Bank Loans, Self-Financing, and Grants in Chinese SOEs: Optimal Policy under Incomplete Information

Bank Loans, Self-Financing, and Grants in Chinese SOEs: Optimal Policy under Incomplete Information

JOURNAL OF COMPARATIVE ECONOMICS ARTICLE NO. 24, 140–160 (1997) JE961412 Bank Loans, Self-Financing, and Grants in Chinese SOEs: Optimal Policy und...

199KB Sizes 0 Downloads 48 Views

JOURNAL OF COMPARATIVE ECONOMICS ARTICLE NO.

24, 140–160 (1997)

JE961412

Bank Loans, Self-Financing, and Grants in Chinese SOEs: Optimal Policy under Incomplete Information1 YOUNG LEE University of Michigan, Ann Arbor, Michigan 48109 Received September 5, 1995; revised October 11, 1996

Lee, Young—Bank Loans, Self-Financing, and Grants in Chinese SOEs: Optimal Policy under Incomplete Information This paper offers a model of the allocation of fund in Chinese state-owned enterprises (SOE) and provides an empirical test of the theory using firm-level data. The paper explains why bank loans and grants coexist with self-financing, which SOEs take out loans, and why subsidies on loan interest payments exist. The model is based on heterogeneous SOEs, asymmetric information, sales taxes, and quota requirements. The results show that reforms of enterprise finance must come as a package, suggesting that the interlocking nature of reform measures should be considered in deciding the direction of further policy modification. J. Comp. Econom., April 1997, 24(2), pp. 140–160. University of Michigan, Ann Arbor, Michigan 48109. q 1997 Academic Press Journal of Economic Literature Classification Numbers: P21, P27, P31.

1. INTRODUCTION The purpose of this paper is to model how funds are allocated in Chinese state-owned enterprises (SOEs) and to test this model empirically using firmlevel panel data. I examine the allocation system of sources of finance in Chinese SOEs from the perspective of an optimal policy under incomplete information. The basic idea behind my analysis is that, under incomplete information, optimal policy will be a combination of direct restrictions on investment based on information available to the government and a voluntary adjustment mechanism allowing SOEs to adjust their volume of investment.

1

I am grateful to Roger Gordon for invaluable discussion and encouragement. I also thank the editor of the journal, an anonymous referee, Robert Dernberger, and Charles Brown for extremely helpful comments, and Carrie Cihak and Sarah Turner for editorial suggestions. Of course, I am solely responsible for any remaining errors. 0147-5967/97 $25.00 Copyright q 1997 by Academic Press All rights of reproduction in any form reserved.

AID

JCE 1412

/ 6w0c$$$$81

140

03-24-97 14:53:22

cea

BANK LOANS IN CHINESE SOEs

141

Voluntary adjustment mechanisms, that is, bank loans whose amount can be largely decided by SOEs, are needed to reduce the rents arising from asymmetric information. This dual track system or policy combination can be found in other areas, such as labor allocation, and it is one of the key features of the Chinese economic reform. The questions addressed in the paper are the following: which SOEs borrow money? Why do they borrow? Why do bank loans and grants coexist with self-financing? Why does the government subsidize interest payments? These questions are addressed by examining the incentives of the government and SOEs in a two-stage game.2 The paper is organized as follows. In Section 2, I examine the incentives of SOEs and the government. In Section 3, I test the hypotheses derived in Section 2. More specifically, in Section 3.A, I test whether SOEs taking out loans are systematically different in terms of marginal revenue product (MRP) of capital and total factor productivity (TFP) from those not taking out loans. In Section 3.B, I test some determinants of each source of finance and estimate the optimal loan interest rate. I conclude this paper by discussing the policy implications. 2. THE MODEL My theoretical analysis is based on a ranking of enterprises’ sources of finance in terms of opportunity cost. The opportunity cost to the firm of government grants is zero, that of self-financing is 1 / r, where r is the market interest rate, and that of bank loans is 1 / (1 0 t i)rb , where t i is the enterprise income tax rate, and rb is loan interest rate.3 The opportunity cost of bank loans to the firm is smaller than 1 / r because there exists a subsidy on interest payments and interest payments are tax deductible. Under this opportunity cost structure, SOEs first prefer government grants, then bank loans, and finally self-financing.4 This ranking of opportunity costs of loans and self-financing has already been noted in Hussain and Stern (1991).5 They argue that the allocation of funds in China provides an incentive for SOEs to substitute loans for self-financing. They also note that the effective loan

2

Lee (1993) emphasizes the need to take agency problem approaches in analyzing China’s enterprise reform. 3 In China, the credit market is not well-developed. However, given the possibility of depositing extra money in a bank or of lending it to other firms, self-financing clearly has an opportunity cost. The rates available for such lending are referred as the market interest rate in this paper for simplicity of exposition. 4 Opportunity costs described above ignore the possibility that the government might play the role of an owner of SOEs and care about their efficiency in addition to its revenue. However, adding this complexity does not change the ranking of sources of finance by opportunity cost. 5 I thank the anonymous referee for directing me to this literature.

