Who Gets Credit? - World Bank Group

13 downloads 87 Views 105KB Size Report
in Allocating Credit to Chinese SOEs. Under the best of circumstances, financial systems find it difficult to mobilize savings to the most productive endeavors and ...
Who Gets Credit? The Behavior of Bureaucrats and State Banks in Allocating Credit to Chinese SOEs Robert Cull and Lixin Colin Xu* The World Bank

Robert Cull, DECRG, World Bank 1818 H Street, N.W., Washington, DC 20433 Email: [email protected] Phone: 202-473-6365

L. Colin Xu, DECRG, World Bank 1818 H Street, N.W., Washington, DC 20433 Email: [email protected] Phone: 202-473-4664

Fax: 202-522-1155

Fax: 202-522-1155 Abstract

We explain variation in the sources of finance for a sample of Chinese state-owned enterprises from 1980-95 using their general characteristics (including sector, physical capital, number of employees, and lagged productivity), the level of government responsible for oversight, and their reform experience. Direct government transfers were negatively associated with past productivity throughout much of the period. Despite institutional arrangements that severely impeded effective financial intermediation, bank finance was positively linked to both past productivity and some types of reform. In particular, reforms that offered managers additional discretion and those that enabled managers to self-select and thus expose themselves (and their employees) to greater risk, were positively associated with acquiring bank finance.

Keywords: China, SOE Reform, Signaling, Banking, Financial Policy. JEL Classifications: G2, P2, D8, L2.

*

This research was partially funded by the World Bank’s Seed Fund for the Study of Transition Economies. We thank the Economics Institute of the Chinese Academy of Social Sciences for providing the data set. The survey was carried out by the Institute of Economics under the Chinese Academy of Social Sciences in collaboration with economists from the University of Michigan, UCSD and Oxford University. We acknowledge the financial support of the Ford Foundation in developing this data set. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the World Bank, its Executive Directors, or the countries they represent.

Who Gets Credit? The Behavior of Bureaucrats and State Banks in Allocating Credit to Chinese SOEs Under the best of circumstances, financial systems find it difficult to mobilize savings to the most productive endeavors and to monitor those firms and individuals that receive funds. The well known Greenwald and Stiglitz (1986) theorems demonstrate that imperfect information and the associated incompleteness of markets ensure that financial markets are, in general, not constrained Pareto efficient. In China, moreover, resolving such informational asymmetries may not be an important priority for the formal financial system. Bank loans and direct government transfers are routinely used as tools to accomplish political and social objectives. For banks, the results have been catastrophic – twenty percent of all bank loans are now reported as nonperforming.1 That figure, which may be a conservative one, implies $200 billion in losses, equivalent to twenty percent of annual GDP. If the estimate is correct, the country’s banks have negative overall net worth. A basic purpose of this paper is to describe which state-owned enterprises (SOEs) received the bulk of government transfers and bank loans from 1980-95. We focus on SOEs because of their pronounced impact on industrial output during the period. Although SOE share of total industrial output declined from 77.6 percent in 1978 to 28.8 percent in 1996, those enterprises still employed 57.4 percent of urban workers and possessed 52.2 percent of industrial fixed assets at the end of the period.2 Our strategy is to explain variation in SOEs’sources of finance using their general characteristics (including sector, physical capital, number of employees, and lagged productivity), geographic location, the level of government responsible for oversight (city, county, province, or central), and reform experience. Our approach enables us to answer such basic questions as whether direct government transfers were, in fact, used to sustain the least productive SOEs, and whether close ties to the central government (as opposed to provincial, county, or city governments) meant increased access to finance.

1

The estimate of non-performing loans comes from Chen Yuan, deputy governor of the People’s Bank of China (reported in Washington Post, Nov. 22, 1997).

2

Lin, Cai, and Li (1998).

1

As importantly, we hope to make a contribution to the literature on signaling. Recent research indicates that changed incentives brought about through the SOE reform program contributed to increased productivity in some instances.3 An important, though unintended, byproduct of adopting certain reforms may have been the information it conveyed to financial intermediaries (namely banks) about the future productivity of an SOE. Spence (1976) notes that all signaling devices involve self-selection, and that, for a signal to be effective, it must be unprofitable for sellers of low-quality products to imitate it. That is, “high quality” sellers must have lower costs for signaling activities.4 We argue that some of the reforms offered to SOEs involved self-selection and acted as signals that enabled better credit risks to separate themselves from worse ones. Given the banks’limited discretion over credit allocation (described below), it is an open question whether they could make use of this information in their lending decisions. If they did, however, this suggests a potential role for devices that enable some self-selection in the reform of SOEs, even when government intervention in the financial system is pervasive. The results may be important not only for China but for other transitional economies dominated by state-owned enterprises in need of quick reform. In section II we briefly review the evolution of the banking sector and of SOE financing from 1980 to 1994. Section III discusses related literature and the hypotheses that stem from it.

3

Guthrie (1997) finds, for example, that transferring control of the vast majority of enterprises to local or municipal authorities, whose funds are quite limited, has improved SOEs’ incentives by imposing harder budgets on them. Harder budget constraints and increased competition gave many SOEs strong incentives to make the most of the increased autonomy offered them under the reform program. Groves et al (1994) attribute much of the 4.5% annual increase in TFP from 1980-9 for a sample of SOEs to greater autonomy and incentives. Xu (1997) identifies particular reforms such as decentralizing the rights to control wages, make production decisions, and appoint new managers as key factors in SOE productivity improvement. While research by Li (1997), Wu (1995), and Liu and Liu (1996) makes it clear that technological progress played an important role in SOE productivity growth, that progress too may have resulted, at least in part, from the improved incentives provided by the reform program.

4

Spence notes that what he calls signals, taking the seller’s point of view, others have called screening or sorting, looking at things from the buyers standpoint. Here we refer to the agreements over production discretion made between SOEs (sellers of credit risk) and the government as signals of SOE productivity to banks (buyers of credit risk). We assume that banks did not actively engage in screening, but rather relied on those signals generated by the reform process. The Spence article is the introduction to a symposium on screening and signaling that contained some of the classic articles in the field including Jaffee and Russell (1976) on credit markets, Rothschild and Stiglitz (1976) on insurance, and Salop and Salop (1976) on labor markets. All of these papers trace their roots to Akerloff (1970) on informational asymmetry, lemons problems, and the resulting breakdown of markets.

2

Section IV contains our empirical analysis where we describe the SOE characteristics that determined the flow of finance from different sources. Section V concludes. Section II: Evolution of the Banking Sector It is not immediately clear that banks’lending decisions were much less susceptible to political influence than was the allocation of government grants.5 Beginning in the late 1970s, however, the Chinese government started to reform banks, which were all state-owned and did not function as effective intermediaries between savers and investors.6 Prior to that point, SOE investment was financed mainly from interest-free budgetary grants and, to a lesser extent, from retained profits. Banks were not that important a source of funds. The banking sector, moreover, was dominated by the People’s Bank of China, which accounted for four-fifths of all deposits in banks and credit cooperatives and 93% of all loans made by financial institutions. With the progression of other reforms and the resultant rising incomes, household savings rose sharply -- its share of total savings went up from 23% in 1978 to 71% in 1991 (Xie 1992, table 3-13; Qian 1995). A large share of those savings appear to have been intermediated through the formal financial sector -- a recent survey, for example, found that residents of China’s major cities put 50-75% of their assets in domestic banks.7 Perhaps most telling, the ratio of total financial assets to GDP increased from 95% to 232% between 1978 and 1991 (Xie, 1992, table 32 and 3-3; Qian, 1995). The growth of the formal financial sector coincided with a decline in government revenue as a share of GDP due, at least in part, to changing tax rates for SOEs and decentralization: total budgetary revenue dropped from 44% of GDP in 1978 to 33% in 1991.8 Given the growth of the banking sector and declining government revenues (as a share of GDP), SOEs came to rely more and more on bank financing over time while government budgetary grants played a less and less

5

In this analysis, government financing is considered as 100% directed credit, and bank financing a mixture of directed credit and commercial lending.

6

The discussion in this subsection relies heavily on Lardy (1998), Chapter 3, “The Evolving Banking System,” which provides perhaps the best analysis of the evolution of the banking sector. Unless otherwise indicated, all numbers cited in this section come from that chapter.

7

The survey was conducted by the Chinese Business Times (results reported in the Washington Post, November 22, 1997).

8

Statistical Yearbook of China.

3

important role. On a national level, domestic bank loans for fixed investment increased from 13% of total domestic financing in 1981 to 20% in 1990; for our sample of SOEs, the increase in the average share of bank finance was almost identical over the same period. Structural change in the banking sector also occurred, albeit gradually. The People’s Bank of China was separated from the Ministry of Finance in 1978. The government then reestablished the Agricultural Bank of China, the Bank of China, and the People’s Construction Bank of China in 1979. In 1983, the State Council designated the People’s Bank as central bank, and established the Industrial and Commercial Bank to take over People’s deposit-taking and lending business. Industrial and Commercial thus became China’s largest bank accounting for half of all bank lending; it continues to be the leading bank in China. The four major banks (which, of course, excludes the central bank) gradually expanded the scope of their services, and they competed for depositors and lending services, although fixed rates implied little actual interest-based competition.9 Entry of some new financial institutions was, however, permitted. Most were joint stock banks whose shareholders were SOEs or local governments. Some operated at a national level, others in regional niches. None presently threaten the dominance of the four major banks (Table 1). 10 Additional entrants included three smaller national banks (Everbright, Hua Xia, and Min Sheng)11 and a number of regional banks that opened in 1987-88. Also in the mid-1980s, several new types of nonbank financial intermediaries opened and began to grow including urban credit cooperatives, trust and investment companies, finance companies associated with enterprise groups, financial leasing companies, securities companies, and credit rating companies. Despite these structural changes, “policy lending” -- otherwise known as directed credit -has remained a defining characteristic of the Chinese financial system. If we define policy loans as 9

World Bank (1996) notes that, “in all, the government sets over 200 interest rates. In some instances, interest rates are set within a band around a reference rate, with band widths varying by type of financial intermediary and the creditworthiness of the borrower.” p. 29.

