Impact analysis of one donor supported project is still preliminary but appears positive, and suggests that wider adoption of the model is indeed warranted. Some funds from existing programs could profitably be used to this end.

In particular, it is recommended that a large share of the subsidized poverty loan program be utilized to support multisectoral poverty reduction projects in China's poor townships. However, effective implementation of multisectoral projects will require substantial changes in current institutional arrangements for implementing the poverty program. In particular, it will require creation of project management offices with full involvement of the PADOs in the design, supervision and monitoring of projects. Improving the effectiveness of the existing programs described in Section D above will also require institutional changes. For these programs, it is recommended that the PADOs take on greater involvement in the planning, supervision and monitoring of programs, but not necessarily in their actual implementation.

Multisectoral Rural Development Projects Experience with a number of large and small-scale projects in China during the 1990s suggests that one of the most effective means of assisting the absolute poor is through an integrated set of interventions in the form of a multiyear project. The initial experience in China with such projects was through donor supported multisectoral projects, designed and implemented in a participatory manner.

The World Bank supported Southwest Poverty Reduction Project (SWPRP) and the Qinba Mountains Poverty Reduction Project (QBPRP), which are summarized in Box 3.1, are the largest scale examples of this new approach. In most cases, the multisectoral rural development project model includes an integrated program of　investments in (a) upland agricultural development, using menus of field and tree crop and　livestock activities to increase upland agricultural productivity, (b) labor-intensive construction of rural roads, drinking water systems, small scale irrigation, agricultural drainage works, and other　rural infrastructure, (c) provision of off-farm employment opportunities through a voluntary system of enhanced rural labor mobility for the upland poor, (d) institution building and poverty monitoring, and (e) rural enterprise development. Improved access to basic education and health, and separate microcredit components, are also included in some donor supported projects.

Implementing Programs to Meet the Needs of the Poor (II)

The World Bank-supported SWPRP and QBPRP are grassroots poverty reduction　projects assisting 61 nationally-designated poor counties (35 counties in Guangxi, Guizhou and Yumian　under SWPRP and 26 counties in Sichuan, Shaanxi and Ningxia respectively). The projects' main　objectives are to introduce an integrated multisectoral poverty reduction project approach, and to assist some 5.2 million beneficiaries in these poor counties overcome poverty. Greater community participation　has been encouraged by enabling project households and villages to make key decisions during project design and implementation.

The multisectoral project approach comprises assistance for (a) increased upland agricultural productivity by providing project households with a menu of field and tree crop and livestock activities and through the support of provincial and regional applied agricultural research, (b) TVE development through support for labor-intensive enterprises with strong backward linkages to poor households and which meet environmental safeguards, (c) small scale rural infrastructure including roads, drinking water supply systems, irrigation and drainage works, and electrification, (d) greater access to off-own-farm employment opportunities through a voluntary system of labor mobility including a job placement system and　monitoring of worker safety and living conditions, and (e) institution building and poverty monitoring through (i) improving the project management offices' (PMO) capabilities in project design, implementation and supervision, and (ii) strengthening the poverty monitoring system for the project areas by using improved SSB survey instruments. In addition, SWPRP includes a component to improve access to basic education and health services, and QBPRP includes a microfinance component to provide credit services to poor households for production activities.

The projects' two principal risks are their above-average complexity and very heavy demands on institutional capacity for project design and implementation. These risks have been successfully overcome by keeping each of the large number of activities as simple as possible, establishing　a central PMO under LGPR with final responsibility for the projects, and a heavy emphasis on institution building. An independent Quality Assurance Panel assessment concluded that SWPRP implementation has been fully successful with several best practice features. The late-1997 LGPR sponsored Second Working Conference on Poverty Reduction Projects also reconfirmed the overall success of SWPRP and QBPRP, and announced China's intention to extend the integrated multisectoral project　approach to all of China's poor counties. Both projects are enabling significant improvements to the income levels and well-being of the intended beneficiaries, and the detailed poverty monitoring system data are allowing a rigorous quantitative assessment of the projects' impact. Other lessons learned are (a) the importance and effectiveness of the projects' institution building and supervision procedures (which is called the "yanshou" system), and (b) that the direct inclusion of the basic education and health services component had the favorable, but somewhat unanticipated, effect of greatly augmenting community participation. TVE development is the only project component which is performing poorly.

Initial experience with this new approach has been favorable. Ravallion and Chen (forthcoming) have evaluated the impact of SWPRP. The analysis uses SSB data from 2000　project and non-project households to calculate returns to the household from project investment. The data covers three years of project implementation, but because of comparability problems　between data in the first and subsequent years, only data from 1996 and 1997 was useable. The shortness of this period means that only preliminary conclusions are possible at present, but it will be possible to firm up the analysis as data for subsequent years of project implementation are gathered. The control group used in the analysis was a subset of the original SSB control group, and some control villages monitored by SSB were rejected as being too dissimilar to the project villages. This makes possible to use the "double difference" method, which assumes that the　control villages are representative of what would have happened in the project villages without the project, and thereby completely controls for initial conditions. The study found that the average income gain in 1997 represented 12 percent of the mean income in 1996 in the sampled project villages. While more definitive results will need to wait until data is available for the entire project period, it does appear that returns from the project are higher than from government poverty spending alone.

(III)

The Chinese authorities have expressed their intention to roll out the multisectoral　approach over a larger area, and some provinces have begun experimenting with this approach. Plans for domestic multisectoral projects appear to include fewer sectors than the donor funded projects, and would not include social services and labor mobility. Evaluation of the donor-supported projects which include basic education and health care indicates that including these components increases villager's enthusiasm for the entire project. However, supervision of these components taxes institutional arrangements, and it may be more practical to continue to implement these components separately for domestically funded projects. Organized labor mobility has been included in donor funded projects with good result, and domestic programs may wish to consider support for this type of component.

Given the disappointing performance of the subsidized loan program, it is recommended　that China consider using a portion of the subsidized loan funds blended together with grant funds to fund integrated multisectora! projects in the poorest areas. The multisectoral rural development projects are very demanding of design and implementation capacity and, at present, funding and institutional arrangements may not be adequate to enable an effective rollout of this approach.

Better coordination between funding channels is necessary to enable the poverty program to direct sectoral components in an integrated manner to the poorest of the poor. Most critical to the successful implementation of this approach are strong institutional arrangements. Indeed, institution building and strong project management systems have been instrumental to the success of SWPRP and QBPRP. This has included establishment of effective project management offices(PMOs) and work stations (WS) at the provincial, county, township and village levels, and implementation of a rigorous works supervision and acceptance process referred to as the "yanshou" system. At present, the number of staff involved in specific poverty work at the　provincial and county levels are probably sufficient to staff functional PMOs. At the township level, however, staffing is inadequate. Based on the experience of SWPRP and QBPRP and assuming this model is applied to all of China's poor townships, the estimated annual cost of　strengthening township WSs to reasonable levels would be less than Y 100 million, or about 0.4 percent of total annual poverty reduction funding Reducing direct poverty funding by this amount would not significantly reduce the scale of the program, and would likely have a significantly positive impact on the effectiveness of the program.

The effectiveness of the entire poverty reduction program could be greatly improved through much stronger institutional arrangements. Despite its mandate to coordinate the nation's poverty reduction program, the LGPR system does not have (a) sufficient advance access to the detailed program information necessary to insure that all poverty reduction funding benefits the poor, or (b) the staffing to properly oversee the quality of poverty reduction project works and activities. This limited control and inadequate staffing directly contributes to the (a) significant leakage of available funding to works and other uses which have little or no benefit to the poor, and (b) unacceptably large share of the works and activities which do benefit the poor but which are of inferior or substandard quality. The following are suggestions for modifications in institutional arrangements which may result in a more effective poverty program.

(IV)

At the provincial and lower levels, officials often complain that the LGPR system has been given responsibility for ensuring that poverty reduction goals are achieved, but not sufficient control over the funding or the necessary staffing to fully complete this important mission ("Fupinban you zeren, mei you quanii"). The LGPR system has only limited access to data on the projects and activities undertaken through the FFW and MOF grant programs, and at best has only limited influence during the planning stage over the works and activities undertaken with subsidized loans provided through ABC. The current arrangement of rely ing on local bureaus and agencies to implement project works and activities must be maintained, but the LGPR system should play a greater role in the planning of how funds are to be used, and in the supervision and monitoring of their usage. It is therefore strongly recommended that, in order to raise the impact of the funds on the poor and reduce the leakage of funding to alternative uses, the LGPR system should take on a greater planning and supervisory role for the use of the funds.

Forging stronger links with government line bureaus, academic and civic organizations involved in poverty work would increase LGPR's effectiveness in setting policy and implementing programs. For example, LPGR's policy making function could be strengthened by contracting research work to organizations with specialized knowledge, and incorporating research findings into their policies and strategies. It seems particularly　unfortunate that the poverty reduction efforts of several government ministries is focused on their "adopted" poor counties instead of on contributing to key research work and assisting with policy making in their areas of expertise. MOA, for instance, could make a far greater contribution to China's poverty reduction program by working with the LGPR system to create a development strategy for the karst region and to initiate a program of applied agricultural research for mountain areas.

The next generation of poverty work could include contracting the implementation of some small projects to grass roots and civic organizations. For example, Chinese "GONGOs" (that is, government organized NGOs) have proven successful at implementing poverty projects in a number of areas. Yet most of their funding is self-raised, and official poverty alleviation funds are not channeled through them. Experimenting with channeling a portion of the poverty funds through GONGOs and other grass roots organizations could enable the poverty program to try new and innovative approaches, and improve its outreach. Such an approach has proven highly successful in other developing countries, and could be particularly valuable to work in China's minority areas.

Strengthening LGPR's oversight and control over poverty reduction should be complemented by reciprocal measures to increase LGPR's accountability by improving monitoring of the impact of the poverty program and the use of poverty funds. "The monitoring function should be contracted to an independent outside organization. Financial monitoring should trace the flow of funds, and determine whether the use of funds meets LGPR's guidelines.

Impact monitoring could build on SSB's rural household survey, and LGPR should rely more heavily on those survey data in targeting its programs and evaluating their effectiveness.

(V)

Development work around the world has found that allowing stakeholders a voice in project design, management, and evaluation improves results. Such approaches were not tried much in China before the 1990s, and still tend to occur primarily in programs funded by　international organizations. Chinese and international researchers have noted the positive impact on projects when participation is encouraged. Most of China's experiments with participation have been in project identification and preparation. Some projects are systematically developing strategies by creating a dialogue between experts and local beneficiaries. In many　cases these result in a menu of options, which allows projects to adapt flexibly to local needs.　The results of recent Participatory Poverty Assessments (PPA) in several villages in Guangxi, Yunnan and Ningxia are summarized in Annex 4, These village PPAs powerfully document the poor's own understanding of their experience with poverty, and highlight a number of significant weaknesses of poverty reduction activities and projects in their villages.

Less work has been done on building participatory institutions for project management, and the few experiments which have been undertaken have had mixed results. Local officials and experts are sometimes unhappy with what they perceive as a diminution of their authority, and China has very little precedent on which to model such independent institutions. On the other hand, there has been some success with institutions that combine local government and existing social networks, ranging from water management associations to small loan societies. Finally, monitoring and evaluation systems have very rarely tried participatory methods in China, and　tend in general to lack qualitative input. The following observations and recommendations, if applied to future poverty reduction programs in China, might further increase benefits to the rural poor:

Projects that bring management organization down to the village level typically respond more effectively to local needs.

The most successful community participation builds on existing social networks. Small loan　societies, for example, work better in groups that already have close social ties than in ad hoc　combinations with no other interests in common. Participatory　institution building will also work best when it builds on natural community leadership.

Local project officials need training programs both to learn participatory methods and to　understand their goals. These methods are a significant departure from the usual cadre work　style in China, and training programs need to show the benefits to local leaders.

Attempts to build new participatory administrative units can be improved by building the　interests of local government stakeholders into the project or by building more community　input into existing institutions.

Minority community participation is especially important to help projects adjust to local　cultural differences. Representatives from the local Ethnic Affairs offices should also be built　into project management teams in minority areas.