Policy analysis for food and agriculture is an art that is practiced in almost as many different ways as there are practitioners. In its simplest forms, it goes back to biblical times and to corresponding eras in ancient China and India. It is a discipline required in one way or another by all governments. *
In the industrialized countries in recent decades, agricultural policy analysis has come to be associated primarily with reviewing protection and subsidy policies, and with the continuing search for new tradeoffs and additional room for maneuver in the conflict of interests between consumers and producers and between domestic and foreign producers. Agricultural research policy also commands increasing interest.
In developing countries, the issues often are more basic: how to generate growth in the incomes of poor farmers, how to ease the burden of food costs for consumers, how to improve the nutritional levels in the lower income groups, how to generate more foreign exchange via agricultural exports or import substitution. Necessarily, the search for acceptable solutions leads into a broad class of questions, such as the functioning of input and factor markets, land tenure, the performance of public institutions, investment and credit allocations, research and extension policy, and trade, pricing and marketing policy.
The diversity and complexity of the issues has affected the nature of analyses of the agricultural sector. On the one hand, it has created a demand for qualitative assessments and summary overviews of the sector, to help keep policy makers informed without delving into the details and complexities of the underlying analyses. On the other hand, it has created a demand for more quantitatively sophisticated tools of analysis, to reflect the heterogeneity of the sector and more of its interrelationships, and thus to advance our understanding of how the sector may respond to policy initiatives.
In the pursuit of these two divergent approaches, sometimes intermediate approaches are overlooked. In particular, little emphasis tends to be placed on the construction and interpretation of simple economic indicators and coefficients for the sector. Frequently, some of the most basic economic indexes, such as production indexes and inter-sectoral terms of trade, are not calculated', and their possibilities for extension and interpretation are not exploited.
The purpose of this paper is to describe systematically a set of simple uses of the more commonly available data series and surveys. They are simple calculations and interpretations that help develop an understanding of the current structure and historical evolution of an agricultural sector and a nation's food supplies, and they often have considerable relevance to policy concerns. Most of them are fairly well known procedures, but some of their extensions and interpretations do not appear to be so well known. And, surprisingly, for most countries very few of these procedures have been carried out, or at least they are not systematically developed and kept up to date.
Examples of the calculations are coefficients of economic protection rates (for more than one year), comparative advantage indicators, indicators of trends in nutrient availability over time, sources-of-growth calculations, and indexes of the purchasing power of farm households.
These and the other calculations described are in the nature of descriptive statistics. They are intended to be complements, not substitutes, for more sophisticated quantitative analyses. Often it is useful to carry out some of the simpler calculations first, to help identify the more pressing issues, and then to develop more complex models of those issues. And in many cases, the simpler calculations can be updated at regular intervals, to assist in the monitoring process regarding sector performance.
This paper is organized by type of data series, with corresponding discussions of how each series can be used for sector-wide description and analysis, always keeping with the spirit of simplicity and emphasizing economic interpretations. Examples are drawn from recent studies in several countries.
The data series range from time series on production, prices, yields and foreign trade to censuses, household surveys, fiscal data, and data on agricultural inputs and costs.
The paper is intended to be a guide to practitioners in the field, working in agricultural development and food issues. It would not be appropriate to attempt to include all the analyses suggested, or even the majority of them, in a single country study. Rather, the analyst would be expected to choose which analyses were relevant to current policy concerns and were consistent with the informational availability. Quantitative analysts may not find much material here, if any, that is new in a conceptual sense, but it is hoped that they will find some useful techniques and a reminder of the value of doing basic analytics.
Experience in a number of countries has shown that most of the data reviewed here are available in almost all developing countries. Often the most urgent priority is not to collect new kinds of data, but instead to make fuller use of existing data. Nevertheless, the discussion in the paper does lead to some conclusions about priorities for systems of data collection, and those conclusions are summarized in the last section.
All of the calculations proposed can be carried out on microcomputers, with database or. spreadsheet programs. Many also can be developed with a hand calculator and pencil-and-paper spreadsheet (as the author has verified on many occasions]). In cases of censuses and extensive surveys, it is assumed the analyst will be working with published or tabulated results, and not with the raw survey data themselves.
The remainder of the paper is organized as follows. The section following this one lists and describes the principal kinds of economic data used in food and agricultural analysis, and it raises a few concerns regarding the coverage and quality of those data and the formats in which they are most applicable to policy concerns. Section 1.3 then places the statistical procedures in a policy context. Major areas of policy concerns, including distributional and nutritional issues, are listed and linked to the corresponding kinds of procedures with the data.
This discussion helps motivate the subsequent review of data and statistical procedures, but it is kept at a somewhat general, introductory level. After the procedures are presented fully, one of the paper's concluding chapter (6) discusses priorities regarding data collection and management that emerge from the considerations in the other chapters. Emphasis is placed on improved management of existing data flows as well as new efforts at collection of data.