10

The two most important national-level banks initiated in the mid-1980s were the Bank of Communication and the China International Trust and Investment Company (CITIC) Industrial Bank. Neither had to extend “policy loans” (i.e., participate in directed credit programs), and they competed in the services offered by the four major banks. The Bank of Communication has since become relatively strong; by the mid-1990s its assets were one fourth those of the Agricultural Bank of China, the smallest of the four major banks.

11

The Min Sheng Bank, the first owned by private shareholders in China, was licensed in 1995 and opened for business in 1996.

4

only those from the central bank to financial institutions which are used to finance specific projects identified by the State Planning Commission (described below), one-third of total loans outstanding in 1985 were policy loans, and one-fifth in 1995 (Lardy, 1998). Given our relatively restrictive definition, this should be considered a lower bound on directed lending. Lou (1993), for example, indicates that at year-end 1991, policy lending accounted for 67 percent of assets at the Bank of China, 51.2 percent at the Agricultural Bank, 58 percent at the Construction Bank, and 25 percent at the Industrial and Commercial Bank. Portfolios dominated by directed credit appear to have negatively affected the profits of these banks. Lardy (1998) estimates the combined returns on assets (pre-tax profits over total assets) for China’s state commercial banks to have fallen steadily from 1.4 percent in 1985 to 0.3 percent for 1994.12 By comparison, pre-tax profits were, on average, 1.5 percent for Asian domestic banks from 1988-95.13 Moreover, the declining profits suggest that, as the government began phasing out its direct grants, it may have relied increasingly on banks to bail out SOEs in the late 80s and early 90s. In 1993, the government established three new “policy banks” to assume responsibility for those parts of the portfolios of the four large state banks that were devoted to non- commercially oriented loans. In this way, the state commercial banks would be free to pursue more commercial objectives. Previously, their freedom to allocate loans for investment was severely limited by the aforementioned detailed lending program drawn up by the State Planning Commission in consultation with the central bank, the Ministry of Finance, and the State Economic and Trade Association. State commercial banks did, however, have some autonomy in extending working capital loans, which accounted for 60 percent of the credit plan (World Bank, 1996).

12

Source: Chinese Finance and Banking Society, Almanac of China’s Banking and Finance 1993 (Beijing: Almanac of China’s Banking and Finance Editorial Department, 1993, pp. 401, 40.

13

See Claessens, Demirgüç-Kunt, and Huizinga (1997) for profit comparisons across countries.

5

Table 1: Chinese Financial System Bank Type

Number Branches

Total Assets (Renminbi billion)

2,529

1,758.8

Policy Banks State Development Bank of China Agricultural Development Bank of China Export-Import Bank of China

-

90.8 406.9 2.5

State Commercial Banks Industrial and Commercial Bank of China Agricultural Bank of China Bank of China People’s Construction Bank of China

37,039 63,816 12,630 33,979

2,633.9 1,253.1 1,837.9 1,397.5

1,937 54 12 n.a. -

305.0 70.7 20.3 82.9 n.a.

31 11 n.a. 18 9 7 5 n.a.

44.5 43.1 15.4 17.8 44.7 19.8 n.a. n.a.

Central Bank People’s Bank of China

Nationwide Commercial Banks Bank of Communications China Trust and Investment Corporation Investment Bank China Everbright Bank Hua Xia Bank Min Sheng Other Commercial Banks China Investment Bank Guangdong Development Bank Shenzen Development Bank Pudong Development Bank Shenzen Merchants Bank Fujian Industrial Bank Yantai Housing Saving Bank Bengbu Housing Saving Bank

Nonbank Financial Institutions 505.3 50,745 Rural Credit Cooperatives 214.8 5,229 Urban Credit Cooperatives n.a. 391 Trust and Investment Companies 66.1 5,240 People’s Insurance Company of China 4.2 n.a. China Pacific Insurance Company 27.6 29 Finance Companies n.a. 11 Financial Leasing Companies n.a. 87 Securities Companies 43 n.a. Mutual Fund Companies “-” is none. “n.a.” means not available. Data are for 1994. Sources: People’s Bank of China, 1995, China’s Financial Outlook; China Statistical Yearbook 1995; Almanac of Chinese Banking 1995. Reported in World Bank (1996).

The policy banks have not, however, achieved their intended objectives. World Bank (1996) notes that, because policy banks lacked sufficient resources to shoulder all policy lending, state commercial banks continued to bear much of the burden in the mid-90s. Moreover, the state commercial banks indirectly supplied the funds for policy lending because they were frequently asked to purchase the bonds issued by policy banks (Yusuf, 1997). Throughout the period,

6

therefore, we can be certain that the four institutions that comprised the vast majority of the banking system operated under severe constraints when allocating credit. Although our measure of bank credit does not distinguish between credit from state commercial, from policy, or from other commercial banks, it is likely that the state-owned enterprises that we analyze relied on state commercial banks for much of the bank credit that they received. There were, after all, no policy banks prior to 1993, and the other commercial banks were quite small even at the end of the period (Table 1). As noted, state commercial banks’ discretion over working capital loans may have been one lever through which credit allocation and thus productivity could have improved. Not all credit, moreover, was directed through the state plan. Still, given the other constraints on lending activity, our SOEs should provide a stringent test of the ability of banks to improve credit allocation by responding to signals.

Section III: Related Literature, Hypotheses Our data come from the two-part Survey of Chinese State Enterprises, a retrospective survey first conducted by the Chinese Academy of Social Sciences in 1991 and later updated in 1996.14 The data enable us to link a number of SOE characteristics with reliance on either government or bank finance. Throughout the analysis our dependent variables include a dummy variable indicating access to bank (government) finance and a variable for the share of total finance from banks (government). A. Control Variables We include dummy variables to control for the province of each SOE. Certain regions may have more political influence than others, which may affect on the share of external finance they receive from the government. Even in market economies, however, the pace of institutional change in the financial sector is often geographically uneven.15 In regions with less institutional

14

For details about the data, consult Appendix A. Each of the two surveys -- A Survey of Chinese SOEs: 19801989 and A Survey of Chinese SOEs: 1990-1994 – produced a separate data set. The first covers 769 SOEs over 1980-1989. It has been used in a variety of studies, including Groves et al. (1994), though not necessarily for the questions we address here. The second one, a follow-up to the 1980s survey, covers more than 600 SOEs, all of which also appeared in the earlier data set.

15

With respect to the U.S., for example, see Davis (1963, 1965).

7

development (e.g., fewer bank branches) we would expect the share of bank finance to be lower than in others. Disentangling whether differences arose from political factors or uneven distribution of financial innovation will not be possible, but the provincial dummies should, at least, help us hold regional factors constant. We also include dummy variables for year to control for macroeconomic fluctuations that may have affected credit decisions. Additional explanatory variables in Table 2 concern the level of government that oversaw the SOE’s operations. Our working hypothesis is that, all else equal, SOEs with closer ties to the central government should have had better access to government finance because central authorities controlled most budgetary resources. We might expect, therefore, that in the government finance regressions, the estimated coefficients would be largest for centrally-governed SOEs, followed next by provincial enterprises, and, finally, by city and county enterprises. With respect to bank finance regressions, we expect that SOEs that had strong ties to the central government (and thus softer budget constraints) had less need to seek bank finance. In general, we expect the hypotheses for explanatory variables in bank finance regressions to be exactly opposite those for the same variables in government finance regressions. To avoid repetition, we couch our hypotheses only in terms of the dependent variables regarding bank finance in the discussion that follows. Remaining control variables are included to capture government motives for imposing specific reforms on certain SOEs. By controlling for these motivations we can better isolate any signaling on the part of SOEs that stemmed from adopting a particular reform. Xu (1998) finds links between the adoption of some reforms and poor past productivity. He argues that the government’s desire to equalize performance across SOEs prompted them to grant additional autonomy to managers of the weakest firms. Including past productivity as an explanatory variable controls for both the direct effect of past productivity on bank finance and any indirect effects from the government equality motive that might bias coefficients for our signaling variables. We included three additional variables from Xu (1998) to account for other government motives. The coefficient of variation (mean divided by standard deviation) of value added per worker within an industry in a given year was included to control for the riskiness of an SOE’s business environment. Firms in riskier environments had less discretion over production than

8

others; the government rather than the SOE manager was, therefore, charged with bearing these risks. Managers at larger firms (as measured by the number of workers) were also found to have less discretion than others due, it was thought, to government’s desire to maintain control of firms where more employment was at stake.16 These larger firms tended also to produce a “strategically important product (such as steel or energy) to the government.”17 Finally, a positive link between an SOE’s capital to labor ratio and discretion over production was found. Firms with greater capital intensity were, apparently, better able to make use of added discretion in increasing productivity (and thus revenue to the government).18

B. Signaling Variables Some SOEs agreed to performance contracts with the government (under the so-called Contract Responsibility System). Under these contracts, which typically lasted for three to five years, managers were granted some discretion over production in return for achieving certain targets. Contracts typically specified the distribution of value added between the state and the SOE, performance requirements such as the minimum annual expenditure on capital maintenance, the number of new products to be developed, the volume and the price of the output to be delivered to the state, the dependence of the chief executive officer’s compensation on the

16

This governance motive is also cited as an important explanation for the differences between private firms and enterprises under bureaucratic control (Shleifer and Vishny, 1993).

