sectors.) -> Expect diff. CV ≠ 0. Patterns of TOL coefficients across any two regions are similar. furthermore. in degree of state ownership across sectors within any one region. regions: sectors different. differ across regions. across sectors) across regions.If they. then some sectors should attract more state attention than others. + Possibly some diff. Regression of TPL on sectors.Question 2: For one specific linkage indicator. (in distribution of state own. --. LSD. little across regions. do the values differ across sectors and across regions? -. Central region as special case: negative profit effects.

904 0.417 0.541 0.741 0.494 0.112 1.747 0.943 0.347 0.505 0. except the North-Southwest value (1% level) Northeast 1.348 0.678 0. Other 0.485 10.662 0.702 0.V.351 0.394 0.482 17.609 0.317 0.798 0. Wood 0.512 + + + + + + + 0.528 0. 29 ^ Weights consist of sectoral value added.799 Beijing-T.618 0.785 0.684 0.389 0.560 0.868 0.530 0.427 0.605 0.837 6.378 0.272 0. Transp.389 0.314 0.794 0.169 0.517 1. 1. Electronic 0.552 0.736 + + 4.826 Central 1.406 1.022 0.502 0.602 0. Mining 0.760 14.433 0.593 7.117 0. 0.288 0.^ Correlation coefficients: all signific.739 0.424 0.748 0.185 0.488 0.000 0.725 0. the squared deviation of a particular sector’s linkage coeff.893 0.449 Average 0.876 0.715 0.418 + + + + 3.535 0.424 0.291 0.000 0. of var.923 0.500 0. Non-metals 0.292 0.595 0.496 0.889 0.407 0.923 0.892 0.652 Coeff.361 1.501 0.787 0.556 + + + 8.493 Weighted C.818 0.953 1.564 0.387 0. Italics mark those non-industrial sectors with a linkage coefficient at least as high as the third-highest linkage coefficient in industry.633 0.786 13.660 0.790 0.773 0. eq.920 0.646 0.351 0.506 0.344 0.000 0.513 0.603 0.^ 0.520 0. 1.813 0.000 0.491 0.477 0.463 0.809 + + + 12.542 0. Utilities 0.156 0.913 + 0.360 0.647 0.483 0.524 0.851 0.362 0.812 South 1. Foods 0.000 0.467 0.801 0.619 0. Machinery 1.946 0.246 16.813 0.686 0.798 0.640 +0.558 0.508 0. Commerce 0.400 0.291 1.613 11.906 0.568 0.834 North 1.714 0.854 0.1% level.494 0. from the weighted ave.324 0.814 0.609 0. Construction 1.536 0.826 0.806 Northwest 1.555 0.312 0.396 0.552 0. Agric.389 0. In the calculation of the standard deviation (used in the coefficient of variation).769 0.000 0. 0.909 0.693 0.221 0.791 0.749 0.512 0. Metals 0.722 East 1.465 0. Chemicals 0.834 0.769 0. at the 0.838 0.304 0.333 Weighted Ave.551 0.898 0. + .529 0. 8 Regions (1997) North.415 0.362 0.714 0.799 0.286 2.461 0.000 0.509 0. Textiles 0.338 0.582 15.652 0.636 0.000 Superscript + denotes the three sectors within industry (industry comprises sectors 2 through 14) with the highest linkage coefficient.852 Southwest 1. linkage coeff.656 0.665 0.258 0.833 0. 0.966 +0.745 0.615 0.836 0.770 0.881 9.901 0. Other serv.417 0.712 0. 17 Sectors.021 0.624 0.BeijingNorth East South Central NorthSoutheast Tianjin west west 1.551 0.270 0.Total Output-Linkage Coefficients.596 0. Paper 0. was weighted with the sector’s share in overall VA.794 0.581 0.788 + + 5.364 0.

