VULNERABILITY ANALYSIS OF CHINESE COUNTIES

Similar documents
Regional Inequality and CO 2 Emissions in China: a consumption-based MRIO approach

Analysis of air quality trends in 2017

Regional disparity and Mitigation cost for carbon policy in China Assessment based on multi-regional CGE model

Changes in Area and Quality of Cultivated Land in China

The Development of Smallholder Inclusive Business Models in China. Prepared by Gu Rui AII/CAAS December,2015. Hanoi

Energy and Pollution Efficiencies of Regions in China

Research on the Ability of Regional Industrial Sustainable Development

Understanding CCS in China s Mitigation Strategy using GCAM-China

Method for Calculating CO 2 Emissions from the Power Sector at the Provincial Level in China

Dynamic Coupling Development of Regional Socio-economy-Energy-Environment in China

IMPACTS OF CLIMATE CHANGE AND GRAIN SECURITY IN CHINA

Chapter 4. China. 4.1 Overview

China Curtain Wall Cladding Materials Review

Regional Differences and Dynamic Changes in Rural China: the Study of 1996 and 2006 National Agricultural Census

Workshop Management Office: Fairlink Exhibition Services Ltd.

Measurements for Forest Ecological Benefit in China

CHINESE AQUACULTURE: A COMPARATIVE ANALYSIS OF THE COMPETITIVENESS OF REGIONAL AQUACULTURE INDUSTRIES

The Accounting Methods of the Local Government Department Output by Factor Analysis

Abuilding and Planning Tunnels in China. China and Latin America

Impacts of emission reduction target and external costs on provincial natural gas distribution in China

Study on Application of Factor Analysis in Regional Environmental Assessment

Improving Energy Productivity: The Policy Practice of China. Dongmei Chen China Energy Conservation Association Nov.17, 2015

Open Access Empirical Study on Ecological Niche Evaluation on Regional Construction Industry in China

Cropland Mapping and area estimation method in CropWatch. Nana Yan, Miao Zhang, Bingfang Wu, Bo Chen RADI, CAS June, 2015

Climate Change Policy Target Setting and Implementation Process in Japan and China

Impacts of China s GPA Accession

TO U R. by CCIFC & MOFCOM-CICPMC. China Cities of the Future

Promoting Energy Efficiency in China: The Status Quo and Way Forward

CEEP-BIT WORKING PAPER SERIES. China's regional carbon emissions change over

The Basic Situation of Regional Grain Trade in China

China Emission Trading Scheme : Policies and Challenges

Study on the Minimum Wages Increases in Beijing

Analysis on Comparative Advantage in the Production of. Major Grain Varieties in Different Areas of China

An Empirical Research on Industrial Structure Optimization of Provincial Area Based on Two-oriented Society

Current Status of Chinese Alumina Industry and SAMI s Technical Solution

Land Use Changes and Economic Growth in China

Supplement of Inventory of anthropogenic methane emissions in mainland China from 1980 to 2010

Human capital and energy in economic growth Evidence from Chinese provincial data

Strategy for Archival Management in the Digital Age

China Market Report 2018

APERC Report: Understanding Energy in China Geographies of Energy Efficiency

麗豐股份有限公司 (4137.TT) Chlitina Holding Limited Chlitina Holding Limited. Group Introduction

Agricultural Science and Technology Innovation Efficiency based on DEA Model: Empirical Analysis of Efficiencies of Regions, Provinces and Anhui

China s Electric Power Industry and Its Trends

Mid-term Business Plan 2019

Long-Term Energy Demand and Supply Outlook for the 31 Provinces in China through 2030

China Corn & Corn Seed Industry Report,

Table 4.1. CropWatch agroclimatic and agronomic indicators for China, October 2016-January 2017, departure from 5YA and 15YA

Analysis of Carbon Emission Efficiency for the Provinces of China YU Dun-yong 1 ZHANG Xue-hua1,*2

Total-Factor Energy Efficiency in China s Agricultural Sector: Trends, Disparities and Potentials

Energy-saving Potential Study on Telecommunication Base Station Free Cooling With a Thermosyphon Heat Exchanger in China

Demonstration Zones of Agricultural Modernization by Mr. Qian (chief agroeconomist

Department of Applied Economics and Management Cornell University, Ithaca, New York USA

Patterns of Domestic Grain Flows and Regional Comparative Advantage in Grain Production in China

4. Internal Convergence and China s Growth Potential

An Alternative Approach to Measure Regional Comparative Advantage in China s Grain Sector

ENVIROMENT PROTECTION AND NATURAL GAS DEVELOPMENT IN CHINA

Estimating Chinese Unified Carbon Market Size

Assessing blue and green water utilisation in wheat production of China from the perspectives of water footprint and total water use

Polyamide & Intermediates

The Empirical Research on Independent Innovation Competencies of Enterprise R&D Departments