17

Xu (1998), p. 546.

18

Again, our primary motivation for including these variables is to hold government motives constant to obtain unbiased coefficients for our signaling variables. We recognize, however, that each of them may also have more direct links to our financial variables. For example, the view that banks, through their threat not to renew credit, can often exercise far more influence than other forms of corporate governance (e.g., shareholders) is put forth in Stiglitz (1985). We might expect a positive relationship, therefore, between past productivity and bank finance. In addition, SOEs with complex production technologies may have relatively high capital to labor ratios. Complex production technologies are likely associated with greater informational asymmetries between borrowers and lenders, which would imply less bank credit for SOEs with greater capital intensity. Predictions regarding the riskiness of the business environment and our financial variables are less clear-cut. In sectors with high coefficients of variation in value added per worker, revenue streams may be more erratic and SOEs may be more inclined to seek external finance rather than rely on internal funds. On the other hand, external sources such as banks may be less willing to provide credit to these firms due to their risky business environment. Finally, with respect to SOE size, we see no obvious direct link to our financial variables.

9

performance of the firm, and a link between the SOE’s total wage bill and its profit level – that is, an ex-ante wage elasticity with respect to profit. Managers’information advantages made it difficult for the government to distinguish efficient from inefficient SOEs. Perhaps to induce firms to reveal their levels of efficiency, the government offered performance contracts that differed in incentive intensity and insurance coverage. One measure of incentive intensity was the firm-level pay sensitivity to profits. We presume that managers’utility functions increased with monetary rewards (which are equal to a lump-sum transfer plus shared profit) and decreased with effort level. “High sharing” contracts offered high pay sensitivity to profits (a high-powered incentive) and a low amount of fixed transfer (little insurance). “High insurance” contracts offered large transfers and a low wage elasticity with respect to profits. To achieve separation between high and low-quality parties, it must have been costly for inefficient firms to mis-represent their type. For the signal to be effective, therefore, managers that knew that their firm was unlikely to achieve high profits must have maximized utility by choosing contracts with high transfers, low wage elasticities, and little effort. Managers of relatively efficient SOEs had to maximize utility through low transfers, high wage elasticities, and greater effort. In addition, some firms may have opted for no performance contract whatsoever. In our framework, these would be extremely inefficient SOEs for which even a “high insurance” contract would expose them to risks not faced under the centralized status quo. However, most firms had adopted performance contracts by the end of the period, which may have decreased their usefulness as a signal over time (absent additional information on provisions).19 More generally, it is not immediately clear that these contract provisions were effective signals for lenders. First, provision agreements involved both the SOE managers and the government. Government actions, therefore, may have inhibited true self-selection. As noted, to address that problem, our control variables include some that capture government motives for imposing certain contract types on some SOEs. A second, more fundamental problem is that the trade-off between insurance and profit-making incentives may not have been as well specified as we suggest, or simply may not be detectable in our econometric specifications given the 19

Although only two percent of our SOEs had agreed to performance contracts by 1984, the figure had grown to 42 percent by 1987, and 88 percent by 1989.

10

government’s multiple objectives in crafting performance contracts with SOEs.20 A third concern is that the lending constraints faced by banks (under the state credit plan) made it difficult for them to respond to signals. Our econometric specifications should be viewed, therefore, as joint tests of the hypotheses that performance contracts involved self-selection, could effectively separate efficient from inefficient firms, and could be acted upon by banks. The variables we use to capture these signaling effects are the firm-level wage elasticity with respect to profits (a continuous variable) and a dummy variable indicating whether an SOE had a performance contract. To the extent that these were reliable signals about a firm’s future performance, we expect both to have been positively correlated with access to bank loans. We also include profit retention rates as potential signals. SOEs faced two retention rates: a base rate for profits up to some pre-specified base level and a marginal rate on anything above that. We view the base retention rate as a transfer mechanism and the marginal rate as an incentive device for SOEs. Provided that base profit levels were comparable targets across firms, and that the government created some trade-off between base and marginal rates, the rates may also have been valuable signals to banks.21 We expect higher marginal rates (lower base rates) to be associated with increased bank finance.22

C. Additional Reform Variables There may have been other reform variables that affected banks’lending decisions but did not act as signals. For example, banks may have viewed SOEs with managers that had discretion to set wages more favorably than those that did not possess this motivational tool. We find it doubtful that such discretion could act as a signal because managers of inefficient firms could 20

Control variables including dummies for year, province, and level of government oversight should, perhaps, control for some of these objectives.

21

Base profit was usually some weighted average of past profits or simply one-period lagged profit. See The Research Group for the Chinese Firm System Reform, System Reform Committee, Management Responsibility System in Practice, (Beijing: Economic Management Press, 1988).

22

Because the marginal and base retention rates are negatively correlated, we included only one at a time in our regressions. We present the results using the base retention rates for the 1980s. In the 1990s, neither the base nor the marginal rate was available. The closest substitute was a simple dummy variable indicating whether the SOE had a performance contract. We recognize that the dummy provides far less information than the rates would have. Indeed, it provided so little additional information relative to the retention rates in the regressions for the 1980s that we dropped it.

11

adopt this reform without exposing themselves to additional risk. Little information should have been revealed through this type of self-selection, therefore, as all managers had an incentive to obtain added discretion. However, banks did observe that the manager had convinced the government to grant him this discretion. Because we control for other government motives as described above, reforms like wage discretion might be viewed as capturing aspects of government decision making not necessarily related to those motives. To the extent that, controlling for its own interests, the government granted such discretion to those SOEs that would make the most of it, we might expect a positive relationship between bank finance and additional managerial discretion. Indeed, Xu (2000) finds that, all else constant, discretion over wages was positively associated with productivity (both levels and growth rates) in the 1980s. We describe other non-signaling reform variables below. 1. Added Discretion In addition to a dummy variable for wage discretion, we include variables that capture other aspects of managerial discretion. For the 1980s, we include a dummy variable for production autonomy indicating whether SOE managers could make production decisions in any of six broad areas: value of output, physical quantity of output, and choices of product, technology, production scheduling, and exports. Because the 1990s data offered a much richer array of questions regarding production autonomy, we used a principal component index of dummy variables in many areas such as investment, foreign trade, employment, and asset transactions. Because wage discretion was a component of this index, we did not include it in specifications where the autonomy index appeared. We expect all autonomy variables to be positively linked to bank finance (and thus negatively to government finance), at least to the extent that government granted autonomy to those firms that would make best use of it (ceteris paribus).23 A similar argument applies to the share of output produced under the state plan -firms with high shares had less discretion than others, and thus may have been perceived as relatively unattractive candidates for loans.

23

Like the dummy variable for managerial wage discretion, Xu (1997) finds a positive, significant link between production autonomy and productivity (both levels and growth rates) for the 1980s.

12

2. Market Structure and Orientation We also include a variable that measures an SOE’s mark-up ratio on its output. While the variable has little to do with the reform program per se, it may have been information that benefited banks. Controlling for productivity, an SOE with a high mark-up ratio may have been competing in a relatively protected market. Since firms with substantial market power are often better able to repay loans than others, commercially-oriented banks find them attractive clients. To the extent that China’s banks were commercially-oriented and had information on market power, we expect a positive correlation between market power and bank finance. For the 1990s, unfortunately, we lack the data necessary to construct a mark-up ratio. We do, however, have the share of output devoted to exports. This information was also probably available to banks, and international competitiveness was probably viewed favorably in lending decisions. Although it captures a very different aspect of market structure than mark-up ratios, we also expect the share of exports to be positively linked to bank finance. Finally, banks could also observe whether an SOE had a new manager. To the extent that turnover during this period signaled a shift from a “bureaucrat” to an “entrepreneur,” bankers may have interpreted changes as an indication of enhanced future productivity. More generally, however, his predecessor’s firing may have provided a new manager a more credible threat of losing his position if he did a poor job. Since a manager that loses a job incurs a loss of income and reputation in the managerial labor market, new managers may have had incentives to work harder and invest more time in improving worker skills than did incumbent managers. On the other hand, a new manager may have found it difficult to re-establish the relationships that his predecessor had with bankers. Because lending decisions are dependent on information, and new managers have yet to establish a track record, banks may have been somewhat reluctant to extend credit to them. We have no strong a priori prediction about the net effect of new management on bank finance. A summary of our explanatory variables and hypotheses appears in Table 2.

