(Item 4 of the Agenda)
11. Five papers from Thailand, Indonesia, the Philippines, CABIG and Eurostat on existing methodologies for collection of farmer income data were presented in this agenda sub-item.
The Farmer Income Statistics Survey in Thailand
12. In STAT-INCOME-4, Ms Sudjai Chongvorakitwatana, Senior Economist of the Division of Farm Households Socio-economic Research, Bureau of Agricultural Economic Research, presented the features of the process on how farmer income statistics were generated in Thailand.
13. She pointed out that farm income data was collected through the Socio-economic and Labour Force Household Survey usually conducted every two years as a data support in the monitoring and assessment of the attainment of goals of the National Economic and Social Development Plan. The survey uses a two-stage stratified sampling design with farming activities as the stratification variable, the villages as the primary sampling unit (PSU) and the households as the secondary sampling unit (SSU). A total of 3 000 villages representing 4.41 percent of all villages are proportionally allocated to the seven strata (agricultural activities). In a sample village, the household is a qualified respondent if they reside in the village for over 6 months and engage in farming activities (not necessarily within the village). Eight households are drawn in a sample village: four are used as the samples while the other 4 serve as possible substitute samples. The survey collects a wide variety of information on various income sources of the farmer as well as the possible determinants/correlates of such incomes.
14. She indicated that the growing cost of survey operations had resulted in requests for budget increases in every survey round. The major users of the data were policy-makers, especially in the agriculture sector, wishing to formulate options that could raise farmer income and improve their well-being. Ms Chongvorakitwatana indicated that farm household income averaged around US$3 600 a year.
15. She added that two steps were being considered to improve the collection of socioeconomic and labour force data in the future. First, the use optical scanning technology in digitizing the data to improve efficiency in data processing. Second, the conduct of a major survey every two years and a minor survey (i.e., with a reduced sample size) inbetween for updating purposes.
16. When asked about techniques to encourage farmers to participate in the surveys, she clarified that some farmers were paid for their time. The experts noted that increases in budget for socio-economic surveys were difficult to pursue due to budget allocation priorities within the countries. They agreed that requests for budgetary increases could be facilitated when the decision-makers were convinced on the importance of the data produced from the survey. The experts also noted the growing importance of offfarm income, which in Thailand averaged at around two-thirds of the total farm income. It was noted that non-farm income was a way of hedging against seasonal income and vulnerable harvests as well as a result of national programmes promoting diversification of income sources. The Chair agreed that measurement of non-farm income was an important issue and even raised the issue of what aspects of farmers' income should really be measured, which would depend on the intended end use of the data. Regarding optical scanning, some experts mentioned negative experiences due to poor scanning quality. In some countries, the technique resulted in double work as scanned survey data was often unclear.
Methodology of Data Collection in Farm Income Surveys: Indonesia's Experience
17. Mr Ardief Achmad, Director of Agriculture Statistics, BPS-Statistics Indonesia, presented STAT-INCOME-5 describing the Indonesian experience in generating farm income data. He highlighted the need for farmer income data since a big proportion of income at national and regional levels came from agriculture. The 2004 Farm Income Survey (SPP04), a part of the Agricultural Census of 2003, was the latest exercise undertaken by Indonesia.
18. Covering 1.42 percent (or 357 770 farms) of the total agricultural households, the survey used a two-stage probability proportional-to-size sample design with census blocks within the villages as the primary sampling units and the agricultural households as the secondary sampling units. Data collection through a face-to-face interview was completed in one month. According to Mr Achmad, farm household income was less than US$1 000 a year in Indonesia as compared to US$4 626 of income per household in the whole country. Non-farm income for the agricultural household was estimated at about 30.54 percent of the total income.
19. He cited limited budget, trained enumerators and other skilled personnel as the main constraints for the undertaking of farm income surveys in Indonesia. He noted that a country like Indonesia required around 9 000 enumerators for the Farm Income Survey and 200 000 enumerators for the Agricultural Census. He added that before the economic crisis in 1997, the survey was undertaken every three years but frequency became a problem thereafter. The above constraints were compounded by the large number of small farm households (over 25 million, averaging 0.3 ha) and the geographic barrier posed by farms located in remote islands of the Indonesian archipelago. In the latter, survey costs rose sharply and data quality suffered, leading to the replacement of samples.