652 0.010 0.658 ***0.Beijing.042 0.026 0.011 0.030 -0.047 0.060 13.025 0.543 0.037 0.038 0.026 0.035 0.006 0. Electronic 0.043 0.033 17.033 0.008 0.000 0.542 0.009 0.048 -0.008 0.058 0.023 0.470 East 1.607 0. of var.411 0.035 -0.036 -0. ** 5%.022 0.482 ***0.485 0.000 ****0.000 *-0.023 0.748 -0.012 0.035 0.390 *0.043 0.019 0.031 -0.021 0.051 0.072 + 4. Metals 0.009 0.Total Profit-Linkage Coefficients.037 0.025 0.052 0.011 0.011 0. Agric.049 -0. Chemicals 0.516 0.015 0.595 ***0.012 0. 0.014 0.024 -0.033 0. Machinery 0.379 0.056 0.341 0.053 0.021 0.056 + 0.038 0.000 ****0.^^ 0.1%.408 *0.033 0. Ave.047 0.039 +0.025 9.000 ****-0.001 0.024 0. 0.274 -0.571 0.032 0.012 0.024 0.039 0. 0. 17 Sectors.094 0.284 Southwest 1.067 0.017 0.031 -0. Commerce 0.051 0.010 0.440 0. 30 .307 -0.024 0.045 0.018 0.002 + + + -0.401 Weight.021 0.002 + + 0.024 11.030 0.042 +0.833 0. Construct.039 0.457 **-0.016 Average 0.979 0.027 0.019 2. Mining 0.036 0.094 0.032 0. eq.026 + 7.043 16.136 Northwest 1.026 Coeff.031 +0.303 -0.030 0. 8 Regions (1997) North.069 0.050 + + 0.025 12.713 0.037 + 0.660 ***-0.047 0.V.051 0.022 Weighted Ave.680 South 1.755 0.050 -0.657 North 1. Wood 0.052 0.050 0.039 -0.040 0.046 6.025 -0. Textiles 0.042 0. Utilities 0.021 0.049 0.286 0. 0.500 **0.040 0.562 ***0.031 0.031 0.851 ***0.028 -0.041 0.033 0.048 0.020 3. ^^ Weights consist of sectoral operating surplus.011 0.022 Weight. Paper 0. *** 1%.031 -0.058 0.036 0.041 -0. **** 0.038 0.V.032 + + + + 5.067 0.475 0.043 0. 1.057 0. Non-metals 0.028 0.^ 0.372 0.577 ****0.015 + 0.005 0. Foods 0.019 0. C.332 ***0. Other serv.062 0.034 -0.020 -0.018 0.058 + 0.023 0.167 Beijing-T.695 **-0.335 0.030 0. Transp.743 0.395 0.009 0.036 8.039 0.862 0.046 0.032 0.677 Central 1.061 0.037 10.442 0.^ Weighted C.063 0.030 0.030 0.025 0.006 15.030 0.^^ 0. 0.014 0.044 0.627 0.013 0.041 -0.364 Correlation coefficients Northeast 1.052 -0.000 **0.063 +0.000 Significance levels: * 10%.421 *0.766 **-0.052 +0.North East South Central NorthSoutheast Tianjin west west 1.000 0. Other 0.029 + + 14.047 0.037 0.053 0.016 -0.

1998-2000: SOE reform program! Statistical breaks: 1998 SOEs vs. 1997-2000. Evidence b. 2005.Stay close to date of input-output table (1997): one period before.End in 2003 b/c 2004 economic census. SOSCEs (state-owned and state-controlled enterprises) 1998 redefinition of directly reporting industrial enterprises 2003 adoption of new sectoral classification 31 . two after --. State ownership (i) Ownership data State ownership by sector available for industry: industrial state-owned enterprises (SOEs) as subset of “directly reporting industrial enterprises” Data on value added available since 1993 (first year of poor quality): choose periods 1994-1997 [1997 = input-output table].4. 2000-2003 --.scarcity of value added data --.

Robustness: Similar results for changes in sectoral shares of SOEs during a period (in relative or absolute terms) Similar results for both measures of state ownership: SOE share in sectoral value added sector’s share in SOE value added of the province Findings for LSD. INT. => State accounts for large share of output in low-linkage sectors. GSD no different Questions… Do high linkage effects not come with rapid economic growth in China? Is the link between ownership and linkage more subtle than a correlation coefficient can capture? 32 .(ii) Linkage and state ownership (1994. 1997. 2003) (Regional) TOL. 2000. TPL and (provincial-level) state ownership not correlated across sectors.