Chapter 2, comprising three sections, is devoted to analyses based on the principal time series. Chapter 3 contains sections dealing with surveys and censuses and other data of a structural or cross-sectional nature, and chapter 4 then reviews analyses that require combining data from different sources (food balance sheets, measures of economic protection). In that chapter, nominal and effective protection coefficients and domestic resource cost coefficients are reviewed in some detail. Chapter 5 discusses the topic of timely information series. Then chapter 6, as noted, turns to the theme of data collection and management
To a plant researcher, tabulations of field trials with different varieties and different input packages may be the most important kind of agricultural data. A. nutritionist may review dietary patterns for a sample of households. A macroeconomist may view the sector through the window of an input-output table, which details agriculture's transactions with the rest of the economy, usually at a fairly aggregative level in terms of product groups. Because their characteristics and levels of aggregation are so different, these different data sets are not readily related to each other. In a sense, this example illustrates the typical circumstance in which different kinds of agricultural specialists and macroeconomic specialists find it difficult to talk to each other meaningfully about policy.
A policy dialogue between the groups, particularly between sector and macro specialists, is facilitated by a data base which is sector-wide in scope. The selection of data series for this paper has been guided primarily by this consideration. The data must contribute in some way to sector-wide analysis, and preferably to the kind of analysis which also will be meaningful when reviewed at the economy-wide level. The agricultural sector is particularly sensitive to macroeconomic policies, and many food and agricultural policies are inextricably linked with macro policies.
At the same time, for strictly sectoral policy, it is important to understand trends at the product level and also by types of producers and consumers. The classifications of producers that are most relevant vary from country to country. In some cases, they are regional (Peru, Nigeria, for example), in some cases they concern the irrigated-nonirrigated distinction (Mexico), in some cases they start with the plantation-smallholder distinction (Indonesia), and frequently they concern farm size distinctions. The data series discussed below contribute in various ways to illuminating behavior according to these categories, as well as contributing to the linkage with data and analysis at the macro level.
Almost all data in the agricultural sector refer to products (including inputs), factors (principally land, labor, reproducible capital and irrigation), and types of households or farms. Some kinds of data refer to two or more concepts. For example, a farm budget (structure of costs of production) usually is defined with respect to both a product and a type of farm. Investments escape a little from this classification scheme. While they refer to types of reproducible capital (marketing and storage facilities, irrigation infrastructure, etc.), they often also have a spatial identification but not necessarily a product identification. An irrigation project serves a particular region but may be used to raise many crops.
Fiscal data, referring to government revenue collections and expenditures in the sector, also have their own frame of reference. They are reviewed in section 3.3 below.
The principal time series normally are product-based: production, yields, prices (at different stages in the marketing chain), exports and imports. In these cases, indexes and aggregate data often are not published, and the only way for the analyst to arrive at aggregate figures (such as the sectoral balance of foreign trade) is to add up data on individual products, weighting when appropriate (as for price indexes).
Those data that document factor availability and household characteristics, including food consumption and nutrition, typically come in the form of cross sections: surveys and censuses. The main examples are household income and expenditure surveys, agricultural censuses and population censuses. Published compilations made from surveys for purposes of reviewing consumption patterns usually group households according to income strata, but they also may include distinctions by sector of employment or occupation of the household members, as well as by location (rural or urban, town or large city) and other criteria.
However, it is important to note that such compilations are made at long and irregular intervals in most developing countries, and in some cases they exist only for one year or even not at all. Even in the case of agricultural censuses, often the period of time between censuses is very long, frequently in excess of ten years and sometimes in excess of fifteen years.
Thus, while the readily available data series may relate the annual levels of production and exports of, say, corn and bananas, if a policy maker wishes to know what has happened over time to net incomes or nutrition levels on, say, nonirrigated farms of less than five hectares in District X of the North Coast, then indirect estimates have to be made. This situation is unfortunate and suggests the possible usefulness of repeated annual surveys, on a small scale, of selected types of farms and households. This topic is taken up again later in the paper.
In terms of procedures for measurement, a time series on net agricultural income (= value added = sectoral GDP) is a derivative concept. Viewed from the product side, it is computed as the difference between the value of all agricultural outputs and the cost of inputs (valued at their appropriate prices). But a cautionary note is in order here. When the national income accounts are compiled, income is imputed by reference to gross output levels, and the measurements of the ratio of inputs to outputs usually are updated only infrequently. Also, outside of a few major products in the sector, the income computations do not take into account the effects on the sector-wide aggregate input-output ratios of changes over time in the composition of the sector's output.