AID

JCE 1412

/ 6w0c$$$$81

03-24-97 14:53:22

cea

142

YOUNG LEE

interest rate was negative for several years during the 1980’s as the result of high inflation and the tax deductibility of the repayment of principal. The opportunity cost to the government of each source of finance is 1 / r minus the opportunity cost to SOEs because the total opportunity cost for the government and SOEs is always equal to 1 / r. Therefore, the government ranks sources of finance in an order that is the reverse of the SOEs’ preferences. These rankings give us the basic behavioral rules for SOEs and the government. SOEs try to maximize government grants and minimize selffinancing. Knowing this, the government will impose a ceiling on grants and require SOEs to provide at least some minimum amount of self-financing. Only when the desired investment cannot be covered by the sum of grants and compulsory self-financing will SOEs consider taking out bank loans. There are four key assumptions. First, it is assumed that there exist two types of SOEs. One is with relatively high MRP of capital (Type A), and the other is with relatively low MRP of capital (Type B). Second, the government is assumed to be unable to distinguish Type A SOEs from Type B SOEs because of incomplete information about their initial capital stocks or production functions. Nevertheless, the model assumes that the government knows the distribution of types of SOEs. Third, the government is assumed to maximize a weighted sum of net total revenue and efficiency. It is natural to assume that the government cares about net total revenue because most government roles, such as the provision of public goods and services, need funding. Also, the government wants to improve efficiency because net total revenue will increase with the efficient operation of the economy. Last, quota requirements, the sales tax rate, and the income tax rate are assumed to be used for other policy considerations and hence are taken as exogenous in the model. Quota requirements are assumed to be exogenous because the government uses them to ensure a stable supply of key commodities. In China, taxes based on output quantity, such as the sales tax and the resource tax, have been in place since 1979, and they have varied substantially across firms. These taxes are taken as exogenously given because the main consideration in setting these tax rates was to offset the inequality arising from distorted prices (World Bank, 1990, p 29). Using a sales tax to correct price distortions and using other policy tools, such as a self-finance requirement and a loan interest rate, to achieve an investment level desired by the government is certainly a feasible policy combination. Furthermore, imposing different sales tax rates on firms to attempt more than offsetting distortions in prices would worsen the allocation of capital unless loan interest rates were firm specific. Applying firm-specific interest rates to offset difference in sales tax rates is not desirable because firms could use borrowed money for purposes other than investment. Optimal t i would be 1 in models, such as my model, where the incentive effect of t i is ignored. When the effect of t i on managers’ and workers’

AID

JCE 1412

/ 6w0c$$$$82

03-24-97 14:53:22

cea

BANK LOANS IN CHINESE SOEs

143

incentives to work is added to the model, an optimal t i less than 1 can exist. I assume that t i can be determined from this incentive consideration and I accordingly take it as exogenous. The first best solution, introducing loans at r and letting SOEs make investment decisions, is not an optimal policy even with complete information. Since self-financing is the cheapest source of finance for the government, the optimal policy with complete information would be to impose a self-financing requirement equal to the investment desired by the government, defined in Section 2.B by Eq. (7), and to set rb high enough so that no SOE borrows money. Grants can supplement self-financing for the SOEs that are losing money and whose own desired investment is artificially low as a result of government interventions such as a high sales tax rate, high quota requirements, or low government list prices. When the government does not have complete information about SOEs, inducing loans becomes a part of an optimal policy. With incomplete information, the government can be better off by allowing SOEs, or agents who have the information, to choose freely the amount of loans at a given rb . By doing so, the government can reduce the rents from asymmetric information by using loans as a revelation mechanism while imposing its preference on SOEs by setting the self-financing requirement optimally based on information available to the government. The optimal rb will be smaller than r under reasonable conditions because, otherwise, explicit and implicit sales taxes lower the effective rate of return and cause SOEs to underinvest. This intuition is captured in the following two-stage game.6 In the first stage, before borrowing is observed, the government sets a maximum for grants, IV g, and a minimum for self-financing, I s, for each SOE, and rb , which is applied to all SOEs. In the second stage, SOEs are allowed to choose freely the amount of bank loans under these restrictions.7 Once these rules of the game are fixed, the game can be solved by backward induction. The second stage of the game is analyzed in 2.A. The first stage of the game is analyzed in 2.B. 2.A. The Second Stage: The Incentives of SOEs To operationalize the analysis of SOEs’ incentives, the following additional assumptions are necessary. First, SOEs are assumed to maximize the present 6 In the pre-reform era, both types of SOEs applied for the largest possible government grants. Because the government could not distinguish one type of firm from the other, grants might end up being allocated equally across SOEs. If we further assume the total amount of government grants was equal to optimal aggregate investment, the actual amount of government grants would have been lower than optimal level for Type A SOEs and larger than optimal level for Type B SOEs. 7 In a multi-period context, the government would be able to exploit the information revealed from a firm’s borrowing decision by imposing a larger I s for SOEs which borrowed money in the second stage. This would in turn inhibit the SOEs from revealing the information. This ratchet effect would weaken the function of loans as a revelation mechanism.

AID

JCE 1412

/ 6w0c$$$$82

03-24-97 14:53:22

cea

144

YOUNG LEE

value of cash flow. Second, it is assumed that SOEs take as given the output quota, Y g, quota price, p g, I s, and IV g when deciding how much to invest and produce.8 Third, for simplicity, capital, K, is assumed to be the only input used in production. Y Å F(K),

F* ú 0,

F 9 õ 0.

