Decentralization can complement market liberalization by strengthening incentives of agents to exploit local information in response to market signals. In China, however, banks centralized lending authority following financial reforms in the mid-1990s. We offer a new theory of financial decentralization in which centralization provides a credible commitment not to refinance bad projects by reducing available information. Using data from Chinese rural financial institutions, we empirically assess the determinants of decentralization and the likelihood of collateral seizure, strongly confirming the predictions of the refinancing model.
We conclude that the inability of financial systems to exploit local information in weak institutional environments may limit the efficiency of financial intermediation despite financial market liberalization.
JEL Codes: G2, D8, P3
Keywords: banking, decentralization, refinancing, transition, China
William Davidson Institute Working Paper 461
In recent years, financial liberalization has been a main focus of reform in developing countries and more recently in transition economies (e.g., Haggard and Lee, 1995). Financial liberalization transforms a heavily regulated system into a market-oriented one by reducing barriers to entry, reducing government influence over credit allocation, and increasing reliance on market-determined interest rates. Conventional wisdom holds that decentralization of control rights goes hand-in-hand with market liberalization. With greater decision-making authority, local managers have greater incentives to exploit local information in response to market signals, which increases the efficiency of resource allocation. Effective use of local information may be especially important when there are large information asymmetries between central and local managers. In banking, this is likely to occur in weak institutional environments in which credit ratings, high quality appraisal and auditing services, standardized reporting systems, and well-developed legal systems are absent.
Surprisingly, in China most banks responded to financial reforms in the mid-1990s by centralizing rather than decentralizing control rights. This contrasts with the history of key reforms in China’s agricultural and industrial sectors that decentralized decision-making authority to households and firm managers as the role of markets increased (Naughton, 1995). But it is consistent with evidence from developing and transition economies that financial liberalization and decentralization may have unintended effects if underlying institutional factors are left unaddressed (Cho, 1986; Buch, 1996; Koford and Tschoegl,
1997; Schmidt, 1998).1
This paper provides a new theory of financial decentralization that explains why decentralization may not always be desirable, especially in developing and transition economies. The model’s key insight is that by improving local information, decentralization can reduce the ability of lenders to credibly commit not to refinance bad projects, the effect of which is to soften the budget constraint of borrowers. In the context of a corporation, Cremer (1995) shows that more information may hurt the principal’s ability to refuse renegotiation. Berglof and Roland (1998) find that when liquidation costs are sufficiently high, lenders may lack the credibility to liquidate financially distressed projects, leading to soft budget constraints. In this paper, we present a model in which centralization can provide a credible commitment not to refinance bad projects, which improves project performance and loan repayment by increasing the effort incentives of firm managers. In China, financial centralization increased as concerns about refinancing grew in the mid-1990s when the economy slowed and increasing numbers of firms encountered financially difficulties.
Our argument uses the key insight from Dewatripont and Maskin (1995) to come to an exactly opposite conclusion. They argue that decentralization in the form of a division of large banks into small ones can serve as a commitment device that helps lenders harden the budget constraint of borrowers. This is because, unlike large banks, small banks are incapable of refinancing ongoing projects independently. However, other potential lenders may be unwilling to refinance projects when they have poor information about firm quality.
Entrepreneurs with questionable projects will anticipate that refinancing is less likely and will not seek financing in the first place, or will exert greater effort to make projects successful. In both our model and that of Dewatripont and Maskin, credible commitment not to refinance is achieved by taking lending authority out of the hands of bank managers who have better information. However, in our case this is accomplished by centralizing lending authority rather than dividing large banks into small banks, which Dewatripont and Maskin describe as “decentralization.”
The refinancing problem is likely to be particularly important in transition economies because poorly developed institutions make liquidation of collateral costly, so that borrowers are more likely to anticipate refinancing loans if projects go bad. In China, “fishing” projects are very common in which outstanding debt serves as bait for attracting additional loans because banks cannot get the old loans back unless they are willing to make more new loans (Li and Li, 1996). This dynamic may help explain the growth in debt-asset ratios in both state-owned enterprises and rural collective enterprises in the 1990s (Lardy, 1998; Park and Shen, 2001). Lack of accounting standards often observed in transition economies may exacerbate the problem (Buch, 1996). For example, if banks are able to roll over loans to hide repayment problems, then liquidation of projects may lead directly to large accounting
losses which may be viewed as unacceptable to bank mana gers. This makes refinancing more attractive, undermining the credibility of threats to liquidate projects once defaults occur.
In this paper, we use bank-level data on managerial decision-making authority in a large transition economy to test whether theory can explain observed heterogeneity indecentralization of lending authority in financial institutions. The unique data set was collected in surveys of rural financial institutions, enterprises, and local government officials conducted by the authors in 1998 in southern China. Surprisingly, we found no existing empirical research on the validity of different theories explaining financial decentralization. The question is important for validating the assumptions of existing theory, and practically for assessing the efficiency of financial resource allocation.
The ongoing financial transition in China offers an excellent setting for empirical study of financial decentralization. China is an interesting case because banks recently became commercialized but supporting institutions are not fully developed, leading to rich variation across space in the extent of decentralization. Financial reform in China aims at transforming financial institutions from government-run banks to independent financial intermediaries (Qian, 1994; and Lardy, 1998). Important aspects of China’s financial reform in the mid-1990s were the introduction of bank competition among state-owned banks, the strengthening of profit incentives for managers, and the transfer of policy loans to newly established policy banks (Park and Sehrt, 2001). Individual banks have been allowed to
decide for themselves whether or not to decentralize lending authority to local bank managers.
In China and elsewhere, agency problems provide an alternative explanation for centralized decision-making. Local bank managers may collude with or be influenced by local government leaders, who, as social planners of local communities, internalize not only the economic benefits of running firms but also non-economic ones such as enterprise and employment creation, and potential tax extraction (Svejnar, 1990; Jin and Qian, 1998).2
Policy lending and soft budget constraints are a notorious problem plaguing financial institutions in transition economies (Kornai, 1986). However, we study decision-making in banks at the lowest administrative level in China, the township, where such problems are considered to be less serious (Qian and Roland, 1998). In China, financial reforms strengthened managerial profit incentives, and our survey found that policy influence on lending fell substantially during the mid-1990s. The timing of centralization thus is not consistent with the expected reduction in moral hazard problems over time (Park and Shen, forthcoming).3
The problem we focus upon in this paper is the allocation of lending authority between upper level management, the center, and lower level management, the local. In section 2, we present two models, a costly information model and a refinancing model, to illustrate how different factors affect optimal decentralization in a two-tier hierarchy. This yields a set of theoretical predictions on the effect of key parameters on the decentralization decision. Section 3 presents the evidence from China. We first introduce the data, then describe financial decentralization and commercialization in China’s rural financial sector, discuss the estimation strategy, and present the results of our empirical a nalysis of the
determinants of decentralization and the propensity for bank managers to seize collateral.
Section 4 concludes.
Our main findings are that the desire to credibly commit against refinancing theoretically can explain centralization, and that this explanation is strongly supported by the Chinese data. We conclude that in imperfect institutional environments, centralization can help resolve agency and commitment problems, although likely at substantial cost in terms of lost information and reduced incentives of local managers. This outcome suggests that rapid commercialization of banking systems in transition economies will not automatically lead to substantial improvement in financial intermediation unless deeper institutional
reforms occur as well.
We present two complementary models of decentralization. The first, which we call the costly information model, is a one-period model that illustrates how the decision to decentralize is affected by the benefits and costs of using local information. Simple extensions examine how collateral, financial competition, and agency problems (i.e., local government influence over lending) affect optimal decentralization. We then present amodel with multiple periods to examine how the refinancing problem alters key predictions about the model parameters. Both models assume that central managers are profit
2.1. The Costly Information Model
We consider a local economy with N firms whose types are uniformly distributed.
Afirm’s type q Î[q ] 0 1 , is the probability of project success, where q0 is the least profitable firm type in the economy. We assume that all firms demand loans and that firms are only able to repay loans when their projects are successful. Therefore, q is also the expected rate of repayment. For simplicity, we assume that all firms require one dollar of investment to operate.
download file disini: