3. CROSS-SECTIONAL AND STRUCTURAL DATA

3.1. Agricultural Censuses and the Distributional Effects of Policies

Agricultural censuses are practically the only source of information on levels of capital stock in the sector, and consequently they are essential for exercises in estimation of production functions on a sector-wide basis. Likewise, they are the only source for data on land tenure and farm size distributions; those distributions are perhaps the most frequently cited data from censuses. They have other important uses as well, such as serving to estimate the incidence of pricing policy changes on farm household incomes, and providing the information for measuring the efficiency of agricultural land use by farm size class.

Unfortunately, they tend to be a rather underutilized source of data, in part because their publication normally is quite delayed (for a census study to be issued five to seven years after the date of the questionnaires is not at all unusual), and because there is such a long time lapse between censuses. If an analyst in Peru wished to use census data, as of this writing, he or she would be using data seventeen years out of date!

If the author of this paper had a wish list for additional agricultural data in developing countries, high on that list would be censuses, or comprehensive surveys covering the same kinds of variables, at five year intervals—with rapid post-survey processing.

3.1.1. Measuring the Distributional Effects of Pricing Policies

One of the more valuable aspects of a census (or comparable survey) is that it links cropping patterns with characteristics such as farm size, region, and the presence of irrigation. These data can be utilized directly to build a “policy effects matrix” for pricing policies, that is, a set of calculations that show, at a first order of approximation, the distributional effects of a change in any product price in the sector. The distribution can be defined in terms of farm size, region, or any other characteristic of interest. The building of the matrix can be done by hand, for it mostly involves rearranging a relatively small number of figures from tabulations that would have been published in the census already.

The first step in building the matrix is to construct a table showing the land allocations over crops by farm size (or by other distributional characteristic), as exemplified in Table 2 from the cited Honduran study (Garcia et al., 1988). Then, information on yields and prices must be obtained. Usually that is also contained in the census; if the prices are not included, they can be obtained readily elsewhere, and it can be assumed that they do not vary over farm size class. With the additional information, another table of the same format is constructed, with gross income per crop per farm size class in the cells, instead of acreage.

The Honduran study added data from another source on costs of production per crop and farm size class, thus converting the matrix into yet another table on net farm incomes, structured according to the same format (shown in Table 3). Now the stage is set for analyzing hypothetical policy variations. A price variation of x percent can be hypothesized for selected crops or, for example, for all tradeable crops to represent the effects of a devaluation, and the cells in the policy effects matrix are filled with the resulting increments in income. The distributional pattern is apparent in the totals by farm size class.

If data on costs of production also are available, then similar experiments can be performed with input prices, including wage rates. Table 4 shows a hypothetical experiment for the Honduran case that represents the effects of a ten percent devaluation plus “moderate wage restraint.”

The principal caveat to these calculations is that they do not include substitution effects on the supply side, in response to the price changes. But in terms of their effects on the farms' net incomes, substitution responses are likely to be second order effects, that is, of a smaller order of magnitude than the static effects measured in the matrix. Recall that each incremental unit of output is displacing some other output, so net income for the farm will not change as much as the production levels of some crops. If it is felt that the substitution responses have quantitatively important effects on incomes, then the analyst should develop a formal model, using this matrix as one element in the model.

The possible dynamic effects of price changes on the choice of production technology also are excluded in this simple analysis, but again that is a question to be approached with other techniques of analysis. The caveats notwithstanding, it can be quite useful to be able to supply quickly to policy makers the kind of first-order distributional estimates that are contained in these census-based exercises.

Table 2. AVERAGE CROPPING PATTERNS IN HONDURAS BY FARM SIZE, 1974
(hectares per crop per farm)

Notes: The table does not include the part of the farm that is in pastures, forests, or non-agricultural uses;
* indicates less than 0.01 ha.

^aHere plantains include other members of the banana family (otros guineos), except bananas proper.

^bSapodilla (zapote), oil palm, chestnut, papaya, peach, quince, pepper, soybeans, rattan, forage sorghum, and others.

Source: Garcia et al. (1988).

Table 3.
HONDURAS: SOURCES OF NUT FARM HOUSEHOLD INCOME BY FARM SIZE GROUP, 1975
(lempiras/year)

Notes: Incomes are net of all costs except family labor. Figures in parentheses show the percentage composition of total net household Income.