(1)

Fourth, the depreciation rate is assumed to be zero. Under this assumption, capital after investment is simply the sum of initial capital and investment, K Å K0 / I b / I s / I g,

(2)

and SOEs can produce F(K) in every period after the investment is made. Last, quota requirements are assumed to depend on total output, Y g Å q(Y),

(3)

where q* ú 0. To save notation, denote (1 0 pg)q(Y), or the amount of implicit sales tax, by Q(Y). Now, the maximization problem SOEs face is V Å Max (1 0 t i) I b,I g,I s

H

J

(1 0 ts)[F(K) 0 Q(F(K))] rb b 0 I 0 Is r r

(4)

subject to Ib § 0, Ig £ IV g, and Is £ Is, where ts is the sales tax rate. For exposition, define the desired investment by SOEs, I* SOE, by F*(K0 / I* SOE) Å

rb . (1 0 t )(1 0 Q*) s

(5)

Note that I* SOE depends on rb , so the government can choose I* SOE by changing rb . s g The behavior of a SOE will depend on the sign of I* SOE 0 I 0 IV . If it is positive, the SOE will take out bank loans and become plan-unconstrained.9 Otherwise SOEs will not borrow money and invest only the sum of I s and s g IV g. Because I* can be directly or indirectly determined by the SOE, I , and IV government, the government can actually choose the volume of loans. 2.B. The First Stage: The Incentives of the Government To examine the incentives of the government, we need to specify its objectives, tools, and the information available to it. The government is assumed 8

As a result, I ignore the possibility that SOEs can devote resources such as money and management effort to getting favorable restrictions ex ante and the possibility that the restrictions can be evaded ex post. 9 Being plan-unconstrained means that SOEs participate in a market transaction. In the paper, being plan-unconstrained simple means that SOEs borrow money. For the general concept of plan-unconstrained, refer to Byrd (1989, pp. 180–181).

AID

JCE 1412

/ 6w0c$$$$82

03-24-97 14:53:22

cea

BANK LOANS IN CHINESE SOEs

145

to maximize a weighted sum of net total revenue and efficiency. Let u be a weight for net total revenue. Net total revenue is defined as the sum of the government’s share of SOE profit, sales tax revenue, and proceeds from reselling appropriations less extra costs associated with each source of finance. Since the Chinese government plays not only the role of tax collector but also the role of owner of SOEs, the government’s share of SOEs’ profit consists of income tax revenue and dividends, or t i 1 Profit plus a[(1 0 t i) 1 Profit 0 I s], where a is the government’s share of after-tax profit. To save notation, denote (t i / (1 0 t i)a) by T. Efficiency is represented by economic profit calculated at the true market prices of each source of finance. Note that true market prices of Ib, Is, and Ig are 1, because we take the present value of interest payments. As for the tools of government might use, IV g, I s, and rb are assumed to be available. It is also assumed that I s cannot be set higher than the after-tax current profit, CP, of an SOE, and that rb cannot be larger than r because SOEs can borrow money from a third party at r. I examine the incentives of the government under two alternative assumptions about its information about the type of SOE. First, I assume that the government has complete information about the type of SOE. Second, I assume that the government cannot distinguish one type from the other. Using subscripts to denote the type of SOE, the problem the government faces with complete information is Max

V

g

s g s IA ,IU A,IB,IU B ,rb

H F

Å u TA

G

S D

(1 0 t s)(YA 0 QA) rb b ts(YA 0 QA) QA r 0 rb b 0 IA / / 0 I A 0 aI sA 0 I gA r r r r r net total revenue from SOE A

/ TB

F

G

S D

(1 0 t s)(YB 0 QB) rb b ts(YB 0 QB) QB r 0 rb b 0 IB / / 0 I B 0 aI sB 0 I gB r r r r r net total revenue from SOE B

/ (1 0 u)

H

YA / YB 0 I bA 0 I sA 0 I gA 0 I bB 0 I sB 0 I gB r

J

efficiency

subject to KA Å K0,A / I bA / I sA / I gA, KB Å K0,B / I bB / I sB / I gB,

AID

JCE 1412

/ 6w0c$$$$82

03-24-97 14:53:22

cea

J

146

YOUNG LEE

I sA £ CPA , I sB £ CPB , I sA Å I sA, I sB Å I sB, I gA Å IU gA, I gB Å IU gB, s g I bA Å max {I* SOE,A 0 I A 0 IU A, 0}, s g and I bB Å max{I* SOE,B 0 I B 0 IU B, 0}.

(6)

For exposition, define the desired investment by the government, I* gov, by F*(K0 / I* gov) Å

[ua / (1 0 u)]r õ r. u{a / (1 0 a)[t i(1 0 ts)(1 0 Q*) / ts(1 0 Q*) / Q*]} / (1 0 u)

(7)

Note that I* gov is not a variable that the government can choose. In fact, I* gov is optimal investment by the government when investment can be fully financed through I s. I* gov turns out to be larger than the socially optimal investment level, implying that the government tends to overinvest when the investment can be fully financed through I s. Complete information. The solution depends on whether the CP constraint Vg is binding. When it is not binding, optimal I s is equal to I* gov, optimal I is 0, s and optimal I* SOE is smaller than I , which implies no borrowing occurs. When the CP constraint is binding, optimal I s is equal to CP and SOEs will be induced to borrow money if ÌV ÌI* SOE

Z

˙ at I SOE ÅCP

is positive. Whether the CP constraint is binding or not, optimal IV g is zero. However, it can be shown that when heterogeneous sets of SOEs exist, grants will be given to SOEs with large ts or Q* and a binding CP constraint under certain conditions. Incomplete information about the type of SOE. Suppose that SOEs are identical except that Type A SOEs have production function F(K) and Type B SOEs have production function G(K) such that F*(K) ú G*(K) for given K. For simplicity, also suppose that the CP constraint is not binding. With incomplete information about the type of SOE, the optimal policy will be to

AID

JCE 1412

/ 6w0c$$$$82

03-24-97 14:53:22

cea

BANK LOANS IN CHINESE SOEs

147

impose one restriction on all SOEs. Optimal I s lies between I* gov,B and I* , and Type A SOEs will be induced to borrow money. This implies that gov,A the MRP of capital of firms using loans would be larger than that of firms not using loans. This is one of the key hypotheses tested in the empirical analysis. This solution suggests incomplete information as the reason the government cannot fully substitute I s for I b. Optimal rb , which is the flip side of optimal I* SOE, is determined by Eq. (8) derived from rearranging Eq. (5) and the first-order conditions: optimal rb Å (1 0 ts)(1 0 Q*)r 0 u(1 0 T)(1 0 ts)(1 0 Q*)I b

Ìrb ÌI* SOE

ú (1 0 t s)(1 0 Q*)r.