414 -0.170 -0.036 -0.248 0.459 -0.168 Guangxi -0.368 *-0.193 -0.572 *-0.375 -0.556 Southwest -0.123 -0.492 -0.760 **-0.164 South -0.727 ***-0.687 ***-0.235 -0.516 -0.135 -0.174 -0.410 -0.342 0.192 -0.313 0.310 -0.105 -0.021 0.268 ***-0.682 ***-0.303 Zhejiang -0.635 -0.559 -0.402 **-0.265 -0.320 0.595 The 13 sectors covered are the 13 industrial sectors (sectors 2 through 14) out of the total of 17 sectors.080 0.124 -0.521 **-0.569 -0.145 -0.450 Shanghai -0.242 -0.168 -0.082 0.154 -0.139 -0.044 -0.326 -0.079 0.099 -0.159 -0.221 -0.333 Ningxia Xinjiang -0.024 -0.701 -0.267 -0.360 *-0.231 -0.265 -0.152 -0.175 -0.003 -0.026 0.476 Hubei -0.094 -0.031 -0.291 -0.475 *-0.143 -0.290 -0.035 0.243 -0.022 -0.624 -0.717 -0.544 Guizhou -0.218 -0.001 Chongqing -0.040 0.384 -0.702 ***-0.061 0.481 Total output-linkage 1997 2000 Total profit-linkage 2003 1994 1997 2000 2003 -0. 33 .254 0.259 -0.245 Hainan 0.148 -0.209 -0.209 0.042 0.161 0.085 Jiangsu -0.093 -0.309 Gansu Qinghai *-0.221 -0.630 ***-0.177 Yunnan -0.082 **-0.277 -0.330 -0.048 -0.344 -0.102 Guangdong -0.219 ***-0.159 -0.157 -0.604 -0.020 *-0.400 Sichuan -0.052 Northwest -0.119 -0. State Share 1994 *-0.303 -0.285 *-0.179 -0.575 **-0.Correlation Coefficients Across 13 Industrial Sectors.230 -0.697 Henan *-0.481 **-0.713 -0.097 -0.045 -0.464 **-0.409 *-0.167 -0.412 -0.694 ***-0.636 Shandong ***-0.224 North ***-0.070 -0.444 -0.062 Central -0.252 -0.707 ***-0.072 0.084 -0.380 -0.445 0.431 Tibet Shaanxi -0.228 -0.591 -0.375 *-0.315 Fujian -0.248 -0.242 0.415 *-0.644 -0.039 -0.295 Heilongjiang -0.357 -0.507 -0.304 -0.718 **-0.401 **-0.439 **-0.113 -0.218 -0.165 -0.124 Neimenggu -0.526 -0.288 *-0.647 -0.508 -0.079 -0.219 -0. An empty cell means that no data are available on the output either of the directly reporting industrial enterprises.353 -0.505 -0.334 -0.720 ***-0.336 -0.246 *-0.150 0.701 **-0.244 **-0.770 0. or of both.331 -0.251 Anhui -0.690 Shanxi -0.717 -0.642 ***-0.300 -0.714 East -0.149 -0.475 **-0.421 ***-0.234 -0.335 Liaoning ***-0.240 -0.337 Jilin -0.200 Hunan 0.041 -0.397 -0.451 -0.528 *-0.308 Beijing-T.023 -0. Total Linkage Vs.415 ***-0.195 -0.694 ***-0.030 -0.017 -0.317 Jiangxi -0.148 -0.261 0.112 -0.248 -0.027 -0.201 -0.025 -0.644 Hebei **-0.263 -0. *-0.429 -0.729 ***-0.335 **-0. or of the state-owned enterprises.690 ***-0.747 ***-0.379 **-0.151 -0.176 -0.139 -0.408 -0.161 0.768 **-0.170 -0.415 -0.080 0.022 -0.031 -0.689 ***-0.059 0.179 -0.441 Beijing Tianjin **-0.016 -0.405 -0.241 -0.373 -0.236 -0.086 -0.007 -0.533 **-0.143 -0.066 -0.124 Northeast -0.697 ***-0.488 -0.

