In most developing countries, there are a number of ways in which existing flows of statistical information can be made more useful for policy analysis and other kinds of research. In many cases, improving the management of existing information should be given priority over collecting new data, or launching new surveys to improve data collection. Usually improved management is less expensive than new surveys, and it can enhance both the quality and the timeliness of the statistical information.
Unfortunately, there are many instances in developing countries in which systems for agricultural data collection and management have deteriorated in the 1980s, as compared to the situation in the 1970s. Cross-sectional surveys are conducted less frequently, statistical publications are slower in coming out or have been dropped, some data series have been dropped, and indirect indications are that the quality of the data is poorer in some cases.
This situation appears to have resulted mostly from the general deterioration in economic conditions in the developing world in this decade, one consequence of which has been a significant decline in real salary levels and other resources in many government statistical agencies. While of course there are exceptions to this statement, it is true in enough countries that it would appear to merit systematic attention by international agencies. A continuing decline in the quality of statistical information eventually will lead to a decline in the quality of policy analysis and advice. These trends come at a particularly unfortunate time, as the structural adjustment and sectoral adjustment programs that are increasingly common are quite intensive in their requirements for information and analysis.
Improvements in data management systems can be grouped into the following major areas: 1) management of the data collection process, 2) organization and interpretation of the data, and 3) management of the publications process.
Sometimes completely new procedures for collection of information at the field level are implemented without giving due consideration to the possibilities of working within the existing system and improving its functioning. A leap to a new technology of data collection is made rather than attempting to improve the old one. Examples are the area frame sampling surveys of agricultural production that were found to be very difficult to implement in some countries (Honduras and Peru, for example). As a result of the attempt to move completely to the new system in Honduras, before the operational problems were solved, the traditional system for estimating production levels was weakened substantially, and users of data were left in worse condition than if the new system had not been proposed.
While new sample frames can contribute to improving the quality of the information, improvements also can be obtained with the existing frames by upgrading the performance of the survey team. In this regard, three steps are recommended:
Developing a practical manual for field estimators, specific to that country, covering topics such as how to estimate the implicit price of crops like sugarcane that sometimes are harvested as part of a vertically integrated operation for production and milling; what form (state) of each crop should be used for the price quotations; and the recommended practices for obtaining production estimates at the farm level.
Holding more extensive training courses for field estimators, with refresher courses at intervals.
Developing procedures for cross-checking, on a spot basis, the field estimates and monitoring the performance of field estimators. Casley and Lury (1982, pp. 110–111) present suggestions on this question.
As simple as these procedures are, it is surprising in how many countries they are absent in agricultural statistical work. Thus the first item on the agenda for upgrading the quality of agricultural statistics would be to ascertain if these procedures are in effect and to what extent they need to be implemented or strengthened.
Another aspect of managing the collection of data is ensuring that the coverage is appropriate. Production or price estimates that cover only a handful of major crops are not likely to provide guidance on the main trends in the sector, as discussed in section 2.1 of this report.
Organization of data starts with questionnaire formats. It must be borne in mind that 1) questions that are meaningful to the interviewer may not be to the respondent, and 2) there is a limitation on the respondents' time or willingness to dedicate time to the survey (again see Casley and Lury [1982, p. 89]). Equally, in order to facilitate the processing of the data, the questionnaires should be tied into the desired formats of the tables of results. In reality, the first step in designing a data collection effort is designing blank tables to be filled in through the survey.
In the case of surveys that already are underway or designed, a step that is sometimes omitted is proofreading the raw listings of the data that are generated by the computers after the data have been transferred from questionnaires to computers. There is a tendency to read only the tabulated results from the computers. The author encountered an experience in northern Nigeria in which ex post comparison of the questionnaires with the computer listing of the questionnaire information revealed that in almost fifty percent of the cases there had been a translation error in going from questionnaire to computer input! Often it was an error in transforming the units of measure of the variable, sometimes it was an incorrect variable identification in the computer coding (such as the type of fertilizer), and often it was a simple typographical error.
After the data are appropriately organize and checked at this level, the question arises as to the kinds of aggregations and transformations to present. Earlier chapters of this report have suggested the value of indexes of farmgate prices gross output (production) for agriculture and livestock, in total and by product groupings. It would be a straightforward matter to generate those indexes from the statistical agencies computerized data sets by product, but frequently it is left to researchers to do so.
It would be even more valuable from a policy viewpoint if the publications of production data were accompanied by indexes such as the overall economic productivity of agricultural land (section 2.1.4), and if the publications of farmgate price data were accompanied by real price indexes (section 2.2).
The same agencies could develop time series of food balance sheets and the associated calculations of the availability of calories and proteins (section 4.1).
When estimates of nutrient availability are developed from household surveys, significant differences in the estimates can result from increasing the number of products included in the computations. Garcia et al. (1988, pp. 48–49) found that expanding the number of products from 23 to 186 made a considerable difference in the estimates of average calorie availability per person per day. For some strata, the difference was as much as 300 calories per day.
In cases of standardized official data, such as fiscal expenditures, the value of the information is greatly enhanced if organized and presented according to functional classifications rather than simply by agency and department, as discussed in section 3.3.
A different aspect of data organization is the way in which different statistical agencies coordinate their work. It is unfortunately all too common to find different “official” time series for the same variable, series which can diverge by large margins. (The only thing more frustrating to an empirical researcher than data of low reliability is a set of different estimates for the same concept!) In cases like this, the most appropriate remedy would appear to be a small-scale but high-level statistical oversight commission that is empowered to change the procedures of statistical agencies. Such a commission is being established in the Dominican Republic as of this writing, precisely because of these concerns.
