The most widely used time series concept at the sectoral level is that of production. Here production refers to gross output, and not value added. In developing agriculture it sometimes is important to distinguish between two forms of gross output—production and marketed output—for often the component of production retained at home is significant.
The main policy-related questions about production that usually arise are: 1) How fast is production growing? 2) Is it growing fast enough to meet the increments in food demand? 3) Is it growing fast enough to significantly reduce food imports (or increase agricultural exports)? 4) Which products are the main sources of growth, and which are lagging? 5) To what extent has the production growth depended on expansion of the area cultivated, and to what extent on increases in yields? 6) Which regions (or farming regimes) are contributing the most to output growth, and which are not experiencing growth? It may be noted that the question about area and yields is directly relevant to the allocation of the public budget between investments in land (including irrigation) and investments in research, extension, and farmer organization.
Behind these questions there are many whys, but the whys are much easier to answer, or at least speculate about in an informed way, if the foregoing questions are answered in numbers.
The first step in providing answers is to calculate a production index for the sector for each of the years in the period under consideration. The index is defined according to the usual procedure in which the level of gross output of each crop is weighted by its corresponding farmgate price. The formula may be written
| (1) |  |
where
Xt is the index of production in year t,
xit denotes the physical output of product i in year t,
xib denotes the same concept for the base year t = b, and
pib is the base-year price at the farmgate level for product i.
It can be seen that the element in the formula that changes over time is the physical production. The numerator of the formula is the constant-price value of (total) agricultural production for the corresponding year, i.e., the value as if prices had not changed. The denominator is the same concept for the base year; it converts the overall expression into an index. The units of the index are pure numbers; since both the numerator and the denominator have value units, their units cancel each other.
Cautionary Notes
Three cautionary notes are in order regarding the application of this formula, two mechanical and the other more substantive. The mechanical points are:
A reminder that the summations have to be performed before the division; sometimes, in writing the formula from memory it is written so that the division is done first, and that leads to an incorrect expression.
There is a tendency sometimes to weight outputs with value shares rather than with prices. Another way of looking at expression (1) is that it is composed of the sum of physical outputs with corresponding weights
| (1a) |  |
The use of value share weights, instead of the weights in (la), is perhaps understandable in agricultural work, because there are some crops with high unit prices (such as coffee), and the use of price weights may appear to give disproportionate weight to those crops. However, in such cases the unit yields are low, and so quantities are low in relative terms, and the total term in the numerator of (1) is not as large as the use of price weights would suggest.
In any event, use of value shares (or value added shares) as weights is not correct, for it leads to an index with units of measure equal to quantity units. An index should be a dimensionless concept. A production index refers to the weighted quantity in one period relative to the weighted quantity in another period, so the units of measure cancel each other.
The more substantive point is that it is useful to include as many products (crops and livestock products) as possible in forming the index. A recent study in Peru (Ccama et al., 1987) found that the eleven “national programming crops” which have been the chief focus of policy, and which have been the crops included in the official production index, accounted for at most 43 percent of the total value of crop output. To obtain a reasonably robust measure of sector performance, experiences in several countries suggest that it is important to include at least twenty crops in the index, and at least five livestock products.
One of the important crops that is commonly omitted from the index is cultivated pastures of various kinds (alfalfa, ryegrass, etc.), although the principal feed crops (sorghum, barley and yellow corn, for example) usually are included. In many countries the combined contribution of cultivated pastures is significant enough to make them the number one crop in value. In part, this omission reflects a fairly widespread tendency to place more policy emphasis on food grains and industrial crops than on livestock and feed crops. But, owing to the effect of higher income elasticities of demand for meat, milk and eggs, the livestock feed and forage crops often are the fastest growing in output.
Choice of Index Type and Base Year
Another issue concerns the choice of the form of the index and the base year. Index number theory deals with the options of a Laspeyres index, in which the base year is fixed, or a Paasche index, in which the base year is variable. In the paasche index, the base year is the current year, that is, the weights are different for each year of the index. Equation (1) is a Laspeyres index. A number of investigators have preferred to use a a third option, the Fisher index, which is intermediate between the other two indexes.
A Paasche index is likely (though not necessarily) to give a higher growth rate than a Laspeyres index, because the base-year weights in the latter tend to get out of date. Outputs in a given year depend partly on prices in that year. As prices change over time, outputs tend to grow more rapidly for those goods whose prices are rising. If price weights are taken from a recent year, those goods for which prices have been rising will have more weight than' they would have had under an earlier, fixed set of weights. Thus a Paasche index tends to reflect the reality that the sector's output composition is likely to have shifted over time in favor of the products whose prices have been increasing more rapidly.
In practice, the Paasche and Fisher indexes rarely are used except in research papers. In published statistics, the Laspeyres index is used. The base year tends to be the one that is used for other official economic accounts, and it is updated from time to time. If a choice is available, a more recent base year normally is preferable, because in that way the production index is constructed from the perspective • of more recent, and hence more pertinent, relative prices.
Decomposing the Production Index
The discussion of the role of feed products raises issues of decomposing the index into its constituent parts, that is, of computing separate indexes for different product groups. the question arises as to how the products are to be grouped. Is it by biological criteria (species, subspecies, etc.), by consumer demand criteria (groupings of substitutes and near-substitutes), or by other criteria? Is cotton an industrial crop or an oilseed? DO annual fruits belong with tree fruits (substitutes in demand) or with vegetables (substitutes in supply)? There is 'no single best classification for this purpose. The answer depends very much on the planned uses of the data, on the kinds of agricultural strategies that may be analyzed.
