Project County Selection of WFP Interventions
Under Country Programme 2001-2005
Han Zheng
WFP/IFAD China
Vulnerability Analysis and Mapping (VAM) Unit
July 2002
1
Background
In the WFP China Country Programme of 2001-2005, 25 counties in south of Gansu province are identified as candidate counties of WFP Gansu Intervention. This identification was agreed between WFP and MOA, based on 1) the VAM 2000 China County Analysis, which put forward 459 counties as most vulnerable and eligible for further study, and 2) the project counties suggested by the Government.
The 25 counties are located in the central and south part of Gansu, which is on the stretches of Minshan Mountains bordering the Tibetan Plateau and extends onto the Loess Plateau and Qingling Mountains. More than 80% of the total area is mountainous and suffering from frequent natural disasters.
Due to limited resource, further targeting is needed to scale down the project area and concentrate the food aid on the most needy people, therefore, further county and township targeting is required1.
General Situation of the Candidate Counties2
South of Gansu is home of quite a number of national poverty-stricken counties. The 25 candidate counties for WFP intervention are among the 43 priority counties designated by the Chinese Government for poverty reduction.
According to the data collected from the 25 counties, the total population of the 25 counties is more than 8 million, living in an area of 6 million hectares. Rural labor force accounts for about 50% of the rural population. The average mountainous area within a township is 63% and the cultivated land is only 0.14 ha per capita. Various natural disasters hit the area frequently and the agriculture is backward. The average per capita grain production in the 25 counties is less than 250 kg in latest five years and the average rural net income of 2001 is about 1100 Yuan.
The main crops of food supply are wheat, maize, tuber and some beans. Rice is rarely planted in some counties on the south border of Gansu where climatic conditions is a little better. The crop-planting pattern differs largely in different counties and the yield of each crop varies. The average wheat yield per mu is 130 kg, and the yield of maize is 200kg. The annual variability of grain is about 0.2, means the grain production is not very stable between years. In some township, this indicator can reach 0.6. Chart 1 shows the average composition of grain output in the 25 counties. Wheat production accounts for one-third of the total grain production, however, maize and potato have a proportion of about 50% of the total outputs.
1 Maps in this report are prepared by Yu Jing, VAM Assistant of WFP China. 2 All analysis of this report are based on township data provided by Gansu PMO.
2 Chart 1 – Composition of Grain Production in Candidate Counties
Composition of Grain Production
11%
7%
34%
wheat/grain maize/grain tuber/grain bean/grain 24% others/grain
24%
Husbandry is important in the life of farmers in this region. The per capita pasture land is about 1.4 mu. However, the livestock raising shows no increase in the number of goats and/or sheep, and big animals raised by each household. No trend of increase in pig raising as well. In average, each household has one draught animal, and two households have three pigs and goats and/or sheep. Chart 2 shows the situation.
Chart 2 – Livestock Raising in the Candidate Counties in Last Five Years
Livestock Raising
1.80
1.60
1.40
1.20
1997 1.00 1998 1999 2000 0.80 2001
0.60
0.40
0.20
0.00 big animal per household goat&sheep per household pig per household
3 The data of adult education show that female education is worse than that of men. Although there are quite a proportion of adult have gone to primary school, there are more than 40% of female never entered school. The primary education today is also worrying because the drop out rate is more than 10% for both boys and girls.
Chart 3 – Adult Education in Candidate Counties
Adult Education
never go to school
female adult average
primary school graduates
0.00 5.00 10.00 15.00 20.00 25.00 30.00 35.00 40.00 45.00 50.00
Infrastructure is backward in this region. About one-third of townships don’t have irrigation, while the irrigated land is about 15% of the total cultivated land in the townships that have irrigation. Only 25 percent of administrative villages have tap water, 70% of administrative villages have a village clinic.
Methodology
1. County selection criteria
It is well known that in China the intra-county disparity is huge in terms of economic development and physical conditions, thus the best way to targeting the vulnerable area at this stage is to cluster the townships and identify the most vulnerable ones. The counties with high percentage of vulnerable townships or vulnerable people should be considered for assistance, and the vulnerable townships within the county can be project townships.