The income accounts are invariably compiled by central statistical agencies. Agricultural sector institutions do not independently estimate time series on agricultural income.
From the factor side, net income is, conceptually, the sum of factor earnings, summed over all factors owned or controlled by the household. But this latter definition is not very operational, for it is not possible to observe directly the factor income (or rents) attributable separately to the farm's land, irrigation supplies, reproducible capital and family labor.
For these reasons, this paper underscores the use of production (gross output) series more than income series, although the latter are more descriptive of the economic wellbeing of producers and should be reviewed as well. The role of both kinds of series is developed in chapter 2.
In the usual treatments of agricultural production, and other concepts such as agricultural exports, there is a long tradition in agriculture of focussing on a few key products, usually grains and “traditional” exports, and an associated tendency to base conclusions about the sector's performance over time on these selected “major” products. This paper presents evidence showing that using a few major products can be quite unrepresentative of the sector as a whole. Therefore it can conceal the main sources of dynamism, and the main problem areas, in the sector.
The paper suggests moving toward analysis which uses more comprehensive aggregates and indexes of sector behavior, so that statements can be made about the sector as a whole.
A major reason why a time series perspective is needed in evaluating the current agricultural situation in a country is that it is a sector with a history of ups and downs. Agricultural production and prices characteristically fluctuate quite a bit over time, so an accurate reading of the “current” situation may be obtained only by looking at, say, the record of the last five years. For some considerations, recourse to time series is unavoidable.
Many of the time series concepts are best presented in real terms, that is, corrected for the economy-wide rate of inflation. Questions associated with the choice of deflator are discussed below. Beyond. those questions, some cross-sectoral deflation possibilities are mentioned,' and cases are noted in which it is preferable to deal with data in current prices, or in both current and constant prices. For example, when agriculture's share of national export earnings is computed, it is better to do it in current prices, to reflect the current situation of world market prices and also to avoid clouding the computation with index-number problems.
If there is a question about the contribution of shifts in the external terms of trade to agriculture's export share, then the share calculation can be done with international prices, both in current prices and in base-year prices.
Some of the data and associated coefficients, such as the DRC (domestic resource cost coefficient) refer to a single year, but in these cases the paper points out that the coefficients are quite volatile over time. Hence it is wise, before making policy-related interpretations, to compute multi-year averages of the coefficients. In other cases, such as regional-average and national-average cropping patterns, farmers' behavior tends to be more stable. Therefore survey and census data on those subjects may be used for quite a few years after the date of the survey.
In closing this section, a word is in order about the quality of the available data. That topic frequently is a major preoccupation in developing agriculture. Normally, the cross-sectional survey data are considered to be more reliable than the time series data. The latter often are based on “expert opinion” in each locality, rather than on objective sampling. In some cases, the principal time series are so unreliable that there is a consensus that they cannot be used at all (for example, when different government agencies have reported “official” production series that differ by as much as a factor of two for the same crop, as for rice in Honduras and cassava in Nigeria).
In those cases, the analyst may wish to construct inferences about production growth, through information on imports and estimates of probable growth in food demand. Or it may be productive to assist in developing better measurement procedures. Or it may be best to base virtually all analyses on cross-sectional data, thereby omitting some topics.
But in most cases the time series are in fact used. For better or worse, they are the basis for the official opinions on sector performance. In these cases, it is important to contribute to better use of those series. At the same time, efforts may be made to improve their quality. Sometimes it turns out that intensive analysis of the time series raises enough questions about them, or puts them sufficiently in the spotlight, that efforts to improve data collection are given new impetus.
Whatever the concerns over the data, it is vital to use them in a systematic, logical way. Problems with data need not be compounded by slipshod analysis!
The purpose of this section is to sketch some typical areas of concern to policy makers and to indicate the associated parameters, or simple data analyses, that would help address the concerns or at least clarify the issues. References to the rest of the report are provided to index the parameters and analyses to particular sections of it.
Although a policy maker may possess considerable intuitive insight into his sector, frequently there is a need to develop a systematic overview, in numbers, of the performance of the sector. And when an international agency or other outside agency considers the possibility of formulating programs in the sector, the overview is even more likely to be needed.
The overview has a historical dimension and a cross-sectional dimension. The numerical history is largely based on production and price indexes (sections 2.1 and 2.2), and derivatives thereof, plus information on trends in agricultural exports and imports (section 2.3).
On the production side, the information required usually concerns how fast production is growing; whether it is growing fast enough, in senses defined below; what the sources of growth (or lack of it) are; and what is the distribution of growth over products and regions. In addition to these kinds of measures, this paper suggests that it can be useful to compute a measure of the growth over time in the total economic productivity of agricultural land. It also suggests the importance of making the production indexes and related indexes as broadly based as possible, in terms of the number of products covered.