Perhaps the classification most frequently used is based on demand characteristics, for the logical reason that a good deal of sector analysis has to do with reconciling supply and demand. A typical exhaustive classification may include the following categories, or sane aggregation of them:
food grains (low income elasticities; major sources of both
calories and proteins for the poor; frequently face strong competition from imports);
legumes (important for protein in the diet; usually characterized by low rates of yield improvement);
roots and tubers (important staples in West Africa and some other regions; often income elasticities are very low or even negative; subject to larger-than-usual measurement errors in regard to production levels, because harvests take place throughout the year for some of these crops, as for cassava);
industrial crops (oilseeds, fibres, tobacco, sugarcane, etc.);
traditional exports (usually perennial crops; the list varies by country; it may include coffee, palm oil, rubber, cocoa, bananas, sugarcane, tea, for example);
feed grains and forage crops (excluding natural pastures);
fruits and vegetables (medium and high income elasticities; includes some non-traditional exports);
other crops (may be disaggregated for, e.g., major spice-producing countries);
livestock products (often lesser livestock products are overlooked, but in total they can be important—hides and skins, offals, small animals such as goats, rabbits and other rodents);
Different degrees of disaggregation are required for different purposes. For example, in almost all cases it is worthwhile to decompose the production index initially into components corresponding to crops and livestock products. Within crops, the policy interests normally will dictate separating into their own groups food grains, feed crops, industrial crops and traditional exports. It might be noted, however, that policy often seems to assign undue importance to food grains. Most pricing policies are directed at that product group, and yet it is not uncommon to find (with the simple data procedures of section 3.1) that they account for only 10–20 percent of farm household income, even on smaller farms.
Having defined the appropriate decomposition, carrying it out numerically will begin to indicate where the sector's main problems and prospects lie. By the same token, it is worthwhile to make calculations of the growth rate of the production index(es) by different subperiods of time. In calculating the production indexes themselves, the analyst has not really finished the job, for the concept of interest is the growth rate of production and food supplies, not the value of an index, which after all is an arbitrary number in its absolute level, and it has no unit of measure.
Care has to be taken not to make the subperiods too short, for the fluctuating nature of agricultural output can lead to misleading growth rates over short periods of time. In most cases, unless a structural break is expected at a particular point in time, it is preferable if the growth rates are calculated for periods of at least six years, and it is better yet if they are eight to ten years in length.
The preferred way to minimize the problem of random variations in production influencing the growth rate calculation is to use statistically fitted estimates of the trend growth rate. For this purpose, the function to be fitted (by ordinary least squares regression) is
| (2) | ln Xt = a + bt, |
where
the symbol ln denotes the natural logarithm (base e);
a and b are estimated parameters; and
b = r (the trend growth rate).
(Annex 3 provides a derivation of this equation.)
Thus a semi-logarithmic regression will yield the trend growth rate. The parameter b will have values like 0.03, for 3 percent annual growth, or -0.02, for a negative growth rate of 2 percent.
The goodness-of-fit (R2) of this equation is not of particular interest, but it is useful to look at the t-value of the estimated parameter b, to see if the resulting growth rate has statistical significance. In the case that it doesn't, then after abstracting from fluctuations it can be said that there is no discernible trend, either positive or negative. Often the t-value will have an acceptable level of statistical significance even though the R2 is quits low.
It may be noted that, for purposes of deriving growth rates that characterize an entire period, fitting a linear trend equation won't work, for it implicitly has a variable growth rate.
Productivity can be defined in many ways. Often the differences arise from the factor (labor, land, capital) to which the output growth is attributed, or how it is allocated to the different factors. One of the useful attributes of an exercise in estimating production functions is precisely that it assigns shares of incremental output to each of the factors. It may be noted in passing that the same result can be achieved with a linear programming model, often with a more disaggregated list of factors (such as land by season), via the dual solution or shadow prices. Also, methods have been developed for estimating the joint increase in productivity of all factors taken together (“total factor productivity”); the interested reader may wish to consult, for example, Christensen (1975).
In the case of agriculture, for the simpler exercises in descriptive statistics, normally the concern about productivity is whether crop yields have been increasing fast enough, according to decision makers' criteria. In other words, the concern is over productivity per unit of land, since the amount of cultivable land often is a more limiting factor than labor availability.
In the aggregate, agricultural growth can be decomposed precisely into three factors: expansion of the cultivated land area, increases in yields per unit of land, and changes in the aggregate composition of agricultural output. If the possibilities for expansion of the cultivated land are not sufficiently encouraging, growth policy will have to concentrate on the determinants of yields and cropping patterns.
Thus the first calculation to make in this regard is to construct the time series of unit yields for all the crops included in the production index, by dividing output by the corresponding land area for each crop. Which definition of land area is the most appropriate, area planted or area harvested? The difference is accounted for by cases of total loss of the crop, owing to exceptionally bad weather (severe drought, floods, etc.) or devastating local infestations by crop pests. Use of the area planted allows for this kind of risk. The fluctuations in yields over time will show all sources of production risk. Use of the area harvested as the denominator gives a higher average value of the yield, for it allows only for those risks which diminish yields but do not eliminate them completely.