277 -0.124 -0.481 -0.701 0.072 -0.445 0.200 -0.770 0.252 0.022 Shanghai -0.575 -0.218 -0.295 -0.326 -0.041 -0.342 -0.070 -0.544 **-0.702 0.441 -0.572 *-0.333 -0.248 -0.230 0. Total Linkage Vs.177 -0.334 -0.300 -0.635 -0.124 0.113 -0.476 -0.085 0.003 -0.099 -0.022 -0.151 -0.079 -0.097 -0.285 *-0.152 0.149 -0.242 0.415 Tianjin **-0.265 -0.401 **-0.159 2003 *-0.179 -0.254 0.375 *-0.408 Hainan 0.025 -0.045 -0.243 -0.701 **-0.290 -0.450 -0.035 0.315 -0.143 Yunnan -0.747 ***-0.261 0.559 Ningxia Xinjiang -0.105 -0. State Share 1994 2003 1994 Beijing *-0.112 -0.042 Hunan 0.380 -0.026 0.717 Hebei **-0.368 Jilin -0.694 ***-0.082 Neimenggu -0.176 Guangxi -0.475 *-0.379 Tibet Shaanxi -0.521 **-0.228 -0.636 -0.148 -0.123 -0.690 ***-0.024 -0.760 **-0.439 **-0.526 Liaoning ***-0.079 -0.221 -0.384 -0.Correlation Coefficients Across 13 Industrial Sectors.124 34 .335 **-0.303 -0.268 ***-0.248 ***-0.516 -0.102 -0.591 -0.309 -0.337 **-0.154 -0.031 0.086 0.066 -0.689 Henan *-0.139 Chongqing -0.192 -0.336 Zhejiang -0.001 0.080 -0.421 ***-0.143 -0.080 Hubei -0.195 -0.429 Jiangxi -0.020 **-0.459 -0.508 -0.310 0.148 -0.630 ***-0.644 -0.167 -0.313 0.218 -0.375 -0.507 Heilongjiang -0.161 Shandong ***-0.235 Guangdong -0.082 *-0.061 Total output-linkage 1997 2000 Total profit-linkage 1997 2000 **-0.084 -0.291 -0.492 -0.157 -0.697 ***-0.168 -0.027 -0.288 *-0.400 -0.231 -0.193 Jiangsu -0.505 -0.682 0.335 *-0.248 -0.317 -0.694 ***-0.713 -0.170 -0.139 -0.016 -0.150 0.021 Fujian -0.267 -0.165 -0.263 -0.093 -0.044 -0.242 Anhui -0.624 -0.373 -0.331 -0.415 -0.245 -0.251 -0.475 Guizhou -0.259 -0.175 -0.119 -0.236 **-0.768 Shanxi -0.094 Gansu Qinghai *-0.303 -0.464 **-0.161 0.040 -0.219 ***-0.039 -0.444 -0.647 -0.412 -0.431 -0.017 -0.697 ***-0.642 ***-0.353 Sichuan -0.

360 *-0.164 South -0.) Total output-linkage Total profit-linkage 1994 1997 2000 2003 1994 1997 2000 2003 Northeast -0.031 -0.179 -0.727 ***-0.644 -0.244 **-0.488 -0. Significance levels: * 10%.357 -0.410 -0.409 *-0.023 -0.048 -0.720 ***-0.595 The 13 sectors covered are the 13 industrial sectors (sectors 2 through 14) out of the total of 17 sectors.717 -0.414 -0.304 -0. *** 1%.265 -0.246 *-0.168 -0.405 -0. An empty cell means that no data are available on the output either of the directly reporting industrial enterprises.330 -0.320 0.402 **-0.533 **-0.201 -0.170 -0.159 -0.569 -0.451 -0.397 -0.174 -0.718 **-0.528 *-0.344 -0.209 -0.729 ***-0.209 0.059 0.030 -0. *-0.687 ***-0.415 ***-0.714 East -0. or of the state-owned enterprises.308 Beijing-T.219 -0.052 Northwest -0.707 ***-0.062 Central -0. 35 .234 -0. State Share (cont.007 -0.481 **-0. ** 5%.145 -0.690 ***-0.221 -0.604 -0.240 -0.Correlation Coefficients Across 13 Industrial Sectors.224 North ***-0. or of both.036 -0.241 -0.135 -0.556 Southwest -0. Total Linkage Vs.

4. and real GDP growth Regress provincial real GDP growth on provincial industry-wide total linkage coefficients (summed regional sectoral linkage coefficients weighted by provincial sectoral value added. Linkage. 13 industrial sectors) ownership variables interactions of total linkage and ownership variables and more 36 . Evidence c. state ownership.

-3 TOL and SOE share Allowing TOL coefficient to differ between three periods makes no difference.9. TPL instead of TOL (note: Central region) 5. periods) -> same negative impact on prov. +2 percentage points growth 7. Include SOE share 6. 064 unweighted mean across provinces. Include TPL (pure): TPL-SOE effect disappears! 37 .54.Summary of findings (Table) 1. Include SOE share 8. with as weight the sectoral share in provincial economy-wide industrial SOE value added SOE focus on high-linkage sectors beneficial for provincial economic growth: on average. growth in 2nd period Impact of TPL: columns 4-9 4. TPL-SOE For each province: sum (regional) TPL across industrial sectors. Mean TOL value ~ average SOE share (0.