Section IV: Empirical Results A. Summary Statistics: Changing SOE Financial Structure

13

Although the financial structure of Chinese SOEs changed gradually, the cumulative effect of those changes has been pronounced. Over the 15-year period, the share of government finance for the SOEs in our sample declined considerably as banks became the primary external source for funds (Figure 1). At the beginning of the 1980s, government provided roughly 25% of SOE finance; from 1990 to 1994, that figure had declined to about 2 per cent. The share of bank finance, by contrast, increased from 15% in 1980 to more than 32% in 1987; since that point, the percentage has hovered between 20 and 30%. Data linking the level of government oversight with SOE finance provide one indication that, relative to bank finance, government finance was awarded based on non-economic criteria. Throughout the 1980s, SOEs overseen by the central government were the only ones to receive a larger share of finance directly from the government than from banks (Figure 2). In that decade, although the share of government finance declined for centrally-governed firms, it never dipped below 30%. For those same SOEs, the share of bank finance, while rising, never exceeded 15%. The further an SOE was from central oversight, the more it relied upon bank finance. On average, provincially governed SOEs received about 40% of total finance directly from the government in 1980 and only 6% from bank loans. Throughout the mid- and late 80s, however, provincially governed SOEs received about 20% of total finance from each source. For city and county-governed enterprises, the balance tipped in favor of bank finance -- banks supplied about 20% of the average firm’s finance in 1984, a figure that grew to 30% late in the decade. Their average shares of government finance declined from about 20% in the early 1980s to only about 10% late in the decade.

14

Table 2: Variables Used in Estimation Variable Names

Dependent Variables Bank Finance

Description

Hypoth. Link to Bank Finance

Either share of SOE’s total finance that was received from banks or a dummy indicating access to any bank finance. Either share of SOE’s total finance that was received directly from government or a dummy indicating access to any grants.

--

Controls for Govt Motives Lny Lnk LnL Var(Lny)

Natural log of value added per worker in prior year. Natural log of capital to labor ratio. Natural log of the number of workers. Coefficient of variation in Lny within an industry-year cell

+ ? ?

Location/Governance Jiangsu, Jilin, Shangxi, Sichuan Central, Provincial, City, County

Dummy variables indicating the province in which an SOE resides. Dummy variables indicating the level of government responsible for overseeing the operations of the SOE.

?

Government Finance

--

Explanatory Variables

Signals Wage Elasticity Marginal Profit Retention Performance Contract

Ex ante wage elasticity with respect to profit at the firm level. Marginal profit retention rate in year t, set before the year begins. (Hypothesis Reversed for Base Profit Retention Rate). Dummy variable indicating whether SOE had performance contract.

Additional Reforms Autonomy

Autonomy Index

Planned Quantity Discretion Over Wages New Manager Markup Ratio Export Share

+ for Central

+ + +

+ Dummy variable equal to 1 in all years after the government delegated production decision rights to the manager of a firm in several areas (used in 1980s regressions). A principal component index of dummy variables of managerial autonomy in many areas such as investment, foreign trade, employment, and asset transactions. Share of SOE’s output under the government’s mandatory plan. Dummy variable equal to 1 when managers had discretion to set wages within the SOE. Dummy variable equal to 1 if new manager took over operation in year t. Difference between output price and price of inputs, divided by price of inputs. Share of total output exported abroad.

+

+ ? + +

15

% bank financing: 1980-94

% govt financing: 1980-94 0.35

0.30

0.30

0.25

0.25

0.20

0.20

0.15

0.15

0.10

0.10

0.05

0.05

0.00

0.00

year

year

Figure 1. The Evolution of SOE Finance: (Source: based on the authors’ calculations from our data set)

These differences in financial arrangements do not appear to be driven primarily by the industrial composition of centrally-governed firms nor by their geographical location. It may have been, for example, that centrally-governed firms were clustered in industries to which banks preferred not to lend, or were clustered in provinces where banks were not well established. However, when one looks only at data for SOEs in individual industries or specific provinces, the same pattern emerges -- the further an SOE from central control, the less its reliance on government finance.24 %govt financing: central govt % govt financing: city govt

% govt financing: province govt % govt financing: county govt

1

%bank financing: central govt % bank financing: city govt

% bank financing: province govt % bank financing: county govt

.35

.9 .3 .8 .25

.7 .6

.2

.5 .15

.4 .3

.1

.2 .05 .1

80

82

84

86

88

90

92

year

94

80

82

84

86

88

90

92

94

year

Figure 2: Evolution of SOE Finance, By Level of Government Oversight

24

Calculations available from the authors.

16

The large decline in government finance depicted in the figures did coincide with changes in the ways SOEs financed themselves. In the 1990s, neither the centrally-governed SOEs nor any other group received more government than bank finance. City and county governed enterprises, which received about 10% of their finance from government sources in the late1980s, received closer to 5% of total finance from government according to the 1990s survey responses. Provincial and central governed SOEs exhibited similar declines -- provincial enterprises went from an average share of government finance near 20% in 1988 to one of 5-10% in the 1990s. Centrally-governed enterprises declined from a 25% government share in the late 1980s to just under 10% in the 1990s. In one potentially important respect, however, the pattern of government finance did not change. In the 1990s, SOEs closer to central control continued to receive more government finance than other enterprises. Again, that pattern holds up when one looks at individual industries such as machinery. There were also changes in shares of bank finance. On that measure, provincial and central governed enterprises experienced 7-10 percentage point increases -- large gains, but not as big as their 15-20 percentage point losses in government finance dating from the late 1980s. In the 1990s city and county SOEs maintained their 1980s average shares of bank finance (near 25%). In short, shares of bank finance for provincial, city, and county governed SOEs became more tightly clustered in the 1990s. Perhaps this is another indication that bank finance was increasingly awarded based on non-political criteria among non-centrally governed enterprises. At the least, the data make it clear that centrally-governed SOEs were somewhat slower to turn to banks as a source of finance, and that central governance is a necessary control in the regressions that follow. Indeed, in all of our regression results centrally-governed SOEs had significantly less (more) access to bank (government) finance than other SOEs.

Regression Results Results regarding allocation of bank and government finance in the 1980s appear in Table 3 and Table 4; for the 1990s in Table 5 and Table 6.25 The dependent variable is either a dummy

25

We dropped observations if the productivity measure was missing, the share of government or bank finance was missing, or if the available panel for that firm was substantially shorter than the full period. In the end, we used two unbalanced panel data sets, one for the 1980s and one for the 1990s. We did not combine these

17

indicating whether an SOE had access to bank (or government) finance, or a continuous variable bounded by zero and one indicating the share of total finance received from banks (or government). We generally opted for simple estimation techniques – logit for specifications where access to finance was the dependent variable, tobit for those where share of total finance was the dependent variable. Standard panel estimation techniques (e.g. fixed effects regression) were inappropriate because the typical SOEs’access to finance and, to a lesser extent, its share of finance from specific sources did not vary sufficiently over the period. Rather, SOEs that received bank finance in one period tended to receive it in all others. While the average share of government finance did decline throughout the period, SOEs that had relatively more government finance early in the period also had more later in the period. Our primary aim, therefore, was to use simple techniques to describe the differences between “bank-sponsored” and “governmentsponsored” SOEs.

1. Bail-out Responsibilities In the 1980s there was a strong positive association between past productivity and bank finance variables (Table 3). Over the same period, there was a pronounced negative relationship between past productivity and government finance (Table 4). The estimates suggest that a 10 percent increase in lagged productivity would (a) increase the probability of obtaining bank loans by roughly 2.5 percent and the share of bank finance by 1 percent, and (b) reduce the probability of government finance by roughly 1 percent and the share of government finance by 0.7 percent. We interpret this as strong evidence that throughout the 1980s government was forced to bail out the weakest SOEs, those that banks were less likely to extend credit. Despite their constraints and limitations (described in Section II), banks did appear to lend to SOEs that were, on average, relatively productive. By the 1990s, the situation had changed. The negative partial correlation between past productivity and government finance had been replaced by a strong positive correlation (Table 6). The magnitudes indicate that an increase of productivity by 10 percent would increase the probability of obtaining government financing by 3.5 percent, and the share of government

two data sets because they did not contain all of the same variables, and because we lacked reliable price deflators for the transition year between the data sets (from 1989 to 1990).

18

financing by 3.5 percent. As the level of direct government transfers declined in the late 80s and early 90s, the relatively few SOEs that continued to receive finance had relatively high past productivity. These SOEs were the exception rather than the rule. Of the 2503 observations in the 1990s data set, only eighty-three indicated that an SOE had access to government finance. In the 1980s nearly half of the active observations indicated that an SOE had access to some government finance. That figure had been declining throughout the period but, by the 1990s, the few SOEs that continued to receive direct government transfers had relatively high levels of past productivity. Although SOE bail-out responsibilities had apparently been shifted to banks, the positive relationship between bank credit variables and past productivity was still evident in the 1990s. The estimated coefficients were somewhat smaller than they had been for the 1980s data--an increase of productivity by 10 percent would increase the probability of obtaining bank financing by 1.5 percent, and the share of bank financing by 0.7 percent--which may be one indication of the increased burden placed upon state-owned banks. The persistence of the positive relationship is an indication that some combination of factors – including increased entry by private banks, the creation of policy banks, and state-owned banks’continuing discretion over loans for working capital -- was able to offset the increased burdens faced by many banks so that, on average, the typical recipient of a bank loan still had better than average past productivity, and thus at least some bank loans appear to have been awarded based on some commercial principles. Our estimates do not, of course, tell us anything about how much stronger the association between bank finance and past productivity would have been in a private banking system free of government interference in lending.