For these reasons, the area planted is the preferred measure, but the area harvested is more often tabulated in the production surveys.
Double Cropping
The sense in which area is measured needs to be clarified. If a hectare is planted and harvested twice in a year, does it count as one or two hectares? Normally it counts as two. The crops may differ between the two seasons. Therefore an irrigation project that enables farmers to grow two crops a year where previously they had only one (in the “wet season”) effectively expands the cultivated area. Whatever the convention used, it should be noted.
Aggregate Yields
Normally these definitional questions will have been resolved by the statisticians who compile the series on agricultural production and yields. For the analyst, there is one small step to take, on the basis of the published data, that is often omitted and yet is important. That is to develop a series on the aggregate yield of the cropping sector, with corresponding series for each group of crops.
It can be done simply by weighting the physical yields by the corresponding areas of all the crops. But that is not meaningful. To say that the sector's average yield increased when the area in sugar cane (with, say, yields of 70 tons per hectare) increased at the expense of the coffee area (with yields of 0.5 tons per hectare) is not a very illuminating statement.
The more interesting way to develop the aggregate yields is to weight the physical yields by (constant) prices, thus developing a series on the economic yield per hectare of cropland. This measure is of direct concern to policy makers, for it should increase if the sector's performance to improve. Ideally, it would be desirable to calculate a series on the net economic yield per hectare, subtracting the costs of production, but time series on all inputs usually are not available; the growth rate of the gross economic yield may be used as a proxy for the growth rate of the net economic yield.
The most direct formula for the economic yield is as follows:
| (3) |  |
where
Yt is the sector's overall economic yield in year t;
xit is the physical output of product i in year t;
pib is the farm-gate price in the base year for product i; and
ait denotes the area in crop i in year t.
Notice that the crop yields themselves do not have to be used in this formula; they are implicit in it.
In the event of double cropping, increasing the winter acreage, for example, via irrigation has the effect of increasing both xit and the total area cultivated in year t (the sum over i of the ait ). If the crop i planted on that winter acreage has a higher price than the weighted average for products in the sector, then Yt will increase; otherwise it will decrease. Mixed cropping changes the xit but not the total area in year t. If measured correctly, the introduction of chili peppers on a mixed corn-and-beans plot would reduce the amount of corn and/or beans and increase the amount of chili produced. Whether or not this would increase the economic productivity would depend on the relative yields of those crops (in the mixed cropping mode) and their unit values, or prices.
It is not an exaggeration to say that the concept of economic productivity and the area cultivated are the two sources of agricultural growth, in the aggregate. Of course they in turn may respond to many economic and climatic influences, but from an accounting viewpoint they are sufficient together to track and describe the sector's growth.
The economic yield subsumes changes in the sector's crop composition. Sometimes, in fact, it can decline over time because a policy of emphasis on self-sufficiency in food grains leads to replacement of higher-value crops with lower value crops. This may have been a factor in the stagnation of Peru's agriculture in the last decade or two (Ccama et al., 1987).
Thus the economic yield bears watching over .time, but curiously it is not often calculated.
As in the case of the production index, the economic yield can be calculated for each crop group. It also can be calculated for other breakdowns, such as between irrigated and nonirrigated land. In the latter case, can the overall growth in the economic yield be lower than each of its components, lower than the yield growth rate on each of irrigated and nonirrigated land? Yes! Suppose irrigated yields are much higher than nonirrigated yields, for each crop and in the aggregate. Also suppose that nonirrigated yields are increasing faster (as has been the case in Mexico in the last decade), and that proportionately more land is being put under nonirrigated cultivation. Then the weight of production will be shifting progressively to nonirrigated areas, which have the lower yields at any point in time. Hence the overall yield growth rate can be less than either component.
The same result can occur when the overall yield growth, or the yield growth for a given crop, is being decomposed into its regional components. In that sense, the overall yield growth rate can understate the true rate of technical progress at the farm level. Nevertheless, it still stands as an accurate characterization of the (gross) economic productivity of the nation's total farmland.
The decomposition of the production index by product group is a kind of sources-of-growth exercise. A more systematic way to approach this question is to calculate the percentages of the sector's output growth over a given period that can be attributed to specific products. It yields statements such as “Rice accounted for 12 percent of the sector's output growth, and coffee 21 percent, over the period 1975–86.”
The formula for calculating product i's share contribution to growth over the period from year o to year t is as follows:
| (4) |  |
The numerator in expression (4) is the increment in the constant-price value of output for product i over the period under consideration. Points o and b need not coincide. Point o is the beginning year of the period over which the sources of growth are measured, and point t is the final year of that period. Point b is the base year—the year whose prices are used for the valuations in constant prices.
For products whose output is declining, the numerator will be negative and therefore S. will be negative, assuming the total increment in sectoral output (the denominator) is positive. Therefore the sum of positive values of S. will be more than 1.0 (more than 100 percent) if overall growth is positive, because the products with increasing output have to compensate for the decreasing products before total growth becomes positive. Taking into account all products i, including those with negative Si,
| (5) | Si = 1.0 |
When this exercise was performed for Panamanian agriculture, some surprising results were found.* Sugar cane accounted for about 25 percent of total growth in agriculture and livestock over the period 1960–75, and then its contribution turned negative, at about -21 percent. The contribution of rice fluctuated around 10 percent. And for the period 1975–85, fully 50 percent of the growth was accounted for by the poultry subsector, although its share of the level of total 'output was well under 10 percent.
