Are China's Financial Reforms Leaving the Poor Behind - Harvard ...

Draft August 2001Are China’s Financial Reforms Leaving the Poor Behind?Loren BrandtDepartment of Economics, University of TorontoAlbert ParkJFK School of Government, Harvard UniversityDepartment of Economics, University of MichiganWang SanguiInstitute of Agricultural Economics, Chinese Academy of Agricultural SciencesPaper prepared for the conference on“Financial Sector Reform in China”JFK School of Government, Harvard UniversitySeptember 11-13, 20012

Are China’s Financial Reforms Leaving the Poor Behind?1. IntroductionSince economic reforms began, the Chinese state has shifted the locus of controlover the allocation of resources from the fiscal to the financial ledger (Sehrt, 1999).Under the planning system, funds were mobilized by setting prices to concentrate profitsin state-owned industry (Naughton, 1995). During the reform period, market competitioneroded the profitability of SOEs, and the government struggled to establish an effectivetaxation system to generate revenue (Wong, Heady, and Woo, 1995). However, byasserting its control over the banking system, the state was able to continue directingnational resources to support its most important political constituency--urban workers instate-owned enterprises (Brandt and Zhu, 2000). This strategy was facilitated by therapid growth of individual private savings in state-owned banks, which have come todominate China’s national savings (Kraay, 2000). 1The financial system has increasinglybecome the key mechanism by which the government attempts to realize its distributionalgoals.Government allocation of resources in China as well as in many other developingcountries has been biased against rural areas and in favor of urban areas, undoubtedlywidening urban-rural disparities (Putterman, 1992; Park and Rozelle, 1998; Lipton,1984). Transfers through the banking system are only one aspect of this urban bias,which also includes extraction of taxes, fees, pricing policy, procurement quotas, and1 According to official data, most of the increase in deposits was by urban residents, although the urbanruraldistinction is not always clear because rural residents may deposit funds in any bank. In 1998, thepercentage of individual deposits from rural areas (defined as deposits in Rural Credit Cooperatives) was20 percent, compared to 26 percent in 1990. Urban deposits in 1998 were 8.1 times that in 1990, ruraldeposits were 5.7 times as great. On a per capita basis, the deposits of urban residents in 1998 were nearly10 times that of rural residents. Individual deposits accounted for 55 percent of all deposits in all financialinstitutions in 1999, with the second largest category being enterprise deposits (34 percent).3

analyze changes in the degree of financial intermediation by rural financial institutions inrich and poor areas. This is an important question because it addresses the keyefficiency-equity tradeoff of market reform. It is now well-established that financialintermediation plays an important role in growth, so that current allocations of financialresources shape future patterns of inequality (Gertler and Rose, 1996; Levine, 1997; Kingand Levine, 1993; Levine and Zervos, 1998; and Rajan and Zingales, 1998). Givenrapidly rising regional income inequality in China and the concern shared by many thatcredit access is a key constraint to development in China’s poorest areas, betterunderstanding of the distributional consequences of financial reform takes on addedimportance. This is especially true given the lack of any systematic empirical study ofthis issue.In this paper, we study changes in financial intermediation and financialperformance over time using a unique data set from surveys of rural financial institutionsconducted by the authors in two rich, coastal provinces (Zhejiang and Jiangsu) and in twopoor, interior provinces (Sichuan and Shanxi) in 1998 and 1999. The paper focuses onthe financial role played by the two main rural financial institutions in China and ourresults should be qualified by an awareness that we do not consider informal borrowingor lending and saving in other formal or informal financial institutions. However,informal networks tend to be segmented and fragmented, and informal organizations thatreach large scale have not been tolerated. We therefore believe the role of ABC branchesand RCCs to be pivotal in achieving effective financial intermediation in rural areas.Focusing on rural financial institutions has several motivations. First, rural Chinaremains home of 70 percent of China’s population. Rural financial institutions are the6

highest in Zhejiang (4470 yuan), followed by Jiangsu (3650 yuan), Sichuan (2060 yuan)and Shanxi (1558 yuan). 5In each province, a representative sample of counties was selected from differentregions of the province and in areas with different levels of economic and industrialdevelopment. Eight counties were selected in Jiangsu, seven in Zhejiang, and six in bothSichuan and Shanxi. Within each county, four townships were randomly selected afterstratifying by industrial output per capita. In total, 108 townships in 27 counties werechosen, and all ABC branches and RCCs in these townships were surveyed. Not alltownships have ABC branches so the sample of ABC branches is smaller than the sampleof RCCs. In each county, separate interviews also were conducted with the county ABCbranch and the county RCC association, the headquarter banks for township ABCbranches and RCCs. In each township, additional interviews were conducted withgovernment officials and a sample of enterprise managers.Most of the information was garnered from face-to-face interviews with bankmanagers, and available historical data on assets, liabilities, income, and expenditureswere copied from accounting books. The survey focused on the period between 1994 and1997. Because the surveys in Sichuan and Shanxi were conducted a year later, anadditional year of data was collected (for 1998).The data have several unique features. First, the data cover a period of majorreform in China’s financial system and in the operation of China’s rural financialinstitutions. Second, the surveys provide for the first time very detailed information onfinancial performance, regulation, and governance of China’s rural financial institutions,5 The 1998 China Statistical Yearbook reports 1997 rural income (living expenditures) per capita of 3684(2839) yuan in Zhejiang, 3270 (2488) yuan in Jiangsu, 1681 (1440) yuan in Sichuan, and 1738 (1145) yuanin Shanxi.9

which is of great interest to those concerned about China’s rural credit markets andChina’s broader financial system reform. Third, the surveys in rich and poor provincesprovides an excellent basis for a regional comparison of financial performance and bankbehavior, facilitating a study of the distributional consequences of financial reform.3. Financial reform and China’s rural financial institutionsChina’s two main rural financial institutions, the Agricultural Bank of China(ABC, or nongye yinhang) and Rural Credit Cooperatives (RCCs, or xinyongshe),accounted for 14 and 13 percent of deposits in all financial institutions in 1998 and 16and 10 percent of all lending. Thus, together they account for more than one fourth of alldeposits and loans in China. 6The ABC is one of China’s four specialized banks, with thelargest branch network among specialized banks, extending to most but not alltownships. 7Most lending is working capital loans for state commercial enterprises, butsignificant shares also are lent to township and village enterprises (TVEs), 8 and toagriculture (including households). The RCCs are cooperatives in name only (not ingovernance). Since 1996, they have been under the administrative supervision of thePeoples Bank of China (previously under the ABC). RCCs are the only financialinstitutions with branch outlets extending to nearly all townships as well as many6 Until 1999, rural cooperative foundations (RCFs, or nongcun hezuo jijinhui) could also be found in manytownships, especially in richer areas (38 percent of all townships in 1996). The RCFs were quasigovernmentorganizations under the administrative supervision of the Ministry Agriculture. In 1996, RCFdeposits were estimated to be one ninth that of RCCs. A national survey found that 24 percent of RCFloans went to TVEs and 45 percent to households. RCFs often had significant involvement of townshipofficials, lack legal status as financial institutions, and often charged effective interest rates above regulatedrates. Following a 1998 State Council circular, they were dissolved in 1999.7 The Agricultural Development Bank of China (ADBC, or nongye fazhan yinhang) separated from theABC as one of three policy banks established in 1994, with branches at the county level. In 1996, theADBC accounted for 10 percent of national lending. Ninety percent of loans were for agriculturalprocurement, mainly grain.8 Following Chinese conventions, we define TVEs to include both collective and private enterprises.10

villages. They are by far the most important source of formal credit in rural areas,providing over half of all loans to rural enterprises but lending also to households.Before examining the performance of local ABC branches and RCCs, we firstintroduce relevant aspects of the political, economic, and regulatory environment forlending. First, although most policy loans in the rural sector (mainly for agriculturalcommodity procurement) were transferred to the Agricultural Development Bank ofChina (ADBC), ABC branches and RCCs were not immune to local political pressures,especially from local government officials keen to support revenue-generating industrialprojects. ABCs also carry a small portfolio of explicit policy loans (including povertyalleviation loans) and may be asked to support state marketing agencies. Nonetheless,most managers in both institutions report that profitability has been the most importantcriteria for approving loans even by 1994. This commercial orientation increasedsteadily over time, so that by 1997 profits had overwhelming become the main statedconcern of bankers (Park and Shen, 2000). In recent years, managerial pay has beenmore closely tied to profits through a higher share of income paid in the form of bonuses.Overall, we should expect less policy lending in rural financial institutions than in otherlarge specialized banks and policy banks, and for policy lending to decline over time.Second, despite efforts to increase profit incentives, most bank managers operatein a highly circumscribed regulatory environment. All deposit and lending interest ratesare controlled at below-market clearing levels, adjustable by small margins. Also, interbankloans are regulated both in terms of quantity and price. After a frenzy ofuncontrolled inter-bank lending in the early 1990s, the government shut down the interbankmarket in 1993. Currently, most transfers of financial resources are through vertical11

deposits of surplus funds in higher level branches of the same bank. Thus, townshipABC branches deposit funds with the county ABC branch, and RCCs deposit funds withthe county RCC association (lianhehui). The ABC is a national bank, but transfersamong RCCs typically occur only among branches within the same counties. Many bankmanagers have an approved credit quota based on a county plan or on an agreementlinking lending amounts to the amount of deposits. In most cases, the quota can beadjusted if the manager can justify the need for more funds. Managers also are oftenrestricted in the size of loans that can be issued without approval from the county ABCbranch or RCC association, and local discretion in individual loan decisions hasdecreased over time (Shen and Park, 2001). The county banks play an important role inregulating and supervising township bank branches, although the township banks remainindependent accounting units. Managers have almost no options for fund use other thanlocal lending and deposits in higher level banks. The internal interest rates for suchdeposits (usually set very low), the credit quota, and the riskiness of local projects thusare major factors affecting loan allocation decisions.In Table 1, we provide summary balance sheet information from the ruralfinancial institutions we surveyed. Dependence on deposits as a source of fundsincreased from 58 percent in 1994 to 64 percent in 1997, suggesting a decline rather thanan increase in interbank lending. This dependence is particular high, and growing in poorareas, suggesting that deposit mobilization may play an important role in determining theamount that financial institutions in such areas can lend. The overall size of the financialinstitutions, as measured by total assets, is nearly ten times as great in the richest quartile12

of regions than in the poorest quartile. The rural financial institutions in Zhejiang areparticularly large, and ABC branches are considerably bigger than RCCs.Most rural lending is to firms, not farmers. Seventy-four percent of outstandingloans went to firms in 1994; by 1997 the share had fallen slightly to 71 percent (Table 1).However, in poorest quartile where presumably there are few viable enterprises, less than40 percent of loans go to enterprises. RCCs, which in general lend more to firms, 83percent in 1994, are reducing their loans to firms while ABC branches are increasingthem. There is also a movement in coastal areas to increase lending to privateenterprises, due in large part to rapid privatization of collective enterprises during themid-1990s (data not presented).For Sichuan and Shanxi, we have a more detailed breakdown of lendingcategories (Appendix Table 1). Enterprise lending is predominant in Sichuan whilehousehold lending is dominant in Shanxi. Household lending also comprises a muchhigher share of lending in poorer townships (over two thirds in the poorest quartile in1998), no doubt due in part to the lack of enterprises in such areas. Except for in therichest quartile, a greater share of loans went to households in 1998 than in 1994. In therichest quartile, lending in 1998 is marked by greater firm lending. Also, loans toagriculture and other loans declined in poor townships but not rich townships. Finally,RCCs lend a much greater share of funds to enterprises than ABCs. This is likely due tosignificant lending by ABCs to marketing agencies such as supply and marketingcooperatives (gongshaoshe) and grain trading companies (liangshi maoyi gongsi).Overall, we observe a greater willingness of financial institutions to lend tohouseholds in poorer townships, and that this relative tendency has become greater over13

time. This is not to say that households found it easier in 1997 than in 1994 to obtainloans, as total lending declined over this period. These findings are consistent with otherresearch showing that a relatively high proportion of farmers in poor regions have at leastsome access to formal credit (Park and Wang, 2000). 9Ironically, it may be households inricher areas that have greater difficulty obtaining individual loans because banks so preferlending to enterprises. However, richer households have greater savings (and deposits)and so generally are more able to self-finance common expenditures such as for fertilizerand school fees. From this perspective, household credit demand does not necessarilyrise with income, unless rural residents shift to self-employment activities that requiregreater financing.4. Changes in financial intermediation in rich and poor areas from 1994 to 1997Financial intermediationTo assess whether China’s financial reforms are leaving the poor behind, werequire a financial performance measure that will enable us to compare the extent offinancial intermediation in rich versus poor areas over time. Financial development isfrequently measured by the amount of financing normalized by the value of economicoutput, e.g., the ratio of total loans to GDP. In this paper, we restrict attention to formalfinancial intermediation in the form of credit, excluding equity financing and informal9 One piece of supporting evidence for unmet household credit demand was the success of RuralCooperative Foundations (RCFs, or nongye hezuo jijinhui) in the 1990s. These quasi-formal institutions,under the administrative jurisdiction of the Ministry of Agriculture, tended to be established in moreindustrial areas and lent mainly to households. However, there is also evidence that they merely crowdedout household lending by incumbent RCCs and so did not increase overall household lending (Park, Brandt,and Giles, 2000). RCFs were banned by the government in 1999.14

orrowing. 10A consistent finding in cross-country studies is that the level ofdevelopment as measured by GDP per capita is positively associated with the degree offinancial depth (Gertler and Rose, 1996). Park and Sehrt (forthcoming), however, findthat this relationship does not hold for provinces in China, implying regional barriers toefficient intermediation.Because GDP is not calculated at the township level in China, we adopt adifferent measure--the ratio of outstanding loans per capita to rural income per capita.Income per capita is a net rather than gross measure and does not include retainedearnings in enterprises or other income that does not accrue to households. It is expectedto be highly correlated with, but to rise less than proportionally with GDP. Relative to ameasure using GDP, this could lead us to overstate the relationship between the level ofdevelopment and depth of financial intermediation.We also make a second adjustment. Many banks in China have large amounts ofoverdue loans with dubious repayment prospects. If loans that should be written off aslosses are kept on the books, they can present a misleading picture of the true extent offinancial intermediation and net flow of funds. Fortunately, our survey providesinformation on overdue, or non-performing loans. We can thus divide total outstandingloans into “performing” and “non-performing” loans. Even though some overdue loanscould be providing a meaningful intermediation function, we consider the amount ofperforming loans relative to income to be a superior measure of effective financialintermediation. We report financial intermediation rates using both total outstandingloans and performing loans (Table 2).10 During this period, TVEs could not be listed in capital markets. Informal borrowing is probably at leastas important as formal financing for households in rural areas (Brandt, Giles, and Park, 1997) but relativelyunimportant for firms.15

We find that from 1994 to 1997 the extent of financial intermediation stayed thesame in terms of total loans but actually fell in terms of performing loans. Even usingtotal loans, the data show that richer provinces (Zhejiang and Jiangsu) and richertownships witness no clear change from 1994 to 1997 but that the poorer provinces(Sichuan and Shaanxi) and townships show a slight reduction in intermediation.Comparing the ratio of performing loans to income in 1994 and 1997, these differencesare much more pronounced, with the fall in intermediation in poor areas quite sharp(from 0.30 to 0.23 in the poorest quartile), and continuing into 1998. The annualreduction in performing loans relative to income is 13 percent in poorest quartile areasbut only one percent in the richest quartile. There is some evidence that intermediationby ABCs is increasing but that by the RCCs is decreasing.To see clearly the different factors that influence the extent of effective financialintermediation in a locality, we decompose our financial intermediation rate in thefollowing way:PL PL L D= x x , (1)Y L D YwherePL is performing loans,Y is income,L is total outstanding loans, andD is deposits.The rate of effective financial intermediation is the product of the share of performingloans in total loans, the loan to deposit ratio, and the ratio of deposits to income. Thesethree variables measure loan performance, net fund flows, and deposit mobilization,16

espectively. In the following sections, we review the empirical evidence on changes ineach of these factors, highlighting differences between rich and poor areas. We thensummarize the relative importance of each factor in explaining changes in intermediationover time.Loan PerformanceNon-performing loans include three categories: overdue, inactive (overdue formore than 2 years), and dead (overdue for more than 3 years or confirmed for otherreasons to be unrecoverable). Most non-performing loans fall in the first category butnearly half fall in the two more delinquent categories. Overall, the share of outstandingloans that are overdue in the full sample rises from 17 percent in 1994 to 24 percent in1997 (Table 3). However, the overdue rates are substantially higher in poorerprovinces. 11In Sichuan, overdues increase from 25 to 32 percent, and in Shanxi, theyrise from 30 to 40 percent. In 1998, the percent of outstanding loans overdue stabilizes to28 percent in Sichuan but continues on an upward trajectory in Shanxi, reaching 48percent. Such repayment rates are financially unsustainable if banks are to remainsolvent without major infusions of outside capital. And they may understate the truemagnitude of the repayment problem if banks roll over loans through “evergreening” orotherwise conceal repayment problems (Lardy, 1998).These percentages of non-performing loans appear comparable to the 22 percentfigure cited by Lardy (1998) for China’s four large state-owned banks in 1995. We alsocan compare the overdue rates to those found in other surveys. The 1996 national village11 The full sample average reflects the greater number of financial institutions surveyed in the coastalprovinces and the greater amount of funds per financial institution.17

survey found an average overdue rate in 1995 of 21 percent and the CRPS survey foundan average overdue rate in 1996 of 39 percent. These results are roughly consistent withour estimates of the non-performing loans in the 4 sample provinces, and reinforce ourfinding that repayment problems are significant everywhere but substantially worse inpoorer regions.However, although the levels of overdues are greater in poorer townships, the rateof increase in overdues is as rapid in Zhejiang, 8 percent, and in the richest quantile oftownships, 12 percent, as in poorer provinces and townships. As a percentage of theinitial share of bad loans, the increase is most rapid in the richest quartile, 32 percent peryear, followed by the poorest quartile at 14 percent. The bad loan problem striking ruralfinancial institutions in China’s coastal areas is the result of poor bank governance(incentives of managers) in some areas, the deteriorating profitability of rural enterprises,and the increasing frequency of strategic default on loans (Park and Shen, 2000). Theseproblems may be related to overall slower growth, intensifying market competition, andunrealistic over-investment in the early 1990s.We also note that the deterioration in loan performance appears to be greater forRCCs than ABC branches, and that based on data from Sichuan and Shanxi thedeterioration is very sharp for TVEs but nonexistent for household loans (typically foragricultural production or self-employment). In Sichuan and Shaanxi, by 1997 nonperformingloans accounted for 38 percent of outstanding TVE loans compared to 23percent in 1994. Meanwhile, overdue household loans were stable at 32 and 33percent(Table 4).18

Are funds flowing out of poor areas?We measure the net flow of funds by the ratio of outstanding loans to deposits ineach financial institution. A financially self-sufficient bank should have a loan-depositratio less than one because of reserve requirements (13 percent nationally but higher forsome local branches, especially RCCs before 1996 when they were regulated by ABCs).Higher loan-deposit ratios imply that more local resources are being lent locally ratherthan being intermediated for use in other areas, such as through bank deposits inheadquarter banks or other banks, or through inter-bank lending. 12If funds tend to flowout of poor areas, we would expect such areas to have lower loan-deposit ratios. Banksborrowing funds from their headquarters or from other banks could have loan depositratios exceeding one.Table 4 presents loan-deposit ratios broken down by province, income per capitaquartile, and bank type. Overall, more funds flowed out of rural areas in 1997 than in1994, as evidenced by the fall in the average loan-deposit ratio for the full sample from0.71 to 0.64. For all financial institutions nationwide, the ratio of loans to deposits fellfrom 1.01 to 0.91 over this same period (State Statistical Bureau, 1999). 13Thus, ruralareas consistently are net contributors of resources for lending elsewhere, but the degreeof this bias did not shift significantly in the mid-1990s. These loan-deposit ratios aresimilar to those found in a 1996 national survey of 94 RFIs in six provinces and in the1997 China Rural Poverty Survey conducted in six poor national poor counties in12 In addition to deposits, funds can come from the bank’s own capital, net borrowing, and other sources.Because we are unable to clearly separate other sources and uses of funds into those that originate or areinvested locally and those that are borrowed from or invested elsewhere, we treat deposits and loans as theonly purely local source and use of funds. A better measure of net outflow would account for all sourcesand uses of funds.13 The high loan-deposit ratios for all banks and ist subsequent decline largely reflect relending by thePeople’s Bank of China to state-owned banks, which through the 1980s accounted for as much as 30percent of funds, but since the mid-1990s has tightened considerably.19

different provinces, both conducted by the authors (Table 5). The national village surveyfound a sharp reduction in loan to deposit ratios from 0.94 in 1988 to 0.68 in 1994(recovering to 0.73 in 1995), while the CRPS finds a smaller but still large fall from 0.73in 1988 to 0.61 in 1994 and a continued decline to 0.54 in 1996.The loan-deposit ratios in poorer townships are lower than in richer townships(Table 4), and the mean ratio in the CRPS is lower than in the national village survey(Table 2). However, the differences among provinces are not significant in 1994. Poorareas likely have a more difficult time attracting outside funds and bank managers mayalso have a greater reluctance to lend their own funds locally. Table 1 reports the shareof total funds that are deposits. Two strong patterns emerge. The first is that poorerareas are much more reliant on their own deposits than richer areas. The second is thatall areas are more reliant on own deposits in 1997 than in 1994. Credit tightened in thefinancial retrenchment period beginning in 1994, and a new emphasis was placed on loanquality.Among provinces, the decrease in the loan-deposit ratio is largest in Jiangsu (14percent), while Shanxi’s loan-deposit ratio actually increases slightly (by 3 percent, Table4). It appears that over time funds are less, not more likely, to flow into richer provinces!The pattern is even more striking if we look at the change in loan-deposit ratios byincome quartile. The loan-deposit ratio fell by fiver percent in the richest quartile andgrew by five percent in the poorest quartile. Are projects better in poorer areas? Or arebranches in richer townships being “taxed” by state financial institutions to mobilizefunds for use elsewhere? The latter explanation hardly seems consistent with China’scommercialization reforms.20

Deposit and Income GrowthBoth deposits and incomes grew rapidly in rural China from 1994 to 1997, aperiod of higher commodity prices and healthy output growth. Deposits grew by 12percent per year and rural incomes grew by 11 percent (Table 6). While average growthin real incomes was similar across provinces over the three-year period (1994 to 1997),poorer townships grew substantially faster than richer townships. Income per capita inthe richest quartile of townships increases by 7 percent per year while in the poorest itgrew by 13 percent. This is consistent with the convergence hypothesis of theneoclassical growth model, and suggests we may be mistaken to assume that goodprojects do not exist in poor areas. However, despite lower income growth, the depositgrowth rate is higher among the higher income groups (13 percent in the richest quartileand 8 percent in the poorest). These differences are consistent with a greater propensityto save with higher income. Loans grow even more slowly than deposits, and nonperformingloans grow faster in both the richest and poorest quartiles. The poorest areasare thus characterized by high income growth (13 percent), low deposit growth (8percent), even lower loan growth (7 percent), and lower growth still in performing loans(minus one percent) because of the rapid growth of non-performing loans. Realperforming loans for the poorest quartile is slightly negative even though real incomesare growing fast. Overall, the results suggest failing intermediation in poor areas drivenby a failure of available resources to keep up with income growth and high amounts ofnon-performing loans. Richer areas have more room to maneuver because they can moreeasily mobilize deposits, but here too, poor loan performance is undermining the ability21

of banks to provide new loans. The richest quartile has moderate income growth (7percent), high deposit growth (13 percent), moderate total loan growth (9 percent) butlow growth in performing loans (2 percent).The very uneven growth in deposits in rich and poor areas explains the apparentlyinconsistent earlier findings that the loan-deposit ratios are falling in rich areas but notpoor areas, but that financial intermediation (loans/income) is holding steady in rich areasbut falling in poor areas. The change in fund flow or loan repayment is not obviouslyworse in poor areas, but deposit growth is not keeping up with income growth, whereasin rich areas deposit growth is far exceeding income growth.Decomposing the sources of change in financial intermediationFrom (1), it is straightforward to decompose the rate of change in effectivefinancial intermediation as the sum of the growth rates of its components:PLlog(Y9797PL94PL97PL94L97L94) − log( ) = {log( ) − log( )} + {log( ) − log( )} (2)Y L L D D9497{log( D97 ) − log( D94)} + {log( Y97) − log( Y94)}949794The difference in logs for each variable is equal to the exponential growth rate, which canbe annualized by dividing by the number of years. Table 7 summarizes the growth ratesin (2) and reports the percentage of the overall rate of change in effective financialintermediation accounted for by each factor.The results confirm the earlier findings that much of the relative deterioration ofeffective intermediation in poor areas is due to poor loan repayment and a failure to22

mobilize deposits even as income rise. Of the 13 percent annual fall in effectiveintermediation rates in the poorest quartile during the period, loan performance accountsfor 10 percent, the loan to deposit ratio–5 percent (it increases), and increasing depositsrelative to deposits accounts for 5 percent.We also estimate a set of regressions of financial performance indicators on thelog of real income per capita, comparing the results for 1994 and 1997 (Table 8).Because of our concern that differences between rich and poor provinces may be drivenby differences in provincial policies or other unobserved provincial characteristics ratherthan income differences, we run the regressions with and without controls for provincialdifferences. In all regressions, we control for differences in bank type. To control forboth provincial differences and bank type, we include dummy variables which areinteractions of province and bank type dummies. This specification is very general inthat it allows bank type to have different effects in different provinces, and the effects ofincome per capita are identified by variation among banks of the same type in the sameprovince.The results confirm many of the previous findings and help to quantify therelationships. Controlling for provincial differences, higher income per capita issignificantly associated with higher intermediation rates, measured by either performingloans or total loans divided by income. This positive association is greater in 1997 thanin 1994. A one percent increase in income increased the effective intermediation rate by0.15 percent in 1994 and 0.21 percent in 1997. However, dropping provincial controls,the relationships are no longer statistically significant.23

We also examine the components of intermediation rates. The effect of incomedifferences on loan performance increases from 1994 to 1997. With provincial controls,a one percent increase in income per capita reduces the share of non-performing loans by0.08 percent in 1994 and by 0.21 percent in 1997. When we drop controls for provincialdifferences, we find roughly similar results. Second, there is a consistent negativerelationship between income level and the loan-deposit ratio, but this shows upsignificantly only in the specification without provincial controls, suggesting that thesedifferences are due in part to differences in provincial characteristics other than averageincome (e.g., quality of alternative investment projects, financial regulation). Fourth,higher income per capita is associated with a more than proportionate increase in depositsper capita, with a one percent increase in income per capita associated with depositincreases of 1.35 percent in 1994 and 1.79 percent in 1997. Without provincial controls,the effects are smaller but still significant. Finally, we note that there is a shift towardgreater lending to firms in richer townships and move away from household lending, andthis sensitivity to income also increases over time.Comparing the results of specifications controlling and not controlling forprovincial differences suggests that the inequities associated with income differences areeven greater within provinces than between provinces. In other words, relative to a richtownship in the same province, a poor township is even worse off than relative to anequally rich township in another province. This may be because in another province(most likely richer) fewer deposits and loans are likely to be made in the rural financialinstitution because of greater financial competition and better alternative projects. In thesame province, rich townships get more deposits and most funds are lent locally, and the24

differences are even more stark. Provincial differences appear to be muting the measuredresponse of per capita deposits and loans to income, which may be due to greaterfinancial competition in these areas. One plausible explanation is that households incoastal provinces are more likely to deposit funds in other financial institutions or inother investment vehicles.5. Bank governance and loan performanceThe high and increasing percentage of non-performing loans is an essentialcomponent of the intermediation failure we have described in poorer areas. Theimportant question is: How much of this can be linked to problems in bank governance?In this final section of the paper, we examine the links between governance structures andnon-performing loans, and changes in governance structures themselves.From the branch bank surveys, we capture governance in a variety of waysincluding: loan authorization limits of branch managers; ex-ante bonus incentives ofbranch mangers measured as a percentage of the manager’s base wage; the percentage ofloans with collateral; government pressure to renegotiate overdue loans; quarterlyreporting requirements of branches; and the time required to resolve overdue loan issuesthrough the courts. A priori, we associate more powerful income incentives, the use ofcollateral, freedom from government intervention, stricter and more frequent reportingrequirements, and the ability to take legal action against defaulting borrowers as keyingredients of better governance structures. Managerial autonomy to make loans can alsobe a key component of a bank governance, especially when combined with incomeincentives that penalize bad managers for making bad loans and reward them for25

decisions that increase branch profitability. Unconstrained autonomy however can leadto moral hazard problems that might undermine banking profitability and efficiency. Weexpect better governance, in turn, to translate into a lower percentage of non-performingloans.Simple descriptive statistics on all of the governance-related variables aresummarized in Table 9, which breaks the data down by income quartiles of thetownships. We report both levels in 1997 and changes between 1994 and 1997. Themuch larger sample size for the former reflects the more complete information we havefor 1997. 14In some important respects, poor areas appear to be handicapped by poorerinstitutional practices. The size of loans that can be approved is much lower than in therichest areas, collateral requirements and reporting requirements are more lax, andgovernment influence on lending is higher. However, there are other aspects that favorpoor areas. Bonus incentives are stronger, perhaps reflecting budgetary shortages, andthe time it takes to complete lawsuits is shorter, perhaps because there are fewer casesbacklogging the system given lower levels of economic activity. In general, changes ingovernance variables is occurring more rapidly in richer areas compared to poorer areas,suggesting that reform is lagging in remote areas.Table 10 reports the results of OLS regressions analyzing the percentage of nonperformingloans in 1997 as a function of bank governance in 1997, and the change innon-performing loans between 1994 and 1997 as a function of governance structurespresent in 1994. We recognize the multiple endogeneity problems associated withrelating institutional change and economic performance, so treat the relationships more as14 Frequent managerial turnover in the bank branches handicapped our efforts to obtain information for1994.26

correlations rather than making causal inferences. Columns 5-8 of Table 10 report theresults linking non-performing loans in 1997 to bank governance. In each of theregressions, we also include township per capita income as an additional control variable.For comparison, column 5 reports the results of the bivariate regression of nonperformingloans on income. 15The regressions are nicely suggestive of the role ofgovernance. We find that the percentage of non-performing loans is negatively correlatedwith branch manager bonus incentives, the use of collateral, freedom from governmentpressure, and authorization limits (insignificantly), and positively correlated with thelength of time required to resolve cases legally. The percentage of non-performing loansis also significantly lower in higher income areas. This may be picking up overall higherloan quality, and possibly unobserved dimensions of local governance structures that arecorrelated with incomes.In column 7, we include as an additional explanatory variable the interactionbetween loan authorization limits and managerial bonus incentives. As in column 6, theindividual coefficients on these two variables remain negative, and now are significant at5 percent. The positive coefficient on the interaction term, however, suggests that acombination of high authorization limits and powerful incentives contributes to aweakening of the quality of the loan portfolio. One interpretation for this result is that thecombination leads to excessive risk-taking on the part of branch managers that ultimatelylead to an increase in the percentage of non-performing loans. It also suggests the needfor higher authorities to properly balance incentives and rights. In column 8 we include asan additional regressor information on quarterly reporting requirements of the bank15 The results are robust to including provincial fixed effects, controlling for branch type, and to excludingthe income variable.27

anches. Note that our sample size drops to 57 because of missing observations. Thesign on the variable is as predicted, but the coefficient is not measured with muchprecision and is statistically insignificant. The remaining coefficients are similar inmagnitude to those reported in columns 6 and column7.In columns 1-4, we analyze the effect of governance structures present in 1994 onchange in non-performing loans between 1994 and 1997. We also include townshipincome in 1994, and in column 4, add the level of non-performing loans in 1994. Limitedinformation on governance in 1994 restricts our attention to a much smaller set ofvariables than used above, and to a relatively small sample of RCC and ABC branches.The results are informative however. First, governance seems to matter: The increase innon-performing loans is smaller in townships in which government influence is less,there is more frequent reporting to higher level county branches, and managers have morediscretion in extending new loans. Second, the increase is less in townships with higherincomes. Since the change in non-performing loans is not correlated with the growth inour income measure (result not reported here) over the same period, one interpretation forthis result is that the income variable is picking up “unobserved” dimensions ofgovernance at the township level that are positively correlated with incomes. Third, wefind that the increase in non-performing loans over this three-year window is lower inbranches with a higher stock of non-performing loans of 1994. This may reflectadditional actions being taken by these branches to address the non-performing loanproblem. Note that the inclusion of the stock of non-performing loans in the regressiondoes not significantly affect the remaining coefficients.28

In Table 11 , we examine changes in governance levels between 1994 and1997. We keep the regressions simple and examine the effect of incomes in 1994, thelevel of governance in 1994, and the stock of non-performing loans in 1994 on thechanges in governance between 1994 and 1997. Three key questions motivate thisexercise: first, are the changes in governance greatest where they are most needed, i.e.branches with the highest stock of non-performing loans; second, are governancestructures converging across localities; and third, are poorer areas lagging behind in thepace of reform, all else held constant?In general, it appears that there is some convergence going on as suggested by thenegative coefficients on the governance variables in 1994. The overall trend appears to betoward better governance, with the greatest changes occurring in those areas with lowermeasures of governance in 1994. 16For three of the five governance variables (loanauthorization, incentives, and reporting requirements), it is also the case that theimprovement in governance is positively related (but not statistically significant) to thestock of non-performing loans in 1994. In other words, the changes are greatest wherethey are probably most needed. However, government pressure on branches appears tobe increasing in areas with a higher stock of non-performing loans. The likely reason forthis correlation is that it reflects government efforts to obtain relief for firms under theiradministration. Contrary to our expectation, we also find that the increase in the use ofcollateral appears to be lower in areas with a higher stock of non-performing loans.Finally, it also appears that with the exception of bonus incentives, the changes ingovernance are occurring most rapidly in higher income areas.16 This result should be interpreted carefully, however, because a regression of a change in a variable on itslagged value automatically imparts a negative bias. A number of the variables are also truncated.Unfortunately, there is not much we can do to correct these problems.29

What are then the implications of our analysis for poorer areas? Several thingsbode well. First, localities with higher stocks of non-performing loans appear to beslightly more aggressive in carrying out reform. Since it is poorer regions that have thehighest stock of non-performing loans, this suggests that governance reform in theseregions may be responsive to the pressures of non-performing loans. Second, theincrease in non-performing loans is smaller the higher the stock of bad loans in 1994.This may be picking up some of the benefits of the improvement in governmentstructures reported above. Yet possibly counteracting some of these trends is thetendency for changes in governance to be greatest in localities with the highest incomes.Despite better overall governance in these regions, the need for governance reformremains high.6. ConclusionsWe summarize some of the main findings from these simple empirical exercises:1. The ratio of loans to income, a measure of financial intermediation, was stable in richareas but fell in poor areas. The ratio of performing loans to income, a better measure ofeffective intermediation, fell in all areas, but much more rapidly in poor areas.2. Non-performing loans are significant in all areas, but greater in poor areas. They haveincreased over time in all regions, but faster in the richest and poorest areas. While theshare of household loans that are non-performing has remained high but relatively stable,the share of TVE loans that are non-performing has increased rapidly to exceed that ofhousehold loans.3. Loan-deposit ratios, measures of net fund outflow, are lower in rural financialinstitutions than in all banks and lower in poor areas than in rich areas. They havedeclined over time at a rate similar to that of all banks, but have declined more in richerareas.4. Income per capita grew faster in poor areas relative to rich areas, but deposits grewslower.30