02 (1. of SOE share TPL-SOE * dummy 1994-97 TPL-SOE * dummy 1997-00 TPL-SOE * dummy 2000-03 Central region dummy ****-7.76 61 ****-7.03 (63.37) **-4.16 (185.43) -0.53 ***5.92) -105.41) (1.10) ****147.51) -2.46) ****7.50 -1.59) ***1.85 (3.31) -1.59 (1.57 # of obs.98 (0.95 (26.58) -1.74 29.80 4.43 (1.08 (27.54) VII ****7. *** 1%. of TOL TPL * dummy 1994-97 TPL * dummy 1997-00 TPL * dummy 2000-03 Coeff.67 (29.43) ****-8.77 38 61 61 IV ****4.58) ***1.39 0.38) (62.62) (1.62) *6.69 (0.61 (0.51 -207.03) -0.59) (28. of var.43 (1.47 (3.47 ****-7.11 (0.52 (1.71) (170.36) R-square 0.30 (0.08) 0.26 (1.57) (1. ****5.49 (181.20 (147.06) 0.78) (129.54 (1.52 (28.71 (27.51 (0.81) ****112.22) ***81.64 0.26) -16.80) ****3.62) ****1.64 (25.33 (1.17) (1.49) ****97.02 (1.90) ****4.27) 0.67 0.47) -3.54 (1.75 (1.42) 42.00) -37.52 (1.47) V ****6.7 ****118.91) (1.71) ***3.78) *252.40) ****7.0) (28.13) 0.62) ***100.48 (1.40) (1.09 (64. of TPL SOE share * dummy 1994-97 SOE share * dummy 1997-00 SOE share * dummy 2000-03 Coeff. of var.04) ****118.15) ***190.44 (150.94) (1.52 **-3.07 (1.62 (0.76) (3.59 (1.10 (2.36 (1.28) VI ****5.46) 0.19 ***4.79 (1.86 ****4.41) ****2.13) (164.76 (0.43) 1.66 (0.72) I II III ****11.00 (0.55) (1.04 187.25) ****146.1%.44 (1.82 (1.63 0.88 (1.28 (30.42 (2.7 **64.82 ***168.72 (3.33) -1.13 272.95 (25. of var.55 (4.10 (1.40) 21.99) (1.38 (1.13) 0.96 (1.98) **-3.65 (1.33) (24.47) ****-8.85) -2. ** 5%.54) ****6.87) *-2.56 61 90. **** 0.84 20.34 (3.68) 0.60 -36.62 (3.63 65 ****4.7) (27.22 (26.Explaining Provincial Growth (dependent variable: provincial real GDP growth) Constant Dummy 1994-97 Dummy 2000-03 Industrial sector variables TOL Coeff.04 (1.76 (0.91) -10.49 (1.54) IX ****7.36 (27.54) 1.18 -0.54) 2.74) *-2.0 ****126.43) (1.50) ****1.58 ****10.04) (136.43) VIII ****4.04 (1.27 -64.20) ****121.54) (58.35) ****5.48) -0.83) ****3.05 ****12.88) ****3. 65 61 61 Significance levels: * 10%.36 (26.70 61 .6) (28.

and SOE value added) if include provincial TPL. SOE value added in this sector / prov. test) Exception: 2nd regr. SOE value added) -> positive. industrial value added has positive impact on growth share of SOEs in these 3 sectors has negative impact growth rate of SOE real value added in these 3 sectors has no impact (also if weight SOE growth rates by linkage coeff.: heteroscedasticity. significant in second period only significance vanishes if include controls (SOE share. sectoral SOE real growth rate * weight (weight: prov. it dominates all other results 39 . not corrected (robust standard errors would be smaller) Ramsey RESET tests with 3 fitted values (2nd to 4th power): no misspecification Use TPL of all 17 sectors (then can’t include ownership values): similar results Use TPL-SOE-growth: sum across sectors of: regional sectoral TPL * prov. White’s heterosc.Robustness Residuals ~N. TPL) Focus on 3 industrial sectors with highest TPL value: share of these 3 in prov. homoscedastic (5% sign. Jarque-Bera.

) Capital intensity Schooling Infrastructure (railway or highway network) State share in always high-linkage construction sector -> mostly insignificant (and some potentially endogenous). L -> economic growth) endogenous (profit opportunity -> economic growth -> K. core findings unchanged Include measures of capital and labor? Omit because they are intervening variables (profit opportunity -> K. L) or 40 . share of agriculture in GDP or empl.) Include additional variables: Development level (GDP per laborer.Robustness (cont.

some significance for INT (b) Via assets? SASAC task: protect and increase state assets -> All 8 regions show a positive correlation between the state share in sectoral industrial assets and TOL (5.) 5 regions for TPL -> Results are weaker if sectoral share in state assets is used -> Almost no correlation for LSD. National level (2008. GSD.) -> Correlation for share of state in all investment in a sector (10% sign.Discussion: Why does the Chinese state not strategically retain (or increase) state ownership in high linkage sectors when these have a positive impact on growth? (1) State is interested in making good use of linkage effects. 1% sign.) in only 2 of the 8 regions -> Similar findings for LSD. lack of data for 1994-2003): Relate state share in investment to regional TPL values. GSD. but not via SOEs? (a) Via current period investment? No provincial-level set of sectoral investment data by ownership. by sector -> No correlation for sectoral share in state investment (10% sign. INT 41 Conclude: some attention of state to linkage effects via assets? = cumulative past policies .

Tend to have high linkage. as well as in natural monopolies. --.Utilities are high TPL (in top 3 sectors) in 2 regions. --.) (2) Industrial policy (a) 10th Five-Year Plan 2001-05 distinguishes between 5 groups of sectors (i) Military industry remains overwhelmingly state-owned. (v) In other industries. Electric machinery and electronic communication equipment? --. and high technology industries. telecommunications.High TPL in one region. -> Not a compelling match. (ii) In public goods industries and services. machine building.Sector “others”? Not a high-linkage sector. (iv) The state should play a driving function in key high technology areas. --. (iii) SOEs should continue to hold a dominant position in industries of great importance for the “strength of the nation. the state should hold a controlling stake. the state is not assigned any specific role.” such as the petroleum. automobile. (b) 9th Five-Year Plan 1996-2000: no specific sectoral instructions for SOEs [next slide] 42 .Discussion (cont.

same holds for sectoral share in state value added // sectoral SOE profitability + no correlation of sectoral profitability (SOEs. profit plus taxes per unit of equity) --. INT.” including reorganization of SOEs to prevent the loss of state assets. … In the 1980s nearly all discussion was centered on reviving the profitability of SOEs. 2003: Positive correlation between state’s share in sectoral value added and profitability of all directly reporting industrial enterprises in a sector (profit per unit of equity.) 43 and (8 regions’) TPL. and a strengthening of enterprise management. = Focus on profitability? SOEs belong to one of four levels of government. LSD.) (b) 9th Five-Year Plan 1996-2000: no specific sectoral instructions for SOEs Focus on introduction of “modern enterprise system. GSD . all ent.Discussion (cont. 2) (2) Industrial policy (cont. Availability of input-output tables at local levels? Scope for input-output planning within the political system? Departmental interests Wong (2009): “official documents from the 1980s and early 1990s revealed no sustained discussion of how to identify areas or services where the government’s role could or should be reduced. each maximizing its own utility.” National data.

not output linkage (2) Distribution of TPL coefficients across sectors does not matter Not availability or unavailability of extreme profit-creating opportunities matters. less so in 1997-2000. growth via linkage effects --. and none in the third period (2000-03) Compare Hsieh and Klenow (2009): 1998-2005: SOEs have on ave. (3) The Chinese state does not strategically retain (or increase) state ownership in high linkage sectors and thereby promote economic growth (4) The SOE share in sectoral value added has a negative impact on economic growth in 1994-97. the faster its growth. Conclusions (1) The greater the degree of linkage in a Chinese province. but that high-linkage sectors are large. And it is profit linkage that matters.5.What if the state in its SOE policies had paid attention to linkage values? (6) Is the Party-state incapable of organizing an overarching development program? 44 . 40% lower within-industry TFP than private ent. Time trend of within-industry TFP dispersion: SOEs have significant positive effect in 1998 and 2001 but no longer in 2005 (5) Cannot justify state ownership as generator of econ.

Sign up to vote on this title
UsefulNot useful