An alternative expression for the growth shares in (4) can be written on the basis of the growth rate of each product's output and the static share of that product in sector outputs
| (6) |  |
where
ri is the growth rate of the output of product i over the period considered;
wi is its average share of the (constant-price) value of sectoral output over that period; and
r is the growth rate of the (constant-price) value of sectoral output over the same period.
Although Si and Si* are conceptually the same, their values will not be exactly equal owing to the use of averages in (6), but they will be approximately equal. The Si * also sum to 1.0, subject to a small approximation error arising from the use of the average output shares.
These formulas give a measure of the dynamic importance of individual crops, in a way that cannot be perceived simply by looking at the fact that one crop grew at two percent per year while another grew at five percent per year.
The Contribution of Price Changes
However, formulas (4) and (6) do not convey the import of price changes. Agricultural prices fluctuate notably over time. In general, at a world level they tended to rise relative to the prices of other classes of goods and services throughout most of the 1970s, and then to fall in the 1980s. The variations among agricultural products substantial. To capture the price effects, expression (4) can be restated to show the sources of growth in current-price incomes of producers. This is accomplished by replacing xitpib with xitpit, and xioPib with xiopio in (4)*
Disaggregating Growth by Yield and Area
The other kind of sources-of-growth analysis for agriculture partitions the overall growth in the cropping sector among improvements in yields and expansion of the cultivated area, as mentioned above. For this purpose, it is useful to disaggregate the economic productivity, or economic yield (equation (3)) into its components of physical yields and shifts in the aggregate crop composition. Thus in this case the sources of growth are three: area expansion, yield improvements, and changes in crop composition. (See Solis, 1970, for an early application of this approach.)
The share contribution of area expansion to total growth is the ratio of the growth rate in cultivated area to the growth rate of sectoral output:
| (7) | Sa = ra/r |
where
Sa is the percentage contribution of area expansion to the sector's output growth;
ra is the growth rate of the cultivated area for the period; and
r is again the growth rate of the (constant-price) value of sectoral output over the same period.
The share contribution of physical yields is the weighted sum of yields, divided by the growth rate of sectoral output:
| (8) |  |
where
Sy is the percentage of sectoral output growth attributable to growth in physical yields;
aio is the area in crop i in the initial year of the period; and
ry,i is the growth rate of the yield of crop i over the period.
Then the share contribution of changes in the crop composition is found by residual:
| (9) | Sc = 1.0 - Sa - Sy |
When Sc is positive, it measures the contribution to growth that stems from shifting out of products with a lower value per hectare and into those with a higher value per hectare.
For a sector whose output is growing at, say, 3.5 percent per year (which may be approximately the rate of growth of food demand in an economy growing at four percent per year), yields and area may each grow about 1.5 percent per year (Sa = Sy = 1.5/3.5 = .43), and shifts in crop composition may add about 0.5 percentage points per year (Sc = .14). It is unusual for aggregate yields to increase at much more than two percent per year, barring a major technological change, so if the area expansion factor drops below one percent per year, the problem is evident.
To clarify strategic thinking for an agricultural sector, it is helpful to set out various combinations of growth rates that fulfill equation (9), and to make experiments at a conceptual level with the influence of different factors such as irrigation on sector growth. Such an exercise illuminates both the possibilities and the constraints to the sector's future growth.
It should be borne in mind that changes in crop composition can make a negative contribution to sectoral growth, if lower-value crops substitute for higher-value crops, as sometimes happens if basic grains displace traditional exports at the margin.
In most, but not all, countries farmgate price data are routinely collected and tabulated. And usually wholesale and retail commodity price data are collected as well, though these data may be collected only for the capital city or for major urban areas. For reviewing sector-wide price trends, the first step that is lacking in many cases is the formation of a farmgate price index. As in the case of production, the overall index enables the analyst to make judgments about agricultural price movements as a whole, and to compare them with price movements in other sectors of the economy.
The farmgate price index looks very similar to the production index of equation (1):
| (10) |  |
The symbols are as defined for equation (1). Again the index is a Laspeyres index. The difference is that here the element that changes over time is the price term in the numerator, and the levels of output are the weights for the prices, rather than viceversa. (Fisher and Shell [1972] provide a theoretical review of price indexes.)
The same cautionary notes apply here as to production indexes; weighting prices by shares in output value won't work!
In dealing with price movements, the absolute level of prices, or of an index, is not as meaningful as a relative price. It can be relative to international prices, or, in the case of changes over time, relative to movements of other domestic prices, but it is a relative concept that is most interesting from an economic viewpoint.
Consequently, the next step in dealing with equation (10), before attempting to interpret its numerical expression, is to divide it (deflate it) by some other price index. A deflated price often is called a “real” price, analogous to the case of real GDP.
The choice of deflator usually is limited by the information available. Each deflator has its own interpretation. The most commonly used ones are as follows:
(1) An index of agricultural input prices. When the farmgate price index is deflated with this index, its movements over time indicate movements in the profitability of agricultural production, assuming constant levels of productivity. In developing countries, this kind of real farmgate price is more applicable to the commercial farmers, as subsistence farmers use few purchased inputs. It is sometimes called the agricultural terms of trade, for it measures agricultural price movements relative to the movements of those industrial goods that farmers purchase (for production purposes). For subsistence farmers, one might use an index of goods they consumer, as suggested in paragraph (4) below.
(2) An index of nonagricultural producer prices (ex-factory prices). Deflation with this index would result in a more general measure of how agricultural producers are faring relative to other producers in the economy. Unfortunately, such an index usually is not available, or if it is, it refers only to the manufacturing sector (and does not include services).
(3) The wholesale price index (WPI), or the nonagricultural component of it. This deflator is often used in the absence of an economy-wide producer price index, the reasoning being that farmgate prices are closer to wholesale prices in concept than they are to retail prices.
(4) The consumer price index. When this deflator is used, the purpose is not to compare producer prices in different sectors, but rather to measure the purchasing power of farm prices, in terms of the goods that a household would purchase. It sometimes is pointed out that the preferable deflator for this purpose would be a rural consumer price index, as the consumer baskets of goods differ significantly between rural and urban areas. Nevertheless, the rural and overall consumer price indexes (CPIs) are likely to move approximately together over time, so for reviewing the time path of the purchasing power of rural households, use of the overall CPI may be acceptable. Besides, in most developing countries, the only CPI available is an urban one.
(5) The GDP deflator: the ratio between current-price GDP and constant-price GDP. This index is commonly used as a deflator for other price indexes in the economy, but its interpretation is not entirely clear. The bundle of goods represented in GDP is the set of all final goods, in the input-output sense, including capital goods and exports. In other words, it includes the sales of automobiles but not the steel used in making automobiles; and the sales of flour to consumers but not the sales of what to flour mills. Most agricultural goods are intermediate goods; they require at least some processing before being sold to consumers. What is the meaning, therefore, of deflating the farmgate price index by an index of the prices of all final goods in the economy? In some writings, it is customary to refer to the GDP deflator as the “price of value added.” While in the mathematical, or accounting, sense the meaning of this phrase may be clear, it is not intuitively obvious in the economic sense. If labor were the only primary factor in the economy, then wage income would be the only kind of value added, and the price of value added would be the wage rate, averaged and adjusted for skill differentials. But a large share of value added in an actual economy comprises rents to fixed factors, that is, capital income from diverse kinds of capital. The price of capital income is not an intuitively evident concept. (An interest rate is a annual price of the stock of capital, not of capital income, which is a flow.) Therefore, this interpretation of the GDP deflator does not clarify its meaning.
The main reason for using the GDP deflator to construct a real farmgate price index is because it appears to be a fairly comprehensive measure of prices in the economy, and indeed it does refer to all final goods. But the meaning is clearer in the case of deflation by the consumer price index (to measure the purchasing power of farm households' income) or by the wholesale price index (as a proxy for a producer price index, and therefore to be used for constructing a measure of the intersectoral terms of trade).
As an empirical matter, consumer price indexes almost invariably increase more rapidly than wholesale price indexes. The gap between the two increases over time; only in this way can total real income in the retail trade sector increase over time. This means that movements in consumer prices of food products are not necessarily indicative of movements in incentives to producers.
The real farmgate price index may be written in symbols as (11) PRt = Pt/Dt
| (11) | PRt = Pt/Dt |
where D signifies the index of the selected deflator. The measure PR. is the key indicator of whether farmers are getting better off or worse off, per unit of output produced.
However the concept is measured, analyses of this kind are beginning to suggest that real farmgate prices in most developing countries increased throughout most of the decade of the 1970s and then decreased in the 1980s. In the majority of countries, the downward turning point occurred during the period 1978–81, occasionally as early as 1977. The primary determinant of these trends was the pattern of behavior of international commodity prices, which in turn has been partly attributable to the global slowdown in economic growth, and hence in demand, in the 1980s. The international trends were transmitted in varying degrees to different countries, owing to differences in the commodity composition of output and in domestic policies on pricing, trade and the exchange rate.
With any deflator, an index of the form (11) may be called a measure of the intersectoral terms of trade, or the domestic terms of trade, although perhaps that label is most appropriate when the deflator is either an index of input prices (“prices paid by farmers”), the producer price index, or the wholesale price index. This concept of the terms of trade is distinct from a country's international terms of trade, which refer to the evolution on world markets of the ratio of the prices for its exports to the prices it pays for its imports.
As in the case of the production index, the price index (11) may be defined separately for each commodity group. The questions that can be addressed with the aid of the disaggregated indexes are then the following: Are the cropping patterns evolving in a way indicated by the relative price trends (relatively more output of the products whose prices are increasing the most)? If not, why not? To what extent have domestic policies (subsidies, taxes, trade restrictions, exchange rate policy) determined the relative price trends? Are those price trends consistent with international price trends? If not, why not? And if not, is it the intention of policy to have some domestic relative prices move differently than their international counterparts (taking into account the arguments about potential loss of economic efficiency in resource allocation)?
Most countries have a consumer price index which is computed separately for food and non-food items as well as for consumption as a whole. This index is, of course, at the retail level. It is more useful if consumer-level food price indexes are developed for each household income stratum, based on the different consumption baskets of the strata, but this disaggregation is not usually available. Then the trends in food prices could be compared with estimates of the trends in incomes of each stratum, to see whether the groups of households are better off or not, and by how much.
This same question can be analyzed for the entire agricultural sector via the data from the national accounts. In standard practice, what is called “real agricultural income” is nominal (current-price) value added in the sector, deflated by its own sectoral GDP deflator. However, to measure how well off farming households are, it is useful to deflate the nominal value of agricultural GDP by either the overall consumer price index or the overall GDP deflator (or the nonagricultural GDP deflator).
When this was done for Colombia for the period 1978–84 (Norton, 1985), a striking result emerged. While real agricultural GDP, defined in the traditional way, had been increasing at a modest rate on average (2–3 percent per year), the purchasing power of agricultural GDP, as measured by nominal agricultural GDP deflated by the GDP deflator, actually had been declining. This simple result helped clarify for policy makers why the farming sector had been resisting so strongly some proposed policy reforms that would have made them even worse off.
One small moral of this story is that, while economists typically collect published series on real GDP, almost before any other data, it is well not to forget to collect the nominal series as well. They may have their uses.
A related exercise was conducted recently for Peru (Ccama et al., 1987), in Which the nominal farmgate value of agricultural output was deflated by the consumer price index, to derive a measure of the purchasing power of farm outputs. The empirical finding was that this measure had declined by more than 30 percent between 1979 and 1983, and by the end of 1986 it still remained well below the 1979 level. That trend certainly aggravated significantly the existing rural poverty.
In conclusion, cross-sectoral deflation procedures can be applied fruitfully to output and GDP series (nominal series) as well as to price indexes themselves.
The first questions to be posed regarding agricultural foreign trade usually concern the extent of agriculture's contribution to the economy's total commodity export earnings, its share of the commodity import bill, the magnitude and sign of the net agricultural foreign trade balance, and the rates of growth of agricultural exports and imports. There are some composition issues and price issues as well, as commented below.
Foreign trade data are among the more reliable data in most economies, as they are gathered by a single institution, the customs bureau, at relatively few ports. All tabulations normally are based on an initial six-digit classification of the commodities. Problems arise here from the user's viewpoint. There are literally hundreds of agricultural commodities. in the six-digit lists for exports and imports, even in relatively simple economies. Often the dividing line between agricultural and manufactured goods is not entirely clear, in the area of semi-processed and processed items. And it is not always clear which subcategories are most appropriate for some composite products.
It is difficult to make aggregations of these data, and for that reason they usually are not made. Therefore the analyst is left with two kinds of tabulated information: the national accounts data on total exports by sector (not always available), and the reported exports and imports of principal products. This information is not sufficient to analyze changes in the structure of agricultural foreign trade, nor to identify the most rapidly growing product groups in trade. Often there are groups whose present importance in trade is relatively small but whose growth rate is quite high.
In a recent Honduran study (Garcia et al., 1988), the six-digit trade data were tabulated (by hand, as the computer tapes were not available). It took a few person-weeks to tabulate exports and imports for two selected years; the results are shown in Table 1. Given that the main interest centered on structural questions, the data were left in current domestic prices. In spite of the effort required, the task appears to have been worthwhile, for several conclusions emerged from the table, for example: 1) Agricultural exports in Honduras are growing much more rapidly than imports; 2) The net agricultural trade balance is strongly positive and increasingly so; 3) Apart from the expected items of bananas and coffee, the largest increases in the value of exports came from fish and livestock products, sugar (I), pineapples, vegetable oil products (palm oil), and tobacco; 4) Extrapolating from the table, future export growth may be slower, because many of the main items face an increasingly unfavorable international market environment: sugar, bananas, palm oil, tobacco, and possibly coffee; 5) The largest increase in food imports came from dairy products (a phenomenon associated in the study with the internal pricing policy for milk), followed by wheat (mostly PL 480) and animal feeds; 6) These items can be expected to continue to increase rapidly in import volume, so the study foresaw a possible weakening of the favorable agricultural trade balance in the future.
Table 1. AGRICULTURAL FOREIGN TRADE IN HONDURAS, 1975 AND 1984 (in thousands of current lempiras)
| Imports | Exports | |||||
| Product Group | 1975 | 1984 | Growth Rate (%) | 1975 | 1984 | Growth Rate (%) |
| Livestock and fish products | 34,851 | 82,109 | 10.0 | 22,728 | 144,719 | 22.8 |
(milk) | (24,445) | (63,563) | (11.2) | (1) | (464) | (n.a.) |
(seafood) | (1,363) | (4,839) | (15.1) | (20,616) | (99,764) | (19.1) |
| Cereals, flour | 55,393 | 58,884 | 0.7 | 683 | 5,702 | 26.6 |
(wheat) | (19,551) | (34,278) | (6.4) | (0) | (0) | (n.a.) |
(corn) | (13,894) | (12,437) | (-1.2) | (0) | (2,789) | (n.a.) |
| Fruit, nuts | 3,581 | 8,035 | 9.4 | 130,706 | 522,206 | 16.6 |
(bananas, plantains) | (0) | (0) | (n.a.) | (123,312) | (469,188) | (16.0) |
(fresh pineapples) | (0) | (0) | (n.a.) | (1,669) | (28,274) | (36.9) |
| Deans | 239 | 1,824 | 25.3 | 1,712 | 3,324 | 7.7 |
| Roots, vegetables | 2,482 | 3,061 | 2.4 | 734 | 1,865 | 10.9 |
| Sugar | 750 | 831 | 1.1 | 15,126 | 57,722 | 16.0 |
| Starch | 276 | 566 | 8.3 | 2,651 | 7,669 | 12.5 |
| Oilseeds, veg. oils | 1,414 | 10,135 | 24.3 | 894 | 23,408 | 43.7 |
| Animal feeds | 2,743 | 12,711 | 18.6 | 1,225 | 221 | -17.3 |
| Coffee, cocoa, tea | 708 | 1,777 | 6.8 | 114,094 | 336,708 | 12.8 |
| Beverages | 2,245 | 6,425 | 12.4 | 136 | 114 | -1.9 |
| Tobacco | 545 | . 2,917 | 20.5 | 15,065 | 31,872 | 8.7 |
| Natural fibres | 4,940 | 4,860 | -0.2 | 9,145 | 15,631 | 6.1 |
| Rubber and gum | 2,293 | 6,169 | 11.6 | 797 | 537 | -4.3 |
| Wood and pulp | 1,413 | 2,467 | 6.4 | 83,914 | 90,942 | 0.9 |
| Others | 1,247 | 22,544 | 37.9 | 3,174 | 5,081 | 5.4 |
| TOTAL | 115,119 | 225,433 | 7.8 | 403,248 | 1,249,243 | 13.4 |
Notes; Oilseeds and vegetable oils include coconut products. Fruit and nuts include jams and jellies. Roots and vegetables include mushrooms and vegetable soups. Sugar includes sugar beets and sweets.
Source: Garcia et al. (1908).
This example illustrates one useful way to handle agricultural trade data. It would have been desirable to make the tabulations for more years, as imports and exports both fluctuate substantially from year to year on an item-by-item basis, but nevertheless the data for the two years were helpful.
Another frequently watched indicator is the import share of total agricultural (or food) supplies. This ratio may be calculated on the basis of the previous trade tabulations, both for principal products and for product groups. (For data on this concept for 21 countries, see Economic Research Service and the University of Minnesota, 1983.) These data also are compiled in the course of developing the food balance sheets discussed in section 4.1 below. However, sometimes there are some special problems to be resolved in developing the ratio of imported food to total food supplies—see section 2.5.3 below.
Price data for exports and imports come from the same original source as the quantity data do. The six-digit trade listings usually report quantities and values; the user has to perform the divisions to get the prices. For exports, the prices are f.o.b., while for imports they are c.i.f. These border prices are especially relevant for calculations of economic protection rates (see section 4.2); it also is helpful to compile time series on border prices for principal traded products, to be able to compare them with trends in domestic prices.
If national data on border prices are not available, internationally-compiled series may be employed. However, using the main international series on commodity prices is not as satisfactory as pulling the prices directly out of the country's own trade data. First, when the international data are used, it is necessary to find information on the shipping and handling margins relevant to transporting the product to the country from the point of the international price quotation (New Orleans, Liverpool, etc.). Second, the country may well obtain its supplies from other places, and the transport costs from those places may not be readily obtainable. Third, and more importantly, the prevailing mixture of long-term and short-term contracting for imports often means that the effective price obtained is significantly different from the standard international price, apart from transport considerations.
The international series are useful for reviewing trend growth rates in international prices, and organizations like the World Bank that make commodity price projections tend to tie them to these international series.
When it is not feasible to obtain the country's own six-digit trade data, a useful substitute for roost major products is the FAO's Trade Yearbook, which does contain data at the country level and not just the international commodity series. It lists the quantities and values of exports and imports, for many products and for virtually all countries. Its main drawback is that it comes out with a time lag of more than a year and a half, but normally there is a considerable lag also in the publication or distribution of national data on trade. A study of the implicit trade prices in that FAO publication will show a surprising amount of variation from year to year.
Another virtue of this trade yearbook is that it can be used to develop approximate estimates of the appropriate border prices for commodities that are not yet traded by the country concerned. Ah average of import prices for that commodity for neighboring countries will give an indication of what the price would be if imports were to commence in the country being studied.
From a viewpoint of monitoring food and nutrition policy, perhaps the most important time series is the one that reports total food supplies, including both imports and domestic production, over time. A time series of at least five or six years, and preferably ten years or more, is needed to be able to perceive the trends with confidence, because changes in inventories often are quite marked from year to year.
Compiling such a series requires dealing with some practical problems. The first one is to convert production data from a basis of agricultural years to a basis of calendar years, so that they will correspond in time to the imports. Often this task will have been performed by an official statistical agency (the Central Bank, if none other), but if not the analyst will have to ascertain the harvest months of each principal food crop and then make the conversions.
The second problem to confront concerns the form in which the import data are supplied. In some cases, time series on both quantities and prices for individual food imports are not published, but rather only a value series in current domestic prices. Prices and quantities may be published only for a handful of principal food items, but that is not sufficient when a measure of the total availability of food is being developed.
The concept of food availability is meaningful only in quantities or (for aggregates) in constant prices. When only current-price series are available, a deflation procedure or another approach is required. But what is the appropriate deflator? In one case, the author tested deflation by the national accounts import deflator, but that turned out to be quite misleading, for agricultural imports constituted a small part of total imports of goods and services, as is usually the case. No other deflator is readily evident.
The above-mentioned option of compiling import tabulations from the raw six-digit data (section 2.3.1) is not really viable for a time series, for it is exceedingly laborious even for two or three years.
The only practicable option may turn out to be recourse to the FAO's Trade Yearbooks, which do report both prices and quantities for imports for a large number of items. Tabulating the data from those yearbooks also is laborious, but it can be done for ten years of information in about a day. And given its value for nutritional analysis (see section 4.1), that is a day well spent. The major obstacle when working in the field may be obtaining a long enough series of the yearbooks. Each edition contains three years of data, but it is necessary to go back in time, and even the local representations of the FAD are not guaranteed to have a very long series. This is a case in which it may be worth doing a little preliminary data collection before travelling to the field.
The import price data in the yearbooks are denominated in dollars, and of course the domestic production value data come in local currency. In making the conversion to constant local prices, it is important to use the exchange rate of the base year. If there were multiple exchange rates prevailing in that year, it is worth inquiring which rate(s) applied to food imports; often it is the lowest rate (to the dismay of local farmers!), but there is no certainty of that.
Since the domestic supplies of food will be tabulated in constant prices, is also is necessary to use the dollar import prices of the base year before converting them by the exchange rate and adding them into the domestic supplies.
Conceptually, the c.i.f. import prices are not really comparable to farm-gate prices (section 4.2), but for the purpose of developing the series on the value of food supplies at constant prices it may be necessary to live with that imperfection.
Now the index of total food supply can be computed by applying an extension of the production index formula (1), where the quantities now will include imported foods. In the case in which import dollar prices and quantities for individual foods are used, then the formula will be as follows:
| (12) |  |
where
Ft is the index of total food supplies;
xit is the domestic production of product i in physical units;
mit is the quantity of product i imported;
pib is the base-year farm-gate price for product i;
dib is the base-year dollar c.i.f. import price of product i;
eb is the base-year dollar exchange rate for the country; and the subscript t refers to the year.
If food and nutrition are the primary concerns, then the analyst may wish to exclude from (12) non-food agricultural products, such as cotton fibre and, for major coffee exporters, coffee.
An Alternative Procedure for the Food Supply Index
If the required series on individual import items are not available, then an indirect procedure may be feasible. If a current dollar value series of total food imports (or total agricultural imports) is available, then it can be used. But those totals will have to be expressed in constant base-year prices through a deflation procedure. Which deflator in this case? An international wholesale price index.
Ideally it would be a weighted wholesale price index for all the country's international trading partners (the United States, France, Argentina, etc.), where the weights are the shares of food imports arriving from each of those countries. In practice it may have to be the u. S. wholesale price index or an index for another exporting country if the U. S. is not the major supplier of food to the country being studied.
If the joint-trading-partner index is to be compiled, it also must reflect the shifts in exchange rates among those partners, such as the devaluation of the dollar vs. European countries in recent years, or the devaluation of the Argentine austral vs. the dollar. The formula for the appropriate international wholesale price index, to be used to deflate the current-dollar values of our country's imports, is then as follows:
| (13) | Wt = Σjwjtejthj |
where
Wt is the computed international wholesale price index relevant for deflating our country's food import data in dollars;
Wjt is the wholesale price index for trading partner j;
ejt is the exchange rate of trading partner j expressed in units of its currency per dollar (=1.0 for the U. S.); and
hj is the share of our country's food imports supplied by trading partner j.
Motes: The subscript t again refers to the year; the foreign wholesale price indexes w. must first' be converted to a common base year, which will be the same base year b to be used for expressing the constant-price value of domestic production; that conversion is achieved simply by dividing each initial value of each wholesale price index w. by its corresponding base year value w...
Let us call the previously-compiled current-dollar series on food or agricultural imports Mt ; then the formula for the index of food supplies is
| (14) |  |
where the symbols are as defined above and W is given in equation (13).
Equations (12) and (14) are not strictly comparable expressions. In (12), imports are valued in their own constant prices of the base year; in (14), current-price imports are deflated by economy-wide price indexes in the trading partners. However, they measure approximately the same concept, and the availability of data may allow application of only one of them. If there. is a choice, equation (12) would be preferable, for in it the imports are treated in exactly the same way that domestic production is.
Food Supplies per Capita
Whichever formula is used, one last step should be remembered: the index should be divided by an index of the country's population, to transform it into an index of per capita food supplies. In dealing with food supplies, the per capita concept is the relevant one. (There are a number of developing countries for which per capita food supplies have in fact declined in the 1980s, so implementing formula (12) or (13) and dividing it by the population index may be important in defining national policy issues.)
Once the index is expressed in per capita terms, the analyst may wish to develop some graphs such as the one shown in Figure 1. This report does not talk about graphical presentations, but it is hard to exaggerate their, value as a tool for communicating the results of the analysis.
The other main policy-oriented use of time series on quantities imported (mit) is to construct series on the apparent aggregate availability of nutrients for the population. This topic is taken up in section 4.1.
Figure 1. EL SALVADOR: THE SUPPLY OF AGRICULTURAL GOODS PER CAPITA, 1970–1986
(excluding cotton fibre and coffee)
Source:: Lievano and norton (1988).