(8)

Optimal rb turns out to be larger than the socially optimal loan interest rate, (1 0 ts)(1 0 Q*)r, which implies Type A SOEs underinvest. Equation (8) can be further rearranged under an additional assumption of F(K) Å AK a to give optimal rb Å

(1 0 ts)(1 0 Q*) r. 1 0 u(1 0 T)(1 0 ts)(1 0 Q*)(1 0 a)(I b/K)

(9)

In Section 3.B, Eq. (9) is used to estimate optimal rb in China. The main implications of the analysis of the incentives of the government are as follows. (1) Even though I s is cheaper than I b from the government’s perspective, the government cannot fully substitute I s for I b without complete information about the type of SOE. This implies that bank loans will be used as a revelation mechanism, and SOEs with higher MRPs of capital will be induced to take out loans at a given rb . The amount of total investment of plan-unconstrained SOEs depends negatively on t s, Q*, and rb (see Eq. (5)). (2) Optimal I s depends positively on observable proxies for I* gov. This implies that I s is positively related to before-tax profit and negatively related to the capital–labor ratio. I s also depends positively on t i, t s, and Q* (see Eq. (7)). (3) Optimal IV g depends positively on t s and Q*, and negatively on CP. (4) The optimal rb will be smaller than r under reasonable conditions because otherwise explicit and implicit sales taxes lower the effective rate of return and cause SOEs to underinvest. 3. EMPIRICAL ANALYSIS The data come from surveys conducted by the Institute of Economics of the Chinese Academy of Social Sciences (CASS). Annual data for 1980–

AID

JCE 1412

/ 6w0c$$$$82

03-24-97 14:53:22

cea

148

YOUNG LEE TABLE 1 DISTRIBUTION

No. of SOEs All 5 major industries a

OF

SOEs

BY

GROUP

Group X

Group Y

Group N/A

Total

190 (127)a 112 (72)

342 (140) 187 (79)

237 138

769 437

Numbers in parentheses are the number of SOEs used in the estimation of MRP.

1989 for 769 SOEs in four provinces covering major industry groups give details of SOE production, costs, and the nature of the relationship between SOEs and the government. For a detailed description of the data refer to Grove et al. (1994, pp. 189–190). 3.A. The Determinants and Effects of Borrowing I divide the 769 Chinese SOEs into two groups according to when they used bank loans. The first group, Group X, consists of SOEs that have been borrowing money since 1980 or 1981. Group X can be treated as the group of plan-unconstrained firms. The second group, Group Y, consists of SOEs that have not borrowed at all during the sample period. Group Y can be treated as the group of plan-constrained SOEs. There are some SOEs that could not be grouped into either X or Y either because they began to borrow from banks sometime between 1982 and 1989 or because their financing behavior is not clear. Table 1 reports the distribution of SOEs by group. To estimate the MRP of capital and TFP for each group, I follow the method used in Gordon and Li (1994). The equation I estimate for each group is DYts DKts DLts DMts Å b / bK / bL / bM / e, Ys Ys Ys Ys

(10)

where Y is real output, K is net real fixed assets, L is the number of workers, M is real intermediate inputs, and D indicates change in the corresponding variables from year s to year t.10 For the construction of data, refer to Appendix B. I chose to focus on the change over 4 years, rather than the year-by-year changes, to reduce any bias arising from the potentially complicated timing 10 Adding other inputs into the theoretical model would not have changed the main result that firms borrowing money have a higher MRP of capital than firms not borrowing money, which is the hypothesis tested in this section. Even when labor and real material are added to the model in the foregoing section, it is still true that firms with a higher MRP of capital for given capital, labor, and real material take out loans.

AID

JCE 1412

/ 6w0c$$$$83

03-24-97 14:53:22

cea

BANK LOANS IN CHINESE SOEs

149

TABLE 2 ESTIMATES

OF THE

BEFORE-TAX MRP

OF

CAPITAL IGNORING DEPRECIATION

1984–1980 1985–1981 1986–1982 1987–1983 1988–1984 1989–1985 1989–1980 All Group X Group Y Five industries Group X Group Y

0.18 (0.12)a 0.05 (0.02)

0.30 (0.11) 0.03 (0.01)

0.20 (0.12) 00.01 (0.02)

0.16 (0.04) 0.08 (0.05)

0.10 (0.13) 0.08 (0.05)

0.17 (0.03) 0.14 (0.06)

0.15 (0.05) 0.06 (0.02)

0.66 (0.27) 0.05 (0.05)

0.70 (0.16) 0.02 (0.00)

0.43 (0.15) 00.06 (0.03)

0.30 (0.04) 0.14 (0.05)

0.08 (0.05) 0.10 (0.05)

0.25 (0.08) 0.18 (0.10)

0.13 (0.04) 0.02 (0.05)

a Numbers in parentheses are White’s heteroskedasticity-consistent estimators for standard errors.

of the impact of additional inputs on output (Gordon and Li, 1994). Two sample sets are used, all industries and five industries: electronics, chemicals, clothing, construction, and machinery.11 The regression results, shown in Table 2, indicate that the MRP of capital for Group X is larger than that for Group Y, and the difference of MRPs between groups decreases over time (Fig. 1), which is consistent with the predictions of the model.12 Estimates of the MRP of capital with an arbitrary 5% depreciation show a similar pattern but the difference is smaller. In calculating TFP, I follow the method used in Gordon and Li (1994). I chose 1980 as the base year and 1987 as t Å 1 because the years 1988 and 1989 were a recession period in China. Annual TFP growth for the period between 1980 and 1989 is estimated to be smaller than that for the period between 1980 and 1987 and in the range 0.5–0.9%. Observations were arranged such that observations belonging to Group X have the index n Å 1, . . . , i, and observations belonging to Group Y have the index n Å i / 1, . . . , i / j. TFP of Group G at time t is defined by TFPG,t Å

( Yt , ( b Kt / ( bLGLt / ( bM G Mt K G

(11)

11 The reasons these five industries are used as a sample are twofold. First, they are the five industries with the most SOEs, excluding miscellaneous categories. Second, the same industries are used in Grove et al. (1994). 12 When we estimate the MRP for each industry, the MRP for Group X turns out to be larger than that for Group Y in four out of five industries.

AID

JCE 1412

/ 6w0c$$$$83

03-24-97 14:53:22

cea

150

YOUNG LEE

FIG. 1. Before-tax MRP of capital, five industries. Doted line is one standard-error range.

where t Å 0, 1; G Å X, Y, or X < Y; summations are over the corresponding range for each Group G; and b values come from Eq. (10) for the corresponding Group G. The weighted average TFP at time t is defined by TFPW,t Å

i/j Yt (nÅ1 . i K i L (nÅ1 bXKt / (nÅ1 bXLt / (inÅ1 bM X Mt i/j i/j i/j / (nÅi/1 bKYKt / (nÅi/1 bLYLt / (nÅi/1 bM Y Mt

(12)

The TFP growth using TFPX
AID

JCE 1412

/ 6w0c$$$$83

03-24-97 14:53:22

cea

BANK LOANS IN CHINESE SOEs

151

gains are due to allocative efficiency gains. When allocative efficiency gains for each input are calculated using the method in Appendix A, allocative efficiency gains are estimated to be 0.13, 0.03, and 0.00 for capital, labor, and real materials, respectively. The introduction of bank loans might explain why most of the allocative efficiency gain comes from capital reallocation. 3.B. Some Determinants of Sources of Finance and Optimal Loan Interest Rate To test some determinants of self-financing, grants, and loans, I estimate the following equation using the ordinary least square (OLS) method for two samples, one with all observations, the other with observations with positive value of the dependent variable:13 Its, I gt , or I bt Å a0 / a1Pt01 / a2Kt01 / a3(Kt01/Lt01) / a4tit01 / a5tst01 / a6(qt01/TSt01) / a7TD / a8IND / a9LOC / et.

(13)

P is profit, K/L is the capital–labor ratio, q/TS is the ratio of quota sales to total sales of the major output, TD are time dummies, IND are industry dummies, and LOC are location dummies. Two profit measures are used in regressions, before-tax and after-tax profit. The summary statistics of I s, I b, and I g are reported in Table 3. Panel A shows that I s has been the primary method of financing and that a mix of I s and I b and a mix of I s and I g have been widely used. Panel B shows that the average loan has more than doubled over the sample period, while the average amount of self-financing and grants have not changed much. Table 3 shows that loans have been substituted for grants since the early 1980’s. The results of the OLS estimation with before-tax profit are reported in Table 4.14 As conjectured, I s is positively related to before-tax profits and to t i, and negatively to K/L. Second, I g is negatively associated with P and positively with t s. This result is consistent with the conjecture that grants tend to be given to SOEs with losses that are the result of adverse government interventions. Third, I b is positively associated with P and negatively with q/TS. The adjusted R2 of the regression for bank loans turns out to be relatively small, implying the potential omission of important explanatory variables that are observable to the SOEs but not to the government and to researchers. One variable left out is the true MRP of capital, as suggested in Section 3.A. 13

Because many observations are censored, especially those for loans and grants, the Tobit model is appropriate. Regression results using the Tobit model, which are not reported here, are very similar to the OLS results for the sample with all observation. 14 Regression results using after-tax profit are very similar except that the estimated coefficient of P changes to negative in self-financing regression. This might indicate that the income tax rate is more strongly associated with before-tax profit than Is is.

AID

JCE 1412

/ 6w0c$$$$83

03-24-97 14:53:22

cea

AID

JCE 1412

/ 6w0c$$1412

03-24-97 14:53:22

cea

341 358 385 392 417 436 435 402 405

Averagea

Self-financing

64 75 194 200

1289 1243 1238 1231 1231 1323 1382 1309 1539

Standard deviation

I s and I b

Unit is 10,000 Yuan at a 1989 value.