2. Signals and Separation Our signaling variables performed largely as we expected. In the 1980s, SOEs that adopted performance contracts with a relatively high elasticity of wages with respect to profits tended to receive more bank finance than did others (Table 3). The estimates suggest that an increase of this elasticity by 0.20 would increase the probability of bank financing by 7 percent. Note that the wage elasticity was not significant in affecting the share of bank financing; neither did it affect government financing significantly. Those that adopted contracts with high base profit retention

19

rates received less bank finance.26 The results are consistent with the hypothesis that these contract provisions permitted self-selection among SOEs, and that the signals conveyed helped to resolve informational asymmetries between banks and SOE managers. Signaling should be reflected, perhaps, more in access to bank finance rather than the share of bank finance. The significance of these two variables only in access equations thus reinforces our interpretation of them as signals. SOE reforms that foster genuine self-selection, therefore, may be worth pursuing if only for the information they provide. Results are not, however, nearly as pronounced for the 1990s (Table 5). Although the estimated coefficients for wage elasticity and for a dummy indicating whether the firm adopted a performance contract are positive (as expected), neither are significant.27 The estimated coefficients for the wage elasticity variable are, moreover, much smaller than they were in the 1980s. In the 1990s base profit retention rates were phased out as policy instruments, and so we cannot test whether the strong negative correlation between those rates and bank finance persisted into the 1990s. We suspect, however, that performance contracts generally became less effective signals as their use became more widespread. Nearly nintey percent of the SOEs in our sample had adopted a performance contract by the mid-1990s. To the extent that these contracts were imposed upon enterprises using a “one size fits all” approach, the self-selection of the 1980s may no longer have been possible. The reliability of these reforms as signals, therefore, may have greatly diminished over time.

3. Other Reforms As noted above, we do not consider these other reforms signals because their adoption did not expose SOE managers to any obvious additional risks. Banks did, however, observe that the SOE manager had convinced the government to grant him a right, one that not all other managers enjoyed. We need to control, therefore, for those aspects of the reforms reflecting those government objectives that had nothing to do with enhanced SOE productivity (as described 26

As expected, results are reversed for government finance. Firms with high base retention rates had better access to government transfers; those with high wage elasticities had less access. However, the wage elasticity coefficient does not quite achieve significance at conventionally accepted levels (Table 4).

27

There was, however, a strong negative relationship between the performance contract dummy and government finance variables in the 1990s (Table 6).

20

above, following Xu (1998), we use lnyt-1, lnkt-1, lnL, and the coefficient of variation of lny in the industry-year cell as proxies for these objectives). Such factors must be held constant to obtain unbiased estimates of the relationship between bank finance and productivity-related aspects of reform. Controlling for such factors, we expected the reform to convey information to banks. We expect (and find) a positive link between many non-signaling reforms and access to bank credit, especially those that relate to increased discretion over production decisions. In the 1980s, for example, SOEs whose managers had discretion over wages had better access to (and shares of) bank finance (Table 3). To a lesser extent, our crude dummy for production autonomy also performed as expected -- in some specifications, SOEs with autonomy had higher shares of bank finance. There was not, however, a consistent relationship between share of output dedicated to the state plan and bank finance variables across our specifications. Yet, with that one exception, reforms that granted additional discretion to SOE managers were associated with greater bank finance in the 1980s.28 By the 1990s, the situation was different. As expected, our index of production was positively associated with bank finance variables and the share of output under the state plan was negatively associated with bank finance. However, neither of those variables was significant in any of our specifications (Table 5).29 Like our signaling reforms, many of the reforms regarding increased discretion had been adopted by the vast majority of Chinese SOEs by 1989 (Xu, 1998). It is likely that they provided less information to bankers in the 1990s than they had provided in the mid-1980s. Also, there was no positive relationship between productivity and most of the other reforms (results not reported); the positive relationship existed for most of the 1980s, however (Xu, 2000). In the 1980s, the presence of a new manager was associated with less access to bank finance, though not significantly so (Table 3). Perhaps the destruction of a relationship with the prior manager outweighed any potential benefits banks perceived from new management. SOEs with new managers did have better access to government finance (Table 4) than other enterprises; this

28

Discretion was also typically negatively associated with government finance (Table 4).

29

Results for government finance are contradictory with respect to these discretion variables (Table 6). On the one hand, the index of autonomy was negatively associated with government finance (as expected). On the other share of output under the state plan was negatively associated with government finance. We put less faith in our government finance specifications for the 1990s, however, because so few SOEs received any government transfers.

21

could reflect the government’s attempt to substitute government financing for bank financing when a lack of information about the new managers was perceived by the government. It is, however, difficult to know whether this was because banks had not established an adequate relationship with the new manager or because the government provided extra financial support during management transition periods. By the 1990s, management turnover was not significantly associated with any of our financial variables. While we mention them, our sense is that the management turnover results are not central to our story. Results regarding variables that describe the product market faced by SOEs are somewhat more central. In the 1980s, the mark-up ratio, our proxy for market power, was positively associated with both bank finance variables (Table 3). This indicates that banks were cognizant of firms’operating margins in their lending decisions. As was also expected, the mark-up ratio was negatively associated with government transfers (Table 4) – enterprises with relatively low margins were more likely to be the responsibility of the state, which provides further support for our bail-out analysis. For the 1990s data, we were unable to compute a mark-up ratio. Export share, a measure of international competitiveness was, however, positively associated with bank finance (Table 5).

Control Variables The variables included to capture government motives, both in credit decisions and in granting reforms to SOEs, were also not the central focus of our analysis. That said, they explain substantial variation in our dependent variables and thus warrant some mention. Results for the past productivity variable (“lnyt-1”) were already discussed in the sub-section on bail-outs. SOEs with relatively high capital to labor ratios (“lnkt-1”) received less bank finance than did others in the 1980s (Table 3). Our interpretation is that informational asymmetry increases as production processes become more complex, and thus banks become less willing to lend. Further support comes from the strong positive correlation between capital intensity and government transfers (Table 4).30 In the 1990s, however, there was no significant relationship between capital intensity and bank finance, and the relationship between government finance and capital intensity was 30

Another interpretation is that high capital-labor ratios were observed in heavy industries, which were less profitable and thus less worthy of bank finance. We have, however, controlled for lagged productivity.

22

negative. There were, however, so few SOEs that received direct government transfers in the 1990s that we do not find this last result too troubling. The number of employees at an SOE, which we included because Xu (1998) had found that government tends to maintain its control of larger enterprises, was positively and significantly linked to both bank and government financial variables in both the 80s and 90s. A simple interpretation of that collection of results escapes us. Furthermore, we had included the coefficient of variation in lny within an industry-year cell as a proxy for risk of the business environment. Our guess was that the government would provide support to SOEs in industries that were relatively risky if banks were less willing to lend to them. While we did find a negative correlation between bank finance and the coefficient of varation of lny for the industry-year cell for the 1990s, the association was positive in the 80s. Moreover, the relationship with government finance was negative in both the 80s and 90s. Either our variable does not measure what we intended it to, or our hypotheses were poorly thought out. Our most straightforward defense is that our main results hold with or without the inclusion of our proxy for market risk. Finally, both in the 1980s and the 1990s, SOEs under the oversight of the central government were more likely to receive government finance and less likely to receive bank finance.31

Section V: Conclusions We find a pronounced shift in the way the poorest performing SOEs were financed from 1980-95. In the 1980s, bail-outs were accomplished through direct government transfers. By the 1990s, few SOEs received such transfers, and those that did were relatively productive SOEs. These bail-out responsibilities must, therefore, have devolved to state-owned banks. Although we did find a significant positive relationship between bank credit variables and past productivity in both periods, estimated coefficients were larger in the 1980s than in the 90s. We interpret this as evidence, albeit somewhat indirect, that bail-out burdens were increasingly imposed upon banks. The result does not augur well for the future of the Chinese banking system.

31

The base category in the 1980s estimations was SOEs under the oversight of “other” government. In the 1990s, the firms under this category no longer existed, thus the SOEs under the oversight of county government was used as the base category.

23

From a more academic perspective, however, the positive correlation between bank finance and SOE productivity could be interpreted as good news. Despite the constraints faced by these banks, they still managed to allocate credit to SOEs that were, on average, relatively productive. We view this as evidence that, even in relatively difficult circumstances, banks’ abilities to economize on the costs of gathering and processing information can offer advantages over direct government credit transfers (provided, of course, there is some incentive for bankers to avail themselves of the information). In the Chinese case, state-owned banks’discretion over working capital loans, the portion of total credit not directed under the state plan, and the link between bank employee compensation and loan portfolio income all likely played a role in maintaining the positive association between bank credit and SOE productivity. Other factors such as the introduction of policy banks (to handle directed credit) and the arrival of new private banks also played a role, but developments on neither of those fronts had been dramatic as of 1995. We are by no means endorsing the Chinese system as a model of banking efficiency. We have no doubt that the disparities in productivity associated with bank credit and direct government transfers would be much more pronounced if Chinese banks had been privately owned and bank managers had full discretion over lending. We merely point out that, even in the least advantageous circumstances, banks were superior to government bureaucracies in gathering and processing the information necessary to select relatively productive firms. Information generated through the SOE reform process was, it appears, an important part of what was gathered and processed by banks. In particular, SOE reforms that offered managers additional discretion over production, and those that enabled mangers to self-select and thus expose themselves (and their employees) to greater risk, were positively associated with acquiring bank finance. These results, especially those pertaining to self-selection, suggest that in statedominated systems where dramatic structural change (i.e., privatization) is not possible in the short run, it may be helpful to offer a menu of reforms that differ in the level of risk imposed upon SOE managers. The choices made by those managers may offer information that could benefit financial intermediaries. We have analyzed the determinants of credit allocation using data on the financial structures of SOEs. A potentially fruitful avenue for future research would be to analyze the

24

same issues using data from Chinese banks. We could, for example, examine whether bank portfolios actually changed in response to information generated through the reform process, or whether it was new private banks that were more likely to respond to signals than state-owned banks. As in any study of one country, moreover, there is always a question of how generalizable are its conclusions. While China is of sufficient interest (and size) that this concern need not be an overriding one, we would rest easier if others came to similar conclusions for other economies in transition.