On the price side, the pertinent concept is relative prices. A price level, or even its rate of change, is not very informative by itself. A policy maker will want to know how agricultural incomes have changed in terms of their real purchasing power over the range of goods and services in the economy. Similarly, the relevant measure of producers' incentives is the movement in output prices relative to input prices. The paper discusses differ deflators for the purpose of developing measures of “real prices,” or relative prices.
Regarding foreign trade, the main issues typically are: i) are agricultural exports growing at a satisfactory rate, and which export products are growing fastest (and slowest)? ii) are agricultural imports, especially food imports, growing faster than domestic production, or faster than agricultural exports? and iii) how important are agricultural exports and imports in the overall balance of payments of the economy? These are straightforward questions, but answering them may require a fair amount of data manipulation, as discussed in section 2.3.
At a general level, agricultural policy can be visualized as concerned with efficiency and distributional concerns, where the latter particularly include food and nutrition questions, subject to the constraints of resources, technology and institutions in the sector, and subject to the overriding macroeconomic constraints of the government budget and the balance of payments.
Distributional issues sometimes concern the distribution of land holdings but more often the effects of proposed policy changes on the incomes and consumption levels of different groups of farmers (section 3.1). From a viewpoint of food policy, they concern the effects of price changes and other policy changes on the real cost of food to different groups of consumers, urban as well as rural (usually stratified by incomes); see section 3.2. These two kinds of distributional analyses permit an evaluation of the main beneficiaries of proposed changes in agricultural and food policy, and also an identification of the groups in society that will bear the main costs of the policy.
There is a dynamic aspect to distributional questions also. For example, some studies have attempted to determine whether the nutritional status of the population is improving or deteriorating over time (section 4.1). This involves computing trends in aggregate food supplies, and a necessary step is the reconciliation of import and production data (section 2.3).
A policy maker reviewing questions of food and nutrition policy will want cross-sectional information on food consumption and nutrition (section 3.2), to be able to identify target groups as well as to be able to quantify the effects on the welfare those groups of potential changes in policy. In addition, information on the structure of fiscal subsidies in the sector will be required (section 3.3). It is not always immediately evident how much of the subsidies allocated even to state marketing boards goes to specified staple foods., And the magnitude of those subsidies will give an indication of the fiscal flexibility in the sector, in light of the macro budget constraint.
The analytic side of food policy studies sometimes leads into assessments of the effects of food aid on incentives for domestic producers. That analysis is greatly facilitated by having the time series indexes of domestic production and prices that were mentioned above. And it also leads to questions of trends in household purchasing power; here the cross-sectoral deflations discussed in section 2.2 can be useful.
When policy makers review efficiency questions, they tend to think in terms of productivity at the crop level, that is, in terms of new input packages. But an equally important aspect concerns the allocation of land, labor and other resources over crops (and livestock products). That allocation is influenced by relative protection rates over crops. Section 4.2 discusses measures of economic protection and comparative advantage, and how they are applied.
With regard to the macroeconomic constraints, a challenge for sector policy makers, one that arises with increasing frequency in the context of structural adjustment programs, is how to raise the equivalent amount of fiscal revenues, or more, while reducing some of the distortive effects of existing taxes (section 3.3). As a rule, commodity taxes are highly distortive of economic incentives, but they are the ones most used in agriculture.
When fiscal changes are proposed in agriculture, the broad concern is how they will affect the triad of farmer incentives, the efficiency of resource allocation in the sector, and the cost of food to groups of consumers. More generally on the fiscal side, policy makers need adequate classifications of expenditures and revenues, classifications that coincide with the categories of programmatic decisions (also section 3.3).
Finally, it may be noted that for some policy purposes the timeliness of information is critical. In many instances, there is more scope for well-focused timely information surveys (chapter 5).
Other kinds of quantitative insights are useful for sharpening policy makers' perceptions of the status and options in the sector, and they are developed at various points in the text of the report. The final section of the report deals with some issues in setting priorities for data collection, in the context of the rest of the report. Policy makers do not often become involved in determining those priorities; perhaps this report can close somewhat the gap between decision makers and suppliers of data.
In beginning an economic overview of an agricultural sector, the selection of techniques of analysis will depend very much on the expressed concerns of policy makers. However, in most cases probably the three most basic tools for assessing the economic state of agriculture are a production index, an index of real farmgate prices (both in series at least ten years long), and a set of nominal protection coefficients, the latter calculated for each of the last three or four years. When food and nutrition policy is a dominant concern, the descriptive analysis should begin with time series on apparent consumption (section 4.1) and cross-sectional estimates of nutrition by household group (section 3.2). Each economic analyst tends to develop preferences for particular parameters or ways of presenting data, but most find that these three pieces of information are indispensable for understanding the sector from a policy perspective.