518 524 543 571 584 598 600 616 612

1981 1982 1983 1984 1985 1986 1987 1988 1989 a

No. of SOE with I s ú 0

197 207 213 233

1981 1982 1988 1989

Year

I s Only

Year

PATTERN

108 114 120 105

I s, I b, and I g

203 221 241 255 306 330 331 349 343

No. of SOE with I b ú 0

204 211 228 301 334 412 568 532 419

Averagea

Bank loans

Average of each source of finance

149 128 89 74

421 334 352 612 662 782 1639 1824 1834

16 18 24 25

302 290 292 287 296 290 255 232 203

No. of SOE with I g ú 0

I b Only

GRANTS

Standard deviation

AND

Number of SOEs

Method of financing

SELF-FINANCING, BANK LOANS,

I s and I g

OF

TABLE 3

425 415 427 517 515 467 487 452 450

Averagea

Grants

31 34 12 11

I g Only

1774 1502 1618 2158 2261 2090 2008 1746 1546

Standard deviation

15 14 10 12

I b and I g

152 YOUNG LEE

AID

JCE 1412

/ 6w0c$$1412

03-24-97 14:53:22

cea

233.93** (95.22) 00.14** (0.03) 0.15** (0.00) 23.02 (14.07) 37.60 (43.57) 337.80 (208.07) 07.72 (50.86) 2198 0.54

For all 449.32** (144.02) 0.22** (0.05) 0.15** (0.01) 59.16** (19.51) 23.11 (58.14) 129.13 (236.27) 047.18 (71.86) 1201 0.54

For I b ú 0 89.57** (42.70) 0.14** (0.01) 0.05** (0.00) 041.37** (5.89) 42.85** (18.84) 80.13 (95.37) 31.51 (22.59) 2392 0.84

For all 95.94** (44.07) 0.12** (0.01) 0.05** (0.00) 043.88** (5.92) 36.37* (18.99) 62.20 (94.39) 32.01 (22.94) 2178 0.86

For I s ú 0

Self-financing

121.10 (76.11) 00.31** (0.02) 0.10** (0.00) 13.45 (10.62) 03.12 (35.26) 312.57* (170.01) 1.22 (40.78) 2288 0.53

For all

128.84 (141.09) 00.32** (0.03) 0.10** (0.00) 71.16** (17.59) 011.01 (53.06) 1108.40* (588.70) 22.20 (75.92) 1192 0.56

For I g ú 0

Grants

EQ. (13) (DEPENDENT VARIABLE Å TOTAL INVESTMENT, SELF-FINANCING, GRANTS,

TABLE 4

Year, industry, and location dummies are included in regression but not reported. * Statistical significance at the 90% level. ** Statistical significance at the 95% level.

a

No. of observation Adjusted R 2

qt01/TSt01

tst01

tit01

Kt01/Lt01

Kt01

Pt01

Intercept

FOR

Total investment

REGRESSION RESULTS

a

For all

179.38** (75.20) 0.20** (0.03) 0.03** (0.00) 01.68 (10.24) 019.17 (30.45) 017.71 (124.74) 0110.80** (37.30) 1239 0.25

For I b ú 0

Bank loans

BANK LOANS)

50.21 (42.46) 0.13** (0.01) 0.01** (0.00) 17.05** (6.25) 08.06 (19.62) 03.42 (94.56) 058.85** (22.70) 2275 0.18

OR

BANK LOANS IN CHINESE SOEs 153

154

YOUNG LEE TABLE 5 OUTPUT SHARING RULE

Intercept Change in output No. of observation Adjusted R 2 a

All

Group X

Without grants

With grants

06850 (2441)a 0.82 (0.007) 394 0.968

01892 (2878) 0.38 (0.036) 120 0.488

0777 (3836) 0.68 (0.055) 123 0.555

08484 (3105) 0.82 (0.008) 271 0.975

Numbers in parentheses are standard errors.

Note that the smaller R2 for the loan equation is consistent with the model, which argues that loans work as a revelation mechanism and, hence, depend on several variables that are not observable to the government. The estimation results should be interpreted with care because t i, ts, and q/TS could be optimally chosen by the government; hence endogeneity among the explanatory variables is a potential problem. Optimal loan interest rate. To estimate the optimal rb , we need to know the value of variables in the right-hand side of Eq. (9). For ts and Ib/K, I use means for plan-unconstrained SOEs, 0.08 and 0.1, respectively. I assume that u, a, and T are 0.5, 0.3, and 0.6, respectively. I estimate Q* in the following section, and an estimate of Q* for plan-unconstrained SOEs, 0.023, is used for calculating optimal rb . Using all these values, the optimal loan interest rate is estimated to be around 90% of r. This implies that the government should give approximately a 10% subsidy on interest payments. If rb is set much smaller than the optimal level, SOEs will overinvest. To get an estimate for Q*, I estimate Dq Å g0 / g1DY / e,

(14)

where Dq is change in the size of the quota sale of the major output from 1980 to 1989, and DY is change in the size of the total sale of the major output from 1980 to 1989. Four samples are used, a sample with all SOEs, Group X, a sample with SOEs that have never used grants, and a sample with SOEs that have received grants at least once. Regression results show the existence of an output sharing rule (Table 5) between the SOEs and the government. Estimates of Q* for Group X, for SOEs without grants, and for SOEs with grants turn out to be 0.023, 0.041, and 0.043, respectively, when the mean value of pg, 0.94, was used in the calculation. The regression results of Eq. (14) are consistent with the conjecture that SOEs receiving grants are more likely to sell a larger share of their output to the government.