20. The Experts agreed that data collection in remote areas was an issue in several Asian countries. The difficulty in accessing remote areas easily contributed to high survey costs, forcing a significant reduction in sample sizes and affecting therefore precision. One Expert suggested the used of localized sampling in such areas, reducing sample size and survey frequency.
Data System for Farm Income in the Philippines, from Collection to Use: Strengths and Weaknesses
21. In STAT-INCOME-6, Ms Maura S. Lizarondo, Assistant Director, Bureau of Agricultural Statistics (BAS), Department of Agriculture in the Philippines discussed how her bureau collected farmer income data. She identified two sources of data: the Costs and Returns Survey (CRS) and the Integrated Farm Household Survey (IFHS). The CRS (targeting specific producers of commodities) was planned to be conducted every five years for the benchmark with annual updating. However, budget and logistics problems restricted the planned frequency due to the large number of crops, fish species and animals commonly produced throughout the country. The IFHS, in turn, was intended for implementation every two years, but since 1988 there have been only three surveys conducted so far, the last one in 2003.
22. She pointed out that the IFHS covered all provinces (domain) and farm households were selected using a two-stage sampling design. It aimed to produce detailed information on the dynamics in which the farmers generated income. CRS intended to collect the financial structure of producing certain agricultural commodities and not total farm income. The samples were selected purposively to capture different segments of the producing population for certain crops. She added that the large volume of commodities, the highly volatile crop production cycle and the increasing cost of survey operations yielded many constraints in the implementation of the two surveys and in the collection of farm income data. The farm household income was found to be at around US$2 125 in 2002-2003, with offfarm and nonfarm income accounting for 36 percent of the total.
23. Ms Lizarondo said that a recent review on CRS and the IFHS had raised questions such as: What else can these statistical inquiries offer as statistics and indicators of farmers' welfare through time? What statistical data can be appropriately updated to indicate farmers' welfare? What other types of data presentation can be made out of the survey data? Is there a way to streamline the surveys to make them more affordable and frequent? She added that an in-depth analysis of the recent rounds of the CRS and the IFHS was necessary to learn some insights on how to improve the data systems, including the possibility of integrating them. Strategies were being identified in generating data for the various demands given the available data.
24. In the ensuing discussion, some Experts argued that while integration of surveys might potentially conserve the limited resources, data quality could suffer because of the possible response burden it could create on the respondents. The issue of whether to measure farmer income in the context of household welfare or a market-oriented entity was raised. It was clarified that the goal of the consultation was more directed towards the generation of income in a household welfare context.
Ideas and Suggestions from CABIG on Farmers' Income Data
25. In STAT-INCOME-7, Mr Jo Cadilhon, Marketing Officer, FAO RAP, Bangkok, put forward some ideas of the Commercialization and Agribusiness Interest Group (CABIG). He noted that the main interest of the group was the dynamics that happened beyond the farmgate and how these impacted on producers' management practices.
26. He suggested considering the collection of data on various employment sources among rural households to provide information on the extent of off-farm activities. He said that information on the different production outlets, different prices involved and marketing arrangements, and the nature of the buyers were also needed. In the surveys, he suggested to consider stratification by farming systems and cited the EU network of farm income statistics (see next presentation) as a useful model.
27. Although agreeing in principle on the usefulness of beyondfarm data, the Experts recognized that the generation of such information would be very difficult and costly in Asia. They were of the view that data collection systems in developing countries, unlike in Europe, were already heavily burdened with budget constraints. With regards to information related to the type of marketing contracts, the Experts indicated that it might be more appropriate for large commercial farmers than for the mostly subsistence farmers in Asia.
Farmer Income Data for Decision Making in the EU
28. Mr James Whitworth, Head of the International Statistical Cooperation of the Statistical Office of the European Communities (Eurostat), presented, in STAT-INCOME-8, the flow of farmer income data in the EU. Mr Whitworth said that Eurostat did not collect but compiled all the data provided annually by all member states. He noted that the special character of the European Commission on the right to propose legislation and monitor compliance with the law facilitated Eurostat work. Legislation on farmer income data generation compelled all member countries to submit data for the Eurostat to compile.
29. He explained that farmer income data came from three different sources: Farm Accountancy Data Network (FADN); Economic Accounts for Agriculture (EAA); and Income of the Agricultural Household Sector (IAHS). FADN used a uniform questionnaire that collected data on crop areas, livestock inventory, labour force, and other physical and structural information on the farm. In addition, economic and financial data were also collected. The EAA was intended to analyse the production process and primary income generated by it. The IAHS monitored year-on-year changes in total income of agricultural households at the aggregate level in the member states. It also monitored the changing composition of income.
30. Mr Whitworth informed the Experts that before the FADN, member states conducted surveys based on farm accounts, and as such, had already established their own sampling plans. The technical sophistication of such plans, however, varied among member states. He recognized that while the participation of farmers in account keeping was voluntary, the number of participants was gradually increasing.
31. In the subsequent discussion, Mr Whitworth clarified that the Eurostat defined the output (statistics) while member states decided how to collect the data. However, countries used a standardized format of reporting to facilitate compilation of Europe-wide data. The Experts noted the degree of sophistication and efficiency of the EU agricultural statistics. However, they also noted that, unlike in Asia and the Pacific, EU countries were more homogeneous, facilitating farm data collection and standardization. Furthermore, EU farmers might be also more willing to supply or kept accounting data since they benefited from EU subsidies. When asked about data quality from new EU member states, Mr Whitworth clarified that it was addressed during membership negotiations and by appropriate training.
32. Farmers' income data can be generated from available surveys and other data sources. In this agenda sub-item, two papers from Australia and India discussed how information from surveys can be integrated to generate farmers' income data.
Monitoring Farm Financial Performance through Surveys
33. In STAT-INCOME-9, Mr Vince O'Donnell, Manager of Commodity Outlook, Australian Bureau of Agricultural and Resource Economics (ABARE), Department of Agriculture, Fisheries and Forestry, presented the surveys that Australia used in monitoring the financial performance of farms. He said that ABARE and the Australian Bureau of Statistics (ABS) collaborated in the collection of agricultural statistics, with the industry providing up to 50 percent of the funding for some data collection activities undertaken by the former. He informed the Experts that surveys were conducted every year for the cropping, beef, sheep and dairy industries and less frequently for other industries. He said that ABS was the principal organization responsible for statistics in Australia. In addition to the agricultural census conducted every five years after the population census, ABS also conducted an annual commodity survey. Both these ABS surveys were focused primarily on production.
34. Mr O'Donnell explained that ABARE's surveys collected an integrated schedule of financial, physical and socio-economic variables. Most ABARE surveys target commercial farms with agricultural operations of more than AU$40 000. Commercial farms account for the largest proportion of Australia's farm output. The main purpose of the ABARE farm surveys was primarily as input into policy analysis by government and industry to support decision making. The resulting research database also supports economic understanding of the rural sector and assists in measuring productivity.
35. He said that a list of the entire population of farms served as a base for sampling the population. Surveys were either regular (broadacre industries, dairy industry) or occasional (forestry, winegrowers, fisheries, and other industries). The Australian Taxation Office's business register, through the ABS, provided the frame for agricultural survey (formerly from the census). The frame is matched to the agricultural census which includes identifiers, industry classification, indicator of size, and geographic classification. He explained that agriculturally, Australia was divided in three broad zones: pastoral (5 000 `commercial' farms), wheat-sheep (54 000 `commercial' farms) and highrainfall (57 000 `commercial' farms) zones.
36. The sampling plan is developed with the aim of estimating means, changes and distributions at various levels. The rotation scheme of dropping around 25 percent of the sample and maintaining the other 75 percent for the next round produces a panel data and allows time series analysis to be done. Stratification initially involves a three-way classification: state, ABARE region and industry. Farm size is also included in the classification. Nonresponse has not been a major issue but there are at least two reserve selections made for each primary selection to ensure the sample remains representative.
37. Mr O'Donnell noted that sample weights were generated for each sample farm and constrained to sum to population totals of key variables (supplied by ABS). Key variables included the population count, number of livestock and areas sown to key crops.
38. He informed that new developments in farm income surveys included the use of geospatial data, which linked geographical and other scientific data with financial performance. He noted that online databases (Agsurf-programme) were also available and could be used to view estimates (average per farm) for variables collected in the survey. He added that ABARE survey data could also provide support in the analysis of the relationship between productivity growth and environmental protection, climate change, water allocations, and access to new technologies (e.g., GMO).
39. Responding to the Experts, Mr O'Donnell clarified that the survey data was collected through a mixture of face-to-face and telephone (relatively simple and straightforward data) interviews. The decision of which method to use depended on the complexity of the data to be collected. For many interviews data were entered directly onto computers. He noted that dataconsistency checking started from the field. Information from the farm was crosschecked with those from other sources, e.g., accounts and marketing outlets. Following that, data were subject to intensive electronic probing. Nonresponse rate was noted to be low particularly for farms being interviewed in subsequent years. He said that individual farmers could get online access to survey results and compare their data to the average within their regions.
40. The Experts noted that support for data collection in Australia came not only from the public sector but also from the contribution of the private sector who found utility in the data. They believed that this could be an approach to the budgetary constraints in developing countries, which affected the regular collection of farmers' income data. They said that a more intensive advocacy campaign was yet to be done to encourage potential beneficiaries of the data to contribute in funding to data collection.
41. The Experts praised Australia's approach to provide information back to the farmers. They noted that the inability of some institutions to put farmer income data into the mainstream of data collection could be contributing to the difficulty in collecting farmer income data. Weak data could be explained by the lack of awareness among the farmers on their potential benefits from using the data. The Experts agreed that the challenge was on the identification of appropriate venues to disseminate the information to the different users including the data providers themselves.
42. The Experts noted the growing use of geospatial tools and satellite images in the collection and cross-checking of farm data. They also noted that these tools were becoming less expensive and its applications improving. The experts agreed that it would be worthy to explore the utilization of satellite technology in the improvement of collection and dissemination of farmer income data.
Developing Appropriate Survey Methodologies to Obtain Reliable Income Data of Farmers: Challenges and Plausible Ways and Means
43. In STAT-INCOME-10, Mr Gurucharan Manna, Deputy Director General, National Sample Survey Organisation, Survey Design and Research Division of the Ministry of Statistics and Programme Implementation of the Government of India, presented the challenges and plausible solutions in the generation of farmer income data. He said that income data was a valuable input in the understanding of farmers' conditions, but difficult to collect and often under-reported. He suggested the use of consumer expenditure or the integration of consumption and savings data as proxies for farmer income.
44. To illustrate his point, Mr Manna described a pilot survey conducted in 1983-1984 that adopted a stratified two-stage design with villages as the primary sampling unit (PSU) in rural areas and urban blocks as the PSU in urban areas. The households served as the ultimate sampling units. In each PSU, the sample households were equally divided into three groups: Set 1 was enumerated with income data only; Set 2 with consumption and savings data; and Set 3 with income, consumption and savings data. When compared household income with consumption plus savings, averages were found to be similar in urban areas, but very dissimilar in rural areas. The average farm household income was 30 percent lower (Set 3) or more (Sets 1 and 2) than the average of consumption plus savings, suggesting under-reporting of income.
45. Mr Manna explained that a Situation Assessment Survey (SAS) was later conducted in 2003, covering all Indian farm households. The survey collected data on land possession, assets, access to modern technology and income, among others. A stratified two-stage design was used and data was collected in two visits to reduce the problem of memory recall. The survey covered 51 770 households from 6 638 villages. A farmer was defined as one who possessed some land and was engaged in agricultural activity on any part of the land in the last 365 days. Average household monthly income (2 115 rupees) was found to be about 76 percent of average household monthly consumption expenditure. The extent of divergence between the two estimates varied across states, with 13 out of 18 major states reporting income lower than consumption. Although other non-farm income such as remittances was not included, it was deemed as negligible source of discrepancies.
46. To address under-reporting in income, Mr Manna put forward several suggestions, including use of a sampling frame mixing a list frame (LF) and an area frame (AF), with LF ideally for large farms; use of appropriate stratification before sampling of households/farms; organization of the questionnaire into manageable blocks; collection of data in successive visits to minimize memory bias; estimation of a correction factor for income based on data on income, consumption and savings collected from a from a sub-sample of households; and creation of public awareness among the respondents about the utility of income data.
47. The Experts noted the problems associated with under-reported income and agreed on the need for a study in the subject. It was further noted that one possible reason for income to be lower than consumption was that some components of income might not be thoroughly accounted for. The Expert pointed out that in household income measurement, underreporting was usually encountered because of the difficulty in collecting information on non-cash income and expenditures. Some Experts noted that the framework in the Living Standards Measurement Study (LSMS) could be considered since it provided a systematic procedure of imputing non-cash income and expenditures.