5. Financial institutions in poor areas showed a greater willingness to lend to householdsand this relative trend strengthened over time.6. The non-performing loans problem is closely tied to institutional practices, or bankgovernance. While banks with greater repayment problems are reforming faster allthings equal, it is still the case the richer areas tend to have better governance and to bereforming faster.If we consider the motivating question for the paper and its title, the most accurateanswer is affirmative, but not for the anticipated reason, i.e., greater fund flows out ofpoor areas. While the increase in non-performing loans is undermining financialintermediation throughout rural China, the problem is worse in poor areas because ofslower deposit growth and more serious repayment problems. However, the way themotivating question is posed may be misleading because these changes are not so muchthe result of financial reform as the inability of reform to reduce non-performing loanseven as many regulations (especially interest controls and restrictions on interbanklending) continue to constrain the decisions of bank managers and link lending volumeclosely to local resources (deposits). If this, in fact, is what is happening, or rather nothappening, there is still ample scope for new changes and new distributionalconsequences as China continues to liberalize the financial sector in anticipation of WTOentry.To draw policy implications, it is important to understand the reasons for poorloan performance, especially in poor areas. If poor performance reflects the lack of goodprojects, the efficiency-equity tradeoff becomes quite stark, and policy-makers might bebest advised to improve the economic environment to make investments more desirable.However, if poor performance reflects institutional failings of financial institutions inremote areas, either because of stifling regulation or poor management, there may be31

greater scope for reforms that can advance both equity and efficiency goals. While manyidentification problems make our empirical results more suggestive than definitive, we dofind evidence that both the quality of projects and institutional practices matter. Some ofthe trends in the speed of institutional reforms are encouraging, others disquieting. Butthe evidence suggests there may be scope improved performance with possibly beneficialdistributional consequences from continued reforms.Experience in other countries suggests that the poor are capable of saving at highrates and that they are both credit-worthy and have demand for loans (Morduch, 1999).But savings responds most to convenience and security, while lending to the poor cannotbe profitable unless interest rates are allowed to be higher than in other areas (tocompensate for the higher administrative costs). The early experience of microfinanceinstitutions operating in China’s poor areas has been very positive, supporting the notionthat institutional reform and greater regulatory flexibility may encourage financialinstitutions to lend successfully in poor areas (Park and Ren, 2001).Thus, prognosis for the future depends very much upon reforms that have yet tooccur in China. If inter-bank markets are liberalized without interest rate liberalization,the effects could be quite negative for poor areas, but more open entry and pricingpolicies could be quite helpful. Underlying institutional development to better enforcecontracts, share credit histories, increase the effectiveness of collateral and guarantors assecurity for loans, increase the quality of managers and loan officers, and provideappropriate incentives will support more effective financial intermediation in both richareas and poor, but their importance to lending in poor areas may be even greater becauseof the greater institutional challenges to banking with the poor.32

ReferencesBesley, Timothy. 1994. “How do Market Failures Justify Interventions in Rural CreditMarkets?” The World Bank Research Observer 9(1): 27-47.Brandt, Loren, and Xiaodong Zhu. 2000. “Redistribution in a Decentralized Economy:Growth and Inflation in Refor China,” Journal of Political Economy 108(2): 422-439.Brandt, Loren, and Xiaodong Zhu. 2001. From Intermediation by Diversion toDisintermediation in China, University of Toronto, mimeo,China Financial Yearbook 1999 (Beijing: China Financial Press).China Statistical Yearbook 1998 (Beijing: China Statistics Press).China Statistical Yearbook 1999 (Beijing: China Statistics Press).China Statistical Yearbook 2000 (Beijing: China Statistics Press).Gertler, Mark, and Andrew Rose. 1996. “Finance, Public Policy, and Growth,” inGerard Capri, Jr., Izak Atiyas, and James Hanson, eds., Financial Reform: Theory andExperience (New York: St. Martin’s Press), pp. 115-128.King, Robert, and Levine, Ross. 1993. “Finance and Growth: Schumpeter Might BeRight.” Quarterly Journal of Economics 108, 3:713-37.Kraay, Aart. 2000. “Household Saving in China,” TheWorld Bank Economic Review14(3): 545-570.Lardy, Nicholas. 1998. China’s Unfinished Economic Revolution (Washington, D.C.:Brookings Institution Press).Levine, Ross. 1997. “Financial Development and Economic Growth: Views andAgenda.” Journal of Economic Literature 35, 2:688-726.Levine, Ross, and Zervos, Sara. 1998. “Stock Markets, Banks, and Economic Growth.”American Economic Review 88, 3:537-58.Lipton, Michael. 1984. “Urban Bias Revisited,” Journal of Development Studies 20(3):139-66.McKinnon, Ronald. 1994. “Gradual versus Rapid Liberalization in Socialist Economies:The Prble of Macroeconoic Control,” Proceedings of the World Bank Annual Conferenceon Development Economics 1993 (Washington, D.C.: The World Bank), pp.63-94.34

Naughton, Barry. 1995. Growing Out of the Plan: Chinese Economic Reform, 1978-92(Cambridge: Cambridge University Press).Nyberg, Al, and Scott Rozelle. 1999. Accelerating China’s Rural Transformation(Washington, D.C.: The World Bank).Park, Albert, Loren Brandt, and John Giles. 1997. Giving Credit Where Credit is Due:The Changing Role of Rural Financial Institutions in China, William Davidson InstituteWorking Paper No. 71.Park, Albert, and Minggao Shen. 2000. Joint Liability Lending and the Rise and Fall ofChina’s Township and Village Enterprises, mimeo.Park, Albert, and Changqing Ren. 2001. “Microfinance with Chinese Characteristics,”World Development 29(1): 39-62.Park, Albert, and Scott Rozelle. 1998. “Reforming State-Market Relations in RuralChina,” Economics of Transition 6(2): 461-480.Park, Albert, and Kaja Sehrt. Forthcoming. “Tests of Financial Intermediation andBanking Reform in China,” Journal of Comparative Economics.Park, Albert, and Sangui Wang. Will Credit Access Help the Poor?: Evidence fromChina, mimeo.Putterman, Louis. 1992. “Dualism and Reform in China,” Economic Development andCultural Change 40(3): 467-493.Rajan, Raghuram, and Zingales, Luigi. 1998. “Financial Dependence and Growth.”American Economic Review 88, 3:559-86.Sehrt, Kaja. 1999. Banks Versus Budgets: China’s Financial Reforms 1978-1996, Ph.D.dissertation, Department of Political Science, University of Michigan.Shen, Minggao, and Albert Park. 2001. Decentralization in Financial Institutions:Theory and Evidence from China, mimeo.Wong, Christine, Christopher Heady, and Wing Woo. 1995. Fiscal Management andEconomic Reform in the People’s Republic of China (Hong Kong: Oxford UniversityPress).35

Table 1Rural Financial Institution Balance Sheet Summary Statistics, 1994 and 19971994 1997N Assets Dep./ Loans/ Of Assets Dep./ Loans/ Of(millionyuan)Assets Assets which:firms(millionyuan)Assets Assets which:firmsFull sample 143 47 0.58 0.42 0.74 82 0.64 0.41 0.71ProvincesZhejiang 39 76 0.49 0.37 0.73 145 0.58 0.39 0.71Jiangsu 45 21 0.70 0.49 0.82 55 0.67 0.38 0.83Sichuan* 23 20 0.73 0.51 0.74 33 0.79 0.52 0.70Shanxi* 36 14 0.81 0.54 0.57 21 0.88 0.60 0.52Income quartilesI 31 95 0.53 0.38 0.81 172 0.58 0.37 0.75II 32 48 0.51 0.39 0.66 94 0.57 0.37 0.66III 34 31 0.71 0.51 0.83 44 0.83 0.58 0.78IV 33 12 0.83 0.46 0.37 18 0.95 0.53 0.35Bank typeABC 52 62 0.39 0.30 0.57 117 0.45 0.31 0.61RCC 91 38 0.76 0.53 0.83 62 0.84 0.52 0.77Note: All means are weighted by the amount of funds.*In 1998, the share of funds from deposits was 0.87 in Sichuan and 0.92 in Shanxi.*In 1998, collective firms, private firms, and others accounted for loan shares of 0.69, 0.02, and 0.28 in Sichuan and 0.47, 0.07, and0.46 in Shanxi.36

Table 2Financial Intermediation Indicators, 1994 and 19971994 1997 Annual %change1994-97N Loans/Inc.Perf./Inc.Loans/Inc.Perf./Inc.Loans/Inc.Perf./Inc.Full sample 127 0.38 0.31 0.38 0.28 0.01 -0.04ProvincesZhejiang 30 0.49 0.43 0.50 0.40 0.00 -0.03Jiangsu 45 0.27 0.21 0.26 0.19 0.01 -0.04Sichuan* 21 0.31 0.24 0.28 0.19 0.01 -0.02Shanxi* 31 0.60 0.43 0.56 0.35 0.00 -0.08IncomequartilesI 31 0.41 0.36 0.45 0.34 0.04 -0.01II 34 0.31 0.25 0.31 0.26 -0.01 -0.02III 31 0.40 0.31 0.35 0.24 -0.03 -0.09IV 31 0.43 0.30 0.40 0.23 -0.02 -0.13Bank typeABC 48 0.35 0.29 0.40 0.31 0.05 0.01RCC 79 0.40 0.33 0.37 0.27 -0.02 -0.07Note: All values are in 1994 yuan, deflated by using provincial consumer price indices.Means for per capita variables are weighted by township population. Means forloans/income and performing loans/income are weighted by total income (income percapita x population).*In 1998, loans p.c., performing loans p.c., and deposits p.c. were 467, 339, and 870 yuanfor Sichuan, and 731, 387, and 1153 yuan for Shanxi. The ratio of loans p.c. to incomep.c. and the ratio of performing loans p.c. to income p.c. was 0.27 and 0.19 for Sichuanand 0.50 and 0.27 for Shanxi.37

Table 3Share of Outstanding Loans that are Non-performing, 1994 and 19971994 1997 AnnualN OVER INACT DEAD Total OVER INACT DEAD Total %∆Full sample 145 0.09 0.07 0.01 0.17 0.13 0.08 0.02 0.24 0.15ProvincesZhejiang 39 0.08 0.03 0.01 0.12 0.13 0.06 0.01 0.20 0.18Jiangsu 47 0.07 0.12 0.03 0.21 0.09 0.10 0.05 0.25 0.12Sichuan* 23 0.16 0.08 0.00 0.25 0.19 0.10 0.03 0.32 0.07Shanxi* 36 0.16 0.12 0.03 0.30 0.21 0.15 0.04 0.40 0.14IncomequartilesI 31 0.06 0.05 0.01 0.13 0.15 0.08 0.02 0.25 0.32II 34 0.11 0.07 0.01 0.19 0.08 0.06 0.03 0.16 -0.07III 34 0.14 0.09 0.02 0.25 0.16 0.12 0.05 0.32 0.08IV 33 0.15 0.11 0.05 0.32 0.23 0.18 0.03 0.44 0.14Bank typeABC 53 0.08 0.09 0.01 0.17 0.08 0.09 0.03 0.20 0.00RCC 92 0.10 0.06 0.02 0.17 0.17 0.08 0.02 0.26 0.23Borrowertype**:TVEs 36 0.15 0.06 0.02 0.23 0.15 0.21 0.03 0.38Households 51 0.18 0.13 0.01 0.32 0.15 0.15 0.03 0.33Note: All means are weighted by the amount of outstanding loans.Definitions: OVER=overdue loans, INACT=inactive loans (overdue more than 2 years),DEAD=loans with no expectation of repayment because of the following reasons: a) theborrower has died or cannot be located; b) the borrower has gone bankrupt; or c) the loanis more than three years overdue.*In 1998, Sichuan and Shanxi overdue loans were 0.28 and 0.48 of outstanding loans.**For Sichuan and Shanxi only.38

Table 4Loan-Deposit Ratios, 1994 and 19971994 1997 Annual %∆N AllLoansPerf.LoansAllLoansPerf.LoansAllLoansPerf.LoansFull sample 142 0.71 0.59 0.64 0.48 -0.07 -0.03ProvincesZhejiang 38 0.73 0.64 0.67 0.54 -0.08 -0.04Jiangsu 45 0.70 0.55 0.56 0.42 -0.09 -0.06Sichuan* 23 0.70 0.53 0.66 0.45 -0.04 -0.05Shanxi* 36 0.66 0.46 0.69 0.41 -0.02 0.00Income quartilesI 31 0.71 0.61 0.63 0.47 -0.11 -0.05II 32 0.77 0.63 0.66 0.55 -0.05 -0.04III 34 0.71 0.53 0.70 0.48 -0.02 0.01IV 33 0.55 0.38 0.57 0.32 -0.05 0.05Bank typeABC 51 0.75 0.61 0.66 0.53 -0.06 -0.03RCC 91 0.70 0.57 0.62 0.46 -0.08 -0.03Note: All loan values are divided by total deposits. Mean loan-deposit ratios areweighted by the amount of deposits.*In 1998, Sichuan and Shanxi, loan-deposit ratios were 0.57 and 0.63, of whichperforming loans were 0.41 and 0.33.Table 5Loan-Deposit Ratios from Other Surveys of RFIs, 1988, 1994-1996N 1988 1994 1995 19961996 National Village Survey 80 0.94 0.68 0.731997 China Rural Poverty Survey 19 0.73 0.61 0.55 0.54Note: All means are weighted by the amount of deposits.39

Table 6Deposits and Income Per Capita, 1994 to 1997NDep.p.c1994 1997 Annual %∆Income Dep. Income Dep.p.c. p.c p.c. p.cIncomep.c.Full sample 125 1002 1860 1487 2515 0.12 0.11ProvincesZhejiang 30 1572 2357 2377 3301 0.12 0.11Jiangsu 43 816 2160 1309 2844 0.14 0.11Sichuan* 21 593 1242 741 1612 0.10 0.09Shanxi* 31 785 871 998 1202 0.08 0.11Income quartilesI 31 1853 3169 2801 3910 0.13 0.07II 32 772 1876 1304 2709 0.15 0.12III 31 718 1308 901 1844 0.09 0.11IV 31 590 748 771 1111 0.08 0.13Bank typeABC 47 889 1882 1488 2526 0.15 0.11RCC 78 1076 1846 1485 2508 0.10 0.10Note: All values are converted to 1994 yuan using provincial CPIs. Means for depositand loan growth rates are weighted by base year starting values. Mean income p.c.growth is weighted by 1994 township population.40

Table 7Decomposition of Changes in Financial Intermediation(annual growth rates, 1994 to 1997)Perf.loans/incomeAllloans/incomePerf.loanshareLoans/depN PL/Y L/Y PL/L L/D D/POP Y/POPFull sample 125 -0.04 0.01 -0.04 -0.03 0.12 0.11ProvincesZhejiang 30 -0.03 0.00 -0.05 -0.03 0.12 0.11Jiangsu 43 -0.04 0.01 -0.03 -0.06 0.14 0.11Sichuan* 21 -0.02 0.01 -0.04 0.00 0.10 0.09Shanxi* 31 -0.08 0.00 -0.07 0.05 0.08 0.11Income quartilesI 31 -0.01 0.04 -0.06 -0.05 0.13 0.07II 32 -0.01 -0.01 -0.01 -0.04 0.15 0.12III 31 -0.09 -0.03 -0.04 0.01 0.09 0.11IV 31 -0.13 -0.02 -0.10 0.05 0.08 0.13Bank typeABC 47 0.01 0.05 -0.03 -0.02 0.15 0.11RCC 78 -0.07 -0.02 -0.05 -0.03 0.10 0.10Dep.p.c.Inc.p.c.Table 8Financial Performance and Level of Economic DevelopmentEstimates from OLS Regressions of Performance Indicators on Income Per CapitaDependent variableWith controls forprovincialdifferences(1)No controls forprovincial differences(2)N 1994 1997 1994 1997126 **0.15 ***0.21 0.01 0.03Performing loans per capita/income per capitaLoans per capita/income per 127 *0.14 ***0.24 -0.05 -0.03capitaNonperforming loans/total loans 131 **-0.08 ***-0.21 ***-0.14 ***-0.20Loans/deposits 129 -0.32 -0.52 *-0.41 **-0.42Deposits per capita (1994 yuan) 125 ***1.35 ***1.79 ***0.91 ***1.03Loans to firms/total loans 118 ***0.28 ***0.41 ***0.33 ***0.38Loans to households/total loans 156 ***-0.16 ***-0.28 ***-0.30 ***-.40Note: Independent variable is log of income per capita (in 1994 yuan).*10 percent significance level, **5 percent, ***1 percentSpecification (1) includes province-bank type interaction dummy variables.41

Specification (2) includes dummy variable for bank type only.42

Table 9Governance by Income QuartilesGovernance in 1997 Change in Governance, 1994-1997Income QuartilesIncome QuartilesVariable I II III IV I II III IVLoan9.64 0.83 1.82 0.93 5.00 0.31 -0.83 -1.21AuthorizationBonus Incentive 0.56 0.48 0.47 0.63 * * -.09 0.34Collateral 49.6 53.1 62.3 38.7 33.2 31.1 16.1 7.2Gov’t Pressure 3.71 3.58 3.88 3.45 1.2 1.1 1.5 0.28Rep. Require. 0 -.6 -.1 0Time to End 4.71 4.35 5.49 1.98 * * * *Lawsuit (months)Number of Obs 14 24 26 20Notes: Loan authorization limits are reported in 10,000 RMB. Bonus incentives measure the exantebonus in event of target fulfillment as a percentage of the base wage. Collateral is thepercentage of loans to TVEs secured with collateral. Government pressure is measured on adecreasing scale from 1-6.Table 10Governance and Non-Performing Loans in Township BranchesChange in Non-Performing LoansNon-PerformingLoans in 1997Income, 1994 -0.102(-2.70)-0.128(-3.28)-0.127(-3.76)-0.120(-4.92)Income 1997 -0.102(-5.05-0.111(-5.82)-0.110(-5.66)-0.105(-4.55)Loan Authorization,1994-0.001(-0.21)-0.008(-1.49)-0.014(-2.39)Loan Authorization,1997-0.001(-0.41)-0.016(-1.78)-0.012(-2.24)Bonus Incentive,1997-0.080(-1.46)-0.112(-2.02)-0.063(-0.98)Authorization*Bonus 0.041(1.85)Loans w/ Collateral,1997-0.002(-2.70)-0.002(-2.57)-0.002(-2.16)Gov’t Intervention,1994-0.024(-1.54)-0.044(-2.96)-0.040(-2.85)Gov’t Intervention,1997-0.029(-2.01)-0.032(-2.16)-0.022(-1.37)Rep. Requirements,1994-0.077(-2.77)-0.090(-3.27)Rep. Requirements,1997-0.051(-0.89)Time to End Lawsuit 0.010(2.69)0.010(2.97)0.010(3.20)Non-PerformingLoans, 1994-0.485(-3.06)Number of Obs. 59 59 46 46 84 84 5743

R 2 0.10 0.14 0.31 0.49 0.39 0.41 0.41 0.4044

Table 11Changes in Governance between 1994 and 1997Loan Authorization Bonus Collateral Gov’t Pressure Rep. RequirementsIncome, 1994 1.633(1.62)2.300(1.81-1.014(-2.57)-1.062(-3.0210.79(2.58)5.12(1.130.389(2.45)0.106(0.840.028(1.54)0.074(1.7)Loan Authorization, 1994 -0.714(-7.06Bonus, 1994 -0.488(-1.82)Loans w/ Collateral, 1994 -0.377(-3.71Gov’t Pressure, 1994 -0.349(-4.49)Rep. Requirements, 1994 -0.363(-2.80)Non-perf. Loans in 1994 2.33(1.47)0.844(1.65)-25.94(-1.01)-1.05(-1.47)0.276(1.36)Number of observationsR 2 73 69 25 25 47 45 92 84 94 800.11 0.52 0.26 0.48 0.10 0.30 0.06 0.27 0.01 0.3145

Appendix Table 1Loan Composition, Sichuan and Shaanxi, 1994 and 19981994 1998N Firms HH Ag Other Firms HH Ag OtherFull sample 59 0.32 0.15 0.22 0.31 0.38 0.20 0.16 0.26ProvincesSichuan 23 0.48 0.09 0.19 0.24 0.50 0.18 0.16 0.17Shanxi 36 0.19 0.20 0.25 0.36 0.24 0.27 0.11 0.38Income quartilesI 13 0.42 0.08 0.22 0.28 0.52 0.09 0.07 0.32II 14 0.33 0.04 0.36 0.27 0.29 0.16 0.29 0.27III 15 0.19 0.29 0.12 0.40 0.22 0.38 0.07 0.32IV 14 0.05 0.51 0.09 0.35 0.11 0.68 0.02 0.19Bank typeABC 16 0.13 0.15 0.22 0.51 0.11 0.20 0.12 0.57RCC 43 0.45 0.15 0.22 0.18 0.48 0.24 0.14 0.15Note: All means are weighted by the amount of outstanding loans.46