CHAPTER 4 EMPIRICAL RESULTS

Total Energy Consumption Control based on Environmental ZSG- DEA

Session 3: Enhancing gas supply and diversification New sources & markets

IMPLEMENTATION OF THE COAL CAP PLAN: LONG TERM IMPACTS, URGENCY AND EFFECTIVENESS

11.481J / 1.284J / ESD.192J Analyzing and Accounting for Regional Economic Growth Spring 2009

The Scenario Analysis of Shale Gas Development on the sufficiency of pipeline network in China by applying natural gas pipeline optimization model

Strategies and Actions of Biodiversity Conservation in China

VOSTEEN Consulting GmbH Thermal Engineering and Environmental Protection

Operation Situation by the Number of Mines (MT=million tonnes)

THE EFFECTS ON THE AISAN AND WORLD GAS MARKETS OF THE CHINESE GAS MARKET EXPANSION

China s Potential of Grain Production Due to Changes in Agricultural Land Utilization in Recent Years

China s Energy Management System Program for Industry

Towngas China (1083.hk) 2016 Final Results Presentation

To: Business Editor April (For immediate release)

Analysis of Total Factor Efficiency of Water Resource and Energy in China: A Study Based on DEA-SBM Model

Regional efforts to mitigate climate change in China: A multi-criteria assessment approach

SUSTAINABLE WATER RESOURCE USE IN ASIA CLOSING WORKSHOP. China 2012,Bangkok

The Efficiency Improvement in Low-carbon Technology Innovation of Chinese Enterprises under CDM: An Empirical Study based on DEA Assessments

Research on China s Regional Cultural Industries Efficiency Based on Factor Analysis and BCC & Super Efficiency Model

ADB Economics Working Paper Series. Impacts of Climate Change on the People s Republic of China s Grain Output Regional and Crop Perspective

advantage in Chinese grain production

China s operating steel capacity increased in 2016, despite efforts on overcapacity

Efficiency Analysis of Chinese coal-fired power plants: Incorporating both the undesirable and uncontrollable variables in DEA

The methods and effecting of Bounty Plan. ----case study on the labor migration in the poverty areas in China

Journal of Informetrics

Research on Applications of Data Science in Macroeconomics

China s Accession to WTO

CDM Country Fact Sheet : China

OVERCAPACITY, OVER-WITHDRAWAL: HOW TACKLING COAL POWER OVERCAPACITY CAN EASE WATER STRESS

The spatial exposure of China s infrastructure system to flooding risks in the context of climate change

China Methanol Industry: Story of Coal and MTO

Coal in China. Carlos Fernández Álvarez Senior Coal Analyst. Paris, 14 January 2013

THE PERSPECTIVE OF WATER SUPPLY AND DEMAND FOR SUSTAINABLE DEVELOPMENT IN CHINA

Factor Price Equalization and Economic Integration in China

Performance Measurement of Healthcare Service and Association Discussion between Quality and Efficiency: Evidence from 31 Provinces of Mainland China

China s Ecological compensation policy

The research of low-carbon industrial cluster in China based on location quotient method

China s Renewables Curtailment and Coal Assets Risk Map

The First Year of China s Twelfth Five Year Plan: Success or Failure for Climate Change Efforts?

Transcription:

VULNERABILITY ANALYSIS OF CHINESE COUNTIES WFP/IFAD China VAM Unit June 2003 1

Background In 1997, when vulnerability analysis and mapping was first introduced to China, an analysis of all the Chinese counties was conducted. Based on the 1990-1995 data collected from the National Statistical Bureau (NSB), the analysis clustered the most vulnerable counties, and identified three eligible areas for WFP/IFAD assistance in the near future with the agreement of the Ministry of Agriculture (MOA) of the Chinese government (see annex 1). WFP/IFAD projects in Shaanxi, Hubei and Guangxi were formulated afterwards upon the agreement. Three years later, when the Country Strategy Outline and the Country Programme of 2001-2005 was being prepared, the WFP/IFAD China VAM Unit updated the county vulnerability analysis of 1997 in May of 2000. Based on the updated analysis, 62 counties in Shanxi, Ningxia, Gansu and Xinjiang province were identified as the potential project counties after discussion with MOA (see annex 2), and resource allocation was preliminarily planned for each province. In the above two round of analysis, the indicators used were mainly refer to the risks of food security, such as population growth, food availability and land availability. Information on the coping ability of a county was not complete, even the income information was not available for all the counties in the analysis. In order to monitor the changes of food security and vulnerability of Chinese counties and refine the previous analysis, in 2003 1, the VAM Unit conducted another round of county analysis for all the counties in China. The preparation of this analysis started in late 2002. Latest information was purchased from NSB, and a database was set up in MS ACCESS for data storage and indicator calculation. 1 The analysis was done by Ms. Han Zheng, national VAM officer of WFP China, with assistance of Ms. Yu Jing, VAM Assistant. 2

Socio-economic Data Data used in the analysis Because NSB s statistical focus of counties was on food supply in the 1990s, the data that purchased from NSB for the first two analysis weighed heavily in favor of the counties performing poorly in agriculture, which means the counties having food supply problems would come out of the analysis as more vulnerable with no exclusion of those that were quite developed in industry or business. Every so often, these counties had a poor agricultural performance. There was nor complete income information to counteract this influence of the NSB data composition. Therefore, kinds of manual exclusion had to be done to take out counties famous for industry or business but with a weak agricultural sector from the list of the most vulnerable counties, for they were able to have enough food through purchasing. In 1999, NSB modified the previous data composition of the published county information. Several agricultural indictors were excluded, such as crop-sown area, the sown area and cultivated land. In the meantime, some social-economic indicators were added, they were: GDP value, students in school, number of hospital beds, number of households with telephones, etc. Although the information about agricultural production is no longer so complete as they were in the past, the available data are enough to assess food security level. It s worth mentioning that the newly added data about social sector are very useful and can prevent the analysis from being agricultural biased for they can help us to get a better view of a county s ability to cope with food insecure risks. In the 2003 analysis, most of the indicators are calculated with data from 1999 to 2001 except for the food production that is an average of 1997 to 2001. The reason for this arrangement is that a five-year average of food availability can give a more realistic picture eliminating dramatic vibrations caused by natural disasters in one of the years. Details can be found in Annex 3. GIS Data One thing should be stressed here is that unlike the other two rounds of analysis, GIS data is introduced to supplement the socio-economic data in 2003 analysis. As the GIS data is collected through satellites or digitized according to existing maps, all the information are immune from artificial manipulation, which could happen to the socio-economic data. Moreover, as GIS information covers mainly the physical conditions and the availability/access to infrastructure, the combination of the independent GIS data to the analysis could broaden the understanding of the root causes of vulnerability as well as help to refine the results. Details are in Annex 4. Counties included in the analysis The NSB data of 2000 and 2001 provide information of 2076 and 2073 counties. However, the NSB data of 1999 does not have information for the 72 counties in Tibet. In an attempt to include all the counties in the analysis, data of the year1999 is considered missing when computing average through years for Tibet. 3

A small number of counties listed in the NSB data set do not have enough information. For example, some cities in the Northeast of China are included in the data set without enough information about rural areas. After cleaning the data set, 2052 counties are included in the analysis. 4

Indicators The indicator selection is based on the concept of vulnerability, which is a combination of food security risks and the inability to cope. Thanks to the enlarged database from NSB and the introduction of GIS information, it is possible this time to measure the overall vulnerability level of a county. The indicators of Food Security Risk are: Annual growth rate of rural population Population density Per capita area with the most adverse natural conditions Coefficient of variation (C.V.) of per capita grain production Percentage of rural population The above five indicators collaboratively depict the pressures of food availability and the current level of risk that the county is facing. The annual growth rate of rural population is an indicator describing the trend of food demand pressed on the outputs of the land. Population density is a static reflection of the food demand. Due to the fact that no information about disaster damages is available, the C.V. of per capita grain production can be considered as a proxy, which is a reflection of the vibration of food supply. Percentage of rural population links directly with food security risk because the rural population has to maintain their food availability on the land output. The higher the data of these indicators are, the bigger the risk of food insecurity is. Per capita area with the most adverse natural conditions is a combination of five GIS images with a resolution of 1km x 1km: elevation, slope degree, land quality, land coverage and rainfall. By using geographical data analysis software, all the information contained in the images is categorized good, ok, bad and worst classes comparing with the requirements for agricultural production, then combined to become a P Index measuring the agricultural condition in an area of 1 square kilometer. The per capita size of bad areas is computed by extracting the size of bad and worst areas in a county out of the P Index image by using geographical data analysis software. The indicators of coping ability are: Average annual per capita grain production Per capita meat production Grain yield Per capita power of agricultural machinery Percentage of households with telephones Number of hospital beds per 10000 people Number of students in school per 10000 people Per capita area of least development The first three indicators are outcome indicators of food availability; the higher the food supply is less insecure. Per capita power of agricultural machinery is a proxy 5

measurement of the application of agricultural machinery that should be proportionate to the overall capacity in agricultural production and disaster resistance, thus this indicator is considered very important for us to understand the capacity of fighting against food insecurity risk. Percentage of households with telephones is a good indicator of the richness of a region, and it is highly correlated with per capita GDP, which is a composite indicator of output and consumption level. Number of hospital beds and number of students in school can play a role in understanding social development level, in terms of health facilities and education prevalence. They can not be regarded as very precise indication, because more information is needed to draw a complete picture, also the distribution of hospital beds and students between urban and rural areas is not clear. Nevertheless, the inclusion of these two indicators will refine the analysis to certain extent. Per capita area of least development is another GIS generated indicator. It covers four aspects of infrastructure, which can be a portrayal of development level: brightness of lights, distance to big roads, big rivers and towns. The composite index generated after combining the four sets of information is called D Index, and is categorized into good, ok, bad and worst classes, the same as the categorization of P Index. The per capita data used in this analysis is computed by summing up the bad and worst areas then divided by population. Altogether 13 indicators are used in the analysis, with five for risk assessment and eight for coping ability measurement. Compared with the indicators used in the 2000 analysis (see annex 3), this analysis has included more indicators of social development, which will make the results more reliable and objective. 6

Data processing Methodologies All the data purchased from NSB is imported into a pre-designed MS ACCESS database. Queries are established to compute necessary indicators. In order to map the analysis results later, each county is assigned a code that is used in mapping software. The map code is used to link all the tables and queries in the database. Clustering For the analysis, the clustering methodology that has been used in the previous analysis 1997 and 2000 is applied in this analysis. Cluster Analysis is a multivariate analysis technique that seeks to organize information about variables so that relatively homogeneous groups, or "clusters," can be formed. The clusters formed with this family of methods should be highly internally homogenous (members are similar to one another) and highly externally heterogeneous (members of one cluster are not like members of other clusters). This methodology has proven successful in classify a large number of observations and generate profiles for each cluster. With this method, it is much easier to understand the characteristics of various regions and make out the causes of vulnerability. The aim of clustering in this analysis is to classify the counties in China by the selected indicators. The 2000 plus counties are to be classified into a certain number of classes, depending on their similarity. Each cluster will have a different profile, which can be interpreted into a level of vulnerability/better-off. The cluster profile will also let us know the characteristics of each cluster so that the causes of vulnerability can be identified. Before clustering, all data were standardized first, and then the non-hierarchical method was adopted using the early warning version of ADDATI. Ranking Z-score ranking is also used to cross check the clustering results. Instead of giving clusters, the z-score ranking can give a composite index of each county. Three indexes are computed: Risk Index, Coping Index and Vulnerability Index. The sums of the z-score of the indicators under category Risk and Coping ability are Risk Index and Coping Index respectively. The Vulnerability Index is the Coping Index minus Risk Index because the higher the Risk Index is, the more vulnerable is. Then categorization is done to classify the counties into five groups according to the indexes, with the same number of cases in each category. The counties fall into the bottom categories (ranks 1 or 2 or 3) are more vulnerable than those in the top ones. Z-score ranking is carried out with SPSS, which provides descriptive information of the indicators as well. GIS data processing is done with Idrisi and MapInfo; both are famous geographical data analysis software. 7

Screening After running the clustering analysis and interpret the clustering profiles, a further screening of the most vulnerable counties needs to be done. The reason for this is that the cluster profiles can only give us the average of each indicator for each cluster. It is not surprising that in the most vulnerable clusters there are a certain number of counties exceeding the vulnerable criteria. In this process, the z-score ranking results are referred to. The percentage of rural population is also considered at this phase to exclude the places where rural population is lower than 65%, showing rural economy may not be dominant. The screening criteria for the most vulnerable clusters first focus on following aspects: the counties should be within the most vulnerable clusters and within the bottom two ranking categories. This way of screening give out the most vulnerable counties that falling into the worst categories by either clustering or ranking. In the meantime, the grain output, which is the key indicator for food security, should not exceed 400 kg per person per year. The vulnerable counties, which are less vulnerable than the above identified the most vulnerable counties, should fall into the bottom three ranking categories and the vulnerable clusters, and with a per capita annual grain production below 400 kg and percentage households with telephones below 36%. These counties are less vulnerable than the first group because they either have less risk or higher possibility of coping with the risks. The last groups are the less vulnerable counties, which have an average per capita grain production between 400kg and 500kg but a percentage of households with telephones below 15%. Considering the adverse natural conditions, some counties in Xinjiang and Inner Mongolia are included although the food production per capita could be higher than 500kg. The counties in the category, although they could maintain sufficient food supply, the possession of telephones shows that the income of these counties could be very low; hence they can become vulnerable when serious disasters happen. The counties out of these three levels of vulnerability are not considered vulnerable at this stage. It should be noted that intra-county disparity can be huge in some areas (the township level targeting analysis that have been done for the WFP/IFAD project counties are examples), therefore the non-vulnerable counties listed by this analysis can also have pockets of very vulnerable areas. 8

1. County clustering Results After the two-stage clustering (exploratory and optimization), the 2050 counties are classified into 12 clusters. The number of counties in each cluster varies from more than 580 to 6 with the first seven clusters including more than 90% of the analyzed counties. The profiles of each cluster are as shown in Table 1. Table 1 Cluster Profiles of County Analysis 2003 Risks Cluster Number of Counties Average Annual Growth Rate of Rural Population (99-01) (%) Percentage of Rural Population (2001) Population Density 2001 (person/sq. km) Per Capita Area of Adverse Physical Conditions (sq.km) (total population) Coefficient of Variation of Grain Production (97-01) Coping Ability Average Per Capita Annual Grain Production (97-01) (kg) (rural population) Per Capita Meat Production (2001) (kg) (rural population) Grain Per Yield Capita 1999 Power of (ton/ ha) Agricultu ral Machiner y 2001 (kw/perso n) (rural populatio n) Percen tage HHs with Teleph ones 2001 (%) No. of Hospital Beds Per 10000 Total Populati on 2001 No. of Student s in School Per 10000 Total Populati on 2001 Per Capita Least Develop ment Areas (total populati on) (sq.km/ person) 1 568 0.09 85 309 0.004 0.39 636 70 4.8 0.4 31 17 1566 0.006 2 458 2.11 92 288 0.004 0.38 500 62 4.2 0.3 19 11 1707 0.006 3 231-0.37 81 505 0.001 0.41 625 60 5.2 0.7 59 24 1502 0.003 4 224 0.00 75 197 0.013 1.11 1493 124 5.9 0.7 35 20 1411 0.016 5 173 2.08 90 666 0 0.35 621 59 5.1 0.7 33 12 1954 0.001 6 153-0.28 88 560 0.002 0.36 899 102 5.6 1.5 45 15 1852 0.003 7 106-3.61 61 500 0.033 0.41 803 99 5.5 1.1 74 40 1455 0.034 8 71-0.33 65 139 0.029 0.78 2997 298 7.2 1.4 39 21 1594 0.035 9 26 6.06 90 1216 0.005 0.41 314 36 5.6 0.4 88 16 1806 0.005 10 22 10.44 91 2 0.947-9999.00-9999 250-9999 0.3 16 23 996 0.901 11 12 2.06 87 579-9999 0.37 518 51 5.6 0.6 58 12 1704-9999 12 6 130.36 86 777 0.032 0.33 523 70 4.4 0.5 66 19 1520 0.032 Total/ average 2050 1.46 82 420 0.005 0.49 703 74 4.5 0.6 35 16 1672 0.007 In Table 1, the yellow cells highlight the values far worse than the average. The clusters with a higher number of highlighted indicators, the more possible the cluster is vulnerable. The Most Vulnerable Clusters It can be seen that Cluster 2 and 10 have six or seven indicators out of thirteen performing worse than the overall average, thus these two clusters are considered to be the most vulnerable. However, the two clusters show different characteristics in terms of vulnerability. Cluster two, which encompasses about 450 counties, has low grain and meat per capita production, which tell us that the food availability is relatively low. The other two key indicators of Cluster 2, the percentage of households with telephones and the number of hospital beds per 10000 people, are also very low compared with other clusters. In general, the coping ability of this cluster is very weak. The big proportion 9

of rural population is also noticeable in Cluster 2. By studying the composition of Cluster 2, it is found that more than 90% of the counties falling into this cluster are located in the central or western provinces of China, few are from costal areas. From Map 1, it can be found that most of the counties in Cluster 2 are located in the central part of China, also quite a few in southern part of Tibet. One can tell the difference between Cluster 10 and Cluster 2 almost at a glance. Cluster 19 is a typical husbandry cluster without grain production. In addition, the annual growth rate of rural population is very high, however, the number of students in school and the percentage of households with telephones are very low. It should be noted that the counties in this cluster suffer an adverse natural condition as well as under-developed infrastructure. All these mean this cluster is very vulnerable with high risks of food insecurity and low ability in coping. Counties in this cluster concentrate on the Tibetan Plateau across the border of Qinghai and Tibet. A few others in Inner Mongolia, Sichuan and Xinjiang. Vulnerable Cluster Immediately after Cluster 2 and 10, it is Cluster 5 and Cluster 1. Cluster 5 has an average agricultural performance and a low husbandry output. Population density of this cluster is very high but the health facility is less developed than many other clusters. Although the coping ability of this cluster is higher than that of Cluster 2, the overall capacity is still not satisfactory. This is also a cluster with very high population growth rate, showing the vulnerability could increase in the future if the coping ability remains at the same level. The counties in this cluster are mainly located in the middle and lower reaches of the Yellow River. Cluster 1 is an average cluster in terms of almost all the indicators except the number of students in school is a bit lower than the overall average, it is also the largest cluster with nearly 600 counties. The counties in this cluster are less rural-populated than the other vulnerable clusters and the population is increasing slowly. This cluster is regarded as vulnerable because the coping ability of this cluster is not very strong, although the risks of food security that this cluster faces is not so fierce as the most vulnerable clusters are facing. The largest number of counties in this cluster shows that a big proportion of counties in China share similar characteristics in terms of vulnerability. Clusters 11 and 12 are consisted of counties with a bit strange values. For example, the annual growth rate of rural population of Cluster 12 is over 130%. But examining the average values of the key indicators for income and social facilities, it is not difficult to find that the food supply is not very high although these two clusters have a certain degree of capability in dealing with risk. Non-Vulnerable Clusters Other clusters are considered not vulnerable because either their performance in agriculture and husbandry are very good, or their social development level is high, or they have both. 10

Cluster 3 is a cluster with very good social development marked by more than 58% of households with telephones and more than 23 hospital beds per 10000 people. The population is decreasing and the rural population is only 80% out of total population. This is a cluster encompass many of the counties along the coastal lines of China. Clusters 6 and 7 also have strong socio-economic capacities like Cluster 3. The difference is that they have a better food supply. In terms of population growth, Cluster 7 faces less risk than Cluster 6 and 3. Cluster 4 and Cluster 8 enjoy the highest food supply among all the clusters with per capita grain production over 1400kg and meat per capita production over 120kg. These two clusters are also not rural-concentrated clusters and the population density is low. Counties in these two clusters should be much better off than other counties. Counties in these two clusters are mainly located in the northeast and northwest of China where fertile and large pieces of land are available. Cluster 9 might be considered the most vulnerable were it not for two factors: the highest population density which is over 1200 person per square kilometer and the very high percentage of households with telephones; while the agricultural performance of this cluster is unexpected low. These contradictory values imply that this is a cluster that the income and development level is high but food production is not one of the dominant income sources. Second or tertiary industry can be the main pillar of local economy. Counties in this cluster mainly scatter in the coastal provinces. Map 1 shows the geographical distribution of the 12 clusters. 2. County Screening Clustering Results of Chinese Counties 2003 Heilongjiang Inner Mongolia Jilin Xinjiang Tibet Province Boundary County Boundary Qinghai Gansu Sichuan Liaoning Beijing Tianjin Hebei Shanxi Ningxia Shandong Shaanxi Chongqing Guizhou Henan Jiangsu Anhui Hubei Shanghai Zhejiang Jiangxi Hunan Fujian Yunan Guangxi Guangdong Taiwan Clusters 12 11 10 9 8 7 6 5 4 3 2 1 Hainan 11

As explained in the part of Methodology, a further screening of the counties in the vulnerable clusters is conducted. The analysis finally identified 350 counties as vulnerable in China. Among which, 154 counties are regarded as the most vulnerable. Their geographical distribution shows that the majorities are located in the central and western provinces in China, especially in Guizhou, Tibet, Shaanxi, Qinghai, Gansu, and Sichuan. Hebei, Shanxi, Anhui, Jiangxi, Henan, Hunan, Guangxi and Ningxia also have the most vulnerable counties to different extent. There are 62 counties identified as vulnerable counties. 134 counties are found to be least vulnerable according to this analysis. Like the most vulnerable counties, the vulnerable and less vulnerable counties are also mainly distributed in backward provinces in China. Table 2 Distribution of the Vulnerable Counties in China Province Number of Total Number In Which Counties in of the Number of Number of the Vulnerable the Most the Less Province Counties in Vulnerable Vulnerable the Province Counties Counties Number of the Least Vulnerable Counties Percentage of the Vulnerable Counties in the Province Beijing 5 0 0.0 Tianjin 4 0 0.0 Hebei 138 19 5 11 3 13.8 Shanxi 96 23 8 9 6 24.0 Inner 83 5 1 4 6.0 Mongolia Liaoning 43 0 0.0 Jilin 39 0 0.0 Heilongjiang 66 0 0.0 Shanghai 3 0 0.0 Jiangsu 56 0 0.0 Zhejiang 62 0 0.0 Anhui 61 9 2 5 2 14.8 Fujian 59 0 0.0 Jiangxi 77 6 4 1 1 7.8 Shandong 92 0 0.0 Henan 110 7 3 3 1 6.4 Hubei 64 2 2 3.1 Hunan 87 12 3 1 8 13.8 Guangdong 76 0 0.0 Guangxi 78 23 12 3 8 29.5 Hainan 17 0 0.0 Chongqing 26 0 0.0 Sichuan 138 26 4 8 14 18.8 Guizhou 78 54 26 2 26 69.2 Yunnan 119 41 20 1 20 34.5 Tibet 72 31 18 6 7 43.1 Shaanxi 87 21 11 3 7 24.1 Gansu 75 36 24 5 7 48.0 Qinghai 39 22 11 4 7 56.4 Ningxia 17 7 2 5 41.2 12

Xinjiang 83 6 6 7.2 Total 2050 350 154 62 134 17.1 It is apparent that the western provinces like Guizhou, Gansu, Ningxia, Qinghai, Yunnan and Tibet have the highest percentage of vulnerable counties (over 30%). Shaanxi, Shanxi and Guangxi, although have less vulnerable counties, the percentage is still over 20%. Hebei, Hunan, Anhui and Sichuan have more than 10% of counties regarded as vulnerable. Provinces of Xinjiang, Jiangxi, Hubei, Inner Mongolia and Henan have less than 10% of counties belong to the vulnerable category. In terms of the number of the most vulnerable counties, Guizhou, Gansu and Yunnan are on the top. The list of the vulnerable counties is attached in Annex 4. Map 2 gives the location of the vulnerable counties by vulnerability level. Vulnerable Counties in China Heilongjiang Inner Mongolia Jilin Xinjiang Qinghai Gansu Liaoning Beijing Tianjin Hebei Shanxi Ningxia Shaanxi Shandong Henan Jiangsu Vulnerable counties(350) Province Boundary Tibet Sichuan Anhui Hubei Shanghai Chongqing Zhejiang Jiangxi Hunan Guizhou Fujian Yunan Guangxi Guangdong Taiwan Hainan 13

Comparisons In order to monitor the changes in food security, comparisons between the results of this analysis and the previous analysis of 2000 is made. In addition, the list of vulnerable counties is compared with the Government priority county list of poverty reduction programme. 1. Compared with the results of Analysis 2000 In this analysis, altogether 350 counties in 18 provinces are identified as vulnerable counties of different level. In the analysis 2000, there were 459 counties in 20 provinces identified as vulnerable. Obviously, the overall food security situation in China has become better and fewer counties are vulnerable to food security risks. However, there are 276 counties remaining within the vulnerable category, in which 103 belong to the most vulnerable lists according to the findings of this analysis. These counties are mainly located in Sichuan, Guizhou and Yunnan. Provinces like Shaanxi, Ningxia, Qinghai, Xinjiang, Hunan, Guangxi and Gansu also have a few counties still being very vulnerable. There are a few counties, not identified vulnerable in the last analysis, fall into the vulnerable group in this analysis. These new vulnerable counties spread relatively evenly between provinces. Encouragingly there are about 175 counties getting out of the vulnerable list. These counties can be found mainly in Hebei, Jiangxi, Henan, Hunan, Sichuan, Yunnan, Shaanxi and Xinjiang. The counties having been excluded from the vulnerable list account for more than 38% of the 459 vulnerable counties identified in 2000. Map 3 and Table 3 gives the comparison details. Map 3 Vulnerable Counties Identified in 2000 and 2003 Heilongjiang Inner Mongolia Jilin Xinjiang Qinghai Gansu Liaoning Beijing Tianjin Hebei Shanxi Ningxia Shandong Shaanxi Henan Jiangsu Tibet County Boundary Province Boundary Counties identified as vulnerable in 2003 analysis only (71) Counties identified as vulnerable in 2000 analysis only (171) Vulnerable counties in both 2000 and 2003 analysis (279) Sichuan Anhui Hubei Shanghai Chongqing Zhejiang Jiangxi Hunan Guizhou Fujian Yunan Guangxi Guangdong Taiwan Hainan WFP China Office, July 2003 The boundaries and names shown and desinations used on this map do not imply official endosement or acceptance by United Nations 14

Table 3 Comparison of the Vulnerable Counties Identified by Analysis 2000 and Analysis 2003 Province Number of Counties in the Province New vulnerable counties identified by analysis 2003 Counties remaining in the vulnerable list of 2003 Counties getting out of vulnerable list Beijing 5 0 0 Tianjin 4 0 0 Hebei 138 5 14 3 Shanxi 96 4 19 15 Inner Mongolia 83 5 0 Liaoning 43 0 0 1 Jilin 39 0 0 2 Heilongjiang 66 0 0 Jiangsu 56 0 0 Zhejiang 62 0 0 Anhui 61 3 6 5 Fujian 59 0 0 Jiangxi 77 0 6 12 Shandong 92 0 0 Shanghai 3 0 0 Henan 110 3 4 13 Hubei 64 1 1 4 Hunan 87 0 12 16 Guangdong 76 0 0 Guangxi 78 6 17 10 Hainan 17 0 0 Chongqing 26 0 0 4 Sichuan 138 5 21 19 Guizhou 78 5 49 6 Yunnan 119 1 40 21 Tibet 72 19 12 4 Shaanxi 87 0 21 24 Gansu 75 2 34 7 Qinghai 39 10 12 3 Ningxia 17 0 7 1 Xinjiang 83 2 4 10 Total 2050 71 279 180 Map 3 2. Comparison between Analysis 2003 and the Priority Counties of Poverty Reduction of GOC In 2001, the Chinese Government release of list of 592 counties as the priority of poverty alleviation programmes. As the data used in this analysis are from the year 1999 to 2001, it is interesting to compare the analysis results with the GOC list, although we don t know exactly the indicators that GOC used. 15

Table 3 Comparison between the Vulnerable List of WFP 2003 and GOC List Province Total Vulnerable Priority Number Percentage of Number counties counties for of counties of identified by GOC counties falling into Counties analysis poverty falling into both WFP list 2003 alleviation the two lists and GOC priority list / 2003 vulnerable list Beijing 5 0 0 0 Tianjin 4 0 0 0 Hebei 138 19 39 17 89.5 Shanxi 96 23 35 19 82.6 Inner Mongolia 83 5 31 2 40.0 Liaoning 43 0 0 0 Jilin 39 0 8 0 Heilongjiang 66 0 14 0 Jiangsu 56 0 0 0 Zhejiang 62 0 0 0 Anhui 61 9 19 7 77.8 Fujian 59 0 0 0 Jiangxi 77 6 21 4 66.7 Shandong 92 0 0 0 Shanghai 3 0 0 0 Henan 110 7 31 6 85.7 Hubei 64 2 25 2 100.0 Hunan 87 12 20 9 75.0 Guangdong 76 0 0 0 Guangxi 78 23 28 17 73.9 Hainan 17 0 5 0 Chongqing 26 0 14 0 Sichuan 138 26 36 11 42.3 Guizhou 78 54 50 43 79.6 Yunnan 119 41 73 36 87.8 Tibet 72 31 0 0 0.0 Shaanxi 87 21 50 20 95.2 Gansu 75 36 43 33 91.7 Qinghai 39 22 15 12 54.5 Ningxia 17 7 8 7 100.0 Xinjiang 83 6 27 6 100.0 Total 2050 350 592 251 71.7 16