2. Data and map collection
4
In accordance to the above principle, which has proved successful, township data should be collected from all the 594 townships in the 25 candidate counties. The data collection form was prepared by WFP Beijing based on the field observations of a VAM mission to that region. Based on the county data analysis and the field visit in the candidate counties, it can be seen that vulnerability of people in this areas is highly related to food production, livestock raising, income sources and social development. People in this area are living heavily on their own production, with limited supplementation of migration income. Livestock raising is popular but at a very small scale, mainly for cash income. Social infrastructure, especially road, drinking water and hospital, is backward and hinders the local development. Many people living in mountains at a high altitude above sea level don’t have much access to markets and other social facilities. Therefore, information about agriculture, income, disaster, husbandry, road, clinics and education are collected.
Gansu Provincial PMO provided all the required data to WFP Beijing on time and worked closely with the VAM Unit on data verification and revising. In the meantime, maps of the 25 counties with township boundaries and the location of administrative villages are digitized and processed in order to illustrate the final analysis results.
3. Data verification and revising
Data verification is essential to the analysis. After the data is sent to Beijing, they are imported to a MS ACCESS database for data verification. There are about 150,000 figures in the data set and the first round checking marked less than 10% skeptical or wrong data, which is much less than some other project provinces. The skeptical or wrong data are sent back to Gansu for revising and clarification. The checking-and- revising circle went twice until the data set is clean and reliable.
4. Indicator Selection
The indicator selection is in accordance with WFP definition of vulnerability, which is a composition of risk to food security and inability to cope with the risks. The higher the risk and the lower the ability to cope, the higher the vulnerability of population/area is. Social development level is also considered a part of the coping capacity. In order to obtain a comprehensive profile of the area and distinguish the vulnerable townships from the others, indicators covering demography, farming, husbandry, income, natural disaster, social infrastructure and education are calculated. By screening the indicators, only those that are available for most of the townships are studied further. For example, the indicators about natural disasters have to be excluded although they are important for that they are available for only a small number of townships. For the indicators of farming, husbandry and income, average annual growth rate and/or coefficient of variation between years are computed to show the growth trend in recent years.
Following indicators are included in further analysis.
5 Table 1 – Indicators Used in Township Vulnerability Analysis of Gansu
No. Indicators Category 1 Annual growth rate of rural population Risk (97-2001) (%) 2 Average per capita grain production Risk (kg) (97-2001) 3 Coefficient of variation of per capita Risk grain production (97-2001) 4 Per capita cultivated land (ha) (2001) Risk 5 Per capita agricultural land (ha) (2001) Risk 6 Per capita pasture land (ha) (2001) Risk 7 Per capita rural net income (yuan) Coping ability (2001) 8 Annual growth rate of rural net income Coping ability (97-2001) (%) 9 Average no. of goats & sheep per Coping ability household (98-2001) 10 Average no. of pigs per household (98- Coping ability 2001) 11 Annual growth rate of pig raising (98- Coping ability 2001) 12 Annual growth rate of goats and sheep Coping ability raising (98-2001) 13 % Of non-grain sown area/crop sown Coping ability area (2001) 14 %Village with tap water (2001) Social development 15 % Of primary dropouts (2001) Social development 16 % Of primary female dropouts (2001) Social development 17 % Adults never going to school (2001) Social development 18 % Female adults never going to school Social development (2001) 19 % Villages with complete primary Social development school (2001) 20 % Of villages with village clinics (2001) Social development 21 Dependent ratio Social development
The correlation between indicators is computed to find out the relations between indicators (Annex 1). From the correlation sheet, statistically significant correlations between certain pairs of indicators are not found. But the general relations can be seen. For instance, there is a positive relation between the average per capita grain production and average grain yield. The negative relation between per capita net income and primary dropout rate is also very reasonable. The results of correlation revealed the complexity of the causes of vulnerability in this region.
5. Methodologies of Township Analysis
6 The objective of the analysis is to cluster the 594 townships into groups with different characteristics. A programme of ADDATI, developed by the Venice University, is used. Through data standardization, correlation analysis and Principal Component Analysis (PCA), the programme clusters all the townships into a number of groups according to the manually set criteria, for example, number of clusters, number of partitions.
After PCA, active variables, which will influence the clustering results, are identified. The other indicators are regarded as supplementary variables, which help to understand the township profiles.
Table 2 – Active Variables of Clustering
No. Indicators Abbreviation in Clustering 1 Annual growth rate of rural population grpop (97-2001) (%) 2 Average per capita grain production pcgrain (kg) (97-2001) 3 Coefficient of variation of per capita cvgrain grain production (97-2001) 4 Per capita agricultural land (ha) (2001) agriland 5 Per capita rural net income (yuan) income (2001) 6 Annual growth rate of rural net income grincome (97-2001) (%) 7 Annual growth rate of pig raising (98- grphpig 2001) 8 Annual growth rate of goats and sheep grphyang raising (98-2001) 9 % Of non-grain sown area/crop sown nongrain area (2001) 10 %Village with tap water (2001) tap 11 % Of primary female dropouts (2001) fedrop 12 % Female adults never going to school femaleno (2001) 13 % Of villages with village clinics (2001) clinic
After the clustering by the programme, the profiles of each cluster are interpreted and further screening of the townships in vulnerable cluster is done. The number of vulnerable townships in each county, as well as the number of rural people living in the vulnerable townships are calculated in order to compare the concentration of vulnerable people in different counties.
Analysis result
1. Township profiling
The 594 townships of 25 candidate counties were grouped into 15 clusters. The table below shows the profiles of each cluster. From the table we consider Cluster 1, 3, 4, 5, 7, 8, 10, 13,14 could be considered for further study, because the majority
7 indicators of these clusters show a worse situation than the weighted average of all the townships. Each of the vulnerable cluster has several problems, but all of them don’t have enough food production, and the food gap can not be easily compensated by their poor income. Some of the clusters even have a poor or deteriorating husbandry, some have a high population growth pressure. The education or drinking water is worse compared with the good clusters.
Cluster 11 has very low food production, however, the income of that cluster is high and growing at a rapid speed. Cluster 6 and 12 are rich clusters with high food production, high income and good husbandry, although there are some problems with the drinking water. Cluster 2 is a large group with average situation in almost all aspects except village clinics.
Table 3 – Cluster Profiles
Cluster no. of no.of grpop Pc Cv Agri In gr grphpig gr non tap Fe Female clinic town indicator grain grain land come income phyang grain drop no ships s worse than average 1 76 7 0.73 226.5 0.16 0.56 703.6 6.35 -2.3 -2.01 20.26 9.76 32.38 36.88 37.23 2 76 6 0.19 301.8 0.16 0.8 1215 13.47 -2.1 -4.59 18.23 21.76 17.79 37.85 39.43 3 75 7 0.17 247.1 0.14 0.63 714.4 6.64 5.54 5.08 14.47 15.48 12.05 63.48 87.6 4 67 8 0.72 181.9 0.32 0.47 699.1 10.31 -1.27 -8.18 18.87 12.94 8.67 57.84 91.17 5 59 7 1.26 281.6 0.12 0.3 888.7 2.52 -9.57 -3.94 22.14 15.4 6.96 24.65 90.91 6 46 3 -0.13 354.8 0.11 0.37 1350 9.74 2.15 4.77 26.04 9.56 11.39 26.18 88.62 7 43 7 0 228.6 0.13 0.27 1118 7.37 -4.83 -5.55 20.04 79.29 6.53 47.44 89.97 8 35 8 1.43 169.0 0.33 0.38 734.2 3.04 -0.67 4.46 17.52 5.84 27.12 23.08 92.93 9 29 3 0.78 153.6 0.17 0.12 1845 11.63 -2.25 -1.82 35.36 55.85 2.84 21.53 100 10 28 9 0.85 190.0 0.24 0.59 625.2 4.11 -2.87 -2.39 9.36 81.41 34.77 79.69 32.61 11 25 5 1.3 191.6 0.36 0.17 1481 13.5 3.35 -0.68 25.16 10.92 14.88 31.23 97.08 12 16 3 1.12 437.2 0.07 0.44 2018 10.61 9.68 9.57 27.26 7.16 3.03 8.45 100 13 12 9 1.27 269.4 0.23 0.25 951.8 6.89 -22.13 -76.79 17.65 24.9 4.13 34.45 76.35 14 6 7 -0.45 304.7 0.06 27.63 638.9 3.18 -1.7 -10.26 5.18 27.8 8.36 81.82 40.67 15 1 6 0.95 209.0 0.25 0.28 1170 15.92 2.33 2.63 -823.8 9.52 1.88 45.99 85.71
Average 594 13 0.67 244.3 0.19 0.5 1062 8.15 -1.42 -2.62 19.15 23.05 13.74 37.81 80.83
The clustering results are illustrated in Map 1, on which the townships are colored according to the clusters they belong to. The color of red series represents the vulnerable clusters, while the green series stands for good ones. Some counties, such as Weiyuan, Tongwei, Zhouqu and Wen Xian, are dominated with green or blue colors, in that the townships of those counties are vulnerable compared with others. The counties, such as Li Xian, Danchang, Wudu, Lintan, Dongxiang, are in red or pink color, showing the townships in these counties are grouped into cluster 1, 3, 4, 7 and 8, which are considered vulnerable. There are some counties that red or green colors cover almost equal areas of the county, for example, Qin’an and Huining.
Within the red series, which highlighted the vulnerable townships, the characteristics of the townships are not exactly the same. In other words, different color can represent different characteristics, or tell different reasons for vulnerability. Take Dangchang county as an example. Townships in the west and east falling into cluster 4 (brown), which are poor in agricultural land, has low grain production and income. In addition, husbandry in these townships is not increasing. There are problems with
8 education and drinking water as well. The townships in the north belong to cluster 3 in pink, which is better in general than cluster 4, but adult education is worse. The other townships in yellow are quite different from the other two clusters, because the population growth rate of these townships is very high, but their income is decreasing, although the average income of these townships are not very low and the grain production is ok.
Map 1-- Clustering Results
Vulnerable Clusters in 25 Candiate Counties in Gansu
Huining
YuzhongYuzhong
JishishanJishishan DongxiangDongxiang
LinxiaLinxia GuangheGuanghe
ZhuanglangZhuanglang HezhengHezheng ZhuanglangZhuanglang TongweiTongwei
Weiyuan Weiyuan LongxiLongxi
Qin'anQin'an ZhangjiachuanZhangjiachuan
LintanLintan Gangu ZhangZhang XianXian WushanWushan
MinMin XianXian
County Boundary
DangchangDangchang LiLi XianXian Vulnerable cluster 1 (76) XiheXihe 2 (76) 3 (75) 4 (67) 5 (58) 6 (46) ZhouquZhouqu 7 (42) 8 (36) 9 (29) 10 (29) 11 (25) 12 (16) KangKang XianXian 13 (13) WuduWudu KangKang XianXian 14 (6) 15 (1) WenWen XianXian
WFP/IFAD China VAM Unit, July 2002
9
For each county, similar in-depth description of the townships can be extracted from Map 1 and Table 3. In order to underline the major causes of vulnerability in each township, a further classification based on the clustering results was done.
According to the concept of vulnerability, all the used indicators were categorized into “”Risk , “Coping Ability”and “Social development”(see Table 1). If within each category, a cluster has more than 50% of indicators falling below the overall average, it is considered that this category can be a cause of vulnerability for this cluster. As a result, the causes of vulnerability can be represented by a combination of alphas:
• R -- high risk of food insecurity • C -- low coping ability • S -- backward social development
Another letter was added to the vulnerable clusters (in grey) to represent the aspect, which is the worst, compared with other indicators of the cluster profile. The alphas are:
• f -- very low grain production • m -- very low per capita net income • h -- poor husbandry • e -- very poor education
Table 4 -- Causes of Vulnerability of Each Cluster
Cluster Risk Coping Social Overall (4 indicators) (5 indicators) development (4 indicators) 1 1 3 3 CS-m 2 0 2 3 S 3 1 3 3 CS-e 4 3 3 2 RCS-fhe 5 2 4 1 RC-h 6 1 1 1 7 2 3 1 RC-f 8 4 2 2 RS-f 9 2 1 0 10 2 4 3 RCS-fe 11 4 0 1 12 2 0 1 13 3 5 1 RC-h 14 0 5 2 CS-m 15 4 0 2 RS
10 Map 2, showing the further classification results, is a helpful tool for understanding the causes of vulnerability in each township, as well as county.