AID

JCE 1412

/ 6w0c$$$$83

03-24-97 14:53:22

cea

BANK LOANS IN CHINESE SOEs

155

4. CONCLUSION By examining the incentives of the government and of SOEs, this paper demonstrates that a mix of sources of finance and a subsidy on interest payments can be an optimal policy under certain conditions. With complete information about the type of SOE, I b is dominated by I s. However, with incomplete information, inducing I b becomes a part of an optimal policy. I g is used for money-losing SOEs suffering from adverse government interventions such as high quota requirements. The empirical part of the paper suggests that such allocation mechanisms of financing sources exist in China. The MRP of capital of SOEs borrowing money is larger than that of SOEs not borrowing, and there is an allocative efficiency gain from capital reallocation. These findings suggest that bank loans worked as a revelation mechanism and improved the efficiency of capital allocation in the 1980’s. The empirical analysis also indicates that TFP has grown annually at 1.9–2.4% between 1980 and 1987. This paper has the following policy implications. First, reforms should be considered as a package. Isolating one reform measure from others and evaluating its effects separately could lead to misleading results. For example, one might focus on grants and find that SOEs using grants are losing money. Based on this observation, one might argue that the government should abolish the grants system. The above argument is incorrect because grants might have their own rationale, as I have suggested in the paper.15 Second, to improve the allocative efficiency of capital, the incentives of the government and SOEs to overinvest should be reduced. To lessen the government’s incentive to overinvest, t i, t s, and Q* should be lowered. This is possible when other tax sources, such as personal income tax, are developed. Making I s tax-deductible could be a short-run solution to reduce the government’s incentive to overinvest. SOEs tend to overinvest when rb is not large enough and the soft-budget constraint prevails. The overinvestment due to the soft-budget constraint could be reduced by raising rb and hardening the budget, requiring a series of related reforms. APPENDIX A Simple Decomposition of Total Efficiency Gains into Allocative Efficiency Gains and Technical Efficiency Gains A simple methodology to decompose total efficiency gains into allocative and technical efficiency gains is based on the following observation. The 15 The debt crisis of Chinese SOEs in the 1990’s may be related to naive grants-to-loans conversion. If grants have been a part of optimal government policy and the government can shift a part of the cost of bad loans to banks, the government will convert grants to loans even when the MRP of capital is smaller than the loan interest rate. If banks are forced by the government to lend money to SOEs with a low MRP of capital, it is not surprising to see a debt

AID

JCE 1412

/ 6w0c$$$$83

03-24-97 14:53:22

cea

156

YOUNG LEE TABLE 6 SIMPLE ILLUSTRATION

OF

Group X

Case I Case II

PROPERTY

OF

TWO TFP MEASURES

Group Y

TFPW

TFPXUY

K0

K1

Y0

Y1

bX

K0

K1

Y0

Y1

bY

tÅ0

tÅ1

bX
tÅ0

tÅ1

1 1

2 3

2 2

4 6

2 2

1 1

2 1

1 1

2 1

1 1

1 1

1 1

1.5 2

1 3/4

1 7/6

unweighted TFP growth calculated for a sample consisting of heterogeneous groups includes both technical efficiency gains and allocative efficiency gains.16 On the other hand, the weighted TFP growth calculated using each group’s MRPs does not include allocative efficiency gains. Therefore, by simply taking the difference of these two TFP measures we can distinguish allocative efficiency gains from technical efficiency gains. The methodology introduced here can be applied to the case where we have a strong prior that each group is a set of homogeneous firms, yet the groups differ. One good example is grouping firms by industry. I claim that TFPX
S

D

TFPX
(15)

crisis. Other explanations for the debt crisis are: (1) the deterioration of profit due to competition and using loans for covering losses instead of making investment; and (2) a low real interest rate for loans and soft-budget constraint. 16 If a group is not a set of homogeneous firms and allocative efficiency gains occur from reallocation of resources between firms in this subgroup, these allocative efficiency gains appear as technical efficiency gains.

AID

JCE 1412

/ 6w0c$$$$83

03-24-97 14:53:22

cea

BANK LOANS IN CHINESE SOEs

157

Rt Å wK,trK,t / wL,trL,t / wM,trM,t,

(16)

where

wZ,t :Å

(

i/j nÅ1

b

K X
i/j (nÅ1 bZX
(17)

for Z Å K, L, M, and rZ,t :Å

i/j bZYZt (inÅ1 bZXZt / (nÅi/1 i/j Z (nÅ1 bX
(18)

for Z Å K, L, M. We can interpret Rt as the index for allocative efficiency; accordingly the growth rate of Rt is the allocative efficiency gain. Equation (15) clearly shows that the growth rate of TFPX
17 An alternative definition for allocative efficiency gain for each input is (wZ,1rZ,1 0 wZ,0rZ,0)/ wZ,0rZ,0. One problem with this alternative definition is that this index changes when the relative importance of one input to total inputs, w, changes. Actually the index decreases if the amounts of other inputs increase even though the allocation of the input does not change. This is not a desirable property.

AID

JCE 1412

/ 6w0c$$$$83

03-24-97 14:53:22

cea

158

YOUNG LEE

Labor The annual average number of employees is used without any modification. Real Materials, Fuel, and Power Inputs Inflation rates for raw materials are constructed by using information in the survey. In the survey, SOEs are asked to report detailed information about their two major raw material inputs. They report the quantities of raw materials needed to produce one unit of output, the government list price, and the market price of raw materials and the quantities of each raw materials input bought at government list prices and at market prices. Using this information, I construct an inflation rate for each SOE. For the SOEs for which I do not have enough information to construct an inflation rate, I use the industry average inflation rate. Fuel and power inputs are discounted by the inflation rates for fuel prices reported in State Statistical Bureau (1991, p. 209). Real Capital Input Measuring real capital input entails an inevitably complicated process, implying that the constructed variable might be less reliable. To begin, information about inflation rates for capital prices is scanty. In the survey, managers of SOEs are asked to quote price changes between 1980 and 1984 if they bought the same machinery in 1980 and 1984. For the years after 1984 they are asked to quote yearly price changes if they bought the same machinery for two years in a row. I can only construct one inflation rate for all industries because the amount of information available is not sufficient to construct an inflation rate for each industry. Moreover, I constructed inflation rates for each of the years between 1981 and 1984 by assuming that machinery prices increased at a constant rate during this period. As a part of capital input, I also include expenditure on structures. Real expenditure on structures is calculated using the inflation rate for building materials. Another complication associated with measuring real capital input is how to measure depreciation.18 I tried to use the information available in the survey to construct depreciation only to find that the depreciation figures reported in the survey seem very arbitrary. I decided to ignore depreciation figures reported in the data set. Since ignoring depreciation could bias the estimate of the MRP of capital, I construct another real capital series by assuming an

18 Casual evidence suggests that the effective lifetime of capital is much longer in China than in the United States. Gordon and Li’s estimate for the implicit depreciation rate is only 0.034% (Gordon and Li, 1994).

AID

JCE 1412

/ 6w0c$$$$84

03-24-97 14:53:22

cea

BANK LOANS IN CHINESE SOEs

159

arbitrary 5% depreciation for all SOEs.19 Constructed real capital series show that the average capital–labor ratio for Group X has been growing faster and converging with that of Group Y. APPENDIX C: NOMENCLATURE A a CP F( ) G( ) Ib Ig Is IU g Is I* SOEs I* gov IND K K0 K/L LOC L M pg q(Y) q* Q(Y) Q*

a coefficient in production function the elasticity of output with respect to capital (F(K) Å AK a) after-tax current profit production function production function amount of bank loans amount of government grants amount of self-financing the maximum amount of government grants the minimum amount of self-financing the desired investment by SOEs the desired investment by the government industry dummies capital initial capital stock the capital–labor ratio location dummies labor intermediate inputs quota price quota (ÅY g) the derivative of q(Y) (1 0 pg)q(Y), implicit sales tax through quota requirement (1 0 pg)q*(Y), the derivative of implicit sales tax rate with respect to Y q/TS the ratio of quota sale to total sale of major output R the index for total allocative efficiency rb interest rate on loans 19

Ignoring depreciation could underestimate the MRP for SOEs with large capital stock which are more likely to belong to Group Y. The reason is that a large capital stock means a large portion of investment is used for the replacement of old capital, hence investment does not necessarily result in an increase in output. To test this possibility, I ran experimental regressions assuming a 5% depreciation rate for all existing capital. Introducing an arbitrary 5% depreciation rate for all SOEs could correct the bias pointed out above, however, it introduces another bias due to the fact that the amount of the depreciation depends on the unreliable initial capital stock figures.

AID

JCE 1412

/ 6w0c$$$$84

03-24-97 14:53:22

cea

160

r T TD Y Yg a b g D e u P ti ts w rZ,t

YOUNG LEE

the market interest rate the government share of SOE’s profit (Åt i / (1 0 t i)a) time dummies output output sold to the government the government share of SOEs’ after-tax profit the MRP of corresponding inputs a coefficient in output sharing rule regression change in corresponding variables error term the weight to net total revenue in the government’s objective function profit enterprises income tax rate sales tax rate the relative importance of one input to total input the allocative efficiency index for each input REFERENCES

Byrd, William, A., ‘‘Plan and Market in the Chinese Economy: A Simple General Equilibrium Model.’’ J. Comp. Econom. 13, 2:177–204, June 1989. Gordon, Roger, and Li, Wei, ‘‘The Change in Productivity of Chinese State Enterprises, 1983– 1987.’’ University of Michigan, mimeo, 1994. Groves, Theodore, Hong, Yongmiao, Mcmillan, John, and Naughton, Barry, ‘‘Autonomy and Incentives in the Chinese State Enterprises.’’ Quart. J. Econom. 109, 1:1–27, Feb. 1994. Hsiao, Katherine, Money and Banking in the Chinese Mainland. Taipei, Taiwan: Chung-Hua Institution for Economic Research, 1984. Jefferson, Gary, Rawski, Thomas, and Zheng, Yuxin, ‘‘Growth, Efficiency and Convergence in China’s State and Collective Industry.’’ Econom. Dev. Cultural Change 40, 2:239–266, Jan. 1992. Hussain, Athar and Stern, Nicholas, ‘‘Economic Reform in China.’’ Econom. Policy 12, 141– 178, April 1991. Lee, Keun, ‘‘Property Rights and the Agency Problem in China’s Enterprise Reform.’’ Cambridge J. Econom. 17, 2:179–194, June 1993. State Statistical Bureau, Statistical Yearbook of China 1991. Beijing: Zhongguo Tonggi, 1991. White, Halbert, ‘‘A Heteroskedasticity-Consistent Covariance Matrix Estimator and A Direct Test For Heteroskedasticity.’’ Econometrica 48, 4:817–838, May 1980. Wolken, Lawrence, C., ‘‘The Restructuring of China’s Banking System Under the Economic Reforms 1979–1989.’’ Columbia J. World Bus. 25, 1 & 2:53–63, Spring/Summer 1990. World Bank, China: Revenue Mobilization and Tax Policy. Washington, DC: World Bank, 1990.

AID

JCE 1412

/ 6w0c$$$$84

03-24-97 14:53:22

cea