Source; Garcia et al. (1988).

Table 4.
ESTIMATED DISTRIBUTIONAL EFFECTS ON NET FARM HOUSEHOLD INCOME OF A TEN PERCENT DEVALUATION WITH MODERATE WAGE RESTRAINT
(lempiras/household/year)

Notes: 1) Figures in parentheses refer to percentage changes in, first, net household agricultural income, and then of total net income, that result from the hypothetical policy change.

2) Moderate wage restraint is interpreted as a 5% wage increase while all product prices rise by 10%.

3) Input prices (except hired labor) as well as output prices are assumed to rise by 10%.

Source: Garcia et al. (1988).

3.1.2. The Efficiency of Land Use by Farm Size Class

A related use of the census information is to estimate the overall cropping intensity of each farm size class, and. the economic productivity per hectare, also by farm size class. When this was done in the Garcia et al. study, the important result was found that farms of 20–50 hectares had a forty percent lower cropping intensity than farms of 0–2 hectares, and the difference in economic productivity per hectare was even greater. These results were used as a basis for a recommendation for a land tax, for in those circumstances even a uniform tax would tend to induce some of the larger units to sell off part of their land to smaller holdings. A. progressive tax would have stronger results in that direction.

In sum, two characteristics of the agricultural censuses make them especially useful: their comprehensive coverage, so that statements can be made about the incidence of policy on a sector-wide basis, and their linkage of production variables with other variables such as farm size and region. Descriptive policy analysis can benefit from greater use of these sources of information.

3.2. Household Surveys and Their Uses for Food and Agriculture

Unlike censuses, household surveys are extensively exploited by economic analysts for a variety of studies, concerning consumption parameters, the determinants of household income, the income distribution, the determinants of fertility rates, and many other socio-economic research topics. Given the frequency of use of household surveys, this paper will confine itself to a few brief remarks about some uses which have priority for agricultural and food policy analysis. These surveys can provide critical information to help design complementary food and agricultural policies. Often a focus of agricultural policy in developing countries is the provision of better incentives to producers, but as long as extreme poverty and malnutrition exist, such policies need to be accompanied by targetted food subsidies or other programs to alleviate the nutritional deficiencies in the population.

3.2.1. Consumption Parameters

In keeping with the spirit of this paper, the discussion focusses on uses of the surveys that can be undertaken on the basis of previously tabulated survey results. However, it is worth mentioning in passing that there are two kinds of statistical analysis with the raw survey data that are of especial interest for agricultural and food analysis, particularly when food assistance programs are being assessed. The first is a set of tabulations of average and marginal propensities to consume food, by income stratum, by rural-urban distinction, and by other relevant disaggregations, such as occupation of head of household. Those propensities should be calculated for each major food product and for all foods taken together. The second is a set of price and income elasticities of demand for agricultural products, utilizing any of a variety of statistical approaches. (For a good agricultural reference on demand estimation, see George and King, 1971.)

With the set of tabulated average propensities by product and household group, it is a straightforward matter to calculate the incidence of agricultural pricing policy, or of food subsidy policies, on consumers, in a first order approximation much as was suggested for the case of producers, with agricultural census data. These data lend themselves directly to calculating the ratio of food subsidies to income by household group; in this way, Trairatvorakul (1984) showed for Thailand that removal of the food subsidies implicitly would affect different groups of poor households differently.

With a matrix of demand elasticities, the most important substitution effects in consumption, in response to hypothetical changes in relative prices, can be worked out as well. Alderman (1984) suggests a direct policy implication of own-price elasticities: that the cost-effectiveness of a food subsidy program will be in proportion to (the absolute value of) the elasticity.

Often these propensities and elasticities will have been developed through other studies. If not, the policy analyst may want to consider recommending a special study to do so, if the survey already has been carried out. Among the more policy-relevant elasticities that can be calculated are the calorie-income and calorie-protein elasticities (Alderman, 1984; Knudsen and Scandizzo, 1982).

3.2.2. Nutrient Availability

Once these initial survey results are available, another direct and rather important use of them is to compute tables of nutrient availability by income stratum for rural and urban areas. The required complementary information consists of tables of nutrient conversion factors for the foods used in the diets of the region. Such tables are available for all regions of the world. The procedure consists of multiplying the average propensities to consume, for each food {expressed in quantity units of the food, not values), by the corresponding coefficients of nutrient content (usually calories and proteins, but other nutrients also if desired). Sums can then i be performed for each household group, to derive estimates of the total per capita household nutrient availability. This procedure is quite simple, and clearly it is an essential step in linking nutrition concerns with pricing policy and other agricultural policies.

While income strata of ten are the organizing framework for data from household surveys, other kinds of tabulations may be more useful in some circumstances. Franklin, Harrell and Parillon (1985) provide an interesting example of a functional classification of households (for Panama) for arraying nutritional information in a way that is particularly helpful for designing nutrition programs.

If a primary goal of food subsidy policy is to assure greater levels of nutrition for the most disadvantaged groups in society, then it is important to identify the foods which are the most appropriate vehicles for such a policy. It also is important to evaluate the cost effectiveness of alternative subsidy strategies in achieving the nutrition goals. For these undertakings, the survey data on consumption patterns and the calculations of nutrient availability are indispensable. (The value of a time series of household surveys for food policy analysis is shown in the paper on India by Evenson [1986]).

3.2.3. Sources of Farm Household Income

Another aspect of the household surveys that is of considerable interest for agricultural policy analysis is the information on the roles of farm income and off-farm income by farm size and region. Tables with these data help in assessing the effects of wage policies and rural works policies. And among other things, they are inputs into an evaluation of whether the net effect of an increase in the prices of the basic food grains is positive or negative for the welfare of small-farm households. If off-farm income forms a large enough share of total household income, then the farm is a net purchaser rather than a net seller of basic grains, and hence will be adversely affected by price increases.

3.2.4. Other Uses of Household Surveys

A related issue that sometimes is clarified by household survey data is the composition of the farm household labor force. How old do the children have to be before they work on the farm? How much time do children and the spouse put into farm field work? If the head of household works off the farm, how much time is he available for work on his own farm?

Household surveys usually are the only source of information on a statistic of considerable interest in analyses of low-income agriculture: the amount of home retentions of the harvests, versus the amount sold.

If the household survey also collects information on farm production patterns and yields, then another field of analysis is opened up. For example, the influence of household characteristics such as educational levels, age of household head, and household wealth on the adoption of improved production technologies can be assessed with statistical analyses.

The broader issue concerning household surveys is that, in spite of their widespread use by economists in general, they tend to be ignored by agricultural production economists. The brief comments of this section suggest that for purposes of analyzing agricultural and food policy options, those surveys can be extremely useful. Moreover, much of the policy-related analysis can be carried out on the basis of published or tabulated survey results that are normally made available by other specialists.

3.3. The Sector's Fiscal System

In the fiscal area, the question is not so much one of data availability as how to organize. those data: what are the questions to be analyzed? The fiscal system is composed of four main components: current account expenditures, capital account expenditures, fiscal subsidies, and fiscal revenues. The requisite data typically are available in the official government accounts, but it may be a sizeable undertaking to assemble a complete numerical picture of the system, because of the large number of institutions involved. It is not uncommon for expenditures in agriculture to be made by twenty or more official agencies, sometimes many more when the agricultural parastatal agencies are taken into account.

There is a long tradition of reviewing the system of public expenditures in agriculture, with a view to making them more efficient or effecting various types of fiscal reforms. Generally, much less attention has been paid to the revenue collection side, as agriculture normally is not very heavily taxed, at least in developing countries (although it often is taxes implicitly by pricing policy). It is beyond the scope of this paper to delve into the many issues of a fiscal nature; the purpose of this section is to suggest ways of organizing the data on public sector activities, for both expenditures and revenues, that shed light on important policy issues.

3.3.1. Functional Classifications of Fiscal Expenditures

Specialists dealing with public sector accounts have invented functional classifications of public expenditure, to complement the normal classifications by agency and program. A functional classification is derived from an economic concept of the role of the expenditures. For example, all expenditures on transportation facilities will be grouped together, regardless of which cabinet agency makes them. Unfortunately, in developing-country agriculture, such classifications of public expenditure rarely exist on a comprehensive basis, although tallies are made of, for example, all expenditures on irrigation.

The functional classifications can be compiled, although to do so requires a commitment of staff resources on the part of the host government. Strictly speaking, it is only through a functional classification of all expenditures in the sector, for several successive years, that it is possible to review the implicit priorities in public agricultural outlays, and to see whether i) they accord with expressed policy statements, and ii) they have changed over time. Therefore, whenever it is possible, it is well worth the effort.

Table 5 shows such a compilation for Mexico, made through the efforts of the Ministry of Programming and Budgeting in Mexico. It took several person-weeks of labor by Mexican civil servants to pull together those two pages. Taking that table as an illustrative case, several conclusions are immediately apparent. First, real expenditure in the sector declined by more than fifty percent between 1980 and 1983. Second, the investment budget declined proportionately much more than the current account budget. (That is a common experience in periods of economic crisis, unfortunately for growth prospects; see Fischer, 1986.)

Third, at least in agriculture, the traditional distinction between investment outlays and current outlays is not very meaningful. Human capital formation, livestock programs, forestry programs, and agricultural research, all of which deal with capital formation, have greater current expenditures than capital expenditures in most years. In contrast, the function of administration and planning, which classically is not a capital formation activity, contains significant amounts of “investment” expenditures. (True, there is a distinction between an expenditure on a capital good— buying a machine—and capital formation—building the machine—but recall that in national accounts the sum of “investment outlays” of the type in Table 5 is called gross fixed capital formation. In this sense, the usual budgetary accounting categories can indeed be misleading if we want to know to what extent public “investment” programs are creating capital, physical or human.)

Table 5. EXPENDITURES OF THE MEXICAN MINISTRY OF AGRICULTURE BY
PROGRAM, 1980–1984
(in billions of 1977 pesos.

* Figure less than 0.1 billion.

Source: Celis and Norton (1985).

Fourth, in spite of policy statements by the Mexican Government that the emphasis on large scale irrigation was to be reduced, in 1984 those expenditures occupied a larger share of the budget than they did in 1980. And fifth, in spite of the commitment to give greater priority to developing agriculture in the Mexican tropics, a strategy which is acknowledged to require more emphasis on research and training, as opposed to physical investments in infrastructure, the real outlays for the functions of “technical assistance and coordination in (agricultural) districts” and “human capital formation” declined as much as the total budget did.

Of course, the years 1982–1984 were the first years of the severe and continuing Mexican economic crisis, so the Government didn't have much room for manoeuver. Hence the point of this example is not to be critical of its actions. Rather, the point is that only a compilation by purpose rather than by administrative agency permits clear identification of those issues. Time series of functional classifications of expenditures can be very useful tools, especially if they cover all the agencies dealing with the sector.

3.3.2. Fiscal Revenue Accounts

On the revenue side of the fiscal accounts in agriculture, the main issues are three: l) Do the revenue collection activities distort incentives in agriculture away from efficient patterns? 2) Are there alternative revenue collection schemes which would result in the same total revenue collection and be less distortive of relative prices? 3) Can additional real revenues be collected without prejudicing either the poorest farm families or the network of incentives? This last question arises with increasing frequency in the context of structural adjustment programs.

The two kinds of taxes which most frequently affect agricultural incentives are export taxes and the implicit taxes created by the price setting policies for major parastatal agroindustries, such as sugar mills, wheat mills, and oilseed processing plants. Sometimes; revenue collection needs are the main force driving the setting of those policies. Exchange rate policy, if it leads to an overvalued domestic currency, can create another implicit tax on agriculture,and it can be quite large. See, for example, the World Bank's 1986 World Development Report.

In these cases, the relevant data to assemble are the tax rates, along with estimates of what the agricultural prices would be in the absence of those taxes. In exploring alternative schemes for revenue collection in agriculture, it is helpful to make a systematic list of actual revenue sources in the sector, with estimates of the amounts collected from each source. That list is more useful if it can be supplemented by a companion list of what may be called policy-induced rents. Those rents are created by public policies, but they go into private hands rather than into the Treasury. For example, there are cases in which public agricultural lands are leased at fees below what the land markets would indicate, and then the lessees in turn sub-lease them at market rates. A private economic rent is generated by these circumstances; had the public leasing fee been closer to the market rate, that rent could have been captured in the public budget. A more common example is found in the practice of exempting irrigators from water charges that would recover the cost of the irrigation facilities or, more to the point, that would represent the marginal value product of the irrigation water.

Policy-induced rents represent the potential revenue sources for the government's budget. Table 6 shows examples of such rents for Haiti, with data as of 1982 (taken from Norton, 1984). The table also shows revenues which were collected but did not go into the regular budget in that era in Haiti, as a further demonstration of the scope for change in the pattern of collection of budgetary revenues.

In effect, the table provided evidence to counter assertions that further revenue collections in Haitian agriculture were not possible. It also served as a basis for discussions of the possibilities for reducing some distortive taxes and replacing them with other, less distortive, sources of revenue. The main point here is that simply organizing the available data in an appropriate way can be quite useful for analyzing policy.

3.4. The Labor Force and Agricultural Inputs

3.4.1. The Agricultural Labor Force

Labor is agriculture's most fundamental input, apart from land, but it is its least well documented. Estimates of the agricultural labor force are made at long and irregular intervals, via population and agricultural censuses, but in developing countries there do not exist direct estimates of employment in agriculture, on a sector-wide basis, which reveal the degree of employment over the year or for the year as a whole.

Table 6. THE AGRICULTURAL FISCAL. SYSTEM IN HAITI

^a US$8.0 million in 1983.

^b Includes advance of US$2.9 million to the Government but does not include capital reserve fund of US$2.6 million

Source: Norton (1984).

Employment in agriculture is highly seasonal and sporadic within a season. With the technologies characteristic of developing agriculture, most annual crops require 30–100 person-days of labor in a season of five to eight months (150 to 240 days). Farmers and their families may be well occupied with artisanal projects, local occupations in the service sector, or seasonal migratory labor, but there do not exist estimates of their total annual employment, except in case studies for particular localities.

Indirect estimates have been made for entire sectors of agricultural employment, via knowledge of cropping patterns and vectors of input coefficients for each crop and each technique of production, usually with linear programming models that embody those coefficients. Such estimates are useful but they are made only for particular points in time and hence do not provide a time series view of the evolution of employment in agriculture.

For studies of time trends, the investigator is left with estimates of the labor force. The age composition of the agricultural labor force is worth examining, for not all family members can provide adult-equivalent labor, particularly in field work. Depending on the issues at hand, it may be worth converting the total labor force into an adult-equivalent measure.

An implication of this situation is that time series on agricultural unemployment or underemployment cannot be constructed. However, there are two aspects of the labor force that merit attention: the sign of its growth rate, and calculations of its productivity.

The sign of its growth rate refers to the fact that, in all countries with expanding economies, eventually the agricultural labor force ceases to grow and then starts to shrink, owing to the influence of urban-rural migration and an urban pole of attraction that is ever stronger with respect to the rural population and labor force. In countries with a relative scarcity of agricultural land, which includes most countries in the world, this turning point is a watershed with social significance, for it marks the date after which the average farm size may start to increase, after having been continually decreasing for decades and perhaps centuries. While farm size structures may not respond immediately to changes in the rural population, certainly the amount of cultivable land per rural family starts to increase after this turning point is reached.

In many cases, that turning point still is many decades away, in spite of continuing rural-urban migration. But when projections indicate it is closer in time, possibly within a decade, as they did in Mexico in the late 1970s, it may be worthwhile to focus on this issue with a simple rural-urban projection model of population and the labor force, utilizing census data supplemented by more frequent population surveys if they are available. Preferably the model would have a regional dimension, for the turning point in rural population may not occur at the same time in all regions.

The more rapid growth of the population and labor force in urban areas can mean that labor productivity is growing more rapidly in agriculture than in nonagriculture, in spite of generally higher overall economic growth rates in the nonagricultural sectors. In this context, labor productivity in agriculture can be measured approximately as the ratio of real sectoral GDP to the rural labor force. It is an approximation because rural areas also contain a non-agricultural labor force, especially in the service sectors.

Nevertheless, the growth rate of this productivity measure may be taken as indicative of the growth rate of agricultural labor productivity. In some cases, the censuses provide enough information to distinguish the purely agricultural labor force. However, if the census is carried out only in one period of the year, and if the season. of that period differs over censuses, then problems of intercensal comparability arise with regard to the agricultural labor force data, as was the case in Mexico for the censuses of 1950, 1960 and 1970.

The Honduran study cited above (Garcia et al., 1988) found evidence that labor productivity was growing more in agriculture than in urban sectors, and that real per capita incomes probably were declining in urban areas, over the 1970–85 period, owing to the large influx of marginally employed workers to cities and towns. If this trend were to continue for a long enough time, it probably would dampen rates of rural-urban migration, but it might be even longer before the absolute numbers of migrants each year tapered off. (In Honduras, as in other developing countries, the average urban income per capita still is much higher than the average rural income per capita.)

3.4.2. Data on Agricultural Inputs

Data on inputs are used mostly in farm budget studies and computations of economic protection rates, for selected points in time. (See sections 10 and 13 below.) Time series on sales of inputs, in both quantities and prices, can be very useful, but they are difficult to compile and thus usually are available only on a partial basis.

If time series were available for all inputs of any economic significance, then one obvious use would be to make a more direct estimate of value added in the sector, by subtracting the inputs from the production series. With series on input prices (and quantities for one base year, to be used as weights), then another measure of the intersectoral terms of trade could be constructed. And the input quantity series themselves could be used as indicators of the overall rate of technological progress in the sector.

In most countries, the primary sources of data on inputs are private suppliers of those inputs. The difficulties are basically three:

1. the suppliers may not be willing to divulge information on their sales;

2. even if they are cooperative, unless a systematic effort has been made over time, the time series may be short, because usually the suppliers' records do not go back many years; and

3. for some classes of inputs, the amount of product differentiation is very great, and many products are discontinued after a few years or will have been introduced for the first time in very recent years.

This last observation is particularly true in the cases of non-fertilizer agricultural chemicals (pesticides and fungicides) and agricultural machinery. The number of different agricultural chemicals typically mounts into the scores, and sometimes into the hundreds. Hence it is rare to find a time series and indexes refering to prices and quantities of agricultural chemicals applied (except fertilizers).

A systematic program of statistical documentation on inputs could be developed in most countries, with the cooperation of the producer associations for those goods, but it would require commitments of government staff members on a regular basis, and many governments have not made that commitment in this area. In the meantime, some partial series normally can be found, especially for the main chemical fertilizers, for irrigation water, and for short-term agricultural credit from banking institutions.

Market Imperfections

A principal policy concern in the area of inputs is the presence of local monopolies and oligopolies in supply, which tends to elevate the inputs' prices over what they would be in a competitive supply environment. This circumstance occurs with greater frequency in countries with small domestic markets. In Panama, for example, Conklin (1986) found that none of the 22 major agroindustries had more than three firms, and roost had only one or two. In some other countries, public monopolies are as pervasive as private ones, or more so, in this area. While the public monopolies may not price their products the way private ones do, the record increasingly shows they suffer from other kinds of problems.

A quantification of the possible degree of market distortion from the presence of monopolies is difficult without knowledge of the industry's supply and demand curves, but a time series can help detect the existence of a problem and its magnitude. In the Dominican Republic, a comparison of the time series of international and domestic prices for major fertilizers revealed an increasing gap between the two series, and hence indicated an increasing loss of competitiveness on the part of domestic suppliers. The problem had been unsuspected previously until these time series were collated and interpreted.

Other Issues on Inputs

The other time series regarding fertilizer that is basic for analysis of the sector is the ratio of fertilizer to cultivated land area, which is taken as an indicator of the rate of modernization of agricultural production. Preferably, this is compiled on a nutrient basis (nitrogen, phosphate, potash), for the mix of fertilizer products may change over time, and each one has its own nutrient composition. The coefficients of nutrient composition are well documented; see, for example, the World Bank's handbook on fertilizers.

Data on inputs in general, and fertilizers in particular, are of course vital to the estimation of agricultural production functions. The main bottleneck in this regard is that researchers prefer to have the data by crop, and there are no direct, sector-wide data on fertilizer applications by crop. The allocations over crops invariably are indirect estimates, and at best would be usable only for a few major crops. Again, the possible utility of intensive, continuing surveys of selected farms.

For livestock, the major input is, of course, feed (apart from capital inputs such as breeding stock, stables and fencing). Veterinary services can have a very high rate of return, but as a component of cost they do not bulk very large. As noted in section 2.1, it can be quite useful to form a time series on all feed products, including those which are imported, to develop an appreciation of the speed of growth of demand for these products. (Among other things, this will help in making import projections.) The main difficulty here is that some crops serve as both livestock feed and human feed, especially when retained on the farm, so recourse may have to be made to occasional cross-sectional surveys on this point to try to separate the proportions of the two components.

3.4.3. Short-Term Agricultural Credit

Information on institutional agricultural credit normally is fairly well tabulated and interpreted. The dominant questions concern the trends in the real volume of credit extended to agriculture and the real interest rate charged. In inflationary periods, interest rates for agricultural loans tend to become low or negative in real terms, and this issue often is a subject of debate with international financing institutions. A reference point for the volume of credit can be established by computing for each year the ratio of total institutional short-term credit to GDP. Then the same ratio can be calculated for the agricultural sector, and a comparison. between the two time series of ratios indicates roughly whether the revealed policy on intersectoral credit allocations is changing over time.

By the same token, it is useful to know what private moneylenders are charging farmers, and for which crops and technologies farmers are willing to pay those interest rates. This provides a benchmark for evaluating institutional lending policies. Small-scale, inexpensive surveys of farmers can be designed to specifically capture information on this issue.

Another handy broad measure is the ratio of the nominal volume of institutional credit to total costs of production in agriculture. The accepted rule of thumb for an adequate supply of credit in developing agriculture is 40 percent of the the total costs (with labor costs included). This rule can be applied either to the group of institutional borrowers or to the sector as a whole, but in the latter case a rough idea of the volume of non-institutional credit is needed.

Simple Statistical Analyses of Agricultural Credit

Aside from constructing these simple indicators, credit data tend to be underexploited for purposes of analysis. In this section, two examples of credit analyses are mentioned, from the previously cited study of Colombian agriculture (Norton, 1986), as illustrations of additional uses of the information on credit. The first example concerns the role of credit in the agricultural economy: does it promote technological progress, in the sense of helping farmers make the transition to higher-input, higher-yield techniques of production, or does it simply sustain the more commercial farmers at their existing levels of production and input intensity?

In the study of Colombia, correlation coefficients were calculated for the average amount of credit received per hectare and the rate of change of crop yields over a twelve-year period. It was found that there was no correlation between the proportionate amount of credit allocated to a crop and its rate of yield improvement. Of coarse, it is possible that within the set of farmers growing a given crop, those who received the most credit, proportionally, were those who improved their yields the most. But in terms of the allocation of credit over crops, which to some extent is a policy variable in Colombia, the conclusion appeared to be that credit was not a significant vehicle for technical change.

The second example refers to a different kind of data on credit, that which is found in the lending records by project of the agricultural banks. In the Colombian case, a file was assembled on about 200 agricultural loans, on the basis of the loan feasibility studies carried out prior to bank approval of the loan. All the feasibility studies contained economic and financial profiles of the farm or ranch. Those profiles reported the operation's stock of fixed capital and the average annual amount of working capital necessary for the operation. Thus it was possible to specify a production function that depended on labor, working capital and fixed capital, all expressed per hectare.

The estimation exercise showed that short-term capital had a productivity several times higher than long-term capital. Field experts questioned subsequently found this result plausible and understandable, for once the fixed capital facilities are in place, access to working capital yields very high returns. Nevertheless, the results provided guidance for reformulation of the policy on interest rates by term, and they illustrate one use of yet another source of information: the farm-level feasibility studies conducted by agricultural banks.

3.5. Farm Budgets

3.5.1. Characteristics of Farm Budgets

There is a long tradition of analysis based on farm budget data,' for the most part concerned with the economic and financial feasibility of different cropping patterns or techniques of production. These studies typically are conducted at the micro (farm) level, sometimes at the regional level.

Farm budgets are compiled by various kinds of institutions: agricultural banks, agricultural insurance companies, agricultural research and extension organizations, and special commissions or study units at the regional level. Some organizations update them regularly over time for each major producing locality and each crop or combination of crops (or livestock products). More often, they are available sporadically and on a selective basis by product.

A farm budget is a list of the costs of production. Preferably it displays both the quantity and the unit price of each input. The more detailed the list of inputs, the more useful the farm budget. Occasionally the list will include inputs by month, but normally they are expressed on the basis of a crop season, and the analyst must have recourse to other sources to find the monthly calendar of inputs. The timing of inputs is important to know for three types of inquiries:

1. those concerning the monthly cash flow situation of the farm over the year,

2. those that calculate the rate of return to short-term credit (which is borrowed for less than a full year), and

3. those that investigate the feasibility of alternative cropping rotations within the year.

To be useful, the budget must identify the locality of the farm and the yield of the crop(s). It is even more useful if the local farmgate price of the output is given, so that regional-average or national-average prices don't have to be applied.

Some farm budgets include the normal requirement for short-term credit, with the interest payments included in the list of costs of production. If they don't, the analyst may have to add this information. A general • cautionary note is that many farm budgets exclude the general category of miscellaneous and unforseen costs. Those costs typically amount to about ten percent of the total costs of production, so again it would be worthwhile to add them if they are not already present.

The other major item that varies in its treatment is land costs. Some budgets exclude them, on the ground that the farmer owns his land or has usufruct rights through a cooperative or other institution. If the budgets are to be used in the framework of a linear programming model (which involves a higher level of modelling expertise than is assumed in this paper), this omission is acceptable, for the model will compute the annualized opportunity cost of land (its net earnings) via the so-called dual solution. But if the budgets are to be used, for example, in studies of economic protection rates (section 4.2), then the implicit land rent must be added. Usually field inquiries can provide an estimate of prevailing land rents by type of land in each region, even if the land rental market is extraofficial. For a detailed treatment of farm budgets, see Brown (1979).

3.5.2. Uses of Farm Budgets

Farm budgets have many uses, for example: 1) calculation of the rates of return to investment projects, especially irrigation projects, via assumptions on the post-project cropping pattern in the region and the net income coefficients per crop from the budgets; 2) calculation of the first-order effects on farm profitability of specified changes in input and output prices; 3) calculation of rates of economic protection (see below); 4) development of estimates of region-wide or sector-wide supply response and the incidence of policy, via linear programming models; 5) analysis of the monthly requirements for irrigation water of alternative cropping patterns, for planning water releases in irrigation districts; 6) analysis of the requirements for other inputs that are associated with hypothetical changes in cropping patterns, for planning input deliveries; and 7) analysis of the rate of employment and underemployment in agriculture. It is not too much to say that the farm budget is the most basic source of information for any analysis concerned with costs of agricultural production or the role of inputs.

They are used as well for the evaluation of the economic feasibility of new techniques of production that emerge from the experiment stations. In these cases, it is important to supplement the budgets with information on the riskiness of the old and new technologies; for that purpose, more specialized statistical techniques are needed (see, for example, Anderson, Dillon and Hardaker, 1977).

A related exercise that can be conducted is to calculate the rate of return to short-term credit that enables the farmer to adopt the new technology. The difference in total costs of production between the new and old technology can be obtained immediately from the budgets, along with the difference in output levels. Once a price is assigned to the output differential, then it can be divided by the cost differential to derive the period return associated with the investment in the additional costs of production. With information on the seasonality of inputs and the harvest, the period return can then be converted into an annual rate of return.

The working assumption here is that the new short-term capital finances the differential in production costs. Experience with these kinds of calculations also suggests that there is a very high rate of return for short-term agricultural lending in most cases.

Formal treatises on methods of agricultural policy analysis make extensive reference to the role of farm budgets. The currents here are brief, but they indicate that these budgets can be used in a variety of analytic exercises, both simple and involved. The relation between farm budgets and input-output analysis is explored briefly in Annex A.)

It should be noted that for some of the analyses mentioned in this section, it is necessary to know how crops combine and rotated (and the role of fallow land.) In other words, a description Pearson, chapter, 3.)