It can be seen that about 70% of the vulnerable counties identified in this analysis match with the government priority counties for poverty alleviation. It should be mentioned that Tibet is not included in the government list so the overlapping rate of Tibet is zero. The difference between the two lists is understandable, because WFP identifies the vulnerable counties in terms of food insecurity risks, while the GOC list is based on the poverty situation of a county, which is often measured by income. It is recommended that the counties that fall into the two lists should be paid more attention in WFP programming as well as other interventions. Map 4 Comparison between GOC Priority County List and WFP Vulnerability Countiy List Heilongjiang Inner Mongolia Jilin Xinjiang Qinghai Gansu Liaoning Beijing Tianjin Hebei Shanxi Ningxia Shandong Shaanxi Henan Jiangsu Tibet Province Boundary County Boundary Counties identified as vulnerable in 2003 analysis only (99) Counties designated as priority by GOC only (341) Counties both identified as vulnerable and as GOC priority (251) Sichuan Anhui Hubei Shanghai Chongqing Zhejiang Jiangxi Hunan Guizhou Fujian Yunan Guangxi Guangdong Taiwan Hainan 17

Conclusions 1. The overall vulnerability level of Chinese counties is going down. Although the indicators used in analysis 2000 and in this analysis are different to some degree, the overall principles guided the two analyses are the same, which is the comparison between food security risks and the coping ability. It is found that the number of vulnerable counties is about 5% less in analysis 2003 than in analysis 2000. The vulnerable list is changing dynamically with more counties excluded from the list than those newly included. 2. Regional disparity in terms of the number of vulnerable counties is significant. The eastern provinces of China are exempt from vulnerable counties in this analysis. Among the central and western provinces, the percentage of vulnerable counties varies from 3% to nearly 70%. In general, the western provinces have more vulnerable counties than those in the center. This means that although there have been great improvement in rural economy, provinces like Guizhou, Gansu, Ningxia, Shaanxi, Tibet and etc. still lag behind many other provinces and need great support and assistance from both domestic and overseas agencies to help them alleviate poverty. 3. Grains supplies of Chinese counties have changed better. More than 90% of the counties in the analysis have their per capita grain production increased in the period of 1997 to 2001. The grain supplies go down in only 40 counties. In 2003 analysis, there are about 350 counties with per capita grain production below 400kg, but in 2000 analysis, the number was about 570. For those that have an extremely low food production below 250kg, in 2003 analysis the number is about 100 counties, while in the 2000 analysis there were more than 140 counties. According to the Paired Samples Test of per capita grain production 2, the geometry mean of per capita grain production of the period of 1993-1998 is 595kg, while that of the period of 1997 to 2001 is 704kg. The difference between the two periods is statistically significant (t=126.857). This apparently shows that the grain supplies in the Chinese counties have greatly improved in recent years. Frequency Difference of Per Capita Grain Production between 93-98 and 97-01 14000 12000 10000 8000 6000 4000 2000 0-200 - -180-180 - -160-160 - -140-140 - -120-120 - -100-100 - -80-80 - -60-60 - -40-40 - -20-20 - 0 0-20 20-40 40-60 60-80 80-100 100-120 120-140 140-160 160-180 180-200 200-220 220-240 240-260 260-280 280-300 kilograme 2 Per capita grain production of 2003 analysis and 2000 analysis are weighted by rural population of 2001. The same is with the comparison of rural population growth rate and per capita meat production. 18

Discussions 1. The data used in this analysis can better reflect the comprehensive economic capacity of a county and its ability in coping with food insecurity risks because information related to income, such as percentage of households with telephones is available. Although the information about public health and education are still weak, adding the number of hospital beds and students in school per 10000 people have greatly helped to broaden our understanding of a county. 2. In this analysis, the GIS information is used for the first time in China and this attempt is regarded as a good start for making more use of GIS data in economic analysis. The correlation study shows that there is no significant relation between socio-economic data and GIS information, including areas of adverse physical conditions and areas of least development. This means that these GIS information can not only help to give us a full understanding of the counties, but also refine the analysis results. This result shows that GIS information and socio-economic information cannot substitute each other. 3. This analysis focuses on the rural population in a county. However, some indicators, especially those related to social development, can not be disaggregated by urban and rural areas. Considering that most of the counties have more than 85% of rural population, the indicators calculated by the number of total population are still comparable among counties. 4. The methodology that used in this analysis has combined the results of ranking and clustering. This is the first time that both the results have influenced the final list of vulnerable counties in China, because in the past, ranking was only used as a reference. The attempt to adopt these two methods is aimed to explore more comprehensive ways in vulnerability analysis. 5. Due to the changes of NSB data sets after 1998, the indicators used in this analysis are not exactly the same as that used in Analysis 2000, which has made the comparison between the two analyses not easy. However, the comparison between per capita grain production is still meaningful to tell us the improvements that have taken place in many counties. 19