Appendix A. The Data Sets The first data set we use is A Survey of Chinese State Enterprises: 1980-1989. It covers 769 SOEs in 21 cities of four provinces (Shanxi, Jilin, Jiangsu, and Sichuan). The 769 firms constitute a stratified random sample of all SOEs in manufacturing. There was substantial variation in the size of these SOEs: the median SOE had 930 employees, the SOE at the 10th size percentile had 304, and that at the 90th percentile had 3175. The data set has two parts. Part one is a quantitative table filled out by the accountants of an enterprise. It includes 321 variables covering details about products, costs, wages and labor utilization, investment, financing, fixed assets, profit distribution, taxes, prices, and material inputs. Part two is a questionnaire answered by the manager of the enterprise. The manager answered questions about performance contracts signed with the government, the relationship between the enterprise and the government, production autonomy, the characteristics of the management, and so on. The second data set, which covers 1990-1994, is a follow-up to the original. About 680 of the original SOEs remained in the data set. There was still a quantitative table to be filled out by the accountant and a questionnaire for the manager, although the entries used were somewhat different. For instance, marginal profit retention rates no long existed, and there were more indicators about different types of managerial autonomy, which partly reflected the changing path of reforms. An advantage of the second data set is that it has far fewer missing observations than the first one. Appendix B. Construction of Key Variables for the 1980-89 Data Set In constructing these variables, we have followed other users of this data set, especially Li (1997), and Gordon and Li (1995). All quantities (value added, capital stock) are expressed in 1989 market values. We assume that the 1989 prices reflected best the opportunity costs of the resources. Capital Price Indexes and Capital Stock The survey contains answers to questions about the inflation rate of the mixed price of equipment between the periods 1965-1975, 1975-1980, 1980-1984, and for each year between 1985 and 1988. Based on these answers we computed average inflation rates for equipment. For 1980-1984, we assumed equal yearly inflation rates. For 1989, since we did not observe equipment inflation, we used the output inflation rate in the machine industry as a proxy.

25

Since the survey did not provide information on prices of buildings or plant, for that inflation measure we used the percentage increase in aggregate construction costs compiled by the State Statistical Bureau. This series has also been used by Li (1997). We computed the composite price index for capital goods by averaging the equipment price index and the buildings and plant price index, the weights being the investment expenditures on equipment and plant. We based our measure of capital stock on capital assets “for productive use”, which includes plant and equipment for industrial production. (In contrast, capital assets “for nonproductive use” are mainly buildings and expenditures on dormitories, cafeterias, employee housing, and other social welfare functions.) Following Li(1997) and Gordon and Li (1994), we did not use the net value of capital stock as the base to compute capital stock because it “tends to exaggerate the increase in enterprise capital stock during the sample period in which the inflation rate was high, because the accounting rate of depreciation was artificially low and the depreciation was based on historical costs.” (Gordon and Li, 1994) Realized investment at year t is imputed by subtracting the nominal value of productive capital assets at the end of year t-1 from that at the end of year t. The reported investment, usually different from our imputed figures, is not used because it measures the value of capital expenditure (rather than capital formation) in a given year. It includes, e.g., expenditure on ongoing construction projects; while it excludes prior investment projects completed in the year. Assuming that investment occurs smoothly over the course of a year, we can compute the capital stock in 1980( Ki,80), the initial year, as K i ,80 = 0.5( Ki*,79 + Ki*,80 ) P89K / P80K where K it* is the productive capital asset in year t, and Pt K is the cumulative price index for the composite capital goods.32 The capital stock for the following years is then constructed by the following formula: K P89K * P89 * K it = Ki ,t −1 + 0.5I i ,t −1 K + 0.5I i ,t K , t = 81, K ,89 Pi ,t Pi ,t −1 * where I is the imputed realized investment. With this procedure, there are still a little more than 700 missing Kit. Their values are imputed as the industry-year averages for 36 industries. Price Index for Value Added The price index for value added is based on the price indexes of output and material inputs. Let Pvt be the price index of value added in year t, and PQt be that of output, and PMt be that of intermediate inputs. Let Qt denote output units, and Mt input units. By definition, the Laspeyres price index of value added is computed as follows: PVt PQt Qt −1 − PMt M t −1 = PVt −1 PQt −1Qt −1 − PMt −1 M t −1 32

Ki*,79 , unobserved in the data set, is extrapolated as in Li (1994):

(beginning - of - year total capital) 80

(end - of - year productive capital) 80 (end - of - year total capital) 80

26

Tyler expansion along (PQ t-1, PM t-1) gives the following formula for the percentage price increase of value added based on those of output and of intermediate inputs: PVt Q M ln = t −1 ( PQt − PQt −1 ) − t −1 ( PMt − PMt −1 ) PV t −1 Vt −1 V t −1 (Below we discuss the construction of the output price index (PQt) and intermediate input price index (PMt) ) In the empirical implementation, we value the value added for each year at the 1989 price of value added.) The Output Price Index The survey reports the mixed (plan and market) price index for the firm’s main product. While most firms reported cumulative price indexes, some reported year-to-year price inflation. We checked carefully and corrected those obvious coding errors. When in doubt, we treated them as missing. Consequently, we have around 500 firms reporting a reasonable mixed price index. For the rest of firms, we computed the average year-to-year mixed price inflation rates for their industry-year sample, then assigned that value as the imputed mixed price inflation rate. Then, we converted them to a cumulative mixed price index. We then estimated the market output price index. The survey has information about the sales under the state plan and to the market, and their respective prices. Based on this information, we constructed the market price index for output. Again, firms with missing values for the market price index were assigned their industry-year averages. These price indexes were then used to compute the gross value of output (GVO). The survey reports GVO in current mixed prices. We first obtained GVO in current market prices by multiplying the reported GVO by the ratio of market output prices to mixed output prices in year t. That number was then translated into GVO in 1989 market prices by multiplying it by the ratio of the market price index in 1989 to the market price index in year t. Price index of Intermediate Inputs The data set has detailed information about the plan and the market prices of the two primary materials but it does not provide information about energy and other intermediate inputs. We therefore computed price indexes for intermediate inputs based on the assumption that the inflation rate for intermediate inputs was the same as that of materials. This is reasonable since materials accounted for the vast majority of intermediate inputs. A significant portion of the reported material price variables was missing: roughly 40 percent of the answers were useful. We first computed the mixed price of each material input using the physical shares of the plan and the market inputs. Then we computed the year-to-year Laspeyres index of mixed material prices. Year-to-year Laspeyres indexes of market prices were computed similarly. Again, the missing values were imputed using the industry-year averages. The quantity of intermediate inputs was then computed using these price indexes. We first obtained the quantity of intermediate inputs valued at the current market price by multiplying the reported intermediate inputs--in current mixed prices--by the ratio of the current market price to the mixed price of intermediate inputs. This number in year t was then translated into intermediate inputs in 1989 market prices by multiplying it by the ratio of the cumulative market price index of intermediate inputs in 1989 and that in year t.

27

The Markup Ratio We follow Li (1997) in constructing the mark up ratio. Specifically, M it = ∑

4 j =1

Dij µ j − θ ∑

89 s = t +1

Cis , The first term on the right hand side is the industry-specific

markup ratio, assumed to be the markup ratio for all the firms in four industries (Light, Material, Chemical, and Machine). It is assumed that the markup ratios were identical in 1989 within an industry, but differed across the four. The second term was calculated by assuming that the change in markup ratio was proportional to the change in output prices relative to input prices ( Cit = π it − π itm , π it being enterprise-specific inflation in market prices of output, and π mit , the enterprise-specific inflation in input prices).33 Thus, the markup ratio, though assumed to be a industry-specific constant in 1989, is allowed to vary across firms and over time between 1980 and 1988. Li (1997) estimated it to be 0.158. In addition, µ 1 is normalized to be 1, µ j for material, machine, and chemical industries are estimated to be 0.41, 0.35, and 0.48. These estimates are used to compute M it . It is important to note that the µ j ’s are identified only up to the proportion with respect to µ 1 ; thus, if the markup ratio is 1 for the industry with the smallest markup ratio, the markup ratios for the rest of the industries are (1/0.35) * µ j , respectively. Appendix C. Construction of Variables for the 1990-94 Data Set Material inputs deflators. The 1990-94 data set contains information about the percentage increase in the usage of the two major material inputs. From these, we imputed the average yearto-year increases in material prices. For firms with missing information on these, we imputed by using the average of its industry-size-year cell, with industry being 36 two-digit manufacturing industries, and size being large, medium, and small. Price index of value added. Computed similarly to the method for the 1980s data using reported information on output prices, material input prices, value added, quantity of output and quantities of material inputs. Number of employees (L). Total number of employees minus the number of redundant workers. Capital Stock and Its Deflator. The first capital stock for each SOE is just the value of productive fixed assets in year one. For later years, we used the investment good price index from various issues of the China Statistical Yearbook to deflate incremental investments (i.e., the value of productive fixed assets at year t minus that at year t-1) and added them to the capital stock estimate for the prior period. 33

To see this, note that (see Li 1995) when Pit MCit and its lagged value are close to 1,

Pit P ≈ln( it ) + 1 , which implies MCit MCit Pit P P MCit P MCit ∆P ∆MCit − it −1 ≈ln( it ) = ln( it ) − ln( ) ≈ it − MCit MCit −1 Pit −1 MCit −1 Pit −1 MCit −1 Pit −1 MCit −1 The first term of the last equation is output inflation rate, and the second term is proxied by the inflation rate for intermediate inputs.

28

References Akerlof, George A., “The Market for ‘Lemons’: Quality Uncertainty and the Market Mechanism,” Quarterly Journal of Economics, 84(3), (1970), pp. 488-500. Akerlof, George, 1976, “The Economics of Caste and of the Rat Race and Other Woeful Tales,” Quarterly Journal of Economics, 90(4), (1976), pp. 599-617. Caprio Jr., Gerard, Izak Atiyas, James A. Hanson, and Associates. Financial Reform: Theory and Practice. New York: Cambridge University Press, 1994. Claessens, Stijn, Aslι Demirgüç-Kunt, and Harry Huizinga, 1997, “How Does Foreign Entry Affect the Domestic Banking Market?” mimeo, World Bank. Davis, Lance. “Capital Immobilities and Finance Capitalism: A Study of Economic Evolution in the United States, 1820-1920,” Explorations in Entrepreneurial History, Second Series, Vol. 1, No. 1, Fall, 1963. Davis, Lance. “The Investment Market, 1870-1914: The Evolution of a National Market,” The Journal of Economic History, Vol. 25, No. 3, September, 1965. Gertler, Mark and Rose, Andrew. “Finance, Public Policy, and Growth,” in Gerard Caprio Jr. et al. 1994, pp. 1348. Groves, Theodore, Yongmiao Hong, John McMillan, and Barry Naughton, “Autonomy and Incentives in Chinese State Enterprises,” Quarterly Journal of Economics, 109 (1994), pp. 181-209. Greenwald, Bruce C., and Joseph E. Stiglitz, 1986. “Externalities in Economies with Imperfect Information and Incomplete Markets.” Quarterly Journal of Economics, 101 (May), pp. 229-64. Guthrie, Douglas, 1997, “Between Markets and Politics: Organizational Responses to Reform in China,” American Journal of Sociology, 102 (5), pp. ?. Jaffee, Dwight M., and Thomas Russell, “Imperfect Information, Uncertainty, and Credit Rationing,” Quarterly Journal of Economics, 90(4), (1976), p. 651-66. Jefferson, Gary H. and Thomas G. Rawski, “Enterprise Reform in Chinese Industry,” Journal of Economic Perspectives, 8 (1994), pp. 47-70. Johnson, D. Gale, “The People’s Republic of China: 1978-1990,” San Francisco, CA: ICF Press, 1990. Lardy, Nicholas R. China’s Unfinished Economic Reform, forthcoming, 1998. Lee, Young, “Bank Loans, Self-Financing, and Grants in Chinese SOEs: Optimal Policy Under Incomplete Information,” Journal of Comparative Economics, 24 (1997), pp. 140-60. Li, Wei, “The Impact of Economic Reform on the Performance of Chinese State Enterprises, 1980-1989,” Journal of Political Economy, 105 (1997), pp. 1080-1106. Lin, Justin Yifu, Fang Cai, and Zhou Li, “Competition, Policy Burdens, and State-Owned Enterprise Reform,” American Economic Review, AEA Papers and Proceedings, 88 (1998), pp. 423-27. Liu, Zinan and Guy Shaojia Liu, “The Efficiency Impact of the Chinese Industrial Reforms in the 1980s,” Journal of Comparative Economics, 23 (1996), pp. 237-55. Lou, Jiwei. 1993. “Financial System and Policy,” paper presented to the Harvard Institute of International Development Conference, Cambridge, Massachusetts. McMillan, John, John Whalley, and Lijing Zhu, “The Impact of China’s Economic Reforms on Agricultural Productivity Growth,” Journal of Political Economy, 97 (1989), pp. 781-807. Qian, Yingyi. “Reforming Corporate Governance and Finance in China,” in Corporate Governance in Transitional Economies, pp. 215-250, edited by Masahiko Aoki and Hyung-Ki Kim, the World Bank, 1995. Rothschild, Michael, and Joseph Stiglitz, “Equilibrium in Competitive Markets: An Essay on the Economics of Imperfect Information,”90(4), (1976), pp.629-49. Salop, Joanne, and Steven Salop, “Self-Selection and Turnover in the Labor Market,” Quarterly Journal of Economics, 90(4), (1976), pp. 619-27. Spence, Michael, “Informational Aspects of Market Structure: An Introduction,” Quarterly Journal of Economics, 90(4), (1976), pp. 591-97. Stiglitz, Joseph E. 1985. “Credit Markets and the Control of Capital,” Journal of Money, Credit, and Banking, 17(2), pp. 133-52.

29

World Bank, Policy Options for Reform of Chinese State-Owned Enterprises: Proceedings of a Symposium in Beijing, June 1995, Discussion Paper #335, (edited by Harry C. Broadman), Washington DC, June, 1996. World Bank, The Chinese Economy: Fighting Inflation, Deepening Reforms, Washington DC, May, 1996. Wu, Yanrui, “Productivity Growth, Technological Progress, and Technical Efficiency Change in China: A Three Sector Analysis,” Journal of Comparative Economics, 21 (1995), pp 207-29. Xie, Ping, and others. 1992. Zhongguo de Jinrong Shenhua yu Jinrong Gaigei (Financial Deepening and Financial Reform in China). Tianjin: Tianjin People’s Press. Xu, Lixin Colin. “Control, Incentives, and Competition: The Impact of Reform in Chinese State-Owned Enterprises,” Economics of Transition, 8(1), 151-173, March 2000. Xu, Lixin Colin. “How China’s Government and State Enterprises Partitioned Property and Control Rights,” Economic Development and Cultural Change, 46(3), April, 1998, pp. 537-60. Yusuf, Shahid. “China’s State Enterprise Sector: Problems and Reform Prospects.” World Bank, mimeo, 1997.

30

Table 3: 1980s Bank Financing Access Access (dp/dX) (dp/dX) Probit: dp/dX Probit: dp/cX No. Obs. 2998 2998 R. Square 0.117 0.117 ln(yt-1) 0.253** 0.248** ( 0.031) ( 0.031) ln(capital-labor ratiot-1) -0.057 -0.063 ( 0.041) ( 0.041) ln(Number of employees) 0.337** 0.333** ( 0.034) ( 0.034) Coeff. of variation of ln(y) for 0.390** 0.400** industry-year cell of a firm ( 0.097) ( 0.097) Jiangsu Province 0.181** 0.201** ( 0.063) ( 0.064) Sichuan Province -0.008 0.010 ( 0.089) ( 0.090) Shangxi Province -0.294** -0.284** ( 0.070) ( 0.070) Central-government oversight -0.882** -0.889** ( 0.254) ( 0.256) Provincial-government oversight -0.301 -0.305 ( 0.240) ( 0.241) City-government oversight 0.172 0.167 ( 0.227) ( 0.229) County-government oversight 0.168 0.140 ( 0.250) ( 0.253) Firm-level wage elasticity 0.353** 0.357** ( 0.136) ( 0.136) Base profit retention rate -0.202** -0.197** ( 0.100) ( 0.100) Production autonomy 0.048 0.050 ( 0.056) ( 0.059) Managerial wage discretion 0.150* 0.154* ( 0.078) ( 0.078) Share of output under state plan 0.094 ( 0.070) The presence of new CEO 0.012 ( 0.063) Markup ratio

Access Share of Bank Share of Bank (dp/dX) Finan. Finan. Probit: dp/cX Tobit Tobit 2996 2998 2996 0.123 0.091 0.104 0.239** 0.111** 0.102** ( 0.031) ( 0.015) ( 0.014) -0.043 -0.055** -0.040** ( 0.042) ( 0.019) ( 0.019) 0.319** 0.118** 0.103** ( 0.035) ( 0.016) ( 0.016) 0.495** 0.145** 0.215** ( 0.099) ( 0.045) ( 0.045) 0.189** 0.029 0.019 ( 0.065) ( 0.030) ( 0.030) 0.028 -0.070* -0.049 ( 0.090) ( 0.042) ( 0.042) -0.257** -0.172** -0.150** ( 0.071) ( 0.035) ( 0.034) -0.831** -0.435** -0.376** ( 0.256) ( 0.123) ( 0.120) -0.263 -0.214* -0.169 ( 0.242) ( 0.115) ( 0.113) 0.157 0.099 0.094 ( 0.229) ( 0.108) ( 0.106) 0.073 0.074 0.027 ( 0.254) ( 0.120) ( 0.118) 0.329** 0.093 0.077 ( 0.136) ( 0.062) ( 0.061) -0.222** -0.054 -0.077 ( 0.101) ( 0.048) ( 0.048) 0.055 0.040 0.041 ( 0.060) ( 0.028) ( 0.027) 0.157** 0.072** 0.075** ( 0.078) ( 0.036) ( 0.035) 0.018 0.027 -0.029 ( 0.072) ( 0.033) ( 0.033) -0.010 -0.025 -0.041 ( 0.063) ( 0.030) ( 0.029) 0.515** 0.376** ( 0.102) ( 0.046) Standard errors in parentheses. *,** represent significance at the 10 and 5 percent levels. Other explanatory variables include year dummies, missing indicators for marginal profit retention rates, the output share under the state plan.

31

Table 4: 1980s Government Financing Access (dp/dX) Probit: dp/dX No. Obs. 2998 R. Square 0.077 ln(yt-1) -0.101** ( 0.030) ln(capital-labor ratiot-1) 0.278** ( 0.040) ln(Number of employees) 0.195** ( 0.031) Coeff. of variation of ln(y) for -0.477** industry-year cell of a firm ( 0.096) Jiangsu Province 0.414** ( 0.063) Sichuan Province 0.406** ( 0.087) Shangxi Province 0.282** ( 0.069) Central-government oversight 0.528** ( 0.244) Provincial-government oversight -0.017 ( 0.236) City-government oversight -0.049 ( 0.222) County-government oversight 0.016 ( 0.245) Firm-level wage elasticity -0.210 ( 0.136) Base profit retention rate 0.230** ( 0.098) Production autonomy 0.040 ( 0.056) Managerial wage discretion -0.141* ( 0.078) Share of output under state plan

Access (dp/dX) Probit: dp/dX 2998 0.082 -0.118** ( 0.030) 0.269** ( 0.041) 0.175** ( 0.032) -0.464** ( 0.097) 0.463** ( 0.065) 0.447** ( 0.088) 0.314** ( 0.070) 0.460* ( 0.246) -0.049 ( 0.237) -0.085 ( 0.224) -0.077 ( 0.248) -0.216 ( 0.136) 0.242** ( 0.099) -0.006 ( 0.059) -0.142* ( 0.078) 0.224** ( 0.069) 0.178** ( 0.062)

Access Share of Gov’t Share of Gov’t (dp/dX) Finan. Finan. Probit: dp/dX Tobit Tobit 2996 2998 2996 0.083 0.065 0.066 -0.113** -0.074** -0.072** ( 0.030) ( 0.016) ( 0.016) 0.261** 0.097** 0.094** ( 0.041) ( 0.021) ( 0.021) 0.180** 0.043** 0.044** ( 0.032) ( 0.016) ( 0.016) -0.507** -0.311** -0.326** ( 0.099) ( 0.051) ( 0.052) 0.469** 0.138** 0.140** ( 0.065) ( 0.034) ( 0.034) 0.439** 0.172** 0.169** ( 0.088) ( 0.045) ( 0.045) 0.299** 0.164** 0.159** ( 0.070) ( 0.036) ( 0.036) 0.443* 0.340** 0.334** ( 0.247) ( 0.130) ( 0.130) -0.063 0.099 0.096 ( 0.238) ( 0.127) ( 0.128) -0.077 -0.005 -0.001 ( 0.225) ( 0.121) ( 0.121) -0.047 -0.061 -0.049 ( 0.249) ( 0.133) ( 0.133) -0.203 -0.121* -0.116 ( 0.136) ( 0.073) ( 0.073) 0.256** 0.140** 0.144** ( 0.099) ( 0.051) ( 0.051) -0.006 -0.011 -0.010 ( 0.059) ( 0.031) ( 0.031) -0.145* -0.029 -0.031 ( 0.078) ( 0.041) ( 0.041) 0.259** 0.035 0.047 ( 0.071) ( 0.036) ( 0.037) The presence of new CEO 0.185** 0.084** 0.087** ( 0.063) ( 0.032) ( 0.032) Markup ratio -0.241** -0.085 ( 0.100) ( 0.053) Standard errors in parentheses. *,** represent significance at the 10 and 5 percent levels. Other explanatory variables include year dummies, missing indicators for the output share under the state plane, firmlevel wage elasticity, and . . .

32

Table 5: 1990s Bank Financing Access (dp/dX) Probit: dp/dX No. Obs. 2503 R. Square 0.037 ln(yt-1) 0.150** ( 0.030) ln(capital-labor ratiot-1) 0.011 ( 0.049) ln(Number of employees) 0.164** ( 0.032) Coeff. of variation of ln(y) for -0.268** industry-year cell of a firm ( 0.073) Jiangsu Province -0.045 ( 0.072) Sichuan Province 0.068 ( 0.079) Shangxi Province -0.127 ( 0.082) Central-government oversight -0.373** ( 0.143) Provincial-government oversight 0.100 ( 0.135) City-government oversight -0.065 ( 0.112) Index of autonomy 0.015 ( 0.031) Share of output under state plan -0.111 ( 0.111) The presence of new CEO 0.025 ( 0.069) Share of export 0.351 ( 0.303) Performance Contract 0.066 ( 0.054) Firm-level wage elasticity

Access (dp/dX) Probit: dp/dX 2503 0.037 0.149** ( 0.030) 0.011 ( 0.049) 0.163** ( 0.032) -0.260** ( 0.073) -0.055 ( 0.073) 0.074 ( 0.078) -0.123 ( 0.081) -0.373** ( 0.143) 0.103 ( 0.135) -0.059 ( 0.112) 0.015 ( 0.031) -0.102 ( 0.112) 0.024 ( 0.069) 0.364 ( 0.304)

0.058 ( 0.054)

Share of Bank Finan. Tobit 2503 0.021 0.072** ( 0.015) -0.010 ( 0.025) 0.045** ( 0.016) -0.132** ( 0.038) -0.032 ( 0.037) 0.014 ( 0.040) -0.044 ( 0.042) -0.099 ( 0.075) 0.080 ( 0.070) 0.032 ( 0.059) 0.001 ( 0.016) -0.054 ( 0.057) 0.036 ( 0.036) 0.293* ( 0.154) 0.039 ( 0.028)

Share of Bank Finan. Tobit 2503 0.022 0.071** ( 0.015) -0.010 ( 0.025) 0.045** ( 0.016) -0.128** ( 0.038) -0.037 ( 0.038) 0.017 ( 0.040) -0.041 ( 0.042) -0.099 ( 0.075) 0.080 ( 0.070) 0.036 ( 0.059) 0.001 ( 0.016) -0.049 ( 0.057) 0.035 ( 0.036) 0.293* ( 0.154)

0.032 ( 0.028)

Standard errors in parentheses. *,** represent significance at the 10 and 5 percent levels. Other explanatory variables include year dummies, missing indicators for the output share under the state plan, for firm-level wage elasticity.

33

Table 6: Government Financing 1990s

No. Obs. R. Square ln(yt-1) ln(capital-labor ratiot-1) ln(Number of employees) Coeff. of variation of ln(y) for industry-year cell of a firm Jiangsu Province Sichuan Province Shangxi Province Central-government oversight Provincial-government oversight City-government oversight Index of autonomy Share of output under state plan The presence of new CEO Share of export Performance Contract Firm-level wage elasticity

Access (dp/dX) Probit: dp/dX 2503 0.155

Access (dp/dX) Probit: dp/dX 2503 0.152

0.347** ( 0.071) -0.466** ( 0.110) 0.217** ( 0.060) -0.263 ( 0.195) -0.462** ( 0.198) 0.346** ( 0.154) 0.205 ( 0.165) 0.725** ( 0.283) 0.414 ( 0.289) 0.151 ( 0.255) -0.100 ( 0.066) -0.425** ( 0.215) -0.050 ( 0.146) 0.098 ( 0.726) -0.243** ( 0.114)

0.359** ( 0.071) -0.496** ( 0.110) 0.209** ( 0.060) -0.264 ( 0.195) -0.516** ( 0.198) 0.296** ( 0.151) 0.118 ( 0.161) 0.774** ( 0.282) 0.460 ( 0.289) 0.158 ( 0.255) -0.113* ( 0.067) -0.421* ( 0.216) -0.039 ( 0.146) 0.270 ( 0.704)

0.047 ( 0.118)

Share of Gov’t Share of Gov’t Finan. Finan. Tobit Tobit 2503 2503 0.141 0.139 0.346** ( 0.078) -0.477** ( 0.120) 0.201** ( 0.063) -0.275 ( 0.199) -0.467** ( 0.201) 0.291* ( 0.159) 0.192 ( 0.168) 0.768** ( 0.294) 0.415 ( 0.297) 0.172 ( 0.261) -0.100 ( 0.067) -0.396* ( 0.219) -0.069 ( 0.147) 0.218 ( 0.708) -0.232** ( 0.117)

0.355** ( 0.079) -0.509** ( 0.121) 0.194** ( 0.063) -0.277 ( 0.199) -0.530** ( 0.203) 0.243 ( 0.156) 0.097 ( 0.162) 0.811** ( 0.295) 0.466 ( 0.297) 0.182 ( 0.260) -0.118* ( 0.068) -0.393* ( 0.220) -0.058 ( 0.147) 0.385 ( 0.690)

0.102 ( 0.120)

Standard errors in parentheses. *,** represent significance at the 10 and 5 percent levels. Other explanatory variables include year dummies, missing indicators for the output share of the state plan and for firm-level wage elasticity.

34