Map 2 -- Causes of Vulnerability in Vulnerable Townships
Major Causes of Vulnerability of Vulnerable Township in 25 Candidate Counties
HuiningHuining
YuzhongYuzhong
JishishanJishishan DongxiangDongxiang
LinxiaLinxia GuangheGuanghe
ZhuanglangZhuanglang HezhengHezheng ZhuanglangZhuanglang TongweiTongwei
WeiyuanWeiyuan WeiyuanWeiyuan LongxiLongxi
Qin'anQin'an ZhangjiachuanZhangjiachuan
LintanLintan GanguGangu ZhangZhang XianXian WushanWushan
MinMin XianXian
County Boundary
DangchangDangchang LiLi XianXian
Causes of Vulnerability XiheXihe CS-m (76) CS-e (75) RCS-fhe (67) RC-h (58) ZhouquZhouqu RC-f (42) RS-f (36) RCS-fe (29) RC-h (13) CS-m (6) KangKang XianXian WuduWudu KangKang XianXian
WenWen XianXian
WFP/IFAD China VAM Unit, July 2002
11
2. Proposed County and Township Selection
There are more than 400 townships falling into the vulnerable clusters for further study. Since a cluster profile represents the average situation of a cluster, while there are disparity within a cluster, strict criteria is introduced to eliminate those township that are relatively more capable to cope food insecurity than others. The criteria is 1) the average per capita grain production over 250kg; and 2) the per capita net income of 2001 is over 1000 yuan. After this screening, 197 townships in 21 candidate counties are remained in the vulnerable township list, with a total rural population around 2.6 million.
Obviously the vulnerable townships are scattering in everywhere of the region. The screening so far has successfully identified the vulnerable pockets, but yet narrowed down the project area. Therefore the next step is to find out the counties with a high percentage of vulnerable townships. Through the following table, we can identify ten counties with about 45% or above townships or rural population falling into the most vulnerable category.
The highlighted counties in the table are those have a higher percentage of the most vulnerable townships than other counties, they are: Dangchang, Dongxiang, Wushan, Li Xian, Wudu, Jishishan, Gangu, Min Xian, Lintan and Qin’an. There are three counties at the bottom of the list don’t have any townships falling into the worst situation. Those that have a low percentage of the most vulnerable townships can be considered not suitable for the interventions of food aid, but the most vulnerable townships should not be forgotten in other projects supported by either the government itself or with international assistance.
The most vulnerable townships in the above 10 counties should be considered to be included in the intervention. The list of eligible townships is in Annex.