From a viewpoint of using statistical information for policy-related studies, one of the main problems is lack of timeliness in the publication of the data. If data on production and farmgate prices are not made available until two years after the harvest, they will have lost most of their policy usefulness. Some cases are more extreme. As of early 1988, no production data for Peruvian agriculture had been released officially for any year after 1979, and researchers who attempted to use unofficial tabulations were informed the data were preliminary and subject to possibly significant revisions. Policy analysis with production data was virtually impossible.
These problems can be solved with an appropriate technology of data management, centered around the use of micro computers and desktop publishing procedures. The time lag in transmission of this new technology to statistical offices in the developing world has been very long, and only in the last year or two have micro computers been introduced to those offices in many countries, generally in insufficient quantities; many others still do not have them. Again, with the fiscal constraints that are prevalent in this decade, this would appear to be an area in which international agencies can make useful contributions. The cost of making large improvements is not great; technology, organization, and training are the key elements.
Delays in publishing official afflict other sectors as well as agriculture. Trade data are notorious for their lags in issuance, even though they are quite important for many policy issues. In this field as well, appropriate data management can speed up the publications process.
The publication of data can take various forms: yearbooks, monthly or quarterly bulletins, special statistical studies. The bulletins are especially appropriate for timely information, such as monthly price data. Yearbooks could be expanded in coverage. Instead of just publishing annual summaries of the monthly data they could, as noted previously, include indexes of production, prices and productivity. They could also include special tabulations such as disaggregations of agricultural trade data, food balance sheets and the availability of nutrients. One of the reasons these things have not been done more often is that the data management and publications process has not generally been guided from a users' viewpoint. And to the extent that it has, production economics rather than food policy has been the guiding influence.
Some countries, such as the Republic of Korea, have a long tradition of publishing high-quality yearbooks of agricultural statistics. The Korean yearbook includes data on the spread of' hybrid seeds, the terms of trade for farmers, input prices and other concepts not usually found in yearbooks.
Apart from timeliness and coverage, another aspect of publications that warrants attention is the form of the publication. When data are processed in micro computers, it now is feasible to make available (at cost) diskettes with complete files, such as the trade data files, or the production data by product and region and year. Of course, the files should be mounted on a standard database or spreadsheet program. That kind of distribution of information would enhance the role of the data in policy analysis and research, and it would enable the analysts to starting working more immediately with the data, thus contributing to more timely analysis.
In sum, there are many ways to improve the management of agricultural data with a view to increasing their usefulness. What is most needed is a focus on these issues by national and international agencies; the technologies and approaches are known, and the resource requirements are not great.
Priorities and preferences for types of data depend on many factors: the policy issues of the moment, the training and background of individual researchers, the cost of collecting the data, and even the nature of applied research programs established in international agencies. Therefore no attempt will be made to establish an objective or general list of priorities, but rather a few suggestions and observations will be made. The suggestions are admittedly based on the author's own experiences, but perhaps they will be relevant to some of the circumstances encountered by other policy analysts and data managers.
In spite of the widening scope of policy analysis in food and agriculture in recent years, one of the most basic tools of analysis remains the farm budget. World-wide, in many country studies each year, including those of international organizations, a considerable amount of researchers' time is spent in gathering scattered farm budget studies. Hence of the the priorities for data collection would be a systematic program of collecting information on farm budgets, with variations by technology and region. Equally important would be a systematic annual publication series of that information. Sometimes the information is gathered on a rather regular basis, but there is no formal publication of it, and finding it is a research task in itself. Incidentally, this is an area in which sample surveys are not needed, but instead well-trained or experienced experts need to visit farms that are more or less representative in each region.
As indicated throughout the text, the production and price series also are basic. If they are non-existent or the coverage is insufficient, priority should be given to correcting those shortcomings. For prices, policy questions require information at three levels: the farmgate, the wholesale level and the retail (consumer) level. The price series can be captured through timely information surveys, whereas the production series can be annual or seasonal (by cropping season).
There is a special issue regarding the collection of production information for root crops like cassava and cocoyams, which are widely used as staple foods in West Africa. The harvest of these crops can occur at almost any point during the year. The traditional survey methodology of sending out the enumerators at one time (the presumed harvest period) during the year or the season is insufficient to gather reliable information. As a consequence, the official data on these crops are notoriously unreliable. It would appear that the only way to obtain reasonable information on the harvest of these crops is through a more frequent (say, monthly) survey of households, in which consumption patterns and the sources of food acquisition are explored. Given the importance of these crops in some countries, this issue merits more attention than it has received to date.
Another pressing need for policy analysis is information on distributional variables that is more up-to-date and gathered at more frequent intervals. The main economic variables of concern here are production patterns and yields by farm size class, consumption patterns by farm size and by income stratum, and sources of household income. There are different ways of addressing this need: more frequent agricultural censuses, more frequent production surveys which also gather information on other variables, and more frequent house-hold consumption and income surveys.
Another possibility is an integrated survey of farm households that develops information on production, income and consumption, complemented by urban income and consumption surveys. But an integrated survey runs the risk of demanding too much time of the respondents. At present, a feasible approach would be to carry out two kinds of surveys at three-year intervals (if not more frequently): farm production. and income surveys, including information on the farm size and the family labor force; and household income and consumption surveys. (An alternative approach is to track the same farms and households year after year.)
The sample size need not be large--some defining characteristics of the farms and households can be ignored as long as the basic ones of location, farm size, family size, and family income are captured. A program of this kind would do more to improve the informational base for food and agricultural policy analysis than almost any other measure.
Finally, the importance of timely information surveys should be restated. The main variable of interest here is prices. Input prices are of interest as well and, as noted earlier, acreages planted by cropping season and crop also help very much in making short-term forecasts of production and therefore of import requirements.