Table 5 – Distribution of the Most Vulnerable Townships
County No.of townships No. of the most % of the most % of rural vulnerable vulnerable population townships townships living in the most vulnerable townships Dangchang 31 24 77.4 77.2 dongxiang 25 18 72.0 66.7 Wushan 20 13 65.0 57.3 Lixian 36 22 61.1 53.0 Wudu 44 26 59.1 61.5 Jishishan 18 9 50.0 47.5 Gangu 20 10 50.0 37.4 Minxian 23 11 47.8 42.2 Lintan 19 9 47.4 48.1 Qin'an 22 9 40.9 45.2 Huining 32 13 40.6 35.3 Zhouqu 22 8 36.4 33.6
12 Wenxian 25 8 32.0 34.4 Zhangjiachuan 19 6 31.6 36.9 Tongwei 23 6 26.1 25.8 Hezheng 14 3 21.4 23.7 Guanghe 10 2 20.0 13.1 Linxia 27 4 14.8 14.3 Weiyuan 20 2 10.0 6.7 Longxi 25 2 8.0 7.4 zhuanglang 23 1 4.3 4.3 Kangxian 28 1 3.6 3.5 Yuzhong 27 ------Xihe 24 ------Zhangxian 17 ------
Table 6 – Major Statistics of the Vulnerable Townships in the Ten Recommended Project Counties
Prefecture County No. of No. of No. of Total area Rural Average per Per capita Per capita vulnerable vulnerable admin. of the population capita grain agricultur net income townships households villages in vulnerable in the production al land 2001 the townships vulnerable (kg) in the (ha) of the (Yuan) of vulnerable townships vulnerable vulnerable the townships townships townships vulnerable townships Longnan Dangchang 24 50250 264 263005 211596 208 0.98 586 Longnan Lixian 22 53525 346 631349 258606 193 0.90 722 Longnan Wudu 26 70979 456 233272.3 291253 182 0.72 716 Linxia Dongxiang 18 32011 165 103393.8 170158 174 0.44 625 Linxia Jishishan 9 19056 70 30013 95250 210 0.18 594 Tianshui Wushan 13 50818 253 152627 228073 197 0.51 598 Tianshui Gangu 10 41492 203 94233 203561 146 0.39 616 Tianshui Qin’an 9 50233 213 67934 251307 160 0.12 594 Dingxi Minxian 11 38760 170 166632 171818 181 0.47 654 Gannan Lintan 9 16117 68 80675 65125 193 0.18 951
Overall 10 151 423241 2208 1823134 1946747 184 0.49 666
Location of the 25 candidate counties and the 10 proposed counties
Following maps showing the location and main indicators of the selected counties and townships. It can be seen that the 10 proposed counties are geographically continuous and the vulnerable townships within the 10 counties are not scattered. The concentration of the vulnerable townships can be helpful for the project designing and management. The ten proposed counties are in four prefectures, namely Linxia, Tianshui, Longnan and Dingxi. All the 10 proposed counties are to the south of Lanzhou, the capital of Gansu province. The southern counties are on the extent of the Qinling Mountains, and the counties such as Qin’an, Gangu and Wushan are on
13 the stretches of the Loess Plateau. The counties in Linxia prefecture are on the edge of the Tibetan Plateau and have a high proportion of minority people.
Map 3 – Location of the Candidate Counties, Proposed Counties and townships
Candidate Counties in Gansu (2002)
Heilongjiang
InnerInner MongoliaMongolia JilinJilin
10 Vulnerable Counties in Gansu Xinjiang LiaoningLiaoning Beijing Gansu TianjinTianjin Hebei Ningxia Hebei Shanxi BaiyinShiBaiyinShi Shanxi Shandong Qinghai
shaanxishaanxi Henan JiangsuJiangsu
Anhui TibetTibet Hubei Shanghai Sichuan Hubei Chongqing Chongqing ZhejiangZhejiang JiangxiJiangxi Hunan Guizhou FujianFujian
Yunnan Candiate Counties in Gansu (25) Guangxi TaiwanTaiwan Guangdong Province Boundary
Hainan YuzhongYuzhong WFP/IFAD China VAM Unit, July 2002 HuiningHuining LanzhouShiLanzhouShi JishishanJishishan DongxiangDongxiang LinxiaLinxia LinxiaLinxia linxialinxialinxia DingxiDingxi HezhengHezheng PingliangPingliang TongweiTongwei WeiyuanWeiyuan WeiyuanWeiyuan LongxiLongxi
Qin'anQin'an ZhangjiachuanZhangjiachuan LintanLintan GanguGangu ZhangZhang XianXian WushanWushan TianshuiTianshui
MinMin XianXian GannanGannan
LiLi XianXian DangchangDangchang XiheXihe LongnanLongnan ZhouquZhouqu
WuduWudu KangKang XianXian
WenWen XianXian Vulnerable Townships
County Boundary
Prefecture Boundary
name Proposed Project Counties WFP/IFAD China VAM Unit, July 2002
14
Suggestions on following steps
1. The final county selection will be determined among WFP, IFAD, MOA and local PMOs through thorough discussions. Following criteria could be considered when the final county selection is made: